The expectations hypothesis of the term structure of interest rates states that the long-term interest rate is an average of the current and expected future short-term interest rates and a constant expected term or risk premium required by bond market participants for investigating in the long-term bond. The forward rate implied by the term structure reflects market expectations of the future short rate and the constant premium. Recent tests of the expectations hypothesis also assume rational expectations on the part of market participants. Most empirical studies of the expectations hypothesis for U.S. interest rates reject the joint hypothesis. The rejection can be attributed to: (a) a time-varying risk premium, or (b) expectation errors by market participants. This dissertation examines the expectation errors argument for the failure of the hypothesis test. In particular, I analyze the impact of learning about a past shift in the interest rate regime on market expectations of the future interest rate over a small sample.; I examine monthly data on U.S. Treasury bill rates over the period from January 1960 to December 1989 to determine various interest rate regimes, and test for regime shifts. My evidence shows that there are several shifts in the interest rate regime. Then, I develop a Bayesian learning model to explain the behavior of market expectations of the future interest rate over the three-and six-month forecast horizon during the period, October 1979 to September 1982. Using a univariate forecasting equation, I find that 20 percent of the variation in the forward rate prediction errors over the six-month horizon can be explained by the model. I extend the analysis to determine if learning about a shift in inflation can explain the forward rate prediction errors over the same period. I use a generalized version of the Fisher equation to determine the impact of learning about a shift in inflation on the expectations of the future short rate. Learning about a shift in inflation can, however, only account for 7 percent of the variation in the forward rate prediction errors. In general, learning about interest rate regimes is important in explaining the term structure over small samples.
展开▼