CECL: Are US banks ready?

Experiences from IFRS 9

To put these CECL challenges in perspective, it may be helpful to look at what happened outside the United States. In the rest of the world, IFRS 9 is the global standard issued by the International Accounting Standards Board to tackle expected credit statements. It was completed and published in 2014, and it has been adopted by more than 140 countries across the globe. Like the CECL standard, it seeks to provide more forward-looking accounting for impairment by moving from a discontinued approach to an expected loss model that considers the size and timing of potential losses.

IFRS 9 differs from the CECL in that it uses a three-step classification to assess the time horizon used for reservation. The first phase – for exposures that work as expected – uses a shorter 12-month credit period. As the credit quality of an exposure deteriorates, it enters the second and third phases based on threshold rules, and like the FASB’s CECL model, it requires an expected life expectancy reserve.

Because IFRS 9 was implemented before the CECL, US financial institutions have the opportunity to learn from their international counterparts. While implementations are unique specifically due to each organization’s internal systems and processes, there are some common themes and practices that emerged:

  • Planning.
  • Data.
  • Models.
  • Control.

Planning: Coordinating resources pays off

Many banks working towards IFRS 9 initially underestimated the amount of work involved. As a result, they slowly started with limited dedicated resources and then found that their timelines are compressed and in some cases forgotten. To combat this, many organizations threw more people at the problem. This was a tactical approach but was not sustainable or cost effective. In order to have a more sustainable approach to CECL, institutions need to take an integrated and holistic view that incorporates people, processes and systems.

IFRS 9 banks also found that the necessary coordination and integration between risk and financing was more crucial than expected. As the effort required was greater than originally estimated, many institutions had to increase their budgets and staffing accordingly. In many cases, the banks also realized that they needed help. There were several cases where the domain knowledge was not deep enough or where the necessary change management skills to organize the organizational and process changes were not strong enough.

In addition, organizations needed to consider the overall sustainability of revised processes in the future when planning for implementation. Are these processes robust? Maintainable? Sound?

All of this suggests a need for U.S. institutions to take a more detailed look at how they address the CECL and proactively plan investments in IT, programs, and resources to avoid potential problems along the way.

Data: Start collecting and preparing all your data as early as possible

Although data challenges may be closely linked to modeling challenges, there are some unique data issues that banks working through IFRS 9 need to address.

First, the requirements for additional data have posed challenges for a number of institutions. For example, cash flow modeling requires the integration of data from both risk and finance to modeling losses and cash flows (cash flows). These data are often placed in different systems, have different data definitions, are often populated at different times, and have varying levels of detail.

And different expertise is needed to clean the data and use expert assessments to customize or expand them, especially in preparing them for credit modeling. This has increased the time it takes to model credit losses, and raised issues for reconciliations and maintenance of audit trails.

An ongoing challenge is dealing with missing or incomplete data. For certain data gaps, institutions have sought publicly available industry data or third-party data. This reliance on third-party data often creates problems with definitions, timing, granularity, and data strain. Another problem is that the accompanying documentation may not be sufficiently transparent. A further challenge with reliance on third-party data is that it may be difficult to justify the use or defend its appropriateness or relevance. For example, industry loss data may not reflect the risk demographics of an institution’s specific portfolio.

One consideration for CECL (but not required for IFRS 9) is vintage analysis. Although modeling at the vintage level is not prescribed, the reporting aspect requires that sufficiently detailed data be maintained throughout the process. When reporting is more detailed than the models it relies on, problems can occur. A significant amount of reporting is required to meet the regulatory requirements as well as the needs of investors and management.

A key insight from IFRS 9 is that there is a need for early, timely analysis and documentation of data (including sources, quality, history and enrichment) and often takes longer and involves more resources than expected. In addition, data availability and quality can also affect the types of models to be used.

Models: Streamlined and pragmatic models are better than ‘perfect’ models

One of the early and often revised decisions that institutions face is deciding whether to use their existing models or whether new models are needed. For example, some organizations initially looked at their point-in-time loss models that were used for stress tests to determine if the level of detail and horizon would be acceptable. In some cases the answer was “yes” and in some cases “no.”

For loss estimation where no model exists or where the level of detail is insufficient, new models must be developed. Some experiments may be required as different model types, segmentation schemes, and assumptions are examined and reviewed. Estimates of lifetime loss can vary widely based on changes in many parameters, so early testing is needed to ensure stability in actual use.

Another decision that institutions face is related to the level of sophistication of their models. Many IFRS 9 banks originally planned to start with highly sophisticated models. Over time, they realized the consequences of this (long development cycles, data requirements, etc.) and became more pragmatic. Given the complexity and time required to review and implement sophisticated expected credit loss models, many banks moved to more streamlined models and construct processes that can be improved over time.

One lesson here for U.S. institutions still implementing the CECL is that perfection should not be an initial goal. Institutions need to start with simple models and introduce improvements step by step. This allows companies to start establishing a workflow and process to deliver and review the potential effects on ALLL from the various inputs and model types. A step-by-step approach will provide flexibility to adapt to changes in interpretations and any future regulatory guidance that arises.

Management: Now start building audit and monitoring functions

Although credit loss modeling is one of the most visible features of IFRS 9, banks that focused exclusively on the modeling aspects of the new accounting standard often ended up backing up and retrofitting their processes to establish proper control. Management of the entire process, including all aspects of data, models, integration and reporting, is crucial to a successful outcome.

Banks that have also built their processes into silos have struggled to gather a comprehensive set of data for review. Having a holistic approach with transparent and understandable processes facilitates auditability, supports repeatability and reduces “key person” risks. And because reserves are a critical component of financial statements, the consequences of failure can be much more immediate and severe (consider the possibility of a re-adjustment of financial accounts in relation to a significant risk assessment).

Given this need for transparency and coordination, U.S. firms need to plan and build robust auditing capabilities, appropriate management and compliance with financial reporting, and create a framework that supports model risk management expectations.