The Bias in the Conventional Test of the Expectations Theory: Resolving the Anomalies at the Very Short End of the Term Structure

While the expectations theory of the term structure of interest rates plays an important role in economics and finance, it has received relatively little empirical support. The most frequently used test, which I call the conventional test, has provided some hope for the expectations theory. When the conventional test is used, the slope of the yield curve predicts the correct direction of long-term changes in the short-term rate and explains a significant proportion of such changes. The failure of the conventional test to accept the expectations hypothesis is usually attributed to the potential for bias from several sources. Rather than explaining the failure of the conventional test; however, this paper accounts for it relative success. Motivated by research indicating that the conventional test yields results that are more favorable to the expectations hypothesis when it is less likely to hold and less favorable when it is more likely to hold, I show that the conventional test is generally biased in favor of the expectations theory. Indeed, I show that this bias led Hardouvelis (1988), Simon (1990) and Roberds, Runkle and Whiteman (1996) to accept the expectations theory in circumstances when it is unlikely to hold. Monte Carlo experiments confirm the low power and bias of the conventional test of the expectations theory. JEL Classification: E40, E52