Method and research questions
Sources
| What | Source | Used for |
|---|---|---|
| Pre-revision roll, Absent-Shifted-Dead list, discrepancy and no-mapping lists | CEO Telangana SIR lists for the five constituencies, compiled elector by elector and aggregated to the polling part, September 2026 | Each elector's outcome, age band, sex, reason code and 2002 link, added up by polling part |
| Religion | Guessed from each elector's name, because the roll does not record religion. A separate classification agrees on 99% of names | Hindu-name and Muslim-name rates |
| Hearing schedules | The same lists, 28 August to 12 October | Rules, slots, daily load |
| 2023 Assembly result | ECI, Telangana general election 2023 | Valid votes and margins |
Every rate, within-part comparison and correlation on these pages is computed from the part-level counts shown in the explorer. No name, EPIC number or address is published.
Definitions
Deletion rate
Electors on the Absent-Shifted-Dead list divided by electors on the roll before the revision, for the group in question. The files cover every elector, so this is a true rate for each group.
Within-part difference
For two groups, the rate of one minus the rate of the other inside each polling part, pooled across parts and weighted by the smaller group in each part. Parts with fewer than 20 of either group are left out. This holds the neighbourhood, the booth-level officer and the housing constant.
Part correlations
Spearman rank correlations across parts with at least 100 electors, within each constituency. The untraceable share uses parts with at least 20 deletions.
Summoned
Kept on the draft roll but listed for a hearing, from the ECI discrepancy list and the no-mapping list together. One elector can be flagged under more than one rule, so the rule counts add up to more than the number of people summoned.
What this shows and what it does not
The lists show how many electors were struck off and from which parts, with each person's age and sex, a religion guessed from the name, and whether they link to the 2002 roll. They cannot show whether any single removal was right. Questions and answers.
The research questions
Each answer is computed from the published data when this page loads.