AuditOne Advisory: On The Horizon – The Current Expected Credit Losses (CECL) Model

AuditOne Advisory

From Bud Genovese, Chairman

Our Co-CEO Jeremy Taylor’s advisory below covers CECL and how to prepare now as it appears it will become a reality sooner than later. Please feel free to forward this informative presentation to any appropriate people in your bank.  And remember, contact us to best serve your risk management needs, as we are the industry leading expert for internal audit and credit review.  Thank you.  –Bud

“Waiting for CECL”

The Allowance for Loan and Lease Losses (ALLL) is a hugely important figure for financial institutions (FIs).  The current, Incurred Loss approach is laid out in the last significant guidance statement on the subject, FIL-105-2006, released back in 2006.  Now that everyone’s finally got themselves more or less comfortable with it, it’s all about to change.  FASB will soon be releasing guidance on the Current Expected Credit Loss (CECL) methodology.  It was expected last year, then Q2/15, now slated for later this year.

The main idea of CECL is that loss reserving should reflect losses expected over the life of a loan and not just losses already incurred (even if not yet recognized), or in standard current practice, losses expected over the next 12 months.  While the 2006 Qualitative & Environmental adjustments (or “Q-factors”) bridge between the historical loss period and today’s conditions and circumstances, you can think of CECL as requiring you to do the same thing out through each loan’s remaining life.  In practical terms, “expected” implies some measure of long-term average losses for that type of loan (i.e., based on its type, industry, location, vintage, etc.), before applying Q-factors. It’s less clear how scenario effects might also be incorporated.  (The literature on this subject generally emphasizes holding reserves against expected losses and capital against the unexpected (or adverse scenario, right-tail) losses.  It remains to be seen how this potential double-counting of reserves and capital will play out.)

Note, too, that FASB is proposing to apply CECL not just to loans but also to securities, thereby replacing the current OTTI (Other Than Temporary Impairment, the credit (not interest rate) effects) approach.  This means that, like loans, they’ll have a reserve against expected credit losses rather than being directly written down to reflect OTTI. Lengthening the relevant time horizon over which potential future losses are to be assessed (from 12 months to life of loan) immediately highlights one critical and much-discussed aspect of CECL: upward pressure on the level of required reserves.  In the absence of details on computation and implementation, it’s too early to determine this with any precision.  Tom Curry, the OCC head, is on the record as suggesting typical increases in the range of 30 – 50%.  That’s material, especially considering that roll-out will overlap with (or follow shortly after) phase-in of increased Basel III capital requirements; the main purpose of the ALLL, like capital, is to absorb losses.  CECL means that booking a loan will require booking the statistically expected loss right away.  But keep in mind the intended counter-cyclicality of CECL: booking higher reserves in today’s more benign macro environment is intended to better buffer institutions against future downturns and necessitate less aggressive provisioning under such less benign conditions.

One more thing before we provide a few practical, pre-CECL suggestions.  We’re all used to thinking in terms of general versus specific reserves (or FAS 5 versus FAS 114, or ASC 450-20 versus ASC 310-10-35).  But CECL essentially applies the impairment (FAS 114) mindset to the non-impaired portfolio, requiring all loans (whether impaired or not) to be assessed based on expected repayment prospects.  However, for practical purposes we’ll still want to think about using homogeneous pools for the good stuff; for most institutions (whatever your size) doing discounted cash flow (DCF) analysis at the loan level is too daunting in terms of the data and computing requirements.  DCF (or net realizable value calculation, for collateral-dependent loans) will likely continue to be worthwhile only for those loans with uncertain repayment/recovery prospects.  Related to this, the projected increase in reserves under CECL are expected to come in the non-impaired portfolio much more than in the impaired.

Again, CECL’s details remain murky, but let’s finish with a few things to keep in mind in preparation:

Yes, higher required reserves, though with many smaller (and non-public) FIs currently holding excess (or “unallocated”) reserves, above what their model calculations now suggest, that effect (in terms of a need for a significant, one-time increase in reserves) may be less pronounced.  Nor is it clear whether that might flow through income (via loss provisioning), versus an adjustment to capital (i.e., like the difference between treatment of mark-to-market losses for AFS versus HTM securities holdings). More likely the latter, though still to be confirmed.

Given the more onerous demands (per #3 and 4, below), it is expected that CECL will have a multi-year phase-in.  That, too, will mitigate the “headline effects” of CECL, though the expected 3 – 4 years must cover the demands associated with not just #3 and #4 but also a parallel-run period.

Data requirements will rise significantly, both in time series (i.e., number of years’ history) and cross-sectionally (i.e., the number of variables).  FIs will be expected to collect loan-level data on a wider range of loss and other loan performance variables than is currently expected.  Internal data on portfolio performance can be supplemented with outside sources such as FDIC and RMA (Risk Management Association).  You’ll also need to be able to tie it in with prevailing external data (e.g., how the local/regional economy was doing), though that’s typically easier to procure.  Ditto for forecasts of such driving, external variables.  A couple of related points:

  1. Loss migration becomes much more important: not just the loss rate by loan type and, if available, risk grade.  It’s also the migration patterns – i.e., movement up and down between grades – by loan type, origination vintage and other identifiers.  You’ll need to be able to assess (with, of course, due supporting data) the probability of a grade 3 loan, for example, being downgraded to 4, and then to 5, and eventually all the way to default.
  2. Community banks and credit unions will be expected (even if not for day one) to follow down the path already taken by larger institutions towards splitting loss rates into probability of default (PD) versus severity (or loss given default (LGD)) rates; the former reflect risk grade/borrower characteristics while the latter are driven by loan structure (in particular, collateral value).

The message here is clear: better start collecting the data now.

While an in-house spreadsheet model has generally sufficed for most smaller institutions under the current methodology, the volume of data and more complex analytics required for CECL may well warrant consideration of dedicated software.  There are various vendors out there with existing models that will no doubt be customized for CECL when details become available.

AuditOne will continue to monitor this area closely.  We have considerable expertise and experience in assisting institutions with their ALLL methodology in both an auditing and consulting capacity.  We will be sure to carry all that forward into the brave new world of CECL, whenever he decides to step forward to introduce himself.

AuditOne LLC sole focus is financial institutions. AuditOne performs many dozens of ALLL audits every year, along with credit review audits delivered by our expert staff auditors that have hands-on, in-depth credit experience.    Please note, too, that many of these audits are done on a completely offsite basis, using our secure client portal for secure sharing of documents.  For more information, please call one of our Co-CEOs: Jeremy Taylor Contact Us), or Kevin Watson Contact Us. Both may also be reached on our Team & Contact page.