Industry Insights

Potential Data Needs & Considerations with CECL

February 2017
Author:  Chris Meriwether

Chris Meriwether



Financial Services

600 N. Hurstbourne Parkway, Suite 350
P.O. Box 22127
Louisville, KY 40252-0127 (40222)


In June 2016, the Financial Accounting Standards Board issued Accounting Standards Update (ASU) 2016-13, Financial Instruments–Credit Losses (Topic 326):  Measurement of Credit Losses on Financial Instruments, which finalized current expected credit losses (CECL) model guidance. Originally proposed in 2012, financial institutions have been anticipating the changes to the allowance for loan and lease losses (ALLL) calculation from CECL, including the effect on financial statements and operations. The finalized standard represents a major shift in accounting for credit losses and could potentially require substantial operational changes in the ALLL estimation process. One of CECL’s potentially largest operational challenges is that data needs to create an expected lifetime credit loss model. Even though today’s allowance estimation methods, e.g., loss rate and discounted cash flow models, are acceptable within CECL, the inputs to those models will need to change for proper standard implementation. Some upfront data gathering is necessary for the methods discussed in Topic 326 for loans held to maturity.

Model(s) Choice Effect on Data Needs

The new standard doesn’t require a certain model or methodology. Instead, it outlines options that behave differently and have different data requirements. The models discussed in Topic 326 are currently acceptable within today’s incurred loss model, but converting these to expected lifetime credit loss models requires additional data that institutions may not be capturing. The additional data needed will be more than just inputs for future forecasts. Before management can evaluate data availability and accuracy, the team first needs to determine what model(s) makes the most sense to consider during implementation. Depending on the complexity and unique risk drivers of an institution’s portfolio segments, management might determine different models are needed for different portfolios. Management may want to start by considering models consistent with current practice as both CECL and regulatory bodies allow approaches that build on existing credit risk management systems and processes and estimation models. For example, if a loss rate model is being used, it may make sense to consider loss rate model options within CECL. However, management should still consider that other models might be more appropriate given the portfolio’s risk and complexity. Management can help make the best methodology choice for the institution by considering multiple models and performing appropriate due diligence to determine how the model reacts to portfolios and the data needs of models.

It’s likely management may determine several viable models, in which case the team will need to perform due diligence to decide what model(s) to build out. When performing this due diligence, management should inventory potential data needed for each model and determine if the institution has internal or external access to all necessary historical data. If there are data constraints, management should create a plan for gathering data going forward. Regardless of the model, these considerations should be made when performing a data inventory:

  • Data Location:  Is the data stored in one or various systems?
  • Data Quality:  Is the data in a usable format, e.g., Excel? Is the data for the whole portfolio and accurate? Are there controls verifying data accuracy, and are they tested and reviewed by the audit function? With various loan systems, is there consistency among data attributes? 
  • Granularity:  Is the data available at the appropriate loan level for all necessary attributes? Do you expect changes in pooling method? If so, is loan-level data by pooling characteristics historically available, and does management believe the information is accurate?
  • Time Accessible:  As it’s possible institutions will need to access loan-level data for a longer historical time period with CECL, management should determine how far back they can obtain quality loan-level data in their loan systems. Limitations may require data-retention policy changes and the use of external data or other solutions when developing models.
  • Historical Consistency in Data:  Have there been acquisition or core conversions in the past that may call into question the reliability or consistency of historical data? 
  • Secure Data:  With the new standard, larger amounts of data will be maintained for longer periods of time. Management should ensure there’s a plan for secure data retention with backups and sufficient storage space available.

With CECL implementation, significantly more data will be needed to perform the allowance for loan losses calculation. To have the information available upon implementation, management should start making plans and gathering data now. Having sufficient data on hand will help management efficiently and effectively transition to the new model.

Contact your BKD advisor if you have questions.

BKD LinkedIn BKD Twitter BKD Youtube BKD Google Plus