Credit Risk from Theory to Application

Abstract

<p> In this thesis, we calibrated a one factor CIR model for interest rate and a two factor CIR model for each hazard rate of 21 firms. The time series of the interest rate and each hazard rate for 21 firms are also obtained. Extended Kalman Filter and Quasi-Maximum Likelihood Estimation are used as the numerical scheme. The empirical results suggest that multifactor CIR models are not better than multifactor Hull-White model. Positive correlations between hazard rate and interest rate are discovered, although most hazard rates are found to be negatively correlated with the default-free interest rate. The 21 filtered time series of the hazard rates suggest that there maybe a hidden common factor shared only by the intensities. Monte Carlo Simulation is conducted both for interest rate and hazard rates. The simulation indicate that both the SKF and the EKF work pretty well as a filter tool but may produce bad estimation for the value of the likelihood function. QMLE works fine in linear state space form model, but it does a poor job in the case of non-linear state space form.</p>