NUMEROUS EMPIRICAL STUDIES of the term structure of interest rates have been published in recent years. Yet despite the considerable interest in the field, little effort has been devoted to a careful comparison and evaluation of the alternative hypotheses, alternative specifications, and diverse results which have been reported in the literature.' In general, the authors of empirical studies have remained content to estimate the parameters of their models and to test them for broad consistency with the data under examination. Unfortunately, it is not possible to compare these alternative models directly by examining the published statistical results since each author has chosen to examine a different sample of data. Their studies span different time periods. Moreover, some have used yearly data while others have employed quarterly, monthly or weekly observations. Most of the empirical studies have concentrated upon the bond markets in the United States, though English and Canadian data have also been analyzed. The studies have dealt with both corporate and government securities and have drawn from a variety of sources for their time series of observations. It is the purpose of this study to evaluate the alternative models and specifications by re-estimating each hypothesis using a consistent body of data. Of course, it will not be possible to prove the "correctness" of any given formulation, since a model that explains one set of observations may fail to explain another set in some subsequent experiment. On the other hand, it will be possible, as we shall see, to reject a number of formulations as inconsistent with our observations of the world and to choose among those remaining the model that best explains the data.
I. THE RoLE OF EXPECTATIONSWith few exceptions there appears to be a basic agreement that expectations of