IRB-banks can use their own estimates of risk parameters when they compute the capital requirements under the New Basel Accord, which has been launched in Korea in 2008. For retail exposures, IRB-banks should estimate the probability of default (PD) a...
IRB-banks can use their own estimates of risk parameters when they compute the capital requirements under the New Basel Accord, which has been launched in Korea in 2008. For retail exposures, IRB-banks should estimate the probability of default (PD) and the loss given default (LGD) as well as the exposure at default (EAD). If downturn or stress EAD is estimated more conservatively than the default-weighted average EAD, banks should apply the former when they compute the regulatory capital.
EAD is composed of two parts: the drawn amount at present and the expected additional amount that could be drawn from the undrawn limit. The ratio of two in the latter is often called "credit conversion factor" (CCF) and the modeling of EAD is tantamount to estimating CCF. This paper studies retail credit card data that covers from the first quarter of 2002 to the fourth quarter of 2005 provided by one of the leading domestic bank. To the best knowledge of the authors, the data have not been studied by the academics before, and interestingly, cover the 2003 Korea Credit Card Crisis period. Models used in the paper have been proposed by Moral (2007) and are extended by Kim (2007). Namely, this paper estimated linear CCF model, conditional CCF model, median CCF model, and quantile CCF model.
This paper finds that, unlike the Korean business card segment studied by Kim (2007), retail credit card segment did not reveal the so-called procyclicality. Therefore, stress EAD could not be based upon the model. On the other hand, this paper finds that the 66.6 percentile CCF model is conservative enough to cover stress EAD experienced during the crisis period.