NEW DELHI: The government’s decision to use an open-ended format to record caste in the upcoming Census could produce unusable data, much like the 2011 Socio-Economic and Caste Census (SECC), says J K Bajaj, a member of the Justice G Rohini Commission for the sub-categorization of the Other Backward Classes (OBCs). In an interview with TNIE, Bajaj said caste identities must be matched against sanitized Central and state lists during enumeration and that the government should have used the Commission’s extensive work on the Central OBC list.
Q) The upcoming Census will use an open-ended format for recording caste names instead of a drop-down menu for Other Backward Classes (OBCs).. What problems could arise from this? Can artificial intelligence resolve the discrepancies later?
I do not see how an open-ended field can produce usable data. There are claims that caste lists will be available at the back end and geo-fenced for each state and district, but enumerators must record exactly what respondents say in the local script. This could generate enormous variations.
Balija, for instance, may be pronounced as Balaja or Balija and recorded as separate castes. Kapu and Telaga may denote the same caste, though there are also different Kapu groups. Names such as Bhandari and Modi can refer to entirely different communities and castes, depending on the region. A name alone cannot establish caste identity; context is essential.
The claim that better AI models can clean up the data later is flawed for two reasons. First, no AI model can correct this kind of error. AI is not a social scientist. Second, the Census officials have no legal authority to club together separate entries after they have been recorded. Without a structured list, mapping these variations during enumeration, the Census could produce a huge number of irreconcilable responses.
Q) Without a drop-down menu, could the caste census repeat the failure of the 2011 SECC, which generated around 46 lakh responses?
If you follow the same method, you will get the same result. The SECC produced around 46 lakh caste responses, but the government later told the Supreme Court that its design was flawed and the data too inaccurate to be used. This exercise could be equally chaotic. Caste classification must happen during enumeration, with respondents confirming the appropriate entry from a structured list. Once an enumerator records whatever a person says, Census officials have no legal authority to treat differently recorded names as the same caste later.
Q) One of the Rohini Commission's terms of reference was to examine and sanitize the Central OBC list. Do you think the Commission’s report could have helped ensure a more accurate caste census?
We spent nearly three years sanitizing the Central OBC list- removing ambiguities and mapping different names and pronunciation variants to numbered caste entries.
The Commission’s sanitized list was particularly relevant to this Census. If it is not used now, it may never be used.
Q) How should the caste question have been designed?
The Central and state OBC lists should have been integrated into the enumeration system. Once a respondent gives a caste name, the device should display the corresponding entries in both lists. The enumerator could then confirm the correct entry with the respondent, resolving ambiguities at the enumeration stage itself.
Q) Why does the changed constitutional status of OBC lists matter for the caste census?
OBC lists were earlier administrative, unlike the constitutionally recognized SC and ST lists. The 102nd and 105th Constitutional Amendments gave OBC lists constitutional status. They should therefore be treated accordingly during enumeration. Otherwise, the amendments' purpose is defeated.
Q) The Rohini Commission report is yet to see the light of day.
The report was submitted to the President. The government’s position is that the executive has not received it and that it is the President’s prerogative whether and when to send it to the government.
This prolonged limbo has another consequence. Changes to the Central OBC list have effectively been stuck. No new recommendation has been processed since 2018. Under the changed constitutional framework, a Presidential list would first have to be issued before further changes can be made, and that was what we had prepared.
Q) Karnataka, Bihar and Telangana have undertaken elaborate caste surveys. Does that show this is feasible?
Yes. Karnataka created a comprehensive coded list of castes and gave people an opportunity to point out if something had been missed. Whatever other controversies there were, nobody eventually raised the issue that their caste had not been counted or had been wrongly counted. Karnataka took the Central list, state list and information from previous and pre-Independence censuses and created a comprehensive state caste list. It is doable. Something similar should have been done for every state and incorporated into the Census device. For OBCs, SCs and STs, we already have lists covering a very large part of the population.
Q) Do the caste surveys conducted by Karnataka, Bihar and Telangana show that a structured count is feasible?
Yes. Karnataka created a comprehensive coded caste list using the Central and state lists as well as earlier and pre-Independence Census records. It also allowed communities to flag omissions, and no major complaint eventually remained about castes being excluded or wrongly recorded. Every state could have undertaken a similar exercise and integrated the resulting list into Census devices. Existing OBC, SC and ST lists already cover a large majority of the population, making this entirely feasible.
Q) Why is reliable caste data important?
A properly conducted caste census would link caste identity with data on education, housing and other indicators, creating a demographic and socio-economic profile of each community. This would give commissions a ready database to assess social and educational backwardness instead of requiring fresh surveys for every caste.
Q) There are concerns that comprehensive caste data could trigger political churn similar to the Mandal period. Is that a reason for caution?
We cannot continue basing our polity on half-truths. Ultimately, we need accurate data to ensure our polity reflects the truth.