THE TERM STRUCTURE OF INTEREST RATES: A TEST OF THE EXPECTATIONS HYPOTHESIS

THE LAST TEN YEARS have witnessed an active interest in the term structure of interest rates. While the proliferation of studies has failed to lead to a widely accepted view, work by D. Meiselman [27], J. Wood [39], F. Modigliani and R. Sutch [28] [29] [30] and others have found evidence tosupport the contention that the term structure is determined by some version of the "expectations" hypothesis.' These studies differ primarily on specific hypothesis about the formation of expectations in the bond market. The expectations hypothesis has also been expanded to include the existence of liquidity, premiums [5] [25]. One of the most novel and important contributions to the term structure discussion in the last several years has been made by R. Roll [32] and T. J. Sargent [36]. Based on the theory of forward prices! developed by P. Samuelson [34] and B. Mandelbrot [26], a very general model of stochastic price change is employed to provide a framework for studying various hypotheses about the term structure. The chief characteristic of the model is the requirement that the bond market be "efficient," to use Roll's terminology. That is, whatever mechanism is used to form expectations, it must be based on all of the available market information. The primary difference between the Roll-Sargent formulation of the expectations hypothesis and earlier formulations resides in the explicit recognition that information be efficiently incorporated into expectations of future rates of interest. According to Sargent, the expectations hypothesis is in reality two hypotheses: (1) the yield structure is determined by expected future rates of interest and (2), expectations are formed within the context of an efficient market. Failure to require that expectations be rational allows the expectations hypothesis to become