Now that you’ve made decisions on your current expected credit loss (CECL) models and understand some of the hidden costs, let’s discuss the next hurdle –implementation. This article will help you understand the scope of work and documentation required in your CECL journey. In our experience working with over 150 institutions on CECL implementations, we’ve noticed the following common challenges.
Challenge 1 - Data requirements and processes
One of the biggest challenges of a CECL implementation comes from its tendency to generate data remediation projects institution-wide. While most allowance was done at the cohort or pool level under allowance for loan and lease losses (ALLL), you can choose to implement exposure level models in CECL.
Our clients see these exposure-level models as the key benefit of moving to CECL. These models will require additional loan characteristic inputs, and the lifetime of each loan, either through contractual or behavioral (based on past experience) maturity. Selecting and supporting a behavioral maturity is a good example of a common assumption, often used for open-ended lines of credit or credit cards.
Looking to get ahead of this potential challenge, collate all instrument-level data you can find within your institution. We often observe that the loan characteristics needed may be stored in core systems, but have received a lower level of scrutiny in terms of completeness and accuracy. Inconsistent inputs by various users in absence of a guidance is the most common issue we tackle. Getting your data game in order will put you ahead of the curve for a quick CECL implementation. That should be welcome news since the standard goes live on January 1, 2023.
You can make assumptions for missing or incomplete data, so let’s talk about some common ways to fill data gaps. Averages are a good starting point and can be supported with data analysis, and when there is no data available, most institutions turn to management assumptions. It is a good practice to discuss management assumptions at the committee level and record the meeting minutes to support conversations with audit and exam teams. CECL provides this latitude to make reasonable assumptions in the short term, while continuing to correct these data issues in the medium term.
If you are a rapidly growing institution (or aspire to be), you should implement some level of historical loss modeling for your CECL calculations. While CECL analysis calculates forward from a reporting date snapshot of your portfolio, multiple historical snapshots may also be required to calibrate these models. We recommend dedicating an additional 8 to 10 weeks to successfully implement these models, and if you are working with a solution provider, ask them about model choices and the impact on timelines.
Challenge 2 – Choosing reasonable and supportable periods and macroeconomic forecasts
Now that your data is in order, it is time to enter the forward-looking adjustments. CECL allowances are required to consider the impact of future economic conditions over a reasonable and supportable period before using reversion to historical long run averages. But there is no definitive guidance on how long this period should be. Many solution providers do not have the ability to offer well-documented forward looking macroeconomic scenarios and their impact on loss levels for your institution.
Your institution can implement an explicit or implicit reversion and make these assumptions at a portfolio or segment level. An explicit reversion means that you define the specific period over which the loss forecast returns to long-term average, for each asset class. When using explicit reversion, each asset class might have its own reasonable and supportable forecast period, and selecting each of these requires documentation and support for your choice. This can quickly create a daunting amount of analytic work to defend your choices, which is the reason many institutions prefer to use implicit reversion. Implicit reversion uses macroeconomic forecasts that converge to a long-term loss rate after an initial forecast period. Using the implicit reversion transfers the documentation needs to economists designing the macroeconomic views.
The last decision under the forecasting umbrella is selecting the number of scenarios to use. Scenario assessments comes in various flavors as does their implementations. An insurance actuary will likely run over 10,000 scenarios to assess the risk of earthquake for each property before making a decision on a premium. Thankfully, a bank only needs to look at a handful - an upside, a downside, a baseline, or a most- likely view. For quick implementation, begin with the baseline and adjust your approach in the future.
The key benefit of CECL is that your selection isn’t set in stone, and can be revisited regularly. While moving the framework from quarter to quarter may be too frequent, annual review and/or adjustments are expected.
Challenge 3 – Changing your allowance process
As you finalize the quarterly process, keep the following factors in mind:
- Auditors want to ensure that senior management was involved in all critical decisions made on modeling and assumptions. Recording your decisions in a memo is a good way to ensure review and approval is tracked.
- Key personnel dependency is a real challenge in today’s market. Personnel redundancy is critical in operating your CECL models and recording ongoing changes and assumptions. Having a single employee as your “CECL guru” exposes you to significant risk, potentially orphaning your CECL process and creating a costly training project to get you back up and running.
- Plan for reporting changes and, get ready to include economic forecast assumptions, including changes to key factors driving your allowance (GDP, unemployment, etc.). If you plan to use peer bank data for loss rates, you may also want to report on the changing performance of your peer group.
When are you “done” with a CECL implementation?
One of the key hallmarks of CECL as the past couple of years have shown us is the volatility in allowance rates and differences from one institution to the next. So, before you can say, “done,” plan on following:
- Educate your teams and set expectations. Ideally, run two parallel quarters, for everyone to get past excitement to acceptance. This is often an adjustment period for the institution, and it often falls to the implementation leads to act as primary educators. Prioritize your implementation so that you can have at least two parallel runs.
- Get in the rhythm of making the decisions, recording them, and finalizing your allowance. You can expect some lagging data remediation to continue at least into 2023. Management assumptions and scenario choices will now become a regular part of your quarterly process.
- Plan for iterative success with ongoing change and improvement. CECL adopters made benchmarking an integral part of the quarterly process. Benchmarking allows you to understand how your peers view the future. Senior management should expect greater involvement than under the incurred method. This has always been one of the major goals of CECL.
Moody’s Analytics credit risk data, models, economic forecasts, advisory services, and infrastructure solutions have successfully supported the CECL implementation of many financial institutions. To learn more about Moody’s Analytics solutions for CECL, visit us today.
