The Global Poverty Line: Past, Present and Future

It is easy to forget how young the global poverty line still is. Though it has been around since I started studying development economics, it’s actually around 10 years younger than I am. While it continues to experience growing pains (befitting its youth), there are also signs that its life (at least in its current form) may be coming to an end. In this blog I explore three sets of challenges in the measurement and use of the international poverty line, and consider five ways to meet these challenges. 

Measuring global poverty at all is a huge achievement

Before going into the challenges, we should celebrate the advances in statistical practice that underlie our ability to talk about “global poverty” in a meaningful way. Two major advances have been key:  the first was the development and roll-out of high-quality household surveys around the world. I recently read Pranab Bardhan’s memoir Charaiveti; the story of household survey development in India occupies many of those pages. Careful measurement in developing countries is requires both good judgement and good systems. The spread of good household surveys has required  Nobel laureates to clarify the conceptual issues that underpin good survey design, and countless unknown enumerators and field managers working through sometimes extremely difficult conditions to implement them. We should not take this for granted. 

The other great advance has been the collection of widespread and comparable price data from around the world; another huge achievement in global statistics. The International Comparisons Programme or ICP (now housed at the World Bank) has grown and improved immensely over time—though, as we will see, it is not without its difficulties. Without it, we simply wouldn’t be able to compare countries on a like-for-like basis and assessing global poverty levels would be a pipe dream. In 1970, the programme covered just 10 countries; today it includes data for 176 countries (having peaked at 199 in 2011). The time spent on the collection and analysis of this data requires hardly bears thinking about. On this statistical bedrock, measuring poverty against a chosen line is relatively easy. Choosing a line was more difficult, but Martin Ravallion and his co-authors opted for pragmatism (with a little flair for marketing) when selecting a dollar a day in the early 1990s. It was pragmatic because it was simply an average of the available lines from a handful of the lowest income countries; stylish because it was simple, catchy and intuitively a very poor standard of living indeed. 

But challenges remain—and may be existential

Having acknowledged progress I now turn to challenges, some of which originate from the very progress I have just celebrated. 

Poverty measurement depends on data; but the places with the noisiest and most frequently missing data are the very places most likely to have large populations =at risk of (or already in) poverty. Afghanistan, Somalia, Eritrea have no internationally-comparable poverty data. All are likely to have substantial populations living in poverty. The most recent data from Sudan is for 2014, and for Venezuela, 2006.   Both places are likely to have seen substantial changes in recent years—and yet our global statistics can say little about them. As the world becomes more prosperous, and poverty declines in more and more places, these blind spots will be increasingly important to our understanding of the level of global poverty.

But measurement issues are not the sole preserve of the most conflict-affected or unstable places. Turning price surveys in places with different consumption habits into a single, globally comparable “currency” is an art as well as a science. And as we continue to learn about strengths and weaknesses of different approaches (which include their cost and ease, not just their accuracy), there have been changes to the methodology adopted by the ICP. This could cause substantial swings in our estimates of the level and distribution of global poverty, since different methodologies do not have geographically consistent effects. This would be problem enough, but when each ICP round is completed, the World Bank typically takes it as a good moment to revise the poverty line. It is hard to escape the conclusion that the poverty line and price data are in fact endogenous to each other, chosen because they create some continuity and order as much as because each are, separately, the right choices. Indeed, the Bank’s own reporting is sometimes muddled here: in their FAQ, they report that moving from $1.90 to $2.15 was purely driven by price data.  In fact, the move also incorporated  a move to incorporate an additional 13 countries national poverty lines, as they were added to the data (though in their defence, the choice of the line as the median of these countries is not particularly sensitive to the number of countries included). 

This is a challenge to the idea of an international poverty line against which we are trying to make progress The reality is that there is no one line, designed to answer one clear question; it has been changing regularly. If you are an official in a developing country or development agency trying to achieve SDG 1, this really matters: the goalposts for eradicating extreme poverty keep changing, and they always have. 

There are other conceptual issues, too. A recent paper by Josh Merfeld and Jonathan Murdoch convincingly argued that because household surveys in developing countries are(though wonderful and utterly praiseworthy) not completely at the frontier of good statistical practice, the “thing” we measure is not actually aligned with our understanding of what poverty is. We tend to think of poverty as a characteristic of people over a given period of time: how many people were poor in a given year. But because of how the surveys are structured, what we actually measure is how much of the year, on average, households spend in poverty. These may sound similar, but the difference can have large consequences. For one, the former is usually quite a bit lower than the latter; that is, even many otherwise not-poor households will typically spend part of the year measurably in poverty. More profoundly, it also implies different policy responses: specifically, interventions that ease consumption smoothing may have a relatively larger effect on poverty as measured, compared to those that focus on lifting average incomes for the very poor. And yet, the latter is closer to the typical conception of “poverty reduction” than the former—though economists would absolutely see the value of each.

And there is a further, more existential challenge. In an increasing number of countries that are poor and economically challenged by almost any subjective judgement, only a minority of the population qualify as extremely poor. In Senegal, the proportion is 10 percent; Mali 20 percent and Pakistan 5 percent. In even Sierra Leone and Liberia the share of the population living in  “poverty” is below 30 percent. And yet, no-one could go to these places and describe the standard of living for the majority of the population as much better than dire.  For these countries, extreme poverty has lost a large part of its informativeness—it says increasingly little about the challenges and difficulties facing the economies overall and even most of the people in those countries facing deprivation. What does the global extreme poverty line offer these places, some of which remain recipients of substantial volumes of foreign assistance? Does it offer continuing policy relevance? 

My colleague Charles Kenny has argued that the global poverty line is most important as a marketing tool: it has focused attention more than actual resources on the world’s poorest and given relevance to the international development project. As time goes on, it will serve these purposes less well if in many of the places needing support, it describes an increasingly small portion of the population, while many others also need support.

What now?

The international poverty line staggers along for now. It remains important because 700 million people in the world live below it; and regardless of how it has shifted over time and the correct conceptualization of what exactly it measures in practice, one thing is clear: they suffer from extreme material deprivation, and there is much we can and should do to relieve it. But despite this, given the needy it omits (both because they cannot be surveyed and because they sit a little above the low-bar line we’ve adopted), it will come under increasing pressure over time. 

There are five possible responses to this pressure.

The first is to do nothing. If we think the 700 million people living in a not-quite-precisely defined form of material deprivation are the principal priority of international action, and a shifting, slightly ill-conceived line is a good marketing tool to remind us of this fact, then no action is needed. The weaknesses will remain, but so will the value. There is much merit to this view; but it does mean extreme poverty becomes less valid and useful for more and more people and places over time. And in the longer run, a global poverty line made by averaging the poverty lines of many low-income countries will disappear when the last low-income countries graduate to lower-middle income status. A related alternative is to use an international basic needs poverty line, of the type proposed by Robert Allen. By allowing for variation in the level of income that denotes extreme poverty by country, it retains informativeness for more countries, and is no longer dependent on the individual poverty lines of low-income countries. Against this, it remains a very low bar for poverty.

The second option is to adopt a much higher poverty line. Indeed, the World Bank already reports on higher lines, precisely because not being in extreme poverty does not mean one isn’t in substantial material deprivation. This approach has been proposed by Lant Pritchett and Max Roser. It’s a simple approach, which would ensure the line has wider relevance. But it also would mean there is less emphasis placed on the very poorest of the poor, who by most conceptions of justice need support most of all. There is merit in being able to identify the very poorest in the world on a comparable basis; a higher line obscures them.

A third option is to use a completely different kind of measure the combines elements of the focus on extreme poverty and the higher line: a prosperity gap approach. The idea is simple: we set a higher line, above which we think a more rounded decent life is achieved, and we measure how far people are from it. The World Bank already report on this for some countries. This still gives more weight to the poorest, but it does not completely neglect the very-but-not-extreme poor. If we wanted an even stronger weight on them, we could adopt a “prosperity gap squared” measure, which then emphasises the plight of the worst-off . There is a lot to like in this approach, but it fails the “marketing” test: poverty gap measures have existed for a long time and never captured the imagination. If the line has been important because it was compelling, gap measures cannot replicate that.

A fourth approach would be to abandon the attempt to make a perfect monetary poverty line, and instead work towards a non-monetary or partly-monetary poverty line. This is not a new idea: it underlies Sabine Alkire’s multidimensional poverty index (MPI). It has the happy benefit that many non-monetary rewards are at least simpler to measure in a consistent fashion: child mortality rates, for example, or access to clean water can be directly defined in a comparable fashion, removing one source of error. Against this, we still need to combine them in ways that reflect local preferences (or choose not to; in either case a controversial choice is being made). A solution might be to pick just one dimension, such as being malnourished, or close to it. This does not, however, solve the problem of data collection, or relevance. Data is still hard to come by in difficult places, and—at least how the MPI is constructed—the bars for exiting poverty are sufficiently low that the relevance of the measure is being outgrown in many places that remain objectively poorly off. Alternatively, ways of carefully combining qualitative and quantitative assessments of poverty are increasingly available, and can provide important additional information. But again, these have existed for a while (in 2004 Ravi Kanbur edited an entire short book about the approach); they have not caught on because they are less intuitive. 

A fifth option is to return to using national poverty lines, returning to our pre-1990s state without a “global” reference point.  This would be useful for countries where fewer people live in extreme poverty, and would tailor relevance to the particular economic circumstances of a country (most poor countries already have their own national poverty lines which often do just this). But again, it falls foul of the marketing value of a global poverty line, and for donors or funders who target development cooperation based on the extent of poverty, a globally consistent measure is still valuable.

Normally, I’d conclude these comments with a clear recommendation. But there may be no good solution to the problems of poverty measurement. (Intellectually, I lean very much towards a prosperity gap-squared measure, but I accept it is unlikely to ever catch on). So extreme poverty will most likely plod on, imperfect, slowly losing relevance; until we eventually stop paying attention to it.

Note: I was recently invited to speak at a workshop concerning ‘issues in the measurement of poverty’ at the Institute for Fiscal Studies. I chose to discuss three problems in the measurement of global poverty. This blog  is adapted from my comments.

This blog has benefited from feedback from participants at that workshop , Stefan Dercon, Charles Kenny and Vijayendra Rao. Any errors or omissions are my own.