This gaper examines the unbiased forward rate hypothesis using an Erro
r Correction Model (ECM). Previous tests have utilized mainly two spec
ifications-differences and levels. The advantage of the ECM is that th
e parameters of the level specification appear in the error correction
term, and hence the ECM provides a direct link to the levels specific
ation. In this paper we fully exploit this link and demonstrate that b
oth specifications yield essentially the same results when the ECM is
explicitly derived from a levels specification and the levels paramete
rs are estimated simultaneously with other parameters of the ECM.