A non-performing loan (NPL) is a loan where the borrower is unlikely to pay or is more than 90 days past due. IFRS 9 impairment staging sorts financial assets into three buckets based on how much credit risk has increased since origination; Stage 3 means the asset is credit-impaired. Banks use these stages to measure expected credit loss (ECL) and to set how they recognize interest income.
Understanding how Stage 3 and NPLs overlap, where they differ, and how they flow through earnings and capital is the fastest way to evaluate a bank’s credit quality. With rates higher and maturity walls approaching, the payoff is clear: you can forecast migration before it hits the income statement.
Stage 3 and NPLs: aligned in concept, different in practice
Stage 3 broadly lines up with NPLs, but the mapping is not perfect. IFRS 9 uses credit-impaired with a 90 days past due default backstop and unlikeliness-to-pay indicators. The European Banking Authority’s NPL regime uses similar anchors but adds forbearance rules, probation periods, and calendar provisioning for capital. When you analyze a bank, reconcile Stage 3 in the financial statements with NPL or NPE disclosures in the risk report. Small definitional differences can explain large swings in earnings and capital.
The staging engine in plain English
IFRS 9 stages are mechanical once policies are set, but their impact is material. Keep this one-page rule set in mind.
- Stage 1: 12-month ECL with interest accrued on the gross carrying amount.
- Stage 2: Lifetime ECL once there is a significant increase in credit risk since origination, with interest still on gross.
- Stage 3: Lifetime ECL with interest recognized on the net carrying amount, which is gross minus the loss allowance.
Default and SICR: the anchors behind staging
Default is an entity-defined concept anchored by a 90 days past due rebuttable presumption and unlikeliness-to-pay events such as a bankruptcy filing, distressed restructuring, material covenant breach with concessions, fraud, or enforcement on collateral. The 30 days past due rebuttable presumption supports significant increase in credit risk at Stage 2. Rebutting either presumption takes strong, consistent evidence and documented governance, not ad hoc overrides.
Stage 3 changes interest income by design
Once an asset becomes credit-impaired, the interest calculation switches to a net basis. Consider a loan with gross amortized cost of 100, an effective interest rate of 5 percent, and a Stage 3 allowance of 40. Interest in Stage 1 or Stage 2 is 5. In Stage 3, interest is 5 percent on 60, or 3. That is a 40 percent drop in reported interest on this asset. The effect compounds across portfolios, so model net interest margin sensitivity to Stage 3 migration before earnings surprise you.
Collateral and day-count: two frequent missteps
Collateral quality does not keep a loan out of Stage 3 if the borrower is unlikely to pay. Collateral reduces loss given default, not the default decision. Day-count mechanics carry weight too. Payment holidays, grace periods, and deferrals pause or reset days past due only if granted before default under documented policies. Administrative changes that lower reported days past due without substance draw scrutiny and invite reclassification later.
Forward-looking ECL needs real governance
All stages require forward-looking modeling across multiple macro scenarios with probability weights and coherent probability of default, loss given default, and exposure at default paths. Overlays are fine when models fall short, but they need rationale, thresholds, approvals, and disclosure. If overlays drive most of the ECL movement, the core models are not doing their job and earnings volatility will rise.
What actually moves loans into Stage 3
Default, under the entity’s documented policy, is the core trigger. The 90 days past due backstop is decisive unless the bank documents a timing anomaly, applies it consistently, and can evidence it. Unlikeliness-to-pay signals include a distressed restructuring, initiation of enforcement, or a bankruptcy process. Note the practical point: collateral does not offset the classification; it offsets the loss once default is established.
Modifications, forbearance, and POCI: how classification interacts
A substantial modification that passes the derecognition test becomes a new asset at fair value with a new origination date. A non-substantial modification keeps the original asset and usually increases ECL. If the concession reflects borrower financial difficulty, that often points to Stage 3. Forbearance under prudential rules is a major NPL driver and often overlaps with Stage 3 treatment under IFRS 9. Purchased or originated credit-impaired assets start life as impaired. They do not move through Stages 1 to 3. The purchase price embeds expected losses and a credit-adjusted effective interest rate governs interest. Buyers of NPL pools should confirm POCI treatment and trace how cash flow updates flow through earnings.
Cure and write-offs: criteria that hold up in audit
Cure from Stage 3 back to Stage 2 requires more than clearing arrears. Banks should see sustained performance, the removal of unlikeliness-to-pay indicators, and a documented probation period. Write-offs occur when recovery in full or in part is not reasonably expected. A write-off reduces gross carrying amount, does not kill legal rights, and can be partial. Clear criteria speed decisions and keep disclosure consistent.
What the data is signaling now
EU banks reported an NPL ratio of about 1.8 percent at Q2 2024. Headline NPL ratios appear stable, but many portfolios saw Stage 2 drift up through 2023 and 2024 as rates rose. That mix shift is the early alarm for Stage 3 migration over the next four quarters. Treat Stage 2 inflows as the canary in the coal mine. It usually sings before defaults hit Stage 3.
Incentives shape the gray areas
Management prefers stability and leans on significant increase in credit risk thresholds, cure rules, overlays, and forbearance tools to smooth outcomes. Auditors and supervisors want unbiased, probability-weighted ECL supported by observable default anchors and consistent judgments. Investors should focus on the gap between Stage 2 and Stage 3, the pace and evidence behind cures, and whether interest is still accruing on loans that are, in practice, not performing.
Capital vs accounting loss: model both views
Prudential rules in the EU add the NPE backstop with minimum provisioning timelines for new NPEs that feed regulatory capital. You can see an exposure with strong collateral and modest IFRS 9 ECL still trigger calendar provisioning add-ons for capital. When you model CET1, run both the IFRS 9 ECL view and the prudential backstop. A conservative capital forecast avoids surprises even when accounting loss looks tame.
A pragmatic diligence playbook
- Default policy: Compare definitions to regulatory anchors and test the 90 days past due application on samples.
- SICR thresholds: Review quantitative triggers and qualitative watchlist criteria.
- Overrides log: Inspect exceptions and governance for rebuttals of the 30 or 90 days past due presumptions.
- Forbearance toolkit: Map concession types, eligibility, and probation rules to NPL and Stage 3 outcomes.
- Data lineage: Reconcile days past due from the data warehouse to the core servicing system, including holiday and deferral tags.
- Behavioral metrics: Analyze restructuring volumes, cure rates, and relapse rates by vintage.
Three quick kill tests
- Stage 2 surge: If Stage 2 inflows run at more than three times Stage 3 inflows for several quarters without a benign macro story, staging or forbearance may be masking defaults.
- Watchlist lag: If watchlist balances are flat while days past due based Stage 2 rises, qualitative capture is weak.
- Overlay dominance: If overlays dominate ECL movement, models need attention and the audit challenge risk is rising.
Structures and hedging that reference staging
Warehouse lines and securitizations often embed Stage-linked triggers. Triggers can divert cash when Stage 2 or Stage 3 ratios breach thresholds. Borrowing bases may haircut Stage 2 collateral and exclude Stage 3 absent eligibility conditions. Attach the originator’s impairment policy to the facility, lock change-of-policy consent rights, and fix reporting cadences to avoid disputes. In loan sales, understanding how private equity funds price non-performing loans is critical to setting bid levels and reserving for loss on sale.
Stage 3 also reduces interest income because of net-basis recognition. If the bank hedges on gross balances, a Stage 3 wave can create accounting mismatches. Test sensitivity of net interest income to Stage 3 migration and to overlay scenarios that could unwind. When risk is transferred synthetically, disclosures around significant risk transfer clarify whether capital relief lines up with loss recognition.
Taxes and cross-framework comparisons
Tax timing differs by jurisdiction. Some tax codes track accounting ECL, while others wait for realization. Model cash taxes accordingly during Stage 3 ramps. Under US GAAP, CECL books lifetime expected losses at origination with no stages, and nonaccrual policies handle interest recognition on troubled loans. Adjust comparatives between IFRS and US banks for these structural differences before you judge relative coverage ratios.
Governance that reduces earnings surprises
An effective IFRS 9 program has clear data lineage, validated probability of default, loss given default, and exposure at default models within a model risk framework, scenario governance, and change control. Internal audit should review staging decisions, overrides, and disclosures end to end. Strong governance lowers earnings surprises and speeds supervisory reviews, which ultimately lowers the cost of capital.
Implementation path that actually works
- Policy alignment: Define default and significant increase in credit risk policies that match risk management and market practice.
- Model build and challenge: Develop PD, LGD, EAD models and scenario paths; validate independently and run challenger models where material.
- Engine configuration: Capture days past due, watchlist flags, forbearance markers, and automated 30 or 90 days past due backstops with controlled manual overrides.
- Overlay discipline: Set documented triggers, approval limits, and transparent disclosure.
- Reporting and controls: Integrate accounting and disclosure modules to produce Stage 1, Stage 2, and Stage 3 roll-forwards, write-offs, recoveries, and IFRS 7 credit risk disclosures.
- Dry runs: Rehearse quarterly closes with audit-ready workpapers and sensitivity analyses.
- Final approvals: Secure auditor alignment on significant judgments and Board approval of the impairment policy before first close.
Common pitfalls that cost time and credibility
- Weak rebuttals: Under-documented rebuttals of 30 or 90 days past due presumptions invite later reversals.
- Collateral misuse: Using collateral value to avoid Stage 3 rather than to set LGD drives misclassification.
- Premature cures: Booking early cures after restructurings without evidence of sustained performance triggers restatements.
Model Stage 3 migration with confidence
Use three lenses for a forecast that management, audit, and investors will accept. Portfolio drivers include sector mix, interest reset schedules, borrower leverage, and collateral liquidity. Policy drivers include significant increase in credit risk thresholds, forbearance criteria, and cure rules. Governance drivers include overlay magnitude, model monitoring results, and prior audit findings. When all three point the same way, your forecast is probably in the right zone for the next four to eight quarters.
Monitor the right signals between quarters
Stage 2 balances, loans 31 to 90 days past due as a share of total, new forbearance volumes, and watchlist entries are leading indicators. Confirm with Stage 3 inflows, write-offs, and actual recoveries versus plan. Compare Stage 2 to Stage 3 transitions period over period. A widening gap deserves attention. In secured books, trends in asset-based lending borrowing base haircuts can also foreshadow staging moves.
Scenario weights and overlays: be explicit
Use market-anchored scenarios with credible alternates, not automatic reversion to trend. If overlay weights change quickly, explain why, show backtesting, and keep disclosures clear. The market reads abrupt overlay moves as a message, so make sure it is the one you intend to send.
Boundary conditions that keep you grounded
Stage 3 equals credit-impaired triggered by default or unlikeliness-to-pay with a 90 days past due backstop. Collateral does not override that classification. Cure requires sustained performance with documented probation. Keep those three in view and the rest of the framework falls into place.
An investor’s shortcut framework
Tie Stage flows to macro and portfolio composition. Compare restructurings granted to restructurings requested to gauge borrower stress. Sample files to confirm unlikeliness-to-pay indicators and ensure contractual changes address financial difficulty, not administrative housekeeping. Then quantify earnings and capital sensitivity to plausible Stage 3 paths. You do not need heroics, just a consistent map from policy to practice and from practice to numbers.
Key Takeaway
Stage 3 and NPLs live in the same neighborhood but they do not share an address. If you track days past due anchors, unlikeliness-to-pay evidence, forbearance rules, and overlay discipline, you can anticipate migration and its impact on net interest income and capital before it shows up in the headline NPL ratio.