The new current expected credit loss standard affects more than just loan books. Under the new update, expected credit loss is recorded through an allowance for loan and lease losses in the financial statements. In contrast to the current “incurred loss” accounting method, the new CECL model requires forward-looking metrics that forecast credit losses throughout the life of a financial asset. Three groups of financial assets are affected: assets carried at amortized cost, purchased credit-deteriorated assets, and available-for-sale securities. The standard presents some unique challenges for structured finance investors due to the complicated and diverse nature of structured bonds. These include gathering of current data, projecting future performance, and mapping potential effects on triggers. Lastly, while the standard does not advocate any particular methodology, there are advantages to a discounted cash flow approach.
In June 2016, the Financial Accounting Standards Board (FASB) released Accounting Standards Update (ASU) 2016-13, which changed the method of accounting for credit loss from an incurred loss approach to a projected loss approach. Expected credit loss (ECL) will need to be calculated on the day of purchase or origination and will need to reflect lifetime loss. At each reporting date, ECL calculations will have to combine historical data, current financial conditions, and future outlooks.
The ASU specifically addresses three different kinds of financial assets that will all be affected differently. They include held-to-maturity (HTM) securities, available-for-sale securities (AFS), and purchased financial assets with credit deterioration (PCD).
The new current expected credit loss (CECL) model will only apply to financial assets measured at amortized cost (AC) and certain off-balance sheet items. More specifically, this includes HTM debt securities, loans, loan commitments, financial guarantees, and net investments in leases, as well as reinsurance and trade receivables.
Financial assets that fall within this scope will need to be pooled together based on similar credit risk characteristics. This is a deviation from the old standard which did not require pooling. This new requirement provides certain challenges, such as creating pooling methodologies and projecting losses for pools of assets instead of individually. This new methodology could generate some probability of default (PD) even for AAA-rated assets on a pooled or collective basis where there might be none on an individual basis.
All losses will be recorded the day of purchase or origination, and the allowance will be based off AC. The allowance will be affected by credit enhancements, which may limit losses. Depending on the nature of the collateral, the fair value (FV) of backing collateral can be reasonably assumed to be recoverable. Credit-enhancing derivatives will only affect ECL when they are embedded in the financial asset (i.e., they would travel with the asset when sold).
When all commercially available means to collect a loan balance are exhausted, the asset is written down to reflect a more permanent credit loss. However, recoveries are recorded when unexpected cash is received.
There are no specific models the ASU requires, but some examples include expected loss rate, vintage analysis, and discounted cash flow.
AFS securities do not measure ECL based on the CECL model. Instead, they use a modified other-than-temporary impairment (OTTI) approach, which requires a discounted cash flow approach. The new method no longer depends on the length of time an asset has been impaired and does not include a minimum threshold for losses. In this regard, the other-than-temporary aspect of the approach has been discontinued. Figure 1 compares the accounting implications of the legacy OTTI methodology with the new impairment approach, while Figure 2 provides an example of the change in calculations. For AFS securities, expected credit loss is measured whenever fair value (FV) falls below amortized cost. ECL no longer reduces amortized cost basis; instead, it is recorded in a contra account which is reassessed every reporting period and can be revised up. This means improvements in ECL will be immediately realized. This will also cause more volatility in ECL reporting. Unlike with HTM assets, pooling of securities is not allowed; assets are assessed on an individual level. Changes in FV that are not attributable to credit loss are still reported in other comprehensive income. Figure 3 shows a comparison of HTM and AFS treatment.
PCDs are assets that have more than insignificant credit deterioration since origination. What constitutes a significant credit deterioration is not explicitly defined, though credit ratings or PD could be used. A PCD is grossed up in value from FV by the amount of expected credit loss. The residual (interest-related) premium or discount is then amortized over time. The ECL calculation has to be reassessed each reporting period. The initial credit loss is reported on the balance sheet, whereas normally it would be reported in the profit and loss statement.
Projecting credit losses for structured security portfolios can be very tricky because characteristics of securities can vary widely, even within the same asset class and vintage. These unique traits highlight the importance of understanding details of each structure, found in deal documents, surveillance reports, and other reports. These are some of the unique challenges:
- SF deals can have complex structures, with various embedded instruments to manipulate the distribution of underlying cash flows.
- Certain adverse credit shocks and events can increase the credit risk of certain tranches, but they can also trigger events that make senior bonds even less risky. For example, if a deal’s payment structure changes from pro rata to sequential, then the most senior bonds are paid before other tranches, improving the chance their contracted payments are received.
- Each SF deal is backed by a unique and segregated pool. These pools of receivables generally would have been originated at different times with different concentrations, reflecting a unique risk profile.
- Collateral is not always purchased before bonds are sold (e.g., collateralized loan obligation (CLO) ramp-up periods). Certain asset classes (e.g., CLOs, credit cards, and student loans) could gain and lose collateral as the deal progresses (e.g., reinvestment or replenishment periods).
ASU 2016-13 does not require any specific methodology for the CECL model but offers examples such as expected loss rate, vintage analysis, and discounted cash flow (DCF). DCF models are the most defensible because they have an expansive set of inputs which generates robust results. These models rely on blended scenarios that larger banks can reuse from Comprehensive Capital Analysis and Review (CCAR) and Dodd-Frank Act Stress Test (DFAST) models (e.g., bank-specific baseline scenario). The standard requires incorporation of reasonable forward-looking assumptions, but a single scenario may miss crucial loss outcomes for SF securities. Blended scenarios that use at least one downside case can better capture losses for SF securities. This applies especially for mezzanine/junior tranches on a loss cliff. SF deals often have contingent characteristics, such as triggers that depend on credit quality. Small changes in economic assumptions could change whether a contingent characteristic is triggered or not. This may have a large effect on the credit quality of a tranche, as a deal might change from a pro rata waterfall structure to one that is sequential. It is important to understand the possibility of these effects and their impacts so they can be properly accounted for in ECL calculations.
SF deals also incorporate other market-based optionality inherent in deals, such as call options, where a case-by-case analysis may need to be performed, for instance, for call likelihood. In some cases, an assessment of this risk may not be possible if stated methodologies do not address these factors. Tranche seniority, thickness, and homogeneity of collateral pool also have large effects as to how different tranches within a deal will perform under different scenarios.
Another reason to use a DCF approach is that it is transparent and dynamically customized. Customization allows an institution to change how it approaches expected loss calculation based on an agreed-upon vision of the future economy.
The downside to DCF models is that they require abundant resources to run cash flow projections, such as credit models, performance data, and economic scenarios. This might limit DCF model use for smaller institutions for which such technical analysis is not feasible.
Although DCF models are resource-intensive, they may be necessary to accurately project ECL for SF portfolios due to the complicated nature of SF securities. Because SF securities are structured in different ways, their risk profiles can differ from deal to deal. Risk may be concentrated at the beginning or end of the life of the deal, depending on the structure of the deal and the subordination of the tranche. Triggers also affect the credit risk of tranches differently based on an economic outlook of the future. DCF models capture the effects of the individual characteristics of each deal. Economic scenario inputs for DCF models can simulate the effects that triggers and other contingent characteristics have on the credit risk for each individual tranche. Thus, to accurately project losses for SF securities that contain various nuances, a DCF model is recommended.
Financial Accounting Standards Board. “Financial Instruments – Credit Losses (Topic 326).” FASB Accounting Standards Update 2016-13. June 2016.
Sohini Chowdhury is a Director and Senior Economist with Moody’s Analytics, specializing in macroeconomic modeling and forecasting, scenario design, and market risk research, with a special focus on stress testing and CECL applications. Previously, she led the global team responsible for the Moody’s Analytics market risk forecasts and modeling services while managing custom scenarios projects for major financial institutions worldwide.
Leading economist; recognized authority and commentator on personal finance and credit, U.S. housing, economic trends and policy implications; innovator in econometric and credit modeling techniques.
David Fieldhouse is a Director of Consumer Credit Analytics at Moody’s Analytics, where he oversees the development of retail loan performance models for financial lending institutions.