The largest banks' release of billions of dollars from loan-loss reserves in the first quarter has triggered insinuations that such practices are veiled efforts to boost earnings, rather than reflections of material improvements in credit quality.

This is a perennial question that comes around every time credit conditions start improving. However, generally accepted accounting principles dictate the rules that result in a roller-coaster ride for bank loan-loss reserving and provisioning exercises.

In the years leading up to the financial crisis, banks were criticized for not having enough reserves put aside for bad loans, and in the immediate aftermath of the crisis banks rapidly built up large reserves to address accelerating loan losses. The buildup may have contributed to the credit crunch, since all those funds squirreled away in reserve could have been lent out instead. However, these outcomes are driven by procyclical reserve accounting standards that introduce unnecessary market distortions and could be significantly reduced by allowing banks to establish reserves over a longer time frame.

Current GAAP accounting practices for loan-loss reserving are guided by Financial Accounting Statements 5 and 114. Under this framework bank reserves must reflect only losses that are "inherent" to their portfolios and are "probable" and can be reasonably estimated from available performance data. Banks typically estimate losses on homogeneous pools of loans using standard statistical loss models over a specified period of time (referred to as the loss confirmation period) and augmented with empirically supportable qualitative factors on the economy, changes in underwriting standards and general market conditions, among others. Events such as the SunTrust case in 1998, where the Securities and Exchange Commission forced the bank to restate its financials due to purported use of reserves to manage earnings, and the Sarbanes-Oxley Act reinforced the use of the incurred-loss approach to mitigate the potential misuse of the reserve process. At the same time bank regulators such as the Office of the Comptroller of the Currency have tended to lean toward approaches that would require banks to maintain certain coverage levels, for example, equal to one year's worth of losses. Such conflicting views between accounting practices and safety and soundness issues introduce additional challenges in reserve setting.

Estimating loan losses consistent with current accounting standards is at times a bit like driving by the rearview mirror. During particularly benign economic times, loan losses tend to be at their lowest point in the cycle, and thus estimates of loss for the reserve exercise will likewise be very low. Conversely, when market conditions deteriorate, losses accelerate, making it difficult for the statistical models to keep up with the sudden rise in loan losses. At that point, banks wind up, usually prodded by their regulator, raising reserves during these high loss periods. So in good years, not enough reserves are set aside for future losses and once the bad times hit, banks race to build large reserves. And once market conditions start returning to normal and losses abate, banks are forced to release reserves as dictated under the incurred-loss reserve accounting provisions.

An approach to reserving that would significantly reduce swings in bank loan-loss reserves that accompany business cycles would be to implement dynamic provisioning for losses featuring a "through-the-cycle" estimation of reserves. Dynamic provisioning takes a longer view of the loss cycle and permits banks to build reserves even during low loss periods in anticipation of higher losses in later years. Spanish banks have used a form of dynamic provisioning since 2000 that incorporates mean reversion – the idea that high and low readings of any metric are temporary and the figure will tend to move to its average over time – to setting reserves. So when a bank's actual reserve levels fall below historical levels, the bank is required to increase provisions for loan losses. Conversely, when actual reserves exceed their long-term average, the bank reduces its loss provision. And while Spanish banks encountered severe stress during the financial crisis, research has found that entering the crisis, Spanish banks held much higher reserves than their U.S. counterparts.

The incurred-loss model entrenched in U.S. reserving practices today unnecessarily amplifies the extremes of economic cycles. During good times when banks have the financial wherewithal to build reserves, accounting standards constrain such actions. And when banks are at their most vulnerable at the peak of a credit crisis, loss-provisioning activity accelerates. Such inherent procyclicality may increase systemic risk, contribute to credit tightening at the wrong time and promote greater risk-taking and asset formation than would otherwise be prudent. Moving to a dynamic providing model would greatly reduce such outcomes.

Clifford V. Rossi is the Executive-in-Residence and Tyser Teaching Fellow at the Robert H. Smith School of Business at the University of Maryland.