Markov chain based PD term structure modelling in an IFRS 9 framework

This masters thesis addresses the quantitative aspect of PD term structure modelling in an IFRS 9 framework. By using credit exposure data from the Italian banking sector, the thesis builds a comprehensive quantitative analysis based on a time homogeneous Markov chain approach for modelling the PD term structure of the private and corporate sector in Italy. The results show, that the assumption of time homogeneity is critical. Therefore using the average matrix for the PD term structure estimation does not fulfill all IFRS 9 requirements for PD modelling. We propose to calibrate the model on the most recent migration matrix for better compliance. Also in this case, the predictive power is likely to be poor after three to four years. As consequence, we stress the need for model validation and model risk management in the context of IFRS 9. Finally we propose approaches to generalize the PD modelling for IFRS 9 purposes for other applications to provide a coherent PD modelling framework.author Thomas Ludwig(VLID)434791