# The Statistic That Improves

Author: Pramod Prasanth
Published: 2026-07-29
Canonical: https://pramod.ch/essays/the-statistic-that-improves
Description: India rations entry to medical education at roughly one government seat for every thirty-two people who sit the exam, and one seat of any kind for every fifteen. The ratio decides how the exam must be built, who can afford to compete, and which numbers the state finds it convenient to publish. In the year the paper leaked and the exam was re-run, it stopped publishing the denominator.

---

The National Institute of Design publishes an admissions handbook for its Bachelor of Design programme. The 2026-27 edition runs to the usual contents. Eligibility, the structure of the entrance test, fee schedules, a calendar of dates. On its fourth page is a table of disciplines at the Ahmedabad campus and the number of places in each.

Product Design, nineteen. Animation Film Design, nineteen. Graphic Design, nineteen. Textile Design, nineteen. Then four more at thirteen apiece: Exhibition Design, Film and Video Communication, Ceramic and Glass Design, Furniture and Interior Design.[^1]

The handbook states this flatly, as an administrative fact, alongside the fee schedule. It also states, in a note under the table, that the number of seats for Indian nationals in the programme including reserved categories is one hundred and twenty-eight, and that the figure for each discipline is derived from that total and rounded to the nearest whole number. So the nineteen is not a judgement about how many product designers the country needs. It is one hundred and twenty-eight, divided.

Note where the reserved categories sit in that sentence. They are inside the one hundred and twenty-eight, not added to it. How a fixed number is shared out is a real argument and a bitter one, and it is not this essay's argument. This one is about the size of the number before anybody divides it.

For the 2024-25 intake, 15,839 people applied to the undergraduate programme and 6,897 to the postgraduate one. Around twenty-two thousand applications, for every NID campus in the country.[^1]

Now hold that table next to a much larger one. In 2025, 2,209,318 people sat the National Eligibility cum Entrance Test for undergraduate medical admission. That figure comes from the National Testing Agency's own result release, which published registered, appeared and absent candidates for seven consecutive years, to the person.[^2] Against those 2.2 million stood roughly 129,000 MBBS seats that admission year, about half of them in government colleges.

Two funnels, two orders of magnitude apart in absolute size, with the same shape. NID's undergraduate programme takes about three in every hundred who register. The government medical system takes about three in every hundred who sit. The ratio is not what separates them. One of them has an expansion story and the other has none.

*Figure: Each square is one applicant in a hundred. The design institute took 22,736 applications in 2024 and the medical examination took 2,209,318 in 2025, so the two funnels differ in size by roughly a hundredfold. The shape does not differ at all.*

This essay is about the one that does. It is about the difference between a ratio getting better and more people getting in.

Between 2019 and 2025 a candidate's chance of a medical seat improved, and three quarters of a million more people were turned away. Both of those come from the state's own figures. Only one of them gets quoted.

The first essay in this pair took the same ratio from the other end, asking not how many places there are but who gets them. This one is about the number itself, and it can be read on its own.

## What India actually built

The obvious response to a ratio like that is to say build more, and the obvious rebuttal is that India has been building at a rate few countries have matched.

Before 2014 the country had 51,348 MBBS seats across 387 medical colleges. That figure is the government's own, repeated across press releases and parliamentary replies for a decade.[^3] By the 2026-27 admission year the National Medical Commission's seat matrix listed 823 colleges and 136,939 seats, and that matrix explicitly excludes the Institutes of National Importance, the AIIMS-type institutions, which are wholly government.[^4] More than twice the colleges. More than two and a half times the seats. Thirteen years.

The National Medical Commission, answering Parliament in February 2026, gave the year-on-year additions: 2,963 seats in 2020-21, then 8,790, then 7,398, then 9,652, then 8,641, then 11,119 in 2025-26. Forty-eight thousand five hundred and sixty-three seats in six years.[^5]

No state refusing to build produces that annexure. Those figures come close to doubling a national medical system inside a decade, and any argument that ignores them is arguing with something other than the record.

Take the concession all the way, then put the other number beside it. Across those same six years, the number of people sitting NEET rose from 1,366,945 to 2,209,318.[^2] That is 842,373 additional candidates against 48,563 additional seats.

Seventeen additional candidates arrived for every additional seat the state built.

Expansion at eight or nine per cent a year, sustained for over a decade, is a serious industrial achievement. Against that denominator it is a treadmill. The gate did not widen. It stayed the same width while both sides of it grew.

> India built a second medical system in a decade and turned away three quarters of a million more people than before.

## Where the expansion went

The aggregate figure conceals what it is made of. Take the National Medical Commission's seat matrix for 2026-27, published on 14 July 2026. It reports two columns for each sector: seats renewed, and seats newly increased.

Government colleges: 441 of them, 61,185 seats renewed, 2,111 increased, 63,296 in total. Private colleges: 382 of them, 65,843 renewed, 7,800 increased, 73,643 in total.[^4]

Of the 9,911 medical seats added to the country that year, 7,800 went to private colleges.

Private colleges now hold 53.8 per cent of MBBS seats on that matrix, against 46.2 per cent in government. That understates government, because the matrix leaves out the Institutes of National Importance, and every one of those is public. Add them back and the government share rises above 46.2 per cent. It does not rise to a majority.

When did it lose that majority? The answer depends on a decision nobody announced.

The previous year's matrix, revised as at 31 March 2025, lists every MBBS college in the country with its management type and annual intake. It totals 780 colleges and 118,190 seats.[^6] Its management column, though, is not the two-way split every downstream report turns it into. It has five values: Govt., Govt-Society, Trust, Society and Private.

Count the rows and they reconcile exactly to the document's own totals. Government colleges hold 58,099 seats. Trusts hold 45,616, societies 8,050, private bodies 4,350. Then come twelve colleges, holding 2,075 seats, filed under Govt-Society. They are the colleges run by the Gujarat Medical Education and Research Society.

Count those twelve as government and government held 60,174 seats in 2024-25, or 50.91 per cent. A majority. Count them as private and government held 58,099, or 49.16 per cent. Not a majority.

Nothing changed on the ground. An administrative label decided whether the Indian state looked like the majority provider of medical education in 2024-25 or a minority one.

The buildings, the faculty, the fee structures and the students are identical under both counts. Only the filing moves.

Remember that mechanism. It is about to appear again, in a different dataset, doing exactly the same work.

## The metric that was never real

For years the government's position on medical workforce adequacy has been that India has cleared the standard. The ratio quoted is one doctor for every 811 people, against a norm of one per thousand.

The construction is public, and the ministry states it in one sentence. There are 13,86,150 registered allopathic doctors and 7,51,768 registered AYUSH practitioners, and assuming that 80 per cent of registered practitioners in both systems are available, the ratio comes to 1:811.[^7]

The availability assumption is applied evenly to both groups, so it is not where the work is being done. The work is done by the word "both". AYUSH practitioners, trained in ayurveda, unani, siddha and homoeopathy, enter the ratio as though interchangeable with MBBS-trained doctors, though they do not staff emergency rooms, operating theatres or intensive care.

Same population, same practitioners, same year. Hold the population at whatever the ministry used, and its own two headcounts give four defensible answers:

| Method | Ratio | Against one per thousand |
|---|---|---|
| Allopathic plus AYUSH, 80 per cent available | 1:811 | clears |
| Allopathic plus AYUSH, all available | 1:649 | clears |
| Allopathic only, 80 per cent available | 1:1,251 | fails |
| Allopathic only, all available | 1:1,001 | fails |

Two of those clear the target and two fail it, and nothing distinguishes them except which bodies you agree to count. The last line is the one to sit with. Count only the doctors the word usually means, assume every one of them is working, and India arrives within a rounding error of the standard it has announced it cleared.

Kiran Kumbhar has gone further and questioned whether the target exists at all. Writing for the Center for the Advanced Study of India in April 2026, Kumbhar argues that this part of India's medical education policy rested on "a non-existent World Health Organization (WHO) recommendation", and that the government "formally declared the achievement of that target in 2022, but only through calculations that once again incorporated AYUSH doctors".[^8] The reclassification is not a detail of the current ratio. It is how the target was declared met in the first place.

Meanwhile, rural India had 5,491 community health centres spread across 757 districts, each meant to hold a surgeon, a physician, an obstetrician-gynaecologist and a paediatrician. That is 21,964 specialist posts. At the end of March 2023, 4,413 of them were filled, one in five.[^9] The seats had been growing for a decade by then.

Twelve Gujarat colleges decided a sector total the same way, by being filed differently. One shifts the numerator by reclassifying who counts as a doctor. The other shifts a sector total by reclassifying who counts as government. Neither requires anybody to state a falsehood.

## The statistic that improves

Which brings the argument to the number in the title.

The seat levels at either end of that window are published. The ministry put MBBS seats at 1,28,875 in December 2025, and the National Medical Commission's own annexure gives the additions year by year, totalling 48,563 seats between 2020-21 and 2025-26.[^3][^5] Subtract the one from the other and 2019-20 stood at 80,312. The intermediate years follow from the same annexure, and that middle stretch is a reconstruction rather than a published table, which I say here so that anyone can take it apart.

Now put the two ends against the candidates, on the state's own appeared figures. In 2019, 1,410,755 people sat NEET against 80,312 seats. In 2025, 2,209,318 sat against 128,875 seats. That is 5.69 per cent against 5.83 per cent.

The success rate improved. The direction is real and it is the direction the state would want, and anybody defending the record can say truthfully that a candidate's chance of a seat is better now than it was before.

This is the statistic that improves: a ratio that can rise because more people are admitted, because fewer people apply, or because both happen at once.

Now the same six years in headcount. In 2019, 1,330,443 people sat the exam and did not get a seat. In 2025, 2,080,443 people did.

That second number needs no reconstruction at all. Take the growth in candidates, 798,563 more people sitting the examination in 2025 than in 2019, and subtract every seat the state added over the same years, 48,563 of them. The remainder is 750,000, and it is the number of additional people who sat the examination and were turned away.

Both are true at once. The rate improved by fourteen hundredths of a percentage point, and three quarters of a million more people were turned away. Policy lives in the rate. Politics lives in the headcount.

> A ministry quotes the rate. A movement lives in the headcount.

But the deeper problem is with the rate itself, and it is a problem of what sits in the denominator.

A success rate divides the seats by the people who applied. People decide whether to apply. That decision depends, among much else, on whether they can fund two years of preparation. Anyone who concludes it is not worth attempting does not appear in the denominator, and their absence raises the measured success rate of everyone who remains.

It is worth doing that arithmetic once, not to claim this is what happened, but to see how little of it would be needed. To move the measured rate from 5.69 per cent to 5.83 per cent by discouragement alone, with no seats added at all, the applicant pool would need to be about 2.4 per cent smaller than it would otherwise have been. That is the entire six-year improvement, produced by roughly one candidate in forty-two deciding not to sit.

For scale: registrations fell between 2024 and 2025, from 2,406,079 to 2,276,069.[^2] That is a drop of 5.4 per cent in a single year, more than twice the shrinkage that would account for the whole improvement.

Why registrations fell that year is not established. It is consistent with discouragement and consistent with several other explanations. What can be said without knowing the cause is structural: seats demonstrably did grow, so discouragement cannot be the whole story, but it is an unmeasured quantity that pushes the headline number in the flattering direction, and nobody collects the data that would settle it.

What the reclassification did to the numerator, discouragement does to the denominator, and it does it without anyone deciding anything. That is what the title of this essay means. A ratio that improves is not evidence that access improved. It is evidence that the ratio improved, and the two come apart precisely when the people it measures start giving up.

## The year the denominator was not published

On 3 May 2026 the National Testing Agency conducted NEET (UG) 2026. On 12 May the Central Bureau of Investigation registered a case over an alleged paper leak, for criminal conspiracy, cheating, criminal breach of trust, theft and destruction of evidence, and under the Public Examinations (Prevention of Unfair Means) Act, 2024.[^10] Arrests followed, including of an expert associated with the agency. The examination was cancelled and re-run on 21 June at 5,440 centres in 551 cities.[^11]

The result came on 16 July 2026. The agency's press release is detailed. It reports 11.21 lakh candidates qualified, and breaks that figure down across all seven categories it recognises, each to the person. It reports that 138 candidates scored above 690 out of 720, that 19 scored above 700, that 1,492 scored 650 or more and 90,780 scored 500 or more. It reports that 56.8 per cent of women who appeared qualified, against 55.1 per cent of men.[^12]

There is a smaller thing worth noticing first, because it makes the larger one legible. The agency's 2025 release, the one with the seven-year table, reports how many people appeared in 2024 twice. On page two the figure is 2,333,297. On page three, in the gender and category tables, the same quantity for the same examination is 2,333,162. The gap is 135 people. Nobody is lying and nothing turns on it, and that is the point: this is what ordinary administrative slippage looks like, 135 people wide, in a document that is otherwise exact to the person.[^2]

Now set that beside what the 2026 release does. For the number of people who took the examination it says "close to 20 lakh".

It is exact where it wants to be impressive, and approximate where it would have to show how many were turned away.

That is the whole of it. No registered figure, no appeared figure, no absent figure. One year earlier the same agency published all three for seven consecutive years, disaggregated by gender, category, language and disability status.[^2]

> The state still counts the people it admitted, to the person. It no longer publishes how many it turned away.

The release does give enough to bound it. If 11.21 lakh candidates qualified, and between 55.1 and 56.8 per cent of those who appeared went on to qualify, then between 1.97 and 2.03 million people sat the examination. Nothing published narrows it further. The 2026 success rate therefore cannot be calculated, only estimated, at around 3.2 per cent for a government seat, or roughly one in thirty-two.

No amount of further research closes that gap. A published document reports the numerator exactly and the denominator approximately, in the year the paper leaked, the police opened a case, the examination was re-run nationally and young people went into the streets over it.

## Four responses in seventy-two hours

What the state did next is a matter of record, and it happened fast.

On 25 July 2026 the education minister resigned and the protests were called off. On 26 July a serving cabinet minister took charge of education in addition to his existing portfolios. Also on 26 July the Prime Minister announced a high-powered task force on examination reform under Nandan Nilekani, its other members being a former Director of the Intelligence Bureau, a former Chairman of ISRO, the Director of IIT Madras, a former Secretary of the Department of School Education and Literacy, and a governance and logistics specialist. Its reported remit is the examination lifecycle: structural, administrative and technology-driven reform of the testing agency, including artificial intelligence and blockchain-based verification.[^13]

On 27 July the Public Examinations (Prevention of Unfair Means) Amendment Bill, 2026 was introduced in the Lok Sabha, by the Minister of State for Personnel, Public Grievances and Pensions. The sponsoring ministry is Personnel, not Education.[^14]

The Bill raises imprisonment for using unfair means from three-to-five years to five-to-ten, and the maximum fine from ten lakh rupees to fifty lakh. For service providers the maximum fine goes from one crore to five crore and debarment from four years to eight. For organised crime the minimum sentence rises from five years to seven and the minimum fine from one crore to ten crore.[^15]

Three features of that response stand out.

The statute being amended is the Public Examinations (Prevention of Unfair Means) Act, 2024. It was in force in May 2026. It is the Act under which the 2026 case was registered. The response to a deterrence regime that did not deter is a more severe version of the same regime.

The measures that would change the odds of being caught, rather than the sentence if you are, are the timing measures. A new section requires investigation to be completed within two months and trial within three months of chargesheet. PRS Legislative Research, in its analysis of the Bill, points out that the Bill specifies no consequence if the investigation deadline is missed, and contrasts this with the Lokpal and Lokayuktas Act, 2013 and the Scheduled Castes and Scheduled Tribes (Prevention of Atrocities) Act, 1989, both of which at least require written reasons for delay.[^15] A deadline with no consequence attached is a direction, not a requirement. Statutory time limits in India bind when missing them triggers an automatic consequence that an interested party will enforce, and nothing here does.

The Bill also requires every state and union territory to designate a Court of Session as a Special Fast Track Court. Designate, not establish. It reallocates courtroom capacity that already exists rather than adding any, and the record of that instrument is available: in the fast track special courts already running, PRS records average trial times for one category of case reaching 1,717 days in Delhi.[^15]

Every one of those responses operates on the integrity of the examination. Not one operates on the capacity of the system behind it.

A ministerial resignation, a portfolio reassignment, a task force and a Bill, inside seventy-two hours.

None of them adds a seat. The number that would have to move stayed where the July matrix left it, at 2,111 additional government places for the year.

## The ratio builds the exam

Filling roughly 63,000 government MBBS seats from around two million candidates means selecting at about the 96.8th percentile.

An examination can only separate candidates where its questions are hard enough to spread them out. Ranking the top three per cent of a two-million-person distribution requires items calibrated to the top three per cent, because everything easier than that produces ties. Examiners do not choose that difficulty as an independent matter of policy. The ratio chooses it for them.

And the band the paper must discriminate within is the band where purchased preparation concentrates.

That coached candidates are over-represented at the top of the distribution is not seriously in doubt. By how much is not measured anywhere in India, because the data that would measure it is not collected. Modelled across plausible parameter ranges, the over-representation comes out anywhere from 1.13 times to 7.04 times, which is a spread wide enough that quoting any single figure from it would be dishonest, so none is quoted here.

What the modelling does support is a direction rather than a magnitude. Holding everything else constant and increasing the number of seats reduces coached over-representation, and it does so across every parameter combination tested, because the effect comes from the geometry of the distribution's tail rather than from the choice of parameters. The probability that an uncoached candidate is selected rises with the seat count in every combination tested, including at the most pessimistic assumption about how strongly the coaching industry would respond by expanding. That industry response damps the gain without reversing it, and shows up as cost: total spending rises while spending per admitted student falls.

Every figure in the preceding two paragraphs is modelled, not measured. It gives a direction and no magnitude.

## The price of the attempt

The cost of preparing is where the ratio stops being abstract.

Two years at a residential centre in Kota costs six to ten lakh rupees. Tuition is the smaller part of that, in the region of 70,000 to 1.5 lakh a year; what takes the all-in figure to a reported three to five lakh a year is living there, hostel and mess and a child who has moved to another state to sit in a classroom. Two years of online-only material costs about ten thousand rupees. Tuition for the whole MBBS degree at a government college runs from twenty-five thousand to a lakh.[^16]

Average monthly earnings for a regular salaried worker were 20,702 rupees in 2023-24, about 2.5 lakh a year. That is the best-paid of the three categories the national labour survey counts, and it covers 21.7 per cent of workers. Kota therefore costs such a household between two and a half and four years of its entire income. That is the conservative version. The 58.4 per cent of workers who are self-employed took an average of 13,279 rupees a month, and the 19.8 per cent who are casual labourers averaged 418 rupees a day.[^16] For those households Kota is not an expense that can be spread across a few years of earnings.

The preparation costs more than the education.

A counter-number belongs here, in the argument rather than in a footnote. Online coaching now runs at roughly three to five thousand rupees a year against Kota's one and a half to five lakh. That is a collapse in the price floor of between thirty and a hundred times, and it has already happened. If the cost of preparation were the binding constraint on who competes, that collapse should be widening participation right now. Whether it is doing so cannot be established, because nobody collects the data. It is an open question.

## Why banning it does not work

The instinctive policy response to all of this is to ban the coaching industry, and it has been tried twice at national scale.

Korea prohibited private tutoring outright in July 1980, and the Constitutional Court struck the ban down in April 2000. In between, the ban simply failed. It "was difficult to enforce, and the private tutoring did not stop", and by 1997, with the prohibition still law, more than 70 per cent of elementary pupils and about half of middle and high school students were taking private tuition.[^17]

China's Double Reduction policy of July 2021 banned for-profit academic tutoring for children of compulsory school age. A study built on 566,152 daily reviews of training institutions, covering Chinese counties from October 2020 to March 2022, found that participation in language tutoring fell while participation in sports training rose, and that the change in the number of institutions rather than any change in demand was the main channel. Spending relocated. A two-wave study following 1,402 primary pupils in one Beijing district from September 2021 to June 2023 found that regulating the training institutions raised subjective wellbeing and lowered academic achievement among low-socioeconomic-status students.[^18]

Both cases concern school-subject tutoring for children of compulsory school age, and India's question is post-secondary entrance preparation for adults and near-adults. The analogy is strong and it is not exact.

The decisive objection is about targeting rather than about outcomes. Mass coaching is visible, licensable, taxable and cheap. One-to-one tutoring in a private house is none of those. A ban removes the tier that a low-income candidate can reach and leaves standing the tier that only a wealthy one can afford. Enforcement is easiest exactly where the equity case for it is weakest.

Prohibition raises the price of preparation, and rising prices concentrate access among those who can pay.

## The thing the state already built

Which leaves provision, and here the position is stranger than it first appears, because the components are already built and paid for.

NPTEL, a joint undertaking of the IITs and IISc funded by the Ministry of Education, describes its own holdings on its about page: more than 70,000 hours of video content, transcribed and subtitled; the most subscribed educational channel on YouTube, at 1.3 billion views; and more than 12,000 hours of English transcripts already translated into regional Indian languages.[^19] That last item matters more than it looks, because translation is usually the thing that kills a national content programme, and it has already been paid for.

The mediation layer exists too, and it is not an afterthought. NPTEL Local Chapters run in partnership with around 4,500 colleges. Each has a single point of contact, a faculty member who recruits mentors to keep students active, ensure assignments are submitted and answer questions. Teacher mediation is not an adaptation bolted onto the model. It is the model.

What does not exist is a preparation corpus for the entrance examinations, aimed at teachers rather than at students studying alone: how to sequence two years, problem sets calibrated to the difficulty the paper reaches, worked solutions, current patterns.

The distinction matters because it decides whether the thing can work. India already has an enormous local private tuition sector, and the residential mega-centres are its top layer, not its whole. A corpus aimed at local teachers is not a new distribution channel that has to be built and populated. It supplies an existing network with the one input it cannot produce at its own scale. It also answers the standard objection to online courses, which is that they demand motivation the learner has to supply alone. Nobody is asking this content to supply motivation. It supplies capability to somebody who already supplies structure.

The costs are not close. The content volume required is around 2,792 hours, about four per cent of what NPTEL has already produced. Production at a mid-range estimate comes to roughly 42 crore rupees, once, and that estimate carries low confidence because the per-hour rate is not published anywhere. One new government medical college costs around 325 crore. Indian households spend on the order of 3,866 crore a year on entrance coaching, every year.[^20]

*Figure: The three figures side by side, on a logarithmic scale, so each gridline is ten times the one before it. The corpus bar is dashed because its estimate is the least certain number in this essay: the per-hour production rate is not published anywhere. The other two are the ones that decide the argument.*

Tamil Nadu supplies the reason to prefer content over centres. The state already runs free NEET coaching through district centres staffed by postgraduate teachers. Reported uptake for the 2026 examination was around 12,000 government-school students against roughly 1.5 lakh candidates from the state.[^21] Rooms and teachers bound centre-based provision. Content-based provision has a fixed cost and then a marginal cost near zero.

Publishing it would not dissolve the coaching industry. Content, method and calibrated mock tests are what a corpus and a set of free percentile-ranked mocks replace. Selection into elite batches and brand are not, and they survive. The largest institutes also hold one advantage that is entirely lawful: observing millions of attempts, they infer regularities in how items are constructed, which no local teacher can do. That advantage lasts only while the regularities do. Publish the item-difficulty distributions those institutes currently reconstruct privately, and it evaporates without a single enforcement action.

Prohibition drives the price of preparation up, and advantage concentrates. Provision drives the price of the baseline to zero, and what the wealthy buy becomes a margin above a free floor instead of a margin above nothing. It also requires no enforcement, which is exactly why the two attempts that relied on enforcement failed.

So the policy that fits the record is not prohibition but provision: a free preparation corpus aimed at teachers, and the item-difficulty distributions published alongside it.

## What stands in the way

The obstacle is not money and it is not capacity. Both are demonstrably present.

NEET is officially set on the NCERT syllabus. The state's formal position is that school is sufficient preparation for it. To publish a national preparation corpus is to concede that school is not sufficient, and therefore that the examination has been sorting candidates on preparation they had to buy rather than on schooling the state provided.

The diagnosis exists. The capacity exists. What is missing is the willingness to say out loud what the gap is.

And the case for the corpus is reasoned rather than evidenced. The research suggesting structured materials help weak teachers most comes from school systems, not entrance preparation, so the transfer is untested. Cheaper local coaching might expand the total market rather than substitute for the expensive tier. And Tamil Nadu's scheme is too new to have been evaluated, while NPTEL has never been assessed for this purpose at all.

None of that argues against trying an intervention costing four per cent of one medical college. It is a reason not to promise what it will do.

## The arithmetic of 2040

Set a target and the required rate falls out of it.

Suppose the aim were that one in ten of the people who sit the examination get a government seat by 2040. Applicant growth cannot simply be extrapolated forward, because 7.76 per cent a year compounded to 2040 implies about 29 per cent of an entire birth cohort sitting one examination, which is not credible. Take a constrained projection instead, with growth decaying towards two per cent, giving roughly 4.6 million applicants by 2040. Reaching one in ten of those requires government seats to grow at about 14 per cent a year, every year, from here.

Fourteen per cent on the current base of 63,296 means roughly 8,900 additional government seats in the coming year.

The number added this year, on the National Medical Commission's own matrix, was 2,111.

That is the falsifiable claim, and it resolves quickly. If the 2027 approvals do not come in at more than four times this year's government addition, the one-in-ten path is arithmetically closed, whatever is announced in the meantime.

*Figure: The required path compounds at about fourteen per cent a year from the 63,296 government seats in the 2026-27 matrix. The current pace is that matrix's own figure for seats added this year, held constant. The 2027 checkpoint is the first year the two can be told apart, which is what makes the claim falsifiable rather than rhetorical.*

## Nineteen

Someone close to me once came within a rank or two of one of those nineteen-place disciplines. That is the whole of my stake in this, and it is worth exactly what it sounds like, which is nothing as evidence.

But it is why the handbook reads differently to me than it used to. The number nineteen is not a judgement about how many people in India can design a textile. It is a statement about how many chairs are in a room, and the room was built to a size decided a long time ago by people who are no longer accountable for it.

Everything downstream follows from the size of that room. The examination has to be built to separate people at the ninety-seventh percentile because that is what the room requires. Preparation for an examination built that way costs more than the degree it leads to. The families who can pay for it appear in the statistics, and the ones who conclude it is not worth attempting stop appearing at all, which makes the statistics improve.

The handbook does not editorialise about any of this. It sets the intake beside the fee schedule and the calendar of dates, and prints all three flat.

The handbook prints the number. The system builds the room. The statistic improves.

Product Design, nineteen.

---

### Glossary

| Term | What it means here |
|---|---|
| **NEET (UG)** | National Eligibility cum Entrance Test (Undergraduate). The single national entrance examination for undergraduate medical admission, conducted under Section 14 of the National Medical Commission Act, 2019. |
| **NTA** | National Testing Agency. The body under the Ministry of Education that conducts NEET and publishes its results. |
| **NMC** | National Medical Commission. The statutory regulator of medical education, and the publisher of the seat matrices this essay counts. |
| **MBBS** | Bachelor of Medicine, Bachelor of Surgery. The undergraduate medical degree NEET admits people to. |
| **AYUSH** | Ayurveda, Yoga and naturopathy, Unani, Siddha and Homoeopathy. Registered practitioners of these systems are counted alongside allopathic doctors in the government's doctor-population ratio. |
| **INIs** | Institutes of National Importance. AIIMS, JIPMER and similar institutions, all government-run. The 2026-27 seat matrix excludes them, which is why the government share in it is understated. |
| **Govt-Society** | A management category in the NMC seat matrix, distinct from both Govt. and Society. In 2024-25 it covered twelve colleges run by the Gujarat Medical Education and Research Society. |
| **Lakh, crore** | One hundred thousand, and ten million. |
| **PRS** | PRS Legislative Research, an independent organisation that publishes analyses of bills before Parliament. Its reading of the 2026 Amendment Bill is used here as its own. |
| **NPTEL, SWAYAM** | The National Programme on Technology Enhanced Learning, run by the IITs and IISc, and the national online-course platform its material feeds. |

---

### Sources

[^1]: National Institute of Design, *B.Des Admissions Handbook 2026-27*, section 3.1, "Intake Capacity at individual NIDs", read directly from the published PDF at [admissions.nid.edu](https://admissions.nid.edu/). Ahmedabad intake: Animation Film Design 19, Graphic Design 19, Product Design 19, Textile Design 19, Exhibition Design 13, Film & Video Communication 13, Ceramic & Glass Design 13, Furniture & Interior Design 13. The handbook states that the number of seats for Indian Nationals in the B.Des programme including reserved categories is 128, and that "the decimal number is rounded off to the nearest whole number for deriving available seats in a particular discipline". Admission is to the programme; discipline is allotted after a two-semester Design Foundation on merit and preference. The other campuses each admit 75 (Andhra Pradesh, Haryana, Assam), with Madhya Pradesh at a total intake of 86 including 11 overseas candidates. Application counts are not in the handbook. They are in NID's [63rd Annual Report 2023-24](https://www.nid.edu/public/documents/KJzdZlCsLk.pdf), page 35, on the 2024-25 admissions cycle: "A total of 15839 applications were received for the B.Des. programme. In all, 15269 candidates (including Overseas candidates) appeared for the DAT Prelims exam for 144 seats (including 19 supernumerary seats in Overseas category) in B.Des.; 75 seats at NID Andhra Pradesh, 75 seats at NID Assam, 86 seats at NID Haryana (including 11 supernumerary seats in Overseas category), and 86 seats at NID Madhya Pradesh (including 11 supernumerary seats in Overseas category)." And: "The total number of applicants for the M.Des. programme across 19 disciplines was 6897. In all, 6612 candidates (including Overseas candidates) appeared for a total of 402 seats in the M.Des. programme." That gives 466 B.Des places in all, of which 41 are supernumerary overseas seats, so 425 for Indian nationals against 15,839 applications: 2.7 per cent. An earlier draft said two in a hundred, which was wrong at every denominator, and the funnel figure has been corrected with it.

[^2]: Registered, appeared and absent candidates for NEET (UG) 2019 through 2025, with gender-wise and category-wise breakdowns: National Testing Agency, [*Press Release: NTA Declares the Result/NTA Scores/Rank of NEET (UG) – 2025*, 14 June 2025](https://cdnbbsr.s3waas.gov.in/s37bc1ec1d9c3426357e69acd5bf320061/uploads/2025/06/2025061472.pdf), page 2. Note that this document reports 2,333,297 candidates appeared in 2024 on page 2 and 2,333,162 for the same quantity on page 3. The 2019 figures are independently confirmed by [PIB, 5 June 2019](https://www.pib.gov.in/Pressreleaseshare.aspx?PRID=1573491).

[^3]: The pre-2014 baseline of 51,348 MBBS seats across 387 medical colleges appears across successive Ministry of Health and Family Welfare replies, for example [PIB, "Update on Medical Education"](https://www.pib.gov.in/PressReleaseIframePage.aspx?PRID=1947690). The December 2025 levels used here, 818 medical colleges and 1,28,875 MBBS seats, are from [PIB, "Update on Medical Education in India", 2 December 2025](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2197614&reg=3&lang=1).

[^4]: National Medical Commission, Public Notice of 14 July 2026, No.U-15029/09/2025-UGMEB, enclosing the [*Seat Matrix for AY 2026-27 for UG Course (MBBS) excluding seats of INIs, as on 13.07.2026*](https://www.nmc.org.in/MCIRest/open/getDocument?path=%2FDocuments%2FPublic%2FPortal%2FLatestNews%2FPublicNotice_Merged_seatmatrix_Secretary_covering.pdf). Grand total row: government 441 colleges, 61,185 renewed, 2,111 increased, 63,296 total; private 382 colleges, 65,843 renewed, 7,800 increased, 73,643 total; overall 823 colleges and 136,939 seats. The matrix is stated to remain subject to revision on appeal.

[^5]: Year-on-year MBBS seat increases from AY 2020-21 to AY 2025-26, totalling 48,563: [PIB, "Update on Medical Education", 13 February 2026](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2227421&reg=3&lang=2), Annexure, from a Lok Sabha written reply by the Minister of State for Health and Family Welfare citing the National Medical Commission.

[^6]: National Medical Commission, [*Revised UG Seat Matrix 2024-25 as on 31-03-2025*](https://www.nmc.org.in/wp-content/uploads/2025/04/Revised%20UG%20Seat%20Matrix%202024-25%20on%2031-03-2025.pdf), state-wise list of MBBS colleges including AIIMS and JIPMER. Document totals: 780 colleges, 118,190 seats. Decomposition by the management column, reconciling exactly to those totals: Govt. 416 colleges and 58,099 seats; Govt-Society 12 colleges and 2,075 seats; Trust 267 and 45,616; Society 54 and 8,050; Private 31 and 4,350.

[^7]: [PIB, "Update on Medical Education in India", 2 December 2025](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2197614&reg=3&lang=1), Ministry of Health and Family Welfare: "As of now, there are 13,86,150 registered allopathic doctors. The Ministry of AYUSH has informed that there are 7,51,768 registered practitioners in the AYUSH system of medicine. Assuming that 80% of registered practitioners in both the allopathic and AYUSH systems are available, the doctor-population ratio in the country is estimated to be 1:811." The same release gives medical colleges rising from 387 to 818 and UG seats from 51,348 to 1,28,875 since 2014. The four-way recomputation in the table is derived from those two headcounts, holding the population implied by the ministry's own 1:811 constant, so each row is checkable against the release without needing a separate population figure.

[^8]: Kiran Kumbhar, ["Does India Have Too Many Doctors?"](https://casi.sas.upenn.edu/iit/kiran-kumbhar), Center for the Advanced Study of India, University of Pennsylvania, 13 April 2026. The quoted phrases are Kumbhar's. The piece also traces the reversal in official discourse, from an assessment in 2000 that the ratio had "already exceeded that required by the country" to confident assertions by 2012 that it was "abysmally low", and argues that India may face an absence of shortage rather than a shortage.

[^9]: Specialist availability at rural community health centres, position as at 31 March 2023, from *Health Dynamics of India (Infrastructure and Human Resources) 2022-23*, the Ministry of Health and Family Welfare's annual publication formerly titled *Rural Health Statistics*. Its release is recorded at [PIB, PRID 2053070](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2053070), which gives the infrastructure totals as at that date but not the specialist tables. The figures used here are as reported by [ThePrint, 10 September 2024](https://theprint.in/health/rural-india-has-an-80-shortfall-of-specialist-doctors-mp-gujarat-tamil-nadu-worst-off/2259874/): "only 4,413 specialist doctors of the requisite 21,964 needed in CHCs in March 2023 were available", a shortfall of 17,551 or 79.9 per cent, and "There are 5,491 rural CHCs in 757 districts of the country". The requirement of 21,964 is four specialists per centre. The publication's own PDF is not retrievable: the PIB release points to the documents section of mohfw.gov.in, which has since been rebuilt and no longer serves it at a stable address. An earlier draft carried four per-specialty vacancy percentages taken from secondary summaries; two of them could not be reconciled with any published series and all four have been cut rather than left standing.

[^10]: [PIB, "CBI Registers Case in alleged Paper Leak in NEET-UG 2026", 12 May 2026](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2260410&reg=48&lang=2), recording that NEET (UG) 2026 was conducted on 3 May 2026 and that the FIR includes offences under the Public Examinations (Prevention of Unfair Means) Act, 2024. On arrests, see [PIB, PRID 2261790](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2261790&reg=48&lang=2) and [PIB, PRID 2261565](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2261565&reg=48&lang=2).

[^11]: [PIB, 27 May 2026](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2266029&reg=3&lang=2), on the re-examination scheduled for 21 June 2026 at over 5,400 centres in 550 cities.

[^12]: National Testing Agency, *Press Release for NEET (UG) – 2026 Results*, 16 July 2026, published via [neet.nta.nic.in public notices](https://neet.nta.nic.in/public-notices/). Reports 11.21 lakh qualified with category-wise and score-distribution detail, and gives the number who appeared as "close to 20 lakh".

[^13]: On the resignation of 25 July 2026 and the calling off of the protests: [Al Jazeera, 25 July 2026](https://www.aljazeera.com/news/2026/7/25/indias-education-minister-resigns-after-weeks-of-cockroach-protests) and [CNN, 25 July 2026](https://edition.cnn.com/2026/07/25/india/india-education-minister-resignation-intl); the resignation letter is reported by [The Week](https://www.theweek.in/news/india/2026/07/25/dharmendra-pradhan-resigns-as-education-minister-hands-letter-to-pm-modi.html). On the successor taking education as an additional charge on 26 July: [Al Jazeera, "Who is India's new education minister Pralhad Joshi?", 26 July 2026](https://www.aljazeera.com/news/2026/7/26/who-is-indias-new-education-minister-pralhad-joshi) and [Deccan Herald](https://www.deccanherald.com/india/pralhad-joshi-takes-charge-as-new-education-minister-holds-review-meetings-4088101). On the task force announced 26 July, its six members and its reported remit: [Outlook](https://www.outlookindia.com/national/pm-narendra-modi-appoints-nandan-nilekani-to-lead-exam-reform-task-force) and [Brut India](https://www.brut.media/in/articles/india/politics/nandan-nilekani-led-exam-reform-task-force-meet-the-6-members-appointed-by-pm-modi). The constitution of the task force is confirmed by the Prime Minister's Office: ["PM announces constitution of high-powered task force on examination reforms", 26 July 2026](https://www.pmindia.gov.in/en/news_updates/pm-announces-constitution-of-high-powered-task-force-on-examination-reforms/). That release carries the announcement only. It gives no terms of reference and no member list in text; both reach the press through an image attached to the Prime Minister's post on X. No government order has been published as at 29 July 2026, which is why the essay says "reported remit". Replace with the order if one appears.

[^14]: [PIB, "Union Minister Dr. Jitendra Singh Introduces the Public Examinations (Prevention of Unfair Means) Amendment Bill, 2026 in Lok Sabha", 27 July 2026](https://www.pib.gov.in/PressReleasePage.aspx?PRID=2290080&reg=48&lang=2).

[^15]: PRS Legislative Research, [*The Public Examinations (Prevention of Unfair Means) Amendment Bill, 2026*](https://prsindia.org/billtrack/the-public-examinations-prevention-of-unfair-means-amendment-bill-2026). The penalty comparison, the requirement that states designate Courts of Session as Special Fast Track Courts, the observation that the Bill specifies no consequence if the two-month investigation deadline is missed together with the Lokpal Act and SC/ST Act comparators, and the fast track special court statistics are all PRS's, and are used here as PRS's analysis.

[^16]: Earnings and shares are read directly from the National Sample Survey Office's [*Periodic Labour Force Survey Annual Report, July 2023 – June 2024*](https://mospi.gov.in/sites/default/files/publication_reports/AnnualReport_PLFS2023-24L2.pdf). Statement 6 gives the percentage distribution of workers in usual status by status in employment, all-India, rural plus urban, persons: self-employed 58.4, regular wage/salary 21.7, casual labour 19.8. Statement 10 gives average monthly wage or salary earnings of regular wage/salaried employees for the four quarters of 2023-24 as 20,095, 20,575, 21,036 and 21,103 rupees; their mean is 20,702. Statement 12 gives average gross earnings over the last thirty days from self-employment as 12,685, 13,044, 13,487 and 13,900, a mean of 13,279. Statement 11 gives average daily earnings of casual labour in works other than public works as 404, 416, 419 and 433, a mean of 418. An earlier draft carried a median of roughly 9,000 rupees a month across all workers; PLFS publishes averages rather than medians and the figure could not be traced to any published table, so it has been replaced with the survey's own category averages. Coaching tuition is read from [Resonance's published fee structure for the 2026-27 NEET courses](https://www.resonance.ac.in/courses/fee-structure-neet.aspx), stated to apply at its Kota study centre and to include GST. Its one-year classroom programme has a total course fee of 70,000 rupees if paid lump sum for an August 2025 admission, rising by month of joining to 110,000 from April 2026; the two-year Class XI plus XII programme is 60,000 in admission, infrastructure and study-material fee plus 140,000 tuition, 200,000 in all. **[VERIFY]** The living component is not a published figure. The all-in three-to-five-lakh a year, the six-to-ten-lakh two-year total and the government MBBS tuition range are from reporting rather than from a fee schedule.

[^17]: On the prohibition and the judgment: Sang Hoon Bae and Kee Ho Choi, ["The Cause of Institutionalized Private Tutoring in Korea: Defective Public Schooling or a Universal Desire for Family Reproduction?"](https://files.eric.ed.gov/fulltext/EJ1413565.pdf), *ECNU Review of Education*, 2024. It records that on 30 July 1980 Korea's National Security Emergency Council announced a ban on private tutoring under the Chun Doo-hwan administration, banning all tutoring outside schools, and that on 17 April 2000 the Constitutional Court held Article 3 and Article 22(1) of the Act on the Establishment and Management of Hagwons unconstitutional. On participation: Sunwoong Kim, ["Private Tutoring and Demand for Education in South Korea"](https://www.journals.uchicago.edu/doi/full/10.1086/648186), *Economic Development and Cultural Change* 58(2), which gives 1980 participation of 13 per cent at elementary, 15 at middle and 26 at high school, rising by 1997 to 70 per cent at elementary and 50 per cent at middle and high school. Kim's paper is read directly at [rrojasdatabank](http://www.rrojasdatabank.info/devstate/southkorea1.pdf): "In 1997, more than 70 percent of elementary school students and about half of middle and high school students are reported to take private tutoring", and, on the prohibition, that "such a ban was difficult to enforce, and the private tutoring did not stop" even after "the apparent failure of the outright ban on private tutoring in 1980". An earlier draft of this essay also gave 1980 participation figures of 13, 15 and 26 per cent by school level. Those came from a search summary attributing them to Kim; they are not in Kim's paper, and they have been removed rather than carried on an attribution that does not hold. Note also that Kim dates the unconstitutionality finding to 1999 while Bae and Choi give 17 April 2000 for the Constitutional Court judgment; the latter is followed here.

[^18]: ["The impact of China's 'Double Reduction' policy on primary school students' subjective well-being and academic achievement"](https://www.sciencedirect.com/science/article/abs/pii/S0738059325001191), *International Journal of Educational Development*. A two-wave longitudinal survey of 1,402 fourth- and fifth-grade students in a Beijing district, September 2021 to June 2023, finding that reduced homework and school-based after-school provision raised subjective wellbeing without cost to attainment, while regulating private training institutions raised wellbeing and decreased academic achievement among low-SES students. On the sector effects, Nan Zhao, Chunyan Shi and Chenyang Wang, ["Effects of the 'double reduction' policy on the commercial tutoring sector in China"](https://www.sciencedirect.com/science/article/abs/pii/S0738059324000117), same journal, volume 105, March 2024, article 102989. Its stated highlights are that the policy "decreased students' language after-school training but increased sports-related ones", that "the supply of training institutions significantly influenced students' decisions for after-school training", that it "reduced anxiety and increased parental focus on family education", and that "students in advanced industrial cities and areas with numerous training institutions were more sensitive to policy changes". The dataset is described in the paper as "566,152 daily reviews of after-school training institutions on Dianping from October 2020 to March 2022". An earlier draft said the Chinese ban pushed wealthier families into unregulated one-to-one tutoring. Neither paper finds that, and the sentence has been replaced with what the sector paper does find. The abstracts, highlights and introductions are readable; both full texts sit behind a subscription.

[^19]: [NPTEL, "About Us"](https://nptel.ac.in/aboutus), read 28 July 2026, which states verbatim: "More than 70000 hours of video content, transcribed and subtitled", "YouTube channel for NPTEL – most subscribed educational channel, 1.3 billion views and 40+ lakhs subscribers", "Translation of more than 12000 hrs of English transcripts in regional Indian languages", and that NPTEL "offers close to 600+ courses for certification every semester in about 22 disciplines". On the mediation layer, [NPTEL Local Chapter](https://nptel.ac.in/LocalChapter): NPTEL "partners with around 4500+ colleges", each with a faculty Single Point of Contact. SWAYAM's cumulative enrolment and certification totals, and the 40 per cent credit-transfer provision, are deliberately not asserted anywhere in the essay. swayam.gov.in was rebuilt and no longer publishes those counters on any page reachable from its homepage, so no figure for them has been used.

[^20]: **[LOW CONFIDENCE]** Four figures carry this label permanently: content volume of roughly 2,792 hours, production cost of around 42 crore rupees, per-college capital cost of around 325 crore, and annual household coaching spend of around 3,866 crore. Each is an order-of-magnitude estimate rather than a measured quantity, the body text says so where it uses them, and the label stays whatever later sourcing turns up.

[^21]: [News Today (Chennai), "TN launches free NEET coaching for govt school students", 11 April 2026](https://newstodaynet.com/2026/04/11/tn-launches-free-neet-coaching-for-govt-school-students/). Free coaching centres in every district, taught by postgraduate teachers, running daily until 2 May 2026, the day before the originally scheduled examination. Around 12,000 government-school students registered for the support, against roughly 1.5 lakh candidates from Tamil Nadu.