Comparing Structural and Reduced-Form Credit Risk Models

Alexander Koivusalo

Thursday, 11 June 2009, 13:30
Seminar room F

Structural and reduced-form models are commonly used by professional risk managers in banks, rating companies and financial institutions for modeling of credit risk. Credit risk is closely related to the distribution of financial losses, that occur when companies fail with their contractual obligation to repay their debts. In our Structural model correlated jump-diffusion processes are simulated to model the value of the companies in our credit portfolio.

The concept of this thesis is to determine how well the reduced-form model reproduces the distribution of portfolio losses from our Structural model i.e. how well will the reduced-form model keep the information from the Structural model assuming that the Structural model generates the real distribution of portfolio losses. We are particularly interested of the tail behavior of the models which corresponds to the worst case scenario. A central question would be if the models used by risk managers before the financial crisis of 2007-2009 really were able to reproduce the tail behavior of heavy tail distributions. Recovery rates and calibration issues are essential for the reduced-form models ability to reproduce the tail behavior of our Structural model. Further we show that constant recovery rate models misrepresents the distribution of portfolio losses.