Provisioning for Appearances: RBI’s Half-Hearted Tryst with ECL  

Expected Credit Loss provisioning was born out of a crisis. The 2008 Global Financial Crisis revealed the bankruptcy of the old “incurred loss” model, under which provisions were recognised only after loans were visibly impaired. By the time recognition happened, losses had already piled up, balance sheets were overstated, and capital cushions proved illusory.

The answer, regulators decided, was to force institutions to look forward. Under IFRS 9, provisions would no longer wait for losses to materialise. Instead, financial firms would estimate probability of default, exposure at default, and loss given default, overlay macroeconomic scenarios, and create buffers in advance. Foresight, not hindsight, would drive capital.

When India adopted IndAS 109, its homegrown equivalent of IFRS 9, in 2018, large NBFCs and housing finance companies were pushed into the brave new world of ECL. On paper, the move signalled global convergence. In practice, it opened a window into India’s discomfort: the standards may be the same, but the governance regimes that make them work elsewhere are absent here.

Dual Accounting Universe
The result is a peculiar schizophrenia. NBFCs now publish accounts under IndAS, with three-stage ECL models:

• Stage 1: Performing loans, where provisions are based on 12-month expected losses.
• Stage 2: Underperforming loans with a significant increase in credit risk, where lifetime expected losses must be recognised.
• Stage 3: Non-performing loans, where losses are fully provisioned.

But the regulation of 90 days past due method of impairment recognition being still in vogue, it is unclear, how such a strict rule-based approach is being married to a subjective principle based approach of IndAS. More importantly, an occasion to hone subjective supervisory assessment across multiple institutions has been lost.

Investors see ECL numbers. Supervisors use IRAC. The coexistence of two provisioning regimes says it all: the RBI is not comfortable with ECL enough to anchor prudential capital.

Deferral for Banks

The story gets stranger when we turn to banks. Unlike NBFCs, which were forced into IndAS-ECL, banks were spared. RBI repeatedly deferred the implementation of IndAS for banks, citing preparedness, systems, and the need for legislative changes. Today, banks still follow the old Indian GAAP and incurred loss model for provisioning.

That asymmetry raises obvious questions. If ECL is the superior system, why haven’t banks, the dominant players in India’s credit market, adopted it? Why impose the more sophisticated, judgment-heavy model on NBFCs, while leaving the systemically critical banking sector under simpler rules?

The answer lies, probably, not in technicalities but in trust. RBI is aware that ECL relies on strong governance regimes: independent auditors, disciplined boards, validated models, reliable macro forecasts, and supervisors who can challenge assumptions. India’s governance fabric does not yet inspire that confidence. Better to stick with the conservative, mechanical IRAC norms for banks than risk procyclical chaos through untested models.

In effect, NBFCs became guinea pigs for global accounting optics, while banks were shielded for reasons of systemic prudence.

This half-hearted approach, when it comes to implementation of ECL for banks have also been apparent in other forms. Banks have been submitting proforma P&L at least for the 5-6 years. But in the absence of standardisation of some treatment across banks, such proforma P&L based on banks’ own judgement are often not as useful as they could have become, in honing supervisory judgements

Theory vs Practice
In theory, ECL promises statistical sophistication. In practice, Indian NBFCs often lack the depth of data required to estimate probability of default, loss given default, and exposure at default. Many specialise in retail or SME lending, where borrower histories are short and volatile.

Instead of robust models, Excel sheets rule. Sceptics feel parameters are guessed, not validated. Where the numbers look inadequate, “management overlays”, euphemism for executive (mis)judgment calls, paper over the gaps. The result is provisioning that looks advanced but rests on shaky foundations.

Auditors, who are supposed to interrogate assumptions, rarely do so. Unlike in Europe, where supervisors and audit committees stress-test models, Indian auditors often are said to defer to management. The Franklin Templeton debt fund fiasco in 2020 showed how credit risk could be consistently underestimated until liquidity evaporated and funds were forced to shut down six schemes.

The problem is not the standard. It is the absence of a governance ecosystem that makes the standard credible.

Nowhere is the gap starker than in data. ECL depends on reliable borrower-level histories: defaults, recoveries, exposures. RBI itself once floated the idea of a Public Credit Registry — a comprehensive borrower-level database to consolidate credit information scattered across CIBIL, CRILC, and other silos.

Had PCR materialised, NBFCs and banks would have had a common backbone for ECL models. Estimates for probability of default and loss given default could have been grounded in hard evidence rather than guesswork. Supervisors, too, would have gained the ability to benchmark and validate assumptions across institutions.

Instead, the project got entangled in politics with alleged turf battles between RBI, government agencies, and private credit bureaus stalling it. Concerns over privacy and control turned the PCR into a contested policy football. The result: India’s credit system still operates with fragmented, inconsistent datasets, forcing lenders to improvise models on shaky foundations.

In effect, India imported a sophisticated provisioning model without building the data infrastructure that makes it work.

Pragmatism or Half Measures?
RBI’s reluctance to rely on ECL is, in one sense, pragmatic. Even globally, critics argue that IFRS 9 ECL is procyclical: in downturns, models force institutions to increase provisions just when credit is most needed, aggravating recessions.

But in India, with weaker governance and more volatile cycles, the risks are greater. By retaining IRAC norms as the prudential anchor, RBI has chosen safety over theory.

Yet this pragmatism has created a muddle. NBFCs must comply with dual rules. Investors see disclosures that regulators don’t trust. Banks remain exempt altogether. The end result is not prudence but opacity.

If RBI believes ECL is untrustworthy, why require it at all? And if it believes in the global framework, why not integrate it fully into prudential norms? It can be argued that steps such as gradually increasing provisions for impaired loans from 15–25% to 40–50%, eventually approaching ECL provisioning, would promote early resolution and better recoveries and would largely capture the positive externalities of ECL even without formally adopting the same.

The deeper truth is that India’s adoption of IndAS was motivated less by domestic prudence than by international signalling. The goal was to reassure rating agencies and foreign investors that India’s financial reporting was “globally aligned.” On paper, IndAS 109 mirrors IFRS 9 almost word for word.

But substance matters more than form. In Europe, regulators actively interrogate ECL models. In India, RBI has stepped back from trusting them. That contrast reveals the hollowness of cosmetic globalisation: global accounting for appearances, domestic regulation for reality.

Governance, Not Just Models
At its heart, the trustworthiness of ECL depends not on the elegance of its mathematics but on the strength of governance.

• Boards must be independent and sceptical, not cheerleaders for management optimism or errand runners to the regulator or the ‘perceived’ proteges of the regulators
• Auditors must challenge assumptions, not rubber-stamp them, given the fact that they can’t escape the flak when things turn out awry
• Supervisors must have both the capacity and the will to question models, not just receive reports.
• Data systems must be deep and reliable, not fragmented across institutions. Most important, collected data should be utilised efficiently by the supervisors

India is still some distance away on all four fronts. The discomfort with ECL is not about the model. It is about the institutions that underpin it.

If India wants ECL to work, three shifts are unavoidable:

1. Data Infrastructure: RBI should spearhead a national credit loss database that captures defaults, recoveries, and exposures across the system. Without this, estimates for probability of default and loss given default will remain guesswork.
2. Audit and Governance Reform: Auditors and boards must be held accountable for validating models and challenging overlays. Supervisory expectations should be explicit and enforceable.
3. Regulatory Integration: RBI must make a choice. Either integrate ECL into prudential norms, as the European Central Bank has done, or abandon the pretence. Running parallel universes only undermines credibility.

ECL under IndAS is, in form, equivalent to IFRS 9. But in substance, it is a pale imitation. For NBFCs, implementation has been uneven, with Excel models standing in for robust systems. For banks, RBI has avoided ECL altogether, preferring the comfort of old prudential norms.

The irony is stark: India’s credit system has global standards on paper but not in practice. Until governance strengthens, ECL will remain what it has been so far — provisioning for appearances, not provisioning for reality. And a less than professional implementation of ECL actually penalises banks, that have honestly made efforts and have sunk money in this regard. They also make balance sheet and P&L numbers incomparable across banks and across time.