ACL Model Validation Part II: Loan Portfolio Segmentation

ACL Model Validation
Part II: Loan Portfolio Segmentation
By: Peter Warmenhoven, Consultant

Financial institutions have transitioned to the Current Expected Credit Loss, or “CECL” method of estimating the Allowance for Credit Losses. Following regulatory expectations, many financial institutions are evaluating the effectiveness of their CECL process and looking for ways to improve the accuracy of their CECL result.  One of the most basic yet important elements of the CECL process to evaluate is portfolio segmentation; the way institutions divide their loan portfolios into pools of loans with similar risk characteristics.  Proper segmentation is the foundation of an accurate ACL estimate.  From each loan segment, loss histories and loan lifetimes must be determined, cohorts can be identified, and relative qualitative adjustments are applied.

FASB ASC Topic 326, the CECL Standard, requires expected losses to be evaluated on a collective, or pool, basis when financial assets share similar risk characteristics. However, neither FASB nor the regulatory agencies have prescribed a specific process for segmenting financial assets into pools for collective evaluation. Depending on portfolio size, composition, and complexity, a pool segmentation that’s appropriate for one institution may not be appropriate for another.  While there is no definitive approach for an institution to segment its portfolio, the CECL Standard does provide these key guidelines:

  • Segmentations or pools should have similar risk characteristics
  • Pools should be as granular as possible while maintaining statistical significance
  • Risk differentiation and economic responsiveness are key considerations
  • Both credit and non-credit related characteristics are relevant
  • Management should exercise judgment when establishing appropriate segments or pools
  • Management should evaluate financial asset segmentation on an ongoing basis

Many institutions, especially community based financial institutions, have segmented their portfolios to some extent according to the Call Report Codes for Schedule RC-C, Loans and Leases.  For Call Report purposes, each institution’s loan portfolio is already segmented into standardized pools according to loan  purpose, collateral type, or borrower. The call code represents a common starting place that virtually all institutions already track.  After first segmenting by call code, CECL pools can be refined by other criteria to break out significant sub-pools, or even by combining segments into larger pools according to shared risk characteristics.

Granularity and significance (in terms of pool size) are often competing priorities and can present conflict when deciding on segmentation. While call report segments may provide a good starting point, digging deeper into loan system data may uncover risk characteristics that warrant further portfolio breakouts. But with more sub-segments, each loan pool becomes smaller and may start to lack significance. Management should exercise judgement to ensure a proper balance between identifying a segmentation that has sufficient granularity to produce a meaningful result. Again, there is no exact method to achieve a balance between granularity and significance.  The conflict in these priorities can be illustrated in a few examples:

  • Bank A has a $500MM loan portfolio and primarily uses Call Report codes to segment loans for its ACL calculation. Bank A holds $75MM of loans secured by “Other nonfarm nonresidential properties” (call code 1.e.2.), which includes diverse income-producing property types such as hospitality, retail, office, and industrial properties. In drilling down on these property types after a period of economic stress, management noted that its loans secured by retail and hospitality properties incurred somewhat higher delinquency and loss rates than other property types. Bank A’s management observed that retail and hospitality property loans do not share risk characteristics equally with other 1.e.2. loans. As economic responsiveness is a key consideration in segmenting loan pools, and for its loan segments to be as granular as possible, Bank A should consider segmenting the retail and hospitality loans into a separate pool that reflects the higher risk. Bank A should also consider whether the new pool (and the remaining pool) is large enough to have statistical significance.
  • In another example, Bank B has a $500MM loan portfolio and primarily uses Call Report codes to segment loans for its ACL calculation. Bank B holds $25MM of HELOC loans (call code 1.c.1.) as well as $8MM of Home Equity Term loans (call code 1.c.2.b.). The bank’s loss history in both segments is very comparable. As both loan types are secured by junior liens on residential real estate and have comparable loss history, Bank B’s management believes these loan types share similar risk characteristics. Bank B should consider combining these two Call Report segments into one segment for ACL purposes.
  • Finally, Bank C has a $500MM loan portfolio and primarily uses Call Report codes to segment loans for its ACL calculation. One of the Bank’s more significant loan segments is owner-occupied commercial real estate (call code 1.e.1.), amounting to $50MM. Within this segment, Bank C has a number of SBA 504 loans, totaling $10MM and having an average loan/value of less than 50%. As the bank has never incurred a loss on an SBA 504 loan, Bank C’s management noted that this group has less inherent risk than other loans in the 1.e.1. segment because of the low LTVs. While SBA 504 loans generally may not share risk characteristics equally with other 1.e.1. loans, and for its loan segments to be as granular as possible, Bank C should consider segmenting the SBA 504 loans into a pool that reflects the reduced risk. Bank B must also decide whether the new pool is large enough to have statistical significance.

As set forth in the Interagency Policy Statement on Allowances for Credit Losses (Revised April 2023), “Management should evaluate financial asset segmentation on an ongoing basis to determine whether the financial assets in the pool continue to share similar risk characteristics.” An asset does not need to receive the same segmentation treatment throughout the asset’s life. Additionally, situations may arise when new or additional segments may be needed. Examples include the introduction of new products, significant changes to the bank’s underwriting standards or practices, or changes in repayment trends.

Lack of data availability is another reality that institutions must address. Loan accounting systems typically track loan data points such as loan type, purpose, collateral type, and borrower type. However, if your institution does not assign a call code to each loan at origination and includes that call code in loan system data, consider making that change in your booking process.

The interagency Policy Statement also states that Management should periodically validate the institution’s loss estimation process, to confirm that the process is appropriate for the institution’s size, complexity, and risk profile. The agencies’ expectation is that the validation process should include procedures for review by a party with appropriate knowledge and technical expertise, who is independent of the Bank’s credit and ACL process.

In 2023 and 2024, our ACL Validation service assisted numerous clients in validating their CECL methodologies and provided ongoing guidance for the ACL process. As the “new” accounting treatment approaches its first anniversary for many community-based financial institutions, some may not have undergone a regulatory review since its implementation, lacking regulatory insights into the ACL process.  An ACL Validation will help identify areas for improvement either prior to or in support of regulatory feedback.

To download a copy, please click here.