an asymmetric reaction of short term rates to changes in the yield spread; moreover, the response of
interest rates to the spread dynamics appears to be conditioned on the level of the term premium.
We thus provide significant evidence that the informative content of the term structure is related to
the uncertainty about the future conduct of monetary policy.
It has been noted by Laurent (1988, 1989), Bernanke and Blinder (1992), that the yield spread, i.e.
the difference between long and short term interest rates describes the expectations regarding the
incoming monetary policy stance. The monetary authority can easily influence the short end of the
yield curve, while long term rates are generally market-driven and do not react hastily to policy
interventions. Furthermore, the long term rate can act as a proxy for the equilibrium short term rate,
i.e. a Wicksellian natural rate; so that, the yield spread can be thought as a measure of the relative
tightness of policy. In this paper, we focus on the unexpected component of the yield spread, i.e. the
term premium, which is regarded as a measure of monetary policy surprise. It has been shown that
the yield spread can be decomposed into two elements, an expectations-based component and a
term premium, which may be thought as the sum of a liquidity premium and a risk premium
(Campbell and Shiller, 1991; Hamilton and Kim, 2002; Favero, Kaminska, and Soderstrom, 2005;
Rudebusch, Sack, and Swanson, 2007). In this paper we assume that the term premium is not only
time-varying, but also regime dependent; in particular, we allow the term premium to determine
distinct regimes, i.e. different states of the world, in which the EH is examined.
The multiple regime model allows us to assess whether the Mankiw and Miron (1986) view, that
the increased unpredictability of short term rates affects the predictive power of the spread, is
supported by data. Moreover, the threshold model is also suitable to verify the idea put forward by
Campbell (1995) who argues that the term structure is more informative about future interest rates
movements when the variability of the expected changes in short rates is higher than the variability
of the term premium. Finally, the two-regime framework provides a useful apparatus to test the
thesis put forward by Thornton (2004); he points out that also the relative variance of short to long
rates is empirically relevant for the success of the EH. Our empirical findings suggest that both the
volatility of short rates and the relative variability of the theoretical spread to the term premium
matter in the empirical analysis of the expectations hypothesis.
The rest of the paper is organized as follows. In the next Section we discuss a selected survey of the
literature. Section 3 describes data. In Section 4 we discuss the empirical model and provide
preliminary evidence of non linearity in the expectations equation. In Sections 5 we present the
empirical results. Section 6 concludes.