Caste enumeration in Census 2027 is both a statistical exercise and a politically sensitive undertaking. Comment.

Census 2027 — India's 16th census and its first fully digital one — will record caste following the Cabinet Committee on Political Affairs decision of 30 April 2025 [1], the first such count since 1931. It is therefore simultaneously a measurement problem and a question of political recognition; its credibility rests on getting both right.

As a statistical exercise

• Scale and design: Phase I (Houselisting) ran April–September 2026; Phase II (Population Enumeration), which carries the caste column, is scheduled for February 2027, using a mobile app, optional self-enumeration and the Census Management & Monitoring System (CMMS) for near real-time supervision [1].
• The taxonomy problem: caste is recorded as declared, not matched to a fixed list. SECC 2011 threw up roughly 46 lakh caste, sub-caste, surname and clan names, compelling an ex-post Expert Group under the NITI Aayog Vice-Chairperson to classify them [2].
• Technology has limits: digital validation catches blank fields and invalid entries, not whether "Yadav", "Ahir" and "Gwala" are one category or three. The Registrar-General's caution to States that Phase II's margin for error is narrower speaks to precisely this class of error [3].
• Human layer: nearly 34 lakh enumerators, mostly school staff, are trained through a cascade, where uniformity of instruction is hard to preserve [1].

As a politically sensitive undertaking

• Delimitation stake: Census 2027 is the first census after 2026, the trigger for redrawing Lok Sabha seats after the freeze was lifted [4].
• Reservation arithmetic: OBC sub-categorisation and quota ceilings will be argued off these figures, so an undercount is a distributional loss, not a rounding error.
• No bounded universe: unlike SC/ST enumeration, which is anchored in the Presidential Lists under Articles 341–342, the general caste column has no notified schedule to validate against, inviting contestation.

Accuracy and acceptability are thus the same problem. Publishing a standardisation directory before February 2027, logging enumerator edits, correcting errors at source through CMMS, and committing to a release calendar would let Census 2027 deliver what SECC could not — caste data that is both statistically defensible and publicly trusted.