Data Fields: What Types of Data Should You Gather for CECL?
The first and most often biggest concern of institutions as they transition to CECL is data – not only how much is required, but what kinds of data. Accordingly, there is no shortage of information available on the various data fields that could be required to estimate losses under CECL, but the data you’ll actually need will depend on other determinations during your transition to CECL estimation, including, but not limited to, how you segment your portfolio and which CECL-compliant methodology, or methodologies, you choose to employ.
Types of data for CECL
Basic loan information. Because you’ll be estimating losses based on the life of each loan, virtually all methodologies will require such basic loan level data as loan numbers, balances, origination dates and borrower information. Loan balance data should include contributors to the amortized cost basis of your loans, including any unamortized premiums or discounts, as well as accrued interest, net deferred fees and costs, and such items as any guaranteed balances. Similar instrument-level information would be necessary for other in-scope assets.
Transaction level data. Includes loss and recovery information that ties back to those loans, such as charge-offs, recoveries, and payments, including pre-payments. You will need the dates of those transactions, as well as when loans have been restructured as TDRs.
Loan status changes. If a loan is on nonaccrual or has been taken off non-accrual, or becomes a TDR, you should have data on these indicators of default probability, including dates of changing status, as well as your definition of what constitutes a default, such as passing certain delinquency thresholds.
Segmentation data. Segmenting by risk within a pool or its subcategories, just as you do today, is appropriate for CECL calculations. Identify the key characteristics that expose the loan segment to risk and the associated data. Ask yourself first how am I going to segment the parent pools, such as by call codes or loan types, and how those will be subdivided. For example, C&I loans might be segmented by industry code, commercial loans by collateral type or loan size, residential loans by geographical locations, and so on. Because you will need certain data fields to estimate your pools, your pool segmentation, and by extension your methodology, could be determined by the types of loan data you have.
Risk data. Credit quality indicators include risk ratings, risk grades and other types of quality ratings, such as days delinquent and updated FICO scores. What you’ll need will be driven in part by your pool segments; for example, risk ratings for commercial loans and delinquency buckets for consumer loans.
Data depends upon methodology chosen
Your methodology will determine which data fields you will need for your CECL estimation. And again, the reverse could be true, that you choose a methodology based on the quantity, quality, consistency and types of data you have. It will be important to keep flexibility in mind, as the result of your initial CECL audits may require you to change your model, which in turn could require different data sets.
From providing integration support to CECL transition guidance, Abrigo Advisory Services works to alleviate the regulatory burden placed on banks and credit unions and help them proactively monitor risk and identify trouble spots in the portfolio. Learn more.
Check out additional information about CECL data in this CECL Prep Guide.