We show that the structural models' out-of-sample predictive performance of the dollar/pound, dollar/mark, dollar/yen, and trade-weighted dollar exchange rates is inferior to that of the simple random walk. However, the application of a Kalman filter to the structural models, which corrects the time-varying properties of coefficients over time, improves the predictive performance of exchange rate models. Our findings suggest that the coefficients of time-series models and structural models which use time-series data have shifted over time. Thus, a model that utilizes this information may increase its predictive performance in in-sample and out-of-sample forecasts.
ASJC Scopus subject areas
- Business and International Management
- Applied Psychology
- Management of Technology and Innovation