Learning and the Long Rate

In this paper we seek to explain the behavior of the long term interest rate in the U.S. over the last 50 years using the idea that investors had only limited information about the process generating short term interest rates. Investors are assumed to behave according to the expectations hypothesis of the term structure but use econometric methods to learn about the appropriate model for short term interest rates and form expectations about future short rates. In particular, they use methods for identifying structural breaks. Although the approach seems reasonably “rational” from an econometric perspective, it differs in important ways from traditional formulations of Bayesian learning. Our approach yields considerably more successful predictions about the behavior of the long term interest rate than existing rational expectations based explanations. In particular, our model provides an explanation for why the long term interest rate has tended to fall when the yield spread is high.