6. Concluding Remarks
This paper deals with the empirical failure of the expectations hypothesis; we suggest that single-
equation models might be inappropriate to analyse the expectations theory. We thus provide
evidence to support the view that non linear models, threshold models in particular, are useful to
examine the informative content of the term structure of interest rates. Economists have provided
evidence suggesting that time-varying term premia may be responsible of the weak empirical
support for the expectation hypothesis. In the expectations equation the term premia effect is
captured partially by the residual component and partially by the intercept of the model. A shift in
the intercept generates a bias in the estimation of the slope coefficient thus affecting the
effectiveness of linear model to test the EH.
We propose a multiple regime framework to analyse the predictive power of the yield spread.
Regimes are assumed to be a function of term premia, which provide with a measure of both the
unexpected stance of monetary policy and agents’ attitude towards risk. We thus suggest that the
expectations model may well be informative in a framework that exploits the countercyclical
behaviour of term premia. We extend previous works by Campbell and Shiller (1991) providing
evidence that the predictive ability of the yield spread is contingent to the level of uncertainty that
reigns in the economy. Results suggest that the informative content of the slope of the term
structure increase substantially once the risk-averse attitude of economic agents is taken into
account. Our results do suggest the presence of important asymmetric effects also in the prediction
of future short term interest rates.
27