Expected Credit Loss Framework for Banks: CAIIB Guide (2026)
The expected credit loss framework is quietly rewriting how Indian banks measure, book and disclose credit risk. Instead of waiting for a loan to actually default before setting money aside, banks estimate probable future losses from day one and provision for them in advance. For CAIIB Risk Management candidates, this forward-looking model is one of the highest-yield topics: it links accounting standards, provisioning policy and capital planning into a single chain of logic.
India's banks currently provision under the RBI's Income Recognition and Asset Classification (IRAC) norms, an "incurred loss" system where provisions rise only after an account slips into the special mention or non-performing buckets. The expected credit loss framework, built on Ind AS 109 (the Indian version of IFRS 9), flips that timeline. This guide breaks down the three-stage model, the PD-LGD-EAD engine, the 12-month versus lifetime distinction, and where RBI's transition currently stands, all mapped to the way the exam questions it.
📊 What the Expected Credit Loss Framework Actually Measures
At its core, the expected credit loss framework replaces judgement-after-the-fact with statistics-before-the-fact. A bank no longer asks "has this loan gone bad?" but "how much do I expect to lose on this exposure over a defined horizon, weighted by the probability of each outcome?" That expected loss becomes the loss allowance carried against the asset on the balance sheet.
The calculation rests on three risk parameters that CAIIB candidates must recall cold. Probability of Default (PD) is the likelihood the borrower defaults over the measurement horizon. Loss Given Default (LGD) is the share of exposure not recovered after collateral and workout, expressed as a percentage. Exposure at Default (EAD) is the outstanding amount expected to be owed at the moment of default, including undrawn commitments likely to be drawn. The headline formula is simply ECL = PD × LGD × EAD, discounted to present value. Because these are the same building blocks used in the internal ratings-based approach, ECL sits naturally alongside the broader risk management framework a bank already runs for capital.
💡 Exam Tip: Remember ECL = PD × LGD × EAD. If a question gives you a 2% PD, 40% LGD and ₹100 crore EAD, the 12-month ECL is 0.02 × 0.40 × 100 = ₹0.8 crore. Examiners love this plug-and-play sum.
🔄 The Three-Stage Model and the SICR Trigger
Ind AS 109 sorts every exposure into one of three stages based on how its credit risk has changed since the loan was first recognised. Stage 1 holds performing assets whose credit risk has not increased significantly; here the bank books only 12-month ECL — losses from defaults expected within the next twelve months. Stage 2 captures assets that have shown a significant increase in credit risk (SICR) since origination but are not yet impaired; provisioning jumps to lifetime ECL, covering expected losses over the entire remaining life of the exposure. Stage 3 is for credit-impaired assets — broadly the equivalent of NPAs — which also carry lifetime ECL.
The stage-2 boundary is where most of the analytical work happens. A "significant increase in credit risk" is not the same as default; it is an early-warning threshold triggered by rating downgrades, sustained overdue behaviour, or macro deterioration. The 30-days-past-due rebuttable presumption is a common backstop. Getting the SICR judgement right is what separates a well-calibrated model from one that either over-provisions or lags reality. This forward-looking discipline is conceptually close to how banks stress their books under the Basel III capital adequacy framework.
⚠️ Common Mistake: Stage 2 does NOT mean the asset is an NPA. It means credit risk has risen significantly since origination — the loan can still be fully performing on repayments. Confusing SICR with default is a frequent exam trap.
⚖️ ECL vs the Incurred-Loss IRAC Model
The sharpest way to understand the expected credit loss framework is to hold it against the incurred-loss system it is meant to replace. Under IRAC, provisioning is triggered by objective evidence that a loss has already occurred — the account ages past due, gets classified, and only then attracts a provision. Critics argued this produced "too little, too late" reserves that amplified downturns. ECL front-loads recognition, smoothing provisions across the cycle.
| Feature | IRAC (Incurred Loss) | ECL (Expected Loss) |
|---|---|---|
| Trigger for provisioning | Loss event has occurred | Loss is expected in future |
| Forward-looking? | ❌ No | ✅ Yes |
| Uses PD / LGD / EAD? | ❌ No | ✅ Yes |
| Day-1 provision on a standard loan | Minimal, rule-based | 12-month ECL |
| Cyclicality | Pro-cyclical (lags) | Counter-cyclical (leads) |
| Basis in India | RBI IRAC circulars | Ind AS 109 / IFRS 9 |
Because ECL provisions can be volatile, they interact directly with a bank's capital and funding plans — a topic examined alongside asset liability management. A sudden migration of loans from Stage 1 to Stage 2 during a downturn can spike the loss allowance, dent the profit-and-loss account, and squeeze capital exactly when it is hardest to raise.
🏦 RBI's Transition and Why It Matters for Banks
As of mid-2026, scheduled commercial banks in India still provision under IRAC for regulatory purposes, even though listed banks already prepare Ind AS-aligned disclosures for their groups. RBI issued a discussion paper on adopting an expected-loss-based approach to provisioning, followed by draft guidelines proposing that banks build their own ECL models, validate them independently, and phase in the transition over several years with a prudential floor. The direction of travel is clear even if the exact switch-on date keeps evolving.
For risk managers, the practical implications are large. Banks must assemble long histories of default and recovery data, build and back-test PD/LGD/EAD models, embed a governance layer for model risk, and reconcile the new numbers against the regulatory floor. The design choices echo those in counterparty credit risk in banking, where the same parameters drive capital, and in market risk capital charge calculations. Even the statistical validation of these models leans on the significance testing covered in hypothesis testing. For a broader set of study notes on the theme, browse the risk management topic hub.
📌 Remember: India has not yet switched banks to ECL for regulatory provisioning — IRAC still governs. ECL is the proposed, forward-looking successor. Answer exam questions on current regulatory status accordingly.
🧠 Practice MCQs: Expected Credit Loss Framework
Q1. Under Ind AS 109, an exposure that has shown a significant increase in credit risk since origination but is not yet credit-impaired belongs to which stage? (a) Stage 1 (b) Stage 2 (c) Stage 3 (d) Stage 4
Answer: (b) — Stage 2 covers assets with a significant increase in credit risk (SICR) that are not yet impaired, and they attract lifetime ECL.
Q2. For which stage is only 12-month expected credit loss recognised? (a) Stage 3 (b) Stage 2 (c) Stage 1 (d) All stages equally
Answer: (c) — Stage 1 (performing, no SICR) carries only 12-month ECL; Stages 2 and 3 move to lifetime ECL.
Q3. Expected credit loss is computed as the product of which three parameters? (a) PD, LGD and EAD (b) VaR, PD and RWA (c) LGD, RWA and CAR (d) EAD, CAR and PD
Answer: (a) — ECL = PD × LGD × EAD (discounted), the standard three-parameter decomposition.
Q4. RBI's current IRAC provisioning norms are based on which loss model? (a) Expected loss (b) Incurred/realised loss (c) Fair value loss (d) Mark-to-market loss
Answer: (b) — IRAC is an incurred-loss model; provisioning follows evidence that a loss has already occurred.
Q5. For a Stage 3 (credit-impaired) asset under Ind AS 109, interest revenue is calculated on which base? (a) Gross carrying amount at 12-month ECL (b) Gross carrying amount at lifetime ECL (c) Net carrying amount (gross minus loss allowance) at lifetime ECL (d) Zero, as no interest accrues
Answer: (c) — Stage 3 assets carry lifetime ECL and interest is recognised on the net carrying amount (gross less the loss allowance).
Want chapter-wise mock tests with 100+ MCQs? Start practising free →
❓ Frequently Asked Questions
Is the expected credit loss framework already mandatory for Indian banks?
Not yet for regulatory provisioning. As of mid-2026 banks still provision under RBI's IRAC norms, while RBI has issued draft guidelines proposing a phased transition to an ECL-based approach.
What is the difference between 12-month and lifetime ECL?
12-month ECL captures expected losses from defaults possible within the next twelve months and applies to Stage 1 assets. Lifetime ECL covers expected losses over the whole remaining life of the exposure and applies to Stage 2 and Stage 3 assets.
What triggers a move from Stage 1 to Stage 2?
A significant increase in credit risk (SICR) since initial recognition — signalled by rating downgrades, sustained overdue status, or macro deterioration, with a rebuttable 30-days-past-due backstop.
Which accounting standard governs ECL in India?
Ind AS 109, the Indian convergence of IFRS 9, prescribes the three-stage impairment model and the PD-LGD-EAD approach to measuring expected credit losses.
Master the expected credit loss framework and you unlock a cluster of connected CAIIB Risk Management questions on provisioning, staging and capital. Anchor the three-stage logic and the PD-LGD-EAD formula, then test yourself under exam conditions with our CAIIB mock tests or dive deeper through the full CAIIB course to convert this theory into marks.
Practice this topic
Take a free mock test, download chapter PDFs, or watch a video class — all included on iibf.store.
