Methodologies Of Extracting Implied Recovery Rate In Corporate Bond

Recovery rate and hazard rate are two key indicators in analyzing credit risk of corporate bond. Although both of them can be achieved by statistics, it’s believed that information extracted from market price is much more meaningful. Pricing models of corporate bond can currently be categorized into three:RFV model, RT model and RMV model according to different definitions of recovery. RFV and RT model can be applied to extract both recovery rate and hazard rate while RMV model can’t. The main innovation and contribution of this paper is to identify the function between recovery rate and hazard rate and to utilize it to extract the recovery rate and hazard rate in RMV model respectfully.Early scholars can only implement numerical method to figure out the corporate bond value in RFV and RT model, which situation has been changed when Bakshi and Madan applied characteristic function to solve out the resolution which not only facilitates the pricing of corporate bond but also offers the possibility of calibrating the parameters including recovery rate and hazard rate in the models.Unfortunately, Bakshi and Madan couldn’t overcome the demerit of RMV model in the extraction of credit risk information either which is that recovery rate and hazard rate are impossible to be differentiated although the result of multiplication is available. Duffie and Singleton (1999) have proposed a number of possible solutions, one of which is to derive the underlying relationship between recovery rate and hazard rate. Inspired by this, this article starts from the classical structural model with and without jump and finds out the implied relationship after some deductions and Monte Carlo simulations, which helps to solve the conundrum in RMV model.So which model should we choose since all of them can be used to extract recovery rate and hazard rate after this article? To answer that, this paper compares the theoretical basis, accuracy of calibration, out-of-sample forecasting errors and computational efficiency of all three models and come to a conclusion that although RFV and RT model performs well in calibration, RMV model has a smallest out-of-sample forecasting error and best computational efficiency.