T here are many ways of modeling the term structure of interest rates, many interest rate models, and many classifications of them. Some models describe the evolution of a given interest rate (usually the short-term rate) and will be consistent by construction with the current value of that interest rate. These models, in general, will not be consistent with the rest of the yield curve, however, and will not "correctly" price (relative to the market) claims as simple as discount bonds; this suggests that the models will do a poor job pricing more complex derivatives. Some of these models use one factor to explain the evolution of interest rates (see, for example, Vasicek [1977] and Cox, Ingersoll, and Ross [1985]), while others employ two factors (Brennan and Schwartz [1979], Schaefer and Schwartz [1984], Longstaff and Schwartz [1992], and Moreno [1996], among others).From the perspective of derivatives pricing, it seems more convenient to develop models consistent with the market yield curve. This is the approach followed by Ho and Lee [1986] (using bond prices) and Heath, Jarrow, and Morton [1992] (using forward interest rates). An equivalent approach is to build models based on the evolution of the short rate (or a function of it), and allow for time-dependent parameters (see Dybvig [1988] and Jamshidian [1988]). These parameters can be calibrated so that (524 observations). The continuous-time unrestricted model is dr = κ(θ -r)dt + σr δ dz. t-statistics in parentheses.