• My institution is required to go live with the CECL accounting standard in 2023. However, we are currently parallel running. Can/should we use your credit modeling/decisioning tools to help us understand the credit risk (for both incurred loss and CECL estimates) of a similarly sized institution we would like to acquire?
• As the COVID-19 pandemic continues, questions continue to arise regarding the handling of reasonable and supportable forecasts. For example, can any forecast be “reasonable and supportable” amid the COVID-19 pandemic?
• For accounting purposes, should I still be using any models that were developed pre-pandemic? Are they still viable in my CECL process?
• How would I calculate an “effective” or “equivalent” value of some of the macroeconomic variables (MEVs) such as the unemployment rate?
• How do I know and explain that my CECL allowance for credit losses (ACL) levels are reasonable during COVID-19 pandemic?
• Do I need to give a different allowance if loans are accommodated?
• How should CECL filers deal with economic data like unemployment that is outside of the historical range and causing loss estimates to explode to unreasonable levels? 
• How do I account for government relief in the commercial real estate (CRE) credit loss model? 
• Does the economic recession and uncertainty brought on by COVID-19 cause an adjustment in the scenario weights?
  

My institution is required to go live with the CECL accounting standard in 2023. However, we are currently parallel running. Can/should we use your credit modeling/decisioning tools to help us understand the credit risk (for both incurred loss and CECL estimates) of a similarly sized institution we would like to acquire?

Yes, you can and yes, you should. Here are two big reasons why:

• This will ensure the pro forma financial plan reflects the CECL approach. Leveraging the forecast component of our models can be particularly insightful for thinking about the merger under multiple scenarios.
• This also pulls forward the operational capabilities needed to provide allowance inputs to financial reporting. CECL makes the first 10-Q as a combined company a much more complex challenge than historical incurred loss estimates.  

As we embark on this reality, which began not even one full quarter into the transition to CECL for most public filers, events unfolded that caused many to question the foundation of the decisions they made during the CECL implementation. These questions remain even for those who delayed CECL adoption as part of the CARES Act (until the end of 2020 or the cancellation of the federal emergency).

The intention of the FASB standard in defining a reasonable and supportable forecast period was to identify the length of the forecast period that a company could defend while creating a forward-looking allowance estimate. In practice, this has ranged from one year to, in some cases, the entire expected life of the asset. Using a one-year reasonable and supportable period as an example, management stated that they could support only the first year of a forecast—but then they would revert to historical values. COVID-19, however, has led many in the accounting industry to question whether any forecast period is supportable. As the first quarter ended, many wondered how high unemployment would go and what the impact of the CARES Act may be. As we move forward, these questions persist as well as concerns over additional waves of outbreaks and what our "new normal" will look like and mean to certain industries.

This volatile short-term market and uncertainty about what a recovery looks like has left accountants questioning what is reasonable and supportable. To answer this, the most helpful tool is reviewing the forecasts themselves. The following table shows the US unemployment expectation based on three economic scenarios produced by Moody’s Analytics. All three are Baseline scenarios as of 2/28/2020, 3/31/2020, and 4/30/2020.

What can be noted is that as the pandemic unfolds, the unemployment projection increases and the time to return to a long-term median elongates. However, it is also noteworthy that in all instances, the projection does return to a longer-term average.

As a result of this outlook, it actually brings into perspective what we have heard from the FASB and regulators as the first quarter was ending. While the short term may be difficult to assess, the reasonable and supportable period should also consider the recovery as well as potential stimulus packages from the government. To do this, you would need to ascertain that the reasonable and supportable period is the original one, and has not been drastically shortened due to these events. This would provide for allowance values to reflect the expected stabilization of the economy, despite volatile and dramatic effects expected in the near term.

Finally, this event has created an unprecedented economic environment in the United States; thus, we do not have a good benchmark to compare to. As a result, shortening the reasonable and supportable period and reverting to historical values (as dictated by the standard) may be less accurate than relying on economic forecasts. 

For accounting purposes, should I still be using any models that were developed pre-pandemic? Are they still viable in my CECL process? 

The events that have unfolded related to the COVID-19 pandemic are certainly new territory for the US banking industry—or at least new territory given our current data capabilities. Therefore, it is safe to say that any models used to measure credit risk did not account for any comparable past events. However, most models do include the Great Recession of 2008, so there is some consideration to drastic changes in the economy. 

Even so, many accountants are asking themselves whether it is appropriate to use such models in their CECL process both during and after the pandemic. Despite the uncertainty, most people in the industry agree that it is still reasonable to use them. What is beneficial is that these models generate a very specific output, with calculations that can be traced and understood. For accountants, this is an important starting point from which to provide these credit metrics. That said, it will be necessary for accountants to understand what is, and is not, included in those model calculations. They will require careful thought and, in many cases, management-level adjustments to factor in the effects of the current market situation. 

Moody’s Analytics has been helping customers with this analysis and giving advice on how to determine what specific adjustments may be necessary for your organization. If you are interested in such assistance, we encourage you to reach out to us using the links on this site. 

How would I calculate an “effective” or “equivalent” value of some of the macroeconomic variables (MEVs) such as the unemployment rate?

There might not be a simple or standard way to compute an “effective” MEV in the same way you would calculate the unemployment rate. Historically, there is a negative correlation between unemployment and income growth. We have never experienced a growth rate in disposable income as in the second quarter of 2020, even when the economy was growing rapidly. If we extrapolate based on the historical relationship, the current rate of income growth suggests a sub-3% unemployment rate. That does not seem reasonable or supportable. Moreover, the growth rate of income is going to turn sharply negative after the stimulus measures expire. 

Further complicating matters are the differences in payment forbearance and deferment across lending products. Because of the additional support the mortgage market is receiving, we might expect the “effective” unemployment rate to be much lower here than the unsecured lending market or the C&I loan market. Going down the path of setting different unemployment rates by product would likely raise a red flag and require significant justification. One way to address the issue is through model modification where income becomes a true driver. However, this effort may not address the issue in a timely manner. A transparent, well-documented, on-the-top forecast adjustment approach may offer a reasonable alternative. This could be informed by running alternative scenarios with varying unemployment assumptions.

How do I know and explain that my CECL allowance for credit losses (ACL) levels are reasonable during COVID-19 pandemic?

Given the high uncertainty of the pandemic severity in the United States and around the world—coupled with the impact on the economy in the short-to-medium term—it is difficult to determine the correct level of CECL ACL for your institution and portfolio. Nevertheless, you can still construct various benchmarks and reference points to put your ACL results in perspective.

• Using characteristics such as time to maturity of your current portfolio and historical loss experience such as net charge-off (NCO), you can calculate an approximation of the realized lifetime loss rate as of various time points during the financial crisis. The FDIC call report is a good source of publicly available information. You can use the NCO history of your banks, a peer group of banks with similar customer profile and business footprints, or the entire banking industry.
• Take the nine-quarter cumulative loss rates for various portfolios disclosed by the Federal Reserve for DFAST stress testing under severely adverse scenarios, and any future disclosures on COVID-19 stressed scenarios. 
• Use industry benchmarks based on the ACL quarterly disclosures of comparable banks.
  

Do I need to give a different allowance if loans are accommodated?

While temporary relief is often a successful strategy for improving loan performance, historical experience suggests that accommodated accounts have higher default rates compared to their pre-crisis level. Lenders should try to capture as many of the features of their accommodation program in their estimation methodology to minimize the role of qualitative adjustments. Nevertheless, it may be reasonable to make an adjustment for any account not making payments.

How should CECL filers deal with economic data such as unemployment that is outside of the historical range and causing loss estimates to explode to unreasonable levels?

Scenario users should consider the total impact of the economic outlook on their borrowers. For example, the impact of double-digit employment rates and the associated decline in wage and salary income will likely be more than offset by direct government assistance and expanded unemployment insurance benefits in the short term. Similarly, forbearance and deferment programs will push out defaults and foreclosures at a minimum. This could lead to a reduction in estimated lifetime losses if the COVID-19 shock turns out to be more of a cyclical liquidity shock than a structural shift in the economy.

Best practices to account for government income support in credit loss models that may have been trained on unemployment rates include:

• Ex post, qualitative overlays reducing the loss forecasts implied by a statistical credit loss model
• Interpolation of forecast credit losses during the second quarter of 2020
• Adjustment of the economic inputs going into credit models, including the calculation of an “effective” or “equivalent” unemployment rate that accounts for the fact that some consumers may receive government transfer payments that offset their loss of wage income
  

Each of these approaches has advantages and disadvantages with no approach clearly superior to the other two. Regardless of the selected approach, CECL filers must be prepared to substantiate that their selected approach is both reasonable and supportable with logic and data.

For more details about Moody’s Analytics scenarios and their use in CECL applications, subscribers are encouraged to consult the documentation on www.economy.com/databuffet/. In addition to narrative descriptions of each month’s published scenarios, Moody’s Analytics gives detailed information on its econometric modeling methodology and an inventory of the key forecast assumptions made each month. Validation and accuracy tracking reports are also available to support users in applying the scenarios for regulatory and financial reporting applications.

How do I account for government relief in the commercial real estate (CRE) credit loss model?

First, you should consider incorporating the latest CRE market outlooks such as those from Moody’s Analytics REIS into the model. While CRE market conditions have historically been strongly correlated with macroeconomic conditions, the massive government stimulus has mitigated the impact of the COVID-19–induced economic crisis on the CRE market to some extent. As a result, the credit risk of CRE portfolios would not be as heightened as it otherwise would be in a similar economic environment.

As part of the CARES Act, many corporations have received short-term loans through the Paycheck Protection Program (PPP), which they can use to repay mortgage interests. This can be incorporated into the CRE credit loss model through cash reserve, which is a common feature of CRE loans. Since PPP loans are expected to be forgiven under most circumstances, they can drastically reduce credit risk at least over their short duration.

Meanwhile, many lenders have implemented payment deferral as a temporary relief for short-term insolvency. Under this program, borrowers can postpone mortgage payments until a later date before loan maturity. As a result, debt service coverage ratio (DSCR) will decrease significantly during the payment deferral even if the CRE collateral is generating much less income. Since DSCR is a key risk factor for CRE mortgages, credit risk will also stay low over the holding period. On the other hand, the refinancing risk will likely increase because the suspended installments are expected to be repaid upon loan maturity. However, this increase may not be drastic on the later dates as the economy will hopefully be in recovery by then.

Finally, you may consider adjusting the refinance interest rates to align with new internal lending policies in the post-COVID-19 environment. Given the extensive government relief, it may be easier for lenders to work out troubled debt. Hence, the refinancing terms may become relatively more favorable such as lower refinancing interest rates, which would lead to lower maturity risk for CRE mortgages.

Does the economic recession and uncertainty brought on by COVID-19 cause an adjustment in the scenario weights?

Not necessarily, since Moody’s Analytics scenarios, updated monthly, are dynamic and shift toward being more optimistic or pessimistic as economic conditions change. So, there is no need to adjust the weights on the scenarios. That said, an institution may change the scenario weights, with reasonable justification, to incorporate a different internal management view.