Caste, Democracy and the Reservation Question

In 1950 India did something no country had attempted. It gave universal adult suffrage, immediately and without qualification, to a population that was overwhelmingly poor, largely illiterate, and organised by a hereditary hierarchy two thousand years old.

Almost everyone expected it to fail. What happened instead is one of the largest natural experiments in political sociology, and its findings surprised every position — including Ambedkar's.

A prediction, and what actually occurred.

The confident expectation in 1950, held by modernisation theorists, by many Indian intellectuals, and by a good number of Western observers, ran as follows.

Caste is a traditional institution. Democracy, industrialisation, urbanisation and education are modern forces. As the modern advances, the traditional recedes. Caste would weaken, then dissolve into class, and Indian politics would come to resemble European politics: parties of left and right, contesting on economic programmes, with voters aligned by occupation and income.

This is modernisation theory (5.1.3) applied to the largest possible case, and it was the mainstream position.

What happened instead: caste became the principal idiom of Indian democratic politics, and it did so not by resisting democracy but by using it.

Caste associations became vote banks. Parties formed around caste coalitions. Candidates were selected by caste arithmetic. Manifestos addressed caste demands. And the categories the colonial census had produced (6.1.1) became the units of political competition.

And the direction of change was not what conservatives feared either. The mobilisation ran upward from below: numerically large lower and middle castes, previously excluded from power, discovered that in a system of one person one vote, numbers are a resource — and that they had them.

Rajni Kothari stated the finding in 1970, and it remains the standard formulation:

"It is not politics that gets caste-ridden; it is caste that gets politicised."

Caste did not capture democracy. Democracy transformed caste — and the transformation ran in both directions.

The theoretical problem is a general one and India is its clearest case.

What happens when a hereditary status hierarchy is placed inside a system that counts heads?

The two available predictions:

Dissolution — modernisation theory's answer. Universal citizenship makes ascribed status irrelevant; individuals vote their interests; hereditary categories lose their function and fade.

Capture — the conservative and elite answer. The hierarchy simply reproduces itself in the new institutions; the dominant control the votes of their dependents; nothing changes but the vocabulary.

Neither happened. And the reason is a fact about the arithmetic that nobody weighted properly in advance.

In a ritual hierarchy, the top is small. Brahmins are around 4–5% of India's population; the "twice-born" varnas together perhaps 15–20%.

In an electoral system, the top being small is a fatal disadvantage.

So the categories persisted and their political value inverted. What had been a mark of subordination became, for numerically large groups, a mobilising resource. The same category that had excluded you from the temple could now deliver a constituency.

Which is a genuinely important general finding: an institution's function can reverse without its form changing, when the rules of the surrounding system change.

Reservations: the design

India's affirmative action system is the oldest, largest and most studied in the world, and its architecture matters.

Three protected categories, with different histories and different politics.

Scheduled Castes (SC) — those formerly designated untouchable. Around 16–17% of the population. Reserved seats in the Lok Sabha and state legislatures, plus quotas in public employment and public higher education.

Scheduled Tribes (ST) — Adivasi communities. Around 8–9%. The same provisions. Note that this is a distinct category with a distinct history: Adivasi exclusion is territorial and colonial as much as ritual, and treating it as a subtype of caste is an error (6.2.5).

Other Backward Classes (OBC) — castes that are not untouchable but are socially and educationally disadvantaged. The most contested category, both in definition and in extent, and the subject of the largest political conflict.

The legislative reservations were originally set for ten years and have been extended by constitutional amendment every ten years since. Employment and education quotas were not time-limited.

The Mandal moment, which is the hinge of modern Indian politics.

The Mandal Commission reported in 1980, recommending 27% reservation in central government jobs and public higher education for OBCs, on top of the existing 22.5% for SC and ST.

It was not implemented until 1990, when the V. P. Singh government announced it.

The reaction was extraordinary. Upper-caste students organised mass protests across northern India. There were self-immolations. The government fell within months.

And the consequences reorganised Indian politics permanently:

The Supreme Court, in Indra Sawhney (1992), upheld OBC reservation, introduced the "creamy layer" exclusion — that the better-off within OBC categories should not benefit — and set a 50% ceiling on total reservations, which has structured every subsequent dispute.

Lower-caste parties consolidated. The Bahujan Samaj Party, the Samajwadi Party, the RJD and the Dravidian parties in Tamil Nadu became major forces.

Christophe Jaffrelot called the resulting shift a "silent revolution": the transfer of political power in the northern states from upper-caste elites to intermediate and lower-caste politicians, over roughly two decades, through electoral means.

And in 2019 a further 10% was reserved for "economically weaker sections" among those not otherwise covered — which breached the 50% ceiling and was upheld by the Supreme Court in 2022, and which is a substantially different principle: economic rather than social disadvantage as the criterion.

Reservations have been in operation for over seventy years and have been studied extensively. The findings are more specific and more mixed than either side of the political argument allows.

Where the evidence is strong and positive:

Political reservation changes what governments do. The best-identified study is Rohini Pande's work using the variation in reserved seat shares across Indian states, which found that SC/ST political reservation increased redistributive spending targeted at those groups. Chattopadhyay and Duflo's study of reserved village council leadership positions — assigned to villages by a rule that approximates random assignment — found that leaders invest more in the public goods their own group prioritises. These are unusually clean identification designs for a question of this kind.

Reservation reduces prejudice about competence. Follow-up work on the same village council reservations found that exposure to a reserved woman leader improved subsequent electoral prospects for women and reduced stated bias — evidence that contact with counter-stereotypical office-holders changes attitudes.

And employment reservation has produced a Dalit middle class in public sector employment, which did not exist before and which is the system's clearest achievement.

Where the evidence is mixed or negative:

Benefits concentrate. Within reserved categories, the better-off capture a disproportionate share — the "creamy layer" problem the Supreme Court addressed and which persists.

Coverage is limited. Reservations apply to the public sector, which is a small and shrinking share of Indian employment. Liberalisation has therefore reduced the reach of the main instrument, which is why demands for private sector reservation recur.

They do not touch the private sector, where audit studies find substantial discrimination (6.2.1).

And the "efficiency cost" question — whether reservations reduce the quality of appointments — has been studied and the findings are not what either side expects. The best work, on Indian Railways, found no evidence of efficiency loss from SC/ST employment reservation, and some evidence of gains in higher-level positions.

And the political consequence nobody designed:

Reservations made caste categories permanently politically salient. A category that carries entitlements will be defended, claimed and contested. Groups campaign to be reclassified as OBC — the Jat, Patidar and Maratha agitations of the 2010s and 2020s each involved large mobilisations by relatively prosperous agricultural castes demanding backward-class status.

Which is exactly the census dynamic of 6.1.1, operating again with much higher stakes: an administrative category becomes a resource, and people organise around it.

Go deeper

The argument about reservations, stated fairly

This is a live political dispute and the sociologically serious versions of each position are worth having.

The case for:

Historical injustice requires positive remedy. Formal equality applied to a population shaped by two millennia of exclusion reproduces the exclusion. Ambedkar's argument (6.2.2): equal treatment of unequals is not equality.

Representation has independent value. A legislature, a bureaucracy or a university without members of a fifth of the population is not merely unfair; it produces worse decisions, because the excluded perspective is absent — which is standpoint theory's claim (5.6.1) applied institutionally.

The evidence supports the political effects. Pande, Chattopadhyay and Duflo, and the attitude-change findings.

And the alternative is not neutrality. Without reservation, appointment runs through networks, credentials and cultural capital (5.8.1), all of which track caste. The choice is between an explicit rule and an implicit one.

The case against:

It entrenches the categories it aims to dissolve. If caste identity carries entitlements, caste identity will persist and be politically reinforced — which the Jat and Patidar agitations demonstrate.

It benefits the already advantaged within categories, which the creamy layer provision acknowledges and does not solve.

It creates resentment among those excluded, particularly poorer members of unreserved groups, which is politically consequential and has been mobilised.

And it addresses the wrong variable. If the underlying problem is economic, an economic criterion would target it more precisely — the argument behind the 2019 EWS provision.

The counter to the last point, and it is the strongest sociological reply: caste disadvantage is not reducible to income. Audit studies find discrimination against equally qualified Dalit applicants; endogamy operates regardless of wealth; and atrocity data does not track poverty alone. An economic criterion would miss what caste does.

The comparative frame

India's system is the world's oldest and largest, and it is not unique.

Malaysia's bumiputera policy, from 1971, reserves quotas in education, employment and business ownership for the Malay majority — affirmative action for a numerical majority, which is a structurally different case and has produced a substantial literature on whether the arrangement is temporary or permanent.

Brazil introduced racial quotas in federal universities from the 2000s, with a system of self-declaration and, controversially, verification panels. The comparison with India is instructive: Brazil's category is phenotypic and contested at the boundary, India's is documented and hereditary, and the verification problems are correspondingly different.

South Africa's post-apartheid Black Economic Empowerment and employment equity legislation targets a majority excluded by law within living memory.

The United States developed affirmative action without quotas — the Supreme Court having ruled explicit quotas unconstitutional in Bakke (1978) — and substantially dismantled race-conscious university admissions in Students for Fair Admissions (2023).

And the comparison yields a general finding: every system faces the same four problems — defining the beneficiary category, preventing capture by the advantaged within it, deciding whether the measure is temporary, and managing the resentment of those excluded. No system has solved all four, and India's is the longest-running attempt.

Caste in the twenty-first century

Three things are simultaneously true, and any account that asserts only one is wrong.

Caste has weakened enormously in some domains. Urban anonymity, education, mobility and legal protection have transformed daily life for millions. The village hierarchy Srinivas described is not the structure of urban India. Untouchability practices in public spaces have declined substantially, though survey evidence finds them persisting at significant rates.

Caste has strengthened in others. Political mobilisation has made caste identity more salient, better organised and more consequential than at any point since independence. Democracy did this, and it was not the prediction.

And caste has been reconstituted in modern institutional form. Networks, credentials, cultural capital, language, and the informal criteria of professional and corporate hiring carry caste into institutions that have no explicit rule about it. The mechanism is Bourdieu's (5.8.1) and the content is caste.

Endogamy remains the constant. It is the least changed feature, the best predictor of the system's persistence, and — exactly as Ambedkar argued in 1916 (6.2.2) — the mechanism that keeps everything else in place.

"Democracy is dissolving caste." Democracy politicised caste and made it more salient. Kothari's formulation is the finding: caste got politicised, not politics caste-ridden.

"Reservations were a temporary measure." The legislative reservations were set for ten years and have been renewed continuously; employment and education quotas were never time-limited.

"Reservations reduce quality." The best-identified evidence — the Indian Railways studies — finds no efficiency loss and some gains. The claim is asserted far more often than it is tested.

"Reservations only benefit the creamy layer." Capture by the better-off within categories is real and documented; the system has also produced a public sector Dalit middle class that did not previously exist. Both.

"An economic criterion would be fairer." Audit studies find discrimination against equally qualified applicants, and endogamy and atrocity data do not track income. Caste disadvantage is not reducible to poverty.

"Caste is disappearing." It has weakened in some domains, strengthened in others, and been reconstituted in modern institutional forms — with endogamy as the constant.

Why it matters

It is the largest natural experiment in political sociology. Universal suffrage granted immediately to a hierarchical, poor, largely illiterate population of hundreds of millions. Both confident predictions failed, and what happened instead — that the hierarchy's categories became the units of democratic competition and inverted in political value — is a finding no theory anticipated.

It supplies the general lesson about administrative categories: a category that carries entitlements becomes a resource, is claimed and defended, and thereby persists. The census produced the categories (6.1.1); reservations gave them value; and now they cannot be removed. That mechanism operates wherever states classify.

It is the world's longest-running affirmative action programme, with the best-identified evidence, and anyone arguing about such policies anywhere should know what seventy years of it has established.

And it demonstrates that a hereditary hierarchy is not simply the opposite of a modern institution. Caste did not resist democracy and was not dissolved by it. The two reorganised each other — which is a better model of how institutions actually meet than the traditional/modern opposition that Part 6 keeps having to dismantle.

The 1950 prediction — that democracy, urbanisation and education would dissolve caste into class — was modernisation theory applied to the largest case, and it failed. Kothari's finding: "It is not politics that gets caste-ridden; it is caste that gets politicised."

The arithmetic explains it. In a ritual hierarchy the top is small — Brahmins around 4–5%. In an electoral system, being small at the top is fatal, so the categories inverted in political value: what had marked subordination became a mobilising resource.

The reservation architecture: SC (~16–17%), ST (~8–9%) and the contested OBC category. Legislative reservations renewed every decade; employment and education quotas never time-limited.

The Mandal moment (1990): 27% OBC reservation announced, mass upper-caste protest, the government falling, Indra Sawhney (1992) upholding it with the creamy layer exclusion and a 50% ceiling, and Jaffrelot's "silent revolution" transferring power to intermediate and lower castes in the northern states. Then the 2019 EWS provision, breaching the ceiling on a different principle.

The evidence: Pande on redistributive spending; Chattopadhyay and Duflo on reserved village leadership changing public goods provision, with near-random assignment; attitude change from exposure to counter-stereotypical leaders; and no efficiency loss found in the Indian Railways studies. Against: capture by the better-off, limited public sector reach in a liberalising economy, and no purchase on private sector discrimination.

And the unintended consequence: a category carrying entitlements becomes worth claiming — the Jat, Patidar and Maratha agitations are the census dynamic of 6.1.1 operating with higher stakes.

Three things simultaneously true: caste has weakened in daily life, strengthened politically, and been reconstituted through networks, credentials and cultural capital — with endogamy as the constant, exactly as Ambedkar's 1916 mechanism predicts.

Scheduled Castes / Scheduled Tribes / Other Backward Classes — India's three constitutional categories for positive discrimination.

Reservation — quotas in legislatures, public employment and public education.

Mandal Commission — the 1980 report recommending OBC reservation, implemented in 1990.

Creamy layer — the better-off within a reserved category, excluded from benefit by judicial doctrine.

The 50% ceiling — the Indra Sawhney limit on total reservation, breached by the 2019 EWS provision.

EWS — Economically Weaker Sections: the 2019 reservation on an economic rather than social criterion.

Silent revolution — Jaffrelot's term for the electoral transfer of power to lower and intermediate castes.

Vote bank — a bloc of voters mobilised by group identity and delivered by intermediaries.

Bumiputera — Malaysia's affirmative action policy for the Malay majority.

One — test the arithmetic argument. Take any hierarchical society and ask what happens to its ranking when heads are counted rather than status weighed. The inversion India experienced is a general possibility.

Two — find a category that became a resource. In your own country, identify an official classification that carries entitlements. Ask whether groups campaign to be included, and what that does to the category's persistence.

Three — read a Duflo study. The village council reservation papers are short, clearly written and use an unusually clean identification strategy. They are a good model of what causal evidence in this area looks like.

Four — compare two systems. Take India and one of Brazil, Malaysia, South Africa or the United States. Check each against the four problems: defining the category, preventing capture, temporariness, and managing resentment.

Five — apply the three-truths test. For any social division in your own society, ask whether it has weakened in daily life, strengthened politically, and been reconstituted institutionally. Frequently all three are true at once, and arguments consist of people asserting different ones.