The recent financial crisis revealed limits of current financial models in modelling loss distributions which in turn caused problems associated with risk measurement and credit derivatives pricing. In this thesis a reduced form model approach based on the CreditRisk+ framework (by Credit Suisse First Boston (1997)) for modelling a loss distribution of a given portfolio of obligors will be presented and implemented using suitable computationally efficient numerical methods. This model allows for stochastic 'loss given default' rates and resolves one of the main drawbacks of the classic CreditRisk+ model -- proper handling of high default probabilities. The latter enables the CreditRisk+ model to deal with high-yield securities, crisis scenarios and long time-horizons. Besides, a Multi-Period CreditRisk+ model together with a Regime-Switching CreditRisk+ model complete the credit model presented in this thesis. These extensions allow one to add flexibility by taking into account a term-structure of input parameters and, in case of the Regime-Switching model, to even model the market-state dependency of input parame- ters. This is a useful addition to the classic model whenever scenario analyses are anticipated or effects of possible crises times are to be accounted for. As in the classic CreditRisk+ framework, default correlation between obligors will be introduced using an appropriate number of risk factors. Whereas in the classic approach the volatility of these risk factors is based on a heuristic approach, we present a structured approach of extracting the sector volatilities from migration matrices. Hereby we aim at improving the credibility of the CreditRisk+ model in front of regulatory bodies. Subsequently, a brief introduction to structural models is given, in particular covering the well-known one-factor Gaussian model and a more recent development, the one-factor NIG model. Both are used to implement a structural model being able to deal with arbitrary credit portfolios. In a last step the CreditRisk+ model is compared with the one-factor Gaussian model in terms of default correlation.      «

The recent financial crisis revealed limits of current financial models in modelling loss distributions which in turn caused problems associated with risk measurement and credit derivatives pricing. In this thesis a reduced form model approach based on the CreditRisk+ framework (by Credit Suisse First Boston (1997)) for modelling a loss distribution of a given portfolio of obligors will be presented and implemented using suitable computationally efficient numerical methods. This model allows for...     »