Does the Expectation Hypothesis Hold at the Shortest End of the Term Structure?

This paper examines the predictability smile at the shortest end of the term structure. The existence of a predictability smile has been well documented: spreads between long rates and short rates are able to forecast subsequent movements in interest rates well, provided the horizon is three months or less or two years or more. The predictive power of the spread at the shortest maturities, however, has not been adequately investigated. This is a potential shortcoming of the existing literature as a projection of the predictability smile to the shorter maturities is not a guarantee that the expectations hypothesis holds. In Japan, a positive spread between the forward and the spot rates has insufficient predictive power for the future spot rate innovations, while a negative spread has near-perfect predictive ability. Further, we provide evidence that this result is not unique to Japan, as we find this "asymmetric predictability" to be a feature of the very short-term money markets of the U.S., U.K. and Italy.