The fallout from the COVID-19 pandemic is underscoring the drastic differences in approaches banks can take when reserving for loan losses under the current expected credit loss model, a new accounting method.

The method requires banks to set aside reserves for expected losses over the life of the loan as opposed to reserving for losses that are probable or already have occurred. CECL requires banks to make forecasts, and during recent earnings conference calls, bank executives described what was driving their decisions.

Some banks were more forthcoming and gave precise details on their assumptions. For instance, Citigroup Inc. assumed unemployment peaking at 15% and GDP falling 35% quarter over quarter, management said on the bank's second-quarter earnings call.

Dallas-based Santander Consumer USA Holdings Inc. assumed peak unemployment of 15% in the second quarter, falling to 10% by the end of 2020 and 8% by the fourth quarter of 2021, management said on the bank's second-quarter earnings call. "Our allowance does not assume any benefit from further stimulus packages the government is currently evaluating," said Fahmi Karam, CFO and head of pricing and analytics at the bank.

Others gave overviews that outlined the approaches they were taking. Short Hills, N.J.-based Investors Bancorp Inc. is mixing Moody's models, using weighted versions of Moody's different scenarios, management said on the bank's second-quarter earnings call. On the New York Community Bancorp Inc. second-quarter earnings call, management at the Westbury, N.Y.-based bank said that property price declines were the main driver of its CECL modeling, more so than the typical factors like unemployment and GDP. "It's qualitative, just because of the history of the bank that has no losses as we look at the portfolio," said CFO Thomas Cangemi on the call.

These discrepancies in assumptions have raised questions about the validity of CECL, and whether it makes sense to compare banks' provisions under the new accounting method.

"CECL is a joke because every bank is different," said Kansas City, Mo.-based UMB Financial Corp. CEO, President and Chairman J. Mariner Kemper on the bank's second-quarter earnings call. Kemper said the bank's commercial and industrial portfolio performance was actually stronger, despite the increased reserves required by CECL forecasts. "It's an obscene and absurd way to think about credit quality," Kemper said.

But the variances are not surprising, said Maria Mazilu, a vice president and senior accounting analyst with Moody's Investors Service in an interview. "We definitely expected from the beginning that CECL will make comparability more difficult," she said. CECL involves more judgment calls, as well as a longer period of time to account for the full life of the loan, Mazilu said.

The economy is also hard to predict, given the pandemic and the government stimulus, said Peter Nerby, a senior vice president and banking analyst at Moody's. "To some extent, the relationship between what their models might predict as charge-offs, and the level of unemployment, because of the effect of the stimulus is making the model results a little bit harder to interpret," he said. "So all of those things lead to reasonable people having different views of what the level of charge-offs ought to be."

There are even differences of opinions on provisions for banks that have yet to adopt CECL and are still using the incurred loss model. McKinney, Texas-based Independent Bank Group Inc. management members said they expect the "day one" CECL adjustment to be about $80 million, but no further reserving after that to adopt the model on the bank's second-quarter earnings call. "We think that the incurred loss model and the CECL model year-to-date are materially the same," said Chairman, President and CEO David Brooks on the call.

But Mazilu said she found that surprising and would not have expected banks using the incurred loss methodology to have reserved as much as banks using CECL. "Over time, the incurred loss model catches up, I wouldn't have thought that by the second quarter it had already caught up with CECL," she said. She said perhaps the two models would be the same by the end of 2020.

Meanwhile, some banks are eschewing economic forecasts completely. Southfield, Mich.-based Credit Acceptance Corp. said that forecasts did not help with the accuracy of the models. "We went actually to even local unemployment data to see if we could figure out how to use that information to forecast more accurately, and it just didn't work," said CEO Brett Roberts on the bank's earnings call. "It's an estimate and I think we've stressed that there's a lot of uncertainty around that."