Will Census 2027 Deliver Reliable Caste-Wise Population Data?
India’s Census 2027 marks a historic shift. For the first time since 1931, the national population count will systematically record caste beyond Scheduled Castes and Scheduled Tribes. The decision, approved by the Cabinet Committee on Political Affairs, responds to long-standing political and social demands for updated data that can inform affirmative action, welfare targeting and debates on representation. Yet as the questionnaire for the population enumeration phase has been notified, a critical question hangs over the exercise: will the resulting caste-wise figures prove reliable enough for policy use?
The answer hinges largely on methodology. Question number 10 on the household schedule reads “Scheduled Caste (SC)/Scheduled Tribe (ST)/Caste.” For SC and ST communities, enumerators will continue to use the established notified lists through dropdown menus. For everyone else, the caste will be recorded as declared by the respondent in an open field. Officials have confirmed that enumerators will simply note what people say, without mapping responses to any pre-prepared national or state list of Other Backward Classes or other categories. This open-ended approach was tested during pre-tests in 16 states and Union Territories and has been retained for the main exercise.
Caste in India is not a simple administrative label. It is a complex social construct expressed through multiple names, sub-castes, surnames, clan identities and traditional occupations. The same community may identify itself differently across regions or even within the same household. Pastoral groups in the north, for instance, have historically returned themselves as Ahir, Goala, Gopi or Idaiyan. Trading communities may report Gupta, Agarwal or simply Baniya. When respondents are free to state their identity in their own words, the resulting data quickly multiplies into variants, spellings and near-synonyms.
This is precisely what happened during the Socio-Economic and Caste Census of 2011. That separate exercise, also based on open-ended self-reporting, produced roughly 46 to 46.7 lakh distinct caste names. By comparison, the 1931 Census, the last comprehensive caste enumeration, recorded about 4,147 castes. The Union government later told the Supreme Court that the SECC figures were too error-ridden to be used for reservations or other statutory purposes. More than 99 per cent of the recorded names had populations below 100, and phonetic variations and enumerator inconsistencies further degraded the quality. The raw caste data was never released for official use.
Census 2027 risks repeating the same pattern. Without a controlled vocabulary or standardised coding at the point of collection, the raw returns will almost certainly contain large numbers of near-duplicates and regional variants. Converting those free-text entries into clean, comparable tables will require extensive post-enumeration cleaning, expert judgment and alignment with the Central list of about 2,650 OBC communities and the varying state lists. That process itself introduces subjectivity. Officials have argued that a dropdown menu would force the state into the role of arbiter of caste identity, which the Census is not designed to do. Census, they note, is a recording exercise, not a verification exercise. Enumerators will not demand caste certificates or challenge self-declarations.
The digital nature of Census 2027 offers some hope of improvement over 2011. Handheld devices, possible self-enumeration options and modern data-processing tools, including AI-assisted matching of variants, could reduce clerical errors and speed up standardisation. State-level caste surveys in Bihar and Telangana demonstrated that a curated list approach can produce more coherent results. Bihar used a predetermined list of around 215 categories and released detailed socio-economic tabulations. Those exercises, however, were smaller in scale and operated within a single administrative framework. Scaling a similar model nationally is complicated by interstate differences in OBC classification and the continuous expansion of state lists.
Political context adds another layer of complexity. Caste data carries direct implications for reservation policy, sub-categorisation of quotas and the identification of the creamy layer. In such an environment, strategic self-reporting cannot be ruled out. Communities may emphasise identities that maximise perceived benefits, or individuals may report differently depending on who is asking the question. Enumerator bias, already a recognised challenge in large-scale surveys, could further influence outcomes.
Proponents of the open-ended method argue that it respects the fluid and self-perceived nature of caste identity and avoids freezing colonial-era or bureaucratic categories. Critics, including opposition parties and several commentators, counter that usable data requires structure. An open question may generate raw material, but without transparent coding rules and public consensus on classification, the final tables risk remaining contested or incomplete. Some have called for a publicly scrutinised list of castes before enumeration begins, arguing that credibility depends on methodological clarity rather than political convenience.
Even if the headcount itself is accurate, the value of caste data depends on what can be done with it. One of the stated objectives is to understand the socio-economic conditions of deprived sections within the backward classes and enable better-targeted interventions. That requires not only population numbers but also the ability to cross-tabulate caste with education, occupation, income proxies and other indicators collected in the same Census. Fragmented or inconsistent caste categories will limit the analytical power of the dataset.
Census 2027 will still be a significant improvement over the near-century-old 1931 figures that have long served as a reference point. Statutory backing under the Census Act, stronger confidentiality protections and the digital backbone distinguish it from the SECC. SC and ST data will remain relatively robust. For the broader caste landscape, however, reliability will rest less on the enumeration phase and more on the quality, transparency and acceptance of the subsequent cleaning and classification process.
India has waited decades for comprehensive caste data. The decision to include it in the official Census is therefore welcome. Yet the choice of an open-ended format, while defensible on grounds of neutrality, places a heavy burden on post-collection work. If that work is rigorous, expert-driven and open to scrutiny, the resulting figures can still serve as a useful foundation for evidence-based policy. If it is not, the country may once again find itself with a large volume of data that is difficult to interpret or apply. The true test of reliability will come not when the enumerators finish their rounds, but when the tables are published and policymakers, researchers and the public begin to use them.
About The Author
