On the afternoon of 25 September, Shri B.L. Santhosh, the highly respected National General Secretary (Organisation) of the ruling Bharatiya Janata Party, did for the Election Commission of India what the Commission has not troubled to do for itself. He set out, in five crisp lines on X, the criteria its software uses to decide which Delhi voters must prove themselves all over again, with the share of notices each accounts for. He then asked which forces want such names kept on the roll, which party and which family built the system that kept them there, and why no intensive revision was held for decades. The nation, he said, wants answers.
The intervention tells us two things. The first is to the party’s credit. It shows how seriously the BJP takes the task of carrying public opinion with it on the new voter list, rather than leaving the field to misinformation, disinformation and noise. The second is less comfortable. When a constitutional body’s criteria have to be explained and defended by the senior-most organisational functionary of the ruling party, because the body itself will neither publish nor defend them, something is rotten in the state of Denmark.
His figures are correct, and I say so first because my quarrel is with the inference drawn from them. In the breakdown of 32.27 lakh Delhi notices reported last week, siblings less than nine months apart account for 2,08,697 cases (6.47 per cent), a parent less than fifteen years older for 1,60,795 (4.98 per cent), six or more children for 1,37,492 (4.26 per cent), a parent more than fifty years older for 80,626 (2.50 per cent), and a grandparent less than forty years older for 72,581 (2.25 per cent). Add them up and they come to about 6.6 lakh notices, roughly one in five. Presenting them as “the reasons” was generous to the Commission. They are a fifth of the reasons. The other four-fifths are voters the software could not map to the 2002 roll, a list that runs to Shri L.K. Advani and Dr S. Jaishankar, and to a notice for Shri Manish Sisodia that cites no reason at all.
The post reads an anomaly in a family’s data as evidence of a bogus voter. I have spent thirty-seven years reading registers, from the patwari’s jamabandi to the electoral roll of the Amritsar parliamentary constituency. An anomaly in a register is first of all evidence about the register.
Garbage In, Notice Out
Begin with the benchmark. The 2002–03 rolls were typed from handwritten registers and digitised in haste. Age was recorded in completed years, usually as declared at the door, not as a date of birth. Spellings wandered from one revision to the next. One reply beneath Shri Santhosh’s own post, from a self-declared supporter of the SIR, put it more plainly than I can: the 2002 list had serious data problems, with mistakes in spelling and age everywhere.
The Commission knew this, which is why it built filters in the first place; nobody sieves a benchmark he trusts. Having chosen a shoddy benchmark, however, it then ran the test the wrong way round. Instead of treating an improbable age gap as a flaw in its own old data, it treated it as a question the living voter must answer.
Consider how the software even knows who a voter’s parents, siblings and grandparents are. The roll does not record kinship. It carries a single “relation name” (father, mother or husband), and the family tree is reconstructed by matching names at an address, or against the relative the voter named as his legacy link. A common name matches the wrong man. A grandparent given as the legacy link is read as a parent, and the voter is flagged for a parent “more than fifty years older”; another reply beneath the post described exactly this happening. The discrepancy is in the linkage, and the linkage is the Commission’s.
The Iniquity of the Fathers
Scripture speaks of the iniquity of the fathers being visited upon the children unto the third and fourth generation. The Commission’s software has found a way to do it with arithmetic.
Take the first-time voter of 2026. She was born around 2008. She has a birth certificate from a municipal registrar, a Class X certificate from a recognised board, an Aadhaar number and possibly a passport. Her date of birth and parentage are better documented than those of any generation of Indians before her. Yet if her grandmother’s age was misrecorded in a register typed in 2003, she receives the notice, and it is she who must queue at the ERO’s office.
Each of the five tests measures a family; none measures the voter. If a parent’s recorded age is impossible, that proves an error in someone’s age record, most likely the parent’s. It proves nothing about whether the child is a citizen, is eighteen, or lives where she is enrolled.
Twins, Child Brides and Families of Six
Look at the five tests one at a time and the demography gives way under each.
Siblings less than nine months apart. Twins are born minutes apart; unless the rule exempts a zero gap, every pair of twins in Delhi is a suspect. A premature child can arrive seven months after an elder sibling. The software groups siblings by a shared father’s name, so it counts children of different mothers as full siblings: the widower who remarried, the man divorced and remarried, the rarer husband with two wives at once. Their children’s birth dates obey no nine-month rule. And two siblings born eighteen months apart can easily carry a recorded difference of zero years in a register that knows only years. A test that demands month-level precision from year-level data produces flags by construction.
More than one wife. Remarriage after a death or a divorce is common to every community. Simultaneous marriage is legally open only to Muslims, and it is rare. The International Institute for Population Sciences, analysing the fifth National Family Health Survey, found polygyny at 2.1 per cent among Christians, 1.9 per cent among Muslims and 1.3 per cent among Hindus, with a national average of 1.4 per cent, down from 1.9 per cent in 2005–06. Nobody should build an argument on numbers this small, in either direction. But the families behind them are real, and they will be caught by every one of these tests: a father decades older than his youngest child, half-siblings born months apart, seven children under one man’s name. An algorithm that cannot tell a second marriage from a forged entry has no business issuing notices.
A parent less than fifteen years older. This is the only test that points to something genuinely improbable. Even so, for the generation now past sixty, marriage at fourteen or fifteen was not unknown, and the likelier culprit remains a wrong age on the parent’s entry. The sitting Chief Minister of Delhi was flagged on this very ground.
A parent more than fifty years older. The rule says “parent”, but biology draws a sharp line between the two. For a mother, a child after fifty is rare, though no longer impossible. For a father, it is commonplace. A man can sire children well into his sixties and beyond, and many do, often on a second marriage to a younger wife. A test that applies the same fifty-year ceiling to fathers and mothers alike has not been thought through.
A grandparent less than forty years older. This implies two generations averaging under twenty years each. Where women married at sixteen or seventeen, as they did across much of north India within living memory, a gap of thirty-five years to a grandparent is ordinary.
Six or more children. This is no discrepancy at all. In the 1950s an Indian woman bore around six children on average, and as late as 1971 the Sample Registration System put the figure at 5.2. An average of five or six means that a great many mothers bore seven, eight or more. For any voter born before about 1975, a family of six children was ordinary.
Every one of these flags is the 2002–03 roll speaking. The Commission chose this roll as its benchmark. It cannot now ask the grandchildren to answer for its handwriting.
Why were these criteria never published? They reached the public through news reports of the Delhi breakdown and a party post on X, never through any notification of the Commission.Perhaps the Commission feared that voters, knowing the tests, would tweak their data to pass them. But how? The whole premise of the notice is that the voter must answer it with documents, and a birth certificate cannot be adjusted to satisfy an algorithm. Secrecy protected nothing except the criteria themselves from scrutiny.
The Supreme Court appears to share the unease. On 22 September the Bench observed that the Commission must clarify what a logical discrepancy is, and asked it to examine whether “additional filters” were needed before issuing such notices. The petitioners’ practical complaint was that documents which say nothing about the relative in question cannot explain an age gap with that relative. The notice asks a question to which no prescribed document is an answer.
Whose Register Is It Anyway?
Shri Santhosh asks which party and which family created a system that kept such names on the roll for decades. With respect, no party compiled these rolls. The Commission’s own Electoral Registration Officers did, revising them every year under governments of every colour. Whether and when to order an intensive revision has always rested with the Commission. And the software does not discriminate by party: the notice list runs from the Advani household to the BJP’s own Chief Minister. That acquits the Commission of conspiracy, only to convict it of not applying its mind.
Down the present road lie lakhs of hearings, a multiplicity of litigation in every High Court with an SIR in its jurisdiction, and eligible citizens disenfranchised because their grandparents were typed wrongly. Delhi’s notices are to be disposed of by 29 October. Punjab’s roll runs on the same software.
My recommendation is to go back to the Constitution. Article 326 says that a citizen of India, eighteen years of age and ordinarily resident, who is not disqualified, is entitled to be registered. The disqualifications are few and the list is closed: non-citizenship, unsoundness of mind declared by a competent court, and corrupt practices or electoral offences. Around that core, the Commission may and must remove the dead, the duplicated and the genuinely shifted, each after the individual hearing the Representation of the People Act, 1950 requires.
The Commission’s own Manual on Electoral Rolls makes the point better than any critic. Under Clause 8.5, a homeless citizen with no documentary proof of residence is to be enrolled once the BLO has visited the spot where he says he sleeps, on more than one night, and found him there, be it a pavement or the underside of a bridge. Earlier this month the Commission relied on that very clause to persuade the Delhi High Court that no homeless person need fear exclusion from the SIR. A Commission that will enrol a man sleeping under a bridge on the strength of two night visits cannot, in logic, ask a first-time voter holding a birth certificate to explain her grandmother’s age.
Everything else, logical discrepancies included, belongs where a sensible administrator would put it: on the BLO’s checklist at the door, as a reason to look twice. A flag is a question the Commission should answer from its own records before it troubles the citizen with a notice.
And to every reader who has received such a notice: your right to be registered is continuing. Even if a hearing goes against you, you may file Form 6 afresh, physically or online, up to the last date for filing nominations.
The criteria are now in the public domain, courtesy of the ruling party. It is the Commission that must explain them.