Better Late Than Never Or Better Hurry?

Posted on July 14, 2022 by Published by

This month’s blog is the second of three installments focused on implementation of the new Current Expected Credit Loss (CECL) requirements, established under ASC 326-20 and the Interagency Policy Statement on Allowances for Credit Losses. The first installment addressed Early Implementation.

As discussed, the early implementation phase would have the necessary “established policies and procedures” for the new CECL program as its outcome/goal. Each organization at this point should be able to confirm that their new policy(ies) meets both the requirements of ASC 326-20 and the Interagency Policy Statement. Further, the selected methodologies and systems should be appropriate for the size, complexity, and risk profile of their institution. And finally, do you have all the resources needed to complete all required steps for implementation, testing, and training? This month we will look at late Implementation, and the final installment in August will discuss Parallel Runs.

Better Late Than Never Or Better Hurry?

Whether your institution opts for a vendor-supplied model or chooses to develop its own, availability of quality data is likely to be the most important part of any CECL calculation and the first step in model development. Not only can the quantity and quality of data impact the methodology selected, but it can also impact qualitative factors in forecasts. Ideally, data collection should cover 10 to 12 years through a full economic cycle. Unfortunately, most banks only have five years of data, thereby limiting methodologies for consideration. Fortunately (or perhaps unfortunately?) CECL allows for the use of third party and peer data in developing ALLL methodologies.

Data quantity will be driven by the complexity of the portfolio and the methodology selected. However, even the simplest methodologies will require sufficient data fields that reflect characteristics of the portfolio segment. More complex portfolios that require a more sophisticated methodology will increase the number of data fields significantly. Given that data quality is an issue for institutions of all sizes, each organization needs to consider developing a data governance program which defines the development, input, security, validation, and use of data.

Implement Methodologies

All institutions will need to develop and implement methodologies based upon portfolio size, composition, complexity, and segmentation. Portfolio segmentation is a critical first step to development of the CECL model. Each segment must be homogeneous and statistically significant. Most, but not all, banks use call codes segmentation. Specialty portfolios may deserve their own segment due to unique risk characteristics. Segmentation of the portfolio should be determined before methodologies for analysis are selected. This recognizes that ASC 326 does not require that all segments use the same approach or methodology.

As a rule, each institution will need to balance the granularity in the portfolio with statistical significance. However, each organization needs to understand each of the methodologies. They should be able to defend the reasons for selecting the method or methods chosen as well as explain why others were not selected.

Types of Methodologies

In addition to issues of data availability and integrity as discussed above, other operational constraints and considerations may impact model selection including loss history, expertise, efficiency in use, and understandability. There are more than a dozen different methodologies to select from when implementing CECL.

These methodologies fall into four broad approaches with varying levels of sophistication as well as different benefits and challenges. From lowest to highest in terms of general sophistication they are: loss-rate, roll-rate, discounted cash flow (DCF), and probability of default (PD). Depending on expertise within your organization, bringing in outside technical assistance in selecting which approach (or approaches) may work best for your institution’s specific needs may be helpful.

Time Is Running Out

Time to complete the CECL project is drawing short. Validation of the model selected would involve about 90 days to run parallel tests against the existing static model. This would happen over at least two quarters of financial performance. Accordingly, each organization should now have made portfolio segmentation decisions. Also, they should have identified needed data, acquired, and scrubbed the data, and developed and tested methodologies for each segment.

In our final installment next month, we will look at finalizing the CECL process. This includes model validation and conducting parallel runs to be ready for full implementation in 2023. The Enlighten Financial team is ready, willing, and able to assist you with navigating the journey to full compliance with the new standards.

 

Richard Rudolph is Senior Consultant at Enlighten Financial, a specialized consulting firm that focuses on loan review and risk management services to community banks and credit unions. Enlighten Financial has made it our business to shed light on the complex financial landscape. We lead clients in the right direction. We work with financial institutions and other providers to mitigate risk. To talk to Rick directly, please call: 920.445.8133.

Tags: ,

Comments are closed here.