The Unbiasedness Hypothesis from a Panel Perspective

Risk premia, peso-problems and market-inefficiencies have been suggested as candidate explanations for the apparent rejection of the unbiased hypothesis. If various explanations interact, a panel approach is called for. In this paper we estimate different panel models, that allow for cross-sectional dependence of exchange rates, for fifteen currencies between 1979 and 1996. We show that the deviation from uncovered interest parity is smaller than commonly presumed. Estimates of the slope coefficients of the forward premium appear to be positive but still significantly different from unity. In addition it is shown that this coefficient is close to unity if only five to ten percent of the largest changes in the forward premium are taken into account. These findings point to the importance of peso-problems and inactivity bands as explanations for the apparent rejection of the uncovered interest parity relationship