The Column That Broke the Last Census

A Column India Has Failed to Design Before The

Centre's notification of the question set for Census 2027's second phase has reopened a design problem India has faced exactly once before, at real scale - and did not solve.

What Self-Reporting Promises Letting a

respondent state their own caste, in their own words, respects a basic principle: caste identity in India is not a fixed administrative category handed down from above, it is something communities and individuals describe differently across regions, sub-groups and generations. A pre-set list cannot capture this. Any fixed dropdown of caste names, however long, will always be missing someone's actual sub-caste, forcing them into an approximate box or an "other" category that undercounts precisely the groups a caste census exists to count accurately. Self-reporting is how other identity questions already work. Religion and mother tongue are self-declared in the Indian census, not assigned from a state-approved list and caste has historically been argued for the same treatment.

What Self-Reporting Has Already Cost

The precedent is not hypothetical. The 2011 SECC used exactly this open self-reporting design for caste and it generated roughly 46 lakh distinct caste and sub-caste name entries - spelling variants, sub-group names, regional variants of the same community and genuine data-entry noise, all mixed together with no reliable way to separate them. The consequence was not a delay - it was non-delivery. The government never officially released the SECC's caste data at all, because classifying that volume of free-text entries into a usable, policy-ready dataset proved unworkable within any reasonable timeframe. A caste census whose caste data cannot be released is not a partial success; for the purpose it exists to serve - accurate information for welfare targeting and reservation policy - it is functionally no census at all.

The Real Choice Being Made

Every caste enumeration design sits somewhere on a line between two failure modes: a closed list that undercounts real diversity or an open list that produces data too fragmented to use. Census 2027, as a digital, tablet-based exercise rather than SECC's paper-based one, has tools SECC did not: real-time coding assistance, a backend caste thesaurus that can suggest standardised entries as an enumerator types and the ability to flag ambiguous responses for supervisor review before the household visit even ends.

Whether the second phase's design actually uses these tools or simply digitises the same open-ended write-in field that broke in 2011, is the question this notification does not yet answer - and it is the question that will determine whether 2027 produces usable social justice data or an even larger unclassifiable list than the one no government has ever released.

Why This Matters Beyond the Form

A caste census that cannot be turned into usable data is not a neutral outcome - it is a decision, by default, to keep reservation policy running on data that is now over three decades old. Getting the caste column right is not a technical footnote to this census. It is the entire point of conducting one.