The Term Structure of Interest Rates: Some Additional Evidence

Abstract

This paper re-examines the evidence relating to the term structure of interest rates. In Section I the explanatory power of Modigliani and Sutch's Preferred Nabitat Model [12] is compared with that of a model developed by Wood [14]. Both models assume that shortand long-term U.S. Government securities are close substitutes. The principal diJ%erence between them is the importance attributed to past interest rates as an indicator of expected future rates. According to the Preferred Nabitat Model, the expected change in future rates may be expressed as the diJ%erence between the current long-term rate and a weighed average of past long rates. The Wood model assumes that at any point in time the best estimate of next period's short-term rate is the current rate; past rates are irrelevant. The data used are monthly and quarterly observations of the yields on three-month Treasury bills and long-term U.S. Government bonds for the period 1951-65. The results favor the Wood model. They provide little evidence that past interest rates are a very useful indicator of expectedfuture rates. In Section II we examine the stability of the Wood model. A regression equation relating the first diJ%erence in the yield on long-term bonds to thefirst diJ%erence in the bill rate isfitted to monthly observations for the period 1920 45 and nine shorter periods. The results are quite striking. They indicate that essentially the same relationship holds for the 1920's and early 1930's, the pre-Accord