We live in an environment in The basics of the theory of modeling credit risk at the portfolio level have been known for around fifteen years, since CreditRiskC, CreditMetrics and Credit Portfolio View were launched by Credit Suisse First Boston, J. P. Morgan and McKinsey, respectively. Since their arrival, these three models have spread quickly and they now dominate credit portfolio management within the financial industry.
When these models became more widespread, therewas also great hope in the industry that credit portfolio models would become an element of the regulatory framework of Pillar 1. However, although the calculation of risk weights in the internal ratings-based approach of the Basel II framework utilizes the theory behind CreditRiskC and CreditMetrics, a comprehensive model for credit portfolio risk has never been accepted by the Basel Committee. The Committee substantiated its rejection of these models by citing serious doubts about the modeling of correlations within these models. It left the door open for credit portfolio models to become part of the regulatory framework later on, once this issue had been solved.A key step forward in the regulation was made when credit portfolio models started to be used as an element for calculating the incremental risk charge (IRC), which addresses default and migration risk in the trading book from 2012 onward.
Risk management based on these models became the standard in all major banks across theworld, and in smaller banks too. But since the breakthrough of these models and the euphoria surrounding it, it has become evident that there are still many little understood areas in the field of credit portfolio modeling and many unsolved issues still remain. With this Special Issue, The Journal of Risk Model Validation attempts to reduce this terra incognita by casting light on some interesting issues. The first paper, byAlfred Hamerle,Andreas Dartsch, Rainer Jobst and Kilian Plank, discusses a highly relevant issue. The authors propose a method for integrating macroeconomic forecasts into portfolio models. This is important because the recent financial crisis has shown that the value of a credit portfolio can change rapidly in a short time. Therefore, more forward-looking modeling techniques have to be developed. Furthermore, the authors’ inclusion of macroeconomic aspects in their method contributes to a more reasonable form of stress testing, providing another aspect of the value of this contribution. The second paper, by Benjamin Bade, Daniel Rösch and Harald Scheule, deals with another area that has attracted much research activity in the past: the dependence between the probability of a default event (PD) and the loss given a default event (LGD), which are the two main drivers of credit risk. The paper is a comparative study of different techniques and approaches to including correlation between PD and LGD that are used in practice. It is a good example of the continuous effort to combine theory and practice that this journal makes.
The third paper, by Marcus R. W. Martin, Helmut Lutz and Carsten S. Wehn, is on the subject of portfolio modeling in the context of the above-mentioned IRC. It is interesting because it analyzes how models can be simplified without losing their predictiveness. Since the final result is semianalytical, the approach offers the opportunity for a better understanding of these types of models. I believe that this paper will be very helpful to all those currently concerned with implementing an IRC model.
Finally, the last paper, by Sebastian Ostrowski and Peter Reichling, addresses the old issue of the relationship between discriminatory power and the fitness of a rating function. It is an issue not specifically related to modeling credit portfolio risk, but an overarching matter for all forecasting techniques. It is to be appreciated that the authors have added new aspects to an old concept and have therefore been able to highlight some limitations of well-known techniques.