Collaborating author: Keith Boom
As we’ve previously discussed, CECL is an upcoming accounting hurdle for many banks, and existing Allowance for Loan and Lease Losses (ALLL) processes will require numerous enhancements to be compliant. In the spirit of building an efficient organization, and since regulators often encourage consistency in capital management across the organization, many banks plan to integrate Comprehensive Capital Analysis and Review (CCAR) and Dodd-Frank Act Stress Test (DFAST) capabilities with CECL while the hood is up.
In the first of this two-part blog series, we discuss the CCAR and DFAST stress tests and how CECL implementation teams can leverage stress testing credit loss models.
CCAR and DFAST Stress Tests
CECL implementation teams must understand what the CCAR and DFAST stress tests are and how they differ.
DFAST requires the Federal Reserve (Fed) to execute a supervisory stress test using data submitted by the banks and requires them to run their own stress tests. Key financial attributes are modeled and evaluated to assess if enough capital exists to remain solvent under differing economic scenarios.
CCAR requires banks to submit proposed capital action plans to the Fed, who assesses whether the bank can maintain minimum regulatory capital ratios during times of stress. Each bank’s stress testing practice also undergoes a thorough qualitative assessment by the Fed.
While DFAST and CCAR tests are complimentary, they are two different tests. CCAR gives banks more options to change capital levels and is only required for large banks. DFAST is required for both large and mid-sized banks, however, requirements for mid-sized banks are less strenuous with lowered expectations for model data granularity, sophistication of estimation approaches, reporting, frequency of tests, and public disclosures.
After seven years of running the stress tests, it’s safe to say these banks have built dynamic stress testing operations, and fortunately, they can leverage some capabilities for CECL. Next, we discuss how the models used in stress testing can be re-calibrated for CECL.
Credit Loss Models
CECL replaces the backward-looking incurred loss model with a forward-looking expected credit loss model. Similar forward-looking credit loss modeling methodologies are in production for stress tests, including roll-rate based models, transition state models and Probability of Default/Loss Given Default (PD/LGD) models.
No single estimation method is required for either stress tests or CECL, and the estimation method used for stress tests is typically chosen based on the bank’s size and complexity. For CECL, using a similar base methodology and segmentation for similar exposures ensures consistency and, as a bonus, these methodologies have already received approval internally by the bank and externally by regulators.
Banks may consider using stress test models as a starting point for exposures that overlap the allowance for credit losses and stress tests. Many components must be re-calibrated for CECL, such as:
- Stress testing models assume a non-static portfolio (e.g., loans entering through purchases and exiting through payoffs) whereas CECL requires a static portfolio (e.g., lifetime loss projections for a specific set of loans).
- CCAR projections are for 9 quarters, while CECL requires a timeframe extended through the life of the instrument.
- Large banks are required to create and use their own scenarios, which encompass forward-looking macroeconomic variables along with scenarios provided by the Fed, and mid-sized banks are only required to use Fed scenarios. Banks will need to evaluate if existing stress test scenarios are appropriate for CECL, or if they should create or source new ones.
- Stress testing models are inherently conservative given their focus on adverse scenarios only. Since CECL results will impact earnings and will be under scrutiny from investors, banks may want to reduce this inherent conservatism and the volatility of results by adjusting estimation assumptions and through qualitative adjustments.
With a better understanding of CCAR and DFAST, your CECL implementation team can begin to think about how to integrate stress test capabilities with CECL. For the modeling component of CECL, be sure your bank understands which CECL exposures already have an applicable stress test model. If so, you can engage your modelers to build incremental enhancements, as opposed to starting from scratch. In our next post we discuss the considerations for leveraging stress test data and IT infrastructure, as well as processes and controls, for CECL.