Introduction
Here’s a number worth sitting with: 9.1 million. That’s roughly how many names quietly vanished from West Bengal’s voter rolls in the months leading up to a state election in 2026â more people than live in Switzerland, close to 12 percent of the state’s registered electorate. Gone, before a single vote was cast.
The Election Commission called it routine housekeeping: cleaning up “logical discrepancies” in the database. Duplicate entries, deceased voters, people who had moved. The kind of unglamorous administrative task election authorities everywhere handle from time to time. Nothing to see here.
Except once you look closely at who actually got removed, the story stops sounding like housekeeping. And once you understand what happened in Bengal, you start noticing the same pattern elsewhere â different countries, different parties, different decades, but the same basic move.
An Election That Looked Normal From a Distance
When the votes were finally counted, the incumbent regional party â which had governed the state for roughly fifteen years â lost decisively to its national rival. From a distance, this looks like an ordinary democratic transition. Voters get tired of one party, they try another. Nothing inherently suspicious about an incumbent losing.
I don’t think that’s the whole story, though. The conventional framework people reach for is “democratic backsliding” â a country was functioning as a healthy democracy, then something went wrong, and things slid backward from a recoverable baseline. I’d suggest a more precise way of thinking about it: nothing slid backward in the usual sense. The playing field was redrawn before the contest even began, using administrative tools so dry and bureaucratically uninteresting that almost nobody outside the directly affected communities noticed until votes were already being counted.
I think of this as technocratic majoritarianism. The form of democracy stays completely intact â citizens still vote, ballots still get counted honestly, courts still function, opposition parties still campaign freely. But the terms of who actually gets to participate are quietly rewritten through paperwork and software, well before the campaign’s noise and spectacle ever begin. It’s a subtler kind of manipulation than the ones democracy-watchers traditionally look for, and that subtlety is exactly what makes it effective.
How a Database Decides Who Counts as a Citizen
The mechanism that removed those millions of names goes by an unassuming bureaucratic label â a “Special Intensive Revision” of the electoral rolls. It works by cross-checking voter records against other government databases, flagging entries where a name, address, or identification number doesn’t quite match across documents. Flagged entries are supposed to receive an individual hearing before being struck off. That’s the theory.
In practice, an extraordinary number of these deletions â by some estimates well over half â happened with no hearing at all. And the people flagged in disproportionate numbers belonged overwhelmingly to one religious minority, concentrated in border districts where documentation has always been messier.
Call it the transliteration problem. Think about how a name gets spelled across decades of record-keeping in a multilingual, postcolonial bureaucracy: one way in colonial-era land records, differently after a border was redrawn and people were displaced, differently again on an old voter card, differently still on a recently issued biometric ID. You end up with several spellings of the same person, generated by a system never built with much care for naming conventions outside the dominant script. Mohammed becomes Mohammad or just Md. Karim becomes Kareem. None of this is fraud. It’s just what happens when a country with multiple scripts and a long history of displacement tries to keep consistent records over a century.
An algorithm doesn’t carry that history with it, though. It sees a mismatch and flags the entry as suspicious. A completely legitimate voter, whose family has lived in the same village for generations, ends up treated as a potential fraud because a clerk decades ago spelled a name differently than a more recent one did.
By election day, an enormous number of appeals remained unresolved. The process technically existed. But if you’re a laborer in a remote border district without easy access to a lawyer or reliable internet, a process that exists only on paper might as well not exist.
Why “ Deception” Isn’t Quite the Right Word
There’s a real temptation to just call all this deception and leave it there. I don’t think that’s the most useful framing, and the distinction matters more than it might first appear.
Deception, in the way most people use the word, implies that someone sat down and deliberately falsified an outcome. Nobody stuffed a ballot box here. The votes that were cast were counted honestly. What happened instead was quieter, and in some ways more unsettling: the population that got to participate was reshaped through a process entirely defensible according to the election authority’s own published rules, even though its effects landed overwhelmingly on one particular community.
Analysts who study hybrid political systems â places where elections happen but the underlying competition isn’t quite fair â have spent decades cataloguing the usual tools incumbents use to tilt the field: controlling media coverage, harassing opposition figures, redrawing electoral districts, packing courts with loyalists. These tactics share something in common: they’re visible. You can point to them, document them, build a case around them.
What happened with this voter roll revision is more like second-order manipulation. Nobody is harassing a journalist or banning a newspaper. The mechanism is buried inside software design choices almost nobody outside a small technical team ever scrutinized, made by engineers who were probably just trying to solve a database deduplication problem without much thought for the downstream political consequences. The discriminatory effect, in aggregate, is completely real. But you would struggle to point to any single individual and say with confidence, this person intended this outcome.
If this pattern sounds familiar, it should. Voter identification requirements introduced in various jurisdictions over the years have worked through remarkably similar logic â neutral on paper, disproportionate in effect, nearly impossible to challenge legally as deliberate discrimination because nobody had to write the discriminatory intent into the legal text. What’s different about the algorithmic version is mostly speed and deniability. A bureaucratic process that used to take years and generate a visible paper trail can now happen in months, hidden behind the reassuring language of database hygiene.
A Party That Built Its Own Trap
It’s worth saying clearly that none of this happened to a ruling party that was otherwise running a clean, well-functioning government. The incumbent party’s own internal decay played a significant role in setting up the conditions for its eventual defeat.
When this party first came to power, displacing a long-ruling left-wing coalition that had governed for three decades, it inherited a particular style of governance: the ruling party doesn’t simply compete for votes every few years, it runs day-to-day civic life â local unions, village councils â all threaded through with party loyalty. The previous government held this together through old-fashioned discipline and a structured cadre system. The new ruling party kept the architecture but discarded the discipline, replacing it with local power brokers operating with substantial independence, welfare resources flowing through informal intermediaries. This worked brilliantly in the short term, but it created a problem that proved impossible to manage: there was no longer any reliable way to distinguish a local operator building the party’s strength from one simply enriching himself.
Eventually the party split into two factions that found it nearly impossible to coexist â a younger wing wanting to professionalize and cut out the local strongmen, and an old guard who actually controlled the ground-level machinery that delivered votes. What most outside commentary misses: those informal networks weren’t a side effect of how this party governed. They were the engine of its power. You couldn’t purge them without dismantling what had kept the party in office.
Corruption scandals involving senior cabinet ministers eventually gave courts legitimate grounds to intervene, and they did. The corruption was real. But the pattern of who got investigated told its own story: central agencies moved quickly against figures still loyal to the ruling party and noticeably slower against politicians who’d recently defected to the opposition. The ruling party found itself unable to fight this framing, since it couldn’t honestly deny the underlying corruption existed.
The Limits of Regional Power
There’s a broader pattern worth pulling out here, because it explains something that puzzles a lot of outside observers: why a regional leader who has proven remarkably resilient against a powerful national party can struggle to convert that local strength into national leadership.
The honest answer is structural rather than personal. A genuine alliance with rival opposition parties nationally would have required this regional party to give up some of its own seats â and with them, the patronage networks tied to those seats. For a political machine built around rewarding local operators with access to government resources, that’s not just politically uncomfortable. It’s close to organizationally impossible. You can build something powerful enough to dominate a single state for a decade. But a coalition built purely around opposing a common rival, rather than a genuinely shared vision, rarely has enough gravity to hold once the stakes shift to a national scale. What works brilliantly at home tends to isolate you everywhere else.
The Communities Caught in the Middle
The religious minority most directly affected by this episode isn’t the single, uniform political bloc that majoritarian strategy often implicitly assumes. It includes urban professionals, agricultural laborers in border districts, and families whose citizenship documentation has remained genuinely uncertain for generations because of how international borders shifted multiple times across the century, generating waves of displacement and incomplete paperwork no single bureaucratic revision has ever fully resolved.
What ties these communities together politically isn’t religious solidarity so much as a shared, rational calculation about survival. Supporting the regional incumbent was rarely about deep affection for that party. It was about avoiding something considerably worse. For families without clean documentation reaching back generations, debates over citizenship law aren’t abstract arguments conducted at a comfortable distance â they’re the practical difference between secure citizenship and the genuine risk of statelessness. And the incumbent party, aware of this, often treated that loyalty as something to be relied upon rather than invested in, rarely translating support into the kind of substantive protection that loyalty might reasonably have earned.
This became visible in a particularly dramatic way during a local uprising in one rural district, when women from marginalized farming communities rose up against local power brokers over land seizures and sexual coercion. These women were not speaking as representatives of any religious community or as symbols in someone else’s political narrative â they were speaking about their own immediate grievances. But within days, national media had reframed the entire episode into a story about a completely different kind of threat. The original demand, for safety and basic justice, was absorbed and reissued as something else entirely. This is a pattern that recurs wherever people without institutional power try to speak up: they speak, and what gets heard is whatever the loudest political apparatus decides it needs to hear.
What This Pattern Actually Adds Up To
Step back far enough, and a single underlying pattern emerges. Marginalized communities in situations like this tend to get squeezed from two directions simultaneously. One squeeze is administrative: software, built without conscious intention to discriminate, ends up systematically removing the most vulnerable voters before a campaign even begins. The other is discursive: whenever these communities voice their actual demands, those demands get hijacked and reframed by whichever political organization is best positioned to convert that energy into electoral capital.
What strikes me most is that neither half of this pattern requires any obvious villain working behind the scenes. Nobody deliberately designed a database glitch to disenfranchise a religious minority â it’s simply what happens when a messy, multilingual paper trail meets an automated matching system never built with much sensitivity to that history. Nobody centrally orchestrated the media’s reframing of a local protest into a national identity story â that’s just what a media ecosystem tends to do reflexively when a convenient narrative presents itself. That absence of an identifiable villain is exactly what makes both halves of this pattern so difficult to fight. Neither one looks like corruption. Both look, on the surface, like ordinary institutions doing their ordinary jobs.
A More General Lesson
I don’t think episodes like this should be read as simple electoral verdicts â voters chose, the better option won, end of story. I think they’re better understood as instances of a newer style of political control, one in which open suppression of democracy is no longer even necessary. You can instead quietly re-engineer who counts as part of “the people” before voting starts, using administrative tools dull enough that almost nobody pays attention, and still accurately describe the resulting process as democracy in the formal sense.
This should concern anyone who pays attention to how democratic systems actually erode, because it suggests the traditional warning signs â banned newspapers, imprisoned opposition leaders, visibly stuffed ballot boxes â may increasingly be the wrong things to watch for. The next wave of democratic erosion is more likely to arrive wearing a duller disguise: a software update, a routine data cleanup initiative, an electoral roll revision nobody objects to in principle, because on its face, who could possibly object to removing duplicate entries from a database?
There’s also a genuinely open question worth holding onto rather than resolving too quickly. Whichever party benefits from this kind of administrative reengineering inevitably inherits its own internal fault lines â disappointed allies, fragmented communities, marginalized groups who don’t fit anywhere within the existing party system’s categories. Re-engineering an electorate doesn’t make the underlying social tensions disappear. It just postpones the reckoning. Whether that postponement can hold indefinitely is the real question worth sitting with â and it deserves more sustained attention than any single election cycle is likely to give it.Â
1/7/2026
Amitava Mukherjee
Copyright@Amitava Mukherjee
Several points raised are worthwhile to consider.
Thank you!