Exchange rate equations based on interest rate rules : in-sample and out-of-sample performance

Using exchange rate data on five currencies vis-à-vis the US dollar, this paper examines the in-sample and out-of-sample performance of exchange rate equations derived from alternative empirical and optimal interest rate rules. These rules could have either homogeneous or heterogeneous response coefficients. Our exercise shows that these exchange rate equations do not offer good in-sample explanatory power consistently across currencies and over time. The relative forecasting performance of these exchange rate equations tend to vary across currencies and over time and bears limited relationship with the relative in-sample performance. When the forecast performance is compared with a random walk model, these exchange rate equations offer no better performance under the usual MSFE criterion but are better when the ability of predicting the direction of change is considered.