Regime-specific exchange rate predictability

Abstract

We study temporary phases of exchange rate predictability in a two-regime threshold predictive regression framework allowing for persistent predictors. Regime switches are triggered by an observable transition variable which relates to media news, expectations, uncertainty and global financial conditions. As predictors for G7 currencies and effective US-Dollar exchange rates, we study various interest rate spreads, yield curve factors, uncertainty measures and deviations from fundamental exchange rate parities. Besides established uncertainty measures, we use a wide range of measures for media coverage and construct uncertainty measures from survey data as transition variables for the activation of the predictability regime. Our results emphasize that short recurring phases of significant predictability are characterized by nonlinear patterns. Phases of predictability are triggered by increased media coverage and high uncertainty with interest rate dynamics emerging as the most important predictor. We find broadly similar results for a contemporaneous threshold analysis where our regressors are allowed to affect the exchange rate in the same period. From a theoretical point of view, we argue that our empirical results are useful for the empirical identification of scapegoat effects and that media coverage and uncertainty affect the exchange rate via the heterogeneity of private signals and the precision of public signals.