CECL: The New FASB Requirements for Loss Accounting

Forecasting credit losses over the entire lifetime
of your loans and leases

The Great Recession caught many financial institutions off guard. To reduce the risk of a similar crisis in the future, the Financial Accounting Standards Board (FASB) and the International Accounting Standards Board (IASB) have adopted new rules requiring lenders to maintain loss reserves sufficient to cover expected future credit losses. As these are accounting rules, not government regulation, they are unlikely to be affected by the current political climate toward deregulation.

FASB’s new Current Expected Credit Loss (CECL) requirements apply to any business issuing financial statements, and to the entire remaining lifetime of every asset that could experience a credit loss, including loans, leases, bonds, trade receivables and off-balance-sheet exposures, such as letters of credit. Current FAS 5/ASC 450 rules are substantially different, only requiring lenders to reserve for incurred losses, losses that are both probable and reasonably estimable, and anticipated to be taken in the next year or two.

The impact of the new rules will be recognized on day one, with quarterly adjustments to the forecast thereafter for every open exposure. The impact varies significantly by institution and credit type; transactions with longer terms are affected more than those with shorter terms. Fitch estimates that CECL will increase the reserves of U.S. banks from $135 billion to about $185 to $235 billion.

FASB does not prescribe specific methodologies, nor are lenders required to incur “undue cost and effort” to obtain future loss forecasts. Forecasts are simply required to be “reasonable and supportable.” Relevant information, such historical experience, current conditions and all reasonably available information, should be considered. The unit of account can be the individual transaction level, or it can be pools of assets with similar risk characteristics (e.g., risk ratings, asset type, collateral type, size, term, industry, geography, vintage). All methods require historical loss data, preferably including a full economic cycle.

“Top-down” loss forecasting methods estimate losses at the portfolio level. For example, quarterly forecasts can be obtained by calculating historical average loss rates for homogeneous risk pools and applying adjusted versions of these rates to segments of the current portfolio. Top-down methods are generally easier; however, these forecasts involve a measure of subjectivity, compromising their accuracy. Furthermore, they are of limited use in contributing to portfolio risk management goals.




“Bottom-up” methods calculate expected losses at the individual transaction level. These methods require more data and effort, but will be more accurate, consistent and defensible over time. Moreover, expected loss data can be combined with pricing and overhead data to calculate profitability by transaction, program or asset type, for use in portfolio management, collections, funding, pricing, target marketing and potentially credit decisioning and structuring. Bottom-up loss forecasting methods generally use the Basel framework of:

Probability of Default x Exposure at Default x Loss Given Default = Expected Loss

PayNet has utilized its massive database of more than 24 million contracts to conduct extensive research on loss forecasting. This database was used to develop a model for equipment finance loss forecasting that is an amalgam of more than 300 underlying models for different asset types and transaction sizes, going out to seven years in the future. In the process, a great deal was discovered about the drivers of loss, which should be helpful to those building their own models, as well as to credit practitioners generally in decisioning and structuring transactions.

The myth is that Loss Given Default is “all about the collateral and its value.” The reality, however, as proved by the data, is that collateral is only one of many factors that determine LGD. In total 10 factors, in three broad categories, were identified (see figure above).

While CECL will be something of a burden for everyone, there is a substantial silver lining in better understanding expected loss.

While many potential factors were tested as part of the research, those listed in the figure above proved to be both statistically significant and explainable from a common-sense point of view. For example, transaction size has a large impact on Loss Given Default. While 15.5% of defaulted transactions under $10,000 were total losses (100% of EAD), only 2.9% of transactions over $1mm were total losses. This makes sense, as very small deals are often not worth the cost of seriously pursuing, repossessing or litigating—whereas a credit manager with a seven-figure deal that’s a total loss might want to think about finding a new line of work!

Credit quality materially impacts Loss Given Default. Among borrowers with very high credit quality, such as hospitals, universities, schools and government, two-thirds of defaults were administrative, “nuisance defaults,” with no loss whatsoever. But even those industries aside, LGD is usually more than twice as much for the lowest quality credits versus the highest quality ones—as they say, “you can’t get blood from a stone.” Similarly, lenders with experience in an asset class knew better what to finance and with whom, and how to maximize proceeds from the collateral if the deal went south—which new market entrants would not know as well.

While CECL will be something of a burden for everyone, there is a substantial silver lining in better understanding expected loss. Just as PDs that provide specific Probabilities of Default are more useful than credit scores that simply rank-order risk, so are ELs better than PDs, as loss is what really matters in lending. Whatever methods lenders decide to use to satisfy CECL, the impact on reserves is potentially large enough that it’s worth addressing these rules early, to manage the impact of CECL when it arrives.