Assessing probabilities of default in Thai corporate bond market

Abstract

This study employs the structural models i.e. Merton, Black and Cox, Longstaff and Schwartz, and reduced-form models, i.e. Jarrow and Turnbull; Constant Hazard Function model, Linear Hazard Function model and Quadratic Hazard Function model to assess the probabilities of default in the Thai corporate bond market. The study covers corporate bonds listed on Thai Bond Market Association (ThaiBMA) and Stock Exchange of Thailand (SET) from January 2001 to June 2006. We then compare the accuracy of default probabilities of each model by using probabilities of default to predict corporate bond price in next 1-month and 3-months. Thus, calculate the value of Mean Absolute Percentage Error (MAPE). Of all the result comparing probability of default value from six models, the average probabilities of default from structural form is lower than from reduced-form in Merton Model which gives the lowest value. For average probabilities of default from reduced-form models: Quadratic Hazard Function model has highest range of all. Comparing MAPE, forecast price in the next 1- month shows that MAPE from reduced-form: Quadratic Hazard Function model gives lowest value followed by MAPE from reduced-form: Linear Hazard Function model, reduced-form: Constant Hazard Function Model, Merton model, The Black and Cox model. For MAPE from Longstaff and Schwartz model has highest value. In MAPE in order to forecast price in the next 3-months, the results correspond with the forecasted price in the next 1-month. Since the mean of MAPE obtained from reduced-form model has lower value than the mean of MAPE obtained from structural form model for all firms. This infers that the probability of default from the reduced-form model is more effective in predicting the bond price than calculated from the structural model