Non-Performing Loan Ratio: Calculate It Without Double Counting

Build the numerator from non-overlapping categories

Suppose a bank defines NPLs as nonaccrual loans plus loans at least 90 days past due that still accrue interest. Assume those categories are $240 million and $60 million, with $15 billion of gross loans in the same reporting scope. NPLs total $300 million and the ratio is $300m ÷ $15,000m = 2%. This is a hypothetical disclosure convention, not a universal definition.

A lower ratio can have different causes

Start with the same $300 million of NPLs and $15 billion loan book. The scenarios below change one item at a time. They show why the numerator and denominator deserve separate attention. A loan sold or charged off has left the reported loan balance; that does not establish that the borrower resumed paying.

Coverage, provisions and charge-offs answer different questions

With a $450 million loan allowance and $300 million of NPLs, allowance-to-NPL coverage is 150%. That calculation is not proof that every eventual loss is funded: the allowance can cover expected losses on performing loans as well. Match the allowance population to the loan book, and state whether a coverage measure uses all loan allowances or only allowances allocated to NPLs.

• The allowance is a balance-sheet valuation account. The provision is a period expense or reversal that changes it; neither is a separate cash reserve. A simplified roll-forward is opening allowance + provision − charge-offs + recoveries = closing allowance, before acquisitions, foreign exchange and other adjustments.
• For example, $400m opening allowance + $100m provision − $70m charge-offs + $20m recoveries = $450m closing allowance. Net charge-offs are $50m. If these are annual figures and average loans are $14bn, the annual net charge-off ratio is about 0.36%. A quarterly rate requires a stated annualization convention.
• Subtracting the entire allowance from NPLs in this example produces −$150m. That does not mean negative credit risk. Do not label it a universal net NPL measure or an amount of unhedged equity exposure; use the reporting institution’s precise definition.

Classification and accounting are not interchangeable

The ECB’s explanation includes both past-due loans and borrowers judged unlikely to repay. The exact threshold, materiality rules and scope come from the relevant reporting framework. NPL, nonaccrual, defaulted exposure and IFRS 9 Stage 3 are related concepts, but should not be treated as identical labels across jurisdictions.

• US CECL generally estimates lifetime expected credit losses for loans within its amortized-cost scope. IFRS 9’s general impairment model instead starts with 12-month expected credit losses and moves to lifetime losses after a significant increase in credit risk; credit-impaired assets also use lifetime losses. The 12-month measure is not merely cash shortfalls during the next year.
• ASU 2022-02 removed the separate creditor TDR recognition and measurement model for CECL adopters and introduced enhanced disclosures about modifications to borrowers in financial difficulty. Do not automatically add all modified loans to NPLs or reuse an old TDR formula without checking the reporting period.

Use comparable evidence instead of universal health bands

A 1% ratio does not by itself establish good underwriting, and 4% does not by itself establish insolvency. Compare loan mix, collateral, vintage, write-off policy, concentrations and the economic environment. Inspect early delinquencies and changes in credit risk as well: the NPL ratio may respond after conditions have already weakened. No fixed allowance-to-NPL percentage proves reserve adequacy for every bank.