Are Bankers’ Decisions Rooted in Hard Data—or Hope?

John DelPonti, Paul Noring and Joseph Sergienko

Our December CECL Reserving and Credit Benchmarking Study will show banks are managing in the face of wild data swings that render their models largely useless

With a bruising 2020 entering its final weeks, a sense of cautious optimism seems to be taking hold in the banking sector. Early indications in banks’ third-quarter filings and recent conversations with banking leaders indicate that many see the economic rebound in the second half of the year as a hopeful sign that the worst is behind them. 

We get a closer look in our third-quarter CECL Reserving and Credit Benchmarking Study. It compares allowance levels, charge-offs, non-performing loans (based on CARES Act classification/reporting relief), COVID deferrals and other key credit metrics across predominantly domestic US institutions.

The newfound optimism, albeit tinged with caution, represents a sharp contrast to the gloomy outlook that pervaded the sector through much of 2020. Whether that optimism proves justified may not become clear for some time. What we can say with certainty is that bankers continue to face a unique and vexing challenge: Radical volatility in this year’s economic data has severely compromised the predictive models they normally rely on to make decisions about their loan portfolios.

Early signs of a bounce back

Our previous study, reviewing second-quarter data and released in September, showed significant COVID deferrals, particularly for commercial real estate loans—where banks were issuing deferrals at double the rate of other portfolios, with one bank reporting 25 percent of its commercial property loans in deferral. With an average of nearly 7 percent of loans in deferral, the picture that emerged was of a banking sector still expecting difficult days ahead. 

The study we’re releasing this month will show whether that has changed and shed more light on the status of banks’ loan portfolios. From what we’re hearing in conversations with banking leaders and seeing in early fourth-quarter filings, COVID deferrals have dropped significantly. And yet, with no indication of widespread, large-scale release of loan-loss reserves, indications are that bankers remain cautious. 

The question then becomes how bankers are making decisions in light of unpredictable swings in the economic data. For instance the third quarter’s 33 percent jump in GDP was clearly an anomaly, tied to the cycle of COVID-driven lockdowns and reopenings. Such anomalies have become the norm in 2020, making the data increasingly difficult to read.

Relying on judgment as COVID plays havoc on data, models

Many models that bankers use to calculate reserves and other key decisions rely on macro measures like GDP and unemployment; this year unemployment jumped from 4.4 percent to 14.7 percent between March and April, then dropped to 6.9 percent in October. Banks’ models aren’t built to handle wild swings like that.

The unusual nature of this economic downturn has complicated matters in other ways. Modelling loan-loss provisions based on a ratio of nonperforming loans, for example, is complicated by COVID deferrals. If deferred loans aren’t classified as nonperforming, the model shows that the bank doesn’t need provisions any greater than at this time last year. For a bank that has a quarter of its commercial real estate loans in deferral, that could be deeply problematic. On the other hand, including COVID deferrals in the model would tell the bank to set loan-loss provisions at an egregiously high level. 

With their models rendered largely unreliable, bankers must use their judgment to make key business decisions. We’ve seen many instances where loan-loss reserves and other metrics were determined by overriding models with qualitative assessments based on bank leadership’s perception that their models simply weren’t reflecting what’s really happening in the market.

Impact on new CECL standards?

In some cases, banks used qualitative overrides to dial back levels of Current Expected Credit Losses (CECL) reserves that were calculated purely based on modelled output. The question in the coming months will be whether the judgment calls used to determine the exact amounts of these qualitative overrides are well grounded and properly documented. 

The same goes for most of the important decisions that bankers are making. Without dependable models to guide them, and in the midst of a pandemic unlike anything the leaders of current institutions have ever seen, there simply are no best practices to guide the way. 

In this circumstance, the institutions with the most robust governance and review processes are the ones most likely to persevere. Strong processes allow banks to add a layer of qualitative assessment to their models in deliberate, consistent ways. That should help them set loan-loss provisions, CECL reserves and other key ratios at the levels most likely to accurately reflect the underlying state of their portfolios.