Time series of global poverty: A note on sources

This page provides additional information on the Global Inequality Project timeseries of extreme poverty from 1981 to 2011. The timeseries relies on recent research on the share of people with incomes below the ‘cost of basic needs,’ defined as the price of purchasing 2,100 calories per day, essentials nutrients, plus a modest quantity of housing, clothing, and fuel for home heating and lighting.

For the year 2011, we rely on estimates of the extreme poverty rate developed by Robert C. Allen (2020). Allen uses price data from the 2011 round of the International Comparisons Programme (ICP) to estimate the price of subsistence goods. He then applies this poverty line to the World Bank’s household consumption surveys.

For the period from 1981 to 2008, we rely on research by Michail Moatsos (2021), published by the OECD. The Moatsos dataset is somewhat less robust than Allen’s figure for 2011. Because ICP price data is not readily available for the earlier period, Moatsos relies on the October Inquiries carried out by the International Labour Organization (ILO). However, the October Inquiries lack price data for several commodities (especially non-food items) and do not cover all countries in all years. Moatsos overcame these gaps in the data using a simple model where unobserved prices are estimated as a function of observed prices and the level of GDP per capita, and by extrapolating to missing years using the Consumer Price Index (CPI). Moatsos then applies this price data to the World Bank’s household surveys.

It is worth noting that Moatsos also provides estimates from 1820 to 2018. However, the estimates outside of the 1981-2008 period suffer from several empirical limitations. The pre-1981 figures are based on income data derived from historical national accounts. These figures are not comparable to the World Bank’s household surveys, as national accounts do not adequately capture changes in non-commodity forms of consumption, such as subsistence farming and access to commons. The post-2008 estimates are also problematic because ILO price data is not available after that year in most cases, and Moatsos assumes basic prices moved in line with the CPI. We therefore only use Moatsos’ dataset in the years from 1981 to 2008, where most countries are covered by household consumption surveys and direct price data. For more information on these issues, see footnote 2 in this recent paper in New Political Economy paper, as well as Appendix I (1.1) in this paper from World Development.

A significant limitation of combining the Allen and Moatsos datasets is that, because they rely on different underlying assumptions about prices, there is often a discrepancy between their estimates of the poverty rate. This graph compares the two datasets in the year 2011. By toggling the graph, you can instead compare Allen’s 2011 figure to Moatsos’ 2008 figure, which is the final datapoint based on direct price estimates for most countries.

The graph shows that for some countries, such as Guinea, Bangladesh, and Kenya, Moatsos’ approach yields higher figures, while for others, including Madagascar, Mozambique, and Yemen, Allen’s estimates are higher. These differences mostly get cancelled out at the global level.

We have chosen to combine the Allen and Moatsos dataset because they are the best available estimates of extreme poverty for the periods they cover. However, given the discrepancies between these two datasets, the timeseries we present should be treated with caution. It is possible that some of the change between 1981 and 2011 is influenced by the change in the dataset rather than actual changes in global living standards. Future research is necessary to overcome this problem.

ve this text inline or in the module Content settings. You can also style every aspect of this content in the module Design settings and even apply custom CSS to this text in the module Advanced settings.