| The hazard rate models used in the recent bankruptcy literature assume the censoring and the default are two independent events, which means the censored company will eventually default. However we believe there will be a portion in the censored group that will be long-term survivors and we propose a mixture model of survivors and risky companies. Moreover this dissertation models the event and the timing of default incident at the same time. For the event of default and the timing of default we utilize a logistic regression. The results have justified the advantage of our model over the standard hazard rate models and proved its predictive power. The companies identified as high default risk by our model proved to deliver extremely low returns in the market. |
