Short-run exchange rate dynamics in South Africa: a microstructure approach

Abstract

This dissertation investigates the short-run variation of the South African Rand versus the US Dollar exchange rate. A market microstructure approach to exchange rate determination is adopted and the central hypothesis is that customer order flow is an important determinant of short-run variation in the exchange rate. Two key objectives are examined. Firstly, the paper looks at the in-sample explanatory power of the order flow as a key determinant of exchange rate variation. Secondly, the practical significance of order flow is evaluated by looking at its out-of-sample forecasting accuracy. The findings from this study suggest that order flow can explain some of the variation in exchange rate; however, a large portion of the exchange rate variation remains unexplained. Further, the out-of-sample performance of the order flow model is found to be inferior to that of the GARCH model for periods up to 6 months and inferior to the naïve Random Walk model at 12 months horizons. The evaluation of a Vector Autoregression model suggests that there are feedback effects from exchange rate returns to order flows. This means that Ordinary Least Squared estimates could be biased and the results thus inaccurate and misleading