ABSTRACT
The use of principal components is a commonly applied statistical technique in the finance literature to determine a reduced number of factors or sources of uncertainty to describe the stochastic movement of the term structure of interest rates over time. Using weekly observations from 1984 until 1988 in the case of US Treasuries, for example, Litterman and Scheinkman (1988), conclude that, based on principal component analysis, a three-factor model explains at least 98 per cent of the variability of excess returns of any zero coupon bond. The first factor essentially represents a parallel change in yields, while the second and third factors represent a change in the steepness and curvature of the yield curve, respectively. These findings have been confirmed by several authors, for example, Garbade (1986), Dybvig (1989) and Heath et al. (1990b).
