Forecasting Foreign Exchange Rate in Colombia Assuming PPP Conditions: Empirical Evidence Using VAR (in Spanish)

Abstract

In this paper we evaluate a set of colombian exchange rate forecasts during the 1995-2005 period, using a Purchasing Power of Parity Exchange Rate Model (PPPER). Our first finding is that the computed forecasts seem to validate the use of this model under certain conditions given that, theoretically, it does a good work in predicting the long-term behaviour of the nominal exchange rate. Our second finding included a comparison analysis of out-of-sample forecasts (saving the 2001-2005 historical data) between the PPP-based forecast models, and the Vector Autorregresive (VAR) ones. The VAR method has a better forecasting performance, according to the RMSE, MAE and U-Theil measures. However, MAPE results measured on the first and second month-ahead forecasts, indicate that the VAR model has the worst performance amongst PPP-based models.

Note: Downloadable document is in Spanish.

Keywords: Foreign exchange, time-series model, financial forecasting

JEL Classification: F31, C22, G17