Does Spread Really Predict the Short Rate? Explaining Empirical Anomalies in the Expectations Theory

Empirical studies often find that the spread between longer and shorter rates does not have predictive power for future longer rates, violating the Expectations Theory (ET). Although the predictive power of the spread for future shorter rates is largely in accordance with the ET, especially when the forecast period is long, researchers often find this holds to varying degrees across samples (country-wise or time-wise). We show this pattern may be due to the powers of all tests depending on interest rates’ maturities and their persistency in small samples. This paper also compares the powers of tests of the ET against the under/overreaction and the time varying term premium alternatives across various maturity combinations, levels of persistency and sample sizes. Tests perform best and are comparable to each other at the shortest end of the term structure, but deteriorate as the distance between maturities of longer and shorter rates increase. However, this deterioration is of varying degrees for different tests and its speed diminishes as we depart from the shortest end. In general Lagrange multiplier and distance metric tests emerge as being the most powerful and least sensitive to interest rate maturities and their persistency.(This abstract was borrowed from another version of this item.)