Credit risk modeling strategies: the road to serfdom?

Abstract

This paper aims at presenting some practical issues in modeling default risk of a single commercial credit counterparty from the perspective of a large retail bank. We define default risk as the probability that a counterparty's intrinsic credit quality deteriorates within a given time horizon such that contractual agreements cannot be honored. This work gives an insight into using scoring/rating models in a credit environment of a large European bank. Contrary to many banks, we did not define the segments in a first step with a view to developing the rating tools in a second step. Our approach has, to some extent, followed a different path. Iteratively, we both defined the borders for a particular segment and selected an appropriate rating tool. More particularly, customer segmentation has been carried out on the basis of various rating tools' goodness-of-fit criteria. The topics cover customer segmentation using goodness-of-fit measures, data measurement levels and optimization algorithms, rating tool calibration to the central default tendency and communication to the end user.

The Impact of Risk Management in Credit Rating Agencies

“…In addition, it was found that accounting ratios and market-based risk measures are more informative for larger companies than smaller companies for assigning ratings. Baestaens's (1999) model aimed to present some practical issues in modelling the default risk of a single commercial credit counterparty from the perspective of a large retail bank. The best rating methodology is very much dependent on the segment to which it should be applied, the nature of the data (qualitative versus quantitative), the classification performance (both within sample and out-of-sample), the grade stability, and the absence of 'black-box syndrome' or the ease of communication towards the end-user.…”

“…Firms should additionally have a risk avoidance model to identify, measure, evaluate, monitor, report, and control or mitigate credit risk on a timely basis (Monetary Authority of Singapore 2013). Baestaens (1999) opined that the rating methodology is dependent on the industry segment, nature of the data, classification performance both within sample and out-of-sample, grade stability, and ease of communication towards the end-user. O' Kane and Sen (2005) suggested that corporations should determine the compensation for assuming the credit risk embedded within the security and the return of credit assets relative to some benchmark of higher credit quality and define, describe, and analyse the main credit spreads for fixed-rate bonds, floating-rate notes, and credit default swaps.…”

Credit Scoring, Statistical Techniques and Evaluation Criteria: A Review of the Literature

“…Lim and Sohn (2007) argue that the way existing models are used is quite worrying, especially at the time when the middle of the repayment term occurs, when it is important to be able to re-evaluate the creditability of borrowers with high default risks for the remaining term (e.g. Baestaens, 1999).…”

“…Nevertheless, applications of credit scoring have been widely used in different fields, including a comparison between different statistical techniques used in prediction purposes and classification problems. These applications can be classified into accounting and finance (Long, 1973; Duliba, 1991; Altman et al , 1994; Sinha and Richardson, 1996; Baestaens, 1999; Pendharkar, 2005; Landajo et al , 2007), marketing (Dasgupta et al , 1994; Kumar et al , 1995; Thieme et al , 2000; Chiang et al , 2006), engineering and manufacturing (Smith and Mason, 1997; Feng and Wang, 2002; Dvir et al , 2006), health and medicine (Warner and Misra, 1996; Nguyen et al , 2002; Behrman et al , 2007) and general applications (Hardgrave et al , 1994; Walczak and Sincich, 1999; Usha, 2005; Nikolopoulos et al , 2007), as noted by Paliwal and Kumar (2009).…”

The relationship between default prediction and lending profits: Integrating ROC analysis and loan pricing

“…However, if this strategy is implemented in the case where the ROCs of two models cross, the expected performance of the combined two models will be superior to either used separately. 10 For example Baestaens (1999) discusses a number of useful approaches to implementing internal rating grades based on commercial models. The objective of the article was to optimize performance over several criteria.…”