There is intense curiosity about the effects of CECL on banks’ financial statements via credit reserves–the ALLL or Allowance for Loan and Lease Losses; so, we surveyed the earnings releases, investors’ presentations and 10Qs and Ks of CCAR banks—plus one more. We present the aggregated results, below, as absolute and percentage changes in the ALLL and the change in the ratio of ALLL to total loans. However, there are two YUGE caveats that we also are described below.

CECL-effect-on-ALLL-percentage-change

CECL-affect-on-ALLL-absolute-change

CECL-affect-on-ALLL-as-percent-of-loans

The caveats:

1. As part of their CECL processes all banks should have model overlays—a.k.a. qualitative adjustment frameworks—to revise strictly-quantitative (modeled) output by applying common sense to model (or data) weaknesses and shortcomings. It’s not clear whether those adjustments were applied to the initial published estimates.
2. Our research was performed immediately prior to the Corona Virus scare, and this has a few important implications.
   • It’s possible—but unlikely—that some of the banks have publicized revised estimates in the meantime—the last week or so.
   •  Most banks, especially the large ones, base short-term loss forecasts on economic and geographic conditions. This is the so-called “reasonable and supportable” forecast period that varies by firm but ranges from one-to-four years.
   • Many banks—probably all—are calculating their initial, official CECL estimates during this month—March, 2020—based on February loan balances and characteristics as well as recently-revised forecast scenarios.¹

Given these caveats, it’s highly likely that the “anticipated” or “business-as-usual” scenarios, today, look much more like adverse or severely-adverse CCAR stress scenarios than they did several months, ago, when banks calculated and published their initial estimates.² We’ll have more to say about Corona Virus and CECL in a separate post.

For now, note that CECL is like “base-case” CCAR with two differences:

• The Fed requires new loan production for CCAR, but CECL is based on Current loans–although future line balances can grow if the the bank can’t unilaterally cancel the line.
• The forecast horizons are different. It’s nine quarters for CCAR, but depends upon the length of the longest loan maturity for CECL. For almost every bank, that going to be more than nine quarters.

So, if a bank’s longest loan maturity were nine quarters from today and no new loans were forecasted in CCAR, then the two should be the same. Unfortunately, neither the banks nor the Fed publish base-case or business-as-usual CCAR results.

Despite its small size, we included Sallie Mae (SLM) with the CCAR banks as its portfolio provides an excellent example of the type of loans that will show much higher reserves. SLM’s estimate is 342% larger than its ALLL under the current “Incurred Loss” method. That large change, relative to other banks’ changes, should make sense.

Without going into all the details of loss forecast modeling, the most precise method estimates three components, or multipliers, each quarter, t, for each customer and/or obligation. During the reasonable-and-supporting forecast horizon, each of these terms may be a represented as a random function of economic and geographical variables. After that, they’re likely to be set as conditional, historical averages.

1. Probability of default (PD_t)×
2. Loss given Default ratio (LGD_t) or the percentage that the bank doesn’t anticipate collecting
3. Exposure at Default (EAD_t) or the forecasted balance

The conditional expected loss each quarter (EL_t) is the product:

EL_t = PD_t × LGD_t × EAD_t.

The fourth factor is the loan term or maturity, T. Roughly–very roughly–the credit reserve for each loan is the sum of these EL_ts. So, again, very roughly, if expected losses were constant over each quarter of the loan’s life, for a term loan that doesn’t amortize, the credit reserve would be T × EL_t.³

Regardless, our simple example illustrates the types of loans that will have the greatest loss reserves:

• Long terms, T
• High probabilities of default, PD_t
• High balances and losses given default, which tend to be related to unsecured loans, where there is no collateral or security

Student loans do amortize–balances decline through time–but with long maturities, the decline is slow. With high default rates, long maturities and no collateral–and, therefore, high LGDs–estimated losses over the life of the loan are high. That’s opposed to, say, mortgages, which may have longer lives but have much lower default rates and collateral values that tend to be either stable or increase through time.

The table below shows the largest portfolios for each bank as well as any statements–made by the bank–regarding the most affected portfolios.

ECT = Earnings Call Transcript
ECP = Earnings Call Presentation
ECS = Earnings Call Supplement
IP = Investor Presentation

CAPITALIZED Firm Name indicates Balance and ALLL are from 2019 Q3 10-Q if no 10-K was available through SEC.gov as of 2/20/20
¹Goldman Sachs’ Loans only accounted for $109B of their $993B in assets
²Assets calculated as the sum of BB&T and SunTrust’s Assets. ALLL calculated as the sum of ALLL for BB&T and SunTrust. ALLL and Asset values as of Q3 2019.

If you are interested in the underlying data, please contact us.