Exchange rate forecasting model performance

In this research, the performance of fundamental exchange rate models, as UIRP and PPP, will be compared, in- and out-of-sample, with each other, with a Taylor Rule based model and with the famous Random Walk model. Former studies show that no single model beats the Random Walk. In this research we see that the Random Walk model still performs better than the PPP, UIRP and Taylor-based model according to their RMSE, MSE and MAE’s. However, the Diebold-Mariano statistic gives a different picture. All models perform better than the Random Walk model over the short and long term forecast horizon according to the Diebold-Mariano squared loss differential statistic.