Which alternative to choose: does the excess sensitivity hypothesis or a time varying term premium explain the failure of the rational expectations hypothesis of the term structure?

Contrary to the predictions of the rational expectations hypothesis of the term structure, empirical evidence suggest that the term spread between long and short rates fails to forecast future long term rates although its forecasts of future short term rates are in the correct direction. This paper examinines which of two popular alternative hypotheses - the excess sensitivity of the long term rate to the contemporaneous short term rate and the rational expectations hypothesis allowing for a time varying term premium - can explain the failure of the pure version of the expectations theory. Using US data at the short and long end of the term structure, the results clearly reject the excess sensitivity hypothesis, and suggest that the anomalies of the term spread can be attributed to the different way that the term premium affect each of the term spread models. It is found that the term premium influencing the term spread model forecasting future long rates is much more volatile than the term premium of the term spread model forecasting future short term rates. In order to remove the term premium biases from the slope of the term spread a simple model of the term premium is employed.