Skip to content

← Economics

Poverty

Chapter 22 · Development

In this chapter 6 sections
  1. Drawing a line
  2. Where the poor are
  3. Life near the line
  4. Testing what helps
  5. Small steps or big change?
  6. What shrank Nepal’s poverty

Between July 2022 and June 2023, survey teams from the National Statistics Office visited 9,600 households across Nepal and asked them, in great detail, what they had eaten and bought: every kilo of rice, from the shop or from their own field, every bus fare, every school fee. From those answers the office worked out how many Nepalis were poor. It gave two answers. Measured one way, 3.57%. Measured another, 20.27%.

Both are right, and nobody was cheating. The two numbers come from the same households and the same spending. They differ only in where the line was drawn, and seeing why is the first step to understanding poverty.

Drawing a line

A poverty line is a sum of money: the least a person needs in a year to get by. Anyone whose spending falls below it is counted as poor. The hard part is deciding what getting by means.

Nepal’s statisticians start with food. A person needs a certain amount of energy to live and work, and for the mix of ages in an average Nepali household it comes to 2,236 kilocalories a day. Then they work out what a calorie costs in the food that families near the bottom actually buy and grow, and price a year of 2,236 kilocalories a day at that cost. That gives the food line: Rs. 35,029 a person in 2022/23.

Nobody lives on food alone, so the statisticians add an allowance for everything else, clothes, rent, soap, fuel, school, by looking at what families whose food spending sits close to the food line also spend on other things. That came to Rs. 37,879. Together they make Nepal’s poverty line: Rs. 72,908 a person a year, which is about Rs. 200 a day.

It is measured in spending, not income, and for a good reason. A farm family’s income arrives in lumps, at harvest, or when a son abroad sends money, while its spending goes on every day, and much of what it eats it grows itself. Spending, with home-grown food counted at its value, gives a steadier picture of how a family lives. Prices and needs differ too: the cost of basic needs, worked out for each part of the country, came to about Rs. 130,000 a year in the Kathmandu Valley’s towns and about Rs. 47,000 in Madhesh’s villages, and the survey allows for that before deciding who falls below.

Now the two answers. The line used for the previous survey, in 2010/11, was much lower: Rs. 19,261 then, worth about Rs. 42,845 in 2022/23’s prices. Against that old line, 25% of Nepalis were poor in 2010/11, and only 3.6% in 2022/23. So the statistics office raised the line, building it afresh from what Nepali families now eat and buy, a diet with more meat, fruit and nuts, and a larger allowance for other needs. Against the new line, 20.27% were poor. The fall from 25% to 20% that is often quoted isn’t a fall at all, strictly: it measures two different years against two different lines.

A line like this is called absolute, because it is a fixed basket: a person either can afford it or can’t, whatever the neighbors have. The other kind is relative poverty, which asks how far a person is behind everyone else. The European Union, for instance, counts as at risk of poverty anyone whose income is below 60% of the middle income in their own country. Under that kind of line, poverty only falls if the poor gain on the middle, and a country where everyone grew richer at the same pace would have just as much of it as before. That isn’t a flaw. A relative line measures something real: being left so far behind that a person can’t take part in the ordinary life around them, the school trip, the wedding gift, the phone everyone else assumes you have. Nepal’s own lines are absolute.

To compare countries, though, the World Bank needs one line for everyone. Its first, in 1990, was a dollar a day: in six of the poorest countries, the national poverty lines had come to about that. The number has risen with prices since. Since June 2025 the international poverty line has been $3.00 a day, in dollars that buy as much in each country as a dollar did in the United States in 2021. It is the middle value of the national lines of 23 of the poorest countries, which happened to be Burkina Faso’s line. By that line about 838 million people in the world were extremely poor in 2022.

By that line very few Nepalis are: about 2.4% in 2022/23. At the Bank’s higher line for countries like Nepal, $4.20 a day, about 9% are below it, and at the line it uses for richer middle-income countries, $8.30, more than half. The World Bank itself says that for judging one country’s poverty, the country’s own line is the right yardstick. The international line is for comparing.

Where the poor are

Nepal’s 20% is an average, and its provinces are far apart.

% of its people poor% of allthe poor010203040Sudurpaschim3414Karnali277Lumbini2423Madhesh2325Koshi1714Bagmati1313Gandaki125Nepal 20% below the poverty line, 2022/23
Figure 22.1 The share of each province’s people below Nepal’s poverty line in 2022/23 (the bars, amber where above the national rate, which is dashed), and each province’s share of all Nepal’s poor people (the right-hand column). Nepal Living Standards Survey IV.

In Sudurpaschim, in the far west, one person in three is below the line. In Gandaki and Bagmati it is about one in eight. Villages are poorer than towns, 25% against 18%, and the gaps inside a province can be wider still: in the Kathmandu Valley’s towns about 7% are poor, in Bagmati’s villages a quarter, and in Sudurpaschim’s villages two in five.

But the rate isn’t the same as the number. Karnali has a high rate and few people, so it holds only about 7% of all Nepal’s poor. Madhesh, with a rate close to the national one, has more people than any other province and holds a quarter of them, more than any other. A program meant to help the most poor people would look first at Madhesh and Lumbini; one meant to help the places where poverty is most common would look at Sudurpaschim and Karnali. Both are fair aims, and they point to different maps.

The same survey shows who is most likely to be poor. Families that live on daily wages from farm work: 38% of them are below the line. Families whose head never went to school: 32%, against under half a percent where the head has a university degree. Large families: nearly 30% of the people in households of seven or more. Families with small children: more than a third of the people in households with two or more children under seven. Poverty even has a season. It was lowest from October to January, the months of Dashain and Tihar, at about 18%, and highest from June to September, at about 22%.

Money isn’t the only way to count. The National Planning Commission also publishes a multidimensional poverty index, which leaves money aside and checks ten things in three groups: health (children underfed or dying young), schooling, and the way a family lives (its cooking fuel, toilet, water, electricity, house and belongings). A person short of a third or more of them, by their weights, counts as poor. On that measure 17.4% of Nepalis were poor in 2019, down from 30.1% in 2014. Its map isn’t quite the money map: Karnali came out worst, with nearly 40% poor, and Bagmati best, with 7%. Each measure catches people the other misses.

Life near the line

So far poverty has been a number. Two economists at the Massachusetts Institute of Technology, Abhijit Banerjee and Esther Duflo, spent years asking what it is like from inside. In a famous paper in 2007 they gathered household surveys from thirteen countries and looked hard at the people living on about a dollar a day: what they ate, how they worked, how they borrowed.

The poor spend most of what they have on food: in the countryside, between 56% and 78% of everything. Many run several small businesses at once, a little farming, a stall, some day labor, and few of those businesses employ anyone outside the family. Part of the reason is that nobody gives them enough money to start one big enough to keep them busy all day. Part is that spreading the work spreads the risk: if the rain fails or the stall has a bad month, something else may come through.

Spreading risk matters so much because almost nothing else protects them. In six of the seven countries where the surveys asked, fewer than 6% of the extremely poor had any kind of health insurance. When someone falls ill, the family pays the doctor and loses the sick person’s earnings at the same time, and the money comes from selling something or borrowing. Borrowing is costly. In Udaipur, a district of Rajasthan in India, people living on less than a dollar a day paid informal lenders nearly 4% a month. Banerjee and Duflo found that this isn’t mainly because the poor fail to repay. It is because lending small sums to people with no papers and no property is expensive to arrange and expensive to enforce, and the borrower pays for that. Saving is hard too. Few of the poor have a savings account, and a small sum is hard to keep from the day’s needs.

Put together, these make the poor’s economic lives the hardest of all: every margin is thin, so one bad event, a flood, an illness, a wedding to pay for, can push a family under the line. Much about their lives starts to make sense, Banerjee and Duflo wrote, once you see how little access the poor have to markets that work well and to good public services.

That is also how poverty lasts. A family short of money eats less well, keeps a child home to work, can’t wait for a better crop or borrow to buy a buffalo; so it earns little next year too. Nepal’s books call this गरिबीको कुचक्र, the vicious circle of poverty. The causes economics courses list feed it from every side: no work, or work that pays too little; illness; old age, or too many young mouths for the earners; and floods, landslides and droughts, which fall hardest on those with no savings.

The ways out that the same courses list attack the circle from its different sides: growth that makes more and better-paid work; schooling and health care, so that a child’s chances depend less on the family’s purse; a minimum wage; and state benefits, such as allowances for the old and the poorest, paid for by taxes that take a larger share from those who have more. The survey gives the size of the task. If every poor Nepali could be handed exactly what would lift them to the line, and not a rupee more, it would take about Rs. 95 billion a year. On average a poor Nepali’s spending falls about Rs. 16,000 a year short of the line, a little over a fifth of it. No government can aim money that precisely, which is why so much of the argument is about what actually works.

Testing what helps

If poverty is made of many small difficulties, perhaps it can be fought one difficulty at a time, finding out exactly what helps. That was Banerjee and Duflo’s idea, shared with a third economist, Michael Kremer of Harvard. In the mid-1990s Kremer and his colleagues began working with schools in western Kenya. Would children learn more with more textbooks? With free school meals? Comparing schools that already had books with schools that didn’t would prove nothing, since the schools with books were probably better off in other ways. So the schools were chosen by chance: a lottery decided which got the books, which got the meals and which got nothing yet. Because chance did the choosing, the groups started out alike, and any later difference could be put down to what they had been given. It is the method medicine uses to test a new drug, a randomized trial, and Thinking Clearly, for the Curious explains it in A Fair Test.

The answers were often surprising. More textbooks and free meals made little difference to how much children learned. What did make a difference was help aimed at the children who had fallen behind, tested in Mumbai and Vadodara in India; programs built on it later reached millions of Indian children. When Kremer and a colleague offered pills against intestinal worms, three quarters of parents gave them to their children when they were free, and fewer than one in five when they cost less than a dollar. Very poor people, it turned out, are extremely sensitive to price when the benefit lies in the future. In villages where a vaccination camp was reliably open, the share of fully vaccinated children rose from 6% to 18%; when families were also given a bag of lentils for each vaccination, it rose to 39%, and the cost of each vaccination halved, lentils and all, because the clinic’s fixed costs were spread over so many more children. Even microloans, much praised, were tested this way, and helped less than their champions hoped, as Money, for the Curious tells in Moneylenders and Microfinance.

One test went to the heart of the vicious circle. A Bangladeshi organization, BRAC, gave the very poorest families, often families that lived by begging, an asset of their choice, livestock or the makings of a small trade, with a little money to live on until it paid, and training and encouragement, for about 18 months. Trials of the idea in seven countries found the families clearly better off in all but one, and in Bangladesh and India still better off seven and ten years later. To Banerjee, that goes to a big question, and the answer very much suggests that these families weren’t poor because they couldn’t do better, but because they were stuck, without the start that would let them climb out.

In 2019 Banerjee, Duflo and Kremer shared the Nobel prize in economics “for their experimental approach to alleviating global poverty”. By then the research network that Banerjee and Duflo helped found counted more than 400 million people reached by programs its researchers had tested.

Small steps or big change?

Not everyone was convinced, and the most searching critic was another economist of poverty, Angus Deaton of Princeton, who had won the prize himself in 2015 for his work on how families spend and how poverty is measured. (The rules Nepal’s survey follows to add up a family’s spending come partly from guidelines he co-wrote.)

Deaton’s objection comes in two parts. The first is about the method. A trial tells you what happened to the people in that trial, in that place, with that organization running it. It doesn’t tell you it will work elsewhere, or why it worked, and a program that helps a few villages can behave differently when a whole country does it, because prices and people’s behavior change when everyone is doing the same thing. A trial, he argued, has no special power to produce knowledge better than careful study by other means; it is useful as one piece of a long effort to understand why things work, not as a quick verdict on what works.

The second part is about size. The great falls in poverty of the last half-century, above all in China and India, were driven, he wrote, by economic growth and by globalization, not by programs tested one at a time. It is hard, Deaton wrote, to tell any story in which trials had a part in China’s escape from poverty. And development, he argues, isn’t something outsiders can engineer, however well they test their remedies: leaving a country’s own politics out of it is seen as a virtue, he wrote, when it is a vice.

The other side has its answers. Banerjee agrees that trials can’t answer most big questions, such as which kind of economy a country should run, but says the other methods mostly can’t either, except by assertion: comparisons across countries are, in his words, nearly worthless for policy, and there is no second China to compare the first with. He also questions what counts as a big question. Whether the poorest are stuck in a trap they could leave with the right start is big, he says, and trials have gone a long way to answering it. And each side gives ground. Deaton writes that nobody doubts the economy as a whole matters, only that the tools for making it grow are few.

So the disagreement is less about whether growth or programs matter, since both sides think both do, than about where reliable knowledge can be found. Nobody can yet say which kind of knowledge will do more for the next billion poor people.

What shrank Nepal’s poverty

Nepal’s own story fits neither side neatly. In 1995/96 about 42% of Nepalis were below the poverty line of the time; by 2010/11, 25% were below a line set higher, and in 2022/23, by the new line, 20%. On the old line it all but vanished. Yet the economy grew at a modest pace over those years, about 4% a year, sixth among South Asia’s eight countries. Something other than fast growth at home did most of the work.

The World Bank took the change from 1995 to 2010 apart. About half of the fall came from what Nepalis earned at home, and much of that from rising wages: pay grew about 4% a year after prices, and farm wages a little faster than others, partly because so many young men had gone abroad that those who stayed were scarcer. Work off the farm counted for more than farming. Money sent home from abroad counted directly for about a quarter of the fall, a third in the villages, and it helped indirectly too, by changing the economy at home in ways that made new work for families with nobody abroad. Families also changed shape, with more earners for each person to feed. For 2011 to 2023 the Bank puts remittances’ direct share at over 30%, as Chapter 32 tells.

Schooling moved alongside. The share of households with a child aged 6 to 15 out of school fell from 55% to 16% between 1995 and 2011, as schools came closer: families whose children walked more than 20 minutes to a primary school fell from 84% to 33%. The survey of 2022/23 shows how closely poverty and schooling go together, though that alone doesn’t show how much schooling caused the fall.

What has shrunk poverty in Nepal, then, is mostly work and wages, a lot of them earned abroad. That is why it is fragile: a family a little above the line can fall back with one illness, one flood or one job lost abroad.

A poverty line, in the end, is a choice of what nobody should have to live below, and as a country gets richer it raises the line, as Nepal did. By the line it set for 2022/23, one Nepali in five is still poor, most often in the far west, in the villages, in families that live on daily farm wages. Their lives are hard in ways that are easy to miss from outside: thin margins, no insurance, dear credit. Economists have learned to test, one careful step at a time, what helps such families, and they still argue about whether those steps can add up to the kind of change that ends poverty for a whole country.

Found a mistake? Please tell us.