ABA Comment Letter to AICPA re: Proposed Statement on Auditing Standards – Auditing Accounting Estimates and Related Disclosures

Sherry Hazel
American Institute of Certified Public Accountants

The American Bankers Association (ABA) appreciates the opportunity to comment on the Proposed Statement on Auditing Standards – Auditing Accounting Estimates and Related Disclosures (the Proposal). We are commenting on this Proposal due to the potential implications the final standard may have on audits of banks. Along with the recent auditing proposed auditing standard on audit evidence, the Proposal addresses key issues that banks will face from auditors as a result of the Current Expected Credit Loss (CECL) accounting standard for the measurement of credit losses. Along with its impact on audit evidence, CECL changes the nature of the accounting estimate – now one that forecasts future credit events that do not conform to any market consensus – and will put to the test many of the estimation concepts discussed in the Proposal. The CECL allowance is expected to often be highly subjective and, therefore, subject to significant possible management bias. The range of reasonably possible CECL estimation outcomes will normally be multiples of materiality. It is through this lens that ABA views this Proposal.

Based on the recent delay of the effective date of CECL for most companies, ABA believes this standard will apply mainly to audits starting in 2023. With that in mind, however, CECL implementation will be a long and costly process for smaller institutions, many of whom are in the midst of making investment decisions related to systems needed to support CECL processes Conflicting messages coming from FASB, the banking agencies, and auditing and consulting firms, have many institutions hesitating in their implementation decisions. The final standard should provide some clarity on auditing expectations.

Specifically, the ASB should address how the final standard would address audits of smaller institutions. The nature and extent of the evidence smaller banks believe they will need to provide to their auditors will highly influence how they will configure not only their CECL estimate models but also the operational systems that feed into their models. Addressing smaller banks, therefore, should be a priority in an Accounting and Auditing Guide. However, just as with what has been proposed regarding “professional skepticism,” we believe this issue should also be directly integrated into future auditing standards, such as this one. This is a major point that was made in our comment letter related to the Audit Evidence proposal.

On the following pages, ABA provides specific points to consider to clarify how an auditor would apply the guidance in the Proposal.

Terminology of Modeling Needs Clarification in Light of Credit Loss Estimates

ABA generally agrees with the definitions of method and model (in paragraph A2, page 29) of the Proposal. However, by describing an overlay merely as an “adjustment” made to the output of the model (in paragraph A40, page 39), we believe this distorts how credit losses are modeled and, thus, may confuse auditors in assessing the various aspects of model risk.

As with other accounting estimates, specific estimation methods will be used as models to estimate credit losses under CECL. Most CECL methods, however, do not fully take into account certain credit risk factors and these factors are then analyzed separately. Adjustments may then be made to the initially modeled results and such adjustments might be in the form of an overlay and (as discussed in A40) made to the output of the model. An example of this is a method that analyzes detailed charge-off data by the age of the auto loan, though does not take into consideration of the current levels of delinquencies in the portfolio. To compensate for this, an additional analysis of historical defaults based on borrower delinquency may then be used by a bank as the basis for an overlay adjustment, which is applied on top of the original estimate. In this case, ABA believes that the “model” consists of both the original method, plus the provision for delinquency. There really is no functional “overlay” to the process.

Another example of an overlay may be due to the historic imprecision of the model used, based on a model validation process. Here, ABA believes that the “model” many not necessarily consist of both the original method and the precision adjustment. If not, the precision adjustment conforms to the adjustment as defined in A40.

On the other hand, risk factors often may be addressed in the CECL model, though not with sufficient precision. The related adjustments may then be applied to outputs of or to inputs to the model. An example is an adjustment made to an economic forecast of unemployment provided by an independent organization, which is an input to the model. Conversely, an adjustment made to the loss given default (based on trends in collateral prices) may be considered to be made to an output of the model or to an input.

Practically speaking, many CECL estimates will be based on various analyses and “sub-models” that adjust the results of the base “CECL model” used in a portfolio. Based on the models and historical data that are used, adjustments from the initially-modeled outputs can be significant. In fact, ABA is hearing from certain CECL implementation consultants that “qualitative adjustments” can often make up 80-90% of the final forecasted loss. Understanding this can change the auditor’s assessment of risk and the necessary governance and internal controls needed to mitigate management bias and estimation uncertainty. ABA also believes that there is an inherent bias that auditors may have toward the accuracy of the base model, with more skepticism toward the adjustments. This would not be appropriate. With all this in mind, we believe a more robust discussion on models and adjustments is needed for auditors to appropriately evaluate the control risks.

Rebuttable Presumptions may be Unintended

Specific sections of the Proposal indicate presumptions that may be unintended:

• “A64. Many accounting estimates do not require the application of specialized skills or knowledge. For example, specialized skills or knowledge may not be needed for a simple inventory obsolescence calculation. However, for example, for expected credit losses of a banking institution or an insurance contract liability for an insurance entity, the auditor is likely to conclude that it is necessary to apply specialized skills or knowledge.”
• “A68. The reasons for the auditor’s assessment of inherent risk at the relevant assertion level may result from one or more of the inherent risk factors of estimation uncertainty, complexity, subjectivity, or other inherent risk factors. …Accounting estimates for expected credit losses are also likely to be subject to high estimation uncertainty and significant subjectivity in making judgments about future events or conditions.”

These statements appear to be rebuttable presumptions that banks will need significant internal controls to address estimation uncertainty and possible management bias and that specialized skills and knowledge will normally be needed. ABA is concerned that these and other examples reflect a biased presumption that the banking industry is inherently more complex than a typical industry or that a credit loss estimate is more complex than other accounting estimates examples listed in the proposed standard. This creates a presumption that banks have a higher level of inherent risk or require a level of specialization that is incremental relative to the range of those assessments in other industries. In fact, assertions by FASB seem to contradict the presumed need for specialized skills or regard for high estimation uncertainty at many institutions:

“There is no expectation that a less complex entity should have to implement a sophisticated model to satisfy the requirements of Update 2016-13.”

ABA recommends that these rebuttable presumptions be clarified or that language be inserted that eliminates the apparent presumption that highly specialized auditors are normally necessary in a bank audit.

The Range of Outcomes should be Limited to those that are Reasonably Possible

In paragraph 12 of the Proposal, in reviewing an entity’s internal controls, the auditor assesses how management:

“…understands the degree of estimation uncertainty, including by considering the range of possible measurement outcomes…”

ABA believes that, in order to maintain consistency with other references to measurement ranges, that the wording be clarified to note the “range of reasonably possible measurement outcomes.”

Thank you for your attention to this matter and for considering our requests. Please feel free to contact me ([email protected]; 202-663-4986) if you would like to discuss this further.

Sincerely,
Michael L. Gullette