The Performance of the Calibrated Leland - Toft Model Howard

A common wisdom about term structure models is that they predict much lower spreads than the observed spreads for investment-grade bond, and most of them tend to overpredict spreads for junk bonds. Among them, the Leland-Toft model is perhaps most controversial. Some studies show it always overpredict spreads, in some cases they can be as high as over 5000 bps, yet other studies show it generates extremely low spreads for investment-grade ratings, hence it is just as poor as other models. We introduce an approach for calibration of the model. We particularly stress the importance of carrying out the calibration properly and show what are the cautions one must take for such a calibration to be successful. Using the calibrated model and some actual bonds data as bench mark, we find the Leland-Toft model performs significantly better than reported by some recent studies. Our findings may help to clarify the controversy and serve as a reference for calibration of other term structure models. In the end, we also provide some general thinking on the calibration and applicability of the structural term structure models. Howard Qi and Chunchi Wu are at Syracuse University and Sheen X. Liu is at Youngstown State University. Contact address: School of Management, Syracuse University, Syracuse, NY 13244. Email:[email protected]; fax:315-443-5457.