An Empirical analysis of structural models of Thai corporate bonds

Abstract

This paper applies four structural models of corporate bond pricing to the Thai bond market. The study compares predicted bond prices with actual bond prices and finds the standard deviations of pricing errors. The four structural models tested are proposed by Geske (1977), Leland and Toft (1996), Longstaff and Schwartz (1995), and Collin-Dufresne and Goldstein (2001). The sample consists of 2,464 observations of bond prices form firms with simple capital structures during 1999-2004. We find that the Collin-Dufresne and Goldstein model generates the lowest pricing errors and standard deviations among the four models using a slightly truncated sample. However, the Geske model provides the lowest pricing errors and standard deviations. In addition, the Geske model understates the credit risk while the other models overstate the credit risk on average. Also, We find that the standard deviations of pricing errors of the four models are quite high compared to pricing errors. The Leland and Toft model has the highest standard deviation on average. Moreover, this study examines the sensitivity of pricing errors to the recovery rate assumptions. The results show a high recovery rate can reduce the pricing errors. Specifically, we test the effect of recovery rate on pricing errors by changing the recovery rate from 30-60% and find that the recovery rate at 60% provides the least pricing errors. Finally, the study finds that each model is quite sensitive to the estimated parameters. For example, the Collin-Dufresne and Goldstein model, using different parameters estimates of the target leverage ratio, generates extremely different results.