India built its first electoral roll in 1950 on a single principle: that a citizen's vote should not wait on a citizen's paperwork. The Special Intensive Revision of 2025 and 2026 reversed that principle and struck tens of millions of names, in proportions that fell heaviest on women who moved at marriage, on Muslims in the districts where suspicion was cheapest to manufacture, and on the poor who never held the documents the Commission chose to trust. The one card the State prints to certify that a person is a voter did not make the list of papers that let you stay one.

Booth 226 at Kendamari, a village in the Nandigram constituency of East Midnapore, had a little over six hundred names on its roll. After the Special Intensive Revision, 321 of those names were placed under adjudication and 172 were struck off. All 172 belonged to Muslims. Some, according to residents who spoke to Outlook, had been marked for scrutiny because their parents had five or more children, a ground for doubting a voter that appears in no section of the Representation of the People Act.

Almost everyone deleted at that booth would have owned an Elector Photo Identity Card, the laminated card with a photograph and a serial number that the Election Commission of India itself prints and hands to registered voters. For most Indians, it is the object they picture when they picture proof that they can vote. During the SIR it counted for nothing. The card the Commission issues to say that you are a voter was not among the documents the Commission would accept to keep you one.

This essay is about that inversion and about who paid for it. The SIR, run first in Bihar between June and September 2025 and then across a dozen states and union territories, was defended as a routine housekeeping of dirty rolls. Read as a piece of administrative design, it does something more particular. It demands documents held by a small and unevenly distributed minority, runs on a compressed clock, and shifts the burden of proof from the State onto the citizen. Filters built that way select along the existing lines of disadvantage. The three that show most clearly in the data released so far are gender, religion, and caste, and where two of them cross, the effect compounds.

The Commission has published no breakdown of deletions by sex, caste, or religion, so the disproportion cannot be lifted off an official table. It has to be assembled from three kinds of evidence: the structure of the process and who is likely to clear it, the overlay of deletion rates onto the demography of particular places, and the handful of constituency-level studies where an independent group did the disaggregation the Commission would not. I have tried to keep each kind of evidence to what it can carry. The claims about women and Muslims are firm and quantified. The claim about Dalits and Adivasis is thinner, and I say where it is thin rather than borrow confidence the numbers do not support.

The roll that was built to include

To see what the SIR reversed, it helps to remember what the first roll did.

When the Constituent Assembly Secretariat began preparing electoral rolls in late 1947, before the Constitution had even been adopted, it was attempting something no country of comparable poverty had tried. The historian Ornit Shani, in How India Became Democratic (Cambridge, 2018), reconstructs the effort from the Secretariat's own correspondence. Around 173 million adults were to be enrolled, most of them poor, the large majority unable to read, many of them refugees who had crossed a new border months earlier with nothing a documents officer would recognise. The franchise was made universal in one step, without the property and literacy qualifications that Britain and the United States had taken more than a century to shed.

The Secretariat had to fight to make it so. Shani documents how, in the winter of 1948, officials found that enumerators in several provinces were recording women not by their own names but as "wife of" or "daughter of" some man, and that hundreds of thousands of women risked being left off the roll for refusing, by custom, to give their personal names to a stranger. The Secretariat instructed that such entries be struck and the women enrolled in their own names, and where that failed, that the women be counted anyway. Refugees were enrolled on the basis of a declared intention to reside rather than documentary proof of where they had come from. The governing judgment, taken deliberately at the founding, was that citizenship came before paperwork, and that a roll which shut out the undocumented poor would not be a democratic roll.

The Special Intensive Revision inverts that judgment while claiming its authority. It asks the citizen to prove, on the State's terms and against the State's clock, the eligibility that the founding roll presumed. That reversal is the substance of the exercise, not a detail of its implementation.

What the Commission asked for, and who has it

The Bihar order of 24 June 2025 required every elector to file a fresh enumeration form. Anyone whose own name, or a parent's, appeared on the 2003 electoral roll could satisfy the requirement by citing that roll and submitting nothing further. The Commission uploaded the 2003 roll, which carried 4.96 crore names, and estimated that close to 60 per cent of the present electorate would therefore need no documents at all.

That estimate tells you who actually bore the weight. Bihar's electorate stood at 7.89 crore when the SIR was announced. If roughly six in ten could clear the bar by pointing to 2003, the documentary demand fell on the remaining four in ten, on the order of three crore people, and within that group most sharply on the subset who could not even link themselves to a parent's 2003 entry. That residual is not a cross-section of Bihar. Anyone who reached voting age after 2003 sits inside it by definition, which tilts it young.

Women are over-represented, because a woman who married into a village after 2003 has no entry on that village's 2003 roll and must reach back to a natal home in another district. Migrants are over-represented too, since the people whose paper trail is thinnest where they vote are the ones who travelled to get there. In effect, the exercise asked the least of those whose lives had stayed put since 2003 and the most of those who had moved, married, or come of age since then.

For that residual, the enumeration form listed eleven documents: a service identity card or pension order for a government or public-sector employee; any government-issued document predating 1 July 1987; a birth certificate; a passport; a matriculation or higher educational certificate; a permanent residence certificate; a forest right certificate; a caste certificate; the National Register of Citizens where it exists; a family register; and a government land or house allotment certificate. Aadhaar, the EPIC voter card, and the ration card were left off. Aadhaar was added as a twelfth document only in September, and only after the Supreme Court pressed, and even then as proof of identity rather than of the citizenship the exercise claimed to be testing.

Hold that list against the first question a development economist would ask: what share of the population actually possesses each item? Bihar is the one state where a recent enumeration answers it, because the 2022 Bihar Caste-Based Survey counted these things. Its findings: 6.11 per cent of the population are graduates, 14.71 per cent have passed class ten, and 1.57 per cent hold a government job. A third of all families (34.13 per cent) live on 6,000 rupees or less per month. Among Scheduled Castes, the graduate share is 3.05 per cent, and 99.49 per cent of Scheduled Caste households have neither a computer nor reliable internet access to download a 2003 roll or fill out an online form.

Set the accepted documents beside those numbers, and the skew is hard to miss. A matriculation certificate is held by roughly one in seven Biharis. A government service card or pension order belongs to the 1.57 per cent in state employment. A passport, a land allotment certificate, a permanent residence certificate each cluster among the propertied, the formally employed, and the educated, and those groups themselves skew by caste. Meanwhile, a PUCL survey of Patna slums in July 2025 found that 91 per cent of respondents held exactly the common documents, Aadhaar and the voter card, that the SIR would not take on their own.

Figure 1. Share of Bihar's population holding selected identity documents, by acceptance under the SIR. The documents the exercise accepted are held by small and caste-skewed minorities; the documents it excluded are near-universal. Sources: 2022 Bihar Caste-Based Survey (state-wide figures for accepted documents); PUCL Patna slum survey, 15 to 21 July 2025 (Aadhaar and voter card). The two samples are not directly comparable.

The sociologist Kamal Sadiq, in Paper Citizens (Oxford, 2009), argued that across the poorer democracies citizenship is lived less as a status than as a stack of documents, and that where documents are scarce or forged, the line between citizen and non-citizen is drawn by whoever controls the paperwork. The SIR turns that observation into policy. It treats the presence or absence of a particular certificate as the operative fact of belonging, in a state where most people hold no such certificate.

There is a second structural point, and for a law-and-economics reading it is the sharper one. To become a new voter, a citizen files Form 6, which accepts Aadhaar, a PAN card, a bank passbook, or a driving licence as proof of identity and address. The SIR demanded a stricter, narrower set of documents to remain an existing voter than the settled law requires to enrol as a new one. The effect was to make an existing right harder to keep than a fresh one is to acquire, which runs against the ordinary presumption of any legal order. An entitlement already held is meant to be harder to strip than to grant, because people have arranged their lives around holding it, and the vote is the entitlement where that reliance matters most.

Underneath both points sits a question about who should carry the cost of establishing a fact. The State already holds the records the SIR was ostensibly checking. Births and deaths are registered under the Registration of Births and Deaths Act. Migration leaves its trace in the destination roll, in ration portability, in employment records. In the language of law and economics, the least-cost provider of proof that a voter is alive, resident, and a citizen is, in almost every case, the State that issued the underlying documents to begin with.

The SIR placed that cost on the citizen instead, and did it on a three-month clock, with a single Electoral Registration Officer handling the forms of more than three lakh people, which the Association for Democratic Reforms fairly called impossible to do with any care. When the burden of proof is placed on the party least able to bear it, and the time to discharge it is short, the process results in failures of compliance, which are later recorded as though they were findings of ineligibility.

The card that could not prove

Return to the EPIC, because the Commission's own reason for excluding it undoes the rest of the list.

Asked why the voter card would not do, the Commission told the Supreme Court that the EPIC merely reflects the current state of the roll and so cannot, on its own, establish a prior eligibility for inclusion. Taken on its own terms, the reasoning is circular, and the circle swallows the whole documentary regime. The card is downstream of the roll, so it cannot validate the roll. True enough.

But every accepted document is downstream of some earlier act of the State. A birth certificate records a prior registration. A passport records a documentary check the passport office already performed. A caste certificate records a prior administrative finding. If "this paper merely records an earlier State decision" disqualifies the EPIC, it disqualifies all eleven. On that logic, none of the eleven fares better than the card. They differ from it only in who holds them, and the poor mostly hold the ones the Commission left off.

The litigation exposed a further turn. Having on 10 July 2025 suggested that the Commission also consider Aadhaar, the EPIC, and the ration card, the Court went on in its interim order of 14 August to direct the Commission to publish the searchable list of the roughly 65 lakh names dropped from the draft roll, with reasons, and to accept claims for re-inclusion supported by Aadhaar. So the identity documents the poor actually hold were admitted at the back door; to argue your way back onto a roll you should never have left, while the front door, the enumeration that decided whether you stayed at all, kept them out. A document good enough to reverse a wrongful deletion but not to prevent one is not a coherent evidentiary standard.

When the Court delivered its final judgment on 27 May 2026, a two-judge bench of Chief Justice Surya Kant and Justice Joymalya Bagchi upheld the Commission's authority to keep the EPIC and the ration card off the accepted list. The card the State prints to identify voters had now been held by the highest court insufficient to identify them for the one purpose that determines whether they could vote.

Women, and the roll that forgot marriage

The plainest design flaw of the SIR concerns how it treated women who had moved at the time of marriage.

Across most of north and eastern India, marriage is patrilocal and village-exogamous. A woman marries out of her natal village and into her husband's, and the documents that anchor her- her natal roll entry, her family's legacy, her own childhood records- stay at the home she left. At the village where she now lives and votes, she is a documentary newcomer. To use the 2003 shortcut, she would have to reach back to a parent's entry in another district or state, and the Commission would not accept parents-in-law as the "relatives" she could point to at her marital home. Her natal records are in a place she has left, and the process would not let her marital household stand in for them. She falls between the two.

This is not a marginal category. The Census of India 2011 migration tables, in the D-series (Table D-3), record marriage as the reason for close to two-thirds of all female migration in the country, compared with a negligible share for men. So when the Commission reads a married woman as "permanently shifted" and strikes her, it is reading the ordinary demographic fact of a woman's life as a disappearance. I made a related argument about how representation is engineered to defer women's inclusion in an earlier piece on delimitation. The SIR runs the same logic in reverse, removing women who had already been counted rather than failing to count them, and it lands on the exact population the Constituent Assembly Secretariat had fought in 1948 to keep on the roll, the married-in woman without papers of her own.

Uttar Pradesh lets us watch the effect in the Commission's own arithmetic. After the SIR the UP roll carried 7.31 crore men and 6.09 crore women, a sex ratio of 834 women per 1,000 men. The Commission presented this as an improvement on the draft roll's 824. Against the starting point, it is a fall, because the roll's sex ratio before the SIR was 877, so the exercise removed women faster than men and left the roll more male by 43 women per thousand. The gap widens once you set it against the population. An independent projection published in The Wire put UP's adult sex ratio for 2026 at about 943, so a roll at 834 under-registers women by roughly 109 for every thousand men relative to the adult population that actually exists. The same analysis found that only about 83 per cent of UP's projected adult population now appears on the rolls at all.

Figure 3. Sex ratio of Uttar Pradesh's electoral roll, before the SIR, at the draft stage, and in the final roll, against the projected adult sex ratio for 2026. The exercise lowered the roll's sex ratio from 877 to 834, widening the gap from the underlying adult population. Sources: ECI and UP CEO final roll, 10 April 2026; adult sex-ratio projection from the Report of the Technical Group on Population Projections, via The Wire, 21 April 2026.

West Bengal is where the disaggregation is best, and it is work published in these pages. The Kolkata-based Sabar Institute, drawing directly on the Commission's own published deletion lists and working with BehanBox, found that 61.8 per cent of the voters deleted or placed under adjudication in the state were women, some 61.9 lakh people, and that women were a majority of deletions in 219 of the state's 294 constituencies. The Commission's own recorded ground for most of these removals was that the voter had "permanently shifted," which for women is, in the main, marriage.

The Bengal data also shows the first crossing of the lines. The female share of deletions was 57.6 per cent on Scheduled Caste reserved seats and 57.9 per cent on Scheduled Tribe seats, against a state average of 53.6 per cent. In absolute terms, the heaviest losses of women fell in Muslim-majority Murshidabad and Malda, where the single constituency of Sujapur lost 68,938 women from the rolls. Six of the nine constituencies in which women were more than 60 per cent of deletions sit in the Adivasi belt of Jungle Mahal, so the women worst hit there were largely Adivasi. Read across the state, the female share of deletions rose wherever the electorate was also poor, Muslim, Dalit, or Adivasi.

Figure 4. Women as a share of voters, deleted or under adjudication, in West Bengal by seat reservation status. The female share of deletions is higher on reserved seats than it is at the state average. Source: Sabar Institute and BehanBox analysis of ECI-published West Bengal deletion data, April 2026.

Where suspicion was cheapest

The evidence on Muslims is the strongest of the three, and in one district it survives the objection usually raised against this kind of overlay.

That objection is the ecological one. A district can carry both a high Muslim share and a high deletion rate without Muslims being the people deleted; the overlay describes places, not persons. Bihar's Seemanchal shows the overlay plainly, its four districts of Kishanganj, Purnea, Katihar, and Araria recording the highest deletion rate in the state at 7.7 per cent, close to double the state average. On its own, the overlay leaves the objection standing.

What answers it is the movement of the Muslim share across the stages of the funnel. Muslims were 24.7 per cent of the roughly 65.75 lakh initially flagged for scrutiny in Bihar, and 32.1 per cent of the roughly 3.23 lakh permanently deleted after ground verification. The share rose as the funnel narrowed, so at the adjudication stage, where an officer's discretion decides who stays, a flagged Muslim was carried through to deletion at a higher rate than a flagged non-Muslim. In Kishanganj, the same disparity appears as a rate: 3.7 per cent of Muslim electors deleted against 1.9 per cent of non-Muslims. That comparison holds the flagging stage constant, so it speaks to what happened to two people who entered the process at the same point and left it differently, and not to the composition of places.

Figure 2. Left: the Muslim share of deletions in Bihar rose from 24.7 per cent of those flagged to 32.1 per cent of those finally deleted, indicating a concentration at the adjudication stage. Right: in Kishanganj, Muslim electors were deleted at nearly twice the rate of non-Muslims. Sources: The Wire analysis of ECI constituency-wise SIR data, 9 October 2025; Yadav and Shastri analysis of ECI data.

The counter-pattern belongs here too, because it cuts against a lazy version of the argument. The single highest district deletion rate in Bihar was not in Seemanchal but in Gopalganj, at about 12 per cent, a district defined by out-migration to the Gulf and to other Indian states rather than by Muslim concentration. Migration and religion are two distinct mechanisms of deletion, and both operated. Conceding Gopalganj to migration leaves the Seemanchal disparity intact, and it shows the analysis is tracking mechanisms rather than confirming a prior.

West Bengal provides the most extreme figure, and its denominator needs to be stated. In Nandigram, the Sabar Institute analysed seven of the Commission's supplementary deletion lists and found that 2,700 of the 2,826 names struck off, or 95.5 per cent, were Muslim, in a constituency about a quarter Muslim. The headline share matters less than the movement behind it. Muslims were 33.3 per cent of Nandigram's December 2025 flagged list and 95.5 per cent of those finally deleted from it, so the concentration sharpened at the adjudication stage, which is where discretion, and its abuse, operate. In neighbouring Bhabanipur, one-fifth Muslim, Muslims were 40 per cent of deletions. Across the state, the absolute losses clustered where Muslims are numerous and could sway a result, near 460,000 deletions in Murshidabad, 330,000 in North 24 Parganas, 240,000 in Malda.

Figure 5. Muslim share at successive stages of deletion in Nandigram: of the population, of the December 2025 flagged list, and of those finally deleted. The Muslim share rises at each stage, most sharply at the final adjudication. Source: Sabar Institute analysis of ECI supplementary lists 1, 2, 3, 4a, 7, 8 and 9, Nandigram, April 2026.

None of this occurred in a rhetorical vacuum, and the history it sits in is recent. The 2019 National Register of Citizens in Assam, the country's last full experiment in documentary citizenship, excluded 19,06,657 people, among them many who could not produce a "legacy" link to a pre-1971 ancestor, a burden that fell heavily on women married across district lines and on the poor whose families had never held such papers. The SIR carries that apparatus into the electoral roll. When, campaigning in Purnia in September 2025, the Union minister Giriraj Singh described the region's Muslims as foreign infiltrators in language meant to dehumanise, and an administrative exercise then deleted Muslims from that region at nearly twice the rate of their neighbours, the two sit together without needing a documented instruction to connect them. I traced how the governing party's eastern strategy rebuilt the categories of welfare and belonging in an earlier essay; recasting the Muslim voter as a presumptive infiltrator is that same category work, applied now to the roll.

The groups the data has not yet named

My claim about Dalits and Adivasis is weaker than the other two, and it is worth being explicit about why before setting out what it is.

The Commission has released nothing that disaggregates deletions by caste. Yogendra Yadav and the psephologist Sanjay Kumar have both noted that detailed data on how the SIR affected Dalits and migrant workers was still awaited even as the picture for women and Muslims came into focus. There is no equivalent for Dalits to the Kishanganj within-funnel comparison. What exists comes in two forms.

The first is the evidence of who holds the documents, already set out: Scheduled Caste graduates at 3.05 per cent of the SC population, the near-total absence of computers and internet, the concentration of the landless and the informally employed among Dalits, all of it mapping onto the documents the SIR demanded and mostly denied them. A Dalit family that has never owned land holds no land allotment certificate. A Dalit household in which no one finished school holds no matriculation certificate. The filter did not have to name caste to select against it.

The second is the migration channel, which catches Dalits and Adivasis heavily because they are over-represented among seasonal and circular migrants. The PUCL enumeration-phase report records that the SIR ran through the peak migration months of November to March, that whole Adivasi families away for work were marked absent or shifted although they were alive and eligible, and that Adivasi names frequently vary across documents in ways that break the matching the process depended on. The Bengal figures give the one hard corroboration available: the female deletion share was higher on reserved seats than off them, so the caste line and the gender line were being cut together even where the caste line alone stays invisible in the data.

I will not claim a precision the numbers do not permit. The defensible statement is that the mechanism selects against Dalits and Adivasis, and that the single disaggregated dataset we have is consistent with it. Whether Dalit deletions can be quantified the way women's and Muslims' can is a question the Commission's silence does not yet let anyone answer, and that silence is not incidental to the exercise.

A cleaning with no measure of clean

Any exercise that calls itself a cleaning invites one question: how much dirt was there, and how much is left. By that test the SIR fails on its own terms, and the failure is the part a development economist can measure.

From its first order onward the exercise was justified by the presence of ineligible names on the roll: the dead, the shifted, the duplicated, and, in the framing that did the political work, the foreign infiltrator. Yet when you read what the deletions were actually attributed to, the infiltrator all but vanishes. Bihar's 65 lakh draft deletions broke down as 22 lakh dead, 36 lakh shifted or untraceable, and 7 lakh duplicate. The former Election Commissioner Ashok Lavasa put the point exactly: there is no public figure for how many people were deleted because they were genuinely ineligible under Article 326, and shifting does not make a person ineligible; it only means they should be enrolled somewhere else. He asked what degree of purity the cleaned roll had reached, and how anyone would measure the health of the new roll against the old. The Commission has not answered.

This is the same structure I described in the Great Nicobar appraisal, where the justification moved from column to column each time one was challenged, so that no single version could ever be put to the test. The SIR's justification behaves the same way. Question the citizenship rationale, and the answer is that most deletions were for death and migration, ordinary housekeeping. Press the housekeeping as disproportionate, and the answer returns to the sovereign duty to remove ineligibles. A justification that relocates whenever any part of it is refuted cannot be tested against evidence, and a justification that cannot be tested is how administrative power places itself beyond review. I made a parallel argument about the State's refusal to let its own accounts be read in Reading the Centre's Books.

The law-and-economics core can be stated cleanly. Any verification regime faces two errors. It can wrongly retain an ineligible voter, a false negative in the cleaning, or it can wrongly remove an eligible one, a false positive. Call the social cost of the first error C1 and of the second C2. Article 326's guarantee of universal adult suffrage is, among other things, a statement that C2 is far larger than C1, because wrongly stripping a citizen of the vote is a graver harm than tolerating one extra name that can be caught at the next revision or challenged at the booth. A verification designed around that ranking would set a high threshold for deletion and a low one for retention, erring on the side of inclusion.

The SIR set a high threshold for retention, a low one for deletion, a punishing documentary standard, and a three-month clock, so its implied loss function treats C1 as the greater cost, which is the ranking Article 326 forbids. That inversion is a policy choice, not a technical necessity, and it produced 47 lakh deletions in Bihar, over two crore in Uttar Pradesh, and, in West Bengal, deletions the state's own officers recorded at 5,46,053 outright removals with a further 60,06,675 names left under adjudication, in the Chief Electoral Officer's press note of 28 February 2026. Wider reporting put the total number of West Bengal removals near 91 lakh, close to an eighth of the electorate, though the net figure after appeals remains contested and should be read as a gross count.

Amartya Sen's argument in Development as Freedom (Oxford, 1999) was that a formal right is only as real as a person's capability to exercise it, and that the conversion of a right into a lived freedom depends on the resources and recognition the person can bring to bear. The vote is such a right. Written into the Constitution, it is universal. Exercised through a documentary filter that a seventh of the population can satisfy, it becomes a capability distributed as unequally as the documents themselves.

The SIR did not touch the words of Article 326. It altered the conversion factor, to the disadvantage of the people the article was written to enfranchise. In eleven Bihar constituencies the number of SIR deletions exceeded the winner's eventual margin of victory, so in those seats the revision stopped being a neutral precondition of the result and became a variable in it.

What the Court saw, and what it set aside

On 27 May 2026, the bench of Chief Justice Surya Kant and Justice Joymalya Bagchi upheld the SIR in a 124-page judgment, reported as 2026 INSC 564. The reasoning, on its own plane, is not weak. The Court held that free and fair elections rest on the integrity and accuracy of the roll and not only on the mechanics of polling day, that the exercise bore a direct nexus to that constitutional goal, that Articles 324 and 327 complement rather than compete, and that Section 21(3) of the Representation of the People Act, 1950, gave the Commission room to order a special revision. It found the process proportionate and, in its own words, not manifestly excessive.

The judgment carries a contradiction at its centre that its critics have fastened onto, and fairly. The Court held that the Commission may examine citizenship for the limited purpose of deciding inclusion, yet may not conclusively determine citizenship, and must refer doubtful cases to the competent authority under the Citizenship Act within four weeks. The difficulty is that the exclusion comes first and the referral second. A voter is struck from the roll on a citizenship doubt the Commission is not empowered to resolve, and only afterwards, if ever, is the doubt sent to the body that can resolve it. The disenfranchisement precedes the determination meant to justify it. Lavasa's follow-up sharpens the same edge: four weeks after the judgment, how many people had actually been referred to a foreigners' tribunal, and where is that number? It has not appeared.

What the Court largely set aside was the incidence. It treated the procedural safeguards it had ordered as a sufficient answer to the hardship, without engaging the demonstrated fact that in West Bengal, under its own supervision, the process had fallen on women and Muslims in the proportions the Sabar Institute had by then documented. A proportionality analysis that never quantifies who bears the burden is proportionality in name only. The Court asked whether the means fit the end. The prior question, whose exclusion those means were producing, it did not press, though for a measure defended in the name of universal adult suffrage that is where the inquiry should have started.

What a revision that included people would look like

A roll does need tending. The dead accumulate on it; people move; duplicates form; and it would be dishonest to pretend the Commission had no legitimate housekeeping to do. The argument here is about who does the tending, in which direction the errors are permitted to run, and what the citizen is made to carry.

A revision built to include rather than to exclude would begin by moving the burden of proof back to the party that holds the records. The State registers births and deaths, seeds Aadhaar into welfare and banking, and maintains the destination rolls into which migrants enrol, so a verification run off the State's own databases, with the citizen approached only to correct a specific flagged discrepancy, would catch the genuine dead and the genuinely duplicated without demanding that a widow in Sujapur produce a matriculation certificate she never earned.

The documents people actually hold would count as strong evidence of identity and residence rather than as suspect for being common; among them are Aadhaar and voter cards. The ADR rightly called it patently absurd to reject them. Recognition of how women move would follow, with parents-in-law and marital households treated as the links they are, in fact. And at every stage the Commission would publish the denominators that let the purification claim be tested: how many names were flagged, how many deleted, on what ground, and, of the deletions attributed to ineligibility rather than to death or migration, how many were referred onward and how many of those referrals were upheld.

The founders' role began with a presumption of inclusion, and it placed the work of enrolment on the State. The revision begins with a presumption of doubt, and it places the burden of proof on the citizen. The people who built the first roll enrolled the illiterate, the propertyless, and the woman who would not give her name to a stranger, on the reasoning that a democracy which waits for the poor to produce their papers has already begun to decide whom it can do without. The Special Intensive Revision has reinstated the papers, and the deletions are its answer to that question.

Further Reading

On the founding of the franchise

1. Ornit Shani, How India Became Democratic: Citizenship and the Making of the Universal Franchise (Cambridge University Press, 2018), on the preparation of the first electoral roll and the fight to enrol women in their own names.

On the SIR process and its legality

1. The Supreme Court Observer case file tracking Association for Democratic Reforms v. Election Commission of India, W.P.(C) 640/2025, and the judgment summary for 2026 INSC 564.

2. ThePrint on the four questions the verdict turned on.

On who was deleted

1. The Wire on the higher exclusion rate for Muslims in Bihar and on why the UP numbers do not add up.

2. Sabar Institute and BehanBox on the gender skew in West Bengal, and Scroll on the Nandigram deletions.

On documents, capability, and the informal citizen

1. The 2022 Bihar Caste-Based Survey, as reported in Business Standard, on how few Biharis hold the papers the SIR demanded.

2. Kamal Sadiq, Paper Citizens (Oxford University Press, 2009), and Amartya Sen, Development as Freedom (Oxford University Press, 1999), on documentary citizenship and on rights as capabilities.

3. PUCL's field reports on the enumeration phase and the Patna slum survey.

Related arguments in this publication

1. Women, Delimitation, and the Design of Representation Deferred.

2. An Island Off the Books: The Development Accounting Behind Great Nicobar.

3. Reading the Centre's Books.

Varna is a development economist and writes at policygrounds.press.