Improving NPAs of Indian Banks: Genuine Reduction or Cosmetic Cleansing?

India’s banking sector today presents a paradox. Official data show that gross non‑performing assets (NPAs) have fallen to multi‑decade lows, yet the journey from a double‑digit NPA crisis to today’s comfortable ratios has relied heavily on large‑scale write‑offs, selective restructuring and limited transfers to Asset Reconstruction Companies (ARCs), alongside genuine improvements in asset quality and recoveries. The key question is not whether NPAs have come down—they clearly have—but how much of this reduction reflects real economic healing versus accounting‑driven clean‑up supported by capital and regulatory tools.

Success story: A dramatic fall in NPAs

From the regulator’s vantage point, the NPA story is one of successful recognition and resolution followed by structural strengthening. Gross NPAs of public sector banks (PSBs) peaked at about ₹8.96 lakh crore in March 2018, with a GNPA ratio of 14.6% for PSBs and around 9–11% for the banking system as a whole, after the Asset Quality Review (AQR) forced banks to recognise legacy stress that had been masked by restructuring and evergreening. By March 2025, PSB gross NPAs had dropped to roughly ₹2.84 lakh crore and a GNPA ratio of 2.58%, while system‑wide estimates place the GNPA ratio near 2.1–2.3%, with net NPAs around 0.5%, the lowest in more than two decades.

The government and RBI explain this turnaround through the “4R” strategy—Recognition, Resolution, Recapitalisation and Reforms. Recognition came via AQR and stricter income‑recognition norms; resolution was driven by the Insolvency and Bankruptcy Code (IBC), strengthened SARFAESI enforcement and Debt Recovery Tribunals (DRTs); recapitalisation injected capital into PSBs to absorb losses; and reforms ranged from governance changes and PSB consolidation to tighter risk management and supervision. RBI’s Financial Stability narrative further highlights declining slippage ratios, higher provision coverage and improved profitability as evidence that the balance sheets of Indian banks are structurally stronger, not just optically cleaner. There is merit in the narrative, but at the same time, the full story is not explained in this. We need to deep dive into the arithmetic.

The arithmetic of clean‑up: write‑offs take centrestage

Beneath the headline ratios, the arithmetic of NPA reduction is dominated by loan write‑offs. Public compilations of RBI and government disclosures show that banks wrote off close to ₹8.9 lakh crore of NPAs in just five years and about ₹10.4 lakh crore over roughly nine years, while cumulative recoveries on these written‑off accounts remained below 20% of the amounts written off. RTI‑based estimates suggest that the effective annual recovery rate on written‑off loans is around 1.6% per year, underscoring how slowly “technical” NPAs convert into actual inflows once removed from the balance sheet.

RBI and market commentary acknowledge that write‑offs have been a “major component” of NPA reduction over the last five years, even as recoveries and upgradations have improved. This means the clean‑up has been real in the sense that losses are recognised, provisioned and absorbed by capital, but cosmetic in that headline GNPA ratios have benefited significantly from removing problem assets from the NPA line via write‑offs rather than from high cash recovery ( fully resolved NPAs).

Role of Resolution channels: recoveries, IBC and ARCs

Alongside write‑offs, India has developed a multi‑channel resolution architecture—IBC, SARFAESI, DRTs and ARCs—that contributes to genuine NPA reduction. RBI and Insolvency and Bankruptcy Board data show that average recovery rates under IBC for large corporate cases have generally exceeded those under earlier mechanisms, even though timelines and haircuts remain challenging. In recent years, a rising share of the annual reduction in NPAs has come from recoveries and upgradations, with some analyses linked to RBI data noting that around 40–45% of GNPA reduction in the latest years reflects actual cash recovery or resumption of regular repayment, not just write‑offs.

However, ARCs—often seen as a key instrument of clean‑up—have played a much smaller quantitative role than is sometimes assumed. At their peak, loans taken over by ARCs were in the low‑lakh‑crore range versus nearly ₹10 lakh crore of NPAs still on bank books, implying that only a fraction of stressed assets migrated to these specialised vehicles. RBI’s evolving regulations, including its Master Directions for ARCs and higher “skin‑in‑the‑game” requirements, have nudged ARCs toward a more resolution‑centric and accountable role, but capital constraints and pricing gaps limit how much they can absorb in any given year; empirical estimates suggest ARC sales typically account for only a single‑digit to low‑teens percentage of annual NPA reduction. Therefore transferring bad loans to ARC books is mere shifting of the bad loan problem, not resolution.

Restructuring and forbearance: shifting, not eliminating, stress

A further layer in the story is restructuring and targeted forbearance, especially during Covid‑19. RBI’s Resolution Frameworks allowed banks to modify repayment terms for viable but stressed borrowers without immediate NPA classification, while schemes like the Emergency Credit Line Guarantee Scheme (ECLGS) provided guaranteed additional funding to MSMEs and other vulnerable sectors. Policy data indicate that ECLGS‑linked support covered around 12% of MSME credit at its peak, preventing a large wave of potential NPAs by temporarily cushioning borrowers through guarantees and moratoria.

These measures clearly contained systemic risk and preserved viable enterprises, but they also re‑categorised part of the stress—from what could have become NPAs into restructured standard assets. While economically rational in the face of an exogenous shock, such restructuring complicates any simple reading of low GNPA ratios as a pure indicator of underlying credit strength. The true health of portfolios depends on how restructured loans perform over time and how quickly temporary forbearance is unwound, which RBI has signalled it will monitor closely through enhanced disclosures and supervisory reviews.

Genuine reduction or cosmetic cleansing : Somewhere in the middle of the Spectrum

Putting these strands together, a balanced interpretation emerges. On the one hand, multi‑decade‑low GNPA and net NPA ratios, falling slippages, higher provision coverage and improved profitability strongly suggest that India’s banking sector is far healthier than a decade ago. AQR forced recognition, IBC and SARFAESI improved recovery prospects, and governance and risk‑management reforms have reduced the flow of new NPAs, particularly in large corporate portfolios.

On the other hand, the scale and composition of the clean‑up matter. Over the last 5–9 years, more than half of the reduction in reported GNPA stocks can be attributed to write‑offs, with ARC sales and restructuring adding another meaningful share, and the remainder driven by recoveries and upgradations. This means headline NPA ratios partly reflect loss absorption and balance‑sheet purging rather than a full economic recovery of bad loans. The official position rightly emphasises recognition, provisioning and resilience; a more critical reading would add that sustaining these gains requires continued discipline in underwriting, monitoring and governance, so that the next credit cycle does not rebuild a similar stock of stressed assets that must again be cleansed at high fiscal and economic cost.

Moving forward: how IFRS 9‑style ECL can help

A forward‑looking IFRS 9 / Ind AS 109 expected credit loss (ECL) framework can make India’s NPA story more genuinely preventive than retro‑actively cosmetic by forcing earlier recognition of credit deterioration, higher through‑the‑cycle provisions and tighter alignment between risk management and accounting. Under current prudential norms for many banks, provisions are still largely rule‑based and backward‑looking: losses are provided once assets cross defined ageing or impairment thresholds, which makes provisioning “too little, too late” and encourages evergreening or regulatory forbearance until stress becomes undeniable.

By contrast, the proposed ECL framework aligned with IFRS 9 / Ind AS 109 requires banks to estimate probability of default (PD), loss given default (LGD) and exposure at default (EAD) across the life of the exposure and hold allowances from day one, escalating as credit risk increases. RBI’s discussion paper and subsequent draft norms outline a three‑stage structure: Stage 1 exposures attract 12‑month ECL, while Stage 2 (significant increase in credit risk) and Stage 3 (credit‑impaired/NPAs) exposures attract lifetime ECL, with regulatory floors for PD, LGD and provisioning percentages to limit model risk and under‑estimation. Banks will be allowed to design internal models, but must subject them to independent validation, back‑testing and strong board‑level oversight, and RBI has proposed a phased transition so that implementation—currently signalled over the next few years—does not destabilise capital positions.

For India’s post‑clean‑up landscape, this shift has three critical benefits. First, it brings forward loss recognition, reducing the incentive to postpone identification of stress and the need for sudden, massive write‑off waves later in the cycle; provisions will rise automatically as early‑warning indicators and macro‑forecasts point to higher expected defaults. Second, it should smoothen the credit cycle, as banks will build buffers in good times based on non‑zero PDs even when realised defaults are low, making them less likely to cut back lending abruptly when losses materialise. Third, by forcing tighter integration of risk models, data infrastructure and governance around ECL, it aligns supervisory expectations, internal risk appetite and front‑line underwriting, making the next generation of NPA ratios more reflective of genuine resilience than of delayed recognition and back‑ended purging.

Annex: Glossary of key terms

Asset Quality Review (AQR)
A targeted supervisory exercise launched by the RBI in 2015–16 to identify stressed loans that had been restructured or evergreened, and to enforce uniform classification of such exposures as NPAs where warranted.pib+1

Asset Reconstruction Company (ARC)
A specialised financial institution registered with RBI that acquires NPAs from banks and financial institutions, typically at a discount, and seeks to resolve them through restructuring, settlement or sale, often financing the purchase by issuing Security Receipts (SRs) to qualified investors.arcindia+1

Debt Recovery Tribunal (DRT)
A statutory tribunal established under Indian law to facilitate expeditious adjudication and recovery of debts due to banks and financial institutions, especially for secured loans, working alongside SARFAESI‑based enforcement.ili+1

Expected Credit Loss (ECL)
A forward‑looking measure of credit loss under IFRS 9 / Ind AS 109 that combines probability of default (PD), loss given default (LGD) and exposure at default (EAD) over a specified horizon (12‑month or lifetime), requiring earlier and more dynamic provisioning than incurred‑loss models.actuariesindia+1

Gross Non‑Performing Assets (GNPA)
The total value of loans for which interest or principal is overdue beyond a regulatory threshold (typically 90 days), before adjusting for provisions; often expressed as a ratio of total advances to track overall asset‑quality stress.ceicdata+1

Ind AS 109 / IFRS 9
Accounting standards on financial instruments that introduce an expected credit loss model for impairment, classify financial assets based on business model and cash‑flow characteristics, and require extensive credit‑risk disclosures.pwc+1

Net Non‑Performing Assets (NNPA)
Gross NPAs minus provisions held against those NPAs; the residual credit risk that remains on a bank’s balance sheet after recognising expected losses through provisioning.statista+1

Provision Coverage Ratio (PCR)
The ratio of provisions held against NPAs to the total NPAs, indicating the extent to which credit losses are already recognised in the profit and loss account; higher PCR implies stronger buffers and lower residual loss risk.rbi+1

Restructured Standard Assets
Loans whose original repayment terms have been modified (for example, extended tenure, reduced instalments or moratoria) due to borrower stress, but which are still classified as standard under specific regulatory frameworks rather than as NPAs.rbi+1

Security Receipts (SRs)
Instruments issued by an ARC to qualified buyers (often including the selling bank) representing an undivided right or interest in the underlying financial assets acquired from the originator; redeemed over time from recoveries or resolution proceeds.arcindia+1

Slippage Ratio
The proportion of new NPAs (accounts turning non‑performing during a period) to the standard advances at the beginning of that period; a key flow‑based indicator of emerging asset‑quality stress.currentaffairs.adda247+1

SUNANDO ROY – On Banking, Finance and Society

Like this:

Discover more from SUNANDO ROY – On Banking, Finance and Society

Leave a ReplyCancel reply

My LinkedIn Posts

Repo Markets Under Stress: Navigating Liquidity Crises and Building Institutional Readiness

Strategic Liquidity Readiness: Embedding Usable HQLA at the Core of the Contingency Funding Plan

Your cart (items: 0)