A defence of the expectations theory as a model of us long-term interest rates

This paper re-examines the empirical content of the expectations theory of the term structure by employing the Campbell-Shiller (1987) methodology to study the behaviour of 10-year/three-month US government yield spreads. The methodology is implemented in two ways. First, theoretical spreads satisfying the expectations theory are constructed from in-sample forecasts of future changes in short rates generated from a small-scale VAR. Second, theoretical spreads are computed from out-of-sample forecasts of changes in short rates with the parameters of each VAR equation updated with a Bayesian procedure. When the procedure is restricted to give less weight to new data than would be the case with OLS estimation over an expanding sample, theoretical spreads computed from out-of-sample forecasts track actual spreads closely in pre-1979 data. This is also the case as from the start of 1984 if data from the 1979Q4-1982Q4 period of non-borrowed reserve targeting are given zero weight when estimating the parameters of the VAR.