Consistency is key when defining Probability of Default
Dec 28, 2016
Probability of default/loss given default (PD/LGD) is widely recognized as a robust method to determine appropriate reserve levels in an institution’s allowance for loan and lease losses. The PD/LGD method has gained popularity of late with specific mentions in Basel II and II, and, most recently, FASB’s CECL guidance. While the method is formulaically simple, the amount of data required to determine values for inputs can present a challenge. However, it is the granularity on losses and comprehensiveness of the estimate that make a move to PD/LGD so attractive for institutions refining their risk management process to maximize insight gained from the process. Consistency in how the granular, loan-level data used for the PD/LGD method is collected and used is paramount in defining the inputs for this calculation and ensuring their usefulness as an accurate reflection of loss in an institution.
Data for defining PD
Probability of Default, or PD, is the average percentage of borrowers who default over a certain period of time. In a recent webinar kicking off Abrigo’ CECL Methodology Series, ALLL Specialist Brandon Russell outlined the recommended data fields institutions should prepare for defining their PD. These include:
- Loan ID
- Customer Balance
- Book Balance
- Segment Identifiers
- Credit Quality
- Charge-offs and Recoveries
The CECL Data Field Guide, which can be accessed as part of the CECL Prep Kit, served as the basis for this discussion and offers a full description of what should be contained in each of these data fields. If institutions start to collect consistent and reliable data in these fields today, they should have sufficient data to calculate an accurate PD by the CECL implementation deadline. But even with sufficient data, there are still choices to be made when defining the Probability of Default.
Defining the Probability of Default (PD)
To determine the input for PD, an institution must settle on the definition of a “default” and determine over what length of time it should be measured. An example of default is a payment that is 90 days past due, but an institution can create their own measure of default drawing on available guidance.
Different banks and providers will define probability of default in different ways. Even in Sageworks ALLL, users have flexibility to define what they consider default events for different kinds of measurements. Once it is defined, the default measurement for probability of default should be internally consistent with the way it’s used in, for example, a loss given default calculation.
While PD/LGD is one of several methods that banks and credit unions may choose for calculating their ALLL under the CECL model, understanding the inputs that go into defining a probability of default within an institution offer valuable insights for institution-wide risk management. An adherence to consistent data-collection methods and a consistent definition of PD provides a solid foundation for the PD/LGD method of calculating loss in an institution.
For more about leveraging the PD/LGD method, watch the on-demand webinar Kick Off to our CECL Methodology Series.
For more about using the PD/LGD method, see whitepaper, The Basics of PD/LGD.
The importance of documenting the PD/LGD method is discussed further in this blog post.