Oops: Economists in Confused Search for the Middle Class in the Developing World

May 31, 2012
This is a joint post with Christian Meyer.With rapid growth in emerging market economies in the last decade, millions of people have entered the new global middle class. That has created new consumer markets in Latin America, Africa, and Asia – a good example: Dunkin’ Donuts going to India – and sparked considerable interest in the global financial and corporate communities.But for economists and students of development, the interesting question is whether a large middle class is not only a consequence of growth but a contributor to growth and other good things in the long run. Does the “middle class” matter, possibly, because its members are more likely to save and invest in the future (in education and directly productive activities), more likely to support market-friendly economic policies, and more likely to insist on accountable, competent and clean government – all adding up to making the middle class  "indispensable" to a stable, prosperous, democratic system?To systematically explore those possibilities (or hypotheses) requires defining the middle class in a manner that is reasonably consistent over time and space. Economists have tended to use income- or consumption-based identification as a proxy for a theory-based definition.  Some such identification is, in principle at least, easier to do using household survey data and less subject to judgment calls than, for one example, the traditional sociological definition of sufficient education, membership in a white-collar occupation, or apparent embrace of bourgeois values.So far, however, there is no agreement among economists on the appropriate income/consumption thresholds to identify the middle class – not even on whether it is relative or absolute income that matters, nor on whether the identification should be based on a country, a regional, or a global standard.  The logic of using some middle portion of a country or global distribution of income/consumption is that the class being identified is called the “middle class,” and the appropriate reference group is the country.  The logic of using absolute thresholds is that the term “middle class” is viewed as identifying a “class,” historically in the Western world and nowadays at the global level, with certain behaviors and characteristics for which a certain level of income or consumption is a good identifier across nations and regions.The graphs below illustrate the, well, confusion among economists about the right thresholds to identify the middle class. Using the income distribution of Brazil in 2009, the first graph plots various identifications (sometimes called definitions) against relative quintiles of per capita household income; the second graph against absolute per capita household income/expenditure in 2005 PPP dollar per day.

Identifying the Middle Class in Brazil, by Income Quintile

Identifying the Middle Class in Brazil, by Absolute Income/Expenditure

In the first graph, note there is almost no overlap at all between Banerjee and Duflo’s middle class and the middle class of Milanovic, Kharas, Birdsall and the World Bank Latin America and Caribbean Department.  That’s because, as shown on the right, Banerjee and Duflo identified the middle class to include anyone over the $2 a day poverty line (and below $10), which includes a very large portion of the population in most developing countries, while others identified the middle class to include people with at least $10 a day in income or consumption! Depending on your point of view, the middle class in Brazil includes everyone in the three middle quintiles (Easterly, 2001) – about 114 million people in 2009 – or everyone with per capita daily income between $10 and $50 (World Bank, 2012) – about 61 million people, with an overlap in the case of Brazil in 2009 of only 36 million people.

In Birdsall (2010) I justified a global identification, based on an absolute $10 threshold at the bottom on various grounds (including Lant Pritchett’s argument for a new global definition of poverty), and at the top a country relative cut-off, excluding from the middle class the richest 5 percent of people.For the LAC region, a World Bank team has developed a LAC-specific identification of $10 to $50. Luis Felipe Lopez-Calva and Eduardo Ortiz-Juarez show that at an income of around $10 PPP a day, non-poor people in the region only have a 10 percent chance of falling into poverty – secure enough for households to be able to care about and save for the future, and to have aspirations for a better life for themselves as well as their children. The $50 PPP per day cutoff is more arbitrary.  In LAC countries it excludes between 2 and 5 percent of the richest people.Using the $10 to $50 thresholds, here is a graph that illustrates the share of population (or relative size within each country) and the absolute size of the middle class in some non-LAC economies. Urban China has the biggest middle class – representing a relatively small portion of its urban population (and of course an even smaller portion of its total population).

Share of population (vertical axis)  and absolute size (bubble area),  2008/2009

Let us know what you think about the various approaches to identification and definition at the national and global level!
Sources and further readings:


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.