摘要：This paper compares the "level" regression of the future spot rate on the current forward rate, which yields a slope coefficient close to unity, to the forward premium puzzle, i.e., a regression of the change in the spot exchange rate on the forward premium, which paradoxically yields a slope coefficient that is frequently negative. We argue that the striking difference between these two otherwise equivalent regressions follows from the existence of a bias together with the non-stationarity of underlying variables. In addition, we contend that non-rationality may potentially explain the existence of the bias that generates the forward premium puzzle.
