Every risk professional in Indian banking eventually asks the same question: is credit risk actually improving, or does it just look that way in aggregate numbers? Based on the Reserve Bank of India’s own published data, the answer is both. System-wide asset quality has genuinely strengthened over the past five years. But the composition of that risk is shifting in a direction that deserves closer attention.
This piece works through RBI’s Financial Stability Reports (FSR), sectoral credit data, and the newly finalized Expected Credit Loss (ECL) framework. Together they show what’s really happening with credit risk in Indian banking between 2020 and 2025. It does not rely on a proprietary survey or projected estimates dressed up as findings. In fact, every figure below is sourced directly to a named RBI report. That distinction matters: in a domain where regulators, auditors, and rating agencies check your numbers, credibility is the entire product.
Three questions structure the analysis. How has aggregate asset quality moved since the 2020 pandemic shock? Where is risk concentrating today, even as headline numbers improve? And what does the incoming ECL regime signal about how Indian banks will need to manage credit risk going forward?
This analysis draws on primary RBI publications, cross-checked across multiple reporting periods for consistency.
RBI Financial Stability Reports (FSR): Published twice yearly. They consolidate Gross NPA (GNPA) and Net NPA (NNPA) ratios of Scheduled Commercial Banks (SCBs), capital adequacy (CRAR), bank-group-wise asset quality, and stress test results. Specifically, this piece uses FSR editions from January 2021 through December 2025.
RBI Sectoral Deployment of Bank Credit data: Monthly data on credit growth to industry, services, agriculture, and personal loans. It’s sourced from 41 banks representing roughly 95% of non-food credit.
RBI’s ECL framework releases: The draft ECL directions (October 2025) and final directions (April 2026). These describe the shift from incurred-loss to forward-looking PD/LGD/EAD-based provisioning, effective April 1, 2027.
One scope note: RBI does not publish a standardized “default rate by loan product” table. Where this article cites loan-category or bank-group figures, they are GNPA ratios: the share of gross advances classified as non-performing. It’s the metric RBI itself uses, and the one directly comparable across periods.
| Period | GNPA Ratio | NNPA Ratio | Source |
| March 2020 | 8.4% | — | RBI FSR, Jan 2021 |
| September 2020 | 7.5% | — | RBI FSR, Jan 2021 |
| March 2024 | 2.8% | 0.6% | RBI FSR, Jun 2024 |
| March 2025 | 2.3% | 0.5% | RBI FSR, Jun 2025 |
| September 2025 | 2.1%–2.2% | — | RBI FSR, Dec 2025 |
| March 2027 (projected, baseline) | 1.9% | — | RBI FSR, Dec 2025 |
RBI’s January 2021 report recorded a September 2020 GNPA ratio of 7.5%, down from 8.4% in March 2020. That was a system still absorbing the pandemic shock. GNPA had fallen to 2.8% by June 2024, then to 2.3% by March 2025. It touched a multi-decade low of 2.1% by September 2025, and RBI projects further improvement to 1.9% by March 2027 under its baseline scenario.
In practice, this reflects five years of balance sheet cleanup: post-IBC resolution of legacy corporate stress, tighter underwriting after the 2018–2020 NBFC stress episode, and stronger capital buffers overall. Meanwhile, system-wide CRAR remains comfortably above regulatory minimums, with public sector banks at 16% and private banks at 18.1% as of September 2025.
In short, aggregate GNPA is a lagging confirmation of underwriting discipline, not a leading indicator. A PD model trained mainly on 2020–2022 stressed data will overstate current default risk. One trained only on 2023–2025 benign data risks understating tail risk in the next downturn.
Explore our Credit Risk Modeling Certification Training for a structured approach to PD estimation across credit cycles.
For instance, PSB GNPA fell sharply from 3.7% in March 2024 to 2.8% in March 2025. Meanwhile, private bank GNPA held roughly stable at 2.8% over the same period, and foreign banks improved from 1.2% to 0.9%.
Even so, this convergence matters. For most of the post-2015 asset-quality-review era, PSB asset quality lagged private banks significantly, largely on corporate exposures. That gap has now nearly closed at the aggregate level. However, remaining risk differs by bank group, which leads to the more consequential finding below.
This is the most important finding for practitioners, because it sits underneath the reassuring headline number. According to RBI’s December 2025 FSR, roughly 53.1% of retail loan slippages now originate from unsecured products like personal loans and credit cards. At private banks, unsecured loans account for nearly 76% of fresh slippages. GNPA on unsecured retail loans stood at 1.8%, versus 1.1% for overall retail advances.
In other words, the 2.1% aggregate GNPA figure blends a very clean secured/corporate book with a smaller, faster-deteriorating unsecured retail book. RBI flagged this as a fintech-adjacent risk, tied to fast credit growth in small-ticket personal loans to borrowers under 35 through digital lending channels.
This pattern, in fact, tracks with operational experience. Unsecured lending has weaker recovery mechanics (no collateral to liquidate, higher LGD), shorter behavioral history on new-to-credit borrowers, and faster origination cycles that compress underwriting review. Moreover, it is the segment where forward-looking provisioning matters most, since unsecured risk builds up quietly between formal NPA recognition points.
As a result, portfolio-level GNPA alone is no longer sufficient. Overall, segment-level GNPA and vintage curves for unsecured retail belong alongside the aggregate number in any board-level risk dashboard.
RBI has issued directions introducing forward-looking ECL provisioning, replacing the incurred-loss model. It takes effect April 1, 2027, for scheduled commercial banks excluding RRBs, Small Finance Banks, and payments banks. ECL provisioning must be based on a bank’s own historical PD and LGD data spanning at least five years, subject to RBI-specified floors. Accounts 30–90 days past due move into Stage 2, a materially earlier trigger than the current framework.
Overall, the shift aligns India’s prudential norms with global IFRS 9 standards. In addition, it requires closer integration between finance and risk functions, as forward-looking macroeconomic scenarios become a formal input to provisioning.
Indeed, this is a direct regulatory response to Finding 3. An incurred-loss model recognizes impairment only after default has effectively occurred. ECL requires estimating expected loss, via PD, LGD, and EAD, well before that point, catching unsecured deterioration earlier in the cycle.
Even so, for banks building this capability, it isn’t a compliance task to fully outsource. RBI has explicitly made a bank’s board and senior management responsible for the adequacy of the ECL framework. Consequently, internal teams need working fluency in PD/LGD/EAD construction, not just the ability to read vendor output. However, it’s worth noting that the standard formula, Expected Loss = PD × LGD × EAD, assumes independence between the three components. In practice they’re correlated: LGD tends to rise in the same downturns that push PD higher. That’s why RBI’s stress tests apply adverse scenarios jointly rather than multiplying baseline figures in isolation.
Explore our Credit Risk Modeling Certification Training to build PD, LGD, and EAD modeling skills ahead of the 2027 ECL transition, or see Understanding Credit Risk: Definition and Types for foundational concepts referenced throughout.
What is the current GNPA ratio of Indian banks?
As of September 2025, SCB GNPA stood at 2.1%, a multi-decade low, per RBI’s December 2025 Financial Stability Report.
Is unsecured lending riskier than secured lending right now?
Yes, and the gap is widening. Unsecured retail GNPA was 1.8% versus 1.1% for overall retail advances, and unsecured products drove over half of all retail slippages.
When does RBI’s ECL framework take effect?
RBI’s ECL Directions were issued 27 April 2026 and take effect April 1, 2027. They apply to commercial banks, excluding small finance banks, payments banks, and local area banks.
Does EL = PD × LGD × EAD fully capture expected loss?
It’s the standard starting formula, but it assumes PD, LGD, and EAD move independently. In stress, they’re correlated — which is why RBI applies adverse scenarios jointly rather than multiplying baseline values.
The data supports a measured conclusion, not a triumphant one. Indeed, Indian banking’s asset quality genuinely improved for five straight years, and RBI’s own numbers back that up without embellishment. However, the same data shows risk isn’t disappearing. Instead, it’s relocating toward unsecured retail lending, addressed through a regulatory shift that will demand more rigorous PD, LGD, and EAD modeling capability than most institutions currently have in-house. For risk analysts, credit officers, and model validators, that combination is telling: improving headline numbers alongside a harder compliance mandate. It’s exactly why 2025–2027 is a build-capability window, not a wait-and-see one.
This analysis is based on RBI’s Financial Stability Reports, Sectoral Deployment of Bank Credit data, and RBI’s ECL Directions (2025–2026). Figures are reported as published at the cited dates; readers should consult original RBI releases for the most current data.
Understanding the theory behind PD, LGD, and EAD is the first step. Building bankable, interview-ready models — in Python or SAS, on real credit datasets, aligned to Basel and IFRS 9 — is what actually moves a career forward.
Explore Dexlab Analytics’ Credit Risk Modeling certification program to build PD, LGD, and EAD models from scratch, work through IFRS 9 ECL frameworks, and learn model validation techniques used by practicing risk teams.
.
Credit and Market Risk, Credit Risk, credit risk analysis, Credit Risk Analytics And Modeling, credit risk analytics training, Credit Risk Modelling, Credit Risk Modelling Using SAS, enterprise risk management certification, free demo session, online certification, online courses, Predictive Analytics, predictive analytics certification, predictive modeling, Predictive Modelling, Risk Analytics Market, Risk Management Courses, risk management courses online, Risk Management in Banking
Comments are closed here.