Bihar: 69 lakh names struck off

Part I: Bihar as a whole

At a glance

The Special Intensive Revision removed 69,10,559 names from Bihar's electoral roll in 2025: 65,59,409 before the draft roll of 1 August and 3,51,150 more before the final roll. That is 9.25 names for every 100 electors on the final roll, across 77,837 polling booths in 243 constituencies. In the Assembly election that followed in November, 4,96,35,462 votes were polled. The names removed equal 13.9% of them, one for every 7.2 votes cast.

Set against the results, the removals are larger than the margin of victory in 174 of 243 constituencies. They exceed the combined winning margin of all 243 seats, 51.4 lakh, by a factor of 1.34, and they exceed the combined winning margin of the NDA across the 202 seats it won, 46.7 lakh. In 66 of the 87 seats the NDA took from INDIA, more names were removed than the NDA won by.

Women are 55.5% of those removed and 47.9% of the roll they were removed from, and their share of the removed is higher than their share of the roll in 38 of 38 districts. Sex is recorded on the roll, so this figure rests on no estimate.

Deletions and coverage

The booth lists account for every one of Bihar's 77,837 polling booths. The table gives each district's removals, removals per 100 electors on the 2025 roll, and its share of the state total.

Removals range from 5.5 per 100 electors in Sheikhpura to 17.3 in Gopalganj. Among constituencies the range is wider, from 4.4 in Darbhanga to 23.6 in Purnia.

Deletions against the margins and the votes polled

Each constituency's removals are compared with two figures from the November 2025 election, the margin by which the seat was won and the votes polled. The comparison makes no claim that these electors would have voted, or how. It measures the size of the revision next to the vote that decided each seat.

The 25 constituencies where the removals were largest against the margin. Sandesh was won by 27 votes and lost 25,665 names.

Across Bihar the removals equal 13.9% of the votes polled. In 68 constituencies they equal more than 15%, in 18 more than a fifth, and in Gopalganj they reach 34.1%. In Digha the names removed exceed the runner-up's entire vote. In no constituency do they exceed the total votes polled.

Could the removals have changed who won?

A name removed is not a vote lost to one side. Some of the removed had died, many would not have voted, and those who would have voted would have split between the parties. Each seat's count here is the names still off the final roll, less those entered as dead, at the seat's 2025 turnout: 21,54,727 electors across Bihar. That count is low, because the names added after the draft include new electors, so every estimate is also run on all the names removed, with nothing added back (32,59,393).

In 56 of 243 seats the removed electors who would have voted were enough to change the result had they backed the runner-up. In 7 of them a lean of under 55% would have done it (Sandesh, Ramgarh, Nabinagar, Agiaon, Balrampur, Dhaka and Forbesganj). In the other 187 seats the margin was larger than every removed vote put together. This is the one figure that needs no assumption about how anyone votes.

The lists do not say how the removed would have voted. Going by how their own polling booths voted in 2024, in the 228 seats the two alliances fought each other, those who would have voted lean 49.0% NDA, 37.8% INDIA and 13.2% other parties. On that reading 4 seats would have changed hands counting only the names still off the final roll (Sandesh to RJD, Agiaon to CPI(ML)L, Nabinagar to RJD and Forbesganj to BJP), and 6 counting every name removed (Sandesh to RJD, Agiaon to CPI(ML)L, Nabinagar to RJD, Forbesganj to BJP, Chanpatia to BJP and Saharsa to BJP).

Weighting instead by how Muslims and others, or women and men, voted in the Axis My India exit poll, Bihar Assembly election 2025, as reported by NDTV on 12 November 2025, 1 and 2 seats change hands on the high count (Jehanabad to JD(U); Dhaka to BJP and Forbesganj to BJP). On every estimate the NDA keeps between 200 and 204 of the 243 seats.

Seats where the winner or runner-up is outside the two alliances get the break-even figure only, because the booth results and the poll both count every other party together. Two of those were close: Balrampur (LJP(RV) over AIMIM by 389) and Ramgarh (BSP over BJP by 30).

Each estimate assumes something the lists cannot show. People marked shifted may vote where they now live, people who moved may vote less than their neighbours, and an exit poll carries its own error. None of it can say whether any removal was right.

Did the vote move where more names were removed?

Between the 2020 and 2025 Assembly elections the NDA parties gained votes across Bihar. If the removals had taken away more of INDIA's voters than the NDA's, INDIA's share should have fallen most where the most names came off the roll. The candidate-wise results for both elections were counted with the parties grouped as they stood in 2025: BJP, JD(U), LJP(RV), HAM(S) and RLM for the NDA, and RJD, Congress, the three Left parties, VIP and IIP for INDIA.

Once each seat's 2020 result is allowed for, the NDA's lead over INDIA grew by between 4.0 and 5.2 points in each third, about the same whether a seat lost few names or many. Before that allowance INDIA seemed to lose most where the fewest names were removed (-2.9 points against +0.2), and part of that is seats giving back in 2025 what they had gained in 2020. Votes cast grew by about the same amount in all three. Comparing each seat only with other seats in its own district, and allowing for its 2020 result, each extra name removed per 100 electors went with a change of +0.31 points in the NDA's lead, most likely somewhere between -0.27 and +0.89. Had every removed person who was alive and would have voted voted against the NDA, its lead should have grown by about 0.66 points more for each extra name removed per 100 electors, and by about 0.33 if three in four of them had. That range takes in no effect at all and both of these, so the vote cannot tell whether the removals cost INDIA its voters. Without the allowance for 2020 the figure is -0.83 (between -1.86 and +0.20). The two models disagree even on the direction, which is itself a reason to read little into either.

This asks only whether the removals moved votes between the alliances. It cannot say whether each name was rightly removed, and a voter wrongly removed has lost a vote whichever way it would have gone. A small lean either way would not show, and five years separate the two elections, in which LJP rejoined the NDA, VIP changed sides and Jan Suraaj entered.

After the final roll

The final roll of 30 September 2025 was published after the claims and objections on the draft. The Election Commission's Format 4B return for it gives, seat by seat and by sex, the names added and removed after the draft and the final roll, and agrees with the Chief Electoral Officer's press note. The roll of June 2025 is published only as a state total, so a seat's roll before the revision is taken here as its draft roll plus the names on its phase 1 list.

After the draft, 21,53,343 names were added on claims and 3,66,742 removed on objections, so the final roll rose by 17,86,601 over the draft: 27.2% of the 65,59,409 names removed at the draft. That rise counts new electors as well as names put back, and the Commission has not said how many of those added had been removed at the draft, so it is the most that can have been put back. The final roll was 47,77,487 names smaller than the roll before the revision, and smaller in every one of the 243 seats.

In 124 of 243 seats the fall that remained at the final roll was still larger than the 2025 winning margin, against 174 when the draft removals are counted alone. Of those 124, the NDA won 92, INDIA 30 and others 2.

Women were 55.5% of the names removed at the draft and 47.9% of those added after it, so their share of the roll fell from 47.9% before the revision to 47.1% on the final roll. Of the 3,66,742 names removed after the draft, 62,441 were entered as dead, 2,12,999 as shifted and 81,334 as repeated, and 9,968 have no reason at all.

Women and the deletions

The roll records each elector's sex, so women's share of the removed needs no estimate. Women are 55.5% of all removals: 38,33,870 women and 30,76,689 men, 7,57,181 more women than men. Their share of the removed exceeds their share of the roll before the revision in 38 of 38 districts, by up to 13.7 points in Kaimur. The roll before the revision is the draft roll by sex plus the names removed at the draft by sex.

Phase 1 removed 7,20,543 more women than men, and 5,64,439 of that difference (78.3%) were entered as shifted. Among those marked shifted aged 18 to 29, 72.6% are women, the pattern that women who move on marriage would produce. The lists do not show whether any of them were enrolled again where they now live. If they were not, they have lost the vote.

Reasons for removal

The phase 1 lists record, against every one of the 65,59,626 names, the reason the officer entered for removing it. Counted as entered: 26,55,900 were marked permanently shifted (40.5%), 22,32,586 dead (34.0%), 9,70,081 absent (14.8%) and 7,01,059 already enrolled elsewhere (10.7%). The counts agree with the booth list in every booth, and the counts of women and men agree with the Chief Electoral Officer's phase 1 figures in every constituency. The 3,51,150 names removed in phase 2 carry no reason in the lists.

A reason is what the officer wrote down after a visit, or after failing to find the elector, and nobody checked it. Someone marked shifted may have been away for work when the officer called, and someone marked absent may have been at home. Of the four, absent rests on the least, because it records only that the officer did not find the elector. Already enrolled is a fair removal only if the other entry is the same person.

Among the removed aged 80 and over, 88.0% were entered as dead, and among those aged 18 to 24, 3.8%. The young were mostly marked shifted or absent.

Taken alone, the names marked absent exceed the 2025 margin in 20 of 243 seats, and the names marked already enrolled in 18. The names marked shifted exceed it in 76 and the dead in 58. The absent share is highest in Madhubani (28.2%), Sheohar (28.1%) and Gopalganj (25.7%).

Two reasons, set against the records

The 22.3 lakh names entered as dead equal about 3.4 years of adult deaths in Bihar. At the Sample Registration System's 2024 death rates by age and sex, the population aged 20 and over has about 6.5 lakh deaths a year. The ages of those entered as dead are close to those of registered deaths. 41.8% of those aged 25 or over were 70 or older, and 39.4% of registered deaths at those ages in 2024. Their sex does not match. Women are 52.0% of those entered as dead but 43.4% of adult deaths, which means 11.6 lakh women where 9.7 lakh would be expected. Women were only 37.7% of registered deaths at 25 and over in 2024, so women's deaths are registered less often and more of them may have stayed on the roll. The other possibility is that more women who are alive were entered as dead. The lists cannot tell the two apart.

Among those aged 18 to 29 marked shifted, 4,58,953 were women and 1,72,765 men, so women outnumber men by 2,86,188. Marriage alone could account for a number that size. In the ten years before the 2011 census, 53.8 lakh women in Bihar moved on marriage, about 5.4 lakh a year, so the excess is about 0.53 of one year's marriage moves. The districts do not fit. Where more women had moved on marriage, women were a smaller share of the young marked shifted (rank correlation -0.51). Marriage can explain how many young women were marked shifted, but it cannot explain where they were. The census counts women in the district they moved to, and its figures are fifteen years old.

Religion

The roll records no religion. The booth lists classify removed names as Hindu or Muslim where the name allows it, and leave 9.0% of names unclassified. The Muslim share of removals is therefore a range: 15.4% counting only names classified Muslim, and 24.4% if every unclassified name were Muslim. Bihar's Muslim population was 16.87% in the 2011 census and 17.70% in the 2023 caste survey. The census share lies inside the range in 18 of 38 districts. In 8 districts the classified Muslim share alone runs more than a point above the census share: Aurangabad, Gaya, Jehanabad, Lakhisarai, Nalanda, Nawada, Patna and Sheikhpura.

The 2011 census is the last count of religion by district, so each district's share is also carried forward to 2025 at the growth rates of its Muslim and total populations between the 2001 and 2011 censuses. That puts Bihar's Muslim share at 17.3% in 2025, close to the caste survey's 17.7%, and moves no district by more than 2.8 points. On the 2025 shares the range holds the population share in 20 districts, and the classified Muslim share is still more than a point above it in 8 districts (Aurangabad, Gaya, Jehanabad, Lakhisarai, Nalanda, Nawada, Patna and Sheikhpura).

Caste

Caste is not recorded either. It can be read, in part, from the surname of the father or husband listed with each removed elector, and two classifications are used because surnames cross caste lines. Classification A places 70.9% of removals in a group; the stricter classification B, which sets aside surnames carried by more than one group, places 35.2%. Only what holds under both is reported here.

A ratio above 1 means the group's share of the removed exceeds its share of the population. Under both classifications, General and SC families are removed at or above their population share and BC, EBC and ST families below it. The two disagree on Muslim families.

Booths

The median booth lost 71 names. 3,418 booths lost 200 or more each, and 142 lost 400 or more. A booth serves on average about 960 electors, so 200 removals is a fifth of a booth's roll.

Border districts

A constituency counts as a border seat when its district touches Nepal, West Bengal, Jharkhand or Uttar Pradesh. Border seats lost 9.81 names per 100 electors, the interior 8.36.

The alliances

Seats the NDA won in 2020 lost 9.58 names per 100 electors, and seats INDIA won lost 8.79. The order stays the same when seats are labelled by the 2024 or 2025 result. Inside the same constituency, booths the NDA led in 2024 lost 2.4 more names each than booths INDIA led. A booth's leading party shows how most of its voters voted in 2024 and says nothing about the removed electors, since every booth has voters on both sides. The seats INDIA lost in 2025 did not lose names faster than the seats it kept. The lists do establish the scale of the removals against the margins, set out above.

Key insights

• 69,10,559 names removed, 9.25 per 100 electors on the final roll, across 77,837 booths.
• Equal to 13.9% of the votes polled in the November 2025 election, or one name for every 7.2 votes cast.
• Larger than the margin in 174 of 243 seats, five times the margin or more in 30, ten times or more in 18.
• Larger than the NDA's combined winning margin in its 202 wins (69.1 lakh against 46.7 lakh).
• 66 of the 87 seats the NDA took from INDIA lost more names than the NDA won by.
• Women are 55.5% of the removed, over-represented in 38 of 38 districts.
• Muslim names are 15.4% to 24.4% of removals, against 16.87% of the population in 2011, and more than a point above the census share in 8 districts.
• In 56 seats the removed who would have voted were enough to change the result, and in 7 a lean of just over half among them would have done it.
• Women are 52.0% of the names entered as dead, against 43.4% of adult deaths in Bihar.
• 3,418 booths lost 200 names or more each.
• Reasons entered in phase 1: shifted 40.5%, dead 34.0%, absent 14.8%, already enrolled 10.7%. 5,64,439 of the 7,20,543 more women than men removed were marked shifted.
• Names marked absent alone exceed the margin in 20 seats, names marked already enrolled in 18.
• The final roll made up 27.2% of the draft removals, and the fall that remained still exceeded the margin in 124 of 243 seats. Women were 47.9% of those added after the draft.

How sure are we?

A pattern seen across all seats can weaken or turn round once seats are compared only with seats like them. Each finding was measured again within each district, within each seat, and with Seemanchal left out.

Two findings survive every check. In most constituencies more names were removed than the margin that decided the seat, on the draft count and on the final roll, and women were removed out of proportion in all 38 districts and 242 of 243 constituencies.

What this shows and what it does not

The lists show how many names were removed, where, who the removed were by sex, age and name, and the reason the officer entered against each phase 1 removal, and they allow the removals to be set against the margins and votes that decided each seat. They do not show whether that reason was true of the person, whether the removal was right, or how the elector voted. Religion and caste are guessed from names, so they are given as ranges or under two classifications. Votes polled are the turnout on the 2025 roll, within about 1% of a second calculation from each winner's vote and share.

Part II: District chapters

One chapter for each of Bihar's 38 districts, in order of removals per 100 electors. Each gives the district's totals and every constituency in it, with the 2025 result.

Gopalganj

Gopalganj lost 3,14,853 names across 6 constituencies, 17.31 for every 100 electors. Women are 60.0% of those removed, against 49.0% of the district's roll. Names classified Muslim are 14.1% to 22.1% of removals, against a 2011 census share of 17.0%. The heaviest constituency is Gopalganj, at 22.72 per 100 electors, and the largest multiple of the margin is in Barauli, where 48,477 names were removed against a margin of 12,374. In 2025 the NDA won 6 of the district's seats and INDIA 0.

Of its 3,10,028 phase 1 removals, the officers entered 42.2% as shifted, 22.9% as dead, 25.7% as absent and 9.2% as already enrolled.

Purnia

Purnia lost 2,99,452 names across 7 constituencies, 14.30 for every 100 electors. Women are 55.5% of those removed, against 48.3% of the district's roll. Names classified Muslim are 26.9% to 42.9% of removals, against a 2011 census share of 38.5%. The heaviest constituency is Purnia, at 23.64 per 100 electors, and the largest multiple of the margin is in Kasba, where 36,240 names were removed against a margin of 12,875. In 2025 the NDA won 5 of the district's seats and INDIA 0, and 2 went to neither alliance.

Of its 2,73,793 phase 1 removals, the officers entered 39.8% as shifted, 31.3% as dead, 21.8% as absent and 7.1% as already enrolled.

Kishanganj

Kishanganj lost 1,56,727 names across 4 constituencies, 13.90 for every 100 electors. Women are 61.5% of those removed, against 48.4% of the district's roll. Names classified Muslim are 54.6% to 70.6% of removals, against a 2011 census share of 68.0%. The heaviest constituency is Kishanganj, at 18.57 per 100 electors, and the largest multiple of the margin is in Kishanganj, where 52,574 names were removed against a margin of 12,794. In 2025 the NDA won 1 of the district's seats and INDIA 1, and 2 went to neither alliance.

Of its 1,45,639 phase 1 removals, the officers entered 40.8% as shifted, 27.1% as dead, 23.5% as absent and 8.6% as already enrolled.

Madhubani

Madhubani lost 3,61,114 names across 10 constituencies, 11.55 for every 100 electors. Women are 54.6% of those removed, against 47.8% of the district's roll. Names classified Muslim are 15.4% to 26.3% of removals, against a 2011 census share of 18.2%. The heaviest constituency is Rajnagar, at 15.04 per 100 electors, and the largest multiple of the margin is in Bisfi, where 30,343 names were removed against a margin of 8,107. In 2025 the NDA won 9 of the district's seats and INDIA 1.

Of its 3,52,229 phase 1 removals, the officers entered 33.5% as shifted, 28.7% as dead, 28.2% as absent and 9.7% as already enrolled.

Bhagalpur

Bhagalpur lost 2,53,818 names across 7 constituencies, 11.35 for every 100 electors. Women are 58.1% of those removed, against 49.5% of the district's roll. Names classified Muslim are 14.7% to 23.8% of removals, against a 2011 census share of 17.7%. The heaviest constituency is Sultanganj, at 14.32 per 100 electors, and the largest multiple of the margin is in Bhagalpur, where 33,936 names were removed against a margin of 13,474. In 2025 the NDA won 7 of the district's seats and INDIA 0.

Of its 2,44,409 phase 1 removals, the officers entered 51.4% as shifted, 25.7% as dead, 12.1% as absent and 10.9% as already enrolled.

Saharsa

Saharsa lost 1,39,670 names across 4 constituencies, 10.76 for every 100 electors. Women are 55.9% of those removed, against 48.4% of the district's roll. Names classified Muslim are 14.3% to 23.7% of removals, against a 2011 census share of 14.0%. The heaviest constituency is Mahishi, at 13.59 per 100 electors, and the largest multiple of the margin is in Saharsa, where 28,198 names were removed against a margin of 2,038. In 2025 the NDA won 2 of the district's seats and INDIA 2.

Of its 1,31,482 phase 1 removals, the officers entered 45.5% as shifted, 26.8% as dead, 16.5% as absent and 11.2% as already enrolled.

Sitamarhi

Sitamarhi lost 2,52,512 names across 8 constituencies, 10.25 for every 100 electors. Women are 52.9% of those removed, against 47.2% of the district's roll. Names classified Muslim are 19.6% to 29.9% of removals, against a 2011 census share of 21.6%. The heaviest constituency is Belsand, at 15.07 per 100 electors, and the largest multiple of the margin is in Bajpatti, where 31,190 names were removed against a margin of 3,395. In 2025 the NDA won 8 of the district's seats and INDIA 0.

Of its 2,44,788 phase 1 removals, the officers entered 33.7% as shifted, 37.0% as dead, 18.6% as absent and 10.6% as already enrolled.

Samastipur

Samastipur lost 2,93,199 names across 10 constituencies, 9.96 for every 100 electors. Women are 55.7% of those removed, against 47.5% of the district's roll. Names classified Muslim are 11.3% to 19.2% of removals, against a 2011 census share of 10.6%. The heaviest constituency is Morwa, at 13.32 per 100 electors, and the largest multiple of the margin is in Morwa, where 35,848 names were removed against a margin of 8,671. In 2025 the NDA won 7 of the district's seats and INDIA 3.

Of its 2,83,585 phase 1 removals, the officers entered 44.7% as shifted, 27.6% as dead, 15.5% as absent and 12.2% as already enrolled.

Katihar

Katihar lost 2,06,621 names across 7 constituencies, 9.93 for every 100 electors. Women are 56.6% of those removed, against 48.3% of the district's roll. Names classified Muslim are 32.3% to 46.7% of removals, against a 2011 census share of 44.5%. The heaviest constituency is Katihar, at 13.74 per 100 electors, and the largest multiple of the margin is in Balrampur, where 30,850 names were removed against a margin of 389. In 2025 the NDA won 6 of the district's seats and INDIA 1.

Of its 1,85,261 phase 1 removals, the officers entered 43.2% as shifted, 31.8% as dead, 15.9% as absent and 9.1% as already enrolled.

Saran

Saran lost 2,88,515 names across 10 constituencies, 9.88 for every 100 electors. Women are 58.4% of those removed, against 48.3% of the district's roll. Names classified Muslim are 9.6% to 15.6% of removals, against a 2011 census share of 10.3%. The heaviest constituency is Amnour, at 15.09 per 100 electors, and the largest multiple of the margin is in Taraiya, where 33,676 names were removed against a margin of 1,329. In 2025 the NDA won 7 of the district's seats and INDIA 3.

Of its 2,73,046 phase 1 removals, the officers entered 41.1% as shifted, 33.8% as dead, 15.3% as absent and 9.9% as already enrolled.

Sheohar

Sheohar lost 28,685 names across 1 constituencies, 9.41 for every 100 electors. Women are 57.0% of those removed, against 47.6% of the district's roll. Names classified Muslim are 15.3% to 25.9% of removals, against a 2011 census share of 15.1%. The heaviest constituency is Sheohar, at 9.41 per 100 electors, and the largest multiple of the margin is in Sheohar, where 28,685 names were removed against a margin of 31,398. In 2025 the NDA won 1 of the district's seats and INDIA 0.

Of its 28,136 phase 1 removals, the officers entered 37.5% as shifted, 26.0% as dead, 28.1% as absent and 8.4% as already enrolled.

East Champaran

East Champaran lost 3,23,982 names across 12 constituencies, 9.40 for every 100 electors. Women are 55.4% of those removed, against 47.5% of the district's roll. Names classified Muslim are 15.7% to 24.7% of removals, against a 2011 census share of 19.4%. The heaviest constituency is Motihari, at 18.65 per 100 electors, and the largest multiple of the margin is in Dhaka, where 18,163 names were removed against a margin of 178. In 2025 the NDA won 11 of the district's seats and INDIA 1.

Of its 3,16,230 phase 1 removals, the officers entered 37.3% as shifted, 33.9% as dead, 16.8% as absent and 12.0% as already enrolled.

Bhojpur

Bhojpur lost 1,97,931 names across 7 constituencies, 9.35 for every 100 electors. Women are 55.2% of those removed, against 47.6% of the district's roll. Names classified Muslim are 7.8% to 12.3% of removals, against a 2011 census share of 7.3%. The heaviest constituency is Tarari, at 10.66 per 100 electors, and the largest multiple of the margin is in Sandesh, where 25,665 names were removed against a margin of 27. In 2025 the NDA won 7 of the district's seats and INDIA 0.

Of its 1,90,953 phase 1 removals, the officers entered 42.5% as shifted, 36.8% as dead, 7.9% as absent and 12.8% as already enrolled.

Siwan

Siwan lost 2,28,135 names across 8 constituencies, 9.28 for every 100 electors. Women are 57.8% of those removed, against 47.9% of the district's roll. Names classified Muslim are 15.3% to 23.5% of removals, against a 2011 census share of 18.3%. The heaviest constituency is Darauli, at 11.18 per 100 electors, and the largest multiple of the margin is in Ziradei, where 19,626 names were removed against a margin of 2,626. In 2025 the NDA won 7 of the district's seats and INDIA 1.

Of its 2,21,556 phase 1 removals, the officers entered 35.8% as shifted, 43.0% as dead, 10.8% as absent and 10.4% as already enrolled.

Vaishali

Vaishali lost 2,35,676 names across 8 constituencies, 9.27 for every 100 electors. Women are 52.0% of those removed, against 47.5% of the district's roll. Names classified Muslim are 10.2% to 17.0% of removals, against a 2011 census share of 9.6%. The heaviest constituency is Hajipur, at 12.06 per 100 electors, and the largest multiple of the margin is in Raghopur, where 35,382 names were removed against a margin of 14,532. In 2025 the NDA won 7 of the district's seats and INDIA 1.

Of its 2,25,793 phase 1 removals, the officers entered 32.6% as shifted, 45.2% as dead, 13.9% as absent and 8.4% as already enrolled.

Supaul

Supaul lost 1,37,607 names across 5 constituencies, 8.93 for every 100 electors. Women are 58.4% of those removed, against 48.4% of the district's roll. Names classified Muslim are 19.0% to 27.2% of removals, against a 2011 census share of 18.4%. The heaviest constituency is Supaul, at 11.29 per 100 electors, and the largest multiple of the margin is in Triveniganj, where 23,705 names were removed against a margin of 5,683. In 2025 the NDA won 5 of the district's seats and INDIA 0.

Of its 1,28,212 phase 1 removals, the officers entered 48.6% as shifted, 25.7% as dead, 13.6% as absent and 12.1% as already enrolled.

Begusarai

Begusarai lost 1,90,817 names across 7 constituencies, 8.91 for every 100 electors. Women are 54.0% of those removed, against 47.4% of the district's roll. Names classified Muslim are 12.8% to 20.4% of removals, against a 2011 census share of 13.7%. The heaviest constituency is Begusarai, at 12.58 per 100 electors, and the largest multiple of the margin is in Matihani, where 33,902 names were removed against a margin of 5,290. In 2025 the NDA won 5 of the district's seats and INDIA 2.

Of its 1,67,659 phase 1 removals, the officers entered 46.6% as shifted, 31.2% as dead, 11.3% as absent and 10.9% as already enrolled.

Aurangabad

Aurangabad lost 1,63,211 names across 6 constituencies, 8.89 for every 100 electors. Women are 56.8% of those removed, against 47.7% of the district's roll. Names classified Muslim are 12.1% to 17.5% of removals, against a 2011 census share of 9.3%. The heaviest constituency is Aurangabad, at 9.83 per 100 electors, and the largest multiple of the margin is in Nabinagar, where 24,857 names were removed against a margin of 112. In 2025 the NDA won 5 of the district's seats and INDIA 1.

Of its 1,59,937 phase 1 removals, the officers entered 41.6% as shifted, 33.4% as dead, 13.0% as absent and 12.0% as already enrolled.

Muzaffarpur

Muzaffarpur lost 2,94,340 names across 11 constituencies, 8.89 for every 100 electors. Women are 53.9% of those removed, against 47.7% of the district's roll. Names classified Muslim are 15.2% to 23.1% of removals, against a 2011 census share of 15.5%. The heaviest constituency is Muzaffarpur, at 11.07 per 100 electors, and the largest multiple of the margin is in Kurhani, where 24,567 names were removed against a margin of 9,718. In 2025 the NDA won 10 of the district's seats and INDIA 1.

Of its 2,82,574 phase 1 removals, the officers entered 36.2% as shifted, 39.3% as dead, 14.4% as absent and 10.1% as already enrolled.

Araria

Araria lost 1,75,087 names across 6 constituencies, 8.84 for every 100 electors. Women are 54.3% of those removed, against 48.2% of the district's roll. Names classified Muslim are 28.4% to 46.7% of removals, against a 2011 census share of 42.9%. The heaviest constituency is Raniganj, at 10.53 per 100 electors, and the largest multiple of the margin is in Forbesganj, where 27,262 names were removed against a margin of 221. In 2025 the NDA won 2 of the district's seats and INDIA 3, and 1 went to neither alliance.

Of its 1,57,731 phase 1 removals, the officers entered 32.3% as shifted, 45.4% as dead, 10.6% as absent and 11.8% as already enrolled.

Patna

Patna lost 4,15,580 names across 14 constituencies, 8.58 for every 100 electors. Women are 50.5% of those removed, against 47.4% of the district's roll. Names classified Muslim are 8.8% to 14.5% of removals, against a 2011 census share of 7.5%. The heaviest constituency is Digha, at 12.58 per 100 electors, and the largest multiple of the margin is in Bakhtiarpur, where 17,924 names were removed against a margin of 981. In 2025 the NDA won 11 of the district's seats and INDIA 3.

Of its 3,95,080 phase 1 removals, the officers entered 39.2% as shifted, 34.0% as dead, 18.6% as absent and 8.2% as already enrolled.

Gaya

Gaya lost 2,53,055 names across 10 constituencies, 8.44 for every 100 electors. Women are 55.2% of those removed, against 48.0% of the district's roll. Names classified Muslim are 13.0% to 21.1% of removals, against a 2011 census share of 11.1%. The heaviest constituency is Tikari, at 11.04 per 100 electors, and the largest multiple of the margin is in Bodh Gaya, where 17,168 names were removed against a margin of 881. In 2025 the NDA won 8 of the district's seats and INDIA 2.

Of its 2,45,458 phase 1 removals, the officers entered 39.1% as shifted, 36.9% as dead, 10.3% as absent and 13.6% as already enrolled.

Banka

Banka lost 1,21,034 names across 5 constituencies, 8.31 for every 100 electors. Women are 52.9% of those removed, against 47.7% of the district's roll. Names classified Muslim are 10.2% to 20.0% of removals, against a 2011 census share of 12.3%. The heaviest constituency is Banka, at 9.70 per 100 electors, and the largest multiple of the margin is in Katoria, where 20,155 names were removed against a margin of 10,986. In 2025 the NDA won 5 of the district's seats and INDIA 0.

Of its 1,17,219 phase 1 removals, the officers entered 46.9% as shifted, 26.0% as dead, 13.9% as absent and 13.3% as already enrolled.

West Champaran

West Champaran lost 2,17,300 names across 9 constituencies, 8.26 for every 100 electors. Women are 54.1% of those removed, against 47.2% of the district's roll. Names classified Muslim are 17.0% to 28.4% of removals, against a 2011 census share of 22.0%. The heaviest constituency is Lauriya, at 10.79 per 100 electors, and the largest multiple of the margin is in Chanpatia, where 14,829 names were removed against a margin of 602. In 2025 the NDA won 7 of the district's seats and INDIA 2.

Of its 1,91,271 phase 1 removals, the officers entered 40.3% as shifted, 37.6% as dead, 10.9% as absent and 11.2% as already enrolled.

Munger

Munger lost 82,185 names across 3 constituencies, 8.19 for every 100 electors. Women are 49.7% of those removed, against 46.9% of the district's roll. Names classified Muslim are 8.0% to 16.2% of removals, against a 2011 census share of 8.1%. The heaviest constituency is Tarapur, at 8.34 per 100 electors, and the largest multiple of the margin is in Munger, where 27,022 names were removed against a margin of 18,750. In 2025 the NDA won 3 of the district's seats and INDIA 0.

Of its 74,877 phase 1 removals, the officers entered 35.4% as shifted, 47.8% as dead, 4.7% as absent and 12.1% as already enrolled.

Jamui

Jamui lost 1,00,975 names across 4 constituencies, 7.92 for every 100 electors. Women are 56.6% of those removed, against 48.1% of the district's roll. Names classified Muslim are 13.3% to 23.8% of removals, against a 2011 census share of 12.4%. The heaviest constituency is Jamui, at 9.54 per 100 electors, and the largest multiple of the margin is in Jhajha, where 20,808 names were removed against a margin of 4,262. In 2025 the NDA won 3 of the district's seats and INDIA 1.

Of its 91,800 phase 1 removals, the officers entered 47.2% as shifted, 30.9% as dead, 10.1% as absent and 11.8% as already enrolled.

Madhepura

Madhepura lost 1,04,304 names across 4 constituencies, 7.65 for every 100 electors. Women are 55.9% of those removed, against 48.3% of the district's roll. Names classified Muslim are 12.3% to 20.7% of removals, against a 2011 census share of 12.1%. The heaviest constituency is Alamnagar, at 9.37 per 100 electors, and the largest multiple of the margin is in Singheshwar, where 22,947 names were removed against a margin of 2,982. In 2025 the NDA won 3 of the district's seats and INDIA 1.

Of its 98,053 phase 1 removals, the officers entered 50.1% as shifted, 34.9% as dead, 4.3% as absent and 10.6% as already enrolled.

Nawada

Nawada lost 1,30,263 names across 5 constituencies, 7.55 for every 100 electors. Women are 51.9% of those removed, against 48.0% of the district's roll. Names classified Muslim are 12.3% to 21.9% of removals, against a 2011 census share of 11.0%. The heaviest constituency is Hisua, at 10.28 per 100 electors, and the largest multiple of the margin is in Rajauli, where 20,850 names were removed against a margin of 3,953. In 2025 the NDA won 4 of the district's seats and INDIA 1.

Of its 1,26,248 phase 1 removals, the officers entered 35.8% as shifted, 41.3% as dead, 7.4% as absent and 15.4% as already enrolled.

Khagaria

Khagaria lost 85,658 names across 4 constituencies, 7.32 for every 100 electors. Women are 54.5% of those removed, against 47.7% of the district's roll. Names classified Muslim are 10.8% to 18.0% of removals, against a 2011 census share of 10.5%. The heaviest constituency is Khagaria, at 8.99 per 100 electors, and the largest multiple of the margin is in Khagaria, where 23,509 names were removed against a margin of 23,415. In 2025 the NDA won 4 of the district's seats and INDIA 0.

Of its 79,534 phase 1 removals, the officers entered 47.9% as shifted, 34.0% as dead, 6.7% as absent and 11.4% as already enrolled.

Darbhanga

Darbhanga lost 2,10,711 names across 10 constituencies, 7.26 for every 100 electors. Women are 56.8% of those removed, against 48.0% of the district's roll. Names classified Muslim are 21.9% to 32.7% of removals, against a 2011 census share of 22.4%. The heaviest constituency is Bahadurpur, at 9.91 per 100 electors, and the largest multiple of the margin is in Gaura Bauram, where 20,236 names were removed against a margin of 5,669. In 2025 the NDA won 10 of the district's seats and INDIA 0.

Of its 2,03,032 phase 1 removals, the officers entered 43.7% as shifted, 30.9% as dead, 13.6% as absent and 11.8% as already enrolled.

Buxar

Buxar lost 93,120 names across 4 constituencies, 7.19 for every 100 electors. Women are 60.3% of those removed, against 48.2% of the district's roll. Names classified Muslim are 6.5% to 10.5% of removals, against a 2011 census share of 6.2%. The heaviest constituency is Buxar, at 9.83 per 100 electors, and the largest multiple of the margin is in Dumraon, where 22,651 names were removed against a margin of 2,105. In 2025 the NDA won 3 of the district's seats and INDIA 1.

Of its 87,629 phase 1 removals, the officers entered 48.4% as shifted, 36.0% as dead, 7.1% as absent and 8.4% as already enrolled.

Jehanabad

Jehanabad lost 57,619 names across 3 constituencies, 7.09 for every 100 electors. Women are 54.1% of those removed, against 47.5% of the district's roll. Names classified Muslim are 9.4% to 15.2% of removals, against a 2011 census share of 6.7%. The heaviest constituency is Ghosi, at 7.66 per 100 electors, and the largest multiple of the margin is in Jehanabad, where 22,793 names were removed against a margin of 793. In 2025 the NDA won 1 of the district's seats and INDIA 2.

Of its 52,918 phase 1 removals, the officers entered 43.1% as shifted, 33.1% as dead, 14.8% as absent and 9.0% as already enrolled.

Rohtas

Rohtas lost 1,57,939 names across 7 constituencies, 7.06 for every 100 electors. Women are 57.3% of those removed, against 47.8% of the district's roll. Names classified Muslim are 9.8% to 15.0% of removals, against a 2011 census share of 10.2%. The heaviest constituency is Chenari, at 8.22 per 100 electors, and the largest multiple of the margin is in Karakat, where 25,162 names were removed against a margin of 2,836. In 2025 the NDA won 6 of the district's seats and INDIA 1.

Of its 1,56,085 phase 1 removals, the officers entered 39.5% as shifted, 41.9% as dead, 7.4% as absent and 11.2% as already enrolled.

Lakhisarai

Lakhisarai lost 51,608 names across 2 constituencies, 6.88 for every 100 electors. Women are 53.8% of those removed, against 47.1% of the district's roll. Names classified Muslim are 5.7% to 14.6% of removals, against a 2011 census share of 4.1%. The heaviest constituency is Suryagarha, at 6.92 per 100 electors, and the largest multiple of the margin is in Lakhisarai, where 26,844 names were removed against a margin of 24,940. In 2025 the NDA won 2 of the district's seats and INDIA 0.

Of its 48,824 phase 1 removals, the officers entered 40.5% as shifted, 37.2% as dead, 9.7% as absent and 12.6% as already enrolled.

Kaimur

Kaimur lost 78,779 names across 4 constituencies, 6.71 for every 100 electors. Women are 61.6% of those removed, against 47.9% of the district's roll. Names classified Muslim are 9.9% to 16.1% of removals, against a 2011 census share of 9.5%. The heaviest constituency is Mohania, at 7.20 per 100 electors, and the largest multiple of the margin is in Ramgarh, where 19,225 names were removed against a margin of 30. In 2025 the NDA won 3 of the district's seats and INDIA 0, and 1 went to neither alliance.

Of its 73,900 phase 1 removals, the officers entered 54.1% as shifted, 33.5% as dead, 5.0% as absent and 7.5% as already enrolled.

Nalanda

Nalanda lost 1,48,457 names across 7 constituencies, 6.60 for every 100 electors. Women are 52.1% of those removed, against 47.6% of the district's roll. Names classified Muslim are 8.4% to 16.1% of removals, against a 2011 census share of 6.9%. The heaviest constituency is Islampur, at 9.01 per 100 electors, and the largest multiple of the margin is in Hilsa, where 24,545 names were removed against a margin of 16,012. In 2025 the NDA won 7 of the district's seats and INDIA 0.

Of its 1,38,308 phase 1 removals, the officers entered 33.0% as shifted, 47.3% as dead, 6.3% as absent and 13.5% as already enrolled.

Arwal

Arwal lost 32,872 names across 2 constituencies, 6.23 for every 100 electors. Women are 59.1% of those removed, against 47.9% of the district's roll. Names classified Muslim are 9.3% to 14.0% of removals, against a 2011 census share of 9.2%. The heaviest constituency is Arwal, at 6.26 per 100 electors, and the largest multiple of the margin is in Kurtha, where 16,073 names were removed against a margin of 5,481. In 2025 the NDA won 2 of the district's seats and INDIA 0.

Of its 30,107 phase 1 removals, the officers entered 39.4% as shifted, 42.0% as dead, 2.9% as absent and 15.7% as already enrolled.

Sheikhpura

Sheikhpura lost 27,148 names across 2 constituencies, 5.47 for every 100 electors. Women are 56.9% of those removed, against 48.1% of the district's roll. Names classified Muslim are 8.2% to 16.5% of removals, against a 2011 census share of 5.9%. The heaviest constituency is Barbigha, at 5.84 per 100 electors, and the largest multiple of the margin is in Sheikhpura, where 13,577 names were removed against a margin of 22,547. In 2025 the NDA won 2 of the district's seats and INDIA 0.

Of its 26,242 phase 1 removals, the officers entered 41.5% as shifted, 41.1% as dead, 4.5% as absent and 12.8% as already enrolled.

Conclusion and what the Election Commission should do

The revision removed 69,10,559 names from Bihar's roll in the months before an Assembly election. The names removed exceed the margin in 174 of 243 seats and equal 13.9% of all votes polled. Whatever reason was written against each name, in most of the state the roll changed by more than the vote that decided the seat.

Women carried a larger share of the removals than their share of the roll in every district, and the lists cannot show whether those women were enrolled again where they now live. The reasons, dead, shifted, absent or enrolled twice, are what officers wrote down, and nobody checked them. 9,70,081 names were removed as absent, the reason that rests on the least, and those alone exceed the margin in 20 seats.

What the Election Commission should do:

1. Publish the reason for each of the 3,51,150 phase 2 removals, name by name. The lists give none, and the Format 4B return gives reasons only by constituency, with 9,968 names left without any.
2. Publish the basis for each entry marked absent or already enrolled: 9,70,081 marked absent and 7,01,059 marked already enrolled in phase 1.
3. Publish every claim and objection filed after the draft roll, booth by booth, with its outcome, and say how many of the 21.5 lakh added had been removed at the draft, so that restorations can be counted against removals.
4. Publish new enrolments by constituency after the revision, so that women removed as shifted can be matched against enrolments where they now live.

Method and sources

• Removals by polling booth come from the phase 1 deletion lists that the Chief Electoral Officer, Bihar published with the draft roll of 1 August 2025, for all 77,837 booths.
• Removals by constituency, phase and sex come from the Chief Electoral Officer, Bihar.
• The final roll is from Format 4B, the Election Commission's return for the final roll of 30 September 2025, as published by the Chief Electoral Officer, Bihar, by constituency and sex. The same return gives the draft roll by sex, which with the phase 1 removals gives women's share of the roll before the revision.
• Religion and surnames of the removed are counted by booth from the same lists, with no individual records kept.
• The reason, sex and age of every phase 1 removal are counted by constituency and booth from the deletion list, with no individual records kept.
• Deaths come from the Sample Registration System Statistical Report 2024 and the Civil Registration System 2024, and marriage migration from Census of India 2011, table D-3.
• The 2025 Assembly results give the winner, runner-up and turnout for each constituency, checked across two separate tables. Where they disagreed, the figure matching the reported result was used.
• The 2024 Lok Sabha result by booth is from Form 20. The other sources are the Census of India 2001 and 2011 for religion by district, the National Family Health Survey 2019-20 for women's literacy, schooling and marriage before 18 by district, the Bihar caste-based survey of 2 October 2023, and the Bihar Economic Survey.

Each booth is counted once. A second program, written separately, recomputes every headline figure from the lists before publication. Votes polled are the turnout applied to the 2025 roll. No name, EPIC number or address is published.