Title: Exchange rate predictability and model incompleteness
Authors: Knut Are Aastveit - Norges Bank (Norway) [presenting]
Herman van Dijk - Erasmus University Rotterdam (Netherlands)
Francesco Ravazzolo - Free University of Bozen-Bolzano (Italy)
Abstract: It is well known that exchange rate fluctuations are very difficult to predict using economic models, and that a random walk forecasts exchange rates better than any economic model (the Meese and Rogoff puzzle). The recent literature has identified a series of fundamentals/ methodologies that claim to have resolved the puzzle. However, although these studies find predictors that provide exchange rate predictability for some countries at specific time periods and specific forecast horizons, typically the relationships, and thus the predictability, are unstable and break down over time. We employ a combined density forecasting framework that allows us to pin down several sources of instability that might affect the out-of-sample forecasting performance of exchange rate models. The latent weights of the combination scheme depend on past forecasting performance and other learning mechanisms and allows for model incompleteness. We show that allowing for these features systematically improve upon the random walk benchmark in an out-of-sample forecasting exercise.