Archive

Credit Impairment Model Proposal

Discussion Paper:  U.S. Banking Industry Model for Credit Impairment (BIM)

Introduction

The incurred loss model under U.S. GAAP and IFRS that governs credit impairment has been criticized following the recent financial crisis.  The two main criticisms raised are: 1) reserve adequacy ( i.e., credit reserve levels at the inception of the financial crises were inadequate to absorb the elevated losses that occurred during the crisis and likely substantially existed at the inception of the crisis) and 2) timing of recognition ( i.e., increases in realized losses led to a combination of elevated charge-offs and large reserve builds at the peak of the crisis that inappropriately reduced market confidence in the banking sector).  Many believe the delayed timing of loss recognition and magnitude of loss, which together reflected a severe and rapid deterioration in credit quality, exacerbated the severity and length of the financial crisis.  All parties (standard setters, regulators, investors, and preparers) agree that improvements to the credit impairment guidance are necessary.  

We believe that the fundamental principles inherent in the incurred loss model are sound and have served the industry, regulators and financial statement users effectively by providing a well understood framework to determine credit-related allowances.  However, over time, the incurred loss model has increasingly been interpreted in a way that has resulted in a significant flaw: allowance calculations based on too narrow a view of the credit cycle. History has shown that the credit profile of financial instruments is highly cyclical, typically with a period of benign loss activity that coincides with the expansion and peak of overall economic activity and credit availability, followed by a shorter and more concentrated period of elevated credit losses.  Narrow interpretations and application of the incurred loss model result in the compression of this cycle by considering only losses estimated over an abbreviated loss emergence period and restricting the use of market trends and other data that would indicate changes in the probability or severity of loss until such deterioration is observable. 

The events of the recent financial crisis put a spotlight on this weakness in the application of the incurred loss model, resulting in the criticism noted above. Although we believe the fundamental principles of the incurred loss model remain sound, some thoughtful and tailored changes are necessary to incorporate the cyclical behavior of financial instruments and lack of transparency around inherent losses prior to the deterioration of the credit environment.

To date, the independent proposals from the FASB and IASB have focused primarily on only one aspect of the problems with the current model.  The FASB's "foreseeable future" model addresses reserve sufficiency through expansion of the loss definition and the time period covered by the forecast by removing the "probable" trigger and expanding the types of inputs that may be considered in a loss forecast.  However, the foreseeable future model suffers from the same flaw as the existing model, as it may be narrowly interpreted with a limited view of both the losses within the emergence period and of breadth and depth of the credit cycle.  Accordingly, we believe the "foreseeable future" methodology, as originally proposed, may not adequately address the weaknesses in the current model and may perpetuate and exacerbate the pro-cyclicality of results while only modestly increasing the absolute level of credit reserves during extended periods of benign credit activity.  Also, recent banking regulator comments lead us to believe this model, if adopted globally would potentially be implemented differently in the U.S. (most likely with longer "foreseeable future" loss forecasting periods) than in other jurisdictions. 

In contrast, the IASB attempted to address the relationship between credit losses and loan pricing and income recognition and, therefore, pro-cyclicality, by introducing a time-proportionate spreading of credit losses over the expected life of the portfolio.  However, this model does not result in a credit reserve that is sufficient to anticipate significant changes in credit loss curves as it also may focus on a narrow view of the credit cycle.  Accordingly, the original IASB approach also perpetuates pro-cyclicality as changes to originally anticipated loss estimates would be recognized retroactively for good book assets and immediately in full for bad book assets.

We believe the compromise proposal set forth in the recent Supplementary Document – Accounting for Financial Instruments and Revisions to the Accounting for Derivative Instruments and Hedging Activities – Impairment, issued on January 31, 2011, was an attempt by the two Boards to combine the independent FASB and IASB proposals in a manner that would address both the reserve adequacy and timing of loss recognition issues with the current incurred loss model.  For the reasons noted above, we believe the compromise proposal does not adequately address these issues. 

We recognize that no impairment model can completely address the cyclical nature of credit risk inherent in financial instruments, and there will always be some level of volatility as we move through the ups and downs of the credit cycle.  However, we believe it is possible to modify and enhance the existing incurred loss model to consider the cyclical behavior of financial instruments and lack of transparency around inherent losses in certain periods of the credit cycle in the determination of credit impairment.  We believe this proposal would result in a better estimate of credit losses related to loss events inherent in the portfolio at the balance sheet date, effectively address the criticisms regarding the adequacy and timeliness of credit loss recognition and provide financial statement users with a more representative view of an entity's financial condition. 

Proposal

Our proposal expands on existing incurred loss practices found within current accounting principles to more effectively estimate inherent credit losses by eliminating the probability threshold, incorporating expected events into the loss forecast and extending the loss emergence period.  Under our proposal, inherent credit losses are estimated using a two-step approach.  Although described in two steps, these components are interrelated and are each necessary to estimate losses inherent in the portfolio.  We have described the components separately and would disclose them separately to provide clarity and transparency of management estimates:

1. A base component (the "Base Component") that represents the estimate of expected inherent losses in the portfolio that are reasonably predictable;
2. A credit risk adjustment component (the "CRA") that represents additional credit losses that are not yet reflected in current credit risk metrics used to estimate the Base Component but are estimated using macro-level factors and are expected to emerge with more transparency as the credit cycle unfolds. 

Base Component

The Base Component is intended to capture expected inherent losses that are reasonably predictable based upon an assessment of historical and current credit information and expected events and conditions.  The Base Component methodology replaces the current incurred loss model with an expected loss concept that incorporates expected events into the loss forecast and extends the loss emergence period to a period over which losses are reasonably predictable.  Uncertainty in the forecasting process, changes in loss emergence periods, and other factors are not explicitly or systematically considered in the Base Component, and as such, the Base Component is by itself an incomplete estimate of inherent credit losses.   The terms "Expected Inherent Losses" and "Reasonably Predictable" are defined as follows: 

Expected Inherent Losses are defined as management's best estimate of losses inherent in the loan portfolio based on a company's credit evaluation process taking into account all relevant current and historical information as well as expected events and conditions.   This is a change from the existing incurred loss definition as the "probable" threshold has been eliminated and expectations of future events can be fully considered to estimate the severity of losses associated with a loss event. Expected Inherent Losses are pro-cyclical, by nature, and reflect the portion of the total allowance for credit losses that can be reasonably predicted in the current environment based on the available evidence and trends.   The elimination of the "probable" trigger in the definition of Expected Inherent Losses is necessary to allow the Base Component to capture a greater portion of the actual losses inherent in a portfolio at any given point in time and align the credit loss recognition methodology with the cyclical nature of the underlying financial instruments.  We believe that a company's best estimate of losses, whether probable or not, is the correct starting point for establishing credit impairment as this information is more reflective of loss estimates used in pricing credit.  We also believe that this articulation of an expected loss concept is preferable to establishing a "more-likely-than-not" threshold of incurred losses because it is better aligned with risk management, and credit loss estimation practices, which generally do not incorporate a probability weighted analysis or a pre-defined level of precision.

Reasonably Predictable is defined as the period of time that losses can be estimated with reasonable confidence.  In estimating the losses that are Reasonably Predictable, several factors should be considered including, the characteristics of the financial instrument or pool of financial instruments, the historical performance of the financial instrument or pool of financial instruments, the current and expected market conditions, and consideration of a company's own credit forecasting processes.  The period of time determined to be reasonably predictable will vary by asset class and may change throughout credit cycles, and will not necessarily be consistent across companies.   

This methodology is not intended to result in the immediate recognition of a full life-of-instrument loss estimate in most cases because it would be unlikely that the Reasonably Predictable threshold would be satisfied unless there is a specific indication of impairment.  For example, if for individual instruments or a specifically identified pools of instruments with specific indications of impairment (e.g., collection of all contractual principal or interest is not expected), all Expected Inherent Losses for that instrument or that portion of the pool of instruments would be considered Reasonably Predictable and the remaining life of loan loss would be immediately recorded similar to current accounting. 

Credit Risk Adjustment Component (the "CRA")

The CRA is a separate component of the allowance for credit losses that is established to address the inherent limitations in a company's credit forecasting process and the cyclical nature of macroeconomic conditions.  Past credit cycles have seen extended periods of benign activity followed by rapid parallel upward shifts in credit loss estimates.  The specific economic and credit conditions that lead to the negative credit shocks often accumulate over a number of years, but often are not readily apparent in the credit metrics commonly used to estimate the Base Component.  For example, a) underwriting standards and loan terms may be eased during benign credit environments; b) favorable economic conditions may mask credit weaknesses of the borrower, c) uncertainty regarding the sustainability of the current economic conditions is often high and d) loss emergence periods tend to extend during benign economic periods.  Each of these factors suggests that credit losses build even during benign credit environments and these losses later become transparent as the credit cycle deteriorates.  Consideration of these factors, therefore, would likely cause the CRA to be highest during these benign credit environments, thus, ensuring that inherent credit losses are appropriately recognized even during such periods.  Conversely, the CRA may not be as high during times of increasing loss rates as the portfolio's loss content is reflected or more apparent in current credit quality indicators and therefore would be more fully captured by the Base Component.

Many critics have concluded that the existing model for credit impairment may not be capable of capturing the portion of losses that have been incurred, but for which, there is no currently observable evidence of credit loss.  We believe the CRA concept more effectively addresses this weakness and enhances the existing incurred loss model from both a balance sheet perspective (by capturing estimates of expected inherent losses that are not readily apparent or observable), and an income statement perspective (by appropriately accelerating the recognition of credit losses into the periods in which they are inherent but not readily observable, and not concentrating loss recognition into the later stages of a credit cycle when losses are observable and can be specifically identified). 

The CRA is intended to capture those losses that are inherent in the portfolio, but due to the nature of the credit cycle, will not become transparent until credit losses begin to materialize.  During the course of a normal credit cycle, the counter-cyclical nature of the CRA will offset some, but not all of the volatility created by uncertainty in the timing and amount of credit losses.  For example, no impairment methodology could have fully addressed the dramatic parallel shift in credit loss curves experienced from 2007 to 2009.  In periods of extreme credit stress, a company may need to increase the Base Component as losses become observable, but may decide a CRA is also necessary if sufficient uncertainty remains regarding the absolute levels of expected credit losses.  In this manner, the CRA addresses both criticisms, reserve adequacy and timing of credit loss recognition, leveled at the existing accounting guidance. 

  • The methodology for establishing the CRA should consider factors including, but not limited to:
  • Current credit metrics and forecast;
  • Historical credit metrics (including stressed loss rates);
  • Management's evaluation of the credit cycle;
  • Other important credit indicators such as borrower behavior and collateral values;
  • Current underwriting standards, loan covenant terms, and other loan characteristics;
  • Recent trends in economic conditions;
  • Portfolio performance, concentrations, and deterioration relative to historical ranges;
  • Changes in loss emergence patterns over a credit cycle; and
  • The level and estimate of imprecision and uncertainty in the factors above.

Many of the factors considered in the CRA would by nature be heavily dependent on management's judgment.  These factors should be fully documented and supported by either market data, where possible, or internal data and analysis, and appropriately disclosed in the financial statements. 

Conclusion

We believe that a credit impairment methodology that estimates credit losses inherent in the portfolio, comprising both a Base Component and a CRA, will address many of the concerns with the existing impairment model, and is superior to all other models proposed to date.  We believe that the application of this methodology will:

  • Generally increase the size of existing credit reserves to more accurately reflect inherent losses in the portfolio, including the risk of deteriorating economic conditions on those inherent loses;
  • Reduce pro-cyclical volatility in the income statement created under the existing model as inherent losses will be appropriately recognized earlier in the credit cycle;
  • Better align recognition of credit losses to those periods where credit losses are inherent in the portfolio, but are latent due to favorable economic conditions; and
  • Provide more useful qualitative and quantitative disclosures to financial statement users through transparent disclosure of the different components of the allowance for credit loss calculation and enhanced information about the key methodologies, assumptions and judgments used in determining those amounts.

We believe that this methodology has a solid foundation in existing accounting principles and credit risk management practices in our industry, and is similar to concepts and practices in analogous circumstances to estimate inherent losses in other industries.


​Questions? Contact Mike Gullette (202) 663-4986 for more information.