At dawn on the Mechi floodplain, the border wakes with the day. Men push bicycles loaded with fodder towards Uttar Pradesh, and a schoolgirl in a blue skirt rehearses her civics lesson. Every household has a story about paperwork – ration cards, job cards, Aadhaar – but it is the smallest card that decides whether they count: the voter slip. In 2020, many arrived at the polling station to find their names had quietly disappeared.
In July 2025, our small independent team began examining whether patterns in Bihar’s ongoing Special Intensive Revision (SIR) of electoral rolls, ahead of the November assembly election, echoed earlier trends. Using official data – 2019 voter-list deletions, 2020 booth-level election results, and the 2022 Bihar Caste Survey – we linked deletions to outcomes and estimated the community composition of removed voters through surname-caste-religion correlations.
Two strands now run in uneasy parallel. The first is our 2020 analysis, which showed that roughly 6.56 million names were deleted across 77,000 polling booths, with about 6,000 booths registering 250 or more deletions each. The second is the 2025 SIR, which ended on September 30 with 7.42 crore electors, down from 7.89 crore at the start. About 65 lakh names were removed, 21.5 lakh added, and 3.66 lakh marked ineligible. The Supreme Court has since asked the Election Commission to explain who those 3.66 lakh are and on what grounds they were excluded.
Both datasets trace the same geography of absence: Bihar’s northern and western border belt – West Champaran, Siwan, Gopalganj, Saran, and Bhojpur – districts adjoining Uttar Pradesh that show the steepest and most persistent changes in the rolls. Here, deletions are not a drizzle but a monsoon burst. In 2020-21, booths along the UP border recorded about 127 deletions each, compared with 76 in interior Bihar – a 1.7-fold difference. Spatial tests confirm that clustering is no accident.
In one Siwan booth, 82% of those deleted were Extremely Backward Class or Scheduled Caste; near Bettiah in West Champaran, 70% were Muslim. These are not anomalies but repeating seams in the map, where the social composition of loss mirrors long-standing vulnerabilities.
Inside the 6,000 high-deletion booths, the profile shifts sharply: Muslims form about 18% of removals (13.5% in low-deletion areas), EBCs 18%, and SCs 5%. In roughly 2,300 of these booths, three-quarters or more of deletions come from Muslim, SC, or EBC households. In hamlets of Araria and Kishanganj, families speak of names struck off for minor mismatches – a surname change after marriage, an uncle listed as dead though alive, a migrant son marked “shifted.” The statistics translate into everyday erasure.
When the 2020 booth-level results were matched with deletion data, one pattern stood out: booths with higher deletions tended to report stronger NDA (National Democratic Alliance) victories. In booths with few deletions, the NDA won about 70%; in the most affected, nearly 80%. What changed was not the scale of victory but whether a booth flipped at all. The composition of deleted voters – disproportionately Muslim, SC, and EBC – suggests that what is presented as an administrative clean-up aligns uncannily with existing social divides.
The new roll inherits the old pattern. The net reduction of nearly 47 lakh names between June and September, the legal scrutiny of 3.66 lakh exclusions, and the recurrence of border-district volatility together signal that deletion remains concentrated, not random. With campaigning underway, a bureaucratic decision in September can determine a citizen’s voice in November.
In a Katihar village, an elderly man clutches his faded slip: “I have voted since Indira Gandhi, but this time my name is gone.” His bewilderment echoes across the border belt. A young teacher near Raxaul says her family voted in 2015 but found their names missing in 2020. In Sitamarhi, women who sign as “Devi” or “Kumari” to hide their caste appear repeatedly in deletion lists. For seasonal migrant families, six months away for work can mean permanent removal. The costs are uneven: for some, an inconvenience, for others disenfranchisement.
What the evidence adds up to
Three facts are clear -
Place matters: Deletions cluster along Bihar’s northern and western borders. Community matters: In the state’s worst-affected booths, Muslim, EBC, and SC voters form a disproportionate share of those removed. Elections reflect these patterns: High-deletion areas in 2020 corresponded to stronger NDA performance and lower Muslim representation in the rolls.
These observations do not assign motive. They do, however, set a standard for stewardship: when an administrative process reliably disadvantages certain communities in certain places, the burden of explanation rests with the State.
The findings reveal a continuum rather than a moment: each roll revision builds on the last, quietly defining who is seen and who is counted. What began as a technical exercise has become a mirror of inequalities in Bihar – of visibility, mobility, and trust. Restoring that trust requires transparency.
Three steps would help: publish booth-wise SIR statistics so independent verification is possible; audit random border blocks and reinstate any names wrongly deleted; and simplify re-enrolment for migrants, widows, and first-time women voters whose documents often carry small but fatal errors.
This analysis draws solely on official aggregates: 2019 voter-list deletions, 2020 Form 20 booth results, and the 2022 Bihar Caste Survey. Spatial patterns were identified by geo-referencing polling booths to border and interior districts. All data were aggregated; no personal identifiers were used.
The ballot is a quiet instrument. Its fairness depends on what happens long before the queue. In Bihar’s border villages, the numbers ask us to look closer – and sooner.
(The writer is an award-winning social researcher)
Disclaimer: The views expressed above are the author's own. They do not necessarily reflect the views of DH.