Vanishing Voters: New Delhi Lost a Quarter of its Voter Roll. Where Did They Go?

INVESTIGATION · DELHI ELECTIONS 2025:

Between the 2020 and 2025 Delhi Assembly elections, the electoral roll of New Delhi (AC-40) shrank by 37,327 names, roughly one voter in four. In February 2025, the BJP’s Parvesh Verma beat AAP chief Arvind Kejriwal there by 4,089 votes.

I found that gap, a roll cut about nine times larger than the winning margin, while working through New Delhi’s electoral data for my own investigation in March 2026. This is not an ordinary seat. Kejriwal had held it since 2013, and losing it marked the end of AAP’s decade in power in Delhi. In this article I lay out my numbers, check them against the published results, weigh the innocent explanations, and ask what one constituency’s roll says about how India keeps its voter lists. A shrinking roll is not proof of theft. But it is a question that booth-level records can settle.

1. The roll: six years of growth, then a cliff

When I tabulated New Delhi’s electors' election by election, the roll grew about 9% between 2014 and 2020, then fell by almost a third before the 2024 Lok Sabha poll. A partial recovery still left it a quarter smaller in 2025 than in 2020 (Figure 1).

In my notes I called the post-2020 loss “very strange." Urban rolls usually drift upward between elections, as this one did for six years; a one-cycle fall of this size needs an explanation the published record does not yet give.

2. The ballot box: AAP’s vote nearly halved

Parvesh Verma of the BJP won New Delhi in 2025 with 30,088 votes to Kejriwal’s 25,999, a margin of 4,089. Five years earlier Kejriwal had won the same seat by 21,697 (Figure 2).

When I marked up the two results tables, two things stood out. AAP lost far more votes than the BJP gained. And Congress candidate Sandeep Dikshit’s 4,568 votes exceeded the winning margin, a reminder that the result had more than one possible cause.

3. The turnout paradox

On paper, turnout in New Delhi improved in 2025. In fact fewer people voted: the rate rose because the roll it is measured against shrank faster than the vote did (Figure 3). The same arithmetic shows where AAP’s lost votes went. The BJP and Congress together gained 6,375 votes, so even if every one came from former AAP voters, at least 14,384 of AAP’s lost votes were simply not cast in 2025. Those missing voters either stayed home or were no longer on the roll, and the published results cannot tell the two apart.

4. Could the removals have decided the seat?

What struck me most is scale: the roll lost about nine times as many names as the margin of victory. But I have to be careful here, because a removed name is not a lost AAP vote. Some removed voters had died or moved, some would not have voted, and those who would have voted would not all have backed AAP (Figure 4).

Figure 4 reduces the dispute to one fact no published record yet gives: how many removed names belonged to living residents who would have voted. It uses the net fall in the roll; if gross deletions were larger and partly offset by new additions, the pool of possibly affected voters was larger too.

5. What AAP alleged before the vote

AAP raised the alarm about New Delhi’s roll weeks before polling. In December 2024 and January 2025, Kejriwal accused the BJP of running an “Operation Lotus” on the constituency: bulk applications to delete existing voters and to add new ones within a few weeks. By AAP’s count the applications touched roughly a tenth of the roll, and the party alleged that new voters were being registered at the official residences of BJP MPs.

The BJP rejected the charge and counter-alleged that AAP was protecting fake entries elsewhere in Delhi, including, it claimed, illegal immigrants. Election officials pointed to the rules: a deletion needs a Form 7 objection, field verification by a booth-level officer (BLO), and notice to the voter.

My figures fit a period of heavy churn. After the 2024 low, the roll grew by a net 8,192 names before the 2025 poll. Net figures cannot show how many names were struck off and how many were added in that window; only the Form 6 and Form 7 registers can.

6. What the rules say should have happened

The Election Commission of India’s position is that rolls are revised in the open. Every deletion is meant to follow a Form 7 objection, a field check by the booth-level officer, and a notice to the voter. Draft rolls go to every recognized party, whose booth-level agents can object, and a final roll can be appealed under Section 24 of the Representation of the People Act, 1950.

On paper, then, a fall on New Delhi’s scale should have left a long and checkable trail: tens of thousands of objections, field reports, and notices. Whether that trail exists, and what it shows, is the question this article cannot yet answer. The rolls themselves are not released in machinereadable form; the ECI cites voter privacy and a 2018 Supreme Court ruling.

7. The innocent explanations

Several explanations fit my New Delhi numbers without any wrongdoing, and I have to weigh them honestly:

• A stale roll being cleaned. Much of New Delhi is central government housing whose residents transfer in and out. Voters who move often stay on the old roll for years, which would also explain the constituency’s low 52% turnout in 2020. A one-time purge of absent, shifted, and dead entries would shrink the roll without removing anyone wrongly.

• Voters staying home. The BJP won Delhi with 48 of 70 seats, so the swing against AAP was citywide. If many of Kejriwal’s 2020 voters simply abstained, the pattern of fewer votes and a falling AAP share would look the same.

• Mismatched data. Wikipedia lists 148,574 registered electors for 2020, which is 2,673 more than my figure, while its own 52.45% turnout implies about 146,100. Service voters or rolls from different dates may explain this; either way the 2020–25 fall is 37,000–40,000.

Two points cut against a purely benign reading. A clean-up of stale entries removes people who were not voting anyway, yet New Delhi cast 15,009 fewer votes in 2025 than in 2020. Across Delhi as a whole, the roll grew from about 1.47 crore to 1.55 crore over the same period, and turnout slipped only about two points.

Caution is needed on my side too. My 2024 figure of 1,00,382 carries the most weight in this story, so it is the first number to verify against the ECI’s segment-wise Lok Sabha statistics.

8. What would settle it

The question is answerable, because every deletion leaves a paper trail. These records, most obtainable from Delhi’s Chief Electoral Officer or through RTI to the New Delhi Electoral Registration Officer, would settle it:

• Booth-wise deletion lists, 2020–2024, with reasons. Deletions spread across government housing point to a clean-up; deletions clustered in a few lower-income booths point somewhere else.

• The Form 7 objection register. Who filed each objection, and how many did each person file? A handful of objectors behind thousands of deletions would be a red flag.

• Proof of notice. Whether each deleted voter was sent a notice and given a hearing before removal, as the rules require.

• Booth-wise results, 2020 and 2025. Did AAP’s vote and overall turnout fall most in the booths that lost the most names?

If deleted names turn out to be overwhelmingly dead, shifted, or duplicate voters who received notice, the drop I called “strange” has a mundane cause. If not, the deletions themselves become the story, and the checks meant to protect voters need explaining.

9. The larger picture

New Delhi is one seat, but what happened to its roll points to three wider problems with how India’s voter lists are kept and reported:

• Turnout figures can mislead. When a roll shrinks, the turnout rate rises even if fewer people vote, as New Delhi’s did in 2025. Reporting votes cast alongside the rate would stop a cleaned-up roll from passing for a more engaged electorate.

• Roll changes can dwarf margins. In a close seat, one revision can add or remove far more names than the winning margin. That makes routine roll maintenance decisive, whether or not anyone intends it to be.

• The public cannot check. Deletions are recorded booth by booth, but they are not published as open data with reasons. A citizen with a notebook, as I was, can spot a drop of 45,519 names but cannot easily find out who was removed or why.

The fix does not require taking sides. If every revision published booth-wise additions and deletions with reasons, a party crying foul and an official defending the roll could point to the same numbers. New Delhi’s missing voters would then be a question with an answer, not an accusation.