The Median as a Better Measure of Development – and Better Than the World Bank’s Shared Prosperity

November 06, 2013

Last month I spoke at the annual Global Empowerment Meeting organized by Harvard’s Center for International Development, which brings together development wonks, entrepreneurs, and international policy leaders. Together with Lant Pritchett (CGD senior fellow and Harvard Kennedy School) and Brookings’ Homi Kharas, I was on a panel that asked: “Does ‘a Dollar a Day’ Really Define Development?”

Lant argued that a focus on a “dollar a day” extreme poverty line is silly and narrow.  Homi, who headed the secretariat to the post-2015 panel report of the UN, counter-argued that eliminating extreme poverty is a politically salient goal that reflects a worldwide consensus on what ought to be done and can be done by 2030.

I took another tack altogether. I argued for median income (or consumption) per person as a sensible, simple, durable measure of material well-being across countries and over time (durable as in salient for all time in all countries, whether low, middle or high-income today) – and also a far better measure of the state of development broadly defined in a country.

I also showed it is “good-enough” measure of inequality, once compared to measures based on means or averages such s as GDP per capita.  That means it beats out the new World Bank measure of “shared prosperity” (the increase in the mean income of the bottom 40 percent of the population). Why? The Bank measure will reflect the growth rate of income for that 40 percent in absolute terms. But compared to what? A measure of “shared” prosperity needs a comparison.  A simple option is the change in consumption at the median compared to the change at some mean such as mean GNI or GNI per capita.

The full slide deck I used for the presentation is based on a forthcoming paper with Christian Meyer here at CGD. Below is the fantastic (thanks to Christian) bubble slide (#3 in the deck).

The slide shows the relationship between median household consumption per capita on the horizontal axis and GNI (gross national income) per capita on the vertical axis (the measure that the World Bank has used to classify economies since its first World Development report was released in 1978). The size of the bubbles in the resulting scatterplot is proportional to the percentage of the population in each country living below $2 a day – a $2 a day poverty rate. 

What does the slide illustrate?

First, median consumption tends to be higher in richer countries measured by GNI per capita; the bubbles are getting smaller as we move towards the north-eastern corner of the chart.  That makes sense. No surprise there. 

Second, one-half of populations in low-income countries and a few middle-income countries survive on consumption levels well below $2 a day. No matter how you define “poverty”, there is lots of it in the world. The median of daily consumption per capita in the developing world is a meager $3 or so (measured at 2005 PPP, in around 2011); one-half of people in the developing world survive on $3 or less.  The median in western Europe is above $50 a day. Among those reading this blog post, my guesstimate of median daily income (consumption will be lower) is probably $100 a day ($36,000 annual income for a single person and about $140,000 for a family of four), and that’s counting all the students and millennials that I hope are reading it.

Third, median consumption conveys information at the country level about both absolute income in a country and the poverty rate. In Mali median consumption is $1.25 a day; if the poverty line (that Lant rejects) is drawn at $1.25 a day, it is obvious that the poverty rate in Mali is 50 percent. In India, median consumption is $1.55 a day; that is associated with a $1.25 poverty rate of 33 percent.

Finally, measures based on national accounts such as GNI or GDP per capita are often misleading in assessing the well-being of the bottom half of any population by consumption or income, and of the state of development in a country defined in a broader sense to include state capability and legitimacy and social cohesion and trust. Consider these examples:

  • Looking along the vertical axis at countries classified as low-income by the World Bank, median daily consumption per capita varies enormously: from $1.30 in Benin and Bangladesh to more than twice as much, i.e. at or close to $3 in Cameroon and Tajikistan.
  • Among countries classified as lower-middle income with GNI per capita of around $1,800, there is Timor-Leste with median household consumption of about $1.50 per capita per day, less than a third of median household consumption in Moldova of more than $5 per capita per day.
  • Among countries classified as upper-middle income (only a few are in this slide, but see slide #4 in the full slide deck), the median also ranges widely: from $1.40 in Angola and $3.40 in South Africa to about $7 in Brazil and $10 in Poland.
  • Looking across the chart along the horizontal axis at about $1.50 daily median consumption per capita, the bubbles represent countries with a wide range of mean GNI or GNI per capita: Malawi, Ethiopia, and Guinea at less than $500; Pakistan, Senegal and Nigeria at about $1000; East Timor at almost $2,000 per capita; Swaziland at almost $3,000; and Angola at almost $4,000 (where the $1.25 poverty rate is 43 percent).
  • At the $2 to $2.50 median, there is Cambodia (GNI per capita of $690), Sudan ($1,110), the Philippines ($1870), rural Indonesia ($2,930) and rural China ($3,620). At just over the $5 median, material well-being for the bottom half in Colombia ($5.11) and urban China ($5.65) is probably similar, though China’s mean GNI (rural and urban) is 50 percent lower.

Maybe Lant and Homi could agree on the median while disagreeing about the logic of a poverty line and a poverty elimination goal. Maybe the World Bank could rethink its measure called “shared” prosperity, and monitor the income or consumption level at the median of every country’s distribution, compared to the mean.

Let me know what you think. 


CGD blog posts reflect the views of the authors, drawing on prior research and experience in their areas of expertise. CGD is a nonpartisan, independent organization and does not take institutional positions.