CECL analyst: If you want to pass a CECL audit, be sure to show your work

Newswatch recently talked with Atul Nepal, a Quantitative Analyst at MountainView Financial Solutions, a Situs company, about the impact of the current expected credit losses (CECL) standard, which was issued in 2016 and will begin to go into effect in 2020.

(1) What is the best model to use for small financial institutions? What about larger institutions?

For any size institution, the best model to use will largely be dependent upon the institution’s data availability. If a financial institution has significant data (volume and quality), that institution will have more modeling options and will be able to segment loan pools for a more granular loan-level analysis. Conversely, a very small institution may not have enough data for an analysis to be statistically relevant and will need to look at testing multiple methodologies and seek out supplemental data for missing historical loss data for each of the asset types. Each methodology will have its own drawbacks, and data limitations may skew the estimate; therefore, institutions should apply multiple methods to identify which methodology will stand up to the scrutiny.   While regulators do not provide prescriptive guidance, they have stated that they expect that most small, community institutions will require only simple modeling techniques.

Generally speaking, I can say that the better the data capture (the more robust and granular), the more sophisticated the modeling techniques an institution can use. This is not an exhaustive list, but small institutions may opt for techniques that are widely used for measuring impairments currently, such as discounted cash flow analysis, vintage analysis, static pool analysis or the average charge-off methods, while mid-sized and large institutions will find success with all of the above listed techniques plus roll rate, probability-of-default, and regression type analysis. Ultimately, you should be able to provide reasonable and supportable loss estimates using necessary methods that represents your institutional risk appetite.

(2) How can financial institutions be sure that their methodology will pass an audit?

Again, CECL is non-prescriptive so a lot of financial institutions have expressed concern about how auditors will assess their methodologies. That said, model and accounting audits are not new and there are a few things institutions should focus on based on other non-CECL audits. These include:

  • Documentation and governance. Stepping back to 5th grade math, financial institutions need to show their work. The more transparency a financial institution can provide through extensive documentation, the easier it is for the auditor to understand and assess its models/methodologies. Documentation should include information about the model selection processes, data generation processes, key assumptions, limitations, and any adjustments made by management. Additionally, the documentation should include how the model/methodology will be applied for CECL calculations. There should also be a model governance process to examine the documentation and process in general to showcase due diligence performed by the model owners.
  • Effective challenge of modeling framework. CECL model/methodologies will be based on expectation of the future. Assumptions of future expectation inherently will add model/methodology risks. The Model Risk Management guidelines published by OCC/FED requires that a financial institution effectively challenge its model/methodology to assess inherent model-related risks. Auditors like to see that the model/methodologies have been effectively validated by qualified independent validators who are not involved in the development process audits. An objective and unbiased model validation will challenge the conceptual design, technical elements, data and assumptions, application and governance framework surrounding the model.
  • Repeatable. Can an independent party come in and repeat the methodology that you’ve used for CECL? The easiest way to find out is to ask an independent party inside/outside the organization to come in and repeat the process and determine whether that individual or provider reached the same conclusion. If that conclusion cannot be met, more work will need to be done to get it right.

(3) If a financial institution did a lot of work with Dodd-Frank Annual Stress Testing (DFAST), how can it leverage that work for CECL?

Many institutions put significant effort into DFAST/CCAR, and there is absolutely an opportunity to put that hard work to good use in CECL. One of the many benefits that came out of DFAST/CCAR implementation is that most of the process had to be documented, validated and audited; therefore, the modeling structure, governance and assumptions used in DFAST/CCAR can also be used for CECL with some modification to account for effective life of portfolio. In fact, DFAST/CCAR institutions have an advantage over other institutions since they will know what to expect from an internal and external audit process to adjust the structural requirements needed to support CECL modeling. 

In addition, DFAST/CCAR institutions used the stress testing exercise to capture data for credit modeling. They maintained a data warehouse to capture and vet significant amounts of data required for nine quarter projections. This has really established a great foundation to aggregate and manage historical data that can be used for CECL. Now, financial institutions that have gone through DFAST can build on that foundation and extend data warehousing to capture an effective life of assets as required under CECL.

Struggling to know which model to use? Contact Us today to discuss your CECL challenges.