The Current Expected Credit Loss (CECL) standard better aligns the allowance with credit risk management and underwriting practices¹, and many financial institutions plan to leverage their stress testing models as a basis for the CECL implementation.

Although entities are permitted to use their current credit risk management approaches as a framework for the CECL measurement², the systems used for stress testing will require significant enhancements before they are appropriate for financial statement reporting. Therefore, for institutions that plan to leverage their stress testing procedures, addressing the operational differences between stress testing and the allowance process³ will be critical to a successful CECL implementation.

1. Establishing controls over the end-to-end reserving process

Unlike regulatory stress testing, the allowance estimate has always been subject to SOX controls over financial statement reporting.

Although larger banks have experience with forward-looking models, many of the practices and procedures used for stress testing will be unacceptable for audited financial statements. For example, model exexution platforms provide the flexibility to swap out models, change assumptions, and compare results between scenarios; this will be extremely beneficial for CECL, as it has proven to be for stress testing. However, most model execution platforms used for stress testing lack the controls required of a financial statement estimate. The model execution platform leveraged for CECL should include the appropriate controls (a full audit trail of model runs, role-based permissions, segregation of duties, and data lineage tracking).

2. Ensuring the accuracy and integrity of credit risk data

Although the data used for CECL may be similar, it will be subject to higher standards of accuracy and integrity than that used for stress testing purposes.

When performing stress testing, few institutions apply a data governance framework comparable to that required of financial statement reporting. According to the “Principles for Effective Risk Data Aggregation and Risk Reporting” (January 2013) by the Basel Committee on Banking Supervision (BCBS), more commonly known as BCBS 239, “controls surrounding risk data should be as robust as those applicable to accounting data.” However, only systemically important banks (SIBs) are subject to BCBS 239, which means that of the 100+ U.S. banks that perform stress testing, fewer than 10% are required to apply the same high standards to both accounting and risk data. As a result, most U.S. banks’ regulatory stress testing processes do not meet BCBS 239 or financial reporting standards for data governance.

When establishing a strong data governance program for the CECL estimate, consider the BCBS 239 principles, which include monitoring risk data for accuracy and integrity, reconciling risk data with accounting data, tracking and recording full data lineage, and establishing a single authoritative source for risk data.

3. Condensing the process into a shorter duration and executing in real time

By the time the Dodd-Frank Act Stress Test (DFAST) scenarios are released in February, the data to which the scenarios are applied have long been finalized; DFAST scenarios are applied to data as of the end of the previous annual reporting period. This is in contrast to the data used to generate the allowance, which is typically in flux while the financial statements are compiled, audited, and ultimately, filed.

Stress testing procedures are conducted over the span of several weeks to months, while the allowance process will be completed in a matter of days in order to close the books, complete the audit, and file the financial statements. DFAST is performed on an annual and semi-annual basis by banks with total assets between $10 to $50 billion and more than $50 billion respectively. The allowance process, on the other hand, will be performed on a quarterly basis at a minimum.

The CECL reserve will be calculated and disclosed more frequently, over a shorter duration, and will leverage “real-time” data. Integrating the models used to generate expected credit losses with the applications used for accounting and reporting will be critical to doing more in less time. Also known as “productionalizing” the allowance process (integrating the execution of the models with the data inputs and reporting in a controlled and repeatable manner), it will be essential to a sustainable and repeatable reserving process.

4. Explaining a potentially volatile estimate to stakeholders

Like regulatory stress testing, the CECL estimate will be sensitive to forecasted conditions. However, because the allowance will represent management’s expectations about the future, rather than hypothetical conditions, it will prompt more questions from investors and other stakeholders, putting more pressure on reporting and analytics.

Unlike the DFAST scenarios issued by the Federal Reserve, the forecasts used for the CECL reserve will be selected by the bank and will differ throughout the industry. Preparers will need to support why a particular forecast represents management’s expectations of the future, in addition to explaining the drivers of change. Further, while the DFAST scenarios are at the national level, the forecast used for the CECL reserve will need to consider the impact to expected credit losses at the regional and local level unique to the bank’s portfolio.

The solution to constructing a narrative that tells a cohesive story is to implement a reporting framework that encompasses data inputs, model results, and multidimensional reports at a granular level in an automated fashion.

The foundation of a successful CECL implementation

A successful CECL reserving process must fulfill all requirements of the accounting close which will require taking a holistic approach that leverages the right framework to get the job done efficiently and effectively.

Banks subject to regulatory stress testing will be familiar with the challenges associated with CECL (e.g. executing forward-looking models and explaining the drivers of change). That said, most stress testing processes are overly manual and the tools are not integrated. Therefore, when leveraging stress testing processes as a basis for the CECL estimate, it will be imperative to bridge the gap between model execution, reporting, and analytics in a controlled manner.