Can structural changes in the persistence of the forward premium explain the forward premium anomaly?