AccessMyLibrary provides FREE access to over 30 million articles from top publications available through your library.
Create a link to this page
Copy and paste this link tag into your Web page or blog:
As banks prepare to satisfy the Basel II Accord requirements, many issues need to be addressed, including the creation of internally calculated estimates for Probability of Default (PD).
Currently in the United States, only a limited number of banks are required to update their systems to comply with Basel II at the Advanced Internal Ratings Based (A-IRB) level. However, many banks outside of this group are choosing to opt in or introduce some of the AIRB components should they choose to participate at a later date. Despite all the criticisms of Basel II and the requirements for compliance, the Accord is based on sound risk-management practices, making it a wise decision for all banking institutions.
For retail portfolios (consumer and small business), PD is viewed as one of the most straightforward risk inputs to be derived. That is a reasonable conclusion, since forecasting the probability of an account's achieving default status in the next 12 months is very similar to forecasting the likelihood of being delinquent (an issue that has been addressed by traditional credit scores for decades). However, does this mean that a credit score currently available can be used for PD? Scores typically have values in the range of 0 to 1,000, so a conversion of the score to a probability number between 0 and 1 is required. But is there more to it than that?
Some organizations have begun to question the assertion that a traditional bureau risk score is adequate for PD estimation. Research by Experian-Scorex has revealed that there are weaknesses in traditional bureau risk scores utilized for PD and that available alternatives produce significantly better results. Regulators, rating agencies and lenders all know that a bureau score value means different things, depending on issues such as product line, bureau data source and lender. The myth of a score of 700 being universally applicable is crumbling fast. Because of this, a validation for a particular data set is always required to determine how to interpret a score for a specific portfolio, along with an adjustment (recalibration) if a particular score-to-bad-rate relationship is required. This validation also needs to be performed separately for each credit reporting agency used, since the score may not denote the same thing for your portfolio with data from multiple agencies. Scorecard developers that provide scores across multiple bureaus attempt to match performance across bureaus, but this is done at an aggregate level. Therefore, the alignment may not be correct for your portfolio.
Although bureau scores are easy to access and use, you must validate their performance for your particular decision environment. For PD, accuracy of forecast is even more important than it is when a score is used for account decisioning. Therefore, it is crucial to test, validate and recalibrate when necessary.
When working with Basel II PD, there is more to the issue than just calibration. Most of the traditional bureau risk scores are built with a 24-month outcome. This means they are optimized to forecast the likelihood of being delinquent over the next 24 months. However, for regulatory capital calculations, a forecast over the following 12 months is required. If you have observed a bad rate from point of origin, you are likely to have noticed that it continues to increase well beyond 12 months--typically flattening out in the 18- to 24-month period. This is why 24 months is used for the outcome period with traditional risk scores.
What does this mean if the same score is used as the basis for estimating PD over 12 months? To test this, Experian-Scorex constructed new bureau scores for PD based on the Basel II requirements of a 12-month outcome period and a 90-plus-day default definition. The performance of two traditional bureau risk scores was compared, calibrated to the Basel II requirements for outcome period and default definition. When using the Basel II-specific bureau score, significant improvements in performance were observed across all three of the product areas--mortgage, other installment and revolving. The results were dramatic, as shown in the following summary: