New Insights on the Forward Premium Regression through a Portfolio-based Approach

Abstract

This study proposes a portfolio-based forward premium regression to estimate its time-varying coefficients and to conduct simultaneous inference. To this end, we sort currency portfolios on forward premiums. The empirical results show much weaker evidence of Uncovered Interest Parity (UIP) breakdown for the currency portfolios than for individual currencies. The main implication is that the effects of two sources of heterogeneity—exposure to currency-specific risk and exposure to common risk—on the forward premium anomaly diminish with diversification through portfolio construction. The study also illustrates that the U.S. fundamentals have persistent predictive power for the risk premium of each portfolio.