Defining and Measuring Poverty

Poverty is measured more, argued about more, and defined less consistently than anything else in this Part.

And the central disagreement is not technical. It is whether poverty means having too little or having much less than the people around you — and that is a question about what a society owes its members, dressed in the clothes of a measurement decision (see 7.1.2).

Both answers have serious arguments behind them. This lesson sets them out, shows what each produces in practice, and explains why the same country can honestly report that poverty is rising and falling at the same time.

Poverty fell. Everyone got poorer.

A country enters a deep recession. Median household income falls by several per cent in real terms over four years — wages stagnate, hours are cut, and the middle of the distribution moves down.

Its official relative poverty measure is the proportion of people living in households below 60 per cent of the median.

The median fell. So the line fell with it.

Households whose incomes had not changed — pensioners on indexed pensions, families on benefits uprated by prices — found themselves above a line that had come down to meet them. They were no better off in any respect. The threshold moved.

The official figure showed relative poverty falling during one of the worst falls in living standards in a century, and it was reported, accurately, as such.

Meanwhile a second measure — a line fixed at an earlier year's real value and only adjusted for prices — showed no improvement. By that measure, more people were falling below a constant standard of living.

Both figures were correct. Both were published by the same statistical office. They measure different things.

And the paradox runs in the other direction too, which is why neither side can be comfortable.

In a country growing rapidly, the median rises fast, the relative line rises with it, and relative poverty can increase while the incomes of the poorest are rising in real terms. Households that are eating better than they did, in a better house, with electricity they did not have before, are recorded as newly poor — because the society around them moved faster.

Which measure is right?

That question has no technical answer, and anyone who tells you it does is smuggling a value judgement past you. What can be said is exactly what each measures, and what each would tell you — which is the whole of this lesson.

Two conceptions, and each defeats the other on a different case.

The absolute conception: poverty is falling below a fixed standard of what a person needs. Its case against relative measures is the recession paradox above — a measure that can improve while everyone gets poorer is not measuring deprivation.

The relative conception: poverty is having so much less than the society you live in that you cannot participate in its ordinary life. Its case against absolute measures is that any fixed basket becomes absurd over time. A standard set in 1900 would today classify almost nobody as poor in a rich country, and it would say nothing about anyone's actual situation.

Both objections land. And there is a resolution, which came from 8.1.3's framework and is the most useful idea in this lesson.

Sen's formulation: poverty is absolute in the space of capabilities and relative in the space of commodities.

The capability is fixed: to be adequately nourished, to be sheltered, to move about, to appear in public without shame, to take part in the life of the community.

What is required to achieve it is not. Appearing in public without shame required different clothing in eighteenth-century England than it does now. Taking part in a community's life now requires, in most rich countries, a phone and an internet connection — not as luxuries but because school, work, benefits, banking and appointments are administered through them. A child without one cannot do the homework.

So the end is absolute and the means are relative — which explains why a fixed basket goes wrong, why a purely relative line goes wrong, and what a well-built measure should be trying to do.

Where the lines came from

Two founding lines, and both should be known because both are still with us.

The first scientific poverty line was constructed in York at the end of the nineteenth century.

Rowntree defined primary poverty as an income insufficient to obtain the minimum necessary for merely physical efficiency — and he meant it austerely. His budget allowed nothing for a newspaper, nothing for a stamp, nothing for a bus fare, nothing for a child's toy, nothing for a subscription to a union or a chapel, and no expenditure on anything not required to keep a body functioning.

He said explicitly that the standard was harsher than any family could actually live by, and he used it deliberately: if even this stringent test found around one household in ten in primary poverty, the finding could not be dismissed as generous accounting. It was a rhetorical design as much as a measurement decision, and it worked — the study transformed British debate about poverty's causes (see 7.6.1).

The second is the American official line, and its history is a lesson in institutionalised arbitrariness.

In the early 1960s an analyst constructed a line by taking the Department of Agriculture's cheapest emergency food plan — designed for temporary use — and multiplying it by three, because a survey a decade earlier had found that families spent about a third of their income on food.

That line, price-adjusted, is still the official American poverty measure.

The multiplier was never revised, although food's share of household budgets has fallen dramatically since, so a modern equivalent calculation would produce a very different figure. It takes no account of housing costs, which vary enormously by region; of childcare; of medical expenses; or of taxes and non-cash benefits.

A supplementary measure addressing several of these has existed since 2011 and has not replaced it, because a great deal of law and programme eligibility is written against the original.

Which is 7.2.2's lesson in its purest institutional form. A measure adopted for convenience becomes the definition of the thing, then becomes the basis of entitlement, and then cannot be changed because too much depends on it.

The main approaches

Five families of measure, and what each adds.

One — relative income lines. A percentage of the median: 60 per cent in most European statistics, 50 per cent in some international comparisons, applied to equivalised household income (see 8.1.2).

Simple, comparable, published everywhere, and subject to the recession paradox. It is best understood as a measure of income inequality in the lower half of the distribution rather than of deprivation as such — which is a reasonable thing to measure and not what its name suggests.

Two — anchored or absolute lines. A relative line fixed at a base year and uprated only for prices. This measures whether the poorest are getting better off in real terms, and it is the right complement to the relative measure. Publishing both is what a serious statistical office does, and the two together tell you far more than either.

Three — deprivation measures. Rather than asking about income, ask directly whether a household lacks specific items and activities — and this is where the most interesting methodological development happened.

The consensual method determines what counts as a necessity by asking the population. A representative survey asks which items and activities everyone should be able to afford; those endorsed by a majority become the list; and a household is deprived if it lacks them because it cannot afford them rather than by choice. That last distinction is the whole design — it separates going without from choosing without, which an income measure cannot.

Its strengths: it tracks the society's own standards, so it updates as they change; it captures households whose income looks adequate but who are in debt, in insecure housing or without assets; and it has democratic legitimacy that an expert-set line does not.

Its difficulty: what a population endorses as a necessity moves with the economic cycle and with public discourse, so the measure's own standard shifts.

A related approach builds budgets from the bottom — convening groups of ordinary people to specify what a household of a given type needs for a minimum acceptable standard of living, item by item, and costing it. It produces a figure with a defensible provenance, and it is far above the relative line in most countries where it has been done.

Four — multidimensional measures. Counting deprivations across health, education and living standards simultaneously (see 7.2.2's story), and identifying as poor those deprived on a weighted threshold. The great advantage is that it identifies a different and often better-targeted set of households than income does; the difficulty is that the weights are a judgement embedded in an index.

Five — official composite measures, such as the European measure combining low income, severe material deprivation and very low household work intensity into a single at-risk indicator. Useful, and it conceals which component is moving unless you look underneath.

Depth, persistence and who is counted.

A headcount tells you how many are below the line and nothing else. Two societies with identical headcounts can differ enormously in how far below people are.

The poverty gap measures the average shortfall — how much would have to be transferred to lift everyone to the line. And a squared version weights larger shortfalls more heavily, so that helping the very poorest counts for more than helping those just below the line. Which of the three you use determines whether a policy that lifts many people just over the line beats one that transforms the position of the poorest few — and that is a value judgement, made by choosing an index.

And persistence matters more than the annual figure.

Longitudinal data consistently shows that far more people experience poverty over a decade than are in it in any single year. Poverty is, for a large share of those who experience it, an event rather than a state — following job loss, illness, relationship breakdown, or the arrival of a child.

But a minority experience it persistently, and that group has a different composition and different needs from the much larger group that passes through. A measure reporting only the annual headcount conflates the two, and policy designed for one will misfire on the other.

And the units problem from 8.1.2 applies throughout: measuring poverty at household level assumes pooling, which the evidence contradicts. Poverty within apparently non-poor households is invisible by construction, and it falls disproportionately on women and children.

The global poverty line: how it is built and what it is contested on.

The international extreme poverty line is constructed from national poverty lines. It takes the poverty lines that the world's poorest countries themselves use, converts them to a common currency, and takes something like their average. It is periodically revised as the price data underlying the conversion is updated — $1 a day originally, then $1.25, then $1.90, then $2.15, each in the prices of a different base year.

Three things to understand about it.

It is very low. It represents the standard of the poorest countries' own thresholds, not a subsistence or decency standard for the world. The World Bank publishes higher lines alongside it for lower-middle and upper-middle income countries, and the counts at those lines are several times larger and have fallen much less.

The currency conversion is contested, and the criticism is specific. The conversion uses purchasing power parities built from broad consumption baskets covering everything an economy produces (see 8.1.2). The goods poor people actually buy are a narrow and different subset, and their relative prices differ from the general basket's. Critics have argued this makes the conversion systematically inappropriate for measuring poverty, and that the resulting count is not comparable across countries in the way it appears to be.

And the count is extremely sensitive to the line. Because the distribution is dense in that range, small changes in the threshold or the price base move the global count by hundreds of millions — which has happened with each revision.

And the finding that survives all of this criticism, stated with the same evenness: on every line and every methodology, the share of the world's population in extreme poverty fell dramatically over recent decades, driven overwhelmingly by growth in China and subsequently India (see 8.1.2). Nobody disputes the direction. The disputes are about magnitude, comparability, and whether a threshold that low is the right thing to be celebrating a reduction in.

Because "poverty rose" and "poverty fell" are frequently statements about a definition.

Five questions.

Relative or absolute — and if relative, relative to what? A percentage of a median that is itself moving.

Before or after housing costs? In countries with high and unequal housing costs this changes the count and the composition substantially, and both figures are usually published.

Income or deprivation? They identify overlapping but distinctly different populations, and where they disagree the disagreement is informative (see 7.8.3).

Headcount, gap or squared gap? The choice determines which policy looks better.

And annual or persistent? A far larger group touches poverty than is in it at any moment.

Two habits worth building.

When a relative measure moves, check what the median did. In a downturn, a falling relative poverty rate is usually a falling median.

And when someone proposes a "true" definition of poverty, identify the value judgement inside it. An absolute line says a society owes its members a fixed minimum. A relative line says it owes them the ability to participate in a standard that rises with the society. Neither is a fact about the world, and the argument between them is worth having openly rather than through the choice of an indicator.

And Sen's resolution is the one to carry. The capability is absolute; the goods required to achieve it are relative. A society in which children cannot do their homework because the household has no internet connection has a poverty problem, and no basket constructed in 1900 will detect it.

A country in recession can honestly report falling relative poverty while living standards collapse, because the line is a fraction of a median that has fallen. And a fast-growing country can report rising relative poverty while the poorest get better off in real terms. Both figures are correct and they measure different things.

Two conceptions, each defeating the other on a case. Absolute measures cannot be right because any fixed basket becomes absurd over time; relative measures cannot be right because they can improve while everyone gets poorer. Sen's resolution: poverty is absolute in the space of capabilities and relative in the space of commodities — the capability of appearing in public without shame or participating in a community's life is fixed; what it takes to achieve it is not.

Two founding lines still with us. Rowntree's primary poverty was deliberately austere — nothing for a stamp, a bus fare or a newspaper — designed so that finding around one household in ten below it could not be dismissed. And the American official line was the cheapest emergency food plan multiplied by three, on a decade-old estimate of food's budget share, never methodologically revised, and now embedded in so much law that it cannot be changed.

Five families of measure. Relative income lines, best understood as measuring inequality in the lower half. Anchored lines, which show real change and should be published alongside. Deprivation measures, including the consensual method — the population defines the necessities, and lacking an item counts only if it is because you cannot afford it — and budget-standard approaches built by ordinary people item by item. Multidimensional measures, which identify a different and often better-targeted population. And official composites, which conceal which component moved.

Depth and persistence matter more than the headcount. The poverty gap and its squared version weight the very poorest differently, and the choice between them is a value judgement made by selecting an index. And far more people experience poverty over a decade than are in it in any year — an event for most, a state for a minority, with different compositions and different needs.

The global line is built from the poorest countries' own thresholds, is very low, is sensitive to the price conversion whose appropriateness for poor people's baskets is contested, and moves by hundreds of millions when revised. And on every line and method, the share of the world in extreme poverty has fallen dramatically, driven by Asian growth.

Absolute poverty — falling below a fixed real standard, price-adjusted only.

Relative poverty — falling below a threshold defined as a fraction of the contemporary median.

Anchored line — a relative line fixed at a base year and uprated for prices only.

Primary poverty — Rowntree's deliberately stringent physical-efficiency standard.

Consensual method — necessities identified by population survey; deprivation counted only where the item is unaffordable rather than unwanted.

Minimum income standard — a budget built item by item by groups of ordinary people.

Material deprivation index — direct measurement of what a household goes without.

Headcount / poverty gap / squared gap — how many are below the line; the average shortfall; the shortfall weighted to favour the poorest.

Persistent versus transient poverty — a state for a minority, an event for a much larger group.

Before and after housing costs — two published figures with different counts and compositions.

International poverty line — a threshold derived from the poorest countries' national lines, converted with purchasing power parities whose fit to poor households' baskets is contested.

One — find both figures. For your country, locate the relative and the anchored poverty rates for the same years. Where they diverge, work out what the median did.

Two — check the housing costs. Find the before-housing-costs and after-housing-costs poverty rates. The difference is a fact about the housing market appearing in a poverty statistic.

Three — build a necessities list. Write down ten things you think everyone in your society should be able to afford. Then ask three people to do the same, and compare — you have just run a small consensual survey, and the disagreements are the interesting part.

Four — do the depth exercise. Consider two policies: one lifting a million people from just below the line to just above it, and one halving the shortfall of the poorest hundred thousand. Decide which is better, then work out which index would say so.

Five — check a global claim. Take any statement about world poverty and find which line it uses. Then look up the count at the next line up, and note how different the story is.