Including caste questions in the census could transform current identity politics into more effective policymaking to address long-standing inequality. Accurate caste data, combined with the demographic and contextual data from the census could shed light on how ?backward? some of the ?other backward classes? actually are. They could quantify variation in economic and social outcomes within caste groups across states and provide a more nuanced picture of what ?backward? means. The data might demonstrate that inter-caste gaps in education, occupation and access to infrastructure are far greater than we think; but they could also diffuse caste-based claims or redefine the list of those eligible for preferential treatment. We don?t know. The production and scrutiny of data could offend those who would prefer to ignore caste, but it could also be a big step towards banishing both the myths and realities of its impact.

The problem is that the data to provoke this transformation are not possible to collect unless the government credibly pre-commits to transform its approach. It has to convince respondents that it will actually use the data to inform a new designation as well as the number of ?backward classes?.

The technical difficulties of representing caste should not be trivialised, but they are surmountable. Lots of castes? Add more options?cumbersome for a paper survey, but not bad in digital one. Multiple caste names for multiple contexts? Offer multiple choice that incorporate the ontology experts have developed and used to critique existing lists. Multiple choice also works for mixed-caste parentage and unwillingness to just assume the official designation of father?s caste. Evolving castes? Complicates comparability over time, but there is always the option of ?other? with write-in. The inevitable lobbying for representation on the list of castes?as has happened with race categories in the US Census?is also a natural corrective mechanism. The data may never be perfect, but they could be acceptable.

The pragmatic difficulty of ensuring that people report caste correctly is harder to tackle. Today?s implicit ?deal? is that a particular list of castes currently considered as disadvantaged has access to quotas, reservations, and other benefits in some proportion to its share of the population. The census will clearly affect the debate by providing an updated official figure for the share of the population. Unless an additional commitment is made to use the data to assess the list of the disadvantaged and adjust the policy response accordingly, this figure is very likely to be dangerously wrong.

Individuals left to themselves might report their true caste, but the chances that they will be left to themselves are slim when rewards for the overall caste numbers are so significant. Advocacy campaigns and initiatives to influence individuals? census responses are a fact of life and are often effective. Ethnic advocacy groups in the US, for example, used YouTube, Facebook, print and TV to urge US residents to respond to the race question in the 2010 Census in ways that will likely build up their numbers. The Arya Samaj?s efforts to encourage Hindu Punjabi speakers to state their language as Hindi rather than Punjabi as a way of blocking linguistic partition of Punjab are credited with reducing the recorded number of Punjabi speakers in Punjab from 60% in pre-Independence censuses to 41% in 1961.

Practice in vote-buying can easily be converted to skill in census-buying. The fact that the data for the new National Population Register are being collected at the same time only makes the job easier. The NPR, ?a comprehensive identity database that would help in better targeting of the benefits and services under the government schemes/programmes,? (to quote from the Census Web site) is separate, but it would not be hard to spread the rumour that answers to one questionnaire could help influence the individual eligibility profile being collected on the other. Such caste data, once collected, would be hard to discredit without damaging the credibility of the entire census. It would have done all of the damage of re-introducing official recognition of caste without any of the benefits of getting a more realistic picture.

Commitment to review the ?backward? label would reduce the incentive and ability to manipulate answers?it is a more complex task to engineer responses to ensure that the number of members of a caste looks large while the economic and contextual characteristics look unusually bad.

If the government is going to open the door by counting caste, it has to promise to march into the room and take a look. The question, then, is how. Committing to a new Caste Commission required to use census among other data to re-assess the level and nature of backwardness in addition to the proportion of ?backward? groups would be a start.

The author is director, Centre for Development Finance at the Institute for Financial and Management Research, Chennai