Though the concept of poverty is old, the idea of measuring it is relatively recent. For most of history, the majority of people were poor. Poverty measurement is only a practical concern when a society becomes rich enough to decide that something ought to be done about its poorest members — which in turn requires the means to locate and identify them. 

Measurement need not be the only approach to doing so. Indeed, before such measurement became feasible, state support for the poor in Victorian England was allocated using an “ordeal” mechanism. Charles Dickens, writing in 1856, described the scene at a typical London workhouse, an institution providing accommodation and employment to the desperately poor: ¹

“Crouched against the wall of the Workhouse, in the dark street, on the muddy pavement-stones, with the rain raining upon them, were five bundles of rags. They were motionless, and had no resemblance to the human form. Five great bee-hives, covered with rags — five dead bodies taken out of graves, tied neck-and-heels, and covered with rags — would have looked like these five bundles upon which the rain rained down in the public street.”

With limited resources and no political appetite for generous and broad-based support, the cruelty was in fact the point of the system: Only a sufficiently miserable experience of relief would encourage the poor to work, overcoming their supposed laziness and leaving only those who genuinely had no option but to appeal to charity. Such a system required no measurement. Though there is more than a touch of the workhouse in modern-day relief schemes, ² it was partly in response to the social outrage at the system embodied by Dickens that measurement of poverty was first seriously attempted.

Not long after Dickens wrote of those five rag-clad figures in the workhouse, social reformers Charles Booth and Seebohm Rowntree adopted similar methods for replacing the ordeal mechanism with a “scientific” means of defining and measuring poverty. This was no small undertaking: Booth’s Inquiry into Life and Labour of the People in London spanned 17 years and produced a detailed map of much of the city, with each street carefully surveyed and classified with respect to its material condition. ³ In an 1887 speech at the Royal Statistical Society, Booth suggested those living on less than 18 shillings per week were the “very poor”; those between 18 and 21 shillings were “poor,” and those earning more were “above the line of poverty.” 

This is the first clearly articulated conception of a poverty line recorded. Yet in Booth’s analysis, he never quite manages to distance himself fully from the Victorian moralizing about the poor — his color-coded maps included entire streets classified as the “Lowest class; vicious, semi-criminal.” 

In Poverty: A Study of Town Life, published in 1901, ⁴ Rowntree went a step further than Booth. He defined a level of income required to achieve “physical efficiency” — a concept we would now recognize as their basic nutritional requirements. He set this level at 18 shillings and 10 pence for a family of two adults and two children (adjusting it up or down based on family size). Any family living on less than Rowntree’s “primary poverty line” would be financially incapable of meeting the minimum needs to maintain their health. 

To forestall opposition to his measure, Rowntree defined these standards so parsimoniously that, as he repeatedly reminds us, his food budget is even less generous than that used in the York workhouse. 

Virtually nothing but the barest necessities for continuing existence were included in his calculation. In his own words:

“A family living upon the scale allowed for in this estimate must never spend a penny on railway fare or omnibus. They must never go into the country unless they walk. They must never purchase a halfpenny newspaper or spend a penny to buy a ticket for a popular concert. They must write no letters to absent children, for they cannot afford to pay the postage . . . Should a child fall ill, it must be attended by the parish doctor; should it die it must be buried by the parish. Finally, the wage earner must never be absent from his work for a single day.” ⁵

Rowntree was choosing such a penurious poverty line so that no reasonable person could oppose some form of relief for those living below it, an approach we will see repeated time and again as the poverty line measurement methods spread around the world.

Rowntree recognized that his line risked being too low: That is, even someone living just above it might be unable to maintain their basic physical health. To mitigate this risk, he did two things. First, he undertook a sensitivity analysis to investigate the extent of poverty at slightly higher “primary” poverty lines. Then, he estimated “secondary poverty” — much less scientifically. His secondary line can only be described as an “eye test.” If a family looked immiserated, they were counted as poor. The results of this approach are striking: 

Poverty rates are highly sensitive to the choice of the specific primary line; and the eye-test method more than doubled the number of households counted as poor. From his own writing, it seems that Rowntree saw secondary poverty as the better measure, but subsequent accounts mainly focus on his more restrictive but “scientific” primary line.

From these very early days of poverty measurement, we see three weaknesses of the poverty line approach that have never been satisfactorily resolved. First, defining poverty as a financial phenomenon narrowed the direct focus of reforms from the broad considerations of what it is to be a fulfilled human to what can be bought or commanded from the market and state. Poverty is almost always defined in such terms now, but debates in philosophy about the idea of poverty and what we owe the poor have never wholly reconciled with this concession to tractability. ⁶

The second weakness of the approach is that it required (and still requires) taking a stand on what we mean by “sufficiency.” Ever since Rowntree, setting a poverty line has meant deciding what a sufficiently “healthy” human life is. But how do we determine this? Is sufficiency simply meeting minimum nutritional requirements? Does it imply separate poverty lines for men and women? How accurately can we really know the minimum requirements for different nutrients, alone and in combination? And what does it mean to be sufficiently healthy? 

Rowntree’s two lines reflected two different ideas of sufficiency. One was a cold mathematical sufficiency, born of nutritional values and costs; the other was what we might describe as “behavioral sufficiency,” accepting that people will use their finances to satisfy a range of needs. Neither is obviously more right than the other. And neither could really be said to get anywhere close to a “good life.” Rowntree’s attempts to ameliorate this problem by setting two lines and using sensitivity analysis are copied to this day, as we will see; they have never been much more successful than his. 

This brings us to the third weakness of Rowntree’s approach: In defining an absolute line of poverty based on physical efficiency, it abstracts from any relation to the society in which poverty is defined against. In practice, a line above which people will reliably maintain their physical needs will involve an allowance for their social and spiritual needs, which can only be defined relative to social norms and their costs. ⁷ A relative poverty line achieves this, but at the cost of comparability; nevertheless most developed countries now use one.

Despite these weaknesses, variants of Rowntree’s approach have spread around the world. In the U.S., in the 1960s, Lyndon Johnson launched his “War on Poverty,” spurring a bureaucratic need to count the poor. His newly created Office of Economic Opportunity adopted a line developed by economist Mollie Orshansky. Much like Rowntree, she defined an “absolute” poverty line against nutritional requirements and the cost of food and basic needs, as well as a higher “behavioral sufficiency” line. Also like Rowntree, her higher line was forgotten and the lower line adopted. ⁸ Around the same time, newly independent India undertook an almost identical exercise, setting a similarly restrictive line based on bare nutritional need. Their concession to variability was to set a second, higher urban poverty line, but this simply reflected a higher price for foodstuffs in the cities. ⁹

Still, the creation of national poverty lines did not imply a concept of global poverty, and still less that it could be measured. Making that leap required huge changes in the global economic environment. Enough countries needed to be free enough of want at home to turn their gaze abroad, while enough countries needed to contain sufficient numbers of poor to warrant counting. 

But it would take decades for the data gathering and conceptualization of poverty to catch up. Much of the story happens in a few rooms of the World Bank building in the 1980s and 1990s.

From today’s vantage point, it seems faintly absurd that the World Bank existed for around 45 years without a clear idea of what the extent of poverty around the world was, or how it could be counted. ¹⁰ Underlying this ignorance were two forces. The first was institutional: For most of its history, the bank’s objective was the economic development of poor countries rather than the welfare of poor individuals directly. ¹¹

The operating assumption often seemed to be that if the World Bank took care of economic development, poverty would follow. In the long run this may be true, but the experience of the 1980s, during the punishing adjustments that heavily indebted poor countries had to undertake, showed that in the short run, not all policies designed to spur long-term growth are equal with respect to the welfare of the poor. The bank decided that its flagship World Development Report for 1990 should be concerned with global poverty, creating a new institutional dynamic.

The second factor was technical: Calculating global poverty required conceptual and empirical advances. As we’ve seen, defining “sufficiency” requires a huge number of judgment calls, and no one at the bank had the appetite to define a sufficient life in a global context. 

The puzzle was solved by the late Martin Ravallion, an Australian economist who would become more closely associated with poverty measurement than anyone else in the world. Ravallion, together with his colleagues Gaurav Datt, Dominique van de Walle, and Elaine Chan, realized that rather than make their own judgment on what constituted sufficient living, they could instead rely on the judgment of poor countries themselves. They would simply take an average of the poorest countries in the world and declare this to be the global minimum of human sufficiency. ¹²

Implementing this simple idea is surprisingly technically complicated. Different countries define their poverty lines in different currencies, which makes sense for a national measure. But to compare them to each other, you need to do more than simply convert them all into a common currency: You need to account for the fact that a dollar in one country buys more than a dollar in the next country. And you need to do the same to incomes across countries so you can judge a Sri Lankan family’s income against one from Liberia and assess both against a common poverty line, so two families with the same measured income would be able to afford the same standard of living. 

Adjusting currencies for so-called “purchasing power parity” was an enormous undertaking: It involved data collection of prices and expenditures at granular level across multiple countries, and painstaking analysis of the resulting data to produce comparable figures. Ravallion’s method was not simply a new idea; it was an idea that would have been impossible to implement at any earlier point in history. Without the data and analysis to conduct comparisons for a sufficiently large number of countries, “global” poverty would have been a valid measure for only the handful the data was available for.

He and his colleagues were lucky in another way, too. After all their data collection and analysis, they found that the average poverty line amounted to roughly $1 a day. The enormous intellectual undertaking that allowed for global poverty to finally be defined in 1990 gave rise to a simple, compelling advocacy tool that would shape international development discourse for the next decade.

It’s hard to overstate how comprehensively the new global poverty metric captured the imagination of the international development movement. I was in the early years of secondary school when Ravallion came up with the dollar-a-day poverty line; by the time I took my first course in international development at university, it was already ubiquitous. A Google Ngram (cropped at 2008, the year the poverty line was updated) gives an indication of the pace with which Ravallion’s poverty line became the default language of international development. Even today, popular books will talk about “dollar a day” poverty or people living on “less than a dollar day.” ¹³

The dollar-a-day metric of global poverty spread so quickly for three reasons. Most obviously, it was a direct measure of what so much international development work was purported to pursue — the material well-being of the poorest people on the globe. Secondly, it was catchy. Anyone could remember the poverty line, and everyone in the rich countries that provided most aid had an instinctive sense that it was very little money indeed. ¹⁴ That, coupled with the startling finding that more than a billion people around the world lived on so little, made it an incredibly effective advocacy tool. 

And because it was needed, memorable, and compelling, it was rapidly adopted as the central target in international development discourse. By 1996, just six years after the first global poverty estimates were released, the OECD’s Development Assistance Committee used it for the headline target of their flagship report on international development, “Shaping the 21st Century: The Contribution of Development Co-operation.” These goals would eventually become the United Nations’ Millennium Development Goals, still headlined by MDG 1, as it was called: to halve extreme poverty.

Indeed, looking at global aid flows, it’s not the establishment of the poverty line that marks an inflection point (in fact, global aid flows were falling for most of the decade after the poverty line was established) but the adoption of the MDGs — after which aid increases at a rapid rate. Still, the act of conceptualizing poverty as a global phenomenon was necessary for the MDGs. Until a single measure of global poverty was proposed and implemented, donors might have had the objective of minimizing poverty, but would have been unable to assess their progress. The dollar-a-day line, and its associated empirical trappings, went beyond measuring what we already did — it opened up new possibilities in how development could be done. 

Since this point, global poverty has been — at least in rhetoric — the overriding focus of international development. In the U.K., the poverty focus of foreign aid is written into law. The very first words of the International Development Act of 2002 are “The Minister may provide any person or body with development assistance if the Minister is satisfied that the provision of the assistance is likely to contribute to a reduction in poverty.” ¹⁵ The World Bank has adopted “twin goals” of eradicating extreme poverty and boosting shared prosperity. The Sustainable Development Goals (the successor to the MDGs) again open with poverty, calling for its end by 2030.

An act of measurement had come to dominate the language of international development. But all acts of measurements involve artifice; and some of it may be damaging.

The move to a single global poverty line has preserved the weaknesses of poverty measurement that have been present since Rowntree’s time. And in taking an imperfect poverty line and elevating it to a central objective of the international development movement, it also introduces new problems: distortions in how aid and development effort is directed and the activities it funds. 

Unlike the national poverty lines in the U.K., the U.S., or India, the global poverty line is not set to any objective or locally specific definition of a sufficient livelihood. Instead, it represents an average from a selection of the poorest countries in the world. This means that a family living at the global poverty line in a relatively richer country may actually be less able to afford basic necessities. In declining to make a direct value judgment about sufficiency, the World Bank imposed a harsher criteria for “true” poverty than many poor countries would have chosen for themselves. ¹⁶

Ravallion and his colleagues knew they were setting an incredibly strict definition of what constituted poverty and, by extension, an incredibly low bar for what constituted escape from poverty. They never intended that policy be made with respect to only one poverty line; indeed they developed several, each of which reflected a level of material deprivation we should not tolerate. In the most recent updates of the global poverty lines, the three poverty lines were set at (in purchasing power parity dollars) $2.15, $3.65, and $6.85. That data remains widely available but largely ignored.

The danger of setting too low a primary poverty line is that “escape” from poverty may not actually imply escape from extreme levels of deprivation or want. Setting the bar too low makes it difficult to see how aid has impacted less extreme kinds of poverty. 

Indeed, the data confirms that the extreme poverty line is a noisy indicator of poverty at higher levels. While countries with widespread extreme poverty also tend to have higher levels of moderate poverty, exit from extreme poverty does not necessarily imply exit from acute material deprivation. There are many countries with low levels of extreme poverty but very high levels of poverty at higher lines. And the national lines, set with respect to a specific context, show a similar clear correlation with substantial variation. The extreme poverty line clearly captures some level of absolute deprivation. but it cannot tell us much else.

None of this tells us if it’s wrong, though. One way of getting a sense of whether extreme poverty is just too high a bar is to spend a few minutes on Dollar Street, Gapminder’s visualization of what poverty really looks like through photo essays taken of families around the world. Everyone’s mileage will vary, but it’s hard to justify a poverty line as low as the one we’ve chosen when you see what life looks like just above it. Take Family 72, in Liberia. They live around the $3.65 line, ¹⁷ in a corrugated metal shack with a plastic tub used both as a kitchen sink and bathtub. If we applied Rowntree’s eye-test method of secondary poverty to Family 72, it’s hard to see how they would clear the bar. If you agree, then extreme poverty is too extreme. ¹⁸

As a result, there are several countries in which extreme poverty is low or even almost eradicated, but where a substantial portion of the population live in what any eye test would describe as poverty. These countries will likely get less aid, and on less generous terms. Efforts to work on their problems will be discouraged by their “invisibility” to progress on the first MDG and SDG (of which more below). 

But how bad is this outcome? Aid is limited; assuming it does some good anywhere it is spent, does it not make sense to spend it in the places with the most extreme deprivation? If we believe that there is diminishing marginal utility of income (and thus going from $0.10 to $1.00 implies a bigger gain to human welfare than going from $1.00 to $1.99), a too-low bar is better than a well-set one, if there is not enough money to go around. ¹⁹

That’s a reasonable position to take, and indeed, I’ve taken it elsewhere. But there’s a weakness to the argument. Budgets can be fungible depending on the scale of a problem. When a challenge is great enough, ways of raising money can be found — as the success of the original dollar-a-day line shows. Today, some 685 million people are judged to be living in extreme poverty. At the $3.65 line that becomes 1.8 billion, and at the $6.85 line 3.6 billion. ²⁰ Is it plausible that more effort would be directed toward solving global poverty if popular discourse suggested that it still affected billions rather than “merely” hundreds of millions? We can’t know. 

The poverty line also creates a direct perverse incentive for donors. Some people living in poverty are just underneath the poverty line, while others are far below it. A donor or NGO seeking to report success in the fight against extreme poverty is incentivized to focus on the relatively “rich” extremely poor, since it is easier and cheaper to bring them above the line. 

Once a family vaults over the line, they cease to count toward the eradication of poverty, even though practically speaking there is basically no difference between someone just below the poverty line and someone just above it. ²¹

These perverse incentives are predicated on the idea that aid can be precisely targeted with respect to recipient incomes. In reality, this is not the case. Aid projects may be launched in areas where poverty is rampant or may be targeted using some locally designed mechanism, but this mechanism will almost never be quite the same as the extreme poverty line. It also assumes that the only way to support the exit from poverty is to focus directly on those living in poverty. Few development agencies think so, even conceptually. 

Nevertheless, some have advocated against spending on economic development on the grounds that the direct beneficiaries are not those living below the most extreme poverty line. Consider this report of the U.K.’s International Development Committee (IDC), a Parliamentary select committee with the remit to scrutinize the U.K.’s development policy and spending:

“In March 2022 we asked BII (then known as CDC) about its investments’ ability to alleviate poverty, particularly extreme poverty . . . We pressed, however, on how well targeted investment could be to help those in extreme poverty, given that their Impact Framework, used to assess potential investments, used the higher $5.50 a day measure to identify low-income populations.” ²²

The point here is not that BII’s approach is obviously correct, it’s that for many, the fact of targeting people at the higher poverty line is a problem. But if we think that extreme poverty is at least partly a response to an insufficiently flourishing economy, such a rigid focus on the specific people below the line may be counterproductive. It is clear that, even if there is a “country fixed effect,” poverty does decline with higher per capita incomes.

Certainly, we would expect that any sensible aid allocation strategy gives more money, other things being equal, to places where extreme poverty is higher, but there is clearly something wrong when the extreme poverty line exerts so much influence that aid focused on destitute people living on a little more money a day is cause for suspicion.

Yet, despite all of these discontents, it remains difficult to think of a much better alternative to the use of a global poverty line. It has three saving graces. 

First, targeting support on the basis of poverty, no matter how penurious the line, will likely focus it where deprivation is greatest. Though better targeting would use more information about incomes than just a poverty line, it’s not a bad shorthand to use. 

Second, aid is never in practice prioritized on the basis of poverty alone. ²³ A great deal of support still does go to places where many people live above the poverty line, on the basis that national development is required to lift those last people up. 

Finally, the measure of poverty itself has been updated several times. The first couple of updates renewed the poverty line by using the lines of the same 15 countries Ravallion used in 1990, but with new national-level data. The most recent update is more substantial, and uses a line derived from 28 low-income countries. What we actually measure is changing, generally for the better. ²⁴

Ultimately, though conceptually problematic, riddled with practical problems, and for all the perverse incentives it may generate, the development and measurement of a global poverty line has changed, and largely for the better, the business of fighting poverty. It has narrowed — substantially — the conception of what poverty is, and with that narrowing has come real-life consequences. It has made poverty an empirically tractable concept, and a rhetorically powerful device for inspiring and organizing action. And with its measurement has come a change in how we support the poor, for both good and ill.