The Chief Statistician is back, in the chair the Registrar left, with a single folder and the look of someone who has been awake as long as you have. “You have the 21:40 folder in your head,” the Statistician says. “This one disagrees with it. I would rather you heard that from me than from a minister at the microphone.”
Folder A has two names on its cover, both of them economists, and the cover sheet says that they are the strongest objection to the line the Labour Secretary’s folder set out at 21:40.
Two routes up
- The objection. Raghuram Rajan and Rohit Lamba argue that India should not repeat the East Asian path of export manufacturing: the world lacks consumers for another China-sized manufacturing powerhouse. Their book bets instead on skilled service exports and domestic demand in education and health; a reviewer for the IMF praised the call for human capital and questioned the case against manufacturing.
- The service lift is real. The Economic Survey puts services growth in 2025-26 at 9.1 percent, up from 7.2, and their share of GDP in the first half of the year at 53.6 percent.
- Where the jobs are. In towns, 62.0 percent of employment is in services.
- The same Survey, on the factory. Advanced manufacturing, it says, “disciplines firms”.
You read the page twice, because it says two things that cannot both be your policy. The 21:40 folder asked why there are so few factory jobs, and what it would take to make more. This folder says the answer may be to go around.
Then you read the Survey’s own box, because the Statistician has marked it with a paperclip.
Other growth drivers—such as infrastructure, capital markets, digital governance, and services exports—can raise incomes, but function best as complements rather than substitutes.
The box is not a case against services. It says what services exports can do, which is raise incomes, and what they cannot, which is stand where the factory stands. The Survey’s reason is about discipline, not arithmetic: exposure to world competition, in its account, leaves little room for regulatory discretion or political access. That is an argument. It is not a measurement, and the Statistician says so.
You put the folder down. There is nothing in it you can test before dawn, and the Statistician has no second page. “I can’t give you a number that ends this,” the Statistician says. “I can give you the numbers that both of them need.”
Folder B is thinner and older-looking. It holds the slow sheet.
The slow sheet
- Reading. In 2024, 23.4 percent of Standard III children in government schools could read a Standard II text. In 2022 it was 16.3 percent; in 2018, 20.9.
- Subtraction. 33.7 percent of Standard III children, in all schools, could do a two-digit subtraction with borrowing, against 25.9 percent in 2022 and 28.2 in 2018.
- Standard V. Government-school reading: 44.8 percent; private: 59.3.
- Human capital index (2020). India 0.49; Vietnam 0.69; China 0.65.
- Mothers with no schooling. Of mothers of children aged 3 to 8, 43 percent in 2014 and 24 percent in 2024.
- Out-of-pocket health spending, as a share of total health spending: 62.6 percent in 2014-15, 39.4 in 2021-22.
- Child stunting. 29.3 percent, against 35.5 in the previous survey; provisional, from the first fact sheet of NFHS-6.
A note in pencil on the cover says what ASER is: a household survey of rural children, conducted one to one, by the same method since 2006. It does not count towns, and the official survey that does reports a faster recovery on standards it says are not comparable.
You do the subtraction before the Statistician can. The 2024 figure leaves 76.6 in every hundred Standard III children in government schools unable to read a Standard II text. That is arithmetic on one survey’s figure. It is also the best reading in government Standard III since the survey began, and still under a quarter. Every state recovered from the 2022 trough; Himachal Pradesh, Uttarakhand, Kerala, Uttar Pradesh, Haryana, Odisha and Maharashtra gained more than ten points.
The slow sheet shows two kinds of line. Mothers’ schooling (2014 to 2024) and out-of-pocket health spending (2014-15 to 2021-22) both moved over many years. Reading is the other kind. The Standard III class of 2024 is eight or nine years old. In 2033 it is seventeen or eighteen, and in the labour market, whether or not anyone taught it to read in between. Your oath year is blank in this book; that date is not.
The Statistician turns the sheet over. On the back, three numbers sit side by side with no caption, and two more below them.
Women's work, three ways
- 32.4 percent. India, 2025, the International Labour Organization’s modelled estimate.
- 41.7 percent. India, 2023-24, the annual labour force survey, up from 23.3 in 2017-18.
- 34.8 percent. India, August 2026, the monthly bulletin, which uses a different reference period and should not be joined to the annual series.
Below the line
- 10.8 percent. Of rural working women, the share in regular wage jobs, in July to September 2025.
- 42 percent. Tamil Nadu’s share of women in registered factories, 2021-22: 6.3 lakh of 14.9 lakh.
“These are not three errors,” the Statistician says, and lays a finger on each. “They are three instruments pointed at the same women.” The first is modelled, the second measured over a year, the third over a week. This is the Thermometer Rule again, and the Statistician does not say its name.
The disagreement is not only about method. The Survey reads the rise from 23.3 to 41.7 percent as inclusion. The International Labour Organization and the Institute for Human Development read much of the recent growth in jobs as self-employment, in which unpaid family workers, mostly women, predominate. The Survey’s own text points the same way: in rural areas 37.5 percent of working women are own-account workers and 34.2 percent are helpers in household enterprises, against 10.8 percent in regular wage work.
So the gauges disagree on how many women work, and nearly agree on how few of them are paid a wage. The 10.8 is the one figure all three instruments and the Survey’s own text can be read against without a dispute.
The last item is not a statistic. Tamil Nadu’s 42 percent is a count of women in registered factories. One reading is that factories hire women where factories are; the evidence on what such jobs do for women points both ways. The Statistician hands you two studies: exposure to Bangladesh’s garment factories led girls to marry and have children later and to stay in school, while Ethiopian factory jobs were quit within months and raised health problems. Two cases point in opposite directions; this book does not draw a rule from two cases.
You come back to the two folders. Folder A wants you to choose a route. Folder B shows that every route needs girls who can read, mothers who went to school, and women who can reach a job and be paid for it. No decision tonight settles services against manufacturing, because the evidence to settle it does not exist yet and cannot before dawn.
You fund what both need. Reading comes first, with an annual, independent, sample-based learning count for Standard III, published by state and district, in the style of ASER, so that you have a number you did not write. Women’s work is counted by the share in regular wage jobs and not by the headline rate. Both start in year one, because the cohort in Standard III will not wait for your second budget; it reaches working age in 2033, after the next election, so none of the credit lands in your term.
On the Triage Sheet two stamps go down: Start the slow things in year one. And: Count the regular-wage share, not headline participation.
Triage sheet: 16 of 40 lines filledBeside the Triage Sheet, the sixth door appears in the margin of the folder, shut.
Door D6 · closedWhat arrives in ten years whatever you do?The Statistician takes both folders and stops at the door. “Everything else you decide tonight,” the Statistician says, “will be judged in this term. This will be judged in the next.” Three hours to the microphone.