Robert Kremer, Sherrill Shaffer
Using several samples of forward exchange rate forecasts for the British pound vs. the US dollar, this article explores the post-sample predictive performance of adjusting the forecasts for recent empirical bias. Numerical accuracy is assessed via both parametric and nonparametric tests, and directional properties are also evaluated. The evidence suggests that simple linear adjustments can yield significant improvements in predictive accuracy, even if the measured bias in the original forecasts is not statistically significant.