South african exchange rate after 2000s: an econometric investigation

This paper is an econometric investigation, an analysis on the difficulty of modeling the south african exchange rate. The aim of our paper is to examine the nature of the existing relationship between the real exchange rate of the Rand, real prices of gold, platinum and the real interest rate differential through different empirical models, and this over the period going from January 2000 to September 2014. Our analysis shows that, over the same period, with different empirical methods, the variables used in the paper can be the determinants of the real value of the south african Rand, but at different horizons. To achieve our goals, long run (Engle and Granger (1987) ; Johansen (1988)) and short run (VAR process (Sims (1980))) analysis have been performed. We come to the conclusion that, the determinants of the Rand change according to the methods used and these do not therefore allow us to have robust results. The long run analysis performed by Engle and Granger approach result to a lack of long run relationship among our variables. To have an robust idea on the lack of long run relationship, we have performed another long run analysis: the vector error correction model (VECM approach) of Johansen which results on the existing of one co-integrating relation among real value of te Rand and their determinants. However, because of the lack of long run relationship resulting of the Engle and Granger approach, we have performed a short run analysis with the vector autoregressive process. We find that only the real platinum price in our study is a short term determinant of the real value of the Rand. The real impact is effective only at the end of the first quarter with a real appreciation of the Rand. The main surprise is the absence of impact of real price of gold shock on the real value of the Rand. Analyze the south african exchange rate through one empirical method/model to find theirs determinants can be biased.