Assessing the Forecasting Performance of Structural Models for the Nominal Exchange Rate : The Colombian Case *

This paper analyses the out-of-sample forecasting performance of three models for the USD/COP nominal exchange rate during the period 1984:I – 2004:I. The sticky price monetary (Dornbusch (1976) – Frankel (1979)) and the Balassa–Samuelson (which gives a central role to the productivity differentials) approaches are used. Additionally, the Purchasing Power Par ity condition (PPP) is analyzed. The forecasting ability of these models is comp ared using a random walk as a benchmark model. The metrics employed in evalua ting the forecasting performance are RMS, MAE, MAPE and U-Theil. It is found that despite the great ability to predict, no model outperforms the random walk. This conclusion strengthens the previous results in the nominal exchange rate modeling literature.