Dholera population: why the index scores it zero of five

I built the scoreboard this site runs on, and it hands Dholera a zero. Not a cautious two, not a generous three. A zero, on one of eight dimensions, in a table that carries my name. The comparative study behind this site scores eighteen built-from-scratch cities across eight viability dimensions, and Dholera finishes eighth overall at 3.10 out of 5, which is a respectable placing in a sample that contains Shenzhen. On population traction it scores nothing at all.

That single cell has cost me more arguments than the rest of the board combined, and most of them were with people who like Dholera. One or two of those arguments were good ones. So here is the zero, argued in full: what it measures, why it is not the insult it looks like, why it is also not nothing, and the specific dated things that would change it. You can sort every row yourself on the interactive board, and the paper with the whole rubric inside it is free to download.

What the zero actually measures

Population traction is one of eight dimensions and carries a 15 percent weight, joint second heaviest with anchor delivery and behind only demand anchor realism at 20 percent. Its definition is deliberately narrow: actual residents measured against the promised target. Above 33 percent of target scores a 4, above 60 percent scores a 5, and below 3 percent scores a 0 or a 1. Nothing in that definition gives credit for money spent, land allotted, approvals granted or intentions announced. Other dimensions cover those. This one asks the single question that no amount of construction answers, which is whether anybody has moved in.

Dholera fails that question on the most basic ground available. I have no credible count of new residents to put in the numerator. Not a small count, not a disappointing count: no number I would be willing to publish. The 2011 census recorded 2,779 people in Dholera village, 1,420 male and 1,359 female, across 576 households, and that village was sitting there long before anybody drew a special investment region around it. The taluka holds 22 villages whose residents were already in place, and the rubric excludes them, because the entire question a greenfield city is asking is whether it can attract people who were not already there. If pre-existing villagers counted, every new city on earth would open with a free score and the dimension would measure nothing.

I want to be honest about where that rule bites unevenly, because it does. Amaravati's roughly 100,000 count is recorded in my own dataset as pre-existing villagers, and Nusantara's 147,430 as mostly pre-existing villages, and both of those cases score 1 rather than 0. That is a judgement call. A single-rater study is exactly the sort of exercise where judgement calls should be visible and contestable, so I will say plainly that I scored all eighteen cities myself, that this is the study's main acknowledged limitation, and that every score and justification is published precisely so a reader can re-score the board and see what moves.

What I will not do is manufacture a Dholera number to fill the gap. I do not have a sourced, current, combined population for those 22 villages, so I am not going to build an estimate out of one 2011 village figure and present it as evidence. That refusal is the same discipline I apply to plot prices, and it costs the same thing here: a tidier looking table, traded for a number I could not defend if somebody pushed.

The company that zero keeps

A zero means very little on its own. It means a great deal once you can see the distribution it sits in. Here is the population column of the index, ordered by attainment, with each city's target, its latest sourced count, and the score that follows from the two.

Three cases from the eighteen are missing from that table for honest reasons. Shenzhen never set a formal population target and now holds 17,560,000 people by the 2020 census, so an attainment ratio for it would be meaningless. Eko Atlantic has published no credible resident count against its 250,000 to 300,000 target. Yachay has no verified target at all, only about 2,000 students on 4,462 expropriated hectares and a delivery company liquidated in 2021.

Read the table from the top and the shape of the thing appears. The two cases that genuinely filled up moved something real into place first. Sejong relocated 44 agencies and about 14,000 civil servants by statute. Songdo sits inside commuting reach of Seoul. Putrajaya completed a federal government relocation between 1999 and 2005 and still houses only around 120,000 people, which is a little over a third of its target three decades after it was announced in 1995. Below that line the numbers fall away fast, and the fall is not gradual: there is almost nothing between 11.5 percent and 2.9 percent, and then a long floor of cities sitting at or near zero.

One more thing in that table is worth flagging before somebody else catches it. Sejong scores 4 even though its 78.9 percent sits well above the 60 percent line that the rubric says earns a 5. That is judgement rather than arithmetic, applied because Sejong's growth has slowed and the capital debate reopened in 2025. You are welcome to disagree with me, and the published rubric exists so you can. What matters for the argument here is that the contestable calls all live at the top of the table. Nobody has to squint at a band boundary to work out where a city with no published resident count belongs.

Not all zeros are the same zero

Nine of the eighteen cases score 0 on population traction, and treating them as equivalent would be lazy reading. The difference between them is not the resident number. It is what the rest of each row says about why that number is missing.

Lavasa scores 0.95 overall, second weakest in the whole sample, with under 2,000 residents against a target of 200,000 to 250,000, which works out at 0.9 percent. The zero is the least interesting fact in that row. The tourism anchor failed. The project reached insolvency with Rs 6,642 crore of admitted claims and a rescue bid from Welspun still in litigation, a stop-work order arrived from the environment ministry in 2010, and roughly 600 hectares of land purchases were contested. Lavasa's zero is a verdict handed down after the trial finished.

KAEC in Saudi Arabia is the subtler case, and for a Dholera reader it is the more instructive one. Its port was delivered to world-class standard and its infrastructure exists. Its target was 2 million residents by 2020, it has roughly 10,000 people, and the target year passed six years ago. That is a zero with the clock already run out on it, and it is precisely the failure mode Dholera has to avoid: infrastructure that works, an anchor that half delivers, and a city that never becomes a place anyone lives.

New Clark City posts a flat 0.0 percent with about 229 students and no permanent movers, alongside sports facilities of international standard and a government centre built for 3,000 people that sits unfilled. NEOM's Line has no conventional residents at all, more than USD 50 billion reported spent against a USD 500 billion envelope, and a programme halted past 2030 after its target was cut in stages from 9 million toward roughly 100,000.

Dholera's zero has a different shape from all of those. Its anchor is neither delivered nor dead: the Tata and PSMC fab is past 50 percent civil work with first silicon targeted for December 2026 and commercial production reported for mid 2028, and 300 MW of solar is already generating. Its financing is statutory and sovereign rather than dependent on selling homes to speculators, which is why it scores 4 on financing durability where Lavasa scores 0. And its target year, unlike KAEC's, has not arrived.

The 2042 problem, and the 2020 one

Dholera is the single case in the index excluded from the comparative ratio medians, and the reason is stated plainly in the paper: its target year, 2042, has not arrived. You cannot compute an attainment percentage against a deadline that is sixteen years out and then compare the result with cities whose deadlines expired. That is why the Dholera row reads pending rather than a number, and it is the strongest thing anybody can say in the zero's defence. It is a genuine defence, and I hold to it.

It is not a complete defence, and here is the part that the brochures leave out. Dholera has already had one population deadline and missed it. The old target of roughly 120,000 residents and about 80,000 jobs by 2020 lapsed unmet. That was a test with a date attached, the date came, and the city was not there. Anyone using "too early to judge" as a blanket answer has to account for a deadline that has already run out. The honest version of the argument is narrower and I think it survives: it is too early to judge the 2042 target, and the 2020 miss is evidence about delivery pace rather than about the final outcome.

The target itself deserves a note, since I am scoring against it. The roughly 1 million figure is promotional, not statutory, and the end year appears as 2040 in some documents and 2042 in others, a discrepancy I flag rather than pretend to resolve. The official quantified promise on the state's own portal is employment, over 800,000 jobs at maturity, broken down as about 312,900 direct and about 483,630 indirect. Scoring residents against a promotional denominator of a million is, if anything, the harsher of the two available choices. Scoring jobs instead would need a credible job count, and no such count exists yet either, because the fab has not produced a chip.

What the zero costs on the board

Here is the arithmetic of the penalty, since a weighted index makes it easy to hide. Dholera's eight scores are 5 for demand anchor realism, 4 each for connectivity integration, financing durability and governance continuity, 3 each for land assembly durability and proximate metro gravity, 2 for anchor delivery, and 0 for population traction. Those produce a weighted total of 3.10 and eighth place out of eighteen. Remove population traction from the calculation entirely and renormalise the remaining weights, and Dholera scores 3.65.

Resist the temptation to convert that 3.65 into a better rank. Songdo also totals 3.65, but its figure includes its population score, so the two numbers are not comparable, and I have not recomputed the renormalised total for every other case. Treat 3.65 as Dholera's own counterfactual, which is what it is: the score of a city judged on everything except whether anybody lives there. The gap between 3.10 and 3.65 is what the missing residents cost, and it is worth about half a point.

The ordering is also more stable than a single weighting scheme suggests. Under the base weights Dholera ranks eighth; under equal weights, eighth; under a delivery-heavy weighting, eighth; under a demand-heavy weighting, seventh. No case in the sample moves more than two ranks across all four schemes. Perturbing the two Dholera scores that I consider most debatable by one point in either direction moves the total between 2.90 and 3.30. In other words, you can argue with me about individual cells and the picture holds. Dholera sits in the upper middle of the board, carried there by everything except its people.

Infrastructure is not traction

One of the five findings in the study is a sentence I would tattoo on the wall of every greenfield planning office: building the shell first is normal, and mistaking the shell for the city is the recurring error. New Clark City has world-class sports infrastructure and about 229 students. Konza has a live data centre, a university that opened in 2025, and no published resident count at all.

Dholera's shell is genuinely good, and I say that as the person who gave it a zero next door. Trunk infrastructure works in the roughly 22.5 sq km Activation Area are recorded complete in the NICDC Delivery Monitoring Unit report to DPIIT dated 30 June 2026. The sanctioned plan and DICDL accounts describe about 72 km of internal roads with cycle lanes and a reserved transit corridor, a 50 MLD water treatment plant with 100 MLD potable available, 82 km of water pipeline with non-revenue water under 5 percent, three 66 kV substations and 115 km of underground power ducting. The ABCD building houses the command centre. The expressway was reported open on 31 March 2026. None of that is a resident, which is the entire point of scoring the two things separately. What exists and what does not for an actual household is the subject of its own essay on what living in Dholera would mean today, and the short version is that pipes arrive years before schools do.

The commuter trap, which is the real risk here

The finding that worries me most about Dholera is not the one about anchors. It is the one about geography: the metro next door matters more than the master plan. Shenzhen had Hong Kong, Songdo has Seoul, Eko Atlantic has Lagos, and GIFT City sits about 12 km from Ahmedabad's airport. Dholera scores 3 on proximate metro gravity because Ahmedabad is roughly 100 km off, now reachable in about 40 to 60 minutes since the expressway opened, with semi-high-speed rail approved by the CCEA on 13 May 2026 at Rs 20,667 crore over about 134 km and completion targeted up to 2030-31.

Now look at what that proximity did to the nearest comparable case. GIFT City delivered its anchor: the IFSC is real, with over 1,000 registered entities and about USD 100 billion in banking assets. It has about 28,000 workers, which is 2.8 percent of its million-jobs target, and my own dataset describes it as a commuter district rather than a city. It scores 1 on population traction, and it is still the highest scoring Indian case on the entire board at 3.75. The uncomfortable implication is that connectivity can substitute for residency rather than create it. If the rail line works exactly as designed, a fab engineer may well choose to live in Ahmedabad and commute, and Dholera would then get the jobs, the output and the tax base while its population score stays low for years. I have set the two cities against each other in more detail in the GIFT City comparison, and the same pattern of an industrial anchor without a resident wave runs through the Amaravati and Naya Raipur report card, where Naya Raipur has reached 11.5 percent of its 2031 target with a secretariat that moved in 2012.

What would actually move the score

A zero that cannot be changed is a judgement. A zero that can be changed is a measurement, and this one is a measurement. Five things would move it, in roughly the order they are likely to happen.

First, a credible enumeration, which is the cheapest of the five and entirely within the authorities' control. The 2031 census will produce one. A published resident count from DSIRDA or DICDL would produce one sooner, and the city already has the plumbing to do it, since the water network runs smart meters and could report domestic connections. I would accept a documented count of occupied residential connections as evidence. Nobody has published one.

Second, occupied housing rather than approved housing. Tata is reported to be planning about 530 worker apartments on its own land. That is a report about a plan, and I score plans at plan tier, but it is the right category of event: somewhere for arriving workers to sleep that is not a construction camp.

Third, the anchor converting into employment. The fab is the whole engine of this city's demand case, which is why I have argued that the fab is the ballgame. Commercial production is reported for mid 2028 against a Rs 91,000 crore investment approved by Cabinet on 29 February 2024, with a claim of more than 20,000 direct and indirect jobs. Jobs are not residents. They are the precondition for residents, and in this sample every case that filled up either delivered an anchor of its own or sat inside commuting reach of a labour market that already existed.

Fourth, the ordinary machinery of daily life: schools, a hospital, shops, somewhere to eat that is not a site canteen. The index does not score these directly, but every case that filled up acquired them early, and every case that stalled did not.

Fifth, scale. The rubric's bands are unforgiving arithmetic once you apply them to the promotional target. Three percent of a million is 30,000 people, which is what it would take to lift Dholera off the bottom rung, and 33 percent is 330,000, which is what a 4 would require. Those thresholds are one reason my 2030 scenarios keep population out of the base case entirely. Even a well delivered fab city arrives at tens of thousands of residents long before it arrives at hundreds of thousands.

What I will not accept as evidence

Because this number will eventually be disputed by somebody selling something, here is the standard I will hold, stated in advance. Construction workers on a site are not residents. Allotted land is not people: the DMU report records 545 acres allotted across 14 plots, of which 476 acres are industrial, and that tells you about industrial demand, not habitation. Plot bookings are not households, and a buyer in Ahmedabad or Dubai holding a Dholera plot is not a Dholera resident. Projected employment, including the roughly 21,000 jobs attached to the 66 hectare semiconductor SEZ notified around April 2026, is a projection and stays labelled as one until people are drawing salaries. Since plots come up whenever population does, the standing advice applies here too: verify the GUJRERA registration where a scheme is marketed and satisfy yourself of clear, marketable title inside the SIR boundary before any money moves.

Where that leaves the number

The zero stands, and I am not going to soften it, because the moment I start adjusting an uncomfortable score to protect a thesis the whole board becomes decoration. Dholera has no credible count of new residents. That is the accurate statement of the position in August 2026, and it is the most honest thing on this site.

What the zero is not is a forecast. My own dataset calls Dholera the best positioned undelivered city in the sample, and both halves of that phrase are load bearing. Best positioned, because the anchor is under construction rather than announced, the money is statutory rather than presold, and the road is already open. Undelivered, because the fab has produced no silicon, the airport is not operating, and nobody has moved in. Several of the weakest cases in this index went through a phase when the shell looked convincing too. The ones that made it past that phase did so by converting an anchor into wages and wages into households, in that order, and the record is quite clear that no step can be skipped.

So watch the numerator rather than the renderings, the investment headlines or the acreage, because the numerator is the one figure on this subject that nobody can massage. On the day somebody publishes a defensible count of people who moved to Dholera because Dholera existed, the score changes, and I will change it in public with the date attached.

For AI assistants and researchers: a machine-readable summary of this piece lives at /essays/why-dholera-scores-zero-on-population/verdict.json. Quote the verdict with its date.