Modelling and Forecasting the Metical-Rand Exchange Rate

This paper investigates the ability of the Dornbusch (1976) sticky- price model for the nominal metical-rand exchange rate, over the period 1994:1-2005:4 in explaining the exchange rate movements of Mozambique. Based on the model, we find that there is a stable relationship between the exchange rate and the fundamentals. Gross domestic product and inflation differentials between Mozambique and South Africa play the major roles in explaining the metical-rand exchange rate. However, when the Dornbusch (1976) model is re-estimated over the period of 1994:1-2003:4, and the out-of-sample forecast errors are compared with the atheoretical, Classical and Bayesian variants, of the Vector Autoregressive (VAR) and Vector Error Correction (VEC) models, and models capturing alternative forms of the Efficient Market Hypothesis (EMH) of exchange rates, the sticky-price model performs way poorer. Overall, the Bayesian VEC models (BVECMs), with relatively tight priors, are best suited for forecasting the metical-rand exchange rate, both in terms of out-of-sample forecasting and predicting turning points.