Testing the Unbiased Forward Rate Hypothesis : Evidence on Unit Roots, Co-Integration and Stochastic Coefficients

In this paper, we demonstrate that the conflicting results found in the literature of tests of the unbiased forward rate hypothesis (UFRH) depend upon the econometric specification used as well as differences in the time period of estimation. It is established that the time series properties of spot and forward exchange rate data rule out certain econometric specifications used to test the UFRH. Specifically, we find that both spot and forward exchange rates for the U.K., Germany, Japan, and Canada have unit roots and are co-integrated. We also find that the co-integrating parameter in the regression of realized spot on forward rates for each currency is approximately one, implying that researchers who use this specification to test the UFRH may falsely accept the hypothesis simply because of the specification used in the test. Using an alternative error correction specification, we find that the UFRH is resoundingly rejected for all currencies and that the coefficients in this specification exhibit temporal instability. More importantly, we find that the evolution of the estimated parameters is becoming increasingly inconsistent with the UFRH with the passage of time. We conclude the paper with an investigation into potential causes for the gross violation of the UFRH