RBI’s ECL Framework for Banks: Complete Guide (2025)
RBI’s ECL (Expected Credit Loss) framework replaces the current incurred-loss IRAC provisioning with a forward-looking model. Banks must classify all loans into Stage 1 (performing — 12-month ECL), Stage 2 (SICR triggered — lifetime ECL), or Stage 3 (credit-impaired — lifetime ECL). Provisioning = PD × LGD × EAD × Discount Factor, adjusted for macroeconomic scenarios. Applicable to all SCBs; minimum 2-year parallel run required before full adoption.
For decades, Indian banks provisioned for loan losses only after a borrower had already defaulted. The logic was simple: recognise a loss when you can see it. This is the incurred loss model — the backbone of India’s IRAC (Income Recognition and Asset Classification) norms that govern how banks classify and provide for bad loans.
The problem? By the time a loss is “visible,” it is often too late. Banks were consistently under-reserved heading into stress periods, requiring sharp one-time hits to their P&L and capital. The COVID-era moratorium and the post-pandemic NPA surge brought this structural weakness into sharp focus.
The Reserve Bank of India has now mandated a structural shift. Under the new Expected Credit Loss (ECL) framework, banks must provision for losses they expect to occur — not just those that have already happened. The model is forward-looking, probability-weighted, and built on borrower-level data. This guide covers everything you need to understand and implement it.
Why RBI Is Moving Away from IRAC
India’s current provisioning system is rule-based. An account becomes an NPA (Non-Performing Asset) at 90 days past due (DPD). Provisions are then applied at fixed rates — 15%, 25%, 40%, 100% — depending on the asset category and how long it has been classified as NPA.
This system has three structural weaknesses that ECL directly addresses:
Provisions are triggered by a default that has already occurred, not by deteriorating credit quality that signals a future default. The damage is done before the system reacts.
An account at 89 DPD carries minimal provisioning. At 90 DPD, it becomes an NPA. The provision jumps sharply — creating P&L volatility and a perverse incentive to delay recognition.
Macro-economic deterioration, sector stress, or a borrower’s declining financials are not inputs into the IRAC provisioning model. Banks can only act after the damage is done.
The ECL framework, aligned with IFRS 9 (International Financial Reporting Standard 9), corrects all three. Provisions are estimated continuously — based on statistical models, macroeconomic scenarios, and borrower-level credit quality — even for performing loans.
Who Does the RBI ECL Framework Apply To?
| Entity Type | Covered? |
|---|---|
| All Scheduled Commercial Banks (SCBs) including PSBs and private banks | ✔ Yes |
| Foreign banks operating as branches or subsidiaries in India | ✔ Yes |
| Small Finance Banks (SFBs) | ✔ Yes |
| Regional Rural Banks (RRBs) | ✘ Excluded (for now) |
| Co-operative banks and Local Area Banks | ✘ Excluded (for now) |
RBI has mandated a parallel run of at least two years before full switchover. During the parallel run, banks must compute both IRAC and ECL provisions simultaneously — reporting both — before ECL figures become the regulatory standard.
The Three-Stage ECL Model
The heart of the ECL framework is a three-stage classification of all financial assets (loans, advances, investments). The stage a loan falls into determines what type of ECL provision is required.
What Triggers Stage Migration: SICR Explained
The movement from Stage 1 to Stage 2 is governed by Significant Increase in Credit Risk (SICR). This is one of the most judgement-intensive — and operationally demanding — parts of the entire ECL framework.
Quantitative SICR Triggers
- 30 DPD rebuttable presumption: An account that is 30+ days past due is presumed to have experienced SICR. Banks can rebut this with strong evidence that 30 DPD is not indicative of credit stress (e.g., a technical delay with no underlying financial deterioration).
- Significant increase in lifetime PD compared to the PD at origination — both in absolute terms (e.g., PD rising from 0.5% to 2%) and relative terms (e.g., PD doubling since origination).
Qualitative SICR Triggers
- Restructuring or rescheduling of the loan (including OTR — One Time Restructuring)
- Placement on internal watchlist or Special Mention Accounts (SMA-1, SMA-2)
- Deterioration in external credit rating for rated borrowers
- Adverse change in business conditions, industry outlook, or geographic risk
- Legal disputes or fraud flags raised against the borrower
Stage Curing (Back Migration)
An account can move back from Stage 2 to Stage 1 if SICR conditions are no longer present. However, RBI requires a probation period before back migration — preventing gaming of stage boundaries where an account is temporarily cured to reduce provisions and then slips again.
How ECL Is Calculated: PD, LGD, EAD
PD — Probability of Default
PD is the statistical likelihood that a borrower will default within a defined time horizon — 12 months for Stage 1, and over the full remaining loan life for Stage 2 and 3. Banks must build PD models using internal historical default data (minimum 5–7 years), credit bureau data, and macroeconomic variables. Critically, RBI requires Point-in-Time (PiT) PD — which reflects current economic conditions — rather than the Through-the-Cycle (TtC) PD used in Basel capital models.
LGD — Loss Given Default
LGD is the share of the exposure the bank will not recover if the borrower defaults: LGD = 1 − Recovery Rate. Recovery depends on collateral type and value, seniority of the claim, and legal recovery timelines (particularly relevant given India’s SARFAESI and DRT processes). RBI requires Downturn LGD — recovery rates estimated under economic stress conditions, not average conditions — making it higher than simple historical averages.
EAD — Exposure at Default
EAD is the outstanding exposure at the time of default. For term loans it is largely known. For revolving facilities — credit cards, working capital limits, overdrafts, cash credit with drawing power — EAD must be modelled using a Credit Conversion Factor (CCF): the proportion of undrawn commitment expected to be drawn before default.
Discount Factor
Future expected cash flows are discounted back to the present using the loan’s Effective Interest Rate (EIR), not a generic risk-free rate. This ensures the time value of money is correctly reflected in ECL estimates for long-tenure loans.
Forward-Looking Information: The Macro Dimension
One of the most significant departures from IRAC is the mandatory requirement to incorporate Forward-Looking Information (FLI) into ECL estimates. Banks cannot rely solely on historical default rates — they must model the future.
Build at least 3 scenarios — Base, Optimistic, Pessimistic — using macro variables: GDP growth, inflation, interest rates, sectoral indices.
Assign probability weights to each scenario (e.g., Base 60%, Optimistic 20%, Pessimistic 20%). Board must approve weights quarterly.
Final ECL = (P_base × ECL_base) + (P_optimistic × ECL_optimistic) + (P_pessimistic × ECL_pessimistic). Reviewed and updated at least quarterly.
Key implication: Provisions move with the macro outlook — even if no borrower has defaulted. A downgrade in GDP forecasts increases ECL provisions across a performing portfolio. This is the policy intent: genuine counter-cyclical provisioning behaviour.
Governance and Model Risk Management
ECL is not a one-time calculation — it requires a permanent institutional infrastructure for model development, validation, and governance.
Board must approve the ECL policy, SICR definitions, FLI methodology, and model validation framework. A Model Risk Committee or equivalent is required.
An independent internal team (separate from model developers) must validate PD, LGD, EAD models, staging logic, FLI scenarios, and data quality on an ongoing basis.
Independent external validation required every 3 years (or earlier if models are materially changed). Must be conducted by a qualified external party.
Loan-level origination data, 5–10 years of historical defaults and recoveries, collateral valuations, macro time series. Data infrastructure is the biggest implementation challenge for mid-sized banks.
Impact on Bank Capital and Profitability
Tier 2 Capital Treatment
Any excess ECL provision over the regulatory backstop floor can be included in Tier 2 capital, subject to a cap of 1.25% of Credit Risk-Weighted Assets (CRWA). This partially offsets the capital impact of higher provisioning.
P&L Impact
Banks with currently under-provisioned portfolios will face a one-time provision catch-up on Day 1 of ECL adoption. RBI has allowed this to be phased in over a transition period rather than hitting the P&L as a single-quarter shock. Going forward, quarterly provisioning will be sensitive to macro forecast changes — a GDP downgrade will increase ECL even for performing portfolios.
Impact Varies Sharply by Bank Type
| Bank Type | Key Challenge | Readiness |
|---|---|---|
| Large PSU Banks | Higher historical NPAs + weaker data infrastructure | Lower |
| Large Private Banks | Calibration of existing risk systems to RBI ECL specs | Higher |
| Small Finance Banks | Concentrated retail/microfinance books — limited default data history | Medium |
| Foreign Banks | India-specific calibration of globally-deployed IFRS 9 models | Higher |
ECL vs IRAC: Key Differences at a Glance
| Dimension | IRAC (Current) | ECL (New Framework) |
|---|---|---|
| Model Type | Incurred Loss | Expected Loss |
| Provision Trigger | Default already occurred | Deterioration in credit quality |
| NPA Recognition | 90 DPD — hard rule | Stage 3 — 90 DPD + qualitative factors |
| Performing Loan Provision | Fixed rates (0.25–1%) | Model-driven, account-by-account |
| Forward-Looking Inputs | None | Mandatory — macro scenario-weighted |
| Provision Path | Cliff-edged at 90 DPD | Gradual, continuous migration across stages |
| Regulatory Override | Rule-based, deterministic | Model-based with backstop floors |
Implementation Timeline
Banks compute both IRAC and ECL provisions simultaneously. Regulatory capital and reported provisions continue on IRAC basis. ECL figures are disclosed in notes to accounts. Banks that are not ready may be required to extend this phase.
RBI will specify the switchover date based on industry readiness assessment. Banks with inadequate models or data infrastructure will be required to continue the parallel run. Day 1 provision catch-up is phased over the transition period.
ECL becomes the regulatory provisioning standard. IRAC norms are discontinued for covered SCBs. RBI has not published a fixed go-live date — the parallel run start date was tied to the final Master Directions notification.
What Banks Must Do Now
- Conduct a data gap assessment — identify missing loan-level origination data, default history, and recovery timelines
- Form an ECL Steering Committee with Board-level sponsorship and a dedicated project office
- Select or build PD, LGD, EAD models — vendor solutions are viable for smaller banks; large banks will likely build in-house
- Define and document SICR policy — thresholds (quantitative and qualitative) must be Board-approved and auditable
- Design the FLI framework — identify macro variables, scenario construction methodology, and a quarterly review cadence
- Integrate ECL computation with CBS — provisioning must be automated, not a quarterly manual exercise
- Begin parallel run immediately upon readiness — do not wait for regulatory pressure; early parallel runs provide more calibration time
What is the RBI ECL framework?
RBI’s ECL (Expected Credit Loss) framework is a new provisioning system for Indian banks that requires provisions to be based on expected future credit losses — not just losses that have already occurred. It replaces the existing incurred-loss IRAC model with a forward-looking, model-driven approach aligned with IFRS 9. All financial assets must be classified into one of three stages, with ECL provisions computed using PD, LGD, and EAD inputs adjusted for macroeconomic scenarios.
What are Stage 1, Stage 2, and Stage 3 under ECL?
Stage 1 covers performing assets with no significant credit deterioration since origination — provisioned at 12-month ECL (minimum 0.25% floor). Stage 2 covers assets with a Significant Increase in Credit Risk (SICR) but not yet defaulted — provisioned at lifetime ECL (minimum 1% floor). Stage 3 covers credit-impaired or defaulted assets, broadly equivalent to current NPAs — provisioned at lifetime ECL on an NPV basis, with a floor equal to IRAC NPA provisioning rates.
What is SICR under RBI’s ECL framework?
SICR (Significant Increase in Credit Risk) is the trigger that moves a loan from Stage 1 to Stage 2. RBI defines SICR through quantitative triggers — including a rebuttable presumption at 30 DPD and a significant increase in lifetime PD since origination — and qualitative triggers such as restructuring, watchlist placement, rating downgrade, adverse industry outlook, or fraud flags. Banks must define their own SICR thresholds within RBI’s guidance and have them approved by the Board.
How is ECL calculated?
ECL = PD (Probability of Default) × LGD (Loss Given Default) × EAD (Exposure at Default) × Discount Factor. PD is estimated using historical default data and macro variables on a Point-in-Time basis. LGD is estimated on a Downturn basis (stress conditions). EAD is the outstanding balance at default, with a Credit Conversion Factor applied for revolving facilities. The result is further adjusted for probability-weighted macro scenarios (base, optimistic, pessimistic) using Forward-Looking Information.
Which banks must implement RBI’s ECL framework?
All Scheduled Commercial Banks (SCBs) operating in India — including public sector banks, private sector banks, foreign bank branches, and Small Finance Banks (SFBs). Regional Rural Banks (RRBs), co-operative banks, and Local Area Banks are excluded from the initial scope.
How does ECL affect bank capital?
Excess ECL provisions over the regulatory backstop floor can be included in Tier 2 capital, capped at 1.25% of Credit Risk-Weighted Assets. The Day 1 provision catch-up impact (for under-provisioned banks) is allowed to be phased in over a transition period rather than taken as a single-quarter P&L hit. Banks with adequate existing provisioning will see a smaller Day 1 impact.
When will ECL replace IRAC for Indian banks?
RBI mandates a minimum two-year parallel run — during which banks compute both IRAC and ECL provisions simultaneously — before full switchover. The earliest realistic full adoption for most banks is FY 2026–27, subject to RBI issuing the switchover notification. Banks that are not ready at the end of the parallel run may be required to continue running both systems until RBI deems them compliant.