Women, religion, caste
Gender
Women among the removed, by district
Each district's share of women among those removed, against women's share of its roll before the revision. A point above the line is a district where women were removed out of proportion.
By the alliance that won the seat in 2025
Women's share of removals in each group of seats.
Women's share of each reason, by age
Every phase 1 removal has one of four reasons against it. For each reason and age group, this shows the share of those removed who were women. Half would be parity.
Two reasons, set against the records
The names entered as dead are compared with the number of adults who die in Bihar in a year, from the Sample Registration System for 2024. The excess of young women marked shifted is compared with the women who moved on marriage, from the 2011 census.
Young women marked shifted, against marriage movement, by district
Each dot is a district: the women's share of those aged 18 to 29 marked shifted, against the share of its women who had moved on marriage in the ten years before the 2011 census.
Religion
The booth files classify each removed elector's name as Hindu, Muslim or unclassified. An unclassified name could be either, so the Muslim share is given as a range, from counting none of them as Muslim to counting all of them.
Muslim share of removals against the 2011 census, by district
The bar is the range from name classification. The dark dot is the district's Muslim share of the population in 2011, and the open dot is that share carried forward to 2025. Where a dot sits inside the bar, the removals match the population.
The 2011 census is the most recent count of religion by district. The 2025 figures carry each district's 2011 share forward at the growth rates of its Muslim and total populations between the 2001 and 2011 censuses. Jehanabad and Arwal were one district in 2001, so both take their joint rate.
Caste
The lists give a surname for each removed elector and for the father or husband named beside them. The father's or husband's surname tells more, because most women are listed as Devi, Kumari or Khatun, which say nothing about caste. Surnames are classified in two ways. Classification A uses the project's mapping of surnames to castes. Classification B is strict and counts only surnames that belong almost entirely to one group, setting aside names such as Singh, Prasad, Ray, Mahato, Mandal, Sharma, Thakur, Chaudhary and Manjhi, which several groups carry. Only findings that hold under both are reported.
How many removals each classification can place
Share of all removals each classification could place, and why the rest could not be placed.
Composition of the classified, against the 2023 survey
Each group's share of the removals that could be classified, under A and B, beside its share of Bihar's population in the 2023 caste survey. Hindu castes are shown separately and Muslims as one group.
Each group's possible share of all removals
Under the strict classification B, the left end counts only names placed in the group and the right end adds every name that could not be placed.
The 25 commonest surnames among the removed
Counting the same surnames on the whole draft roll would turn each surname's share of the removed into a removal rate. Surname counts from the draft roll, even for a sample of booths, would give each caste group its own removal rate, as age and religion have in the Hyderabad analysis.
The caste baseline is the Bihar caste-based survey released on 2 October 2023. It counted 13,07,25,310 people, of whom 36.01% were EBC, 27.12% BC, 19.65% SC, 1.68% ST and 15.52% unreserved, with Muslims at 17.70%. Muslim castes listed under EBC (10.29 points), BC (2.09) and unreserved (4.80) are moved to the Muslim group here, which leaves the Hindu castes at EBC 25.72, BC 25.03 and General 10.72. The strict classification gives a surname to a group only if almost everyone carrying it belongs to that group.