Do Structural Models for Corporate Bonds Work?

Posted: 10 Dec 2008 Last revised: 10 Nov 2014

Date Written: December 10, 2008

Abstract

Recent empirical results on calibrating structural models for corporate bonds have been mixed. Most authors find that such models significantly underestimate the default component in credit spreads, when they are calibrated to observed default frequencies. Others found that structural models tend to generate to high values for the default component when bond prices are used to estimate the asset-value volatility. Using a simulation experiment, Ericsson and Reneby (2005) show that the empirical performance of such models can be improved upon when calibrated by means of maximum likelihood. We test whether this is also the case when a structural model is calibrated to actual bond yields, rather than simulated data. The tests in this paper use the structural model of Leland and Toft (1996), which has been shown to do a good job at explaining observed default frequencies.

A first set of results indicates that the superior performance of the maximum likelihood method is lost when it is confronted with a real-live data set, rather than with simulated data. More precisely, implied default spreads seem too high. However, forcing the model to replicate observed default frequencies, significantly improves the results. We find that in this case the implied default spreads seem to be unbiased estimates of the true default component of credit spreads.

Keywords: corporate bonds, structural models, asset pricing, credit risk

Suggested Citation

Simon, Steven C. J., Do Structural Models for Corporate Bonds Work? (December 10, 2008). Available at SSRN: https://ssrn.com/abstract=1314285

Steven C. J. Simon (Contact Author)

Free University of Brussels ( email )

Pleinlaan 2
http://www.vub.ac.be/
Brussels, 1050
Belgium

Do you have a job opening that you would like to promote on SSRN?

Paper statistics

Abstract Views
389
PlumX Metrics