Modelling and forecasting the volatility of JSE returns: a comparison of competing univariate GARCH models

Abstract

The objective of the study was to identify the best approach participants in South African market participants can rely on to model and forecast volatility on the Johannesburg Stock Exchange (JSE). The study models and forecasts volatility (conditional variance) using the GARCH (1, 1), GARCH-M (1, 1), EGARCH (1, 1), GJR-GARCH (1, 1), APGARCH (1, 1), EWMA models and the South African Volatility Index (SAVI). The research relies on data on the FTSE/JSE Top 40 Index returns from 2007 to 2013. It was found that returns on the JSE are categorized by volatility clustering, leptokurtosis and asymmetry. The EGARCH (1, 1) model outperformed all other GARCH models in modeling and forecasting the volatility. When compared to the EWMA and the SAVI, forecasts given by the EGARCH model best described out-of-sample realized volatility. Finally, the study found that the explanatory power of the EGARCH model on out-of sample realized volatility is enhanced when its forecasts are combined with forecasts from the SAVI.