If we do not get a miracle turnaround with respect to the coronavirus and its impact on markets, the first quarter earnings for most publicly traded community banks (SEC filers) is going to be a disaster because of CECL. In fact, you could not pick a worse time for a group of banks to have to comply with the new accounting standard for estimating loan losses.
The big problem that banks face with CECL in this environment is that the loan loss reserve needs to include the impact of “reasonable and supportable forecasts.” While I don’t pretend to have a crystal ball, I think it’s reasonable to say that the forecast doesn’t look too rosy right now. The likelihood of entering a recession has increased dramatically. This means that the loan loss reserve under CECL will have to be significantly elevated, which was not the case when banks were estimating their practice calculations as recently as a three weeks ago.
This leads to the second problem. Not only will there be massive pressure on the loan loss reserve to increase because of current and forecasted market and economic conditions, but most community banks have not practiced calculations under these conditions. They got ready for a good economic scenario. They weren’t expecting a recession, so they are not ready for one.
As a result, they will have to make Sophie’s choice: (a) quickly re-jigger their loss rate assumptions, which is much easier said than done, or (b) jack up their qualitative factors. Most of the banks will resort to the latter, but that is throwing darts at a board. They will not have confidence in their estimate, nor will auditors, regulators or investors. The one thing these banks can hang their hats on is that they are not alone. Most community banks are in the same boat.
It didn’t have to be this way though. I made the case two years ago for how important it was for banks to integrate CECL with stress testing. In fact, not only should CECL and stress testing be using the same data and the same models, the stress testing results should be used directly in the CECL estimate to help capture the impact of reasonable and supportable forecasts. The reality is that banks don’t experience material levels of loan losses in good times. If a portion of the CECL calculation was heavily influenced by the stress test, then you are in a much better position to estimate loan losses.
While it’s true that the probability assigned to adverse forecasts should and will increase because of what’s going on right now with the economy and markets, at least banks that took this approach are prepared to quickly make an adjustment and know what to expect. The increases in their reserves should be far more manageable, as long as their stress tests are driven by the loan-level risk characteristics of their loans and their underwriting is strong.
Hence, the net impact is both more manageable, but also more accurate. Most importantly, everyone can have more confidence in the calculation, including investors who are understandably nervous right now.
If you are a community bank that is not required to adopt CECL until 2023, do not make the same mistakes that many of the SEC-filer banks are making. The first mistake is assuming that software solves all your problems. It’s just a big giant calculator. You are still on your own. The second mistake is not choosing your methodology wisely. PD/LGD is the only method that is most natural for both CECL and stress testing. SEC filers predominantly utilized WARM or DCF, both of which are highly dependent on historical losses and make little to no connection to loan-level risk characteristics. They should be avoided. Finally, marry your CECL and your stress testing. Simply having both modules in the same software package is not enough. These two processes need to be exactly the same. Same data. Same method (PD/LGD). In good times, the only difference should be the economic scenario they assume. Stress tests always look at a downturn, while CECL calls for an expected forecast. But now as SEC filers are going to learn the hard way, even that difference is going away.
Note: This issue is gaining traction. See this KPMG advisory.