The Contradictions in Global Poverty Numbers

    Laurence Chandy and Homi Kharas

    The World Bank has just issued its latest estimates of global poverty. In 2008, 1.29 billion people lived on less than $1.25 a day (the bank’s global poverty line) while 2.47 billion people lived on less than $2 a day. Poverty is falling across all regions. While expressing caution given the lack of comprehensive data, the World Bank indicated that it has enough information to declare that the first Millennium Development Goal of halving global poverty was met in 2010.

    The World Bank’s numbers for country and global poverty matter. They can affect the allocation of aid dollars. They frame the scale of the poverty challenge confronting each country and more broadly the global development community. They define areas of focus: Africa, fragile states and lagging regions in middle-income countries. And they are used to justify funding expansions of the concessional windows of the multilateral banks and capital increases for these aid agencies.
    Given all this, it is surprising that the empirical basis for country and global poverty numbers is rather weak. Taking the bank’s figures at face value also implies that we have to believe the following:

    These three examples highlight three key difficulties in making global poverty estimates.

    First, it is impossible to say anything meaningful about poverty in a country without having a household survey to explain how income (or consumption) is distributed among its people. The World Bank gets around this by making the assumption that any country with no survey has the same poverty rate as the average for its region. This leads to the peculiar result of North Korea being assigned essentially the same poverty rate as China, from whom it regularly receives food aid. (China dominates the East Asia Pacific regional poverty rate because of its vast population).

    Thankfully, the number of countries for which no household surveys exist is shrinking—a result for which the World Bank deserves some credit given its push for greater coverage and its technical support to countries administering surveys. Nevertheless, those countries that remain without a survey—a group which includes Burma, Zimbabwe and Somalia—are unsurprisingly among those where one would suspect poverty levels are especially high. While small as a share of the global population, these countries may contain a significant share of the world’s global poor.

    What is more, the increased coverage of surveys flatters to deceive. Of the 49 countries in sub-Saharan Africa, a seemingly credible 43 have a survey. Yet only half of these countries have undertaken a new survey in the past six years. Of the 386 million people who are estimated to live on under $1.25 a day in the region, a third are derived by extrapolating from surveys dating from 2005 (the year of the bank’s last global poverty estimate) or earlier.

    Second, surveys need to be reasonably accurate and representative if they are to be used as a basis for estimating poverty. However, the World Bank uses household surveys as an article of faith, even when the data is at odds with other sources of information.

    Unfortunately, inconsistencies between surveys and other data are not uncommon. The case of India is the most famous and widely studied, and also the most important for global poverty numbers by virtue of India’s population size. Survey numbers suggest that the average Indian consumed $720 per year in 2010, while the country’s national income accounts indicate that household expenditure was about two-and-a-half times greater, at $1,673 per person per year. As one might expect, such a discrepancy has dramatic implications for India’s poverty estimates—a difference in the order of hundreds of millions of people.

    Which figure should be believed? Relying blindly on survey data, the World Bank must conclude that growth in India’s household expenditure per capita has been only 1.5 percent per year since the country embarked on its celebrated economic reform program in the early 1990s. This also implies hardly any acceleration from India’s pre-reform period when surveys reported an equivalent growth rate of 1.1 percent.

    By contrast, the corresponding data from national accounts have household expenditure per capita averaging 4.5 percent growth a year over the past two decades, and show a clear break from the period before the reforms when it averaged 1.6 percent a year. The survey data not only deny the impact of India’s economic reforms, but reject the existence of an emergent middle class. According to the survey, less than 1 percent of Indians make it into the ranks of the global middle class, with consumption above $10 a day.

    There even seem to be discrepancies within the survey data itself. Survey data show that meat consumption in rural India has grown at a rate of 4.8 percent a year since just prior to the reforms, while fruit and vegetable consumption grew by 3.2 percent. These trends do not seem compatible with overall consumption growing at 1.5 percent per year.

    The third difficulty with generating global poverty data revolves around the use of Purchasing Power Parity (PPP) estimates. PPP estimates are used to convert survey data, measured in local currency, into globally-comparable data that takes into account cost-of-living differences between countries. The current estimates are drawn from a global exercise conducted in 2005.

    The trouble is that for some countries, most notably China, the PPP conversions have little credibility. China did not permit a random sample of locations from which to survey prices, as was done in other countries. Instead, China restricted data collection to a few urban areas. When the results came in, a few eyebrows were raised: China’s prices were 40 percent higher than what had previously been thought, meaning that Chinese living standards were revised downwards by about 40 percent. If one takes this at face value, along with Chinese growth rates, it would mean that China in 1981 would have been as poor as the poorest country in the world today (except perhaps the Democratic Republic of the Congo), making its consequent economic transformation all the more dramatic. By the bank’s count, China has 173 million poor people consuming less than $1.25 a day. But if the PPP conversion rate was changed back to where it used to be, the poverty estimate would be cut to 69 million.

    The World Bank’s global poverty estimates extend over nearly three decades, with its earliest estimates provided for the year 1981. Throughout this period, the global headcount (based on the $1.25 poverty line) has been dominated by three population groups: Sub-Saharan Africa, India and China. These three account for a remarkably constant three-quarters of the world’s poor—a share which has never deviated by more than three percentage points on either side. Yet poverty estimates for each of the three suffer from glaring problems: insufficient survey data, flawed surveys, and faulty PPP conversions, respectively. If we cannot believe the poverty estimates for Sub-Saharan Africa, India and China, then we cannot believe the World Bank’s global estimates, and we must admit that our knowledge of the state of global poverty is glaringly limited.

    Calculating poverty numbers requires making many assumptions and the World Bank should be commended for making its methodology (and data) available in a transparent fashion. But one should not take the bank’s final figures at face value; there are too many discrepancies with common sense.

    We are not ready to believe that North Korea has the same poverty rate as China; that India only has a middle class of 9 million people; or that China was destitute in 1981. Poverty numbers are too important a target for global development to be left in their current state. Isn’t it time for the development community to organize itself to resolve these contradictions?