Has India ended poverty?

So has India ended poverty?

Only at the lowest meaning of the word. At the World Bank's $3-a-day line, which is about ₹58 per person per day in 2021 PPP rupees, India's poverty rate fell from 27.1% in 2011-12 to 5.3% in 2022-23. That is a large, real fall. It means India is no longer the country where extreme consumption poverty described a quarter of the population.

Move the line to $4.20 a day, about ₹82 in the same PPP terms and the lower-middle-income-country benchmark, and the same data gives a different answer. Poverty still falls hard, from 57.7% to 23.9%, but nearly a quarter of the country remains below that fuller floor. In people, that is about 34.2 crore (342 million). The honest answer is plain: the floor rose. The floor is still low.

What does PPP mean?

PPP means purchasing-power parity. It is a way to translate money by what it can buy locally, not by the market exchange rate. In this article, one international dollar is not the same thing as one US dollar converted at today's forex rate. It is a standardised purchasing-power unit.

That is why the World Bank lines should be read as global comparison lines. In India's 2021 consumption PPP, $3 a day is roughly ₹58 a day, $4.20 is roughly ₹82, and $8.30 is roughly ₹162. These are not today's Indian official poverty lines. They are a way to ask how many people fall below comparable global floors, the same per-person lens that makes India look big in total but poor per person.

The line names also changed in 2025 because the World Bank moved from 2017 PPPs to 2021 PPPs after the International Comparison Program released new price data. The old $2.15 extreme-poverty linePoverty lineA threshold used to decide who counts as poor. Change the threshold and the headcount changes.India looks close to ending poverty at a low line, but far from done at a broader line. became $3.00; the old $3.65 lower-middle-income line became $4.20; and the old $6.85 upper-middle-income line became $8.30. That update changes the price-year and benchmark. It is not a claim that every household suddenly became poorer overnight.

Why does the next floor get so crowded?

India turns percentages into populations. A 5.3% extreme-poverty rate at the $3 line is about 7.5 crore people (75 million). At the $4.20 line, the count is about 34.2 crore (342 million). At the $8.30 upper-middle-income line, the World Bank WDI/PIP series implies about 117 crore people (1,170 million) in 2022.

That is why "poverty is gone" is too loose. It may describe the direction of the lowest line, but it does not describe the scale of remaining vulnerability. If a household is above $3 but below $4.20, it is not destitute by the global extreme-poverty floor. It is also not secure in any normal Indian sense.

How far below the line are people?

The World Bank also publishes poverty gaps. This is the depth measure the headcount leaves out. A headcount asks how many people are below a line. A poverty gap asks, on average, how far below that line people are, counting people above the line as zero gap.

At the $3 line, India's poverty gap is down to 0.8% in 2022. That supports the claim that the deepest poverty has become much shallower. At the $4.20 line the gap is 4.4%. At the $8.30 line it is still 31.1%. So the depth story matches the headcount story: the lowest floor has improved dramatically, but the broader security floor is still far away.

Why is rural poverty still the harder half?

At the $4.20 line, rural poverty fell from 64.9% in 2011-12 to 27.7% in 2022-23. Urban poverty fell from 39.7% to 14.3%. Both are big declines. The rural rate is still nearly twice the urban rate.

That matters because rural India is where low productivity, farm risk, casual work and weaker services meet. The poverty line does not prove those causes by itself. It does show where the burden remains heavier.

What if we used a rich-country line?

The highest line used so far is $8.30 a day, the upper-middle-income standard, and even that leaves about 82% of India below it. Push the line up to a rich-country level and the picture is starker. The World Bank's prosperity line sits at $28 a day in 2021 PPP, roughly the floor it treats as a secure, rich-country standard of living, and the economist Max Roser uses a similar $30-a-day thought experiment to show how low the world's poverty lines really are.

By that line, almost all of India is poor, and almost always has been. The share below $28 a day was 99.8% in 2011-12 and 99.7% in 2022-23, and it was essentially 100% through the 1980s and 1990s. The collapse in extreme poverty is real, but it is concentrated at the very bottom of the distribution. Lift the bar to what a rich country calls a normal life, and the line barely moves.

This is the whole argument in one frame. At $3 a day, India transformed. At $4.20, it improved a lot but still leaves a quarter behind. At $8.30, most of India is below. At $28, almost everyone is. "Has India ended poverty?" has as many answers as there are lines, and the rich-country line is the reminder that India is a poor country getting less poor, not a rich one.

India was one of the engines of the great global fall in extreme poverty. About 1.5 billion people worldwide left it behind after 1990, most of them in fast-growing Asian economies. Roser's newer warning is that this progress is now stalling, because the world's remaining extreme poor are increasingly concentrated in economies that have not been growing. India's own path shows the link plainly: as income per head nearly quadrupled, from about $2,300 in 1993 to $8,600 in 2022 in constant PPP dollars, its $3-a-day poverty rate fell from around 48% to 5% (the growth and poverty relationship). The lesson for India is sober: the fall was never automatic. It came from growth, and it can slow.

How does India compare with its neighbours?

India's headline is that extreme poverty at $3 a day is down to 5.3%. Set against its Asian peers, on the same PIP data and the same 2021 PPP lines, that is middling rather than exceptional. At the $3 line in their latest surveys, China is at 0% and Vietnam at 1.6%, while India (5.3%), Indonesia (5.4%) and Bangladesh (5.9%) sit close together. India crushed the deepest poverty at about Bangladesh's pace, and well behind China and Vietnam.

The comparison comes with an honest caveat. Each country surveys on its own calendar, so the lines do not share clean year-by-year points: China has many survey years, Bangladesh ten, and Indonesia's latest reading is 2024 while the others stop at 2022. Read them as trajectories, not as a same-year race. China at 0% means below this very low line, not that poverty has vanished in every sense; its remaining poverty shows up only at higher lines. The deeper question of why some Asian economies pulled away is the subject of our India-versus-Asia divergence series.

What does the broader line show across Asia?

Move to the $4.20 lower-middle-income line and India's position gets worse, not better. Here India's 23.9% is the highest of the five. China is at 0.5%, Vietnam at 4.2%, while Indonesia (19.9%) and Bangladesh (20.5%) sit just below India. The pattern is the same one the national lines show, now in regional company: India did about as well as its neighbours at clearing the floor of extreme poverty, but it carries a much larger group stranded just above it. The cluster between $3 and $4.20 is where India stands out, and not in a good way.

What if the post-2011 fall is overstated?

This is where the Tinbergen paper is useful. Himanshu, Lanjouw and Schirmer argue that a direct 2011-12 to 2022-23 comparison is unsafe because the newer HCES changed the consumption definition, recall periods, field process and treatment of free welfare items. So they estimate poverty through survey-to-survey imputation instead.

Their finding is not that poverty failed to fall. It is that the fall may have slowed sharply. In their PLFS-imputation exercise, Tendulkar-compatible poverty is around 17.5-19.9% in 2022-23, depending on model, and the post-2017 path is fairly flat. That is a very different story from treating the new HCES as directly comparable and reading poverty as nearly gone.

What does HCES say about the consumption floor?

HCES measures MPCE, monthly per-capita consumption expenditure. Take a household's monthly consumption, divide it by the number of members, and you get the per-person figure.

By that measure, average consumption rose. Rural MPCE, without imputation of free welfare items, went from Rs 1,430 in 2011-12 to Rs 4,122 in 2023-24. Urban MPCE went from Rs 2,630 to Rs 6,996. Some of that is inflation. Some is real improvement. Either way, this is the consumption base on which poverty estimates changed.

This is also where the wage-versus-consumption puzzle enters. The PLFS-derived real casual wage series in our data rises from about Rs 187 a day in 2017-18 to Rs 224 in 2023-24, in 2012 prices. That is progress, but not a boom. Rural-wage series are also much flatter than a simple reading of current-rupee HCES growth. So the article should not pretend every data source is saying the same thing with the same force.

What changes when free welfare items are counted?

HCES now reports consumption with and without the imputed value of items received free through welfare programmes. This matters most near the floor. Free foodgrain is real consumption. A household that eats it is better off than a household that does not receive it.

But it changes the reading. With imputation, we are measuring consumption supported by public transfers, not only what households bought from cash income. That is not a trick. It is also not the same as saying household earning power rose by the full amount.

In the HCES series we are using, imputation adds about Rs 125 per person per month to average rural MPCE in 2023-24 and about Rs 82 in urban India. In 2022-23, the comparable gaps are about Rs 87 rural and Rs 62 urban. Those are averages; the effect can matter more for households close to the floor.

Did the real floor actually rise?

Rising rupee consumption is not the same as rising real consumption. To see the floor in constant prices, we take the published bottom-fractile MPCE from the last three surveys, the NSS 68th round in 2011-12 and HCES in 2022-23 and 2023-24, and deflate each to 2023-24 rupees.

In those constant prices, the poorest 5% of rural India went from about Rs 1,034 to Rs 1,677 a month, a real rise of around 62%. Urban India's poorest 5% rose about 76%, from Rs 1,349 to Rs 2,376. In both sectors the floor rose faster than the median, so the bottom gained a little ground on the middle rather than just keeping pace. That is real progress, and it is the strongest version of the good-news story.

Two honest limits keep it grounded. First, the three surveys are not cleanly comparable, because the recall periods and the treatment of free welfare items changed, so part of the measured rise is a methodological lift rather than pure improvement, the same point the comparability section makes. Second, even after rising, Rs 1,677 a month is about Rs 56 a day. The floor rose. The floor is still low.

Why does work quality matter?

Poverty is not only whether a household crossed a line in a survey year. It is also whether it can stay above the line after a bad monsoon, a fever, a job loss or a rent increase. That is where work quality enters.

ILOSTAT's working-poverty series has fallen, but it keeps a hard fact in view: employment is not the same as security. India's labour market has too much casual work, self-employment with low returns, and informality inside salaried work. A poverty article that stops at the headcount misses this fragility.

Methodology: how to read these numbers

This article is built as a set of floors, not one verdict. The World Bank's $3 line measures extreme consumption poverty using international PPP dollars. In India's 2021 household-consumption PPP, that is roughly Rs 58 per person per day. The $4.20 line is the lower-middle-income-country standard, roughly Rs 82 per day. The $8.30 line is the upper-middle-income-country standard, roughly Rs 162 per day. These rupee figures are PPP translations for comparison, not market exchange-rate conversions and not official Indian poverty lines.

The PPP conversion is approximate and rounded. We used India's 2021 household final consumption PPP conversion factor, about Rs 19.47 per international dollar, then multiplied it by each World Bank daily line: $3, $4.20 and $8.30. That gives about Rs 58, Rs 82 and Rs 162 per person per day in 2021 PPP rupees. We show those rupee equivalents only to make the lines legible for Indian readers.

The national poverty numbers come straight from the source, not from a PDF. The $3, $4.20 and $8.30 headcounts and poverty gaps are pulled directly from the World Bank's Poverty and Inequality Platform (PIP), the database the Poverty & Equity Brief and the WDI API are both extracts of. We pin the PIP data version to the 2025-09-30 release in 2021 PPP, the vintage behind the October 2025 India brief, so these numbers stay reproducible: anyone can re-run the same query and get the same figures. We take survey years only, not interpolated or projected years, because this article's whole argument is about survey comparability. Two honest limits follow. First, PIP serves India at the national level only, so the rural and urban splits, the age and education group cuts, and the published headcount counts in crore still come from the October 2025 brief, not from PIP. Second, PIP and the older WDI series can differ by a rounding step because WDI lags PIP; for example, the 2011 poverty gap at $8.30 is 48.0% in this PIP version against 48.1% in the WDI snapshot we previously used. We have moved the national series onto PIP and left the brief-only pieces clearly marked as such.

The World Bank line names changed because the World Bank moved from 2017 PPPs to 2021 PPPs after the International Comparison Program released new price data. The old $2.15 extreme-poverty line became $3.00. The old $3.65 lower-middle-income line became $4.20. The old $6.85 upper-middle-income line became $8.30. This is a price-year and benchmark update. It does not mean every household suddenly became poorer when the line changed.

The people counts are shown in crores because that is easier for Indian readers. We did not change the underlying World Bank million-person counts. For the $3 and $4.20 lines, we converted the World Bank brief's counts from millions into crores by dividing by 10: 75.24 million becomes 7.52 crore, and 342.32 million becomes 34.23 crore. In prose, we round this to 7.5 crore (75 million) and 34.2 crore (342 million).

The $8.30 count is derived, not copied from the World Bank brief. We took the World Bank WDI/PIP headcount rate for India at the $8.30 line, 82.1% in 2022, multiplied it by the World Bank population series for India, and divided by one crore. That gives about 117 crore people, or about 1,170 million. Because this is a derived count, the article treats it as an approximate scale marker, not as a separately published World Bank count in the India brief.

The official Indian poverty-line section uses the Planning Commission's Tendulkar series and the Rangarajan Expert Group report. Tendulkar is India's last adopted official national poverty line, and the official all-India estimate stops at 2011-12. Rangarajan is included because it is the best-known higher committee benchmark for the same 2011-12 survey year, not because it became the official line.

The post-2011 comparability problem is central. HCES 2022-23 and 2023-24 are not mechanically comparable with NSS/CES 2011-12 because questionnaire design, recall periods, survey process, sampling and the treatment of free welfare items changed. The World Bank adjusts old and new consumption surveys for its international estimates, but even then the comparison is not the same as having an unbroken official Indian poverty series.

The unreleased 2017-18 consumption survey is treated as reported evidence, not as an official table. Media reporting on the leaked "Key Indicators" report said real average MPCE fell 3.7% between 2011-12 and 2017-18, with rural MPCE down 8.8% and urban MPCE up 2.0%. The government did not release the survey, citing data-quality concerns. We include the figures because they explain the 11-year break and the political economy of the poverty debate, but we do not use them as a formal article data series.

The Tinbergen paper is used as a methodological counterweight. Himanshu, Lanjouw and Schirmer do not simply compare the old NSS consumption survey with the new HCES. They use survey-to-survey imputation through PLFS data to estimate Tendulkar-compatible poverty after 2011-12. We chart their three reported model paths as research estimates, not official poverty rates. The point is the range of credible disagreement: poverty fell, but the speed of the post-2011 fall is contested.

The HCES MPCE charts use MoSPI's published 2023-24 tables and related HCES artifacts. MPCE means monthly per-capita consumption expenditure: household monthly consumption divided by household members. It is consumption, not income, savings, assets, debt or wealth. A household can have low MPCE but own assets; another can have temporarily high MPCE funded by borrowing. MPCE also cannot show who inside the household actually consumes what.

For welfare imputation, we compare HCES MPCE without imputation and with the imputed value of items received free through welfare programmes. Counting free foodgrain or other free items is not fake: those items are real consumption. But it changes the meaning of the number because measured consumption now includes state-supported consumption, not only what households bought from cash income. In the HCES series used here, imputation adds about Rs 125 per person per month to average rural MPCE in 2023-24 and about Rs 82 in urban India. In 2022-23, the comparable gaps are about Rs 87 rural and Rs 62 urban.

The bottom-fractile welfare-imputation gap is computed from HCES 2023-24 unit data. We isolate the imputed value of items received free through welfare programmes using MoSPI's own imputation-rate book (free foodgrain and free durables), express it as a share of each fractile's consumption, and validate it by checking that the population-weighted average reproduces the published rural and urban imputation gaps of about Rs 125 and Rs 82. The real-floor-by-round series likewise uses published bottom-fractile MPCE from the NSS 68th round and the HCES press note, deflated by CPI to common prices; it is a distributional measure of the real floor, not a poverty headcount, and it inherits the same survey-comparability caveat as the rest of the post-2011 comparison.

The rupee-cutoff simulation is our own sensitivity exercise using HCES 2023-24 unit-level microdata. For each household, we reconstruct monthly consumption from the unit records, apply reference-period scaling to put weekly and annual items onto a monthly basis, divide by household size to get MPCE, and weight by household multiplier times household size. Then we ask what share of people live in households below chosen monthly MPCE cutoffs such as Rs 2,500, Rs 3,000 or Rs 4,000 per person per month.

The simulation is calibrated to published rural and urban HCES means. After reconstructing MPCE from the unit files, we scale rural and urban records so their weighted means match the published 2023-24 non-imputed MPCE values: Rs 4,122 rural and Rs 6,996 urban. This keeps the microdata distribution aligned with the official published averages while preserving the shape of the reconstructed distribution.

The cutoff simulation is not an official poverty estimate. It uses nominal 2023-24 rupees, not PPP dollars. It does not adjust each state cutoff for local prices. It does not choose a normative poverty basket. It is a sensitivity test: if a reader thinks Rs 3,000 per person per month is too low or too high, the chart shows how sharply the headcount changes when that floor moves.

The state and social-group cutoff charts apply the same nominal MPCE threshold to subgroups. The state chart ranks people by the household's reported state. The social-group chart uses the social group of the household head: Scheduled Tribe, Scheduled Caste, OBC or Others. These are descriptive cuts, not causal estimates. They combine region, education, occupation, land, prices, discrimination and household composition into one observed headcount.

The consumption-share and fractile charts are also HCES consumption measures, not income or wealth measures. Consumption inequality is usually narrower than income inequality, and much narrower than wealth inequality. The poorest rural 5% MPCE chart is useful because it shows the floor behind the average, but it should not be read as a full account of deprivation, debt or vulnerability.

The multidimensional poverty section deliberately uses a different concept. NITI's MPI is a deprivation index across health, education and living standards. It is not a rupee poverty line and should not be added to the World Bank or HCES headcounts. Some NITI points are interpolated or projected because exact matching survey years do not exist, so the broad direction is stronger than the precision of any single non-survey-year point.

The NFHS charts are outcome evidence, not poverty rates. Electricity, water, bank accounts, schooling, insurance and nutrition indicators show whether the basic floor of life has improved. Nutrition indicators such as stunting, wasting, underweight and adequate diet are shaped by food, disease, sanitation, care, health systems and intra-household allocation. They belong in a poverty article because poverty is lived as deprivation, but they cannot be converted one-for-one into consumption poverty.

The health-spending chart uses World Bank UHC financial-protection indicators. These measure vulnerability from out-of-pocket health spending: whether health costs create financial hardship or push households below a poverty threshold. They are not HCES poverty headcounts. Their role here is to show why being just above a poverty line is not the same as being secure.

The work section uses ILOSTAT working-poverty indicators and PLFS-derived wage context. Working poverty is modelled internationally and should be paired with Indian labour-market data before making strong job-quality claims. The PLFS-derived real casual wage series we cite rises from about Rs 187 a day in 2017-18 to Rs 224 in 2023-24 in 2012 prices. That adds an important warning: consumption estimates, wage trends and deprivation indicators are not all saying the same thing with the same force.

The source audit also found official measures that are useful context but too easy to misread as direct poverty rates. The World Bank societal poverty headcount is 21.3% in 2022. Its prosperity gap is 5.4, an average shortfall from a $28/day prosperity standard, not a poverty line. World Bank learning poverty is about 56% for end-of-primary-age children in 2017. The World Bank's October 2025 India brief reports its Multidimensional Poverty Measure at 15.5% for 2022-23, while the WDI API series reports 17.7% for 2022. These belong in the evidence base and caveats, but they should not be mixed into the main poverty headcount unless the article explicitly changes the question.

The main assumption of the article is that poverty in India is best read through multiple lenses: international consumption poverty, old official Indian lines, HCES consumption distribution, multidimensional deprivation, nutrition and vulnerability. The main transformation choices are: converting World Bank counts from millions to crores; translating international PPP dollars into approximate 2021 PPP rupees; deriving the $8.30 count from headcount rate times population; calibrating HCES unit-level MPCE to published rural and urban means; and treating all simulated cutoffs as sensitivity tests rather than official poverty lines.

The right conclusion is not that nothing changed. Too much changed for that. The right conclusion is also not that poverty is over. Extreme poverty fell sharply; broader poverty, deprivation and vulnerability remain.