The Caste Lists

The source behind almost every caste figure still in use. Hutton's chapter on caste and his appendix on the “exterior castes” set out who was counted and why, and Table B gives the count: 50,195,770 people in the exterior castes in 1931, the category that became the Scheduled Castes. Read the definitions before the numbers. Who counted as “exterior” was settled province by province, so an all-India share adds together lines that were drawn in different places by different tests.

The strongest version of the argument that colonial rule made caste into the single category through which India was governed. Dirks does not claim the British invented caste. His claim is that the census and the ethnographic survey fixed identities that had been negotiable and gave them administrative consequences. Read the chapters on the census and on the “ethnographic state”, then read Bayly for the case that he overstates it.

The necessary counterweight to Dirks. Bayly traces how caste hardened between the eighteenth and twentieth centuries through kingship, trade, religious reform and colonial rule together, and argues that Indians were active in shaping the categories the census recorded. For a practitioner the useful lesson is that a caste name on a list is the outcome of a long contest, which is why the same community can appear under different names in different states.

The list that decides who is a Scheduled Caste, state by state. Read paragraph 3 before anything else: as amended in 1956 and 1990, it provides that no person who professes a religion other than Hinduism, Sikhism or Buddhism is deemed a member of a Scheduled Caste. That one paragraph is why the same community name, Dhobi or Nat in Bihar, appears on the Scheduled Caste list and again, marked “(Muslim)”, on the list of backward classes.

The report that put the backward classes into national policy. It listed about 3,700 communities state by state and, working from the 1931 Census because nothing later existed, estimated the backward classes at 52 per cent of the population. Read the chapter on the eleven indicators used to identify backwardness alongside the state lists in Part II: the method and the lists do not always agree, and the gap is where most later litigation began.

The judgment that upheld 27 per cent reservation for backward classes in central government posts, held that caste can be a criterion of backwardness, set the 50 per cent ceiling on reservation, and required the “creamy layer” to be excluded. It also directed the creation of a permanent body to revise the lists, which became the National Commission for Backward Classes. Every Central List entry since rests on this case.

By six to one, the Court held that states may sub-classify the Scheduled Castes and reserve a share of the quota for the castes within them that have benefited least, overruling E. V. Chinnaiah (2004). The decision turns the list from a single category into something a state may divide, which makes the quality of caste-level data a constitutional question: sub-classification has to be justified by evidence of relative backwardness, and in most states that evidence has not been collected.

The clearest statement of the case for a caste census, written during the argument over the 2011 Census. The authors' central point is that not counting caste is as much a political decision as counting it, and that a census which records only the disadvantaged castes leaves the advantaged ones unmeasured and so invisible. The reprint collects the main positions on both sides, which makes it a useful single source for the debate.

The first full caste count published by a state. It put Extremely Backward Classes at 36.01 per cent of Bihar, Other Backward Classes at 27.12 per cent, Scheduled Castes at 19.65 per cent, Scheduled Tribes at 1.68 per cent and the unreserved castes at 15.52 per cent. Read the published tables, not the political response to them: the survey's caste codes are themselves a list, drawn up before enumeration, and how a household chose among them is the question the figures cannot answer.

The first machine-readable set of the lists in one place: the 1931 caste and religion tables, the exterior castes, the Central List of OBCs with the Mandal lists and the state Scheduled Tribe lists, with identifiers linking them. Use it, and use it carefully. In V1.1 the year given for 654 of 2,428 active OBC entries comes from the gazette file number, one to five years before the resolution's printed date, and two rows of the 1931 exterior table do not add to their printed subtotals; both have been reported to the authors. The Union Cabinet decided on 30 April 2025 that the next census will enumerate caste, so data like this is about to have a successor.

The 1931 exterior castes, the Scheduled Caste lists linked back to 1931, the Central List of OBCs by the date of each resolution, and the Mandal lists against today's Central List, charted from the Jolad and Kalyani data. Each section says what its matches can and cannot show, because every link in it is a match of spellings.

Match the Mandal Commission's Bihar list to today's Central List yourself, in Python in the browser. In Bihar, 114 of the 168 Mandal entries share an exact name with the Central List; the exercise is about the rest, where a similarity score ranks candidates such as Churihara and Churihar or Kawar and Kalwar and a person has to decide, with a source, whether they are one community.