ABSTRACT

The real exchange rate of South Africa can be forecasted using the direct or the
indirect methods of forecasting. This article compares the forecasting results of
direct and indirect forecasting of the real exchange rate by using two univariate
models and a multivariate model. The direct models outperformed the indirect
models in-sample and the indirect models generally outperformed the direct
models out-of-sample. Given the closeness of the forecasting results, the modeller
should decide whether it is worth the effort to forecast the real exchange rate
indirectly if similar results can be obtained from a (less time-consuming) direct
method.