Imperfect Knowledge, Temporal Instability and an Uncertainty Premium: Towards a Resolution of the Excess-Returns Puzzle in the Foreign Exchange Market

This paper offers a refinement and explores a resolution of the excess-returns puzzle in the foreign exchange market. We find that the predictions of the forward premium are not negatively biased throughout the three decades of floating, as commonly believed, but rather are sometimes positively biased, negatively biased, unbiased or possess no predictive content depending on the subperiod examined. To explain this modified puzzle, the paper makes use of a recently developed model of the risk premium, which we have called an aggregate uncertainty premium. Our model employs an alternative approach to modeling exchange rate expectations, dubbed Imperfect Knowledge Expectations (IKE), which recognizes that rational agents do form expectations based on imperfect knowledge. Our model also makes use of a dynamic extension of the assumption of myopic loss aversion. We find that our IKE-based approach can account for the pattern of positive and negative biases estimated over three decades of floating rates.