Cleaning Voter Rolls Without Losing Voters

Millions of names left India's rolls. Only a transparent statistical audit can show whether the removals were fair.

Whenever a very large number changes in public life, people ask the same question: how do we know it is right? In India's voter lists, that question now covers tens of millions of names.

Reports compiled from Election Commission data suggest that the Special Intensive Revision removed about 91 lakh names in West Bengal and around 2.04 crore in Uttar Pradesh between late 2025 and April 2026. The exercise, announced in October 2025, covered many states. The Commission says it removed duplicate, dead and ineligible entries. Critics say that many genuine voters, especially poor, migrant and minority voters, were struck off. This week, the argument spilled onto the streets, with protests demanding the removal of the Chief Election Commissioner.

Both sides sound sure of themselves. Neither can prove its case with the information the public has. The way to settle it is not another round of statements. It is an audit built on simple, solid statistics. This article explains what such an audit would look like and why it matters.

Why large deletions are not proof of anything

First, a point of fairness. Large deletions do not by themselves show wrongdoing. India's voter lists have long had problems. People die and are not removed. People move between cities and are listed in two places. Some names are duplicated, and some entries are false. A roll that is never cleaned becomes less accurate, and an inflated roll can be used for fraud. Cleaning is good practice, and the opposition itself has complained about ghost voters in the past.

At the same time, cleaning can go wrong. People can be removed because a form was lost, a document did not match, a house was locked during a visit or a name was spelled differently. People who migrate for work, who live in informal housing or who find paperwork hard are more at risk. A system can have good intentions and still make mistakes that fall unevenly.

So the right question is not "how many were removed?" It is "how many of those removed should not have been?" and "how many who should have been removed are still on the list?"

The two kinds of error

Any clean-up exercise can make two kinds of mistake.

Wrongful removal. An eligible, living voter living at the listed address is taken off the roll. This is the harm that matters most in a democracy, because it can deny a person the vote.

Wrongful retention. An ineligible, dead or duplicate entry stays on the roll. This weakens the accuracy of the list and opens the door to fraud.

A good audit measures both. Looking only at wrongful removal can miss ghost entries that remain. Looking only at wrongful retention can miss voters who were wrongly dropped. Fairness demands that both be counted and that the figures be published.

How a sample audit works

The idea is simple and does not require checking every name. Statisticians use samples all the time, in opinion polls and quality control. If the sample is chosen at random, it can tell us a great deal about the whole.

Here is how it could work.

1. Pick the list. For each state, take the complete list of deleted names, with the reason given for each.

2. Draw a random sample. Choose a few thousand names at random, using a method that anyone can check.

3. Visit and verify. Independent teams go to the listed address, check records and speak to neighbours and local officials to find out if the person is alive, resident and eligible.

4. Count the results. Record how many were rightly removed and how many were not.

5. Report with a margin. Publish the share of wrongful removals, with a margin of error.

To see how powerful this is, consider a simple example. If a team checks a random sample of 10,000 deleted names, the margin of error on the result is about one percentage point. If it finds that 3% were removed wrongly, then the true share across the whole list is likely to be somewhere between 2% and 4%. Applied to a list of 2 crore deletions, that would be a large number of people, and it would be possible to say so with evidence. If it finds the share is well under 1%, the Commission can say that, too, with confidence.

That is what makes a sample audit valuable. It replaces competing claims with a measured number.

Making it fair: splitting the sample

An overall figure can hide uneven effects. A fair audit should break the results down by groups. If wrongful removals are 1% across the state but 8% in one district, or among one kind of voter, that matters.

Useful splits include the following.

• District and constituency. Are errors clustered in particular areas?

• Rural and urban. Do villagers or city dwellers face more problems?

• Migrant-heavy areas. Are workers who move for jobs more likely to be removed?

• Gender. Are women, who may change address after marriage, removed more often?

• Reason for deletion. Are "dead", "shifted" and "duplicate" equally accurate?

The audit does not need to name individuals. It can use anonymised data, protecting privacy while showing patterns.

What else the audit should measure

Counting errors is only one part. Several other measures give a fuller picture.

• Claims and objections. How many people filed claims to be added back, how many were accepted, how many rejected and how many are still pending?

• Time taken. How long does it take for a wrongly removed person to be restored? A person who gets back on the list after the election has still lost the vote.

• Documents required. Which papers were demanded, and how many people failed to produce them?

• Notice. Were people informed before removal, and how?

• Outreach. How many door-to-door visits were made, and how many were repeated if the person was absent?

Each of these shows how easy or hard it is for an ordinary person to protect their place on the roll.

Who should run the audit

Credibility depends on independence. If the Commission audits itself, many people will doubt the result, however careful the work. A better design would include:

1. An independent lead. A retired Supreme Court or High Court judge, or a statutory auditor, to oversee the exercise.

2. Technical experts. Statisticians and survey specialists to design the sample and check the method.

3. Field teams from outside the Commission. For example, staff from universities, state audit offices or other neutral bodies.

4. Observers. Representatives of political parties and civil society who can watch the process.

5. Public methods. The sampling method, questions and tools published before the work starts.

The Commission itself should welcome this. If its work was sound, an independent check is the best way to prove it. If errors are found, they can be fixed before the next election.

What the Commission has said, and what is missing

The Commission's public position is that the revision removed duplicate, deceased and ineligible entries and made the lists more accurate. That claim may well be partly right. But it is not backed by published data that outsiders can check.

There is also the matter of internal concerns. Media reports say that the two other election commissioners raised concerns at least 14 times in ten months about voter-list revisions, registration procedures and digital systems. The Commission has not publicly set out the substance of those concerns or its replies. Releasing them, with limited redactions, would let the public judge whether they were routine or serious.

What the opposition and parties must do

Parties have a part to play, too. They have booth-level agents who can file claims and objections during a revision. Where large numbers of real voters were removed, parties should show what they did to challenge it. If their machinery failed, they should say so.

It also helps if parties support an audit even if it shows the results are good. A genuine concern for voters means wanting to know the truth, not only the answer that suits.

Why the audit matters beyond this election

The most important reason is that elections depend on trust. Voters must believe that the list is accurate and that the count will be fair. When they doubt it, even close results are contested, and turnout and faith in the system suffer.

A transparent audit would help in several ways.

• It would show whether the exercise worked, and where it needs fixing.

• It would protect voters who were wrongly removed, by finding and restoring them.

• It would protect the Commission from unproven charges.

• It would set a standard for future revisions in other states.

• It would lower the temperature of the national debate.

The cost, and the timeline

An audit of this kind is not cheap or instant. Teams must travel, verify and analyse. But it is small compared with the cost of an election, and the benefit is large. A sample audit for a state could be designed in weeks and completed in a few months. The Commission could begin with the states where concerns are greatest, and publish results as they come in, so that findings can inform the next round of revisions.

A better way to argue

For now, the argument about India's voter rolls is a contest of claims. One side says the rolls were cleaned. The other says voters were stolen. Both may contain some truth. Without data, neither can be proven.

The remedy is within reach. Publish the lists of deletions and the reasons. Draw a random sample. Send independent teams to check. Report the errors, in both directions, with a margin of error. Show the differences between groups. Publish the methods so anyone can test them.

If the numbers show the exercise was fair, the country can move on. If they show problems, the country can fix them. Either way, trust grows. A democracy that wants to be confident about its elections should not fear a count of its own voters. It should want one.