Content area
Full Text
Barbells and bullets do not always perform as expected when the yield curve changes shape.
Duration is a useful metric for assessing a bond portfolio's sensitivity to a parallel shift in the yield curve. When the yield curve shift is not parallel, however, two bond portfolios with the same duration may not experience the same return performance. To evaluate differences in expected performance across portfolios, it is therefore necessary to quantify the price impact due to changes in shape, as opposed to merely shift, of the yield curve.
Litterman and Scheinkman [1991] suggest that changes in shape are due mostly to parallel shifts, changes in slope (i.e., twist), and changes in curvature (i.e., humpedness or butterfly changes). Jones [1991] and Willner [1996] indicate that these three types of changes in the yield curve's shape (level, slope, and curvature) are not independent, so correlations among them must be considered when estimating the bond portfolio's expected performance.
Bond portfolio managers pursuing active strategies must therefore consider all three yield curve parameters as well as correlations among changes in these parameters. Consider, for example, yield curve strategies that involve structuring a portfolio to exploit expected changes in the yield curve's shape.
Three basic yield curve strategies are: 1) bullet strategies; 2) ladder strategies; and 3) barbell strategies (see Fabozzi [1996]). A bullet strategy calls for investing in a bullet portfolio, which is constructed so that maturities of the bonds are concentrated at one particular point on the yield curve. A ladder portfolio is constructed so as to have approximately equal amounts at each maturity across a range of maturities (e.g., approximately 10% in each of ten bonds ranging in maturity from one to ten years). A barbell portfolio is constructed by concentrating investments in two different-maturity bonds at extreme ends of the yield curve (e.g., approximately equal amounts in a two-year bond and a twenty-year bond).
We examine the relative performance of these strategies under anticipated changes in the level of the yield curve, and the corresponding implied changes in its slope and curvature. First, using data for U.S. Treasury yield curves, we estimate the three shape parameters of the yield curve (level, slope, and curvature). Second, we estimate how changes in the parameters are correlated with...