VoxEU Column Development Poverty and Income Inequality

A new measure of the global middle class

According to the broadest measure, anyone who is not poor is part of the middle class – that could mean that anyone living on more than $2 a day. This column suggests a more sensible measure: anyone who owns a car. Based on this measure, the global middle class looks quite different.

The swelling middle class in emerging economies is transforming the economic balance of power across the globe. Measuring it, however, is no easy task. The broadest classification of "middle class" suggests the middle class comprises anyone who is not poor, which according to the World Bank means those who earn an income in excess of $2 a day after adjusting for purchasing power (Chen and Ravallion 2010).

That level has now been achieved by more than 4 billion of the world's 7 billion people. Many of these people can buy a cell phone but have no access to a regular power supply. This measure is clearly too low.

In our recent research (Ali and Dadush 2012), we suggest an easier way. In the developing world, buying a car is virtually synonymous with entry into the middle class. In these countries, car ownership separates those with the ability to purchase many other nonessentials from those within the wider population. Car statistics, moreover, are generally reliable and frequently updated. For this reason, we propose the number of passenger cars in circulation serves as a reliable gauge of the size of a country's middle class.

Applying this measure shows that there are many more affluent people in developing countries than had previously been thought and very large numbers of people will enter the middle class in the coming years.

Defining the middle class

There is no widely accepted definition of what constitutes the middle class, and the most common ways of measuring its growth – through looking at rises in income – suffer from a number of flaws. Reflecting rapid growth and low levels of initial income, the middle class in the developing world is certainly rising (Dadush and Shaw 2011).

  • The narrowest classification defines middle class as individuals with an income close to or above the median income in advanced countries – roughly $31,000 per capita or $85 a day at US prices.

Only about 12% of the world's population lives in countries whose average per capita income is higher than that threshold, and only a very tiny minority in developing countries would qualify. Moreover, that level of income is about seven times what marketing studies suggest is needed to buy a car.

  • The most widely used measure of the middle class was proposed in 2002 by Branko Milanovic and Shlomo Yitzhaki, who counted people with daily incomes between roughly $10 and $50, after adjusted for purchasing-power parity (Milanovic and Yitzhaki 2002).

If one uses this definition, there are an estimated 369 million people in the developing G20 economies who qualify as "middle class."

There is a degree of arbitrariness in all these income-based definitions, but even beyond that, they suffer from a number of deficiencies.

  • They are based on infrequently conducted household surveys, which vary enormously in quality.
  • Incomes, even when adequately measured, do not directly reflect private consumption.

Moreover, there is much controversy over appropriate PPP exchange rates for some countries, notably China.1 Comparing incomes across countries at different levels of development represents an enormous challenge. For example, it is obvious that it is impossible to buy even the most basic food and shelter in the United States at $2 a day, the World Bank's definition of non-poor.

So, where is the middle class?

There's a better way to measure the middle class. Cars are big-ticket items that indicate the ability and willingness to purchase many other nonessential goods. Indeed, while the vast majority of households own a car in advanced countries and many own more than one, in developing countries owning a car symbolises relative affluence. Critics may contend that measuring car ownership excludes households that can afford, say, a computer, TV set, or air-conditioner, but not a car. However, because cars in circulation in the developing world are often of very old vintage and correspondingly cheaper – for example, the average passenger car in India is 20 years old, compared with 11 years in the US – this supposed omission is not nearly as large as it seems.

Cars in circulation can therefore provide a measure of the number of middle-class households. After correcting for household size2,  measuring car ownership suggests that the middle class in developing G20 countries is in the range of 550 million to 600 million people – about 50% larger than the number arrived at using the Milanovic-Yitzhaki definition.3

Measuring car ownership also provides us with a more nuanced understanding of where the global middle class resides.

  • In Argentina and China, for example, the estimate of the middle class using cars in circulation matches that arrived at using the Milanovic-Yitzhaki method.
  • But the car-ownership metric suggests that some developing countries – notably Brazil, India, Indonesia, and Mexico – have a much larger middle class than previously believed.

Table 1. Middle class size, using cars and Milanovic-Yitzhaki method

  Average Household Size Passenger Cars (2010, millions) Middle Class, Using Cars (millions) Middle Class, based on Milanovic-Yitzhaki Method (millions)
China 3.1 34.4 106.6 118
India 5.3 13.3 70.5 37
Russia 3 34.8 104.4 57
Brazil 3.7 25.5 94.4 66
Mexico 3.9 21 81.9 37
Indonesia 4.6 10.8 49.7 11
Turkey 4.3 7.5 32.3 17
South Africa 3.7 5.1 18.9 9
Argentina 3.2 7.6 24.3 17
TOTAL   160 582.9 369


Source: Ward's World Motor Vehicle Data and authors' calculations

The car ownership metric suggests:

  • The global middle class has grown much more rapidly than we had previously understood.

For example, according to a World Bank report issued in 2007, the global middle class in developing economies was projected to grow on average annually by 4.5% between 2005 and 2030 (World Bank 2007), but this is less than half of the 10.3% average annual growth of cars in circulation in developing countries from 2006 to 2010. Another study employing the Milanovic-Yitzhaki definition, projects the middle class in China, India, and Russia to grow annually by 9.4%, 5.8%, and 2.4%, respectively, over the coming two decades (Dadush and Shaw 2011). Again, such growth rates are much slower than the growth of cars in circulation in these countries from 2006 to 2010: 35.8% in China, 15.7% in India, and 7.1% in Russia.

  • The ranks of the global middle class appear poised to swell considerably in the coming years.

Cars in circulation per capita rise once income per capita crosses a certain threshold. A cross-country analysis of more than 60 countries suggests that the threshold is around $3,400 PPP. During the 1996-2010 period, for countries with average per capita income between $3,400 and $10,000, the average income elasticity of car ownership is 1.9 (see Figure 1 below).4

About 70 developing countries, home to a combined 4 billion people, lie in the per capita annual income range of $3,400 to $10,000, meaning that a large share of that population is just on the threshold of affluence. In 2010 alone, the BRIC countries – Brazil, Russia, India, and China – added about 14 million cars to their circulation. That figure implies that about 46 million people were added to the middle class, or roughly the population of Spain. The BRICs also accounted for more than half the global increase in cars in circulation in 2010.

Figure 1. Income per capita and elasticity of car ownership*

* Income elasticity of cars is calculated as the ratio of average car ownership growth and average income per capita growth (1996-2008)

Sources: World Bank, Ward's World Motor Vehicle Data, and Authors' Calculations

As one would expect, the correlation between the cars in circulation per 1,000 people and per capita income and per capita consumption is high (0.77 and 0.8, respectively). Reflecting in part differences in the income distribution, the correlation is far from perfect, however. The middle class (as proxied by cars in circulation) is larger in relatively equal societies, controlling for income.5 Moreover, other factors such as space restrictions and policies discouraging car ownership play an important role in some rich countries such as Singapore.

Figure 2. Cars per capita and consumption per capita

Source: World Bank

  • The growth of the world's middle class is coming almost exclusively from developing countries.

Given the stagnation in both population and incomes in advanced countries in recent years, the number of people in the middle class in advanced countries is unlikely to be increasing – and may even be declining (Dadush et al. 2012).

The untapped potential of the middle class in developing countries is clearly evident when comparing car ownership levels there with those in advanced economies. The number of passenger vehicles per 1,000 people in India and China is just 10 and 27, respectively, compared with 502 in Germany and 451 in the United States.

Why this matters

As the ranks of the middle class swell, firms in advanced countries will increasingly focus their efforts on selling products to that middle class. For this reason, trade agreements struck with developing countries may yield more gains for advanced-country exporters than is generally understood, at least over the next several years.

Policymakers should also take heed of this rapid rise in car ownership, as governments in developing countries will be forced to contend with more urban and airport congestion, as well as more pollution and carbon emissions, for example, than would be deduced from aggregate growth statistics. Similarly, exploding demand for government services are likely to present particular challenges for countries where per capita income is still quite modest, tax revenues are low, and government capacity to satisfy such demands is limited.

The middle class also expects government to be representative and accountable, and they are sure to put increased pressure on the nondemocratic systems in many developing countries. Seen in this light, the rising incidence of protests and dissent in China, Russia, Thailand, and the Arab world is not surprising.

A longer version of this article will be published as a Carnegie Paper. The authors are grateful to Branco Milanovic, William Shaw, and Zaahira Wyne for their comments.


Ali, Shimelse and Uri Dadush (2012), “Where on Earth is the Global Middle Class?”, Carnegie Paper (forthcoming).

Banerjee, A and E Duflo (2007), “What is Middle Class about the Middle Classes Around the World?”, MIT Department of Economics Working Paper 07-29, Cambridge, MA.

Dadush, Uri, Kemal Dervis, Bennett Stancil, and Sarah Puritz Milsom (2012), Inequality in America: Facts, Trends and International Perspective, Brookings Press.

Dadush, Uri and William Shaw (2011), “Juggernaut: How Emerging Markets Are Reshaping Globalization”, Carnegie Endowment for International Peace, Washington, DC.

International Energy Agency (2006), World Energy Outlook.

Kharas, Homi (2010), “The Emerging Middle Class in Developing Countries”, OECD Development Centre, Working Paper No. 285.

Milanovic, Branko and Yitzhaki, Shlomo (2002), “Decomposing World Income Distribution: Does the World Have a Middle Class?”, Review of Income and Wealth, International Association for Research in Income and Wealth, 48(2).

Shaohua, Chen and Martin Ravallion (2010), “The developing world is poorer than we thought, but no less successful in the fight against poverty”, Policy Research Working Paper 4703, World Bank.

World Bank (2007), Global Economic Prospects: Managing the Next Wave of Globalization, Washington DC.

1China’s GDP was recently revised downwards by about 40% based on new methods for measuring cost of living in China, adding millions of households to the poverty count.

2In some countries, the middle class household size is smaller than the average national household size. In India, for example, the average size of middle-income households is 4.3 people, compared with 5.3 in all households. In other major economies, however, such as Brazil and China, the difference in family size between middle-income households and the national average is very small. In China, India, and Brazil, using the household size of the middle class, instead of the national average, will reduce our estimate of the middle class by about 22 million people.

3Using Milanovic-Yitzhaki’s definition, those with annual incomes of at least $4,000 in 2005 PPP, we estimated the global middle class in developing G20 economies, which accounts for more than 70% of developing economies’ GDP, at 369 million.

4These elasticity estimates are broadly confirmed over much shorter periods, including over the most recent five year period, 2006-2011.

5This may not be true in low income countries as a certain minimum level of income is necessary to buy a car. Thus, low-income countries with high inequality (that have a larger share at the upper end of its income distribution) may initially have larger car ownership rates than countries with more equal distribution of income.