Last week’s comments by the Financial Accounting Standards Board (FASB) about how they will allow the different levels of complexity in credit loss calculations for lenders of different sizes would seem to be a victory for smaller credit unions and community banks.
Beware the unintended consequences.
Experience from the evolution of the CCAR (Comprehensive Capital Analysis and Review) & DFAST (Dodd-Frank Act Stress Tests) regulations suggest that this may produce a worst-of-all-worlds approach.
FASB is making final edits to the Current Expected Credit Loss (CECL) guidelines. They have promised to release them in June 2016.
The first version of the guidelines rightly recognized that a level of loss can be anticipated from the moment the loan is booked. Knowledge of the loss timing for loans can allow an institution to estimate future losses one or several years into the future. A simple loss model of volume booked times the loss timing function would have produced a forecast accurate enough to sound alarms in 2006 about problems in the mortgage market.
Smaller institutions were not as dynamic as large lenders through the most recent crisis, but that does not mean that their losses are flat through time. Quite the opposite. As our recent analysis showed, macroeconomic adverse selection can explain half of the losses experienced even in the most stable portfolios. Changing consumer demand affects everyone. Even small portfolios can be vulnerable.
However, in recognition of the limited resources and potential cost of the new CECL model, FASB indicated that it will revise the proposal:
"The revised language makes the proposal more flexible, stating that there is no one methodology that has to be used. The board’s intent is that each institution needs to apply the method that is appropriate for its portfolio based on the knowledge of their business and processes.
By allowing community banks to evaluate and adjust their loan-loss amounts using qualitative factors, historical losses, and current systems, such as spreadsheets and narratives, FASB has made important changes to its proposed accounting standard."
This sounds pretty good. So why are we worried?
The new guidelines are all about life-of-loan reserving. Even if small institutions are allowed to use “spreadsheets”, all institutions will need to reserve for the same horizon. That means either rolling back life-of-loan (unlikely) or new models. All institutions will need new models, whether they are simple or complex, statistical or spreadsheet. Then the fun begins.
All models must be validated.
Who will decide if the models are good enough for the size of the institution? With no fixed rules on which institutions must build which kind of models, the validators, auditors, and examiners will be left to adjudicate institution-by-institution. We’ve seen this before.
Regulations fall broadly into two categories, rules-based and principles-based. For example, Basel II is largely rules-based, where specific formulas are implemented to compute regulatory capital. In contrast, CCAR and DFAST are principles-based, where the goal is stated along with some suggestions, but each institution must decide what models should be created. What we’ve seen for the last several years for CCAR and DFAST lenders is mass confusion.
Each review cycle brings a new set of opinions. Last year’s acceptable models may not be acceptable this year, depending upon what the “industry standard” has become in the mean time. Citi was the most visible example of an institution that failed CCAR after previously passing. They were surprised when the goal posts were moved and the previous year’s models were no longer acceptable.
For DFAST, the disarray is worse. The OCC and FDIC say that the models should be simpler and not create an undue burden for the lenders, yet every examiner brings a different background and experience. Even basic statistical principles are in dispute. For example, the American Statistical Society has recently stated that p-values are inappropriate as the sole determinant of model effectiveness, and yet we’ve seen individual OCC examiners state that p-values are all that matter. Other examples abound of examiner personal-preference overturning person-years of effort.
FASB is correct that no single credit risk model is universally accepted, but the more vague the guidelines, the more disarray will come with attempts to implement them. Credit unions and community banks may initially feel relief over being able to use methods similar to previous efforts, but these are still new models for a new regulation. With a principles-based approach to modeling, the validators, auditors, and examiners will be left to decide. And frankly, most of them will be learning right along with the lenders over many years of trial-and-error.
Rather than pushing for flexibility, we believe small lenders should push with all their might for clarity. In the long run, a clear regulation is much cheaper to implement, support, and defend than a vague one.
“Flexibility” may be the most expensive path of all.