How IFRS 9 Expected-Credit-Loss Algorithms Move Loans from 12-Month to Lifetime Losses: SICR, PD/LGD/EAD, Scenarios, Discounting and Stage Boundaries

Reader question: A loan is still paying on time today. Why might its loss allowance suddenly increase from a 12-month expected-credit-loss measure to losses over its full remaining life?

IFRS 9 answers with a change-in-credit-risk framework. At initial recognition and while credit risk has not increased significantly, the entity generally recognises 12-month expected credit losses. If credit risk increases significantly since initial recognition, it recognises lifetime expected credit losses. If the asset becomes credit-impaired, lifetime ECL continues, while interest revenue is generally calculated on the amortised-cost basis rather than the gross carrying amount.

The accounting standard defines ECL through probability-weighted discounted cash shortfalls, not through one mandatory bank formula. Many banks implement the measurement using probability of default (PD), loss given default (LGD) and exposure at default (EAD), combined across time and macroeconomic scenarios. That decomposition is useful, but the accounting target remains the IFRS 9 expected cash shortfall.

What this page owns — and what it does not

This page owns:

credit-risk state + expected cash shortfalls + scenarios + discounting → 12-month or lifetime ECL allowance.

It does not replace LGD estimation, underwriting models, IFRS 9 effective-interest calculations, or model-drift monitoring. Those are adjacent owners.

This is public accounting mathematics, not a credit decision, not an assessment of a real borrower and not personalized financial advice.

The accounting target: discounted expected cash shortfalls

IFRS 9 defines credit loss in terms of the difference between contractual cash flows due and cash flows the entity expects to receive, discounted using the appropriate effective-interest-rate framework.

Measurement of ECL must reflect:

• an unbiased and probability-weighted amount from a range of possible outcomes;
• the time value of money;
• reasonable and supportable information available without undue cost or effort about past events, current conditions and forecasts of future economic conditions.

So ECL is not simply “historical average default rate × current loan balance”.

The common three-stage shorthand

Industry practice commonly describes the general IFRS 9 impairment model as three stages:

Common shorthand	Loss allowance	Interest basis
Stage 1	12-month ECL	gross carrying amount
Stage 2	lifetime ECL	gross carrying amount
Stage 3 / credit-impaired	lifetime ECL	amortised cost / net basis

The stage labels are useful operational shorthand. The actual IFRS 9 logic is driven by significant increases in credit risk and credit-impaired status.

What 12-month ECL actually means

This is one of the most common misconceptions.

IFRS 9 states that 12-month ECL is the portion of lifetime expected credit losses associated with default events that are possible within the next 12 months.

It is not:

• the cash losses expected to occur during the next 12 months;
• the lifetime losses only on loans that management predicts will definitely default in the next 12 months.

If a default occurs in month 10 and causes cash shortfalls extending for years afterward, those lifetime shortfalls can belong in the 12-month ECL measure because the triggering default event occurs within the 12-month horizon.

Lifetime ECL looks across the expected life

Lifetime ECL reflects default risk across the instrument’s expected life and the resulting expected cash shortfalls.

A stylised discrete-time representation is:

Lifetime ECL = Σ_t Expected Cash Shortfall_t × Discount Factor_t.

The expected cash shortfall at each time already incorporates probabilities across outcomes.

PD × LGD × EAD is a common implementation, not the IFRS definition

Banks often operationalise ECL using:

ECL ≈ Σ PD × LGD × EAD × Discount Factor.

More precisely, a multi-period implementation can use marginal default probabilities at each future time, conditional survival, scenario-specific LGD and EAD, and the effective-interest-rate discount factor.

But IFRS 9 does not prescribe one universal PD/LGD/EAD formula. A model that directly estimates probability-weighted discounted cash shortfalls can also satisfy the accounting objective if it meets the standard’s requirements.

A simple Stage 1 illustration

Suppose a loan has:

• EAD = 100;
• 12-month probability of default = 2%;
• LGD = 40%;
• no material discounting for this simple illustration.

A basic one-period estimate is:

ECL = 100 × 0.02 × 0.40 = 0.80.

This is only a teaching simplification. Production ECL can include multiple periods, contractual cash flows, prepayments, cures, collateral timing, revolving utilisation and several macroeconomic scenarios.

Step 1: establish credit risk at initial recognition

The staging decision is relative. IFRS 9 asks whether the risk of default has increased significantly since initial recognition.

A borrower that was risky at origination can remain in the 12-month ECL category if its credit risk has not increased significantly. A borrower that was extremely strong at origination can move to lifetime ECL after deterioration even if its absolute current default probability is still moderate.

This is why origination credit-risk data must be preserved.

Step 2: test for significant increase in credit risk

The SICR assessment can use:

• changes in lifetime or relevant-horizon PD;
• internal credit grades;
• external ratings where relevant;
• delinquency;
• forbearance or restructuring signals;
• adverse industry or borrower information;
• macroeconomic forecasts;
• watchlist indicators;
• changes in collateral or covenant conditions when they inform default risk.

The assessment is meant to be forward-looking when reasonable and supportable information is available.

30 days past due is a backstop, not the whole SICR model

IFRS 9 contains a rebuttable presumption that credit risk has increased significantly when contractual payments are more than 30 days past due.

But IFRS material also stresses that significant increases in credit risk are generally expected to be recognised before an instrument becomes past due when forward-looking information indicates deterioration.

A bank that stages loans only after 30 days past due can therefore react too late.

The low-credit-risk practical expedient

IFRS 9 permits a practical expedient under which an entity can assume credit risk has not increased significantly if the instrument has low credit risk at the reporting date, subject to the standard’s conditions.

This is often associated with investment-grade-style characteristics, but the assessment should follow the IFRS 9 definition rather than a mechanical external-rating label.

Step 3: identify credit-impaired assets

An asset becomes credit-impaired when one or more events have a detrimental impact on estimated future cash flows.

Indicators can include severe borrower financial difficulty, breach/default, concessions for financial difficulty, probable bankruptcy or similar evidence.

IFRS 9 also contains a rebuttable presumption that default does not occur later than 90 days past due unless reasonable and supportable information justifies a more lagging criterion. Institutions align default definitions with their risk-management processes subject to the accounting requirements.

Stage 3 changes the interest basis

For non-credit-impaired assets, interest revenue generally uses the effective interest rate applied to the gross carrying amount.

For a financial asset that becomes credit-impaired, IFRS 9 generally requires interest revenue to be calculated by applying the effective interest rate to the amortised cost, which is the gross carrying amount after the loss allowance.

This connects directly to the existing effective-interest article.

Scenario weighting: one macro forecast is not enough when outcomes are nonlinear

IFRS 9 requires an unbiased probability-weighted amount that evaluates a range of possible outcomes.

A common implementation uses scenarios such as:

• baseline;
• upside;
• downside;
• severe downside where justified.

If scenario s has probability w_s and ECL L_s:

Probability-weighted ECL = Σ w_sL_s.

Scenario weights must sum to one.

Why “ECL at average GDP” can be wrong

Credit losses are nonlinear. A 5% fall in property prices can have little impact on a highly collateralised mortgage, while a 25% fall can move many loans into loss territory.

Therefore:

ECL(expected macro variables) ≠ expected ECL

when the loss function is nonlinear.

Explicit scenario weighting captures some of that nonlinearity.

Forward-looking information is part of the algorithm

Historical defaults alone are not enough. IFRS 9 requires reasonable and supportable forecasts of future economic conditions when available without undue cost or effort.

Inputs can include unemployment, GDP, house prices, interest rates, commodity prices or sector variables depending on the portfolio.

The chosen macro variables should have an economically and empirically defensible relationship with default, recovery or exposure.

Discounting matters

A cash shortfall expected in five years is not the same accounting amount as the same nominal shortfall tomorrow.

ECL therefore incorporates the time value of money using the relevant effective-interest-rate framework.

A model that computes PD×LGD×EAD across future years but forgets discounting can overstate later losses.

Collateral enters through expected cash recoveries, not as a magic deduction

If a loan is secured, expected cash flows can include recoveries from collateral where consistent with the contractual and accounting requirements.

The ECL engine must model:

• collateral value;
• haircuts and sale costs;
• time to recovery;
• probability of enforcement;
• seniority;
• legal constraints;
• correlation between default and collateral value.

This links to LGD modelling.

Exposure at default can change before default

For a term loan, future EAD may decline through amortisation.

For a credit card or revolving facility, utilisation can rise as the borrower weakens.

Therefore EAD is often a future path rather than today’s balance copied into every period.

Loan commitments and financial guarantees also bring off-balance-sheet exposures into ECL measurement under IFRS 9.

Revolving credit facilities are a special horizon problem

Some revolving facilities can expose the lender to credit risk for longer than the contractual notice period because normal credit-risk management does not necessarily withdraw the undrawn commitment immediately.

IFRS 9 includes special guidance for determining the period over which ECL is measured for such facilities.

A generic “maturity date = ECL horizon” rule can therefore be wrong.

Purchased or originated credit-impaired assets are different

POCI assets are credit-impaired at initial recognition.

Their accounting uses a credit-adjusted effective interest rate and recognises changes in lifetime ECL since initial recognition rather than applying the ordinary Stage 1-to-Stage 2 sequence in the same way.

A staging engine must identify POCI status explicitly instead of forcing every asset through one general state machine.

Simplified approach for some receivables

IFRS 9 permits or requires a simplified lifetime-ECL approach for specified trade receivables, contract assets and lease receivables.

Provision matrices can estimate loss rates by ageing bucket and adjust them for forward-looking information.

This is another reminder that “Stage 1/2/3 for every receivable” is not a universal IFRS 9 implementation.

Management overlays are not a substitute for model repair

Banks can use post-model adjustments when models do not capture known risks, for example after a sudden economic shock or structural change.

But an overlay needs:

• clear rationale;
• quantification method;
• governance and approval;
• double-counting checks;
• expiry/review criteria.

A permanent unexplained overlay can hide a broken underlying model.

Evidence polarity: what supports confidence?

Evidence for a reliable ECL calculation includes preserved origination risk, staging rules that respond before delinquency where appropriate, scenario weights that sum to one, PD/LGD/EAD or cash-shortfall models that reconcile to observed outcomes, effective-rate discounting, collateral recoveries supported by evidence, and allowance movements explainable by portfolio and macro changes.

Evidence against confidence includes every asset remaining Stage 1 until 30 days past due, sudden unexplained Stage 2 spikes at quarter-end, scenario weights chosen to hit an allowance target, lifetime PD lower than 12-month PD, LGD that improves automatically in severe downturn scenarios, or ECL that cannot be reconciled to contractual and expected cash flows.

Counterexample: a high absolute PD can remain Stage 1

Suppose a high-yield borrower had a 10% lifetime PD at origination and still has roughly the same credit risk today.

The absolute risk is high, but credit risk may not have increased significantly since initial recognition. The general model can still use 12-month ECL.

SICR is about change in default risk, not simply crossing one universal PD level.

Counterexample: a low absolute PD can move to Stage 2

Suppose an investment-grade borrower’s lifetime PD rises from 0.2% to 1.0%.

The current PD is still low in absolute terms, but the relative deterioration may be significant depending on the instrument and methodology.

A staging system that looks only at absolute PD thresholds can miss this.

Counterexample: 12-month ECL can include losses after month 12

If default occurs in month 8 but collateral recovery does not arrive until month 30, the discounted shortfall after month 12 can still belong in 12-month ECL because the default event occurred within the first 12 months.

This directly falsifies the common “only next-year cash losses” interpretation.

Counterexample: one base-case scenario can understate ECL

Suppose baseline unemployment is 4%, upside 3% and downside 9%. If defaults accelerate nonlinearly above 7%, calculating ECL only at the probability-weighted average unemployment rate can miss the downside tail.

Probability-weighted scenario losses are more faithful to the IFRS 9 objective.

Weak links in implementation

origination-data loss. SICR cannot be measured correctly without an initial risk anchor.

PD-horizon mismatch. A 12-month PD is compared with a lifetime PD without adjustment.

delinquency-only staging. Forward-looking deterioration is ignored.

scenario double counting. Macroeconomic stress enters both PD and an overlay twice.

LGD timing error. Recoveries are not discounted to the reporting date.

EAD static-balance error. Amortisation or future drawdown is ignored.

12-month misconception. Cash shortfalls after 12 months are excluded incorrectly.

Stage 3 interest error. Interest remains on gross carrying amount after credit impairment.

Diagnostics: how to test the engine

• scenario-weight test: probabilities sum to one and results reconcile to scenario ECLs.
• 12-month definition test: include a default in month 6 with losses extending beyond month 12.
• SICR relative-risk test: compare current risk with origination risk, not just a fixed threshold.
• 30-DPD test: verify the rebuttable presumption and earlier forward-looking triggers.
• credit-impaired test: Stage 3 uses lifetime ECL and the correct interest basis.
• discounting test: push the same recovery farther into the future and verify present value falls.
• revolving-EAD test: allow stressed utilisation to increase before default.
• collateral test: reduce collateral values in downside scenarios and verify LGD response.
• backtest: compare prior ECL assumptions with realised defaults, recoveries and exposure paths.
• overlay test: remove an overlay and prove which known risk would become unrepresented.

What would falsify confidence?

Confidence should be withdrawn if the system cannot reproduce the movement from 12-month to lifetime ECL; if origination credit risk is unavailable; if ECL ignores probability weighting or discounting; if downside scenarios produce lower loss without an economic explanation; if backtesting shows persistent underprediction; or if management overlays cannot be traced to specific model gaps.

Alternatives and limits

CECL under U.S. GAAP uses a different lifetime-loss framework and should not be treated as interchangeable with IFRS 9 staging. Regulatory expected-loss calculations under Basel have different objectives again.

Within IFRS 9, banks can use PD/LGD/EAD, discounted cash-flow, roll-rate, provision-matrix or other techniques depending on portfolio characteristics, provided the result satisfies the standard’s measurement principles.

The IASB completed its post-implementation review of IFRS 9 impairment in 2024 and concluded that the core impairment requirements are working as intended, while identifying areas for targeted clarification and disclosure work. In 2026, separate IASB work on amortised-cost measurement continued to consider modification and derecognition questions. Those projects should be monitored as update triggers rather than treated as changes to the core ECL model unless and until final amendments are issued.

How this connects to the surrounding knowledge estate

PD reasoning connects to rating-transition matrices. LGD connects to workout and recovery modelling. Discounting and carrying amount connect to effective interest. Persistent prediction errors feed model-drift monitoring.

Verification and update triggers

Preserve accounting-standard version, origination-risk measures, staging methodology, scenario definitions and weights, PD/LGD/EAD model versions, discount rates, collateral assumptions, overlays and backtesting evidence. Revalidate after major macro regime changes, model redevelopment, portfolio acquisitions, restructuring programmes, default-definition changes or final IFRS amendments affecting impairment.

Primary and high-quality references

• IFRS Foundation, IFRS 9 Financial Instruments.
• IFRS Foundation, IFRS 9 issued standard text, including expected-credit-loss definitions and measurement guidance.
• IFRS Foundation, IFRS 9 Project Summary, including the 12-month/lifetime impairment-stage illustration.
• IFRS Foundation, Post-implementation Review of IFRS 9 — Impairment.
• Basel Committee on Banking Supervision, Guidance on credit risk and accounting for expected credit losses.

Educational boundary: This article explains IFRS 9 impairment mathematics and model mechanics. It does not determine an allowance for any real entity, loan or borrower and does not provide accounting, audit or personalized financial advice.