Abstract
This paper illustrates the interplay between theory development and data analysis by considering the ability of the rational expectations hypothesis to explain the empirical cointegration structure found in the term structure. It finds that although a standard no-arbitrage theory that incorporates rational expectations can explain some of the properties of Treasury Bill yields, this theoretical explanation is incomplete. A broader-based explanation that accounts for government debt and time-varying risk premia can improve predictions of yield movements, relative to those predictions based solely on a bill yield spread.