Why India Grew Fast Without Becoming China or Vietnam

India is a large economy. It is not a rich one.

The 1991 balance-of-payments crisis forced New Delhi to dismantle much of the Licence Raj. Thirty-five years later the country is still growing fast. It is still poor per person.

The World Bank puts India’s GDP at $3.96 trillion in 2025. The IMF’s April 2026 World Economic Outlook puts 2026 output at $4.15 trillion, sixth in the world behind the United States, China, Germany, Japan and the United Kingdom. That is size. It is not prosperity.

World Bank GDP per person in 2025:

	1990	2025
India	$371	$2,703
China	$318	$13,862
Vietnam	~$95	$5,066
South Korea	—	$36,227

In 1990 India was slightly richer per person than China, and well ahead of Vietnam. By 2025 China’s income per person is about five times India’s. Vietnam has roughly doubled India’s level. South Korea is in another league.

The World Bank still classifies India as a lower-middle-income country (Atlas GNI per person $2,760 in 2025, against an upper-middle threshold of $4,636).

This is an explainer of that gap. Growth happened. The East Asian factory path did not.

What 1991 did — and did not do

In 1991 India was close to running out of foreign exchange. The emergency package cut industrial licensing, opened trade, and let the rupee adjust.

That worked at the macro level. Private firms and foreign capital could operate in ways that were illegal or impossible in the 1970s. Software, telecom, finance, and later e-commerce took the opening and ran.

What 1991 did not do is the rest of the story. It did not move tens of millions of people from low-output farms into labour-intensive factories. It liberalised product markets faster than land, labour, and logistics. It built elite colleges faster than it built the school and clinic that East Asia treated as the foundation of the factory floor.

The result is a dual economy. A globally competitive services layer sits on top of a large informal workforce that never entered mass manufacturing.

The ladder India did not climb

East Asia’s sequence was blunt. Take surplus farm labour. Put it into export factories that need hands more than degrees — garments, footwear, toys, simple electronics. As wages rise, move up the chain.

When manufacturing’s share of output and jobs peaks too early and too low, economists call it premature deindustrialisation. Dani Rodrik has used India as a leading case.

World Bank, manufacturing value added as a share of GDP, 2025:

India	China	Vietnam	South Korea
13.5%	25%	25%	27%

India’s manufacturing share peaked near 18% in 1995. It has since drifted down. That is not the profile of a country still absorbing farm labour into plants. It is closer to a post-industrial share, at a pre-industrial income.

The employment chart is worse. The Periodic Labour Force Survey (PLFS) 2023–24 puts 46.1% of Indian workers in agriculture, up from 44.1% in 2017–18. Manufacturing’s employment share fell from 12.1% to 11.4% over the same period. World Bank modelled estimates for 2025 put agricultural employment at about 42% in India, 22% in China, 25% in Vietnam, and 5% in South Korea.

Agriculture, forestry and fishing still produce only about 16% of India’s GDP (World Bank, 2025). A large workforce generating a small slice of output is the definition of a productivity trap. People are busy. The work does not pay.

Services filled the GDP chart. They did not fill the jobs chart.

Human capital: elite colleges, weak classrooms

East Asia combined markets with a state that delivered basic literacy and health before the export boom. South Korea had near-universal primary literacy before its 1960s takeoff. China drove mass literacy and rural public health before and after 1978.

India built IITs, IIMs and the Indian Institute of Science. It underbuilt the school at the bottom of the pyramid.

UNDP’s 2025 Human Development Report (data year 2023):

	HDI	Rank	Category
India	0.685	130 / 193	Medium
Vietnam	0.766	93	High
China	0.797	78	High
South Korea	0.937	20	Very high

ASER 2024, the Pratham rural household survey of 649,491 children in 605 districts, is the classroom test:

• Only 48.8% of Class 5 children can read a Class 2 text (42.8% in 2022; 50.5% in 2018).
• In government schools the Class 5 figure is 44.8%.
• Only 30.7% of Class 5 children can do division.

Enrolment is not the same as learning. A demographic dividend cannot be cashed by workers who cannot read a shop-floor manual.

Amartya Sen’s older point still stands: India copied East Asia’s markets more readily than East Asia’s basic education state.

Why factories stayed small

Trade opening was only half the reform. The other half is factor markets.

Labour. Rules written for a tiny formal sector taught firms a rational lesson: stay small, stay informal, or substitute machines for workers. The old Industrial Disputes Act threshold — government permission to retrench in plants above a headcount limit — is the textbook case. India later consolidated labour codes. Implementation across states is still the live issue.

Land. Unclear titles, acquisition fights, and court delays make a contiguous industrial plot expensive in time as well as money.

Logistics. For years commentators recycled a 13–14% of GDP figure. The official NCAER–DPIIT study released in September 2025 puts logistics cost at 7.97% of GDP in FY24 (₹24.01 lakh crore), down from 8.84% in FY23. That is real improvement under Gati Shakti and freight corridors. It is still not Singapore. Software crosses a border as a file. A container does not.

These are prices. When it is cheaper to stay small, or to hire a graduate in Bengaluru, than to run a 5,000-worker line in a tier-two town, the line does not get built.

Politics preferred transfers to plants

The constraint is not that Indian officials never read a development textbook. Pranab Bardhan described the bind decades ago: industrialists, rich farmers, and public-sector interests pull the state toward short-term redistribution. Competitive elections reward visible consumption — free power, cash transfers, farm-loan waivers, cheap food.

Those programmes have welfare effects. They are not a substitute for primary schools, city infrastructure, or a plant that can hire the next cohort of eighteen-year-olds.

Two other drags sit beside the subsidy habit. The “inspector raj” survived the death of industrial licensing: tax uncertainty, compliance stacking, and official discretion that punishes a firm the moment it becomes formal. And political energy spent on identity conflict is energy not spent on whether a district can actually run a factory.

None of that makes democracy a development error. It does explain why the politically easy policy is rarely the employment policy.

Two clocks: AI and age

The old model assumed high-skill services would keep adding Indian jobs even if factories did not. Generative AI attacks that assumption at the entry level: code maintenance, basic testing, customer support, routine outsourcing.

NASSCOM’s FY26 strategic review is the cleanest official industry series:

• Tech industry revenue: $315 billion in FY26, up 6.1% from a revised $297 billion in FY25.
• Direct employees: 5.95 million, up 135,000 (about 2.3%).
• FY25 net additions were 133,000.

Revenue is still growing. Headcount is barely growing. That is decoupling, not a rumour.

AISHE 2023–24, the Ministry of Education’s higher-education census, shows more than 800,000 engineering graduates in a single year — Computer Engineering alone produced 262,408. The popular “1.5 million engineers a year” line is inflated. Even the real number is large relative to net IT hiring of about 135,000.

The second clock is age. UN World Population Prospects put India’s median age at 29.8 years in 2025. The working-age bulge lasts into the 2040s. After that the window closes. If formal, productive jobs are not created at a scale of several million a year, the “dividend” becomes a youth-underemployment problem. The warning is the East Asian sequence run in reverse: old before rich.

Female work is part of the same clock. PLFS 2023–24 female LFPR (age 15+) is 41.7%, up from 23.3% in 2017–18. ILO modelled estimates for 2024 put female participation at 59.5% in China, 68.9% in Vietnam, and 56.8% in South Korea. India’s rise is real. So is its composition: about 67% of employed Indian women are self-employed, much of it unpaid family labour in agriculture. Counted work is not the same as paid factory or office work.

“China Plus One” is not an entitlement

Companies looking for a second factory base after China have a name for the search: China Plus One. India has proof of concept.

Bloomberg, citing people familiar with Apple’s production, reported that India assembled about 55 million iPhones in 2025, roughly a quarter of global output, up from 36 million in 2024. That is a genuine industrial event.

The harsh comparison is still with Vietnam and Mexico, not with a slogan.

India still does a lot of final assembly on imported parts, including Chinese parts. Vietnam and Mexico sit inside thicker supplier networks and deeper trade agreements (CPTPP, EVFTA, USMCA). India’s Production-Linked Incentive schemes have leaned toward capital-heavy bets — semiconductors, batteries, champion hardware. Those matter for technology security. They do not, by themselves, hire the village workforce the way apparel, leather, furniture, and food processing can.

Department of Commerce data for FY26 (2025–26) make the dependence concrete:

• India–China goods trade: $151.1 billion. China is India’s largest trading partner.
• Indian exports to China: $19.47 billion.
• Indian imports from China: $131.63 billion.
• Goods deficit: $112.16 billion, a record, up from $99.2 billion in FY25.

That is not a one-year blip. It is the trade face of a thin manufacturing base.

China Plus One is a window. It is not a reservation in India’s name. Buyers will go where a container leaves on time, a tariff does not punish the input, and a plant can scale without a court case over the land.

What would have to change

The list is not exotic. It is the part of the East Asian sequence India deferred.

1. Put employment weight on light manufacturing, not only on prestige hardware.
2. Cut inverted duties and high tariffs on the parts Indian exporters must import.
3. Move public money from untargeted consumption subsidies toward foundational literacy, apprenticeships, and basic health.
4. Make the labour codes real in the states that want factories, and build land banks that can actually deliver a plot.

None of this requires India to become China. It requires India to stop treating a services elite and a four-trillion-dollar headline as the same thing as mass prosperity.

Why this matters for India in the world

A country that is large in GDP and thin in manufacturing is exposed in two directions.

It buys a large share of capital goods, electronics, and industrial inputs from China — $131.6 billion in FY26, about 17% of merchandise imports. That is an economic fact with a border attached. Readers of our LAC explainer already know the political half of that sentence.

It also sells the world a story of inevitability: the next China, the next factory floor, the workforce of the twenty-first century. Investors now test that story against Vietnam’s export machine, against a $112 billion deficit, and against whether Indian IT still hires the way it did in 2015.

Great-power politics will not wait for the demographic window. Neither will the firms choosing a second address after Shenzhen.

What to watch

• Manufacturing’s share of GDP and of jobs, not only semiconductor-plant announcements
• Female labour-force numbers: how much is paid employment, how much is distress work at home
• NASSCOM revenue versus headcount
• The China goods deficit, and the import share of electronics, APIs, and capital equipment
• Tariff and FTA choices on intermediate goods
• State-level land and labour implementation — national slogans will not show up in the export data

Bottom line

India liberalised in 1991 and grew. It did not absorb its people into industry the way China, South Korea, and later Vietnam did. Services made the GDP chart look modern. They left the employment chart looking like a developing country.

That split was manageable while the world wanted Indian back offices and while the youth bulge still had time. Generative AI and the scramble to leave concentrated Chinese supply chains close both cushions.

This is an explainer of structure, not a forecast of collapse. The constraint is not that India cannot grow. It is that headline growth and broad prosperity are no longer the same variable.

This is analysis, not news. Headline figures are from the World Bank World Development Indicators (2025), the IMF World Economic Outlook (April 2026), PLFS 2023–24 (NSO/MoSPI), UNDP Human Development Report 2025, ASER 2024, NASSCOM’s FY26 strategic review, AISHE 2023–24, the Department of Commerce (FY26 India–China trade), NCAER–DPIIT logistics cost (FY24), and UN World Population Prospects 2024.