Census, delimitation a 'single spectrum test' for democracy: GN Devy

"MP chori" — that's the term literary critic and public intellectual GN Devy uses to describe how India's ruling coalition, which retained power in 2024 with a wafer-thin majority, has functioned since. Devy argues that the government is quietly working backwards: instead of the Census determining how delimitation should unfold, delimitation appears to be dictating the Census itself — right down to the construction of the new Parliament building, designed to seat more than 800 MPs. The Federal spoke to GN Devy, literary critic and former professor of English, about what this convergence of the Census, the Special Intensive Revision (SIR) of electoral rolls, and delimitation means for India's democracy.

You've voiced deep apprehension about a possible massive structural shift involving the census, the SIR and the impending delimitation — even calling it a "single spectrum test" of democracy. Why does this combination worry you so much?

In our times, we have learnt to ask the wrong questions, so I'm glad you've asked the right one. Let me first tell you what the wrong questions have been. This government was formed in 2024 with a wafer-thin majority — it was short of the halfway mark, and it formed government only with the help of allies. Our Constitution has schedules, including one specifically on horse-trading. That schedule is being violated all the time, and people in the country have started to take it as a new norm.

Also read | Delimitation Bill: Why Stalin and Akhilesh hold the key to BJP’s gameplan

I heard so many debates over the last month about whether the NDA has the numbers to secure a two-thirds majority, without anyone realising how profoundly unethical that very question is. What this government is attempting, if I may use a popular term, is "MP chori" — that is deeply unethical. But I'm glad you've asked me about the census instead.

Normally, the census is a quiet, uneventful phenomenon. People don't worry about it — it just happens, the way the monsoon comes, a little early or a little late. The census arrives every decade in the first year of that decade, completes its work quietly, and delivers results by the second, third or fourth year. For more complicated counts, like language data, results can take five or six years. Then it disappears from the horizon until the next decade begins.

Had that been the situation this decade, things would have been very different. The Registrar General of India was completely ready to carry out the census in 2021 — preparations had been made through 2019 and 2020, schedules were prepared, and the census was ready to hit the road. Then Corona happened, and we all understand why it couldn't take place in 2021.

But here is the thing: every other country that conducts a census every decade completed theirs by 2023. We are the solitary exception. That is the question that should be asked — why wasn't this census carried out in 2023, but postponed to 2026 and pushed to be completed in 2027?

If any schoolchild were asked that question, their presumption would be: how can data from 2027 be meaningfully compared with data from 2011, or with future data from 2031 or 2041? Half the use of a census vanishes when the data becomes non-linear, chronologically speaking. There has to be a strong reason for this — and the government never stated one when it postponed the census.

The real reason goes back to a constitutional amendment made 25 years ago, which said that the question of delimitation could only be reopened after 25 years. So, quite clearly, delimitation may, can, or shall take place after the census. But that 2002 amendment — which followed an earlier one in the 1970s — was itself premised on the fact that population growth rates across Indian states were uneven. Some states were increasing in population, some were lagging, some had gone ahead with population control programmes. That situation hadn't changed by 2001, so the 2002 amendment said: wait another 25 years. The reference date for the current census is 1st March 2027.

The government should have answered a simple question before postponing the 2023 census to 2026: has the population across states stabilised? The answer is clearly no — in some states, the population replacement rate is still around 1.4. And despite knowing the population hadn't stabilised, the government has already shown its intent to increase the number of seats. That's why the new Parliament building was constructed to accommodate over 800 MPs in the Lok Sabha. It appears entirely pre-planned. The census is being used to serve delimitation, rather than delimitation emerging from the census. The logic has been reversed: the census is now following delimitation, instead of delimitation following the census.

There's another worry too. While the census was already underway, the SIR of electoral rolls wasn't even necessary. And alongside the census, yet another high-level committee — on demographic changes — has been formed, which is completely redundant. The census itself tells you, state by state, how many people follow which religion and speak which language. If census data is meant to be the official, reliable dataset for any government, then a second committee set up to investigate reasons for demographic change is entirely redundant.

This committee was formed very recently, after Prime Minister Narendra Modi announced it in his Independence Day speech last year. One member of this committee is the Registrar General of India (RGI) — the same RGI who is already fully occupied running the census. That places an additional burden on him. On the other hand, this demographic committee doesn't have a single trained demographer on it. All of this, taken together, tells the country that the government is using the idea of a religious divide — Hindu versus Muslim — which this demographic committee will study, and align with the same census data, and the SIR will also be aligned to that same census data. If the census data itself were to be doctored, that is exactly the kind of apprehension already in people's minds.

Let me add one more thing briefly: certain communities, such as denotified and nomadic communities, have been struggling for years to be properly counted, and the census isn't responding to that at all. The census has made itself opaque — it doesn't even answer RTIs. It cannot be dragged to court, because the Census Act of India says the census cannot be challenged in any court. It functions almost like a quasi-judicial body once it is announced.

So this census, which would normally be an uninteresting, routine matter for ordinary people — interesting only to social scientists who eventually write papers using the data — has this time become a very big question for everybody who cares about democracy.

You've raised the issue of denotified and nomadic tribes. This is also being called the first caste census since 1931. What's your assessment of how this process is being carried out?

The objectives of the census are stated when the exercise begins, and these are meant to align with United Nations norms and principles for a census. One of the core objectives is to widen and deepen social security access for citizens — the census tries to establish whether people have houses, whether they're alive, what the migration levels are, and so on.

Also read | Why delimitation Bills appear more ominous than originally thought

On denotified tribes specifically: these communities were once notified as "criminal tribes" and restricted to certain areas because they were nomadic. They were denotified in 1952. By then, the schedules for Scheduled Castes and Scheduled Tribes had already been fairly well settled, and no new schedule was created for denotified and nomadic tribes. They were left to languish. They've suffered socially, with very little legal security — hounded out of villages and cities, with no assets, no land and no livelihood.

Several governments have actually recommended addressing this. There was a commission — the Renke Commission — whose first recommendation was to carry out a proper census count of these communities. There was also an NHRC judgment in 2003 that made the same recommendation: carry out a proper enumeration. A commission was also appointed by the NDA government specifically for denotified tribes, again recommending proper enumeration — because going by the 1931 figures and working out the projections decade after decade, the total national population of denotified and nomadic communities today is likely to be very large. That's not a small section of society.

If the census doesn't count them specifically as denotified tribes, then its very objective of widening access to social security is defeated. And despite committees appointed by both UPA and NDA governments recommending this, it still isn't being done. This isn't even a political, party-versus-party problem — every party and every body has said it should be done, but the census authorities simply aren't doing it. Instead, they're saying: we'll count everyone by caste, and then you can take that data home, sit down with a pencil and paper, add up the population of every caste, and work out the denotified tribes (DNT) total yourself.

But that's a separate issue in itself. Some DNT communities belong to the Scheduled Castes, some to the Scheduled Tribes, some belong to neither — and the nature of the injustice and suffering they face is very different from suffering that arises purely out of poverty or caste discrimination. The census isn't accounting for any of this. Even though this is technically the first census since 1931 to attempt this, it could have been done in a far better way. Anthropologists have worked extensively on this subject — they could have been involved, and a full, open discussion could have taken place.

I can say the same about languages, which is my own subject. All along, the census has used an extremely absurd idea — and this isn't specific to any one government — that a language, to be counted as a language, must have at least 10,000 speakers. This is a foolish criterion. Sanskrit, for instance, isn't spoken by more than 10,000 people in this country today. Does that mean what those few hundred people speak isn't a language? That's not scientific — a language is a language when it has a grammar. Nobody in this government would dare say Sanskrit isn't a language, and we know exactly why, for political reasons — in fact, Sanskrit is being promoted more than perhaps any language deserves, almost like it's on artificial respiration.

It's a great language, but languages come into being and disappear naturally — that process should be allowed to happen. After the Grierson survey and subsequent studies — the Grierson survey was conducted between 1898 and 1928, and it recorded 179 languages and 544 dialects — a great deal of further work was done after independence. In the People's Linguistic Survey of India, with which I was associated, we documented the presence of 780 living languages, each with grammars written down. The Census should be taking this into account.

But the last time round, the census declared only 121 languages, pushing everything else into those 121 baskets as mere subsets. We actually have tremendous linguistic and cultural diversity to be proud of — diversity other countries wish they had — and instead of showcasing it, I don't think the Census held enough dialogue on this at all. As someone who headed the People's Linguistic Survey of India, I expected to be consulted by the census on how to go about the language count. I don't think any credible linguist in the country has been consulted. The same is true for caste census.

Whether they will end up mismanaging the entire caste enumeration is a real fear of mine. There has been a People of India series, there have been other caste counts in the past, but the census hasn't shown any evidence of having followed the existing scientific literature. All of this gives me a real apprehension that this census is being carried out only to support delimitation and to feed the demographic changes committee — not to genuinely study and enumerate Indian society as a whole. That's my real worry about this census.

When enumerators go door to door and ask people their caste, how do you think this will actually play out — will people simply state their caste and have it written down, or will there be a specific, structured list to choose from?

In normal times, that would have been a good question to discuss. In these times, it fills me with a great deal of fear, because there are United Nations specifications on the technology to be used during a census — specifications agreed upon by all UN member states, India included, precisely to maintain parity between one country's census and another's. One of those specifications relates to the technology itself. In the past, UN agencies invited by the Indian government have worked closely with the census authorities — the Registrar General's Office — discussing and even helping determine what kind of technology should be used.

Also read | Three grounds on which delimitation laws may be challenged in court

This time, as far as I know, the United Nations has been entirely kept out of the census discussions. That's very unusual. And this technology question matters enormously for the caste census — do all people in India even have access to the same technology the census is going to use for enumeration? Will 35 lakh enumerators receive exactly identical training? The questions asked in the census aren't very many, but the training is fundamentally the most important thing — and that training manual is normally prepared jointly by the census office and the UN office.

This time, the census office has been giving instructions to enumerators over WhatsApp. That is genuinely surprising — it's exactly like what's happening with the SIR and the Election Commission. The RGI is now behaving exactly the way the ECI has been treating the SIR, with instructions going out over WhatsApp. Will every village, every basti, every tanda, every camp, every labour camp, every nomadic group actually be able to register and fill in their precise caste correctly? The answer is no.

There's also a linguistic difficulty — the spelling of caste names in English doesn't always correspond to how they're actually pronounced. It's a genuinely confusing thing in India. Bangla speakers will pronounce a word one way, Gujarati speakers another way entirely. Take our Home Minister's name — we say "Amit Shah," but in Gujarat, it's pronounced differently, without that same "sh" sound. All of this is going to affect the process of collecting and computing the list of caste names and sub-categories. And what's most surprising — almost like something out of Alice in Wonderland, whether you want to call it a dystopia or a utopia — is the timeline.

The census office has said that although the reference date is March 1, 2027, the final data will be out by December 2027 — that's eight or nine months after the census is completed. When you ask how that's possible, the answer given is that because it's all being done digitally, it will take less time. But while digital recording may indeed be faster, sorting and organising the vast number of caste names and sub-names is not something a digital machine can do on its own. It requires a well-trained human mind, exercising real discretion, to determine the names of communities and languages. That takes time.

And if a caste census takes place now, is India likely to carry out another one for another 50 or 100 years? My answer is no. This will almost certainly become the definitive caste census for India for decades, even centuries, to come — and it is not going to be a very perfect one. On the basis of this data, we will be making judgments about who we are, how we're constituted, how we're composed, and about the different orders of justice and injustice in our society. Our national narratives are going to change profoundly once this caste census comes out. To do it this hurriedly means we risk losing our own self-understanding as Indians — that is going to remain a big question mark for a long time to come. We will have missed a great opportunity to genuinely enhance our understanding of ourselves, an opportunity that could have been realised had this been done in a calibrated, careful way — rather than being rushed for the sake of delimitation, or for winning elections.

Let's turn to delimitation. In 2001, Parliament deferred it for 25 years, tying it to the first census after that period. You've pointed out the government missed the chance to conduct that census in 2023, and instead, the new Parliament building's foundation stone was laid in December 2020, months after the Ram temple's foundation ceremony — suggesting this expansion was planned well in advance. Now, the government appears to be proposing a uniform 50% increase in seats across every state, which on paper means southern states aren't penalised in absolute terms. Given that, why do you still consider this approach inappropriate?

The question of delimitation has to be seen in its proper historical context to be able to answer that. In every democracy, there's meant to be a built-in process of increasing the number of representatives in step with a growing population, so that the principle of "one person, one vote, one value" is preserved. But — and I have to say, hats off to the makers of our Constitution, who thought very deeply about this — they built in a caveat. The law says "as far as practicable." That means this rule cannot be applied mechanically. The Delimitation Act of 1952 itself says: please do not apply this in a mechanical way.

There are two ways of doing delimitation. One is to determine the geographical area of a constituency — cartographic delimitation. The other is arithmetical delimitation, based on numbers. The real question before the country today has two parts. First: are there enough representatives for people to be adequately represented, given how enormously our population has grown? We started as a country with 37 or 38 crore people; we're now at over 100 crore — three, four times that. But there's a second question staring us in the face, and that's the question of federalism.

Delimitation is deeply connected to federalism because we have two houses of Parliament. If delimitation is done purely numerically, it directly affects the Lok Sabha and state assemblies. But if it were done on a matter of principle, it should also affect the Rajya Sabha — which is, after all, the House of States. Parliament makes laws for the entire nation, which means all states are affected by them. So do all states have equal representation in Parliament? That question can only be answered by looking at how many MPs each state has in the Rajya Sabha. This government should have paid far greater attention to restructuring the Rajya Sabha and making it more federal-friendly, rather than simply looking at which states have higher populations and can therefore be given more seats.

Now, coming to this flat 50% increase across every state: what is the fundamental, basic principle behind it? Is it appeasement of every state? Is it simply to manage public mood? Is there any legislative or constitutional principle actually being applied here? How can anyone justify a 50% increase on the basis that it delivers greater justice to the people of any particular state than before? The answer is no. So what's the actual need for such an increase?

Take Karnataka as an example — it currently has 28 MPs. If that number increases to 42 through a flat 50% hike, will those 42 people represent Karnataka any better than the 28 do today, especially inside an assembly of 850 or so? The answer, again, is no. This entire idea of a gross, across-the-board increase is like handing out a Diwali bonus to employees — "here, everyone gets 10%." That is not what delimitation is meant to be. And if that's all it's going to be, why do you even need a delimitation commission? Are you going to need a bunch of cartography graduates to simply draw new constituencies?

Also read | How Modi govt hopes to railroad Constitution 130th Amendment Bill amid hurdles

I'll admit, when it was suggested to me during our conversation that the commission might end up merely signing off on maps already drawn up by some back-room team, I found it impossible to disagree.

A government has many roles to play — it's a political organisation when it comes to elections, but it also has to think about the entire nation and about people's representation as a whole. Delimitation is meant to be an act carried out on behalf of the people; the power to decide how many MPs there should be must rest with the people. If a government approaches this only myopically, purely to ensure it returns to power in the next election, that renders very valuable constitutional provisions dysfunctional. This is something neither the previous NDA government nor any earlier Congress government ever dared to do. There's an English saying, isn't there, about fools rushing in where angels fear to tread?

So this isn't really a question of mathematics — of working out 50% of 40, or 50% of 30. It's a constitutional question, and constitutionally, it's a flawed argument — the idea that everyone simply gets the same 50% increase. It's definitely a flawed argument.

We've also seen delimitation play out recently in Assam, where analysts pointed to constituencies that once had fair minority representation shrinking sharply — from around 35-36 seats in a 126-member house to about 20-21. There's a fear this could be replicated nationally, and it comes as the government is also pushing ahead with One Nation One Election, with a committee report already in and legislation now being considered. Looking at all of this together, do you think elections will still have real meaning going forward — or are we heading toward a point where the outcome is already known in advance?

This whole slogan of "One Nation One Election" is a very strange, very awkward formulation. Let me tell you something interesting — there was a Vadnagar-based right-wing political thinker named Vasant Parikh who, back in the 1960s, coined the slogan "one note, one vote." That's the same town Narendra Modi comes from, and Parikh was, in fact, one of Modi's mentors. That idea — "one note, one vote" — has stayed somewhere in Mr Modi's consciousness ever since, and I think "One Nation One Election" carries a similar spirit.

"One nation" itself, I have no quarrel with — I'm proud to belong to this nation, there's no separation between my own consciousness and the nation. But doesn't our Constitution describe India as a Union of States? Those states have people who are empowered to elect their own governments, and those governments are expected to stay in place and serve the people for five years. If, from tomorrow, we simply declare that we are "one nation" and therefore every Assembly election must happen simultaneously with the Lok Sabha election — what happens if the central government collapses and a re-election is called? What happens to the state assemblies then?

The Constitution has already made provisions for all these kinds of emergencies. So there's no real reason why we should deny people their constitutional right to elect governments as and when the Constitution allows, simply because our elections aren't all happening on the very same day. There was also a lecture recently at the Constitution Club, by Justice AP Shah and Justice Madan Lokur, which showed that the government's central claim — that simultaneous elections will save on election expenditure and reduce the burden on the exchequer — isn't actually supported by the data.

If the government genuinely wants to save money, it should turn its attention to things like electoral bonds, the PM CARES Fund, and the sheer amount of money being spent on buying MPs — that's what should be reduced. Instead, we're essentially punishing people by telling them: don't hold an election now, hold it two years from now, hold it whenever we tell you to, hold it two years earlier than scheduled. The basic purpose of a democracy is that there should be elections, and we should be prepared to have them as frequently as they're required — including when a government has failed to do its job and has been reduced to a minority through a legitimate legislative process.

I'm actually of the view — and I know many people have been asking for this — that there should be a right to recall representatives. There's also a method called liquid democracy, adopted in parts of the United States, like Colorado, and in some north European countries, which allows citizens, after electing a representative, to use certain "vote tokens" every month, or every three months, or every six months, to continue expressing their preference. Let there be a forever nation, forever election — not one nation, one election. It should be a forever nation.

There's also another side to this "one nation" slogan that isn't clearly visible right now — and that's Savarkar's own definition of what constitutes a nation. In that definition, only Hindus constitute the nation; everyone else does not. It's a very unitary idea of nationhood, and that is the kind of state the BJP appears to be moving toward. What we actually need is a diverse nation, with diverse, forever elections.