Exchange rate models and parameter variation: The case of the Dollar-Mark exchange rate

Abstract

Iti a recently published paper, Wolff (1987) used varyingparameter estimation techniques, based on recursive application of the Kalman filter, to improve the predictive performance of monetary exchange rate models. Allowing estimated parameters to vary over time was found to enhance the models' forecasting performance for the Dollar-Pound, Dollar-Mark and Dollar-Yen exchange rates. Contrary to earlier results in the literature (Meese and RogofT, 1983), ex-post forecasts for the Dollar-Mark rate compared favorably with those obtained from the naive random walk forecasting rule. In this paper the Wolff (1987) results are reexamined and extended in two directions. First, test statistics are constructed to test whether the monetary models do significantly better than the random walk for the Dollar-Mark case. Second, a true ex-ante forecasting experiment is performed in order to test whether the earlier finding for the Dollar-Mark exchange rate remains valid in a context where exchange rate forecasts are generated only on the