Adaptive forecasting of exchange rates with panel data

Abstract

This article investigates the statistical and economic implications of adaptive forecasting of exchange rates with panel data. The candidate exchange rate predictors are drawn from (i) macroeconomic ‘fundamentals’, (ii) returns/volatility of asset markets, and (iii) cyclical and confidence indices. The proposed forecasting strategy exploits information from many dimensions, since it generates alternative exchange rate forecasts at various horizons from each of the potential predictors using single market, mean group and pooled estimates by means of rolling window and recursive forecasting schemes. The capabilities of single predictors and of alternative adaptive techniques for combining the generated exchange rate forecasts are evaluated robustly by means of statistical and economic performance measures. The results show that combining exchange rate forecasts generated from a wide range of information sets reduces ex-ante uncertainty, improves forecasting precision and leads to better market timing than most single predictors.