Loan Loss Provisions Are a Forecast Recorded as an Expense
Banks must estimate losses on loans that are still performing. The accounting changed in 2020 to require forecasting them earlier, which made bank earnings considerably more volatile.
What a Provision Is
A bank holding a loan portfolio knows some of it will not be repaid, without knowing which loans.
A loan loss provision is an expense recognising expected losses, which builds an allowance reducing the carrying value of the loan book. When a specific loan is eventually written off, the allowance absorbs it rather than the write off hitting earnings at that moment.
Provisioning moves loss recognition forward in time. The economic loss happens when the borrower stops paying. The accounting loss happens when the bank decides it is coming.
The Old Approach and Its Flaw
The previous framework was an incurred loss model: a provision was recognised only when there was objective evidence a loss had occurred.
The consequence was that provisions stayed low while conditions were good and rose sharply after deterioration was evident. Banks reported strong profits right up to a downturn, then took large charges once losses had already materialised.
This was procyclical. It flattered earnings and capital during booms, encouraging lending, and forced capital consuming charges during downturns, forcing retrenchment.
The Expected Loss Models
The frameworks introduced after the crisis require recognising expected losses earlier, incorporating forecasts of future economic conditions.
| Framework | Approach |
|---|---|
| CECL, United States | Lifetime expected losses recognised at origination |
| IFRS 9, international | 12 month losses, moving to lifetime on significant deterioration |
The American approach is the more aggressive. Recognising lifetime expected losses immediately means a bank writing a new loan records a loss on day one, before any payment has been missed.
This produces a counterintuitive result: growing the loan book depresses current earnings, even for high quality lending. Growth is penalised in the reporting period and rewarded later as the loans perform.
Why Earnings Became More Volatile
Because provisions now depend on forecasts of unemployment, growth, and asset prices, a change in the economic outlook moves reported profit directly.
2020 demonstrated this vividly. Banks took very large provisions in the first half of the year based on forecasts of severe deterioration. When the outcome proved far better than forecast, substantial amounts were released back into earnings during 2021.
Neither the charges nor the releases corresponded to loans actually going bad in those periods. They reflected changes in what the models expected, and the swing in reported earnings was considerable in both directions.
The Judgement Involved
Provisions depend on the economic scenarios chosen, the weights assigned to them, and management overlays applied where models are considered inadequate.
Overlays are where discretion concentrates. When models produce results management considers implausible, they adjust. That is sometimes genuinely necessary, since models calibrated on history struggle with unprecedented conditions, and it is also a mechanism for managing reported earnings.
The disclosure of overlays and of scenario weightings is where an analyst should look, because that is where the judgement sits.
What to Watch
The coverage ratio, allowance as a percentage of loans, indicates how much loss the bank has already recognised. A rising coverage ratio means the bank is building for expected deterioration.
Net charge offs measure loans actually written off, which is the realised experience rather than the forecast. Comparing charge offs against provisions shows whether the bank is building the allowance or releasing it.
A bank whose provisions fall well below charge offs is depleting its allowance, which flatters current earnings and cannot continue.
The Bottom Line
Loan loss provisions recognise expected losses in advance, and the shift to expected loss models moved that recognition considerably earlier. The intent was to reduce procyclicality, and the effect is that bank earnings now move with economic forecasts rather than with realised defaults. Read the coverage ratio, compare provisions against actual charge offs, and look for management overlays, which is where the discretion lives.