RBI Expected Credit Loss Framework 2026: What Banks and NBFCs Must Prepare For

On 27 April 2026, the Reserve Bank of India issued its final directions on Expected Credit Loss provisioning. In contrast, the ECL framework replaces India’s current incurred-loss model, under which lenders recognise provisions only after a default event, with a forward-looking approach that requires lenders to estimate and provision for expected losses before they occur.

Furthermore, the RBI Expected Credit Loss Framework 2026 represents a major shift in Indian banking regulation since the adoption of prudential norms in the 1990s. The Reserve Bank of India mandates implementation from 1 April 2027, with a four-year phased transition period until March 2031. This guide explains ECL, required lender capabilities, and its credit risk implications.

What the RBI ECL Framework Change and Why

India’s current provisioning model under the Income Recognition, Asset Classification and Provisioning (IRACP) framework is retrospective.

Consequently, this retrospective approach systematically understates credit risk during benign credit cycles. Moreover, when the cycle turns, it produces sharp and procyclical provisioning spikes. The 2015–2019 NBFC and PSU bank NPA crisis illustrated this pattern vividly.

Accordingly, the RBI considers accounts more than 30 days past due as a rebuttable presumption of increased credit risk. Furthermore, lenders can use forward-looking indicators, financial deterioration, sector stress, and covenant breaches to move accounts to Stage 2 before delinquency occurs.

The Three-Stage Classification System Under ECL

The RBI’s ECL framework introduces a three-stage classification that replaces the binary Standard/NPA division:

• Stage 1 (Performing): no significant increase in credit risk since origination. 12-month ECL provisioning covers the expected loss from default events possible within 12 months. This covers the vast majority of a healthy loan book at any given time.
• Stage 2 (Underperforming): significant increase in credit risk since origination, but not yet in default. Lifetime ECL provisions the expected loss over the remaining life of the loan. Accounts that would previously have been classified as SMA-1 or SMA-2 would typically fall here.
• Stage 3 (Credit-Impaired / NPA): default has occurred. Lifetime ECL provisioning continues, but with reduced gross carrying amount by expected credit losses. This is broadly equivalent to the current NPA classification with provisioning.

The critical operational challenge is Stage 2 identification. What constitutes a “significant increase in credit risk”? Accordingly, the RBI provides guidance that more than 30 days past due serves as a rebuttable presumption of a significant increase in credit risk. Furthermore, lenders can also consider forward-looking indicators, deteriorating financial metrics, sector stress, and covenant breaches to move accounts to Stage 2 before actual delinquency occurs.

PD, LGD, and EAD: The Three Model Components Banks Must Build

ECL calculation requires three credit risk models per asset class:

• Probability of Default (PD): the probability that a borrower will default within a specified horizon (12 months for Stage 1; lifetime for Stages 2 and 3). PD models must be calibrated on a minimum of five years of historical default data per asset class.
• Similarly, Loss Given Default (LGD) measures the proportion of exposure that lenders expect to lose after a borrower defaults, while they account for recoveries from collateral, guarantees, and legal proceedings. Moreover, lenders observe significant LGD variations across product types; therefore, they must estimate LGD separately for each asset class, as secured LAP typically delivers higher recovery rates than unsecured personal loans.
• Exposure at Default (EAD): the outstanding exposure at the time of default. For term loans, this is relatively predictable. For revolving credit lines (credit cards, overdrafts), lenders must model how much of the facility borrowers will draw at the point of default to calculate EAD accurately.

Each model must be independently validated and Board-approved. The RBI requires a three-tier model risk management structure: business line development, independent model validation, and internal audit review.

ECL for NBFCs: Ind AS 109 Already Applies: What Changes?

Furthermore, NBFCs in India above the specified threshold have already been applying Ind AS 109, the Indian adaptation of IFRS 9, since FY2019-20. The ECL framework for banks aligns prudential provisioning with this existing approach.

For NBFCs, the key change introduced by the February 2026 IRACP Amendment Directions is: NBFC asset classification and provisioning must now follow the ECL principles of their Ind AS 109 models, with the RBI providing product-wise minimum provisioning floors as regulatory backstops.

NBFCs that have already built robust Ind AS 109 ECL models are better positioned. Those that have implemented Ind AS 109 primarily as a compliance exercise with simplified models and limited forward-looking information will need to upgrade model quality to meet the RBI’s governance and validation expectations.

Data, Governance, and Implementation Requirements

ECL implementation is primarily a data and governance challenge, not just a modelling challenge:

• Historical data: a minimum of five years of clean, consistent, loan-level default and recovery data per asset class. However, many smaller banks and NBFCs continue to face data quality and consistency issues in their legacy core banking systems. Consequently, they must resolve these issues before they can calibrate robust ECL models.
• Forward-looking information: ECL must incorporate macroeconomic forecasts and scenario analysis. This requires a formal process for selecting economic scenarios, attaching probabilities, and documenting the assumptions used.
• Board oversight: the Board must approve the ECL methodology, the model governance framework, and the significant management judgments in Stage 2 identification. ECL governance is not a technical function; it is a Board-level accountability.
• Under the transition provisions, lenders can phase the difference between ECL-based provisions and existing IRACP provisions over four years (FY2027–2031), which limits the immediate capital impact of the transition.

Key Takeaways

• RBI ECL framework 2026 replaces India’s incurred-loss IRACP provisioning with a three-stage forward-looking Expected Credit Loss model. Final directions were issued in April 2026; implementation is from April 2027 with a four-year transition.
• Three-stage classification: Stage 1 (performing, 12-month ECL), Stage 2 (significant credit risk increase, lifetime ECL), Stage 3 (default/NPA, lifetime ECL).
• Three models are required per asset class: PD (probability of default), LGD (loss given default), and EAD
• Data readiness: five years of clean loan-level default and recovery data per asset class is the most time-consuming implementation prerequisite.
• Although NBFCs already operating under Ind AS 109 have a head start, many will still need to upgrade their model quality and governance frameworks to meet the RBI’s validation expectations.

Frequently Asked Questions

Conclusion

RBI ECL framework 2026 is a structural upgrade to how Indian lenders recognise and provision for credit risk. Lenders face a tight one-year runway before April 2027 because they must build the required data, model development, validation, and governance infrastructure for implementation.

Begin now. By contrast, institutions that treat ECL merely as a compliance exercise before the deadline will likely develop lower-quality models, adopt weaker governance practices, and face greater capital uncertainty. Those that treat it as a credit risk management improvement will emerge with better risk intelligence and more resilient portfolios.