Noting the diversity in portfolio sizes, complexities, as well as practices of applying the current incurred loss methodology, the FASB’s guidance on CECL offers quite a bit of latitude to financial institutions (FIs). It encourages FIs to leverage their current methods and existing systems in the application of a CECL compliant methodology. The general modeling strategies around CECL must incorporate the lifetime losses calculation, segmentation (one of the three pillars of CECL), determination and impact of adjustments, and the integration forecasts.
What is the Cohort Methodology for CECL?
A particular area of flexibility is with the determination of methodologies for the calculation of the allowance. One of the main methodologies FIs are using is the cohort methodology, which, as with all methodologies, requires institutions to make rational and defensible decisions.
The cohort methodology, or “snapshot” or “open-pool analysis,” relies on the creation of cohorts to capture loans that qualify for a particular segment, as of a point in time. They then track those loans over their remaining lives to determine their loss experience.
The most essential step in the utilization of the cohort-based methodology is the determination of the cohort – which starts with appropriate segmentation. The Update states that an “entity should aggregate financial assets on the basis of similar risk characteristics” when evaluating financial assets on a collective basis. Those characteristics include, but aren’t limited to, internal or external credit score, risk ratings, financial asset, loan type, collateral type, size, effective interest rate, term, or geographical location.
Benefits that come with segmentation include a reasonable way for institutions to identify their key vulnerabilities and assess how to manage those risks. Segmentation enables the institution to capture the unique behavioral characteristics that vary the degree of inherent risk or increase the likelihood of loss. The FASB suggests that FIs use at least two levels of disaggregation for their pools, though a third level, such as risk rating or risk grade level, is preferred. In considering the segregation of pools, it is easy to start broad, then sub-segment, as appropriate to what the portfolio allows.
Segmenting the risk pools will also ensure that the loss experience is applied appropriately in conjunction with the estimation of the loss reserves. Once the segmentation is determined, the mathematical applications become more straightforward.
Advantages and disadvantages
The cohort methodology is particularly attractive to many FIs, as it is very similar to what they do now. Most institutions currently use either annualized loss rates, and, may include a “loss emergence period” (LEP) multiplier. FASB has been clear that simply increasing that multiplier to represent the average life of loans in a segment will not be acceptable. A simple multiplier approach would likely greatly overstate the necessary allowance, as loans do not experience losses equally over their life cycle.
Some advantages of the cohort methodology:
- It requires fewer data fields/points to calculate than some other methods.
- It is less prone to “spikes” in losses in a particular year – as stated previously, loans do not experience losses in a linear fashion.
- It can be run on pools with broader risk characteristics than some other methods.
Some disadvantages of the cohort methodology:
- While smoothing out spikes may be advantageous, it can also be difficult to isolate the losses expected in the early years of a loan.
- It is more difficult to revert to the mean for non-forecasted periods than other methodologies.
- It may also require more “qualitative” assumptions for items like changing remaining lives than would be necessary in other methods.
The more quantitative the methodology, the less the reliance on qualitative adjustments. Consequently, with a cohort-based methodology, additional adjustments to the lifetime loss calculation will need to be made based off Q-factors, incorporating both current conditions as well as reasonable and supportable forecasted conditions.
An important expectation with Q-factors in the implementation of CECL will be the appropriate and more specific assignment of Q-factors. Considering the market information, any of the markers (economic trends such as unemployment rates, housing starts, etc.) used in evaluating the nine recommended factors from the Interagency Guidance can have a host of interpretations, each with specific application to different loan types. Quantitative and qualitative factor evaluations should be specific to the appropriate segmented pools.
Adjustments for directionality and magnitude can be determined using historical correlation testing between calculated loss rates and key performance indicators for those internal and external Q-factors. Determining correlations between historical loss experience and internal or external variables can help determine directionality of the impact of that factor as well as exclude factors that may have little or no impact on losses.
- The variables in the CECL allowance calculation rely on a variety of inputs including historical experience, current conditions and forecasted conditions. A recommended first step for FIs is to hold internal discussions reviewing the various methodologies to determine feasibility as well as alignment with existing processes. An important step in the review will be a data analysis to assess if the appropriate data is available to support the identified methodologies.
- As with any changes associated with the implementation of CECL, remember the Three P’s:
- Policy – the framework for the bank’s lending activities, setting the underwriting standards for credit decisions
- Process – the policy establishes the lending process and the responsibilities of the people.
- People – the individuals executing the process
Additionally, be sure to look at the tools used in the execution of the process and the training required to execute the process to come to a loan decision.
- Sample levels of disaggregation can start at the major segmentation, then move toward more granular levels of segmentation: minor segmentation, class segmentation, risk identifier.
- Relevant and appropriate application of Q-factors is required; Q-factors are virtually useless unless applied to a well-segmented portfolio.
- For periods beyond which forecasts of expected credit losses are reasonable, historical loss information may be used to fill in those spaces.