Syndicate content

"HAVEs" and "HAVE-nots": A Simple Global Poverty Target

Aart Kraay's picture

There has been much discussion around the World Bank on the choice of a "global poverty target" that can be used to measure global progress against poverty. To be successful, such a target needs to be (a) simple to understand, and (b) relevant to all World Bank client countries.

                The World Bank already has a well-known "dollar-a-day" headcount poverty measure, which measures the share of the developing world's population living on less than $1/day. This measure has been enormously successful, not least because it is very simple to understand. It is also relevant to many low-income countries, where a significant fraction of the population lives below this threshold. However, it is not very relevant to middle-income countries, because few people in these countries live on less than $1/day.

                Martin Ravallion has recently proposed a poverty target that addresses this problem. It keeps the low "dollar-a-day" poverty line for the poorest countries, and then sets progressively higher poverty lines for richer countries. Choosing higher poverty lines to match those observed in richer developing countries elegantly solves the problem of relevance. Counting the number of people in each country below each country-specific poverty line maintains the simplicity of a headcount measure of poverty. 

                However, an important feature of this approach, noted by Jishnu Das in his review of Martin's proposal, is that it does not treat people equally in monetary terms across countries. For example, Martin's proposal would set the relevant poverty line for China at $2.40/day, and the poverty line for much richer Brazil at $5.50. This means that the World Bank's global poverty target would consider a Chinese citizen living on $3 per day to be "not poor", while a Brazilian citizen living on exactly the same amount would be considered "poor".   This matters if one takes the global poverty target seriously as a guide to policy, or to inform aid allocation decisions. For example, would this mean that the World Bank should support a workfare program with a guaranteed wage of $3/day in Brazil, but not in China?

                Martin's proposal does treat people equally in terms of their utility or welfare. Martin's proposal is based on the idea that people care not only about their own consumption, but also their consumption relative to others in their country. Given this assumption, a person living on $3/day in Brazil feels poorer than he would if he lived in China, because Brazilians on average consume more than Chinese do. This in turn implies that poverty lines in dollar terms should be higher in richer countries. However, this may sit a bit uncomfortably with non-economists unused to thinking in welfare terms. For example, at least some people may find it puzzling that the World Bank's global poverty target would improve if a Brazilian living on $3/day moved to China, just because he would feel less poor there.

A Simple Alternative

                Here is an alternative proposal for a global poverty target that is simple, relevant, and that classifies people as poor in the same way regardless of where they live. First, calculate the fraction of the developing world's population living below a sequence of poverty lines set at $1/day, $2/day, $3/day etc., up to $10/day.   Second, define the global poverty target as the simple average of each of these poverty headcount rates. This graph shows this alternative global poverty target over the past three decades.  

                The "headcount average" approach is very simple to understand. It also is relevant to all of the Bank's client countries, because the highest $10/day headcount included in the average is chosen to match the highest poverty lines currently observed in developing countries.  The "headcount average" approach also counts people as poor in the same way everywhere in the world:  if a Brazilian living on $3/day moves to China, the "headcount average" does not change.

                As a global target, the "headcount average" has another useful feature -- it is a weighted average across people, with more weight assigned to poorer people.   A person living on a dollar a day is poor according to the $1/day poverty line. But she is also poor according to the $2/day line, and the $3/day line, all the way up to the $10/day line, so that she is counted 10 times in the "headcount average". In contrast, somebody living on $2/day would be considered poor only for poverty lines starting at $2/day and up, and so she is counted only 9 times, while a person living on $10/day gets counted just once.

                Giving more weight to the poorest in a global poverty target intuitively sounds appealing for an institution like the World Bank. But does it make sense to give 10 times more weight to a person living on $1/day than to a person living on $10/day? One justification for such weights is that they are exactly the ones used by poverty specialists when they calculate the standard "poverty gap" measure of poverty. In fact, if we use more finely graduated poverty lines ranging from $0/day to $10/day in increments of one cent, the "headcount average" would just be the "poverty gap" with the poverty line set at $10/day. However, the "headcount average" might be easier to explain to non-specialists than the "poverty gap".

                Finally, this approach has the benefit of a catchy acronym -- maybe the "Headcount AVErage" is a good way to distinguish the world's "HAVEs" from the world's "HAVE-nots"?