77 Pages Posted: 12 Jun 2014 Last revised: 21 Jul 2017
Date Written: June 26, 2017
Using methods from machine learning we show that fundamentals from simple exchange rate models (PPP or UIRP) consistently allow to improve exchange rate forecasts for major currencies over the floating period era 1973--2014 at a 1 month forecast and allow to beat the no-change forecast. "Classic" fundamentals hence contain useful information and exchange rates are forecastable even for short forecasting horizons. Such conclusions cannot be obtained when using rolling or recursive OLS regressions as in the literature. The methods we use -- sequential ridge regression and the exponentially weighted average strategy both with discount factors -- do not estimate an underlying model but combine the fundamentals to directly output forecasts.
Keywords: exchange rates, forecasting, machine learning, purchasing power parity, uncovered interest rate parity, monetary exchange rate models
JEL Classification: C53, F31, F37
Suggested Citation: Suggested Citation
Amat, Christophe and Michalski, Tomasz Kamil and Stoltz, Gilles, Fundamentals and Exchange Rate Forecastability with Simple Machine Learning Methods (June 26, 2017). HEC Paris Research Paper No. ECO/SCD-2014-1049. Available at SSRN: https://ssrn.com/abstract=2448655 or http://dx.doi.org/10.2139/ssrn.2448655