Released October 2018
In this paper, we discuss the thorny issue of using industry data when performing a credit loss estimate under CECL.
We have found that the more granular the input, the more predictive the result and that the use of industry-wide data can provide meaningful insights.
OCC Provides Insights Regarding CECL Implementation
The magnitude of the changes required by CECL have created uncertainty regarding how to implement it. The Office of the Comptroller of the Currency (“OCC”) updated its Bank Accounting Advisory Series in August of this year. In the newly updated section 12 D Allowance for Credit Losses, the OCC addresses two of the questions most often asked of us:
1. Am I limited to one model or approach to calculate the reserve?
2. Can I rely exclusively on my own data?
Question 2 – How Should a Bank Measure Lifetime Expected Credit Losses?
OCC Staff Response
A bank will need to apply judgment to select an estimation method(s) that is appropriate and practical for its circumstances to measure expected credit losses. Various methods that reasonably estimate the expected collectability of the bank’s loans and that are applied consistently over time can be used. Acceptable methods include, but may not be limited to, loss rate, roll-rate, vintage, discounted cash flow, and probability of default/loss given default methods. No specific method is required for estimating expected credit losses. Additionally, a bank may utilize different methods for different groups of loans. When measuring lifetime expected credit losses, the bank must consider available information that is relevant to assessing the collectability of its loans. This information may include internal information, external information, or a combination of both relating to past events, including historical credit loss experience on loans with similar risk characteristics, current conditions, and reasonable and supportable forecasts that affect the collectability of the loans over their remaining contractual terms.
We note that Wilary Winn has consistently taken this position and believe the best model or approach depends on the type of loan being analyzed and what a financial institution plans to do with the analysis. We strongly believe that robust discounted cash flow models informed by statistically valid data sets represent the only practical solution for managing credit risk on homogenous loans (e.g., residential real estate, consumer, and C&I) on a prospective basis.
The OCC addressed the issue of valid data sets and model input in Question 7.
Question 7 – Does the Bank Need to supplement its historical loss experience with external (i.e., peer or market) data when determining its ACL?
OCC Staff Response
No. A low level of credit losses over an extended time period is not, by itself, a condition that would necessitate a bank defaulting to, or supplementing its loss experience with, external data. The bank may have a sufficient loss history to use its own experience as a starting point for its ACL, even though its credit losses have been minimal. In this fact pattern, the bank compared the characteristics of its current portfolio with the portfolio characteristics that generated its historical loss data. Because the nature, terms, volume, and underwriting standards of the current portfolio, as well as the bank’s expectations about future economic conditions, were similar to the portfolios and economic conditions that generated the loss experience, the bank will not need to supplement its historical loss experience with external data. Conversely, if the characteristics of the current portfolio or the bank’s expectations about future economic conditions had differed significantly from the portfolios and economic conditions that generated the loss experience, the bank would need to consider whether the use of external data, or appropriately supported qualitative adjustments to its own data, is necessary to appropriately reflect the bank’s expected credit losses.
We note that the OCC response is consistent with our view that robust discounted cash flow models represent the optimal implementation of CECL because the other approaches become very difficult to implement when underwriting conditions change or when a financial institution faces a different prospective economic environment than that which is recently experienced. Loss rate and vintage analyses are particularly difficult to modify given a change in either underwriting or economic conditions or especially when a financial institution faces both.