Abstract
In this dissertation, we make use of a Realised GARCH (RGARCH) model to estimate
currency exchange rate in the USDZAR market. We compare the Realised
GARCH model to the GARCH and EGARCH models as the benchmarks, whilst
analysing the performance of each model using different distributions. The distributions
used in the analysis are the Skewed Student-t, Student-t and the Normal
distribution. It was found that the Realised GARCH model had a better fit than
the benchmarks with the best performance being seen under a Student-t distribution.
Also, we model currency exchange rate using the ARMA(R,M)-Realised
GARCH(p,q) model under a Student-t distribution and find that the ARMARealised
GARCH model, with a realised volatility proxy, under a Student-t distribution
was used to capture the volatility clustering and mean reversion effects
within the USDZAR market. The AIC, SBIC and MSE were made use of in ensuring
that the forecasting performance was accurate and valid. It was found that the
ARMA(2,1)-RGarch(1,1) model produced the better results than the benchmark
models.