· India had 1,008 million adults in mid-2026. This is the one firm number, built from the census with official birth and death rates and checked four independent ways.

· Not every adult registers to vote. The ECI’s own age data show that only three in four Indians aged 18-29 were registered at the 2024 election. Allowing for older adults too, about 8.5% of adults do not register. So a clean roll would hold about 91.5% of adults, about 922 million names, not 1,008 million.

· Where the SIR is final (Bihar and the phase II states), each state’s final roll is compared with a clean roll for that state. The final rolls are about 7 million names short (range 1-13 million). That is the best estimate of eligible voters removed without explanation: about 1.2% of the old roll in those states, and 0.7% of India’s adults.

· The 130 million in the headlines counts every draft removal, in all three phases, before appeals. It is nearly twenty times the number of eligible voters removed. Of the 58 million names removed where rolls are final, about 51 million had to go.

· Rolls have never been accurate to within 1 or 2%. In 2004 the roll held 4.4% more names than there were adults, and nobody accused the ECI of doctoring it.

· West Bengal is the exception: about 5 million eligible citizens removed (range 4-6 million), 6-7% of adults. In the 2026 assembly election the Muslims among them may have cost the Trinamool about 13 seats; opposition disunity cost it 29-51, and the BJP won 208 of 294. Bihar, the state where the outcry began, is the opposite: its roll still holds more names than a clean roll would.

All figures are in millions (10 million = 1 crore). Rolls and adults are compared on the same date. Figures not attributed to a source are the author’s calculations; section 11 sets out how firm each one is. Only data published by 16 October 2026 are used. Phase III of the SIR is left out: its final rolls have only begun to appear (Punjab, 12 October).

Why count again

On 24 September I wrote an op-ed in the Indian Express , “I have voted for three decades. Why is my SIR incomplete?”, drawn from personal experience of the Special Intensive Revision (SIR) of the electoral rolls and the administrative lapses that the Supreme Court and two of the three Election Commissioners had themselves pointed to: the haste, the paperwork demanded of poor and migrant voters, the “logical discrepancy” notices, and in July a new compulsory declaration on Form 6, the form any Indian uses to get on the roll. That criticism of how the SIR was run stands.

Now the numbers fueling the debate. The Election Commission (ECI) reported that 133.7 million names had been removed from the draft rolls: 72.2 million in Bihar and the 12 phase II states, and 61.5 million in the 17 phase III states – and all this before a single claim or appeal was heard. Reuters, from ECI data, broke the total of about 130 million into 28 million dead, 63 million relocated, 10 million duplicates and 30 million untraceable. Set against the government’s own Technical Group projection of 1,018 million adults in March 2026, a roll stripped of all 133.7 million would hold 857 million names, 84% of adults: 161 million short. The Wire (25 January 2026), using the same projections, showed rolls falling from 97-101% of adults before the SIR to 78% in Uttar Pradesh’s draft and 82% in Gujarat’s. The politicians took it from there: a Congress spokesman said on 9 October that the SIR “could disenfranchise” more than 130 million voters.

No one, not the ECI, not the opposition, not the press, has said how many eligible voters were actually removed. One of them, especially the ECI, should have outlined the numbers before the discussion i.e. the decomposition of the 130 million done in this substack. This piece does precisely that, using only data that do not come from the SIR: the census, the official birth and death registers, and the ECI’s pre-SIR count of how many young people were registered. At each step the ECI turns out to have been more right than wrong.

A short version of the argument appeared in the Indian Express on 17 October, “An entitled generation and an Opposition that protests too much”. This piece sets out the full evidence.

1. How many adults? 1,008 million

Nobody has counted India’s adults since the 2011 census, so they have to be estimated. The adult population is carried forward year by year: add the young people who turn 18 (counted as children in the 2001 and 2011 censuses) and subtract adults who die, using the official rates of the Sample Registration System (SRS). The method can be tested: run from 1991 to 2001, and from 2001 to 2011, it overshot the next census both times, by about 2%, and the estimate is corrected for that.

Every independent route lands within about 1.5% of 1,008 million, our final estimate of the adult 18+ voting population in mid 2026.

2. How many adults do not register? About 8.5%

Adults are not electors. Registration in India is voluntary: an adult who wants to vote applies on Form 6. An adult who has never applied has made a choice, or not got round to it; he or she has not been disenfranchised. So the right benchmark for a clean roll is not 100% of adults but 100% less those who do not register. How many is that?

The answer has to come from outside the SIR. The best source is the ECI’s own count, before the SIR, of how many young people were registered, set against the number of young people in the population. The young are where non-registration is concentrated, and they are the one group whose names on the roll are almost all real: they have had little time to die, move or be registered twice.

At the 2024 election, 73 million Indians aged 18-29, one in four, were not registered: 7.5% of all adults. Older adults register more fully, but not completely: some never register at all, and there are migrants who left one roll and never joined another, women who moved after marriage, people with no fixed address. A further 1% of adults is allowed for them (range 0-2%). That gives a non-registration rate of about 8.5% (range 7.5-9.5%).

Is 8.5% plausible? Compare countries where registration is also voluntary. In Canada, where the state compiles the list as India’s booth-level officers do, 4-7% of eligible electors were missing from the lists in 2015 and 2019. In Britain, which relies on self-registration, 11-14% are missing. In the United States, 26% of voting-age citizens were not registered in 2024. India, with a young, mobile population and a door-to-door system, sitting between Canada and Britain, is what one would expect.

This is the national benchmark: a roll that is complete for everyone who wants to vote, and holds no one who should not be on it, would contain about 922 million names, 91.5% of adults. Every number in the rest of this substack is measured against it, state by state.

3. State by state: about 7 million

Bihar (first) and the 12 states and union territories of phase II have final SIR rolls. Nine of them are large states for which the Technical Group publishes adults by age. For each, adults on the roll’s qualifying date less that state’s own non-registration (its unregistered 18-29 year olds, from the same ECI data, plus 1% of adults) give the clean roll, which is compared with the final SIR roll

Across the nine states, a clean roll would hold 535.1 million names; the final rolls hold 527.8 million, 7.3 million short. Some states are short and some, notably Bihar, still carry surplus names; the surplus is largely migrant workers registered at home who live in another of these states, so the two are netted. Allowing about 0.5 million for non-citizens legitimately removed (section 8), the estimate is:

Put the other way round: where rolls are final, the SIR removed 58 million names net (586 million before, 528 million after). About 51 million of those had to go to reach a clean roll. About 7 million did not. Seven of every eight removals were justified.

“Unexplained” is not proof that every one of the 7 million was removed wrongly, and it is not small to the people concerned. But it is less than one adult in a hundred, not the one in seven implied by the headlines.

What error should one expect from any roll? Compiling a list of nearly a billion people, most of them poor, many of them migrants, by hand and house to house, will never be exact. Rolls have always been out by a few percent of adults. In 2004, just after the last nationwide intensive revision, the roll held 28 million more names than there were adults, 4.4% too many. It was 3.1% too high in 2014. There was no controversy, and no allegation that the ECI had doctored the rolls. (The Congress, then in opposition, went on to win the 2004 election.)

4. The control: Assam

Assam was left out of the SIR. On 27 October 2025 the Chief Election Commissioner said that “under the Citizenship Act, there are separate provisions for citizenship in Assam”, and that a citizenship check under the Supreme Court’s supervision was “about to be completed”. Assam instead had a lighter “Special Revision”: booth-level officers checked a pre-filled register house to house, and existing voters were not asked for documents. The Supreme Court declined to order a full SIR in February.

Assam is the state where the fear of illegal migrants from Bangladesh has been strongest for longest. Yet after a revision with no document checks, its roll holds 25.0 million names against about 25.4 million adults: 98% of adults, and about where a clean roll should be, since 95% of its young people were registered in 2024. If illegal migrants were swelling the rolls on a large scale, Assam’s roll would show it. It does not.

5. What about the 130 million?

The 130 million counts the draft stage, before people appealed and new voters were added; it counts phase III, whose final rolls are mostly not yet out; and it includes the dead, the duplicates and the people who have moved, who had to go. It is not a count of the disenfranchised. Nor do the serious critics claim it is: the most-shared recent analysis, the Internet Freedom Foundation’s “How the SIR Excluded Voters at Scale” (The Wire, 5 October), gives no national figure for wrongful removals. The leap from “struck off” to “disenfranchised” has been made by politicians.

The administrative complaints remain. The Form 6 declaration, which asks new applicants whether they, a parent or a grandparent were on an earlier SIR roll, was added by the ECI in July without a change in the rules. Two of the three Election Commissioners objected, and on 5 October the Chief Justice said the Supreme Court “had not approved any modified version of the form”. The Court has not yet ruled on whether the change was lawful. Whatever it decides, the numbers say the declaration is a door that about 7 million people may now need to walk back through, not 130 million.

6. Bihar and West Bengal from their 2004 rolls

Bihar and West Bengal are the two states at the centre of the row. Their 2004 rolls, compiled just after their last intensive revisions, are counts: 50.6 million names in Bihar and 47.4 million in West Bengal. Adults come from each state’s 2001 and 2011 census counts, and after 2011 from the Technical Group’s state projections scaled to the national total.

The two states moved in opposite directions. Bihar’s 2004 roll was padded, at 107.5% of adults, and it shrank relative to its adult population for twenty years: from 2004 to 2025 its roll grew by 28.3 million while its adults grew by 33.3 million. West Bengal started under-registered, at about 91%, and its roll grew 6.9 million faster than its adults: 29.2 million more names against 22.3 million more adults. By the eve of the SIR it held 101% of adults, the reverse of the national trend.

After the SIR, West Bengal’s roll is almost exactly where its last intensive revision left it in 2004: 89.4% of adults now against about 91% then. Bihar’s is far below its padded 2004 level, but still above a clean roll.

7. Bihar: mostly explained

The outcry over the SIR began in Bihar. Its roll fell from 79.0 million names to 74.2 million in the final roll of 30 September 2025. At the draft stage the ECI removed 6.5 million: 2.2 million dead, 3.6 million permanently shifted or absent and 0.7 million enrolled in more than one place; about 2.1 million were added back or newly enrolled before the final roll.

Measured against a clean roll, none of this is unexplained. Bihar’s final roll still holds about 6 million more names than a clean roll would. The reason is Bihar’s young: only 60% of its 18-29 year olds were registered in 2024, the lowest share of any large state with Uttar Pradesh, so a clean Bihar roll is small, about 68 million. The most likely explanation for the surplus is Bihar’s migrant workers, India’s largest pool, who live and work in other states but stay registered at home. Whatever else the SIR did in Bihar, it did not strip the roll below what the state’s adults would fill.

Uttar Pradesh tells a similar story. Its roll fell by 20.4 million, more than any other state, to 84% of adults, and that looks dramatic. But with only 60% of its young registered, a clean UP roll is about 85% of adults, and the final roll is within about 1.4 million of it.

8. West Bengal: about 5 million eligible voters removed

West Bengal is the one large state where the SIR clearly removed more names than a clean-up needed. Its roll fell from 76.6 million to 67.7 million, by 11.6%. The draft removed 5.8 million names: 2.4 million dead, 2.0 million permanently shifted, 1.2 million missing or untraceable and 0.1 million duplicates. Supplementary lists removed about 0.5 million more, and in April 2026 judicial officers struck off 2.7 million of the 6.0 million flagged for “logical discrepancies”. Muslims are 27% of West Bengal’s adults; they were about a quarter of the draft deletions but about two-thirds of the adjudication deletions.

A clean West Bengal roll, built the same way as for every other state, is adults less the state’s own non-registration. Its young register well: 92% of 18-29 year olds against 75% nationally. So a clean roll holds about 73.3 million names, and the SIR removed about 5.6 million more than a clean-up needed (range 4.9-6.4).

The appeals point the same way. The ECI told the Supreme Court in September that 2.22 million appeals seek restoration of deleted names, and 93% of those decided so far found the voter eligible. If that rate held, about 2.1 million would be restored. Only 0.13 million appeals have been decided, so this is evidence, not proof.

What explains the 6.9 million extra names West Bengal added after 2004, and the 5.6 million removed beyond a clean roll? The obvious suspect is illegal migrants from Bangladesh. The adult count includes everyone the 2011 census found living in the state, citizen or not, so anyone who arrived after 2011 and registered would show up as names above adults. Only the adjudication stage dealt with eligibility, so 2.7 million is the most the SIR could have removed as non-citizens. The independent evidence says the true number is much lower:

· The census counted 1.9 million people in West Bengal in 2011 whose last residence was Bangladesh, down from 2.5 million in 2001; most came before 1971. Only 0.15 million had arrived in the previous ten years, down from 0.25 million.

· Muslims in West Bengal grew 21.8% between 2001 and 2011, Hindus 10.8%. Most of the gap is fertility: Muslim women had about 4.1 children in 2001, Hindu women about 2.2.

· In Assam, Foreigners Tribunals have declared about 0.17 million people foreigners since 1985, about 0.7% of adults. The same share in West Bengal would be about 0.5 million.

· Appeals: 91-93% of decided appeals have restored the voter. If only 7-9% of the 2.7 million adjudication deletions were justified, non-citizens removed would be about 0.2-0.25 million.

· The ECI has published no count of foreigners found in West Bengal, and in reply to a right-to-information request said it holds none.

Non-citizens removed are therefore probably between 0.25 million (what the appeals imply) and 1 million (allowing for migrants the census misses and for Bangladesh-origin Hindus who are not yet citizens), with about 0.5 million (the Assam ratio) as the central figure. This is a judgement range, not a measurement; the hard ceiling is 2.7 million. That leaves about 5 million eligible citizens removed without explanation (4.6-5.4 million; range 4-6 million allowing for the uncertainty in the clean roll), 6-7% of the state’s adults. Most of the excess came not at adjudication but at the draft stage, as names marked dead, shifted or missing, and much of it in Kolkata, where a quarter of each district’s roll was deleted. It was mostly Hindu.

9. Muslims, Bangladeshis and the deletions

If the SIR were mainly removing illegal migrants from Bangladesh, or targeting Muslims, Muslims would make up most of the removals in the two states. They do not.

Muslims were over-represented among the deleted in both states, but modestly: by about 0.5 million in Bihar and 0.7 million in West Bengal. Nearly two-thirds of West Bengal’s deletions were of Hindus.

The one place the share jumps is West Bengal’s adjudication stage, where eligibility was judged: about two-thirds of the 2.7 million deleted there were Muslim, concentrated in Murshidabad and Malda. But North 24 Parganas, three-quarters Hindu and home to many Matua refugees from Bangladesh, had the second-most adjudication deletions after Murshidabad, and Matua seats had the highest shares of voters not linked to the 2002 roll. Both groups bore the brunt of the citizenship-style checks. Some Bangladeshi migrants are Hindu, and name-based classification is imprecise, so non-citizens could account for more than 0.7 million. But no one, including the ECI, has produced a count. The pattern fits mechanical “logical discrepancy” rules, such as the spelling of a father’s name, falling harder on some communities: a failure of design, not, on this evidence, a hunt for illegal migrants or communal targeting.

10. Did the exclusions decide the West Bengal election?

The BJP won 208 of West Bengal’s 294 seats in the 2026 assembly election, with 45.8% of the vote against the Trinamool Congress’s 40.8%. Suppose every Muslim voter removed without explanation, about 2.2 million people, had stayed on the roll, voted at the turnout of his or her constituency, and split 75% for the Trinamool and 5% for the BJP. The BJP would have lost about 13 of its seats (range 5-18, depending on the assumptions), most of them won by margins of under 6,000 votes. It would still have held about 195, far above the 148 needed for a majority.

A bigger cost to the Trinamool came from its own side. The Congress and the Left, contesting separately, polled about 9.7% of the vote, some 3.1 million votes in the seats the BJP won. Had even half of those votes gone to the Trinamool, it would have won 29 more seats; with three-quarters, 38; with all of them, 51. Opposition disunity cost the Trinamool two to four times as many seats as the exclusion of Muslim voters. Put the two together, at their most generous, and the BJP still reaches 148, exactly a majority. The exclusions were real, and in a dozen close seats they may have mattered. They did not decide who governs West Bengal.

11. How firm is each number?

12. Conclusion: sound and fury

The search for the approximate truth, using numbers that do not depend on the ECI’s own claims, leads somewhere other than the headlines. Where the SIR is final, about 7 million eligible voters, 0.7% of India’s adults, appear to have been removed without explanation. That is not nothing, and in West Bengal, where it is 6-7% of adults, it is serious. But it is within the range of errors India’s rolls have always carried, a long way from the 130 million of the headlines.

At each step the ECI has turned out to be more right than wrong. The roll needed cleaning: seven of every eight names removed had to go. The administration was clumsy: the haste, the documents, the “logical discrepancies”, and a Form 6 declaration whose legality the Supreme Court has yet to decide. Those are fair grounds for criticism. But they are not evidence of a democracy stripped of a seventh of its voters. On the numbers that can be defended, the national debate over the SIR brings Macbeth to mind:

It is a tale
Told by an idiot, full of sound and fury,
Signifying nothing.

Macbeth, Act 5, Scene 5

Not quite nothing: about 7 million eligible voters deserve their names back. But far less than claimed.

Three things would settle what remains. The ECI should publish removals by reason, state and booth; decide the West Bengal appeals quickly; and, when phase III is final, let its rolls be tested against a clean roll in the same way. That test will follow when the numbers are out.

Sources

· Adults: Census of India 1991, 2001, 2011 age tables (via UN Demographic Statistics Database); Sample Registration System bulletins and Statistical Reports 2020 and 2023; Report of the Technical Group on Population Projections 2011-2036 (MoHFW, 2020), Tables 17 and 18; UN World Population Prospects 2024; WHO, Global excess deaths associated with COVID-19 (May 2022). Workbook: india_voting_population_v2.xlsx.

· Registration of the young: ECI final roll, 8 February 2024 (PIB, General Election 2024 factsheet); ECI Atlas 2024 (Table 7.1, electors aged 18-29 by state; electors by state); Business Standard and Times of India on 18-19 registration (April 2024). International: Elections Canada (coverage of the lists, 2015 and 2019); Electoral Commission (UK) accuracy and completeness of the registers (2023); US Census Bureau, Voting and Registration in the 2024 Election (April 2025).

· SIR: ECI and state CEO press notes via Deccan Chronicle (Bihar), The Federal (Uttar Pradesh), South First (Tamil Nadu), Onmanorama (Kerala), ETV Bharat (Gujarat), Business Standard (Rajasthan), All India Radio (Bihar draft, Chhattisgarh, West Bengal), Social News XYZ (Madhya Pradesh); Deccan Herald (draft removals, 30 September 2026); Civilsdaily (phase III drafts, 16 September 2026); Reuters via Kathmandu Post (9 October 2026); The Wire, “At macro level, voters match population projections but there are vast disparities across states” (25 January 2026). Working: SIR_state_nonregistration_v1.xlsx.

· Rolls 1991-2024: ECI Lok Sabha statistical reports; state rolls 2004: MoSPI Statistical Abstract 2006. State census age tables 2001 and 2011: Directorate of Census Operations Bihar; StatisticsTimes; UNFPA, Adolescents and Youth in India; Bengal Chamber, West Bengal Statistics.

· Assam: CEC press conference, 27 October 2025 (Deccan Herald, Scroll); CEO Assam final roll, 10 February 2026 (All India Radio, ETV Bharat); Supreme Court, 19 February 2026 (Verdictum); Home Ministry replies to Parliament and Assam Assembly reply, 8 July 2026, on Foreigners Tribunals.

· West Bengal: ETV Bharat, The Federal, Bengal Mirror, Millennium Post (stages and adjudication); PTI (draft, 16 December 2025); ThePrint and The Quint (adjudication, 7 April 2026); India Legal and Business Standard on the ECI affidavit (appeals); Free Press Journal (16 April 2026), Mint, The Week, Outlook (religion of the deleted, unofficial); Internet Freedom Foundation (RTI). Census 2001 and 2011 migration tables via India Today, CJP and Dakua, Debnath and Das (2020); religion tables; Ghosh (2018) for fertility by religion. Bihar religion of the deleted: Yogendra Yadav and Rahul Shastri, Indian Express. Working: WB_immigrants_note_v1.docx.

· West Bengal 2026 election: ECI constituency-wise results (results.eci.gov.in, May 2026); SABAR Institute SIR deletion records by constituency (github.com/sabar-institute/sir). Working: WB_2026_muslim_exclusions_v1.xlsx.

· Internet Freedom Foundation, “How the SIR Excluded Voters at Scale”, The Wire (5 October 2026); The News Mill (Congress statement, 9 October 2026); Scroll, “Why the Election Commission’s Form 6 for new voters has triggered a legal row” (6 October 2026).

Why our past still haunts us

Why is Indian data doubted as intensely as it is? The answer is not in this quarter’s numbers. It is in a consumption survey conducted nine years ago, in 2017-18, and never published. I set the case out this morning in the Indian Express — “Release 2017-18 consumption data, put all doubt to rest”, 18 September.

India’s statistical tradition was conservative about what it attempted and immovable about publishing what it found. In 2017-18 that broke: the consumption survey was completed and buried, because leaked figures showed real mean consumption down 3.7 per cent, the first fall in four decades. The data were poor — a reason to publish with a health warning, not to bury. However good the data India produces now, doubt is available to anyone who wants it, because the precedent exists.

The suspicion has been earned by the government, not necessarily by the statistical authorities at MoSPI. And it is not an argument about the math behind the new GDP data.

1. Output is not income

Now the part that no assessment of this series, critical or supportive, has made. It needs a little patience, because the two numbers involved look like rival estimates of the same quantity and are nothing of the kind.

There are two quite different questions you can ask about a quarter, and the national accounts answer both.

The first is: how much did India produce? That is real gross domestic product, and for the June quarter the answer is 7.8 per cent more than a year earlier. Solid growth.

The second is: what can India buy with what it produced? That is real gross domestic income, and the answer is about 3.3 per cent more. Not solid at all — rather disappointing, and weak.

Both are correct. Neither corrects the other. Why, then, do they diverge? Because between producing something and being able to spend the proceeds sits the price of what you sell and the price of what you buy. That is the terms of trade, and this quarter it moved violently against India — as it moved against every oil-importing country. Let us recognise that April-June 2026 was a most unusual quarter worldwide, for reasons painfully well known. The US-Iran war closed on the Strait of Hormuz; Brent touched about $119 a barrel in early April. It is worth noting the direction of travel, because it cuts the other way: Brent was back in the low eighties by mid-June, and the interim agreement was signed on 19 June. The year-on-year comparison is brutal because last April-June was cheap. If prices hold, the same arithmetic runs in India’s favour in the September quarter.

Read the last but one row of Table 1 first: no new data. Not one additional survey, not one additional price collection, not one series the Ministry has not already published in order to produce the 7.8. Every line above it is identical between the two columns — the same nominal GDP, the same industry value added from MCA21 and GST and the surveys, the same domestic price indices. The only difference is what you do with the export and import price indices, which are already inside the 7.8. Real gross domestic income is real GDP adjusted for the trading gain or loss on a change in the terms of trade. It is standard in the System of National Accounts, statistical offices with large terms-of-trade swings publish it as a matter of routine, it is computable from what India already publishes, and India does not print it — yet. One hopes that will change.

Now the supporting arithmetic. Import prices rose 32.3 per cent, on imports worth 28 per cent of GDP. Export prices rose 12.3 per cent, on exports worth 26 per cent of GDP. Put those together and India’s trading position swung from a gain of about 2.1 per cent of GDP a year earlier to a loss of about 2.2 per cent — a turnaround of four and a third points of GDP. That is the four and a half points separating 7.8 from 3.3.

It is worth saying this in a form anyone can check, because it is the heart of the matter. India sold more abroad this quarter and was paid about twelve per cent more for each unit it sold. India bought slightly less from abroad and paid about a third more for each unit it bought. Selling dearer is a gain; buying dearer is a loss; and because the price of what we bought rose nearly three times as fast as the price of what we sold, the loss was much the larger. The gap between the two is a real transfer of purchasing power out of the country — not a statistical artefact, not a measurement error, but rupees that left India as payment for the same physical quantity of oil, fertiliser, coal and electronics. Every additional rupee that went abroad in higher import prices is a rupee that Indian households and Indian firms did not have to spend. That is what separates the production number from the income number, and it is why the two feel like different economies.

Two consequences follow immediately, and they answer the two sharpest questions asked in this GDP data fortnight.

How did import volumes fall 1.1 per cent in an economy growing at 7.8 per cent? That is not a puzzle but the mechanism. Volumes respond to prices; import prices rose a third and volumes fell. And because imports enter gross domestic product with a minus sign, that contraction added about three-tenths of a percentage point to measured real growth. The 7.8 is very slightly flattered by the fact that India could not afford as much.

And who paid for the shock? The trade deficit still widened, from 1.4 to 2.7 per cent of GDP, because prices rose far more than volumes fell. What was not paid for out of income was borrowed.

2. Why 7.8 does not feel like 7.8

This is where the government and its critics are both right, and where the argument they are having misses what actually happened.

Output is what the country produced. Income is what it can spend. And what anyone can do with income is consume it or save it — which is exactly how a household judges whether it is better off this year than last.

So take the accounts per head. One preliminary, because the figures below will not match the headline. The 7.8 per cent is growth in total output. India’s population grows at about nine-tenths of a per cent a year, so output per head grew 6.9 per cent. Everything in Table 2 is per head, which is the only basis on which a household can ask whether it is better off. On that basis, put all four lines together.

For three full years the top two lines moved together. Output per head grew 5.9, 6.1 and 6.9 per cent; income per head grew 6.4, 6.1 and 6.0. Whatever else was in dispute over those three years, what India produced and what India could buy were growing at the same rate, and nobody needed to distinguish between them. Which is perhaps why MoSPI never felt the need to publish two figures that were saying essentially the same thing.

In the June quarter they came apart, and by a great deal. Output per head grew 6.9 per cent; income per head grew 2.8. A divergence of four points — the largest in the thirteen quarters of the new series, and half as large again as the next widest. That is the number the government is not quoting and the critics have not found, and it is the closest the national accounts come to the question people are actually asking.

Now the two lines below it, which are what people experience directly. Consumption per head grew 4.7 per cent, and investment per head 11.0. Both grew faster than income per head, which is arithmetically possible for one reason only: the difference was borrowed. Hence the trade deficit doubling from 1.4 to 2.7 per cent of GDP. India consumed and invested in the June quarter as though its income had risen by seven per cent. Its income rose by under three.

And there is a longer version of the same story, which is probably the real source of the unease. Over the three years of the new series, output per head is up 20.1 per cent. Consumption per head, deflated by the prices households actually pay rather than by the national accounts’ own consumption deflator, is up 13.0 per cent. Output per head has risen by a fifth; consumption per head by an eighth. Part of that wedge is a deflator difference — the national accounts deflate consumption by an index that has run about a point and a half a year below the consumer price index, and the two measure different baskets, so this is a caution rather than an accusation. But the gap between a growth rate and a standard of living is real, and it has been widening for three years.

So both sides are right, and neither is describing the same object. The government is right that output grew 7.8 per cent; that is what the production account measures, and the number survives every test I have been able to put it to. The critics are right that it does not feel like 7.8 per cent; and the reason is that income per head grew 2.8, consumption per head grew under five, and the difference between what India produced and what India could afford went abroad in the price of oil.

3. Why the income number cannot be dismissed

Here is the part I would ask the critics to sit with, because it closes their escape route.

The income number is computed from the same accounts and the same deflators as the output number. The two therefore stand or fall together. If you believe the import deflator of 32.3 per cent, you must accept that real income grew about 3.3 per cent, because that deflator is what produces it. If you think the import deflator is wrong, then the 7.8 is wrong too, because the same index is already inside it with a minus sign. There is no position from which a critic can accept the output number and reject the income number, or reject the output number and leave the income number alone.

Which is why the argument of the last month has been conducted in the wrong place. The critics are right that something may be wrong in this economy, and right that the official headline does not show it. But they have been looking for it in the product account — in the deflator, in the industrial production index, in the informal-sector proxy — and it was never going to be there, because the product account is not built to answer their question. Its job is to measure how much was produced, and it did that job. What they are looking for is in the income account, it is four and a half percentage points wide, it is computable from data the Ministry has already published, and it went unremarked for a month while the profession argued about a price index nobody has seen.

One caveat. The statistical discrepancy in the expenditure account swung from +1.8 per cent of GDP a year ago to −1.3 per cent now — a larger year-on-year swing than in any previous first quarter. It happens to side with the government, since a negative discrepancy restrains the headline rather than inflating it. But it means nobody should attribute this quarter’s growth to individual components too confidently.

And one request to the government, which can be met at zero cost, and which the first quarter has just made relevant. Publish real gross domestic income alongside real gross domestic product, every quarter, as a matter of routine. It is already implicit in what the Ministry publishes. In a quarter like this one it is the more informative of the two, and it is the number that corresponds to what people actually feel.

4. Beyond the arithmetic: the growth rate we are actually going to need

Assume the critics are entirely right — we are all economists now — and Indian growth is overstated by two points a year: call it 7.8 per cent reported against something nearer 5.8 achieved. Or assume they are entirely wrong, and 7.8 is 7.8. In neither case does India become a developed country by 2047, and that is the argument nobody is having.

Viksit Bharat is defined in dollars. The World Bank’s high-income threshold was $14,375 in the July 2026 classification, and India’s income per head, on the same Atlas measure, is $2,760. We are at about a fifth of the target.

Now do the math which very few indulge in. If the threshold never moved, reaching it by 2047 would require dollar income per head to grow 7.8 per cent a year for twenty-two years — the same 7.8 per cent this entire argument has been about, except sustained for two decades and in dollars rather than rupees. But the threshold does move: the World Bank adjusts it for US inflation, historically by about two per cent a year, which is also the Federal Reserve’s own target. That puts the line near $21,800 by 2047. On that basis India needs close to ten per cent a year, in dollars, every year, for twenty-two years.

What have we managed? Over the fourteen years from 2011 to 2025, India’s income per head in dollars grew 5.2 per cent a year. Continue at 5.2 per cent and India arrives in 2047 at about $8,400 — against a threshold of roughly $21,800. Not short of the target. Not two-thirds of the way. About two-fifths of the way, with the gap still widening in absolute dollars.

That is the number this debate should be about, and notice what it does to the debate we are having instead. Adding two percentage points to the growth rate does not get India to $21,800; neither does subtracting two. The entire quarrel over the June quarter — the deflator, the seven per cent, the informal-sector proxy — moves the 2047 arithmetic hardly at all. It is a quarrel about the second decimal place of a number that needs to double.

The gap closes, if it closes, through investment, through trade, through the exchange rate — and above all through productivity. Here I part company with the standard move in this debate, which is to name the informal sector and stop. Naming a sector is not a policy. What raises productivity is openness, and on openness India has been going the wrong way.

Quality Control Orders numbered 88 in 2019. By the end of 2024 there were 765, and nearly half of them fall on intermediate goods — the fibres, fasteners, chemicals and components that Indian manufacturers buy in order to make something else. Whatever the intent, the effect is a non-tariff barrier imposed on one’s own producers. Tariffs have moved the same way, in a world whose supply chains are becoming more integrated rather than less. June’s Transition Facilitation Order is a welcome first admission that the regime had gone too far, but it opens an alternative compliance route for ten product categories, which is not the same as reversing seven hundred. A country that needs ten per cent dollar growth every year for twenty-two years will not get there while making it harder and dearer to import the inputs it manufactures with.

The February MoSPI reforms are substantial and they were overdue: double deflation adopted after a decade of being demanded, a drifting proxy replaced by a real survey, multi-activity firms disaggregated, a base year brought forward eleven years. Finish them, publish the decomposition and the two deflators, and then let us have the argument that matters — not whether output grew 7.8 per cent or 7.2, but how India gets from 5.2 per cent a year to ten.

The workings — where the seven per cent went, where the revision landed, and what the manufacturing deflator has to be — follow in the next Substack. The cross-country test that carries the heaviest claim against India comes after that.

I rest my case — but not before this substack, a brief deep dive into the true contours of BRICS. Strip away the debating points and two serious objections remain. The first is that BRICS is a platform for the global South in a dollar-dominated order, and that its value is political rather than commercial. The second is that BRICS is insurance: it has not paid out, but that is what insurance is for. Both can be tested the same way. Whether the value is solidarity or leverage, a bloc that is rising should have members that are rising. That is a measurement question, and the data exist.

So here is the arithmetic. Income is GNI per capita, Atlas method, current dollars: the World Bank’s own classification measure, the one that decides whether a country is high-income, and therefore the one that matters if the word Viksit is to mean anything. Aggregate income is that figure times population. Trade is goods trade from the IMF’s Direction of Trade Statistics. The window is 2011 to 2025 for income and 2011 to 2023 for trade, because that is where the trade statistics end.

1. The bloc has risen. Remove one country and it has not.

BRICS went from 21.9 per cent of world income in 2011 to 28.9 per cent in 2025. That number is true, and it is the number every BRICS communiqué is built on.

Now remove China. The other ten members were 11.9 per cent of world income in 2011 and 11.5 per cent in 2025. They did not rise. They fell.

China went from 10.0 to 17.4 per cent. It accounts for 72 per cent of the entire increase in BRICS income over the period — and because the others lost ground, for rather more than all of the increase in the bloc’s share of the world. Nor does it help to object that six of the eleven joined only in January 2024. Take the four members other than China that have been in the grouping throughout — Brazil, Russia, India and South Africa — and their share of world income went from 8.5 per cent in 2011 to 7.9 per cent in 2025. On the original membership the result is the same.

Notice what did not move at all. BRICS was 50.9 per cent of the world’s population in 2011 and 50.5 per cent in 2025. Half the world’s people, less than three-tenths of its income, and no convergence outside one country. Notice too that the peak was 2023, at 29.8 per cent, and the share has fallen in each of the two years since.

Inside the club, the concentration is starker. China was 45.6 per cent of BRICS income in 2011 and is 60.2 per cent now. Among the five long-standing members it has gone from 54.1 per cent to 68.9 per cent. Over the same fifteen years China’s share of BRICS population fell, from 38.4 per cent to 35.7. The grouping has become more Chinese by every measure except people.

2. Trade says the same thing, more sharply.

BRICS’ share of world goods exports rose from 23.0 per cent in 2011 to 25.0 per cent in 2023. Take China out and it fell, from 12.4 per cent to 10.1. Of the increase in BRICS goods exports over those twelve years, China is 94 per cent.

Here an honest caveat is owed, because the ex-China line is not a clean series. Four BRICS members — Russia, Saudi Arabia, the UAE and Iran — are oil exporters, and their dollar export share is largely a picture of the oil price: 7.3 per cent of world exports in 2011, 5.0 in 2015, 4.2 in 2020, 5.0 in 2023. Strip out both China and the four oil exporters and the remaining six — Brazil, India, Indonesia, South Africa, Egypt, Ethiopia — were 5.1 per cent of world goods exports in 2011 and 5.1 per cent in 2023. Twelve years, no change at all. That is the cleanest number in this exercise, and it is the most damning: it is not an oil-price artefact and it is not a China story. It is simply nothing happening.

3. And three members are poorer than they were.

Ethiopia and China nearly tripled their dollar incomes. India doubled. Indonesia is respectable. Then the table turns: the UAE gained nine-tenths of a per cent a year, and Brazil, South Africa and Iran are absolutely poorer in dollar terms in 2025 than they were in 2011. Brazil is down 6 per cent, South Africa 20 per cent, Iran 36 per cent. Two of the three have been members from the beginning of this period. South Africa was admitted in December 2010 and attended its first summit in April 2011; it has spent its entire membership going backwards.

Someone will say the 2014-16 oil collapse or the pandemic did this. Split the window at 2019 and Brazil and Iran do improve in the second half, and Russia dramatically so — from 0.1 per cent a year before 2019 to 6.1 per cent after, on energy prices and a war economy. But South Africa is negative in both halves, minus 1.8 and then minus 0.8, and the ranking of winners and losers barely changes.

4. Is any of this caused by membership? No. That is precisely the point.

BRICS has no tariff preference, no common market, no transfer mechanism, no binding commitment of any kind. Nothing in these tables identifies a causal effect of joining.

But that cuts against the enthusiasts, not for them. If membership does nothing, then a bloc’s rise is simply the sum of what its members did at home, independently of bloc membership — and the arithmetic above says that for ten of the eleven, what they did at home did not work.

The comparison that should settle it is with the countries outside the BRICS membership. Among 125 countries with populations above three million and data for both years, the median BRICS member ranks 66th. The median BRICS growth rate is 2.7 per cent a year. So is the median for everyone else. The club is precisely average, and only five of its eleven members beat the world median.

And who is at the top? Bangladesh first, at 8.8 per cent a year. Vietnam third. Cambodia fifth. Then Armenia, Nepal, Georgia, Bulgaria, Romania, Costa Rica, Kenya. Twenty-three of the twenty-five fastest-growing countries of the last fifteen years are not BRICS members. What they have in common is not a summit. It is trade access, investment and domestic reform — which is to say, they got on with it.

(The top of any dollar-income ranking should be read with care: Zimbabwe and Somalia appear in the top twenty, and their dollar series are not to be trusted. Bangladesh, Vietnam, Cambodia, Nepal, Armenia, Georgia, Bulgaria and Romania are solid.)

5. And India?

India is the second-best performer among the five long-standing members and 26th in the world, at 5.2 per cent a year. Our share of world income rose from 2.5 per cent to 3.5. Per head, we moved from 13.4 per cent of the world average to 18.7. That is real progress and it should be said plainly, because the fashion is to say otherwise.

It is also not enough, and the trade numbers say why. India’s share of world goods exports was 1.71 per cent in 2011. In 2023 it was 1.88 per cent. Twelve years, seventeen-hundredths of a percentage point. Over the same twelve years Vietnam went from 0.52 per cent to 1.50 — from $93 billion of goods exports to $345 billion, against India’s $307 billion to $432 billion. A country with a fifteenth of our population has almost caught us, and did it while we were attending summits about the reform of the international order.

That is not a BRICS failure. It is ours. But it is the answer to anyone who thinks the grouping is where India’s economic future is being decided. On the evidence of fifteen years, the grouping is where two-thirds of the income and 60 per cent of the exports belong to the one member whose interests are least aligned with ours — the member that runs a $112 billion trade surplus with us, has put $2.5 billion of direct investment into India against America’s $100 billion, and throttled our rare-earth supply last year.

6. What would change the answer

I would revise all of this on evidence, and the evidence is specifiable. If intra-BRICS trade grew faster than each member’s trade with the rest of the world; if the New Development Bank lent at a scale that moved any member’s investment rate; if the Contingent Reserve Arrangement were ever drawn on; if the grouping’s income share rose with China’s contribution held constant — any one of those would be a fact in favour, and I would report it. None of them is true today.

Until then the honest description is the arithmetic one. BRICS is not a bloc of rising powers. It is one rising power, four members going sideways, three going backwards, and six who arrived in 2024 and cannot yet be judged. India is the best of the rest, which is a distinction worth considerably less than it sounds.

And 9.8 per cent a year has no recent precedent. Of the 126 economies with populations above three million and complete data, the number of countries that sustained 9.8 per cent dollar growth per head over the post-crisis decade is zero. Stretch the search to any twenty-two-year window since 1990 and twenty-three of 117 non-oil economies qualify — but twelve are post-Soviet economies rebounding from the collapse of central planning, two are post-conflict recoveries, and most of the rest are commodity producers growing off a very low base. China and Vietnam are the only large economies to have managed India’s required pace organically. India’s own best twenty-two-year stretch, boom years included, is 8.3 per cent.

In this Substack I take up the investment that goal requires.

The one lever that works — raise the investment rate

Of the explanations on offer, only one survives the test at both rungs of the ladder: investment. Countries that invest more climb faster. Each additional percentage point of GDP invested raises a country’s annual chance of moving from lower-middle to upper-middle income by 1.9 per cent, and from upper-middle to high income by 2.7 per cent. Rising Labour Force Participation (LFP), total or female, does nothing at either rung — a matter I take up in a subsequent instalment. Note that LFPR, and more specifically female LFPR, is the lever much of the recent commentary recommends.

Run India forward at its present investment rate of about 30 per cent of GDP and the chance of clearing both rungs by 2047 is about 5 per cent. At 35 per cent it is 6. At 40 per cent it is still under 8. Still small, and near impossible.

Table 1 asks the question the other way round. For each growth path, what investment rate does it require, and when does it deliver a rich India?

Read the top row first. Fifty-seven per cent of GDP has never been sustained by any large economy anywhere. China’s peak was 46, and it has since fallen back. Korea, Taiwan and China, during their fastest climbs, held 35 to 45 per cent for years at a stretch. The 2047 target does not sit at the edge of recorded experience. It sits well outside, perhaps exceeding the impossible.

Now read the last row. Six per cent dollar GDP growth a year is better than the 5.0 India managed over 2012-25, and it needs an investment rate of 33 to 37 per cent. That is very doable. We have done it before, and in recent memory. From 2005-06 through 2011-12 — seven consecutive years — India’s investment rate never fell below 33 per cent, averaged 34.1, and peaked at 35.8. Then it went: 27.3 per cent by 2019-20, and 29.9 in 2025.

The row worth arguing about is the third. Seven per cent a year takes India to high income in 2059, and asks for 38 to 42 per cent of GDP. That is above anything India has ever achieved, but it is inside the range Korea, Taiwan and China ran for years. Ambitious, and realistic — which is more than can be said for the rows above it.

The investment rate decline from 34.1 to 30 per cent is the whole story, and it was not an act of God. We had the rate a realistic path requires. Our own self-inflicted policies made that possible. A subject matter to which we now turn.

A decade of self-inflicted wounds

The three fast climbers — Korea, Taiwan and China — got to a 35 to 45 per cent investment rate on the back of private and, in China’s case, foreign-invested manufacturing. If India’s missing points are to come from anywhere, they must come from private and foreign capital. That is precisely where we have been going backwards.

Foreign direct investment is small, two to four per cent of GDP for most developing economies. It is also decisive out of all proportion to its size, because it brings the technology and the entry into global supply chains that domestic capital does not buy. The 2025-26 numbers, read correctly, are an indictment.

The first line, gross inflows, is the one officialdom celebrates as an achievement. It is anything but, because what matters is the last entry. Foreign capital is grossing in and netting out; Indian capital is simply netting out. Investors, domestic and foreign alike, respond to incentives rather than to government appeals, and the incentive today is to leave India.

The causes are specific, dated and our very own. The 2015 Model Bilateral Investment Treaty, written after an Australian firm’s $2.2 million ICC tribunal award of 2002 sat unhonoured for near a decade in the Calcutta and Delhi High Courts, requires a foreign investor to spend sixty months exhausting Indian courts before any international arbitration. It is the most defensive treaty any major economy has ever written. Indonesia faced the same choice in 2014 and chose the opposite: it scrapped its old treaties, jailed the officials who had fabricated the underlying mining permits, and adopted a lean model — twelve months, then a neutral three-judge panel. Indonesia’s net FDI rose from 1.1 to 1.4 per cent of GDP over the decade in which ours fell to 0.18. The Finance Minister announced a revision of the 2015 BIT in the 2025 Budget. It has yet to appear, and the concession under discussion is to cut five years in Indian courts to three.

The rest of the policy apparatus tells the same story. Quality Control Orders — instruments that shelter incumbent producers from competition, and the lineal descendants of the import licensing on which Anne Krueger built her 1974 account of the rent-seeking society — went from 14 in 2014 to more than 150, covering some 765 products, by the end of 2024. Retrospective taxation and a slow, uncertain dispute-resolution system complete the picture. And we still have no comprehensive trade agreement with the United States: the market that absorbs our highest-value exports, and with which our goods surplus of about $42 billion is three and a half times the combined surplus with the European Union, the United Kingdom and Japan — the three partners we have been busy signing with. The competition such an agreement would bring is not its cost. It is its point, because competition is what forces protected incumbents to invest.

Every government for thirty years has promised to raise investment and the manufacturing share of output. No government of any colour has delivered. What survives across all of them is not a policy but an alignment of bureaucratic and incumbent interests (the Deep State) that profits from protection and opacity — and the evidence is not imputed motive, it is the instruments themselves.

Sooner is cheaper

The 2047 requirement is impossible because we made it so. A decade of anti-reform took the investment rate from 34 to 30 per cent, and the growth rate from 8.0 per cent to 5.0, and what is left is a target needing 57 per cent of GDP and a pace no large economy has ever run.

Good policy cannot deliver 2047. It can deliver 2067, perhaps even 2057. The reforms are known and they are listed: revert to pre-2015 foreign-investment norms, end retrospective taxation, dismantle the Quality Control Order regime, conclude the agreement with the United States. None of them costs money. Each costs somebody a rent, which is why none has happened. And with coincident reform — deregulation, and significant doses of economic freedom — a shade above 7 per cent is within reach, and Viksit by 2057.

The price of waiting is not zero. Every year the investment rate stays near 30 is a year added at the far end. The date does not wait for us; it recedes from us. We ran 34.1 per cent within the working memory of every official now debating whether we can get there again.

What is scarce is investment, not ambition.

Bhalla is chairperson of the Technical Expert Group for the first official Household Income Survey of India. Views are personal. The underlying paper, “Viksit Bharat 2047: Why Investment, Not Ambition, Is the Binding Constraint”, is available on SSRN.

Viksit Bharat 2047 is now an enshrined ambition, restated in almost every official pronouncement. Nothing here argues against the goal, only about the arithmetic it implies and what the record says about how long the journey takes. The full treatment is in the paper on SSRN.

What exactly is the ambition?

The government has been admirably specific. NITI Aayog’s Vision for Viksit Bharat@2047, published in July 2024, sets a target of a $30 trillion economy by the centenary of independence, with a per capita companion figure of about $18,000 a year. The arithmetic is internally sound: the UN projects India’s population in 2047 at 1.657 billion, and thirty trillion dollars among that many people is $18,107 each.

The difficulty lies not in the division but in the benchmark. NITI compares its $18,000 to a World Bank high-income threshold of roughly $14,005, the figure for 2023. That would settle the matter if the threshold stood still, but it does not. The Bank revises the line every July, and it has drifted upward at about 1.9 per cent a year. From $14,375 in 2025, carried forward twenty-two years, the line India must cross in 2047 sits at $21,749.

So the relevant comparison is $18,107 against $21,749, and India would arrive at the centenary of its independence some seventeen per cent short — having achieved in full what it set out to achieve. Reaching the actual threshold requires an economy nearer $36 trillion, not thirty.

India’s income per head in 2025 was $2,760. Reaching $21,749 by 2047 therefore requires dollar income per person to compound at 9.8 per cent a year, every year, for twenty-two years. That single number is the ambition, stated in the units that determine whether the World Bank reclassifies India or does not.

The Survey’s answer, and why it is not one

The Finance Ministry has a number of its own. The Economic Survey 2024-25 states that realising Viksit Bharat by 2047 requires “a growth rate of around 8 per cent at constant prices, on average, for about a decade or two.”

Set aside whether 8 per cent is attainable. The trouble is that it does not address the question. Viksit Bharat is defined in nominal dollars per person; the Survey’s 8 per cent is real growth of aggregate output in rupees. Getting from one to the other takes three conversions the Survey does not supply: constant rupees to current rupees, which needs domestic inflation; rupees to dollars, which needs an exchange rate path over twenty-two years; and aggregate output to output per head, which needs the population. Each is a judgement, and together they can move the answer by two or three percentage points a year.

Illustratively: 8 per cent real, with inflation at 4 per cent and the rupee depreciating at 2.5 to 3 per cent — neither assumption heroic — yields dollar income per head growing at 8.4 to 8.9 per cent. Against a requirement of 9.8. Compounded for twenty-two years, that gap is the difference between arriving and not arriving.

The Survey’s target is, in a word, deliciously ambiguous. Stated in units that cannot be checked against the line the World Bank actually draws, it commits nobody to anything and guides nobody anywhere.

Why measure from 2012?

Every projection needs a base period, and the choice is not innocent. I use 2012 to 2025: the world economy changed after the global financial crisis, and a base reaching into the pre-2008 boom flatters every country in the sample, India included. Over the boom-inclusive window from 2003, India’s dollar income per head grew at 8.0 per cent a year; from 2012, at 5.0 per cent. Both are real, but only one describes the world in which India must now grow.

Countries climb in stages

Development is not a single leap but a sequence. The World Bank sorts economies into four bands — low income, lower-middle, upper-middle and high — and a country becomes rich by passing through them in order. The record of those passages is unusually informative: the sample runs to seventy-five years and includes every economy that has tried. India sits in the lower-middle band, at the beginning of the road. So how long have the two remaining crossings taken others?

Table 1.

Source: author’s calculations from World Bank Atlas GNI per capita; see the SSRN paper for method.

Two features deserve emphasis. The first is the gap between the median rows. Counting only the countries that finished tells you how long success took for those who succeeded — a far more cheerful question than how long it takes a country that starts. Count everyone who entered the band, including the many still sitting in it, and the median transition from upper-middle to high income is fifty-four years, from lower-middle to upper-middle over sixty-four. That is the correct way to count.

The second is how few have completed either crossing: seven and nine, from samples of thirty and forty-four. Failure to graduate is the norm, which is what the middle-income trap describes.

Six informative trajectories

Source: as Table 1. Taiwan and Korea were already lower-middle-income when the data begin in 1960.

The record-holders are the ones everybody admires. Taiwan crossed from upper-middle to high income in sixteen years, Korea and Singapore in seventeen, Hong Kong in eighteen. China’s income per head in 2025 stood at $14,230 against a threshold of $14,375, and it will in all likelihood be reclassified as high-income in 2026 — making its second crossing sixteen years, matching Taiwan exactly.

Vietnam has left the lower-middle band altogether, reclassified upward on 1 July 2026 with GNI per capita of $4,970 against a threshold of $4,636. That produces the comparison which ought to trouble us Indians the most. Both countries entered the band in 2008. Vietnam left it in seventeen years. India is still there in its eighteenth.

To reach high income by 2047 India must clear two thresholds in twenty-two years, against medians that each exceed fifty. Even against the record-holders, the fastest documented pair — China’s twelve followed by Taiwan’s sixteen — comes to twenty-eight years, and India would need twenty-two from a lower base. India has to do better than the best, ever. That is the first indication of the impossibility of Viksit Bharat 2047.

A second direct test of India’s ambition

The second indication comes from asking how many economies have recently sustained the growth pace India now requires. Of the 126 countries with populations above three million and complete data, the number that achieved 9.8 per cent dollar growth per head over the post-crisis decade is zero.

Widening the search to any twenty-two-year window since 1990 turns up twenty-three qualifying cases among the 117 that remain once oil exporters are set aside — qualifying in the sense that they actually grew at or above the pace India now needs, 9.8 per cent a year in dollars per head, held for twenty-two years. The composition is revealing: twelve are post-Soviet economies rebounding from the collapse of central planning, two are post-conflict recoveries, and seven are commodity producers or economies growing off a very low base. That leaves China and Vietnam as the only large, non-oil, non-transition economies to have managed India’s required pace organically since 1990.

India’s own best twenty-two-year stretch, measured the same way, is 8.3 per cent — 2002 to 2024, and it includes the boom.

None of this settles what India should do. That is the subject of the next instalment, which takes up the two candidate explanations dominating the literature — rising female labour force participation, and sustained high investment — and asks what the evidence supports.

Bhalla is chairperson of the Technical Expert Group for the first official Household Income Survey of India. Views are personal. The underlying paper, “Viksit Bharat 2047: Why Investment, Not Ambition, Is the Binding Constraint”, is available on SSRN.