It’s unlikely that there is a financial institution in America that doesn’t know it will need more data to estimate its allowance under the coming CECL—Current Expected Credit Loss—accounting standard than under current GAAP. It’s been the mantra of auditors, regulators, and advisors since well before the standard was finalized and guidance was issued.
But questions arise. What kind of data? And how should the data be organized?
Data collection must be improved
In a MST Advisory webinar earlier this year, a cursory audit of data submitted by attending lenders revealed that none had the historical data available sufficient to support a CECL-compliant methodology. Indeed, they said that they weren’t collecting data in all the fields they will require, even though the information is likely retrievable from their core systems.
In an MST survey completed in mid-February, only 20% of lenders responding said they currently have sufficient data to implement a CECL-compliant methodology. Furthermore, although 86% said they are still in the initial stage of internal discussions on how to prepare for CECL, another 37% said they already know they will need more or better data than they have.
As well, respondents to an MST 2017 CECL Survey indicate how complicated the data gathering process can be. More than half of the 41 data fields listed in the survey were being targeted by more than half of the respondents.
Of course, the types of data needed for CECL will vary from institution to institution. It will depend on the types of loans being made and how loans are organized into the segments required to estimate expected losses.
Still, a comprehensive list of loan elements—those 41 data fields listed in the MST survey—is a great place to start. From there, a lender can narrow the list to its own requirements.
Based on a study of FASB’s ASU 2016-13, MST developed the following list of recommended loan data elements in four categories: basic loan data, basic segmentation, loan risk segmentation and transaction.
Basic Loan Data Elements
These are the basic loan data fields that should underpin any future CECL or current ALLL methodology. They should exist on any loan in virtually any core system, although they might be identified differently in different systems.
• Commitment amount at origination
• Balance at origination
• Origination date
• Renewal date (important because a loan renewal with a full re-underwriting will constitute a “reset” in the life of the loan)
• Extension date (though it is unclear if a pure extension of a loan without a full re-underwriting will constitute a “reset” of the life of that loan)
• Interest rate at origination
• Current interest rate
• Interest rate type (fixed or variable)
• Current unpaid balance (balance of the loan net of participations, gross of partial charge-offs and non-accrual interest applied to principal)
• Charged-off principal
• Non-accrual interest applied to principal (f the loan has had at any time non-accrual status, the amount of interest collected during that period that was applied to the principal of the loan)
• Current book balance (balance of the loan net of participations, partial charge-offs and non-accrual interest applied to principal)
• Government guaranteed balance (balance of any bank-retained government guarantee—SBA, USDA—that would cover potential losses)
• Current undisbursed commitment amount
Loan Basic Segmentation Elements
Under CECL, lenders will be required to categorize loans into segments or pools based on the characteristics of the loans. You can use the following list of pools to segment loans. Unlike the Basic Loan Data Elements, these fields can be core-specific or institution-specific.
If your core does not include these fields or they are not relevant to your institution, they need not be included in your data collection efforts.
• Loan type
• Purpose code
• Collateral code (generally the primary type of collateral if the loan is supported by multiple types of collateral)
• Group code
• Class code
• Owner-occupied code
• Call report code
• NAICS industry codes
• Property type
• Weighted average lives
Loan Risk Segmentation Elements
These fields can be used to segment loans in a general segment into risk-based categories, or to analyze loans based on specific risk indicators. Some items could be core-specific or institution-specific—or even specific to particular types of loans. For some items, we have listed both the current and original value (or value at the most recent underwriting).
Theoretically, if your bank’s data cover a long enough historical period, the original value will be included in a loan or segmentation data set. However, if your historical data availability is limited, it would be helpful to include the original values here as well.
• Risk rating (all changes and dates of changes)
• Credit score (all changes and dates of changes)
• Current days past due
• LTV (original, current, date of last change)
• DSC (original, current, date of last change)
• NOI (original, current, date of last change)
• Number of times delinquent (30+ days, 60+ days, 90+ days, etc.)
• Account status
• Date to non-accrual
Unlike most of the other data elements, which are point-in-time in nature, some information will be needed at the transactional level. You could potentially have multiple transaction records per loan record.
Charge-off and recovery information
• Loan number
• Transaction date
• Charge-off amount
• Recovery amount
• Loan number
• Data of default
• Default reason (move to non-accrual, 90+ days past due, etc.)
• Balance at default
Combined, these four groups serve as a checklist of data elements key to consider in your data gathering process en route to a CECL-compliant methodology.
About the author
Dalton Sirmans is CEO of MST.
Download a worksheet from MST’s website [Registration required]
- Banking Exchange to Host A Second Panel on Faster Payments
- Branch News: BMO Harris Buys Commercial Office, Citizens Bank Adds WV Locations
- Big Banks Outperform Regions in Mobile Apps, Says Survey
- As Bank Branches Go Digital, How Do They Attract Customers?
- Building A Data-Driven Culture the Right Way: Five Lessons to Build Better Relationships