The World Bank has good news. According to its latest calculations, global poverty has been falling. In 1999, an estimated 1,751,000,000 people suffered from extreme poverty. By 2011, the number had dropped to 983,000,000. World Bank President Jim Yong Kim tweeted excitedly that the rate of extreme poverty may drop below 10 per cent of the world population when the figures are in for 2015. ‘The international community can celebrate’, exclaimed a recent Bank report. Despite the global financial crisis, it had chalked up some ‘robust’ development successes. The un’s Millennium Goals had played an important role in galvanizing efforts to reduce poverty, and that experience would help drive progress towards the Sustainable Development Goal of eradicating it altogether.footnote1 But what is poverty? How do you count the poor?

Since 1980 at least, the World Bank’s answer has been a concept known as the International Poverty Line.footnote2 In 1990 the Bank’s researchers, led by an lse-trained Australian, Martin Ravallion, estimated this at $1.02, the famous ‘dollar a day’, at purchasing power parity—that is, what the equivalent of a dollar could buy at local prices. Anyone falling below that line was counted as ‘extremely poor’. Ravallion and his colleagues derived the figure of $1.02 as a ‘representative’ marker for absolute poverty: of the 33 poor countries they were studying, this was the approximate sum already set as a ‘national poverty line’—often by the World Bank itself—for Indonesia, the Philippines, Bangladesh, Pakistan, Kenya, Tanzania and Morocco. To be precise: this was the figure at which the Bank economists arrived after re-scaling these domestic-currency sums—denominated in Indonesian rupiahs, Filipino pesos, Bangladeshi takas, and so forth—according to local inflation levels, to calculate their ‘equivalents’ for the year 1985; and then, for purposes of international comparison, converting those sums into a common unit of ‘purchasing power’: the 1985 ppp dollar. The purchasing-power values themselves were calculated from data gathered in its 1985 survey by the International Comparison Programme—itself now hosted by the World Bank in Washington, dc—which checks prices of goods in different countries. Each subsequent release of new global ppp data by the icp, usually some years after the date—1985, 1993, 2005, 2011—to which the prices refer, has led to a corresponding reassessment of the International Poverty Line by the Bank’s economists.footnote3 With the release of the 2011 ppp figures in 2014, the World Bank economists have raised the ipl to $1.90, using methods justified at length in a new report by Bank economist Francisco Ferreira and his colleagues.footnote4

The figures on global poverty extrapolated from the Bank’s poverty lines enjoy enormous international legitimacy. Canonized in the Bank’s World Development Reports, they have been used to determine priorities for resource allocation and to assess the relative success of poverty-reduction programmes. They formed the reference of the un’s first Millennium Development Goal, to reduce poverty by half between 1990 and 2015—and were the measure of how well it had succeeded. In September 2015, world leaders at the un’s headquarters in New York pledged to eradicate extreme poverty altogether, one of a set of Sustainable Development Goals adopted to guide their future policies—and once again based on the Bank’s assessment. Since poverty reduction is claimed to be the Bank’s central aim, the data it provides are also a key metric for assessing its own performance. Gratifyingly, the Bank has consistently been able to claim that its figures prove it is ‘on the right track’, even if ‘much remains to be done’.

Yet the Bank’s money-metric approach is open to a host of objections, even on its own terms. The criticisms that have been levelled at the Bank’s methodology since the 1990s over key technical questions—ppps, inflation measures, price variations within countries (in particular the differing costs in rural and urban areas) and the merits of income versus consumption data—apply in spades to its latest iteration, as we will show below. More broadly, it might be asked whether an approach focused only on extreme poverty, or ‘absolute deprivation’, as the Bank terms it, is potentially self-serving. Poverty lines may be set so that hundreds of millions are only just below them, with a change of only a few cents sufficient for large numbers to be ‘lifted out of poverty’, without any substantive change in their position. The Bank assumes a priori that there is no poverty in high-income countries, even though it acknowledges this may not be ‘fully supported’ by the facts.footnote5 Indeed, as discussed below, alternative data show the assumption to be false, especially at more adequate poverty lines.

Further problems arise from inadequate data. For entire regions, including most saliently the Middle East and North Africa, the Bank presently reports no results at all, because of poor survey coverage. In a number of other cases, it appears to have based country estimates on figures deriving from other countries, to account for missing data. Judgements of this type are an unavoidable aspect of applied work in data-poor environments, but while one can sympathize with the necessity to make such calls—and indeed, to be forthright in defending them—the resulting uncertainties must be adequately recognized. There have also been serious institutional problems with the International Comparison Programme, on whose ppp estimates the Bank’s entire poverty count depends. Originating in 1968 as a project of the un Statistical Division, after its 1993 report the icp was the subject of a highly critical un review, which identified failures of management and resources at all levels; hence the long delay before the appearance of its 2005 figures. Although the icp’s latest price collection exercise is its most comprehensive yet, there remain fundamental questions as to how to collect information and undertake comparisons, and also as to what is being measured.

At the root of most of the seemingly unrelated problems with the Bank’s global poverty estimates, however, there lies a single conceptual failure: the lack of a criterion for identifying the poor that is appropriate to the task and has a consistent, substantive interpretation. It is a problem that cannot be solved within the existing approach, but requires an altogether new one. In what follows, we first examine problematic issues in the Bank’s most recent poverty estimates, focusing in particular on purchasing-power parity, inflation calculations and rural–urban variations. We then demonstrate that following the Bank’s own stated approach could lead to an alternate—and much higher—set of poverty estimates, which we report. However, we do not present these as a last word, but rather as a demonstration of the seriousness of the uncertainties involved—and of the need for an altogether new framework.

The Bank’s economists justify their decision to set the latest International Poverty Line at $1.90 a day, in 2011 ppp dollars, with the argument that this will ‘preserve the real purchasing power of the previous line’—that is, $1.25 a day, in 2005 ppp dollars—‘in the world’s poorest countries.’footnote6 But in what sense do the lines in fact correspond? The question applies equally to the Bank’s previous ‘updates’, replacing ipls specified in the base years of 1985 with 1993, and 1993 with 2005. In looking for an answer, we might begin by examining the Bank’s own view of the matter. Its customary argument as to why the new poverty line is equal to the previous one in purchasing power is that the poverty headcount ratio is very similar in both cases. However, as with the Bank’s earlier estimates, such an argument is little more than a non sequitur.footnote7 We may think of the problem this way: suppose that an arbitrary set of new ppp dollars were chosen to translate the ipl into local currencies—or, indeed, an especially perverse set, deliberately chosen, for example, to misrepresent the real level of purchasing power in each country. By starting at a low enough value and then creeping up, one could always find an ipl that would suffice to generate exactly the same headcount as did the previous ppps. Since this argument can be used to ‘rationalize’ any set of ppps, it cannot justify any one choice thereof. What it resoundingly does not show is that the new ppp dollars maintain the purchasing power of the old ones, anywhere—let alone everywhere. In any case, it is evident that this argument cannot be used to justify the original choice of the ipl.