Predictability of Exchange Rate Models: Statistical and Economic Evaluation Approach

This paper seeks to compare the exchange rate predictability of economic fundamental models, including the Uncovered Interest Parity, the Purchasing Power Parity, the Monetary Fundamentals, the Symmetric and Asymmetric Taylor Rules with the benchmark Random walk with drift. Using out-of-sample regression method for monthly returns on eight exchange rates against the US dollar, I compute their statistical and economic values to compare with the benchmark in both rolling and recursive scheme. From the statistical perspective, predictive models generally are unable to outperform the Random walk in out-of-sample forecasting. However, exchange rate investment strategies based on the forecasts from a couple of models do generate economic values for investors. Results are robust when a single currency is excluded from the original portfolio