Modeling Changes in Daily $A Exchange Rates: An Application of GARCH

This paper examines the statistical properties of the logarithmic changes in daily $A exchange rates. For all five exchange rates considered, changes exhibit volatility clustering and highly significant non-linear serial dependence. GARCH models in various forms were estimate and the results are that there are significant GARCH effects and an unexpected change has asymmetric effects on the future volatility of changes in all five exchange rates. There is some evidence of the day of the week effect in the variance of changes, and the announcements of economic news have a significant effect on changes but the effects on the variance of changes are generally insignificant except in the case of USD/$A rate. In general, the GARCH modelling of logarithmic changes in daily $A exchange rates was found to be useful as shown by the significant reductions in the skewness and excess kurtosis and the non-linear serial dependence of the estimated standardised residuals.