Why Caste Cannot Be Reformed

India spends a great deal of effort on the reform of caste. Sensitivity workshops in corporate offices, inter-caste dinners convened by NGOs, diversity initiatives in elite institutions, anti-discrimination modules in universities — the activity is constant, well-funded, and the people running it are not insincere. It also does not work. Ambedkar explained why almost ninety years ago, in a speech that was refused before he could deliver it. The reasons he gave then have not been answered since.

Annihilation of Caste was prepared for the Jat-Pat-Todak Mandal in Lahore in 1936. The Mandal — a Hindu reformist body — invited him to address its annual conference, read the text, then withdrew the invitation when he would not soften it. He published it himself. The argument it makes is that reform of caste is not merely difficult; it is the wrong category of project. Caste is sustained by religious belief and structured to prevent the unity that resistance would require. Behavioural reform addresses neither.

The first move in the argument cuts directly at the reformist strategy of inter-caste contact.

Caste is not a physical object like a wall of bricks or a line of barbed wire which prevents the Hindus from co-mingling and which has, therefore, to be pulled down. Caste is a notion, it is a state of the mind. The destruction of Caste does not therefore mean the destruction of a physical barrier. It means a notional change. Caste may be bad. Caste may lead to conduct so gross as to be called man's inhumanity to man. All the same, it must be recognized that the Hindus observe Caste not because they are inhuman or wrong headed. They observe Caste because they are deeply religious. People are not wrong in observing Caste. In my view, what is wrong is their religion, which has inculcated this notion of Caste. If this is correct, then obviously the enemy, you must grapple with, is not the people who observe Caste, but the Shastras which teach them this religion of Caste. Criticising and ridiculing people for not inter-dining or inter-marrying or occasionally holding inter-caste dinners and celebrating inter-caste marriages, is a futile method of achieving the desired end. The real remedy is to destroy the belief in the sanctity of the Shastras.

— Annihilation of Caste, Section XX, pages 83–84.

Read this carefully and the contemporary reform agenda dissolves. Behaviour is not the cause. Behaviour is downstream of belief, and the belief carries the sanction of scripture. To target the behaviour while leaving the belief intact is to weed without uprooting.

Ambedkar then makes the categorical case. Reforming caste, he argues, is not like other social reforms because it requires people to reject religious principles they take as divinely sanctioned.

What are your chances of success? Social reforms fall into different species. There is a species of reform, which does not relate to the religious notion of people but is purely secular in character. There is also a species of reform, which relates to the religious notions of people. Of such a species of reform, there are two varieties. In one, the reform accords with the principles of the religion and merely invites people, who have departed from it, to revert to them and to follow them. The second is a reform which not only touches the religious principles but is diametrically opposed to those principles and invites people to depart from and to discard their authority and to act contrary to those principles. Caste is the natural outcome of certain religious beliefs which have the sanction of the Shastras, which are believed to contain the command of divinely inspired sages who were endowed with a supernatural wisdom and whose commands, therefore, cannot be disobeyed without committing sin. The destruction of Caste is a reform which falls under the third category. To ask people to give up Caste is to ask them to go contrary to their fundamental religious notions. It is obvious that the first and second species of reform are easy. But the third is a stupendous task, well-nigh impossible.

— Annihilation of Caste, Section XXI, pages 84–85.

The taxonomy is doing real work. Most reforms in democratic life fall into the first two categories — they extend rights, refine practice, or recover a principle that has slipped. The reform of caste belongs to the third. It is a refusal of the religious authority the practice rests on. Any movement that does not understand it is committing this refusal — and instead presents itself as a tidying-up of behaviour — will not survive its first contact with the authority it has refused to confront.

The third argument explains why the system is also structurally insulated from organised resistance. Caste, Ambedkar observes, is not only a division but a graded ranking. Each caste defends the rank it holds above some caste below.

The second reason, why I say the task is impossible, will be clear if you will bear in mind that the Caste system has two aspects. In one of its aspects, it divides men into separate communities. In its second aspect, it places these communities in a graded order one above the other in social status. Each caste takes its pride and its consolation in the fact that in the scale of castes it is above some other caste. As an outward mark of this gradation, there is also a gradation of social and religious rights technically spoken of an Ashta-dhikaras and Sanskaras. The higher the grade of a caste, the greater the number of these rights and the lower the grade, the lesser their number. Now this gradation, this scaling of castes, makes it impossible to organise a common front against the Caste System. If a caste claims the right to inter-dine and inter-marry with another caste placed above it, it is frozen, instantly it is told by mischief-mongers, and there are many Brahmins amongst such mischief-mongers, that it will have to concede inter-dining and inter-marriage with castes below it! All are slaves of the Caste System. But all the slaves are not equal in status. To excite the proletariat to bring about an economic revolution, Karl Marx told them: 'You have nothing to lose except your chains.' But the artful way in which the social and religious rights are distributed among the different castes whereby some have more and some have less, makes the slogan of Karl Marx quite useless

— Annihilation of Caste, Section XXI (continued), pages 86–87. The retrieved excerpt ends mid-sentence; the argument continues in the original text beyond what is reproduced here.

This is the structural genius of the arrangement. Every caste is held by what it stands to lose downwards, not by what it might gain upwards. A reform programme that asks castes to merge into a single front is asking them to give up the small rank they hold — and offering nothing the system has not already designed to outweigh. Class solidarity, the standard organising frame of the twentieth century, cannot do the work. The slaves are not equal, and they know it.

A reader who has watched a workplace inclusion programme work might object that small behavioural changes do accumulate, that inter-caste contact really does soften prejudice, that the structural argument is too absolute. Ambedkar would not deny that behaviour can shift. He would say that without belief change, and without dismantling the graded structure, the shifts are unstable, easily reversed, and politically marginal. His point is not that small efforts are worthless. It is that they will keep being defeated, in entirely predictable ways, until the underlying authority is confronted.

The reason the argument still hits, ninety years on, is that the politics it indicts are still the politics being practised — by states, by NGOs, by corporations adopting their inclusion programs. Ambedkar's verdict that the project is well-nigh impossible without confronting religious authority is not pessimism. It is the honest map of the terrain. Pretending the terrain is flatter than it is has not produced the change reformers hoped for. Mapping it accurately, as he did, is the precondition for any politics that might.