The feasibility of moving to an ALLL model based on PD/LGD or migration analysis is largely driven by the availability and integrity of data.
For a PD/LGD based model, the first question to address would be how the probability of default and corresponding loss given default would be calculated. There is a wide spectrum of methods that are used, particularly for the PD portion. These often involve sophisticated statistical techniques like multi-variable regression analysis to determine correlation between losses and various financial metrics. Many institutions will find this type of analysis too cumbersome and may not have usable data for inputs going back in time. Also, due to the complexity of the analysis, institutions may find it difficult to articulate how the calculation works for audiences like senior management, the board of directors, external auditors and examiners.
For the LGD, institutions often rely on published or subscription based market data for this information. Though more readily available, a drawback with this data is that it isn’t based on the institution’s own loss experience and thus may not be as applicable for all types of banks.
Migration analysis allows institutions to determine specific loss rates for specific risk levels within their FAS 5 pools (typically based on risk rating or risk grade). Though this type of analysis is often more intuitive for bankers than PD/LGD models, migration analysis still requires significant amounts of historical data. As the analysis essentially follows how loans in the portfolio move through risk levels over time (and some eventually to loss), it is necessary to look at all loans in a given pool over the length on the analysis. Any gaps or inconsistencies in this data can cause errors and possibly misleading results.
There does not appear to be a preferred or favored method by examiners, but they will need to understand why the methodology chosen is appropriate for your bank. A good rule of thumb to consider is the “elevator pitch”. If you can’t explain at a high level how your methodology determines loss factors and why it is the right one for your institution, you might consider simplifying your model.
The feasibility of moving to an ALLL model based on PD/LGD or migration analysis is largely driven by the availability and integrity of data.
For a PD/LGD based model, the first question to address would be how the probability of default and corresponding loss given default would be calculated. There is a wide spectrum of methods that are used, particularly for the PD portion. These often involve sophisticated statistical techniques like multi-variable regression analysis to determine correlation between losses and various financial metrics. Many institutions will find this type of analysis too cumbersome and may not have usable data for inputs going back in time. Also, due to the complexity of the analysis, institutions may find it difficult to articulate how the calculation works for audiences like senior management, the board of directors, external auditors and examiners.
For the LGD, institutions often rely on published or subscription based market data for this information. Though more readily available, a drawback with this data is that it isn’t based on the institution’s own loss experience and thus may not be as applicable for all types of banks.
Migration analysis allows institutions to determine specific loss rates for specific risk levels within their FAS 5 pools (typically based on risk rating or risk grade). Though this type of analysis is often more intuitive for bankers than PD/LGD models, migration analysis still requires significant amounts of historical data. As the analysis essentially follows how loans in the portfolio move through risk levels over time (and some eventually to loss), it is necessary to look at all loans in a given pool over the length on the analysis. Any gaps or inconsistencies in this data can cause errors and possibly misleading results.
There does not appear to be a preferred or favored method by examiners, but they will need to understand why the methodology chosen is appropriate for your bank. A good rule of thumb to consider is the “elevator pitch”. If you can’t explain at a high level how your methodology determines loss factors and why it is the right one for your institution, you might consider simplifying your model.