Fundamentals, regimes and exchange rate forecasts: Insights from a meta exchange rate model

Abstract

A ‘meta’ model of the exchange rate combines a range of models distinguished by the drivers of the rate and by regime duration. Alternative model weights are proposed, including those obtained from a novel non-nested hypothesis-testing technique that accommodates periods of stability and slowly-evolving or abruptly-changing regimes involving multiple drivers. Focusing on density forecasts, the meta models perform well, demonstrating that all the sets of fundamentals considered can be useful for forecasting when the model is estimated over an appropriate time frame, but that the ability to exploit the changing relevance of different sets of fundamentals over time is important too.