From incurred loss to expected loss
The previous 'incurred loss' model only let an institution build its allowance once a loss event was probable and could be reasonably estimated. In a downturn that produced a procyclical lag: reserves were lowest just before losses spiked. The FASB designed CECL to front-load that recognition.
Under CECL, the allowance for credit losses (ACL) reflects the credit losses an institution expects over the full remaining life of its assets — from day one. That means a brand-new, fully performing loan still carries an allowance, because some lifetime loss is expected across the pool.
What CECL requires you to estimate
A CECL estimate blends three lenses: historical loss experience, current conditions, and reasonable and supportable forecasts of future economic conditions. Beyond the forecast horizon an institution can revert to long-run historical averages. The estimate must cover the contractual life of the asset, adjusted for prepayments.
CECL does not mandate a single method. Smaller institutions often use approaches such as the weighted-average remaining maturity (WARM) method, vintage analysis or loss-rate methods, while larger ones may use discounted cash flow or more sophisticated probability-of-default / loss-given-default models. The standard is method-agnostic but documentation-hungry: whatever you use, you must justify it.
Why credit unions face particular considerations
Credit unions adopted CECL for fiscal years beginning after 15 December 2022, later than the largest public banks. To soften the impact on regulatory net worth, the NCUA adopted a phase-in that spreads the day-one CECL adjustment over a transition period, so the immediate hit to a credit union's net-worth ratio is smoothed.
Many credit unions also have less historical loss data and smaller modelling teams than large banks, which makes method selection, data quality and documentation especially important. Examiners expect the chosen method to be appropriate to the institution's size and the complexity of its portfolio.
Data, documentation and governance
CECL turns the allowance into a model. That brings it squarely within model risk management expectations: the methodology, assumptions, qualitative adjustments and forecasts all need documentation, governance and periodic validation. Auditors and examiners will probe how forecasts were chosen and how qualitative overlays were supported.
The biggest practical hurdle is usually data: loan-level history, segmentation, prepayment behaviour and the linkage to macroeconomic forecasts. Institutions that manage their portfolio data, model assumptions and the resulting allowance on connected, well-documented infrastructure spend far less time defending the number each quarter.
Living with CECL each quarter
CECL is not a one-time conversion; it is a recurring, judgement-heavy estimate. Each period the institution refreshes its forecasts, re-runs its method, reconsiders qualitative overlays, and explains the movement in the allowance. Volatility in the economic outlook flows directly into the provision and therefore into earnings.
Because the allowance now reflects forward-looking judgement, governance around it matters more than ever. A clear, repeatable process — with versioned assumptions, an audit trail and review sign-offs — is what keeps the estimate defensible to the board, the auditor and the examiner alike.