Recognising trouble before default

In its first-quarter 2020 results, JPMorgan Chase reported $8.3 billion of credit costs, including a $6.8 billion reserve build. The distinction matters. Most of that quarter's reported cost reflected an increase in amounts set against expected losses, rather than loans newly written off during the quarter. The bank linked the build to the prospect of a severe recession as the pandemic spread.

Expected-credit-loss accounting is designed to recognise that deterioration early. A bank need not wait until every doubtful borrower has stopped paying before recognising that its loan portfolio is worth less than the contractual payments suggest. It must estimate the shortfall using the relevant accounting framework. The estimate can move even while many customers remain up to date.

Follow the allowance

The provision is an expense, or sometimes a reversal of expense, recognised over a reporting period. The allowance is the accumulated valuation adjustment remaining at a particular date. For loans measured at amortised cost, it reduces their net carrying amount. Neither is a separate vault of cash waiting to be handed out.

In a simple invented example, a bank starts with a £10 million loan-loss allowance. It records a £5 million provision and writes off £3 million of loans against the allowance. With no recoveries or other changes, the closing allowance is £12 million. The £5 million expense belongs to the period's profit calculation. The £3 million measures loans written off. The £12 million remains as the closing allowance for expected losses.

A write-off uses an allowance already recognised, although fresh information can also require an additional provision. Recoveries of previously written-off amounts and other adjustments complicate real accounts. This is why the reconciliation of the allowance is more useful than a headline saying a bank has merely “put aside” money. US banking agencies explain the allowance's valuation role in their CECL guidance.

Two frameworks with different clocks

Under the US current expected credit losses model, or CECL, covered assets measured at amortised cost generally receive an allowance for expected losses over their contractual life from origination or acquisition. Historical experience, current conditions and reasonable, supportable forecasts inform the estimate. The allowance is subsequently updated as the relevant information changes.

IFRS 9 uses a different general model. Performing loans normally begin with 12-month expected credit losses. A significant increase in credit risk brings recognition of lifetime expected losses, even before the loan becomes credit-impaired. Credit-impaired assets also carry lifetime expected losses, with different treatment of interest revenue. The framework contains exceptions and specialised treatments, so this description concerns the general loan model.

The trap lies in the phrase “12-month loss”. It refers to losses over the life of the loan arising from defaults that could occur within the next twelve months, weighted by their probability. It does not mean simply the instalments likely to go unpaid during that year. A default near the end of the twelve-month window can create a cash shortfall much later.

A comparison of banks must therefore start with their accounting basis, loan mix and measurement horizon. The same provision ratio can carry different information under the two systems.

Read the forecast inside the expense

Suppose an otherwise unchanged portfolio is more likely to suffer defaults under a weaker employment outlook. An expected-loss model can produce a larger allowance before arrears rise. If the outlook subsequently improves, a release can lift profit. That improvement in reported earnings does not mean a new customer paid an additional fee or that the bank made more loans.

The opposite interpretation also needs care. A rising allowance can reflect rapid loan growth as well as worsening credit. An unchanged allowance can conceal offsetting movements: deterioration in one portfolio and contraction in another. The useful evidence lies in portfolio balances, arrears, write-offs, changes in assumptions and the allowance reconciliation.

In the invented bank, the £5 million expense, £3 million of write-offs and £12 million closing allowance cannot be used interchangeably. Nor can a subsequent release be treated as evidence that lending income improved. The allowance reconciliation shows how expectations moved through the accounts; arrears and write-offs help test those expectations against borrowers' behaviour.