2. Literature Review
The expectations hypothesis states that the long term yield can be expressed as the average of
expected future short term yields1. The expectations theory, however, has always found little
empirical support. Campbell and Shiller (1991) conclude “we thus see an apparent paradox: the
slope of the term structure almost always gives a forecast in the wrong direction for the short term
change in the yields on longer bonds, but gives a forecast in the right direction for long term
changes in short term rates”. The weak empirical support for the expectations hypothesis2 has thus
inspired numerous empirical studies.
Campbell and Shiller (1991) and Hardouvelis (1994) suggest that the empirical failure may be due
to an over-reaction of long rates to the expected change in short rates. In addition, Hardouvelis
(1994) believes that large measurement errors can account for the forecast in the wrong direction of
long term rate prediction. Fama (1986), Cook and Hahn (1989), Lee (1995), Tsavalis and Wickens
(1997), among others, argue that a time-varying term premium correlated with the spread can
account for the empirical failure of EH. Froot (1989), however, indicates that a violation of the
rationality principle, rather than a time-varying risk premium, is one of the main reasons underlying
the rejection of EH. McCallum (1994) warns that the rejection of EH might be simply due to a
misspecification of the equation used to test the theory. In particular, he believes, as we do, that
single-equation models may be inappropriate to test the EH. He points out the traditional Campbell-
Shiller equation is misleading to think in terms of the predictive power of the spread. He shows that
estimates of the slope coefficients are inherently lower than one when allowing for both a time-
varying risk premium (first-order stationary autoregressive process) and a specific monetary policy
rule that features interest rate smoothing and responds to the spread dynamics.
Mankiw and Miron (1986) provide with a suggestive explanation of the inability of the spread to
predict future movements in interest rates. They show that the slope of the yield curve seems to
have substantial predictive power to anticipate future short term rate dynamics before the creation
of the Fed. Between 1890 and 1915 the high predictability of short term rates was due to a clear
mean reverting behaviour displayed by the short term policy rate; however, after 1915, the interest
rate smoothing policy pursued by the monetary authority has enhanced the difficulty of forecasting
short term rates, reducing the predictive power of the spread. The random walk path followed by the
1 The expectations hypothesis implies both that the forward rate equals the future spot rate and that the expected holding
period return is constant, i.e. equal, on bonds of all maturities.
2 Thornton (2003) has introduced a suitable terminology to distinguish the empirical testing of the expectations
hypothesis. The regression for predicting short term rate changes over the life of the long term bond is labelled the
conventional test of EH. This test returns a positive estimation of the slope coefficient although less than one; the
conventional test thus gives a forecast in the right direction for long term changes in short rates. The equation to predict
long term rates is called the contrarian test, since it returns negative estimates of the slope coefficient, i.e. a forecast in
the wrong direction. In this paper we focus only on the conventional test of the EH.