Regulating the regulators: Assessing regulation-making frameworks in India's financial sector

by Natasha Aggarwal and Renuka Sane.

Indian regulators, like the Securities and Exchange Board of India (SEBI) and the Reserve Bank of India (RBI), routinely wield quasi-legislative powers. For example, Section 30 of the Securities and Exchange Board of India Act, 1992 empowers the SEBI Board to make regulations. According to its 2024-25 annual report, SEBI issued 104 consultation papers, 61 amendments to its regulations, one new set of regulations, 14 master circulars, and 154 "policy measures." Such regulatory interventions can significantly influence markets and affect economic outcomes. Yet the processes by which regulators design, consult on, and review delegated legislation have been fragmented and, in large part, left to each regulator's discretion.

Traditional safeguards on delegated legislation in India, i.e., parent statutes requiring regulations to be laid before Parliament, prior publication requirements under the General Clauses Act, 1897, and judicial review, provide important accountability functions, but do not regulate the internal process by which a regulator formulates its regulations. They do not require a regulator to identify the problem warranting intervention, weigh alternatives, or assess costs and benefits.

In 2013, the Financial Sector Legislative Reforms Commission proposed provisions to govern regulation-making and, through the Financial Sector Development Council Resolution of 24 October 2013, financial sector regulators agreed to comply with these procedures. Since then, six financial sector regulators (the Insurance Regulatory and Development Authority of India (IRDAI), Insolvency and Bankruptcy Board of India (IBBI), International Financial Services Centres Authority (IFSCA), Pension Fund Regulatory and Development Authority (PFRDA), SEBI and RBI) have each adopted some form of instrument governing how they make regulations, ranging from non-binding concept notes to regulations. More recently, the Economic Survey (2024-25) recommended strengthening regulatory impact assessment; the Securities Markets Code, 2025 proposes statutorily mandating public consultation and periodic review at SEBI; and in March 2026 the Standing Committee on Finance recommended a mandatory regulatory impact assessment framework for the IBBI.

In this backdrop, our paper, 'Regulating the regulators: Assessing regulation-making frameworks in India's financial sector', evaluates the regulation-making frameworks adopted by these six regulators against three principles of good regulation-making - consultation, evidence-based regulation-making, and periodic review - and then assesses a randomly selected 2025 consultation paper issued by each regulator against indicators derived from these principles and from each regulator's own framework.

We find that while all six financial regulators have adopted some form of instrument, these instruments vary considerably in legal form, substantive scope, and analytical ambition. Regulators operating under more demanding frameworks are more likely to clearly identify the regulatory problem in their consultation documents. However, this relationship is not linear: stronger frameworks do not consistently produce stronger performance on more analytically demanding requirements. No regulator, including those whose frameworks expressly require it, included a cost-benefit analysis in its consultation paper, and no regulator assessed available alternatives to direct regulation. Every regulator failed to comply with at least one of its own procedural requirements.

Indian legal frameworks

From 2016 onwards, Indian regulators have progressively formalised how they make their own regulations: IRDAI led the way with a concept note in 2016, followed by IBBI in 2018, IFSCA in 2021, and PFRDA in 2024. In 2025, IFSCA issued an updated and expanded framework for making regulations and subsidiary instructions, SEBI adopted regulations for making, amending, and reviewing regulations, and RBI opted for a non-binding policy framework rather than enforceable regulations.

While all regulators now subject regulation-making to some framework, they diverge along two axes: legal form and substantive scope. On legal form, IBBI, PFRDA, IFSCA, and SEBI have adopted regulations, signalling a commitment to enforceable constraints; RBI and IRDAI, by contrast, have adopted non-binding approaches, suggesting either a desire to retain discretion or a reluctance to subject internal processes to enforceable standards. On scope, PFRDA's framework is narrowly confined to the making of regulations; IBBI, SEBI and IRDAI expand this to include amendments; IFSCA moves further by bringing "subsidiary instructions" within its fold; and RBI adopts the broadest scope, extending its framework to directions, guidelines, notifications, and other instruments. In this context, IFSCA stands out as the strongest: it uses binding regulations, rather than non-binding frameworks, to govern its regulation-making process, and their applicability extends beyond regulations and amendments to subsidiary instructions.

On consultation specifically, most regulators (IRDAI, IFSCA, SEBI, RBI and IBBI) make public consultation mandatory, typically for a minimum of 21 days; PFRDA alone makes it optional, albeit with a longer 30-day window. IRDAI, IFSCA, IBBI and PFRDA publish stakeholder comments and provide responses to them, while SEBI and RBI provide responses but do not publish comments, limiting external visibility into the range of views considered.

On evidence-based regulation-making, the picture is fragmented: IRDAI and IFSCA require both a problem statement and a statement of regulatory intent; SEBI requires only regulatory intent; PFRDA and IBBI require a problem statement but not regulatory intent. Only RBI's framework requires an impact assessment, and only PFRDA and IBBI mandate cost-benefit analysis.

On periodic review, IBBI has the most frequent cycle (three years), followed by IFSCA (five years) and RBI (five to seven years); SEBI and PFRDA require review but specify no timeline, and IRDAI's concept note is silent on review altogether.

Evaluation of consultation papers

We evaluated one randomly selected 2025 consultation paper (for IRDAI, an exposure draft) issued by each regulator, against indicators drawn from the principles of good regulation-making and from each regulator's own framework. Notably, the RBI did not issue a formal consultation paper in the relevant period; the document evaluated for RBI is a circular proposing amendments to its directions, reflecting a broader pattern of the RBI using directions and circulars to make substantive regulatory changes.

The IRDAI Exposure Draft states the objective of its proposal and describes the key features of the framework, but does not clearly explain the problem it seeks to address, does not consider alternative approaches to regulation, and does not include a cost-benefit or impact analysis.

The IBBI Discussion Paper, for each of its three proposals, includes a statement of the problem, a proposed solution, and the draft regulation, but does not identify and assess available alternatives to direct regulation, does not include a cost-benefit analysis, and does not comply with the IBBI Regulations' requirement of an economic analysis, guidance from international standard-setting bodies, or the statutory provision enabling the proposed regulations.

The IFSCA Consultation Paper does not comply with any of the principles of good regulation-making, other than relying on market data as evidence of growth; it refers to fund management entities facing unspecified "operational hassles" without elaborating on what these are, and does not specify the statutory provision enabling the amendments or include guidance from international standard-setting bodies, both required under its own regulations.

The PFRDA Consultation Paper performs comparatively better: it identifies the problem to be addressed, assesses how existing frameworks contribute to the problem, and relies on evidence. However, it does not identify and assess available alternatives to direct regulation or include a cost-benefit analysis, and does not comply with several of its own regulations; it does not specify the statutory provision enabling the proposed regulations, attach a draft of the proposed regulations, include the required economic analysis, include guidance from international standard-setting bodies, or specify the manner of implementation.

The SEBI Consultation Paper does not comply with any of the principles of good regulation-making other than identifying the problem to be addressed; it does not assess how existing regulations contribute to the problem, identify alternatives, or include a cost-benefit analysis, an outcome that closely mirrors the design of SEBI's own framework, which requires only a statement of regulatory intent.

The RBI Circular likewise does not comply with any of the principles other than a rather broad articulation of the problem, and does not comply with the RBI Policy because it does not specify the statutory provision enabling the proposed regulations, or include an impact analysis or guidance from international standard-setting bodies.

Analysis

The results reveal a gap between the formal existence of regulation-making frameworks and their actual operationalisation in consultation documents. There are failures at two levels: compliance with general principles of good regulation-making and compliance with each regulator's own procedural requirements.

More broadly, regulators operating under more developed procedural frameworks, particularly IBBI and PFRDA, which explicitly require problem identification, perform better on basic problem-definition indicators. Both clearly identify the regulatory problem, and PFRDA goes further by examining whether existing regulations contribute to it. In contrast, SEBI and RBI, whose frameworks impose minimal analytical obligations, produce consultation documents that are largely limited to statements of regulatory intent, with little substantive justification. Second, there is inconsistent articulation of the regulatory problem, even at a basic level. While some regulators - such as IBBI, PFRDA, SEBI, and RBI - identify a problem, others (notably IRDAI and IFSCA) fail to do so clearly. Even where a problem is identified, it is often thinly specified and not linked to evidence or to failures in the existing regulatory framework.

However, stronger frameworks do not necessarily translate into stronger performance on more analytically demanding requirements. Despite formal mandates, both IBBI and PFRDA fail to include the economic analysis required by their own regulations. IBBI omits cost benefit analysis, while PFRDA fails to provide economic analysis, draft regulations, international benchmarking, and implementation details. No regulator assesses alternatives to direct regulation or conducts a cost benefit analysis. The absence is uniform and not explained by framework design alone: even regulators whose own frameworks require economic analysis (IBBI, PFRDA, and RBI) fail to provide it. Their universal absence suggests that regulators do not treat regulation as one option among many, but as the default response. As a result, consultation processes are narrowed: stakeholders are invited to comment on how to regulate, but not whether regulation is justified in the first place. This significantly weakens accountability and the quality of regulation-making.

Moreover, every regulator, without exception, fails to comply with at least one of its own procedural requirements. The most consistent gap is the failure to specify the statutory provision enabling the proposed regulation - a basic transparency requirement met only by IRDAI. This omission raises concerns about the legal legitimacy of the proposed regulation, as stakeholders are not informed of the source of regulatory authority. IBBI omits the economic analysis mandated by its framework. PFRDA fails to include draft regulations, economic analysis, implementation guidance, and international benchmarks. IFSCA omits both the problem statement and international benchmarks required under its framework. RBI, similarly, does not provide the impact analysis or international benchmarking contemplated by its policy. These are not merely formal deficiencies. The absence of draft regulatory text, as in the case of PFRDA, prevents stakeholders from engaging with the legal substance of the proposal, limiting consultation to broad regulatory intent. The absence of economic or impact analysis means that the regulatory choice cannot be independently assessed.

Across all six regulators, consultation papers ostensibly function as instruments for presenting pre-determined regulatory proposals, rather than as vehicles for reasoned, evidence-based decision-making.

Reforms

We propose five reforms: (i) legislative amendments to parent statutes that clearly define the scope of regulators' quasi-legislative powers and the processes governing their exercise; (ii) regulatory impact assessment should be made mandatory and comprehensive; all consultation papers should be required to identify the problem or market failure to be addressed, assess whether existing regulations contribute to it, consider available alternatives including non-intervention, and include a cost-benefit analysis; (iii) constituting Regulations Advisory Committees of domain experts, legal scholars and market participants at all regulators; (iv) requiring periodic review of regulations at defined intervals with a clear methodology specifying which regulations are to be reviewed, against what criteria, and within what timeframe; and (v) leveraging technology (for instance, dashboards tracking active consultations and regulators' responses, and automated tools that flag missing elements in consultation papers before publication).

The authors are researchers at TrustBridge Rule of Law Foundation.

Supervising what you cannot inspect

by Maninder Singh Juneja and Renuka Sane.

In traditional financial regulation, supervisors are able to inspect the thing being regulated. For example, a scorecard by a bank was usually a short list of factors that one could comprehend. These factors could be traced to key documents within the bank, banks had a stated rationale for why they were being used, and there was some common sense in making weighted averages. The scorecard itself was static - the same weighted factors were applied over a period of time. Inspecting this process of constructing the scorecard and using it for management decisions was how trust was built. The AI world is different. Models are rented rather than built, change continuously, and behave probabilistically. They cannot be fully inspected even by the institution deploying them let alone by the regulator.

How should we then think of regulation?

One approach is to intensify the traditional approach where regulators demand more explainability, more documentation, and more validation. This will drive up the costs of compliance. But more importantly, this is ill suited to the new world where the technology changes rapidly, where the bank does not control the AI it uses, where there is no clear artefact that the bank can give the supervisor such as a model or a document (Board of Governors et al 2026). When regulators push traditional approaches, banks will respond by choosing AI models which are easy to document rather than the ones best for them, or defer AI deployments altogether. All these are unhappy consequences. What we need are policy makers who understand the live systems of the new world.

In this article, we analyse these emerging problems from first principles. We start from scratch, understand the landscape of market failure in the world of AI in banking, and think about how regulators can grapple with this world.

Our key idea is that AI systems resist replicability. Regulatory strategies that demand replicability will flounder or choke technology deployment. We suggest the regulatory standard applied should be AI deployments that are "supervisable" - the outcomes should be observable, they should be attributable to causes, and reversible by humans.

How AI is actually deployed

Before we get to the puzzles faced by regulators, we need to describe what AI in banking is. This categorisation is not unique to banking. Banking is simply where their consequences are regulated. Five modes span the range.

AI as Tool: AI augments a step, a search, a calculation, a first draft that the human controls. The human is cognitively engaged.

AI as Collaborator: The human and AI co-produce iteratively, and the human participates at every stage.

AI as Recommender: AI generates scores or options, and the human makes the decision. This is the classic human-in-the-loop, but credible only if the human can meaningfully interrogate the recommendation and not degenerate into cognitive surrender.

AI as Preparer: AI does the work and the human signs off. The approval here is closer to a check by a supervisor rather than a re-derivation.

AI as Autonomous executor (agentic AI): AI executes autonomously inside guardrails and the human monitors on exception. Emerging forms of agentic AI for banking run from single-task agents (a payment released, a service query resolved) to multi-step workflows and customer-facing agents that transact.

A central issue here is the true (de facto) role of the human. A reviewer who approves a thousand recommendations a day is not overseeing a model, the model is overseeing her. A feature of any deployment is the measured divergence between the model recommendation (and estimated uncertainty) vs. the human decision. Managers of banks will need to worry about relapses of human behaviour inside the organisation, a bit like how hospital managers worry about bad behaviour by doctors within their organisation.

What AI does to market failure in banking

We now shift gears to look at the standard knowledge on market failure in banking. Regulation may be justified when (and only when) there exist market failures which cannot self-correct fast enough, and there is adequate state capability in banking regulation to be able to correctly identify them and intervene. AI's distinctive feature is that it can cure several classic failures. Better default prediction reduces credit rationing, better fraud detection cuts deadweight loss, richer risk assessment lets banks serve customers they previously could not price, AI advisors help customers avoid some malpractices by the bank. But there are also some new problems that are anticipated.

Information asymmetry: This happens in banking at two levels: borrower-to-lender (adverse selection, hence credit rationing) and firm-to-consumer (product complexity, hence mis-selling). AI narrows the first through alternative data and may widen the second. On one hand, the customer armed with AI can see through many things proposed by the bank which are not in her best interest. But the consumer cannot observe why they were shown a product, offered a price, or steered toward a specific insurance plan. The sales process itself becomes more opaque. Personalised pricing approaches first-degree price discrimination, extracting consumer surplus. A single flawed model can mis-sell to millions simultaneously, converting isolated conduct failures of the pre-AI world into a big correlated event. And redress weakens when a denial comes from a model the firm itself cannot explain.

Systemic externalities from shared infrastructure: Each bank chooses its models, data sources and vendors to optimise its own performance. When multiple entities choose the same ones, the sector's exposures become correlated, which is a cost no individual bank prices in. We list the vulnerabilities below:

Correlated model risk: banks on similar models and the same foundation providers respond identically to an event. The regulator, at the system level, has to manage what happens when institutions move together, because no single entity has experience of such behaviour or of the impact synchronisation adds.

Third party concentration: one vendor's failure propagates everywhere at once. India has already run this experiment, when a ransomware attack on one shared technology provider knocked roughly three hundred cooperative and regional rural banks off the payments network.

Correlated cyber breach: shared stacks mean one exploited vulnerability is every institution's vulnerability. AI lowers the attacker's costs (automated vulnerability discovery, deepfake social engineering) and adds new attack surfaces (data poisoning, model inversion, prompt injection against agents that can move money).

Runs on banks at level 3. The bank runs of old were a queue on the pavement. Then we got to Silicon Valley Bank where over a weekend, customers took away money from the bank. Now we can be at level 3: autonomous agents managing customer cash can turn a shared signal into a self-reinforcing run at machine speed.

The various market failures listed above behave differently across the five modes of AI use. For example, when a human constructs the offer, opaque pricing can get contained. However, when an agent personalises autonomously at scale, this may become severe. Systemic correlation is moderate when AI advises and severe when fleets of similar agents act simultaneously. Any regulation that grades by model type alone, or by use case alone, misses half the object. The next question is the mode of regulation itself.

Regulatory strategy

Regulation can work in two ways. Process-based regulation is ex ante: it prescribes how the firm must operate, defines required controls, mandates oversight, validation standards, limits on autonomy. Outcome-based regulation is ex post: it prescribes ends, fair treatment, solvency, and judges results, leaving the choice of methods to the firm.

Outcome-based regulation is the efficient default. It is technology-neutral, so it does not ossify as methods change; it lets firms find the least cost route to compliance; and it does not require the regulator to understand the firm's production function better than the firm does. But it has important preconditions: the outcome must be observable and measurable; it must be attributable to the firm and, ideally, to the cause; and the harm must be reversible or compensable. Process regulation is the right departure from the default when those preconditions fail and where outcomes are unobservable, harm is catastrophic or irreversible, or damage manifests only systemically or with a lag.

Traditional doctrine treats these preconditions as given: examine the activity, choose the mode. This does not work for AI. Whether an AI deployment's outcomes are observable, attributable, and reversible is an engineering choice, which needs to be settled at design time. We suggest that regulation should mandate observability. This makes it possible to have an "outcome-based supervision" model. There are three ways to ensure observability.

1. Telemetry implies that the institution keeps a track of every decision such that the system records which version of itself it was using, what information it was given, and any time a person stepped in to overrule it. This makes outcomes attributable. That way if a certain group starts getting more (or less) approvals than before, the organisation can evaluate what caused the shift - was it the model, or the group itself. Such a capability is being mandated elsewhere in the world for similar use-cases (European Parliament and Council of the European Union, 2024). The regulator should also consider if it wants to set a minimum common telemetry standard.

2. Boundaries and rollback include putting caps on what the system is allowed to do, rolling out new updates to just a small number of cases first (instead of everyone at once), and having a tested plan for switching back to the older version if something goes wrong. This ensures that if a bad update slips through, it only affects a small slice of decisions.

3. Probes make bias observable. One way is "paired testing": one submits two applications that are exactly the same except for details that hint at things like someone's race or gender, and see if they get treated differently. One can also compare approval rates against the company's own normal levels. Together, these checks can catch an unfair credit model in just a few weeks, instead of waiting years to see who actually pays back their loans. One can also check rejected applicants against credit-bureau data to see which ones got approved by someone else, and how they fared.

These are similar to the idea of decision receipts that record which rules were applied to which facts and in what sequence for every decision made by a government or public system (Srivastava, 2026).

Over time, the supervisor should also build a repertoire of its own test cases drawn from incidents, complaints and examinations across the system and run it against every material AI deployment, much as stress-test scenarios are run against every balance sheet today. This will ensure that what surfaces in one institution becomes a probe for all others. It thus allows the regulator to set its own observability layer.

Process regulation is then reserved for the harms that are systemic, correlated, or irreversible at machine speed. For such events there need to be protections such as circuit breakers that halt things before they cause damage, limits on how much the systems can do on their own, model diversity so they don't all fail the same way, and rehearsed back-up plans.

What follows for the supervisor, the board, and the customer

For the supervisor: The unit of examination shifts from the model to the deployment, and the examiner's question shifts from "show me the validation report" to "show me the behaviour": what boundaries were set, what exceptions were thrown, what overrides were exercised, how far the system drifted from its baseline. Supervisors will also need to find the intellectual clarity to avoid a wide variety of extraneous claims about regulation of AI, e.g. the push for economic nationalism which has nothing to do with market failure.

For the board: A board cannot certify systems it cannot inspect or understand. Its role is to govern the framework which includes the limits on what the system is allowed to do, and making the rules for when a decision must be escalated to a human. The board then needs to continuously check the exceptions and overall performance.

Exception-handling should also be written into policy. If a problem stays unresolved beyond a defined size or time limit, it automatically gets escalated to the board. Internal auditors should double-check that the numbers are real. The board sticks to this audited framework and does not inspect the system directly.

If a board is asked to approve something they can't understand, they will default to saying no. But if you let them govern the limits and the exceptions instead, they can say yes.

The customer: The customer becomes part of the supervisory architecture. The widening firm-to-consumer asymmetry has a structural corrective the pre-AI world lacked: the customer now has AI too. Mainstream assistants abroad have begun connecting directly to users' accounts. If product terms (rates, fees, eligibility criteria) are mandated to be structured and machine-readable, the customer's own AI does the comparing, the explaining, and the policing of mis-selling, continuously and at zero supervisory cost. The redress channel weakened by opacity is restored the same way: an adverse decision should carry its reason to the customer, and what would have had to be different for the answer to change. This restores the ability to contest. The same asymmetry that AI widened, AI-equipped customers can close, but only if regulation hands them the data.

An example

Consider a debt-collection example. When borrowers fall behind on payments, the bank has to decide how to chase each overdue account. Contact methods differ in cost. Automated SMS and IVR (the automated phone system - "press 1 to pay") are cheap; having an employee actually phone the customer is expensive. So the bank builds a model that allocates accounts: cheap automated nudges for most, and the costly human call reserved for the accounts where the model predicts that talking to a person will actually "cure" the account (get it back to paying). Human calling works better but it costs more. The model is rationing an expensive resource to where it thinks it'll pay off.

With AI, the collection systems would retrain continuously or run reinforcement-style optimisation against a live reward (cure rate per rupee spent). The model would be moving on its own, faster than the review cycle, toward a target that is a proxy for what the bank may actually want. Under the conventional "inspect-the-artifact" approach, the model is checked before deployment, signed off, and reviewed on a schedule (say quarterly or annually). However, if AI is an active optimiser, a quarterly review discovers deterioration only after it has touched thousands of accounts. It may drift toward a mis-specified target in a way no one can read off the model itself. Further, if the collections model is a shared vendor product, or several banks fine-tune the same foundation model on similar data, they all go bad the same way at the same time when borrower behaviour shifts. In contrast, in a "supervisability-built in by design" approach, a small slice of accounts is deliberately kept on the previous allocation method, running live alongside the new model. This allows comparison between the old and the new in real time.

Every decision should keep a record of a few things: which version of the AI was used, what action it chose, which group of customers it was dealing with, and how things turned out in the end. This is how it helps: Say the AI's overall success rate is going up, but one particular group is quietly getting fewer phone calls from real people and doing way worse because of it. With all that recorded, that gap shows up right as it's happening. You can compare it against what's going well elsewhere, trace it back to a specific version of the AI, and undo it that same day, because the older setup is still up and running as a backup.

The company should never have to dig into the AI's inner "thought process" to realize its behaviour has taken a bad turn. None of this replaces the normal testing that the firm would do anyway. The difference is that here, the ability to observe what's happening, pin down what caused it, and shut it down are all built right into the system from the start, instead of being things you have to go do to the AI afterward.

Conclusion

AI deployments sit uneasily in conventional mores of banking regulation. The intelligence is rented, the behaviour is probabilistic, and harms can move at machine speed. We suggest that the response should not be more inspection of what cannot be inspected, nor blind faith in outcomes that arrive too late. Efficient regulation will emerge from observability, attributability, and reversibility. Regulators should require banks to build these.

References

Board of Governors of the Federal Reserve System, Federal Deposit Insurance Corporation, and Office of the Comptroller of the Currency. "Supervisory Guidance on Model Risk Management." SR Letter 26-2. April 17, 2026. https://www.federalreserve.gov/supervisionreg/srletters/SR2602.pdf.

European Parliament and Council of the European Union. Regulation (EU) 2024/1689 (Artificial Intelligence Act), arts. 12, 26(6), and Annex III(5)(b). https://ai-act-service-desk.ec.europa.eu/en/ai-act/article-12.

Srivastava, Manish. "Digital Governance Needs Decision Receipts." Episode 73 of Big Ideas. XKDR Forum, June 1, 2026. Podcast, video, 13:06. https://youtu.be/WFX4ITb9yok

Market Reaction to Insider Trading: Evidence from Regulatory Orders in India

by Arjun Gupta, Sonam Patel, and Renuka Sane.

Introduction

Market integrity depends on effective enforcement against market abuse. When regulators credibly sanction violations, they reinforce investor confidence and reduce the risk premiums that markets impose for uncertain governance. In developed markets, evidence suggests that enforcement achieves this objective: SEC enforcement actions in the United States produce abnormal stock price declines of $-0.5\%$ (Persons, 1997), and UK sanctions trigger reputational losses that far exceed the direct penalties (Armour et al., 2017). This is especially true for insider trading enforcement: Persons (1997) documents significant negative abnormal returns following the SEC's announcements of insider trading enforcement actions. (Engelen, 2012) finds that a clear negative abnormal return on the day of even newspaper coverage of the illegal insider trading practice of CEOs.

An open question, however, is whether this pattern extends to India. We investigate this by examining stock price movements around two types of insider trading enforcement actions in India: final SEBI adjudicatory orders and appellate decisions by the Securities Appellate Tribunal (SAT). We focus on insider trading orders as they can be a signal about the quality of the firm's internal governance. If insiders are trading on privileged information, it suggests that boards, compliance functions, and internal controls are weak, leading to investors discounting the stock accordingly. Further, when the firm and its executives face potential penalties, disgorgement, or other sanctions, these can impose direct costs on the firm and may affect its ability to attract capital and talent. The insider trading laws in India are quite expansive, and cover not only connected persons, but also those who just have access to unpublished price sensitive information, or if there have been some minor disclosure violations. All orders, therefore, may not signal governance issues within a firm. We therefore also look at orders by violation severity and type of insider relationship.

Empirical Strategy

We use an event-study methodology to test whether Indian stock markets react to SEBI enforcement actions and outcomes challenged before SAT. We compile a list of individuals and entities against whom an insider-trading order was issued, then identify the companies whose scrip was alleged to have been insider traded, map them to their corresponding order dates (event dates), and use these firm-event pairs to check for market reaction.

Estimation Procedure

We estimate each firm's normal return using the market model over an estimation window of 210 trading days ending 11 days before the event ($t = -210$ to $t = -11$):

\( R_{it} = \alpha_i + \beta_i R_{mt} + \varepsilon_{it} \)

where $R_{it}$ is the daily return of stock $i$ on day $t$ and $R_{mt}$ is the daily return on the Nifty~50 Index. The Abnormal Return (AR) on event day $t$ is the difference between the actual return and the predicted normal return:

\( AR_{it} = R_{it} - \left(\hat{\alpha}_i + \hat{\beta}_i R_{mt}\right) \)

Cumulative Abnormal Returns (CARs) are computed by summing $AR_{it}$ over a 21-day event window centred on the insider trading announcement date:

\( CAR_i = \sum_{t=-10}^{+10} AR_{it} \)

We test whether the cross-sectional average $\overline{CAR}$ is statistically different from zero using a $t$-test; a negative $\overline{CAR}$ indicates an adverse market reaction to the announcement.

Data and Sample

Our sample is drawn from the data set used by Aggarwal et al., (2025). It comprises two types of regulatory actions from 2009 to 2023, restricted to firms listed on the National Stock Exchange (NSE). After removing duplicates and cases with missing stock price data, our final sample contains:

1. SEBI Orders: Final adjudicatory orders; $N = 176$ firm-event pairs.
2. SAT Orders: Appellate Tribunal decisions; $N = 42$ firm-event pairs.

We further look for heterogeneity in market reactions by partitioning the sample along four dimensions of interest:

• Sanction status: Sanctioned ($N = 119$) vs. not sanctioned ($N = 57$). An order may or may not result in a sanction. Here, we examine a reaction based on whether an order resulted in a sanction.

• Violation severity: Major violations ($N = 74$) vs. minor violations ($N = 122$). We classify insider trading violations as Major (e.g., sharing or using unpublished price information for trading) or Minor (e.g., code-of-conduct breaches, delayed disclosures).

• Insider relationship: Connected persons ($N = 44$), deemed connected ($N = 21$), and those with access to UPSI ($N = 19$). Connected persons are loosely defined as those associated with a company (contractual, fiduciary, or employment), while those deemed to be connected persons include their relatives or cohabitants. UPSI access refers to knowledge of information materially impacting the stock price.

• Monetary outflow: Above-median ($N = 59$) vs. at-or-below-median ($N = 60$) alleged illegal gains. Monetary outflow is the total penalty and disgorgement paid to SEBI. We analyze market reaction based on the magnitude of this outflow to see if the amount paid affects the reaction.

Results

Baseline Event-Study Findings

Our event-study results indicate that Indian stock markets exhibit no statistically significant reaction to any type of insider trading enforcement announcement. CARs are indistinguishable from zero across all two regulatory action types at the 95% confidence level, with point estimates close to zero in magnitude. For comparison, SEC insider trading enforcement actions in the US produce average CARs of $-3.47\%$ (Muradoglu and Clark Huskey, 2008).

Figures display the CAR trajectories. In all two cases, the CARs fluctuate around zero with no discernible trend before, during, or after the announcement date.

Cumulative Abnormal Returns (CARs) around SEBI final order announcements ($N = 176$). The shaded region represents the 95% confidence interval.

The result for SEBI final orders is striking: these orders contain explicit findings of misconduct and penalties, yet markets do not react. We discuss four candidate explanations below: high appeal and reversal rates, long enforcement delays, low penalty amounts, and pre-existing credibility discount.

Cumulative Abnormal Returns (CARs) around SAT order announcements ($N = 42$). The shaded region represents the 95% confidence interval.

SAT orders, on the other hand, are more final in nature. They may affirm, modify, or overturn SEBI sanctions, and should lead to a market reaction. In our dataset, they also produce no detectable market reaction. However, this result should be interpreted with caution, given the small sample size. With only 42 events, our test has limited statistical power to detect abnormal returns. It is possible that these may be further appealed at the Supreme Court, but given the small sample size, we do not test for the impact of those decisions.

Subsample Analysis

We examine whether the aggregate result masks heterogeneous effects by partitioning the sample along the four dimensions described above. Across all subsample splits, CARs remain statistically and economically insignificant. Even for high-severity cases, directors trading on confidential information, monetary outflows above the median of Rs.~12.83 lakh, and third quartile of Rs.~5.27 crore, abnormal returns remain proximate to zero. Also, there is no evidence of a significant market reaction even in the subsample of cases where the insider relationship is more direct (connected persons). This suggests the null result is not an artefact of averaging across heterogeneous effects; rather, it is pervasive across subgroups.

One caveat to our analysis is if the true information release occurred earlier (e.g., via media leaks), our tests measure the reaction to information from informal sources rather than to the announcement itself.

Interpreting the results

One interpretation of this result is that markets may rationally discount the significance of SEBI enforcement actions. Several institutional features of Indian capital markets lend support to this interpretation:

1. High appeal and reversal rates: Aggarwal et al., (2025) find that a substantial fraction of SEBI orders (30-41%) are appealed to SAT, and around 50% result in modifications or reversals. Investors who have learned that sanctions are frequently overturned will rationally discount any announced penalty. This is probably compounded by the fact that several SEBI orders are not able to demonstrate the unfair gains or loss avoided, or provide reasons for imposing sanctions as debarment, reducing the credibility of its enforcement actions (Aggarwal et al., 2024).
2. Long enforcement delays: Damle and Zaveri (2022) find a median of over three years between violation and SCN, and a further 18 months to a final order. This is reinforced by the findings of Aggarwal et al., (2025), who find similar timelines for insider trading orders. By the time enforcement is announced, investors may have already moved on.
3. Low penalty amounts: The median penalty is Rs 12.83 lakhs, with approximately 40 cases involving amounts under Rs 10 lakhs. Such low penalties suggest a lower perceived severity of the offense, and consequently signal the market to treat this news as immaterial.
4. Pre-existing credibility discount: If years of weak or delayed sanctions have already led investors to assign a low probability to effective enforcement, individual announcements convey little new information, and markets have stopped paying attention.

Another possibility is that markets receive the enforcement information but do not regard insider trading as material to firm valuation. Under this view, it reflects an investor judgment that insider trading by management is not indicative of broader governance failure or future cash-flow risk.

Conclusion

Indian stock markets exhibit no statistically significant response to insider trading enforcement, in contrast to the negative abnormal returns documented in the US and UK. This result is robust across SEBI final orders and SAT appellate decisions, and persists even for high-severity violations involving senior insiders and large monetary outflows.

The functioning of SEBI entails considerable public expenditure, and the Board has, over time, sought progressively wider powers - including expanded surveillance capabilities. Given this, the question of what is actually being achieved warrants serious scrutiny. A stock price reaction to an enforcement order is one observable signal of whether the market believes the enforcement actions carry some significance. A null result across many orders suggests the market does not view these actions as conveying meaningful new information. It is, therefore, worth questioning if enforcement actions are advancing the goal that justified the expenditure in the first place.

The authors are researchers at Trustbridge Rule of Law Foundation.

What happens when arbitration deadlines are missed

by Prashant Narang and Renuka Sane.

Section 29A of the Arbitration and Conciliation Act 1996 was introduced to deal with delays in arbitration. It sets a time limit for making an award. If that time runs out, parties have to go to court to extend it. The court can also impose consequences for delay, such as reducing fees, awarding costs, or replacing the arbitrator.

Our new working paper studies how this works in practice. It looks at 202 reported orders of the Delhi High Court between 2015 and 2024.

It finds that the Court almost always grants extensions and almost never imposes sanctions.

What the data shows

Out of 202 cases, the court granted extensions in 198 (98%). Only 4 cases were dismissed, and those were on technical grounds. Sanctions were rarely imposed.

• Fee reduction: 0 out of 202 cases
• Adverse costs: 6 out of 202 cases (about 3%)
• Replacement of arbitrators: 4 out of 202 cases (about 2%)

Repeat extensions are not unusual. There are 30 cases where parties came back for a second or later extension. The court granted 29 of them (96.7%). There are no sanctions in these repeat cases.

These petitions also move quickly.

• Median time to decide: 3 days
• Median number of hearings: 1
• About 63% of cases are decided in a single hearing

So the delay is not in the court process. Courts dispose of these matters quickly. But they usually extend time without imposing any consequence.

Why extensions are common

Part of the answer lies in how Section 29A is structured.

For the Court, giving an extension is easy if both parties agree. The court can dispose of the case quickly.

Imposing a penalty is harder as the Court has to find out who caused the delay. It may have to look at the record in detail. It also has to hear the arbitrator before cutting fees. All this is likely to take more time and effort.

It is not surprising that consensual extensions are more common.

What this means for the law

Over time, this pattern shapes how the law works.

Section 29A was meant to push arbitrations to finish on time. It often works as a way to formally extend time after the deadline has passed.

If parties expect that extensions will be granted without much difficulty, the deadline may lose its force.

This does not mean the provision has no value. But it suggests that deadlines work best when consequences are easy to apply.

Looking ahead

If deadlines are not backed by predictable consequences, do they change behaviour?

The paper does not answer this fully. It focuses on what courts do once parties come for an extension. But the pattern is clear. Extensions are routine and sanctions are exceptional.

That may matter for how arbitration timelines are taken in practice.

You can read the working paper here.

The authors are researchers at TrustBridge Rule of Law Foundation.

Evaluating India's Energy Ambitions: Evidence from Electricity Generation Project-Level Data

by Upasa Borah, Akshay Jaitly and Renuka Sane.

India's electricity demand has been growing rapidly, at 9% per annum since 2021. Meeting this demand by 2030 would require around 777 GW of installed capacity, as estimated by the Central Electricity Authority (CEA). At the same time, India has committed to achieving 500 GW of installed non-fossil capacity by 2030. A study by CEEW (2025) finds that meeting this target would require adding around 56 GW of non-fossil capacity every year between 2025 and 2030, failing which India would need an additional 10 GW of coal-based capacity to meet future demand. There is little doubt that renewable energy in India has seen a sharp growth, with 74 GW in 2018 to 162 GW by the end of 2024 (excluding large hydro and nuclear projects), driven by falling renewable energy prices, and policy support like subsidies for developers, waivers on inter-state transmission charges, Green Energy Corridor investments, changes in Green Open Access Rules and various state-level initiatives that signal policy commitment to the sector. In 2025 alone, the country added 45 GW of renewable capacity.

However, the next phase of the transition is likely to be more complex. India is now facing new challenges regarding grid integration and transmission infrastructure, leading to delays in commissioning projects and curtailment of operational projects. As of June 2025, around 50 GW of awarded renewable capacity was stranded due to a lack of buyers, transmission constraints or disputes over land and environmental clearances. This results in time and cost overruns, dampening investor confidence.

In this backdrop, our paper Evaluating India's Energy Ambitions: Evidence from Electricity Generation Project-Level Data studies how electricity generation projects evolve from announcement to completion. Using project-level data from the Centre for Monitoring Indian Economy (CMIE) CapEx database, we analyse 8,540 projects announced between January 1957 and December 2024 to understand how project size, cost, ownership, energy technology and location influence project timelines. We ask,

1. How many projects have been announced and of them, how many have been implemented and completed? What is the time taken?
2. Given the projects currently in the pipeline, how likely is India to meet the 2030 targets?
3. How do factors like project size, geography and developer characteristics influence the completion timelines and probabilities?

From announcement to completion

We find a significant divergence between projects announced and completed: of the total announced conventional (CE) and renewable (RE) capacity, only 15% and 9% have been completed, respectively. Announcement here refers to events like signing of MoUs, inviting bids, seeking approvals or preparing feasibility reports and may differ from official statistics that use alternative definitions of project status (Borah et al., 2025). The next stage in a project lifecycle is beginning implementation, which includes events like awarding contracts, securing financing, obtaining approvals or beginning construction, indicating a deeper commitment of resources. Even among this set of projects that have been implemented, completion rates remain low: 30% of CE and 22% of RE capacity have been completed. The timelines from announcement to implementation and implementation to completion vary significantly among different technologies, with solar and wind having the shortest timelines.

How much capacity will be added by 2030?

We used an accelerated failure time survival model to estimate the completion probabilities of projects currently in the pipeline (i.e. announced or under implementation as of December 2024). Applying a probability threshold of 0.5, i.e. excluding projects with less than 50% chance of completion by 2030, and scaling our dataset to match the capacities reported by the CEA, we find that India is likely to fall short of its capacity targets.

If the current completion trends continue, total installed capacity would fall short of the 777 GW target by around 56 GW for CE and 45 GW for RE. Similarly, for the 500 GW non-fossil target, the projected shortfall is around 77 GW. It is important to note that our analysis does not include new projects that may be announced after 2024. In that sense, our findings imply that meeting the 500 GW target would require announcing and completing 77 GW of projects within the next six years.

Explaining the capacity additions

We find that project characteristics play an important role in influencing implementation and completion timelines:

• Project size: Larger projects take longer to begin implementation and get completed.
• Ownership: Privately developed projects tend to be completed faster.
• Developer ranking: For RE projects, those developed by top firms (by market share) perform better.
• Location: RE projects in certain states such as Gujarat, Rajasthan and Andhra Pradesh complete faster than those in states with weaker RE ecosystems. Location is less important for CE projects.
• Year of announcement: RE projects announced after 2022 have longer implementation timelines compared to those announced before 2018.

These findings hold taking into account disruptions caused by the COVID-19 lockdown, which we explicitly model.

Finally, we compare completion timelines of large-scale solar and wind projects across states with benchmark timelines in the literature and find that even in RE-rich states, large projects face delays in commissioning.

Taken together, our findings suggest that the challenge is not just the announcement of new capacity but ensuring projects are implemented and completed on time. Bridging this gap will be critical to meeting India's future energy goals.

The authors are researchers at TrustBridge Rule of Law Foundation.

Comments on the Securities Market Code Bill, 2025

by Natasha Aggarwal, Pratik Datta, K. P. Krishnan, Bhavin Patel, M. S. Sahoo, Renuka Sane, Ajay Shah and Bhargavi Zaveri-Shah.

Finance is the brain of the economy. It dictates allocative efficiency. The financial system chooses which industries and firms receive capital. This efficiency determines the extent to which investment translates into GDP growth. Getting finance right is critical. The prioritisation of financial reform must be absolute.

The Securities Market Code Bill, 2025 (SMC) marks a substantial advance over the existing Securities and Exchange Board of India Act, 1992, particularly in strengthening governance arrangements and formalising the processes of regulation-making. Importantly, it makes a serious attempt to end the ''circular raj'' by confining the issuance of subsidiary instruments to the Chairperson or senior members of the Board, rather than dispersed internal authorities. Further, it has introduced timelines for investigations and attempted to separate the investigation function from the adjudication function, making the first effort towards a clearer separation of powers. That said, the SMC can make further strides if it focuses on the issues described below.

We now address the issues in relation to specific provisions drafted within the current SMC.

Separation of powers

The SMC raises three related concerns, which demonstrate a concentration of powers at SEBI.

Issue 1: Excessive delegation of essential legislative functions

Clause 96 prescribes imprisonment, a fine, or both as penalties for market abuse (an offence defined under Clause 93). However, Clause 93 also grants the regulatory authority to define new offences within the 'market abuse' category, which would carry the same criminal sanctions. This raises concerns around excessive delegation: the identification of criminal offences is a core legislative function and cannot be delegated. Moreover, such excessive delegation is subject to being struck down in judicial review.

Issue 2: Regulation-making on adjudication

Clause 146(2)(j), read with Clause 17(4), permits SEBI to make regulations on the manner of conducting adjudication proceedings. This should not be done by SEBI itself. SEBI is the agent, and the Parliament is the principal. The Parliament must define the checks and balances on the coercive power of the agent. Otherwise, the agent always has incentives to appropriate more arbitrary power.

Issue 3: Ineffective separation of investigative and adjudicatory functions

Clauses 17 and 27 introduce limited separation of investigative and adjudicatory functions for specific matters. Investigation is an executive function, and adjudication is a quasi-judicial function. A conflation of these two functions in the same individual raises concerns about the separation of powers.

In summary, there is no clear separation of power between the three functions of the regulator. The same regulator is empowered to define the scope of violations and offences, investigate them, enforce them, adjudicate upon them, and impose sanctions for their violations, all under regulations of its own design. This combination blurs the distinction between legislative, executive, and adjudicatory functions and concentrates powers in the same persons.

Proposal:

Remove Clauses 17(4), 92(f), 93(g), and 146(2)(j) from the SMC. Implement strong structural separation between the investigative and adjudicatory functions. One way to do this is to create a distinct career track for adjudicatory officers as Administrative Law Officers (ALO). One SEBI board member should also be designated as an Administrative Law Member, who oversees the functions of ALOs. These officers should be solely responsible for adjudication and must have no involvement in investigative or quasi-legislative functions. Introduce extraordinary safeguards to mandate arm's length operation between investigation and adjudication.

Timelines for investigation and adjudication

Issue: Clauses 13, 16, and 27 introduce timelines for investigation and interim orders. However, provisos allow these timelines to be extended (Clause 27(4), proviso to Clause 13(2)). Additionally, the SMC specifies no timelines for the completion of adjudication proceedings. This allows investigations and adjudications to continue indefinitely, rendering the statutory limits ineffective.

Proposal: Remove the power to extend timelines for investigation. If extensions are retained, mandate the publication of written reasons, subject to mandatory review by the SEBI governing board. Introduce a strict statutory timeline for the conclusion of adjudicatory proceedings. These timelines should be part of the Parliament-specified regulations on the manner of conducting adjudication proceedings that we recommend in our preceding suggestions.

Methodology for calculating unlawful gains

Issue: The SMC requires the determination of unlawful gains by an investigating officer under Clause 13(3), but provides no calculation methodology. This virtually guarantees arbitrary and inconsistent determinations. It defeats the rule of law.

Proposal: Codify standard methods or guidelines for calculating unlawful gains within the SMC. Operationalise these through detailed regulations. Reference the Competition Commission of India (Determination of Monetary Penalty) Guidelines, 2024, as a baseline.

Sanction determination factors

Issue: The SMC lists factors for adjudicating officers to consider while imposing sanctions. Some mirror Section 15J of the SEBI Act, which are unimplementable in practice. Terms like 'impact of the default or contravention on the integrity of the securities markets' (Clause 19(b)(v)) lack precision and invite arbitrariness.

Proposal: Base sanctions strictly on the quantifiable extent of harm caused to specific persons. Codify this methodology. Alternatively, publish binding guidelines detailing specific aggravating and mitigating factors, expanding upon the approach in the SEBI (Settlement Proceedings) Regulations 2018.

Criminal enforcement

Issue: The SMC retains criminal liability, including imprisonment, for some offences. Establishing guilt in Indian criminal law requires proof beyond a reasonable doubt, typically coupled with the requirement to establish intention. This is an inefficient tool for complex financial markets. The boundary between aggressive trading and market manipulation is thin. The threat of criminal sanctions deters contrarian strategies. This reduces market liquidity and harms price discovery. Traditional fraud is adequately covered by the Bharatiya Nyaya Sanhita.

Proposal: Remove all criminal liabilities. Structure sanctions as punitive civil penalties or restorative remedies, scaling to a multiple of the illicit gains. Retain debarment for systemic misconduct.

Power to issue directions

Issue: Clause 23 vests SEBI with open-ended direction-making powers. Moreover, the requirement to record reasons in writing (currently included in Section 11(4) of the SEBI Act) has not been included in Clause 23.

Proposal: Delete Clause 23. Confine non-penal measures to specific, narrowly defined statutory triggers (e.g., immediate asset freezing powers under strict procedural safeguards). All adjudicatory actions must be justified by reasons in writing.

Nominee directors on the SEBI board

Issue: The SMC retains government nominee directors on the SEBI board. Nominee directors prioritise the perspective of their parent departments over market efficiency. They exercise disproportionate influence. Inter-agency coordination should not occur via board representation.

Proposal: Appoint mid-career professionals for fixed terms until a mandatory retirement age. Bind them statutorily to SEBI's specific objectives. Address inter-agency concerns externally through the Financial Stability and Development Council (FSDC).

Commodities markets

Issue: Clause 49 empowers the government to determine commodities eligible for trading. The market must decide which commodities warrant hedging instruments. State determination of eligible commodities is equivalent to the government deciding which firm is permitted to issue equity.

Proposal: Delete Clause 49. Empower SEBI to draft regulations defining objective eligibility criteria for commodity derivatives, identical to the framework for eligible scrips.

Ombudsperson

Issue: Clause 73 empowers SEBI to designate an Ombudsperson. This creates a conflict of interest. The SMC lacks an appeals mechanism for decisions made by the Ombudsperson.

Proposal: Mandate statutory independence for the Ombudsperson. Ensure job security separate from SEBI management. Define a clear appellate process.

Exemptions for PSUs

Issue: Clause 65(2) empowers the Central Government to exempt listed public sector companies from listing and disclosure requirements. This violates Article 14 of the Constitution. State-owned enterprises must face the identical market discipline applied to private enterprises.

Proposal: Delete Clause 65(2). Mandate equal treatment for all market participants.

References

Natasha Aggarwal and others, "'Balancing Power and Accountability: An Evaluation of SEBI's Adjudication of Insider Trading'" (Working Papers, TrustBridge Rule of Law Foundation, 2025).

In Re: The Delhi Laws Act, 1912 (AIR 1951 SC 332).

M S Sahoo and V Anantha Nageswaran, 'Regulatory architecture 2.0: Securities Markets Code marks a decisive shift' (Business Standard, 25 December 2025).

M.S. Sahoo and Sumit Agrawal, "Reimagining SEBI's Consent Settlement Framework" (Chartered Secretary, January 2026).

C.K. Takwani, Lectures on Administrative Law (7th edition, 2023) at page 100.

Bhargavi Zaveri-Shah, 'SEBI does not need unlimited powers – here's what's wrong with the Securities Markets Code' (ThePrint, 5 January 2026).

Bhargavi Zaveri-Shah and Harsh Vardhan, 'Ghost of the Commodities Controller—why India's new financial law feels like the 1970s' (ThePrint, 19 January 2026).

An Analysis of Electricity Outages in Delhi: 2024-25

by Upasa Borah and Renuka Sane.

Introduction

In a previous article, A Review of Outage Reporting by Indian DISCOMs, we examined the state of outage data reporting across India. We studied which distribution companies (DISCOMs) report such data and the variations in the way they do so. A natural next step is to thus look more closely at the available data to understand the kinds of analyses they enable.

This article focuses on the three privately owned DISCOMs operating in Delhi. Delhi's DISCOMs rank below the top 20 in the Ministry of Power's annual ranking of DISCOMs, all three graded B minus in the 13th Ranking exercise in 2025. They are similarly situated in terms of their billing and collection efficiency, power procurement portfolios and costs. There are, however, notable differences in the availability, structure and clarity of their reported outage data.

It is important to note that not all outages at a feeder level translate into outages for consumers due to the presence of redundancy in power systems. Most modern systems can re-route electricity through alternate feeders in case of faults. Understanding whether and how redundancy is accounted for is thus crucial to interpreting outage data. For instance, one of Delhi's DISCOMs, BSES Rajdhani Power Ltd., reports outages at the feeder level, but there is no information on which feeders have redundancy systems or how many outages were rerouted and thus did not cause interruptions for end consumers. On the other hand, Tata Power Delhi Distribution Ltd. reports outage data by zones and the number of consumers affected, allowing us to infer the extent of consumer impact. BSES Yamuna Power Ltd., however, reports outages by division and subdivisions and does not note the feeders or consumers impacted.

Given these data limitations, our analysis does not directly compare performance between DISCOMs. Instead, we study the available data to demonstrate the kinds of insights that can be drawn about the frequency, duration and spatial patterns of outages in Delhi. Specifically, we ask:

1. What is the pattern of outages on the following parameters:
1. Duration and frequency,
2. Intensity,
3. Geography,
4. Reasons for outages
2. What is the relationship between outages and electricity demand?

Methodology

There are four distribution companies operating in Delhi: i) BSES Rajdhani (BRPL) covering the southern and western areas, ii) BSES Yamuna (BYPL) covering the southeast and northeastern regions, iii) Tata Power (TPDDL) in the north and northwest areas, and iv) New Delhi Municipal Corporation (NDMC), which supplies to government buildings in central Delhi. Excluding NDMC, the first three DISCOMs are privately owned and supply to 93% of consumers in Delhi; BRPL supplies to 31 lakh consumers covering an area of approximately 700 sq km, TPDDL supplies to 20 lakh consumers in 510 sq km, and BYPL supplies to 19 lakh consumers in an area of around 200 sq km (Chitnis et al., 2025). In 2024-25, Delhi's electricity requirement stood at 38,287 MU, with peak demand hitting 8,685 MW.

We collected outage data from each DISCOM's website (see Data appendix). Lack of data for NDMC limited our analysis to the remaining three DISCOMs. The reported data includes date and time of outages, durations, areas affected, reasons for outages and measures taken to rectify the issue. However, there are inconsistencies in the data reported by the three. Table 1 summarises the variations in the availability of outage data for the three DISCOMs under study.

Table 1: Availability of data on power outages

DISCOM	Days of data availability	Spatial unit of reporting data	Number of spatial units
TPDDL	April, May, July and August 2024	Zones	12 zones
BRPL	April 2024 to March 2025	Grid and feeder	428 grids, 2,951 feeders
BYPL	April 2024 to March 2025	Division and sub-division	28 divisions and 108 subdivisions

TPDDL data is available only for April, May, July and August 2024. It reports data on zone-wise outages and the number of consumers impacted. BRPL, on the other hand, provides data on grid and feeder levels, without noting how many consumers were affected. Since outages at the feeder level may not always indicate consumer-level interruptions, understanding redundancy systems is important, but data on these was not available. There is also no data on how many consumers are serviced by a grid or feeder. Finally, BYPL reports outage data at the division and sub-division level without specifying feeder details or the number of consumers affected.

Aside from these differences, we also noticed inconsistencies in the way data is recorded, in terms of structure, format and number formatting. We extracted outage data from PDFs, conducted thorough cleaning and reorganisation. Although the datasets included reported outage durations, we recalculated the duration of each outage for all three DISCOMs based on recorded start and end dates and times. In terms of reasons for outages, TPDDL lists six broad reasons, which we retained. In contrast, BRPL and BYPL record a wider and more open-ended set of reasons, which we analysed and classified into six broad categories using text search.

TPDDL: consumer-facing outages

Between April and August 2024 (excluding June), the parts of Delhi serviced by TPDDL recorded an average of around 87 outages per day. Across all zones and feeders, these outages cumulatively amounted to roughly 159 hours of interruptions per day, and affected around 46,000 consumers. Figure 1 shows the daily frequency and total cumulative hours of outages across all TPDDL zones. On most days, outages occurred in 11 of the 12 reported zones.

Figure 1: Aggregate frequency and duration of outages for TPDDL

Over the four months for which data is available, we analysed outage days and duration for each TDPPL zone, and then averaged the results across zones. The median and mean values are presented in Table 2.

Table 2: Average days of outages, intensity and number of consumers impacted in the four reported months

Total number of zones	Number of consumers facing outages (lakhs)		Days of outages		Intensity of outages per outage day* (hours)
	Median	Mean	Median	Mean	Median	Mean
12	4.17	4.75	121	116	8.39	13.59

* cumulative value across all feeders

On average, a TPDDL zone experienced outages on 116 days, affecting around 4.7 lakh consumers. It is important to note that these are aggregate zone-level values, i.e. they do not represent outages faced by an average consumer but rather the cumulative outages across all feeders within a zone, covering multiple subdivisions and localities. For instance, Narela, Badli, and Bawala zones have the highest number of outage days, with Narela having the highest intensity (40 hours cumulatively per outage day) across the various areas in the zone, affecting 9.15 lakh consumers. The total duration exceeds 24 hours because a single zone has several feeders whose outages are aggregated when they occur simultaneously.

Around 16% of all outages reported by TPDDL are due to planned events. Figure 2 shows the share of outages by reason. 71% of outages, accounting for 60% of total outage hours, are due to external factors where the specific cause is not reported. A more detailed classification of these categories would help identify the underlying causes of outages more accurately. It also remains unclear what is included under "EODB compliance" outages, which account for 12% of all outage hours, and "Industrial weekly off" that accounts for 3% of outage hours.

Figure 2: Reasons for outages for TPDDL

BRPL: Feeder-level outages

On an average day, around 48 feeders under BRPL experience outages, amounting to a cumulative total of 50 outages and 119 total hours of interruptions across all feeders. Figure 3 shows the daily frequency and duration of these outages. The highest number of outages occurred on 7 January 2025, when 104 feeders were affected, resulting in a combined total of 386 cumulative outage-hours.

Figure 3: Aggregate frequency and duration of outages for BRPL

Of the 428 BRPL grids, an average grid had around 8 feeders under outages, with a mean of 28 days of outages in a year. Cumulatively, this results in approximately 1.8 hours of interruption per outage day across its multiple feeders. Table 3 presents the median and mean values of feeders under outage, days of outages and intensity of outages across the grids. The median values are lower than the means, indicating that while most grids experience relatively fewer and shorter outages, a few grids have significantly higher levels of outages. For instance, in 2024-25, the most outages occurred in Jaffarpur grid (187 days of outages with a cumulative intensity of 9.9 hours per outage day), followed by Nilothi grid (247 days, 4.6 hours), Mitraon grid (182 days, 6 hours), Hastal grid (236 days, 4.4 hours) and C-Dot grid (185 days, 5.39 hours).

Table 3:Average days of outages, intensity and number of feeders impacted 2024-25

Total number of grids	Number of feeders under outages		Days of outages		Intensity of outages per outage day* (hours)
	Median	Mean	Median	Mean	Median	Mean
428	2	8	2	28	0.75	1.80

* cumulative value across all feeders

Figure 4 shows the share of outages by reason. Planned events account for 54% of all outages and 82% of total outage hours. Fault-related outages follow, making up 31% of outages and 9% of total outage hours. Most outages of BRPL are thus planned rather than caused by unforeseen circumstances.

Figure 4: Reasons for outages for BRPL

BYPL: Area-wise outages

On an average day, BYPL areas recorded 16 outages, with a cumulative duration of 12.5 hours across all affected feeders. Figure 5 shows the daily frequency and duration of these outages. The highest number of outages occurred on 28 June 2024, when 98 outages were recorded, lasting a combined total of about 100 hours.

Figure 5: Aggregate frequency and duration of outages for BYPL

For an average subdivision serviced by BYPL, outages occurred on about 22 days in a year, with a cumulative average of 58 minutes per outage day. The median values are lower at just three days of outages (Table 4), indicating that most subdivisions experienced fewer days of outages, while a few faced disproportionately higher outages. Sonia Vihar recorded the most outages (201 days with a cumulative intensity of 1.86 hours per outage day), followed by Nand Nagri (196 days, 1.84 hours) and Karawal Nagar (179 days, 1.62 hours).

Table 4: Days of outages, intensity per outage day during the year 2024-25

Total number of subdivisions	Days of outages		Intensity of outages per outage day* (hours)
	Median	Mean	Median	Mean
108	3	22	0.92	0.96

* cumulative value across all feeders

Figure 6 shows the share of outages by reason. BYPL has zero outages explicitly listed as "planned". 51% of outages accounting for 47% of outage duration were due to faults, followed by maintenance outages and outages due to infrastructure damage.

Figure 6: Reasons for outages for BYPL

Electricity demand and outages

The lack of consistent and comparable data makes it difficult to analyse the yearly correlation between Delhi's electricity demand and outages. However, looking at BRPL and BYPL's outage data reveals contrasting results. BRPL's daily outage hours show no correlation with Delhi's electricity demand (Figure 7), while BYPL outages are positively correlated, significant at the 1% level (Figure 8).

Figure 7: BRPL outages and Delhi's total electricity demand

Figure 8: BYPL outages and Delhi's total electricity demand

Moreover, when we look at the time when most outages occur, we find similar divergence. Most of the outages of TPDDL and BRPL were recorded to have occurred between 6am to 12pm, which is different from Delhi's peak demand hours which are generally from 2 pm to 5 pm, and 11 pm to 1 am. BYPL's outages, on the other hand, seem to mostly occur around 12pm to 6pm. A detailed share of total outages by time of day is given in Table 5.

Table 5: Proportion of total outages and duration by time of day

Time of day	Share of TPDDL's total outages (%)		Share of BRPL's total outages (%)		Share of BYPL's outages (%)
	By frequency	By duration	By frequency	By duration	By frequency	By duration
12am - 6am	8.4	6.4	7.2	2.3	21.1	22.3
6am - 12pm	38.7	51.5	53.1	73.2	22.0	21.9
12pm - 6pm	37.3	28.6	31.3	21.8	32.5	31.3
6pm - 12am	15.5	13.4	8.4	2.6	24.4	24.6

Conclusion

Our analysis finds that the lack of a common standard and clarity in reporting makes it difficult to draw definitive conclusions about the frequency, duration, and causes of outages in Delhi. There seems to be a substantial number and hours of outages, but in the case of BRPL and BYPL, we do not know how many of those lead to consumer-facing outages, and thereby cannot assess the reliability of supply.

Several other issues also stand out. For example, TPDDL's outage reasons are not clearly defined: what exactly counts as EODB and Industrial weekly off outages? Meanwhile, most of BRPL's outages are marked as "planned". It is unclear if they translate to interruptions for consumers, but it is worth asking why such a large share is planned. On the other hand, BYPL does not report a single planned outage, which seems equally puzzling.

There are also differences in the spatial units used for reporting. That TPDDL reports 12 zones, BRPL 428 grids and BYPL 108 subdivisions implies that TPDDL's higher outages could be due to its larger geographical units. Even between BSES's two DISCOMs, outage data are reported differently, with no information on how many consumers are connected to a feeder or fall under a subdivision, making it difficult to assess the real impact of outages.

While much attention is paid to the financial performance of DISCOMs, it is also important to study the reliability of the electricity they supply. Internationally, countries like the United States and the United Kingdom publish country-wide, disaggregated outage data that enable detailed analyses of reliability, causes and impacts. For instance, studies using US Department of Energy data examine reliability and causes across states (Ankit et al., 2022) and counties (Richards et al., 2024), while data from the UK's National Fault Interruption Reporting Scheme has been used to analyse trends in outages and weather data (Shouto et al., 2024). These highlight the potential of regular, consistent and transparent reporting, which is missing in India.

As we discussed in our previous article, several independent studies in India have tried to estimate outage data, largely through household surveys (Agrawal et al., 2020; Bigerna et al., 2024; Khanna & Rowe, 2024). However, DISCOMs are better positioned to provide granular, feeder-level data in an accessible and comparable form, but as of the writing of this article, they are not mandated to make this information public. There is also no command standard of reporting, which make it impossible to make meaningful assessments. While DISCOMs are investing in redundancy systems and infrastructure, they must also clarify which recorded outages translate into consumer-facing interruptions. Doing so would, in fact, allow for a more accurate evaluation of the measures undertaken to improve reliability.

Aklin et al. (2016) had conducted a household survey in six Indian states and found that not only are outages very frequent, but that increasing the reliability of supply has effects comparable to electrifying an unelectrified household. Improving reliability of supply, however, first requires an understanding of where, when and why outages occur, which in turn requires better data. We recommend adopting a common standard of reporting outage data that includes daily, consumer-facing feeder-level outages, with information on the outage start and end times, durations, reasons, the number of consumers and the localities impacted. A first-level reason can broadly indicate whether an outage is planned or unplanned, and then provide a detailed description of the underlying cause. The data should be updated regularly and historical archives should be publicly available. This would enable more accurate and regular analyses of outage patterns, across DISCOMs and states.

References

Factors affecting household satisfaction with electricity supply in rural India by Aklin, M., Cheng, C. Y., Urpelainen, J., Ganesan, K., & Jain, A., 2016, Nature Energy, 1(11), 1-6.

Stalemate - How Consumers are Losing in the Fight Between the Regulator and Discoms in Delhi by Chitnis, A., Dmonty, A. N., & Singh, D., 2025, CSEP.

Data appendix

The data on outages was extarcted from:

• BRPL, accessed on 2 June, 2025
• BYPL accessed 7 June, 2025
• TPDDL accessed on 7 July, 2025

Delhi's daily electricity demamd was accessed from Grid-India on 7 July, 2025

The cleaned datasets and code used in this analysis are available on our GitHub repository.

The authors are researchers at TrustBridge Rule of Law Foundation. They thank an anonymous referee for useful comments.

Anchor pension policy to its core design principles

by Renuka Sane.

Pensions policy, at its heart, stems from paternalism. When people stop working, they stop earning, and if they haven't saved enough, face the risk of destitution in old age. In most societies, the state feels compelled to step in with tax-payer funded cash transfers in the form of old age pensions. Over time, these commitments have expanded to cover entire populations. Yet financing the consumption of all elderly citizens through welfare transfers is fiscally unsustainable. As a result, pensions policy has focused on how to force individuals to build wealth during their working life. This coercion is justified on the grounds that people tend to underestimate their future needs, and discount the future too heavily. Assessment of pension policy, therefore, must recognise that it flows from the state's decision to compel individuals to save for their own future consumption. The legitimacy of this coercion depends on whether it contributes towards preventing poverty in old age. This article examines pension policy from such a perspective. The design of means-tested transfers for the already impoverished elderly, while significant, is not addressed here.

Four elements of a sustainable pension

This paternalistic foundation shapes the four defining features of a funded pension system. First, participation is mandatory. Second, the savings are illiquid. Third, the structure is low cost. Fourth, it converts savings into a stable stream of income in retirement. These features make a pension different from other forms of saving or investment. Take away one of them, and the system begins to resemble an ordinary investment account rather than a vehicle for old age income security.

Let us begin with the first feature: mandatory participation, where individuals are forced to save a proportion of their monthly income into a pension account. But, compulsion requires an employment relationship so that contributions can be enforced. Extending such schemes to the informal sector where workers move between jobs or remain outside formal payroll systems is a challenge. Further, there is the question of what contribution rate to mandate: if it is set too low, the accumulated savings will be inadequate for retirement; if set too high, it will unduly constrain consumption during working life.

The second feature, illiquidity, is put in place to provide income in old age, not to finance mid-life consumption. This design feature is always contested. Individuals argue that since it is their money, they should be able to access it when they need it. Such demands become greater when contribution rates are very high. Policymakers often give in, allowing partial withdrawals or loans against the accumulated corpus. But withdrawals can leave retirees with inadequate balances, defeating the entire purpose of the mandatory contribution.

The third feature, low costs, is crucial because of the long horizon of pension saving. Fees and commissions, even if small annually, compound heavily over decades. High costs can erode a significant portion of the final corpus. Keeping costs low is especially important because participation in a pension scheme is compulsory; having coerced individuals to save, policy cannot then channel their money into high-cost funds that primarily enrich fund managers. The global experience suggests that keeping costs low requires deliberate policy design. This can be achieved through two ways.

1. Auction-based system for selecting limited fund managers (as was the case of India's National Pension System (NPS)). The larger the corpus with a fund manager, the lower the fees. For example, Vanguard S&P500 ETF has assets of about US$1.5 trillion, and an expense ratio of about 0.03% (3 bps). When fund managers are given a specific mandate to manage a large corpus, costs can be negotiated down. Pension policy, especially when the market is very small, must then decide between the competing trade-offs of multiple managers and choice vs. limited fund managers and low costs.

2. Limit investment options to low-cost passive index funds. Over long periods, index funds typically outperform most actively managed funds, net of costs. Critics argue that restricting investment options stifles innovation and deprives individuals from exercising choice. But the idea of unfettered choice has also been questioned, given that most individuals may not be equipped to make complex financial decisions. These trade-offs become relevant because of the forced nature of savings.

A reasonable middle ground lies in offering a limited number of fund managers who provide a restricted menu of low-cost index fund options. Index funds with equity exposure provide for an upside and international exposure can help reduce risk through diversification.

The final feature, a retirement income, is what completes the cycle. This can be achieved through an annuity, which converts the accumulated balance into a guaranteed stream of payments for life. In the case of an inflation-indexed annuity, the payments are adjusted for inflation throughout retirement. While annuities provide longevity insurance (and can sometimes provide inflation protection), they are often unpopular because they appear to offer poor returns and lack flexibility. Pricing of annuities may be a challenge when the bond market is itself underdeveloped. The compromise in many systems is to mandate partial annuitisation requiring that a fraction of the corpus be used to buy an annuity while allowing flexibility for the rest. The other alternative is to design systematic withdrawal plans that allow for gradual withdrawals from the corpus. This doesn't insure against longevity risk, but is often preferred for its simplicity and flexibility.

It is important to emphasise the word funded when outlining the four features. If taxpayer resources are used to finance retirement, features such as guaranteed returns can be built in. However, such arrangements are often vulnerable to funding pressures over time. If the system is to remain self-sustaining, the investment risk must rest with the individual unless that risk is explicitly priced and paid for.

Pensions in India

The central question for any government, then, is how to achieve these four features. Let's consider how the EPF, NPS fare on these parameters, especially relative to a mutual fund that is not a pensions product.

Feature	Mutual fund	EPF	NPS
Mandatory	No	Yes (for formal sector)	Somewhat
Illiquidity	No	Cannot withdraw 25% of corpus	Yes
Low cost	No	Somewhat	Yes
Retirement income	No	No	Yes

A mutual fund is not mandatory, or illiquid, or low cost. It makes no promises of a retirement income. These features are not expected from a mutual fund, as it is not a pensions product.

EPF

The Employees Provident Fund (EPF) functions effectively on the issue of mandatory contributions, as it is meant for formal sector workers, where employment is defined and contributions are linked to payroll. However, its contribution rate, around 24%, is high, making it burdensome, particularly for low-income workers. Early withdrawals from the EPF have been a persistent concern. The EPFO has recently restricted withdrawals to 75% of the corpus (and hence 25% of the corpus is illiquid till retirement), which is an improvement, but still undermines the purpose of a pension product. The EPFO needs to consider a calibration of the contribution rates. The administrative costs, borne implicitly through an employer levy of about 0.5% of wages, make it relatively expensive. Moreover, it offers no choice in investments and provides a guaranteed rate of return, which limits both flexibility and upside potential. Finally, by paying out a lump sum at retirement, the EPF exposes individuals to longevity risk. From a pension design perspective, the EPF would benefit from reforms across all four foundational elements of a pension system.

NPS

The National Pension System (NPS) did not encounter challenges of coverage when participation was mandatory for government employees. The total AUM of NPS in September 2025 was about US$178 billion (Rs. 15.8 lakh crore), of which 85% was with the three public sector fund managers (SBI Pension Funds Pvt Ltd, LIC Pension Fund Ltd, UTI Retirement Solutions Ltd). The Pension Fund Regulatory and Development Agendy (PFRDA) has capped investment management fees for all pension fund managers, which continue to be some of the lowest in the world (between 3bps - 9bps). As the total corpus grows these may further come down. The NPS permits equity exposure but limits international investments, thereby constraining diversification opportunities. The scheme allows only three partial withdrawals during the entire subscription period and limits the amount that can be withdrawn. It would do well to reserve these safeguards, which reinforce the principle of lliquidity that underpins any pension scheme. It requires artial annuitisation at retirement and offers a systematic ithdrawal plan, with ongoing efforts to design additional tructures that can strengthen income security in old age. These are steps in the right direction.

Despite these advantages, the transition to the Unified Pension Scheme (UPS) has brought forth a fundamental challenge for the NPS: building a base of contributors for whom saving is compulsory. The natural tendency will be to compete in the market for voluntary savings. More recently, under the Multiple Scheme Framework (MSF) fund managers are permitted to design and offer multiple schemes tailored for different subscriber segments. While this will allow more choice for subscribers, it runs the risk of diluting what makes the NPS a pension product. A mature mutual fund industry already caters to voluntary investors, and an excessive focus on marketing voluntary contributions risks undermining the NPS's defining advantage - its low-cost, low investment options structure.

Conclusion

Retirement schemes, whether the EPF or the NPS, are only one component of an individual's broader savings portfolio. Yet, for the portion that is locked into a dedicated pension scheme, fidelity to the four core design principles is crucial. This is especially important as both schemes, and particularly the NPS, consider various reforms related to the design of different schemes, valuation models and withdrawal options. The focus of a pension system should remain on expanding wholesale participation through large-scale group subscriptions, rather than competing directly in the retail savings market.

The author is a researcher at TrustBridge Rule of Law Foundation.