NOTE: This post is the second in a two-part series on CECL. Click here for Part 1, which explains the rationale for CECL and its implications on financial allowance practices.
As financial institutions prepare for the 2020 and 2023 implementation of CECL guidance regulations, most institutions have the same hurdles to overcome. We view these hurdles as strategic opportunities for the institutions that put the time and energy into understanding and developing the best solutions.
Institutions are required to develop forward-looking approaches based on the materiality of their portfolios. While they might be able to rely on simple approaches for immaterial portfolios, more sophisticated approaches are expected for material portfolios. In addition to the modeling approach itself, forecasting future economic conditions as a part of the new approach is no small task, as we’ll explain.
Data and Modeling
In nearly all modeling exercises, data sourcing, preparation and validation represent the majority of the effort. CECL requires more complete and reliable data than the incurred loss methodology for most institutions. Given the need to forecast under any future economic conditions, it’s absolutely critical to leverage data collected during at least one past recession. CECL requires the use of several new data elements that might not have been examined before – including term structure, interest rate and payment type – in order to enable a proper comparison of portfolios with similar risk characteristics.
A bank’s modeling approach will be reliant on data availability, data quality, materiality of each portfolio, and supported analytics. Institutions should re-examine their existing segmentation strategy to ensure segments have similar risk characteristics when viewed in a forward-looking assessment. This diagnosis requires business intuition and a granularity of data that’s not normally available today. In many cases, the qualitative factor (“Q factor”) process will require updates to address data quality and segmentation issues.
Data quality and availability are focal points for both regulators and auditors. As data quality issues arise, financial institutions should ensure appropriate governance and remediation activities are in place to mitigate them. In cases where data is limited, they should ensure that they are beginning to properly collect date for future use. Many institutions will look to external data sources to supplement their internal data. When using external data, be sure the information you use is a suitable proxy for internal data.
Reasonable and Supportable
As we mentioned in Part 1, CECL guidance is not prescriptive on what constitutes “reasonable and supportable.” At a minimum, it’s a thorough analysis over various scenarios. It is data analytics, reliability of economic forecasts, alignment with key planning processes within the institution, and/or modeling approaches that might drive the forecast during the R&S period. Additionally, different portfolios might have different R&S periods.
Careful consideration should be made as impacts of the R&S period are analyzed and a framework is ultimately chosen. Your rationale should be well documented and presented to management for effective challenge – and final approval. The choice of an R&S period will likely have differing impacts depending on what portion of the economic cycle is included. For example, a shorter R&S period as the economy heads into recession is likely to produce a lower loss estimate than one that incorporates a recession’s full depth. Institutions must be prepared to assess their R&S period each quarter as part of the allowance process.
Beyond the R&S period, institutions must revert to historical credit loss experience. CECL guidance does not prescribe the methodology for calculating the historical levels that institutions should revert to or the time period/process by which they should revert to those historical assumptions.
For model-based approaches, reversion can be performed by leveraging forecast inputs or outputs. Reverting at the input level involves determining historical assumptions for econometric inputs and, if applicable, loan characteristics. Reverting at the output level involves assumptions for each component of the loss calculation or for loss rates (e.g., average default history and average loss given default history versus average loss history).
Regarding reversion timing, institutions may choose to revert immediately, creating a one-month “cliff” or “jump” from the final forecast point to historicals, or transition between the two over a defined time horizon. In the second case, they must choose a reasonable and supportable path (e.g., straight line over x periods, curved relationship over x quarters, etc.). As a best practice – and a compliance safeguard – institutions should plan to assess and document the multiple approaches they considered.
Financial institutions can use any number of scenarios, but they need to build out well-governed scenario design processes. Key decision points for scenarios include:
- Which variables should be used during the CECL process?
- Should regional, state or national variables be used?
- Should scenarios be designed internally or by an external partner?
- If using a vendor, does the firm have sufficient expertise within the bank to challenge or confirm the validity of the scenario(s)?
- How many scenarios should be used in the CECL process? If using multiple scenarios, should the weighting of outcomes be applied to produce a single number?
While a single scenario might seem easier, the approach is likely to result in significant volatility. Leveraging multiple scenarios that encompass a range of potential outcomes can reduce volatility within CECL’s allowance process inputs, provided all of them are reasonable and supportable.
The concept that more sophisticated modeling under CECL will reduce the importance of the qualitative aspects of the allowance process is a misconception. In fact, CECL guidance retains most of the qualitative framework common in the incurred loss methodology. In fact, new qualitative components will likely be needed as modeling decisions, model performance over various timeframes, limitations in segmentation, data quality, etc., represent new uncertainties in the process.
For example, newer portfolios or those lacking default/charge-off events might require Q factor consideration even though they’re combined into a larger portfolio that’s been modeled more granularly.
A strong governance program is a must in implementing CECL and requires resources from various disciplines and coordination on the appropriate controls. Governance should be present in all steps of the CECL journey, from design to reporting. Best practices include having a repository for all CECL-related controls, as well as documentation on presentations, meeting minutes, data assessments, and other pertinent information as it relates to the structure of your methodology.
Many institutions that are well along in this process struggle to pull together the entire package when audit activities begin. That’s why maintaining a tightly managed, organized environment and inventory of necessary evidence is well worth the investment of time and effort. Institutions should ensure that all data used in the development of CECL models/assumptions ties to source systems. Based on our experience, several areas of particular focus include:
- Data: Institutions should ensure that evidence is maintained for any data assessment, even if issues were identified. Data quality issues, even if they don’t directly impact the CECL process, should be brought up to the appropriate governance committee, and remediation plans should be initiated. Data quality metrics should be developed and continually monitored for all variables used in the process.
- Model Development: This is perhaps the most challenging governance process, as various levels of upper management will be presented with technical details that they may not be as familiar with. All key modeling decisions, assumptions and limitations should be brought to management for review and approval. Institutions should ensure that management receives sufficient information to support decisions on the R&S period for all products and that they understand the impacts of reversion assumptions. A thoughtful sensitivity analysis should be presented to the allowance committee periodically.
- Scenario Design: As scenarios are likely to have large impacts on the process, financial institutions should ensure sufficient internal subject matter expertise. For example, the use of exotic variables might pose a challenge for appropriate governance.
- Second- and Third-Line Involvement: Institutions should aim to include second and third lines of defense throughout the process. In our opinion, a CECL implementation best practice is to have second- and third-line work begin as processes are finalized to identify any gaps/weaknesses as the first line proceeds. Institutions should be cautious that working in parallel or one step behind poses challenges because future decisions by the first line could raise additional questions/concerns about prior processes; however, the advantages of identifying issues as early in processes as possible and gaining familiarity with early progress far outweighs the downsides, in our opinion.
- Source System Tie-Outs: Institutions should ensure that all internal data ties to source systems and has been reconciled, and that any data hops are verified along the way. Any data used in CECL processes should have verification that the resulting data is correct.
The previous ALLL process left some banks struggling to weather the Great Recession and, in some instances, resulted in government intervention. The new CECL guidance will require a forward-looking outlook, and for many banks that rely on more simplistic models and heuristics to forecast losses, it will be an arduous journey. While it might be tempting to simply repurpose or augment existing methods to produce a loss forecast that’s defendable to regulators, failure to think strategically about CECL will have long-term implications. After all, the purpose of CECL is to ensure banks set aside enough capital to cover losses in the next downturn. Put simply: Implementing CECL means a hit to profits in the form of a larger allowance, so executing effectively and knowledgeably matters.
Banks can create a competitive advantage in their loss forecasting processes by treating this as an opportunity to reimagine loss forecasting methods to use existing data more effectively, leverage automated frameworks, and begin incorporating new techniques like machine learning. The benefit most banks can gain from even incremental gains in accuracy far outweighs the initial investment in technology and processes. With superior forecasting, banks can conserve capital for investment elsewhere – and use their newfound insights to inform strategic moves on managing capital, extending credit to the best borrowers, and pivoting away from the ones who’ll be riskier in the next recession.
Banks that capitalize on CECL to better understand and manage their portfolio can do more than comply with the new standards; they could, quite literally, profit from them.
From data to model development, automation to reporting, and all things governance, Flying Phase can help financial institutions seize the opportunity to turn routine compliance into a competitive advantage.
Click here to contact us and learn more about the personalized approach we’d take on your CECL journey.