The Voter Roll Purge That Helped End Trinamool Rule

On May 4, 2026, Mamata Banerjee, the incumbent Chief Minister of West Bengal, lost her own assembly seat. Bhabanipur, the constituency she had represented for over a decade, went to Suvendu Adhikari of the Bharatiya Janata Party by 15,105 votes, 73,917 to 58,812. Statewide, the BJP won 207 of West Bengal's 294 seats, ending the Trinamool Congress's fifteen years in power across three consecutive terms since 2011, and making the BJP the first extreme right-wing party ever to govern the state. 

The commentary that followed reached for the usual explanations: three terms is a long time to govern anything, the fallout from the RG Kar Medical College case had not faded, and Hindu voters had consolidated behind the BJP in ways they had not in 2021. All of this is likely true. It is also not the first time the BJP and the Election Commission have faced accusations of manipulating the vote in their favor: national opposition leaders made similar claims about the 2024 general election, under the banner of what became known as “vote chori,” or vote theft. None of it is the whole story. A careful analysis illustrates the increasing importance of vote suppression to right-wing politics as well as the perils of centralized administrative power, even in nominally independent bodies.

A brief orientation for readers less familiar with Indian politics. The Trinamool Congress is a regional party, not a national one; it governs only West Bengal, a state of roughly one hundred million people in India's east, bordering Bangladesh. Banerjee founded the party in 1998 and has run the state since 2011, building a base heavily concentrated among Muslim voters (about twenty-seven percent of the state's population) and the rural poor. Just like how the Prime Minister is the head of the central government, the Chief Minister is the head of the state government. The BJP is India's governing national party, led by Prime Minister Narendra Modi, and has spent over a decade trying and failing to break Trinamool's hold on Bengal. The Election Commission of India (ECI) is a constitutionally independent body, roughly analogous to a combination of the Federal Election Commission and every US state's chief election official rolled into one, which administers elections directly rather than merely regulating campaign finance. Its independence from the ruling party is a foundational assumption of Indian democracy, which is precisely why accusations that it has been used as a partisan instrument are as serious as they are.

It is worth pausing on how different this is from anything in American electoral administration. The United States has no single national body that runs elections. That task is left to roughly 10,000 county and municipal offices operating under state law, with voter roll maintenance governed loosely by the National Voter Registration Act of 1993, which requires states to remove ineligible voters but also sets real limits on how and when they can do it, including a ban on systematic removals within ninety days of a federal election. Nothing in American practice resembles a single nationwide agency conducting a compressed, house-to-house re-verification of an entire state's electorate within a six-month window immediately before that state's election. The closest analogues Americans might recognize are large-scale voter roll purges in states like Georgia or Ohio, or the long-running court fights over such purges, which operate at a fraction of the Indian scale—and which are subject to a federal-court oversight that has no real equivalent in the Indian system, where the ECI's decisions face a much higher bar for judicial reversal. Even Jim Crow’s regime of vote suppression in the Southern states was always piecemeal, as a result of America’s decentralized federalist system.

Between October 2025 and April 2026, the Election Commission of India ran a Special Intensive Revision (SIR) of West Bengal's electoral rolls, the state's first such exercise since 2002. It removed roughly 9.1 million voters, about 11.6 percent of the state's electorate, ahead of the election that unseated Banerjee. The question worth asking is not whether SIR alone decided who governs West Bengal. It did not, and I want to be honest about that from the start. The question is whether the purge was large enough, and patterned enough, to have mattered in a meaningful number of the seats that decided the state's next five years.

The mechanics of an inversion

SIR does not work the way most people imagine an electoral cleanup working. It does not simply check names against a database and flag duplicates. It flips the basic presumption that has governed Indian electoral rolls for decades. Previously, a voter already on the rolls stayed there unless the state affirmatively showed cause to remove them. Under SIR, the burden runs the other way: anyone not on the 2003 roll, the last full revision before this one, has to produce documentary proof of eligibility, often including proof of a parent's eligibility as well. For a country where birth registration has been historically uneven (only 57.7 percent of births were registered nationally as of 2003, the very year SIR uses as its baseline), and where enormous numbers of people migrate for work, marry across district lines, or simply never had reason to keep decades-old paperwork, this is a burden that falls unevenly by design, whatever the intent behind it.

West Bengal's version of the exercise added a wrinkle that appears nowhere in the Representation of the People Act: a deletion category called “logical discrepancy.” Officials used it to flag and remove names over transliteration variants in Muslim surnames, the placement of abbreviations like “Sk.” or “Md.,” and age gaps between parents and children common in larger families. None of these are evidence of fraud. They are evidence of what happens when a rigid bureaucratic filter, applied at speed and at scale, mistakes ordinary variation for genuine fraud.

What the data shows and does not show

Three independent post-election analyses have tried to quantify what this meant for the outcome. Scroll compared, seat by seat, the number of voters deleted against the BJP's winning margin, and found that in 105 of the party's 207 wins, deletions outnumbered the margin; eighty-six of those were seats the BJP had never previously carried. Outlook India ran a stricter version of the same test, isolating deletions not explained by a voter's death, on the theory that those are more likely to reflect documentation disputes than genuine roll-cleaning. Even by that narrower measure, non-death deletions alone exceeded the BJP's remaining margin in thirty-one constituencies. Thirty-one seats decided, plausibly, by a documentation dispute is not a rounding error in a state legislature.

A more rigorous test comes from a regression Scroll ran on the relationship between roll shrinkage and the shift in the BJP-Trinamool margin from 2021 to 2026. Each percentage point of roll shrinkage in a constituency was associated with a 0.377-point shift in the margin toward the BJP, a relationship the authors describe as statistically strong. Applied to Bihar's earlier and considerably smaller SIR, the same test found no comparable pattern, which suggests the Bengal effect was not an artifact of SIR in general, but of the particular scale of what happened there. The median Bengal constituency lost 9.6 percent of its electorate; the median Bihar constituency lost 3.0 percent.

This is not simply an assumption that every deleted name was a certain Trinamool vote. The deletions were not distributed evenly across the state: they were concentrated in Muslim-majority and migrant-labor districts, precisely the demographic core of the Trinamool's own coalition, and the same population the “logical discrepancy” category was, in practice, most likely to flag. A purge that fell evenly across the electorate would not predict a seat-level swing toward the BJP; one concentrated in the incumbent party's base would. That the data shows exactly the latter pattern is what makes the statistical relationship more than a coincidence of timing.

I do not want to overstate this. The same analyses that found these patterns also found real limits to how far the argument can go. The BJP's own vote share grew enough, in eighty-six of the 122 seats that flipped from the Trinamool, that the party would likely have won those seats even if every deleted voter had cast a ballot the other way. Turnout hit ninety-four percent, the highest ever recorded for an Indian state election, which complicates any simple story of a suppressed electorate. And no analysis, however careful, can tell us with certainty how the roughly 2.7 million voters still under adjudication when polling began would have actually voted.

Why the nuance is important

Liberal readers, myself included, are drawn to clean causal stories, especially ones that confirm what we already suspect about an illiberal governing party. I understand the pull of “the BJP rigged the election,” and after the data I have just walked through, I understand why people reach for it, seeing the BJP's own electoral record in past general elections alongside allegations of mass voter fraud being put forth by several bodies. But I think the more precise claim is also the more damning one: SIR did not need to manufacture a Trinamool defeat that would not otherwise have happened in order to be a serious problem for anyone who takes universal suffrage seriously. It needed only to be what the evidence shows it was: an administrative procedure whose costs were not distributed neutrally, that fell disproportionately on the incumbent's own electoral base, and that was large enough in a meaningful number of close seats to have plausibly changed who represents West Bengal for the next five years.

The Supreme Court, in upholding SIR's constitutionality on May 27, 2026, drew a careful distinction: the Election Commission can weigh citizenship claims to decide who belongs on a voter roll, but it cannot make a final determination of someone's citizenship. That distinction is legally sound and, I would argue, almost beside the point. A family that loses its vote, and in some Bengal districts its access to welfare benefits tied to electoral roll status, does not experience the difference between “removed from the rolls” and “declared a non-citizen” as a meaningful one. A right that survives only if you win an appeal after the fact, on a timeline set by the state and often too short to navigate (as of mid-July 2026, only about 38,000 of 3.4 million pending appeals had been decided, across just nineteen tribunals, two of whose judges had already resigned), is a right held on sufferance rather than a right held in fact.

The American case

SIR is not only a story about India. It is also a live argument about a design question American readers have their own stake in: whether elections are safer run by one accountable body, or by 10,000 uncoordinated ones.

American election administration is, by design, a mess. Elections are run by roughly 10,000 overlapping county and municipal offices, under fifty different sets of state law, with no single body capable of doing to the American electorate what the Election Commission of India just did to West Bengal's. This is not an accident of neglect; it is federalism doing what federalism is supposed to do, which is to make it structurally difficult for any single actor, however determined, to capture the machinery of elections all at once. It is also, historically, no guarantee of anything. That same decentralization let Jim Crow states run their own SIR in miniature for the better part of a century: poll taxes, literacy tests, and purges administered county by county, safe from federal interference precisely because no single office ever held enough of the picture to be stopped by any single court order or federal law.

So decentralization has already failed once, slowly, unevenly, and for a very long time. But SIR is a portrait of the other failure mode: not slow but immediate, not local but total, a single captured authority reordering an entire electorate's rights inside of six months, upheld by a Supreme Court either unable or unwilling to say no. Which is the more dangerous design flaw is not a settled question, and it should not be treated as one, especially not now. Donald Trump has spent his second term testing how much of the machinery of American government a sufficiently determined executive can bend toward his own ends, and the administration of elections has not been exempted from that testing. A single election authority, like the Election Commission of India, does not exist in the United States. But West Bengal shows, with unusual clarity, what it would look like if one did, and who might end up holding it. That is reason enough to stop treating federalism's inefficiency as merely an embarrassment to be reformed away.

Featured image is All India Trinamool Congress Office - Ganganagar - Mukundapur - Kolkata, by Biswarup Ganguly