N.V1-6.01 | THE LINE IS AN INVENTION

01| Somebody sat down and decided what a poor person needs. That decision became a number, the number became a threshold, and everything built afterwards inherited it — who gets counted, who gets served, and what counts as having succeeded. The line is not a discovery about the world. It is an instrument, and instruments have designers.

02| The Stiglitz–Sen–Fitoussi commission, convened in 2008 to examine the limits of GDP, put the general principle in a single sentence:

"What we measure affects what we do; and if our measurements are flawed, decisions may be distorted." — Stiglitz, Sen & Fitoussi (Report by the Commission on the Measurement of Economic Performance and Social Progress, 2009)

03| Applied to poverty, the distortion has a specific shape, and it is about where success gets recorded. An intervention that lifts a household from just below a line to just above it registers as a complete success. An intervention that leaves the household in the same income band while making it durably less likely to fall registers as nothing at all. Programmes are funded, evaluated and renewed against the first kind of result. That is a plausible mechanism rather than a demonstrated one — the chain from metric to funding decision to programme design has been argued far more often than it has been traced — but nobody has offered a better account of why the sector optimises the way it does.

04| The counter-case deserves stating properly rather than waved past. A simple, stable, internationally comparable income line is what makes it possible to say anything at all about eight billion people, to compare countries, and to hold a government to a number it cannot quietly redefine when the number becomes inconvenient. Every proposal to measure more things is also a proposal to make the measure more discretionary, and discretion over the poverty line has historically been used to shrink it. The 1965 adoption of the lower American threshold is the standing example. This does not resolve. It is the real dilemma in the chapter and we leave it open.

Where progress stops

05| The most consequential empirical claim in any argument about the sector is that it is good at one transition and bad at another: it moves people off the floor of extreme deprivation but not into anything resembling security. Unusually for claims of this kind, it is directly checkable, and the check produces something sharper than the claim.

06| The World Bank revised its lines in June 2025 to 2021 purchasing power parity. The international poverty line moved from $2.15 to $3.00 a day; the line typical of lower-middle-income countries from $3.65 to $4.20; the line typical of upper-middle-income countries from $6.85 to $8.30. The counts for 2022: 838 million below $3.00, 1.6 billion below $4.20, 3.9 billion below $8.30 — 48 percent of the world.

07| These are not gradations of the same thing. The lowest reflects the minimum needed in a typical low-income country for food, shelter and clothing. The highest reflects what national poverty lines in upper-middle-income countries actually cover — a healthy diet, sanitation, electricity, connectivity, education. Moving from one to the other is not climbing a scale. It is crossing between two different definitions of what a person needs.

08| In share terms there has been movement at every level. In 1990 roughly a third of the world lived below the extreme line, another third in the middle band, and a third above the upper line. By 2024 fewer than 9 percent were below the extreme line and 56 percent were above the upper one. In absolute terms the picture inverts, and the Bank states it plainly:

"Because of population growth, the number of people living on less than $6.85 a day globally has barely changed; namely, 3.7 billion in 1990 and 3.5 billion in 2024." — World Bank (Poverty and Inequality Platform, 2025 update)

09| Two hundred million fewer people across thirty-four years, during which world population grew by 2.8 billion. So the accurate version of the claim is not that progress stops. It is that proportional progress occurred at every threshold and absolute progress occurred only at the lowest. People are crossing the bottom line faster than the population grows, and crossing the upper line at roughly the rate the population grows. The band in between — out of deprivation, not into security — has been the same fraction of humanity since 1990.

10| That band is a distinct condition, not a milder version of poverty. The strongest conceptual work defines the middle class not by income adequacy but by the probability of falling back. Ferreira, Messina, Rigolini, López-Calva, Lugo and Vakis, working on Latin American mobility, place the vulnerable class at roughly $4 to $10 a day and the middle class at $10 to $50, with the boundary set by the income level at which the chance of falling back becomes low. A household in the lower band has left deprivation without acquiring resilience. One shock returns it. The dollar figures are calibrated to one region and should not be exported without re-estimation, but the concept is general and it is the analytic footing the last-mile argument has always lacked.

11| Now the part that matters. Almost nothing in the sector is evaluated against that crossing. The trials measure consumption, income, assets, employment, health and wellbeing over one to eleven years. Those are the right outcomes for asking whether an intervention lifts someone out of deprivation. They are the wrong outcomes for asking whether it moves anyone into security, which requires either a threshold crossing or a measured change in downside exposure, over a horizon long enough for shocks to arrive.

12| So the honest close is uncomfortable in a way the usual complaint is not. If the band between deprivation and security has held at a third of humanity for thirty-four years, the interesting question is not why interventions fail to move people across it. It is whether any intervention has ever been evaluated against that crossing at all. On the evidence, essentially none has. The sector may not be failing at the last mile so much as not measuring it — and those are very different problems with very different answers.