OPINION: Preparing For CECL: Six Considerations In Managing The Biggest Change In Bank Accounting History

As the most significant accounting change in banking history approaches, a surprisingly large number of financial institutions remain unprepared for the significant challenges involved with meeting the new Current Expected Credit Loss (CECL) standard.

Because the new rules require estimating an Expected Loss at origination for all loans, CECL could potentially prompt financial institutions to reexamine their lending practices. It’s also may also require additional loan loss reserves to be set aside.

Notably, in a survey of nearly 29,000 bankers, 33% expected their reserves to increase significantly. While the remainder of the respondents were not as concerned with major increases, or expected variation by loan type, not a single one expected to lower reserves under the new regime.

Bankers who have already headed down the compliance path know that CECL is a time-consuming and expensive proposition. For small- to mid-tier institutions with limited staffing, budgets or expertise, the key issue is whether to build a model internally or buy a solution externally.

The Upside of CECL

Although the CECL deadline may seem far off, the first cohort subject to the new standard – publicly traded banks – needs to be running parallel reserve calculations starting this January. While the process can seem overwhelming, banks that implement thoughtfully could realize unexpected advantages – provided they make the most of the opportunity.

The following are six recommendations for maximizing the implementation of CECL.

1. Before doing anything else, examine the quality and quantity of your loan data. Many institutions don’t have a centralized repository for storing loan loss data. That data is generally the most critical item in building, back-testing and validating Expected Loss models. In fact, many banks have difficulty generating a current loan tape complete with underwriting elements necessary for basic analytics because that information isn’t captured in their core systems. If a bank has an exceptional credit culture, it likely doesn’t have a statistically significant sample of defaulted loans required for CECL analysis. Ditto for de novo banks or recently acquired portfolios. Determining what you have, or what you don’t have, will dictate the process.
2. Select a model for predicting losses that best fits your available data.Model-building is always a challenging and time-consuming process. But is especially daunting under CECL requirements because the results must be predictive. In other words, not only does the model have to generate losses for all the various asset types based on individual loan characteristics, but it also needs to forecast future economic and market conditions for the life of the loan. Using those forecasts, the model has to predict a loan’s performance based on how those conditions will alter its characteristics at every pay period until maturity. Part of an auditor’s review of the finished model will include back-testing predicted results against actual loss data.
   
   After available loss history data is processed, the next step is to analyze whether to buy or build the model. If a bank has collected a statistically significant amount of loss history, it can build it internally or purchase commercially available software that will help frame the construction process. Either option is ambitious and time-consuming, but can be done in about six months if software is purchased, longer if built from scratch.
   
   If a bank has little or no loss history collected, there is really no option beyond purchasing a third party, end-to-end solution.
3. Use loan loss data and forecasts that are germane to your portfolio, not national level information. While there isn’t any one methodology that is absolutely recommended, it is generally accepted that CECL model results must be historically relatable to your lending footprint. Actual loan loss data is naturally more useful.  However, deep sets of actual loan loss data and the raw materials required to build forecasting models are very expensive.
4. Model Expected Losses at the loan level, not portfolio level, to optimize implementation and utility. In a recent survey, 70% of bankers planned to use Expected Loss results for portfolio management purposes or credit risk monitoring. To maximize your CECL investment and make it truly useful going forward, implement a model that delivers loan level results. At the loan level, Expected Loss data can help determine whether a specific loan should be held, monitored or sold, depending on its impact to reserves. In contrast, portfolio-level modeling doesn’t offer much insight or opportunity for value-added analysis.
5. Create an auditable plan to keep the model updated and assign staff accordingly.CECL modeling isn’t a one-and-done event; it requires regular updates to loan loss, market and economic data. If you choose an in-house solution or purchase software, make sure there is a team responsible for the updates, with adequate redundancy.
6. If you buy software or use a third-party provider, don’t buy a one-trick pony.Find a suite of functionality that can improve other areas of loan portfolio management and credit monitoring. Some solutions include stress testing, fair value, credit risk scoring, and data visualization tools, among other functions.

However painful CECL may be in the short-term, smart implementation can strengthen a bank’s balance sheet, lending operations and overall risk management.

Kingsley Greenland is CEO of DebtX, the world’s leading liquidity provider for whole loans. Managing Director Will Mercer is responsible for DXCDA, DebtX’s CECL solution.