Who Is Really Stealing India s Electricity
The most familiar image of power theft in India is a hooked wire on a distribution line. The kunda is visible, illegal, and easy to photograph. It gives television crews a simple villain and gives distribution companies, or discoms, a clean enforcement story.
But it does not answer the harder question: who steals the most electricity by volume?
India does not publish that answer in a form the public can test. There is no regular, comparable disclosure of theft by consumer category, feeder, sanctioned load, assessed units, and recovered amount. What we get instead is a blended number: Aggregate Technical and Commercial loss, better known as AT&C loss.
That number matters. India’s AT&C loss stood at 15.04 percent in 2024-25, according to the figure cited in the brief. But AT&C loss is not a theft number. It mixes several different failures into one percentage: energy lost in wires, power stolen, supply that was never billed, and bills that were raised but not collected.
That is convenient for system accounting. It is poor evidence for public blame.
The debate on electricity theft in India often starts with the poor household. It should not end there. A small illegal domestic connection may be widespread, but a single tampered industrial meter can consume more unpaid electricity than hundreds of hooked wires. The public record gives enough hints to question the stereotype. It does not give enough data to settle the issue.
That gap is the story.
AT&C loss tells us the system is leaking, not who opened the tap
AT&C loss is a useful management indicator. It asks a simple question: of the electricity input into the distribution system, how much money did the discom finally collect?
That chain can break in several places.
Power can be lost technically as current moves through lines and transformers. Old conductors, overloaded feeders, low-voltage networks, and poor maintenance all raise technical losses. These are engineering problems, not theft.
Power can also be consumed but not billed. This can happen through direct tapping, meter bypass, meter tampering, defective meters, poor meter reading, billing errors, or unmetered categories where consumption is estimated rather than recorded.
Then comes collection. A discom may bill a consumer but fail to collect the money. That unpaid amount also enters the AT&C framework.
So when a state reports a high AT&C loss, the number does not tell us how much was stolen by a slum household, a commercial complex, a farm pump, a factory, or a government department. It only tells us that the distribution business failed to convert supplied energy into collected revenue.
That distinction matters because policy responses differ.
If technical losses dominate, the answer is network investment, feeder separation, better transformers, and system planning.
If theft dominates, the answer is metering, inspection, prosecution, political backing, and better local intelligence.
If billing and collection dominate, the answer is process reform, public payment discipline, prepaid meters, and disconnection rules that apply evenly.
If unmetered subsidised supply dominates, the answer is metering, direct benefit transfer, subsidy budgeting, and honest accounting.
A single AT&C number cannot choose among those answers. It can only tell us to look harder.
AT&C loss is a warning light on the dashboard. It is not a forensic report.
The public discussion often treats AT&C losses in India as if they were a synonym for theft. That helps nobody. It hides technical weakness behind moral outrage. It also lets large, politically protected, or hard-to-audit consumers disappear inside a system average.
The result is a distorted enforcement culture. Discoms can show raids, photographs, seized wires, and case counts. Those things are not meaningless. But they are not the same as a category-wise theft audit in units of electricity.
And units are where the real story lies.
Case counts can mislead because loads are not equal
A theft case is not a standard unit of harm. One case can be a hut running two bulbs and a fan. Another can be a plastic moulding unit drawing power for machines through a direct supply connection. Counting them equally may be useful for legal reporting, but it is a poor way to understand financial damage.
The enforcement data cited in the brief shows this clearly.
Maharashtra’s discom reported 27,964 theft cases worth ₹66.28 crore in three months. That is roughly ₹24,000 per case on average. A number like that can suggest small-ticket theft. But averages flatten outliers. They do not tell us whether the loss came from many small domestic cases, a handful of high-load commercial and industrial cases, or some mix of both.
Now compare that average with selected enforcement examples.
Reported case | Location or category | Assessed value |
Maharashtra discom theft cases over three months | 27,964 cases | ₹66.28 crore total, about ₹24,000 per case |
Plastic moulding unit with direct supply connection | Mumbai | Over ₹1 crore |
Electroplating shop with direct supply connection | Mumbai | Over ₹1 crore |
Plastics factory | Patna | Over ₹20 lakh |
Cotton ginning unit | Rajasthan | About ₹1.24 crore |
These figures do not prove that industry steals more electricity than households across India. They do prove something narrower and important: the value per case can vary by orders of magnitude.
A photograph of a hooked domestic wire may carry more public anger than a sealed industrial meter box. But the load behind the meter box can be far larger. An electroplating shop, plastic moulding unit, cold storage site, welding shop, rice mill, stone crusher, or cotton ginning unit can draw electricity for hours at loads far above a household connection.
That changes the economics of theft.
A poor household may steal because legal supply is unaffordable, unreliable, or unavailable. That is still theft, but its daily consumption may be small.
An industrial or commercial consumer who bypasses a meter may be avoiding not only the energy charge but also demand charges, cross-subsidy charges, fixed charges, penalties, and taxes embedded in the tariff. The incentive can be large because power is a major input cost. The unpaid units can also be large because the load is large.
This is why the public needs three numbers for every theft category:
Number of cases
Assessed units of electricity
Assessed value and recovery
Without all three, the picture is incomplete.
A category with many cases may not dominate stolen units. A category with fewer cases may impose a larger financial hit. A category with modest assessed value may still matter if recovery is poor. Policymakers cannot rank interventions without separating these effects.
That is also why power theft and discoms’ enforcement data should be read with caution. A raid database reflects where the utility looks, what it can prove, and what it can safely act on. It does not automatically reflect where the largest theft occurs.
The visible illegal connection is often the easiest case to catch. The largest theft may be the hardest case to prove.
The poor household is visible, but not necessarily the largest thief
The kunda became the symbol of theft because it is easy to understand. It involves a physical hook, a wire, a pole, and a household that is visibly outside the billing system. It also fits a long-running political story: honest bill-paying consumers bear the burden while others consume for free.
There is truth in that story. Domestic theft exists. In some areas, it is socially tolerated. In some neighbourhoods, meter readers face intimidation. In others, illegal connections are organised by local intermediaries who collect money from residents while the discom loses revenue.
No serious reform can ignore household theft.
But the data problem begins when the symbol becomes the whole explanation. Older state records have shown domestic connections topping case tables. That can mean domestic theft is widespread. It can also mean domestic theft is easier to detect, easier to count, and more likely to produce many small cases.
A better question is not “which category has the most cases?” It is “which category accounts for the most stolen units and unrecovered revenue?”
On that question, public evidence is thin.
The brief points to another important clue. Punjab figures reported by The Tribune placed theft in rural tubewell areas far above industry. That does not fit the simple “factory bad, poor household blamed” story either. It suggests that the truth varies by state, feeder type, tariff design, political economy, and metering quality.
That is exactly the point. India’s power theft problem is not one story. It is several different stories hidden under one loss number.
In one state or region, domestic hooking may dominate reported cases. In another, farm feeders may carry a large share of unmetered or under-recorded consumption. In an industrial cluster, meter tampering or direct tapping by high-load users may account for a large financial impact. In a commercial district, shops may understate load or manipulate meters. In some places, local government dues and weak collection may be as serious as theft.
A national stereotype cannot substitute for feeder-level numbers.
The household hook remains politically useful because it is visible. It can be raided, cut, filmed, and condemned. It does not require the public to ask how large consumers are inspected, how meters are tested, how assessments are calculated, or why category-wise theft volumes are not disclosed.
It also shifts attention away from tariff politics. Urban consumers paying full bills often cross-subsidise other categories through higher tariffs. If theft and unmetered consumption blend into the same system loss, the paying consumer cannot know which burden they are carrying: genuine subsidy, technical inefficiency, theft, poor billing, or non-collection.
That weakens trust in the entire distribution system.
Unmetered farm supply makes the theft debate harder
Agricultural power sits at the centre of India’s electricity politics. In many states, farm electricity is free or heavily subsidised. In several places, consumption is not metered at the pump level. Instead, utilities estimate use through feeder readings, connected load assumptions, supply hours, or administrative norms.
This creates a measurement problem even when nobody steals.
If a pump is unmetered, the system does not know exactly how many units it consumed. That consumption must still be accounted for. It may be booked as subsidised supply. It may be estimated as agricultural consumption. It may affect loss calculations if input energy and billed energy do not align cleanly. The boundary between poor measurement, subsidy, and theft becomes blurred.
This is not a small accounting issue. Agricultural supply can form a large share of electricity sales in several states. Pumps often draw significant loads. Supply may be rationed by hours, but actual use depends on groundwater depth, crop choice, pump efficiency, local voltage, and farmer behaviour.
When metering is weak, several distortions follow.
Discoms may overstate agricultural consumption to hide other losses. State governments may underfund subsidies because consumption estimates are contested. Farmers may distrust meters because they fear future billing. Paying consumers may assume all losses are theft. Reformers may push technical fixes without facing the political bargain around cheap power.
The result is a category that is neither fully commercial nor fully transparent.
This matters for subsidised electricity in India because subsidy and theft should not be treated as the same thing. A subsidy approved by a state government is a policy choice, even if it is inefficient or poorly targeted. Theft is illegal consumption. Unmetered supply is a measurement failure that can hide either legitimate subsidised use or illegal diversion.
If these are not separated, every side can choose the number it prefers.
A discom can blame theft. A government can blame technical loss. A consumer group can blame free power. A farmer group can blame faulty accounting. An industrial association can blame cross-subsidy. Each may be partly right. None can be tested without better data.
This is why feeder-level energy accounting is central. A feeder that mostly serves agriculture should show input energy, supply hours, estimated or metered consumption, transformer capacity, connected load, and losses. A mixed feeder should not be used to make sweeping claims about farm theft. Pump-level metering, where politically feasible, is even better. Where it is not feasible, feeder metering still gives a sharper picture than a state-level AT&C average.
The goal is not to criminalise farmers. The goal is to stop using farm supply as a fog bank inside which everything else disappears.
Factories, farms, shops, and homes need the same audit lens
A fair theft audit would not start with the assumption that one category is guilty. It would start with energy flow.
How much electricity entered the feeder? How much passed through distribution transformers? How much was billed by consumer category? How much was collected? How much was assessed as theft after inspection? How much of that assessment was recovered? How many cases were compounded, litigated, or written off?
Those questions can be answered. Discoms already collect much of the underlying information. Smart meters, feeder meters, GIS mapping, and energy accounting systems make the task easier than it was a decade ago. The barrier is not only technical. It is institutional and political.
Publishing category-wise theft data would make many people uncomfortable.
It could show that small domestic consumers dominate cases but not units. It could show that certain industrial clusters have unusually high losses. It could show that agricultural estimates are inflated or understated. It could show that government connections do not pay on time. It could show that enforcement is concentrated where resistance is weakest.
That is precisely why the data should be public.
A useful disclosure format would be simple enough for citizens and detailed enough for regulators. At minimum, every discom should publish quarterly theft and loss data with these fields:
Data field | Why it matters |
Consumer category | Separates domestic, commercial, industrial, agricultural, public supply, and other users |
Feeder or division | Shows whether losses are localised rather than spread across a state average |
Number of theft cases | Measures enforcement activity, but not impact by itself |
Assessed units | Shows the volume of electricity involved |
Assessed amount | Shows the financial claim raised by the discom |
Amount recovered | Shows whether enforcement produced cash, not only paperwork |
Connected load involved | Helps compare small and large cases |
Theft method | Separates hooking, meter tampering, bypass, direct supply, and other methods |
Billing and collection efficiency | Distinguishes theft from non-payment |
Technical loss estimate | Keeps engineering losses separate from commercial losses |
This would not be perfect. Theft assessments can be disputed. Meter testing can be challenged. Some illegal consumption is never detected. But imperfect category data is still better than a blended figure that supports every claim and proves none.
Regulators should also require independent checks. A discom’s own theft figures can reflect incentives. If managers face pressure to reduce losses, they may reclassify, under-report, or overstate categories depending on what helps. Third-party energy audits, random meter testing, feeder sampling, and public dashboards can reduce that risk.
There is also a legal dimension. Theft cases often move through special courts, compounding processes, and settlement mechanisms. If the public sees only the number of cases filed, it cannot know how many resulted in recovery. A high-value industrial power theft case that ends in low recovery is not a success story. It is a warning about enforcement weakness.
The same standard should apply to every category. A domestic illegal connection should be counted. A shop bypassing a meter should be counted. A factory drawing direct supply should be counted. A farm feeder with unexplained losses should be investigated. A government office that does not pay should not disappear from the accountability chain.
Fairness requires equal measurement before equal punishment.
Better data would change the politics of electricity reform
India’s power sector has lived for years with a familiar cycle. Discoms lose money. State governments announce support. Tariffs rise for some paying categories. Loss reduction targets are set. Raids take place. AT&C loss improves in some years and worsens in others. The deeper political economy remains.
Transparent theft data would not fix every problem. It would change the terms of debate.
For urban bill-paying consumers, it would show what share of the burden comes from theft, subsidy, technical loss, and non-payment. That matters because anger without detail often turns into broad resentment against poorer consumers or farmers. Data can make the grievance more precise.
For discoms, it would identify where enforcement gives the highest return. If a few high-load feeders account for large unpaid units, inspection teams should not spend most of their time chasing small users for headline counts. If domestic hooking is genuinely the biggest loss source in a division, that should be visible too.
For regulators, category-wise data would make tariff orders more honest. Cross-subsidy debates often assume clean categories of paying and subsidised consumers. Theft complicates that. If industrial consumers complain about high tariffs while some industrial clusters show serious theft, regulators should know. If agriculture is blamed for losses based on weak estimates, that should also be challenged.
For state governments, transparent data would expose the cost of political choices. Free or cheap power is a budget decision. It should be funded openly. It should not be hidden as a distribution loss or passed silently to other consumers through poor accounting.
For honest consumers in every category, it would strengthen the case for fairness. Many farmers, households, shopkeepers, and factory owners pay what they owe. They suffer when their category is blamed as a block.
The main fear is that transparency will create political conflict. But the present opacity already creates conflict. It simply directs it towards the easiest target.
India does not need a moral panic about the poor household’s hook. It needs a theft balance sheet.
That balance sheet should answer five questions in every discom area:
Which consumer category has the most theft cases?
Which category accounts for the most assessed stolen units?
Which category accounts for the highest assessed value?
Which category has the lowest recovery after assessment?
Which feeders show losses that cannot be explained by technical factors or approved subsidies?
Only then can enforcement become proportionate.
The poor household with a kunda should not be excused. Theft raises costs for everyone, damages local networks, worsens outages, and erodes payment discipline. But a system that sees only the hook and not the high-load bypass is not serious about loss reduction.
A theft policy built around visibility will over-punish what is easy to photograph and under-measure what is costly to confront.
The answer to “who is really stealing India’s electricity?” may differ by state, feeder, and season. In some places, households may dominate. In others, farm supply may hide the largest unexplained gap. In industrial belts, a small number of large consumers may cause losses that dwarf thousands of petty cases.
The country cannot know because the country is not shown the data.
Until discoms publish theft by category, feeder, units, assessed value, and recovery, India will keep debating shadows. The kunda will remain the face of theft because it is visible. The factory meter, the unmetered pump, the commercial bypass, and the unpaid institutional bill will stay inside a percentage.
The real test is not whether India can catch the poor household stealing power. It is whether the system is willing to count everyone with the same precision.
