Does conditioning information matter in estimating continuous time interest rate diffusions?

We examine an important aspect of empirical estimation of term structure models; the role of conditioning information in dynamic term structure models. The use of both real world or simulated data implicitly incorporates conditioning information. We examine the bias created in estimating the drift by a specific form of conditioning, namely truncation. Using the theory of enlargement of filtrations we provide estimates of the extent of this truncation bias for commonly used short rate models. We find that this truncation bias causes the drift of these models to have a nonlinear structure.