he readily notes that the empirical evidence is far from conclusive.We examine differences in the effective duration, effective convexity, and the option-adjusted spread resulting from different one-factor no-arbitrage interest rate models. The models considered are: the Ho and Lee (HL) [1986] model; the Kalotay, Williams, and Fabozzi (KWF) [1993] model; the Black, Derman, and Toy (BDT) [1990] model; the Hull and White (HW) [1994]
I. NO-ARBITRAGE INTEREST RATE MODELSThe interest rate models we examine assume that the short-term interest rate follows a certain process that can be represented by a stochastic differential equation. All the interest rate models are special cases of the general form of changes in the short-term rate:( 1) where f and g are suitably chosen functions of the short-term rate and are the same for most models presented here, θ will be shown to be the drift of the short-term rate, and ρ is the mean reversion term to an equilibrium short-term rate. The term σ is the local volatility of the short-term rate, and z is a normally distributed Wiener process that captures the randomness of future changes in the short-term rate.Equation (1) is a one-factor model that gives only the short-term rate (one factor).2 Its first component (the dt term) is the expected or average change in the short-term rate over a short period of time. The second component is the risk term, as it includes the random component dz. All the interest rate models we consider are special cases of Equation (1).
The Ho-Lee ModelThe Ho-Lee model assumes that changes in the shortterm rate can be modeled using Equation (1) by setting f(r) = r and ρ = 0, so that the process for the short-term rate is:Since dz is a normally distributed Wiener process, the HL process is a normal process for the short-term gu @ +w,gw . +w,g} gi+u+w,, @ ++w, . +w,j+u+w,,,gw . +u+w,> w,g} rate. As can be seen from Equation (2), the short-term rate may become negative if the random term is large enough to dominate the drift term (dt). This is a serious shortcoming of the HL model, although it is argued that as long as the HL model provides good prices for bonds with embedded options, it does not matter if some of its assumptions are unrealistic. Another possible drawback of the model, however, is that the volatility of the short-term rate does not depend on the level of the rate, and the short-term rate does not meanrevert to a long-term equilibrium rate, as many practitioners believe would hold in reality.Some of these restrictive assumptions are relaxed in the other models. Note that the distributional properties will have a tendency to bias the values of the embedded contingent claim.The simplicity of the HL model combined with the fact that it provides reasonable prices under many circumstances makes it a very popular interest rate model.
The Kalotay-Williams-Fabozzi ModelThe Kalotay-Williams-Fabozzi model assumes that changes in the short-term rate can be modeled using Equation (1) by setting f(r) = ln (r) (where ln is the natural logarithm) and...