The simple econometric models for the exchange rate, according to recent researches, generates the forecasts with the highest degree of accuracy. This type of models (Simultaneous Equations Model, MA(1) Procedure, Model with lagged variables) is used to describe the evolution of the average exchange rate in Romanian in January 1991-March 2012 and to predict it on short run. The best forecasts, in terms of accuracy, on the forecasting horizon AprilMay 2012 were those based on a Simultaneous Equations Model that takes into account the Granger causality. An almost high degree of accuracy was gotten by combining the predictions based on MA(1) model with those based on the simultaneous equations model, when INV weighting scheme was applied (the forecasts are inversely weighted to their relative mean squared forecast error). The lagged variables Model provided the highest prediction errors. The importance of knowing the best exchange rate forecasts is related to the improvement of decision-making and the building of the monetary policy.
