Abstract
M.Com. (Investment Management)
The focus of this study is primarily based on the significance of forecasting volatility on the JSE Limited. The study investigates the appropriateness of using volatility models to forecast volatility on the Resource 10 (RESI), Financial 15 (FINI), and Industrial 25 (INDI) FTSE/JSE sector-indices classified according to the Industry Classification Benchmark (ICB).
This study uses historical closing values of the three FTSE/JSE indices which are then converted into log returns. Quantitative data are used to investigate whether volatility on the RESI, FINI, and INDI FTSE/JSE indices is correctly specified by ARCH class of models. The data are obtained from McGregor I-NET BFA databases and spans the period from 17 February 2006 to 16 February 2016. The 10 year period is also divided into two 5 year sub-periods and five 2 year sub-periods for each FTSE/JSE index.
This study employs the Autoregressive Conditional Heteroscedasticity (ARCH) model, the Generalised Autoregressive Conditional Heteroscedasticity (GARCH) model, and the Threshold (Generalised) Autoregressive Conditional Heteroscedasticity (TARCH) model. These models are used to generate in-sample forecasts of volatility on the three aforementioned FTSE/JSE indices. The performance of the volatility models used in this study is evaluated based on three statistical loss functions: the root mean squared error, mean absolute error, and the mean absolute percent error.
The results of this study evidence the presence of ARCH effects in the data of the three FTSE/JSE indices. The ARCH, GARCH and TARCH specifications are statistically significant for all indices; though there are some sub-periods of each of the FTSE/JSE indices which show no statistical significance in the parameter estimates of the volatility models employed. There is also evidence of volatility asymmetry in all of the FTSE/JSE indices considered in this study. There is no single superior volatility model between all three ARCH models that specifies the volatility of the FTSE/JSE indices over all the others when the forecasts are evaluated based on the statistical loss functions. However, the TARCH model outperforms the ARCH and GARCH models in most cases. This means that accounting for asymmetries in volatility is important in generating reliable volatility forecasts.