Post by Sapphire Capital on Jul 11, 2008 6:18:28 GMT 4
Benchmarking Discriminatory Power of Credit Risk Rating Models
DARSHI ARACHIGE
Affiliation Unknown
--------------------------------------------------------------------------------
JASSA, Journal of Finsia, Issue 1, pp. 6-12, Autumn 2007
Abstract:
Benchmarking internal credit risk models is one of the key issues faced by Banking Industry with the implementation of New Basel II framework. In validation of such models, main consideration is how well the internal rating systems can rate the risk of an obligor defaulting within a given span of time from the time of rating. Industry focus is currently on two major dimensions, namely, the discriminatory power of internal rating models and strength of their calibration to the cycle average or long term default rates. One challenge faced by many model validation personnel would be to find a proper benchmark for the discriminatory power of their models.
In this article, we demonstrate how we can easily develop such a benchmark through a simple modification to a popular discriminatory power measure. We also compare the outcome of a test based on this modified measure to two other benchmarking approaches through a publicly available set of data.
papers.ssrn.com/sol3/papers.cfm?abstract_id=1137243
DARSHI ARACHIGE
Affiliation Unknown
--------------------------------------------------------------------------------
JASSA, Journal of Finsia, Issue 1, pp. 6-12, Autumn 2007
Abstract:
Benchmarking internal credit risk models is one of the key issues faced by Banking Industry with the implementation of New Basel II framework. In validation of such models, main consideration is how well the internal rating systems can rate the risk of an obligor defaulting within a given span of time from the time of rating. Industry focus is currently on two major dimensions, namely, the discriminatory power of internal rating models and strength of their calibration to the cycle average or long term default rates. One challenge faced by many model validation personnel would be to find a proper benchmark for the discriminatory power of their models.
In this article, we demonstrate how we can easily develop such a benchmark through a simple modification to a popular discriminatory power measure. We also compare the outcome of a test based on this modified measure to two other benchmarking approaches through a publicly available set of data.
papers.ssrn.com/sol3/papers.cfm?abstract_id=1137243