Research @WorldBank.org

Please refer to PovcalNet in the subject line of your email.

Add "*.worldbank.org" to your bowser's "Pop-up blocker settings|Allowed sites" list

Enable your browser's active scripting

Poverty Measures for the Developing World
In assessing poverty in a given country, and how best to reduce poverty, one naturally focuses on a poverty line that is considered appropriate for that country. But how do we talk meaningfully about "global poverty?" Poverty lines across countries vary in terms of their purchasing power, and they have a strong economic gradient, such that richer countries tend to adopt higher standards of living in defining poverty (Ravallion, et al., 2009). But to consistently measure global absolute poverty in terms of consumption we need to treat two people with the same purchasing power over commodities the same way—both are either poor or not poor—even if they live in different countries

This note summarizes the methods used by the World Bank in measuring global poverty, and by PovcalNet in implementing those methods.

Methods used for international poverty measurement
To measure poverty in the developing world as a whole, the World Bank’s “$1 a day” measures apply a common standard, anchored to what “poverty” means in the world’s poorest countries. The original “$1-a-day” line was based on a compilation of national lines for only 22 developing countries, mostly from academic studies in the 1980s (Ravallion, et al., 1991). While this was the best that could be done at the time, the sample was hardly representative of developing countries even in the 1980s. Since then, national poverty lines have been developed for many other countries. Based on a new compilation of national lines for 75 developing countries, Ravallion, Chen and Sangraula (RCS) (2009) proposed a new international poverty line of $1.25 a day. This is the average poverty line for the poorest 15 countries in their data set.

Multiple poverty lines should be used to test the robustness of global poverty comparisons. Chen and Ravallion (2010) used five lines: (i) $1.00 a day; (ii) $1.25; (iii) $1.45, as obtained by updating the previous (1993) line for inflation in the US; (iv) $2.00, which is the median of the RCS sample of national poverty lines for developing and transition economies and is also approximately the line obtained by updating the $1.45 line at 1993 PPP for inflation in the US; and (v) $2.50, twice the $1.25 line, which is also the median poverty line of all except the poorest 15 of countries in the RCS data set of national poverty lines.

Since 2015 October World Bank uses the new $1.9/a day poverty line at the 2011 PPP to estimate global poverty rates using available household survey data from all countries. The new $1.9/a day poverty line is the average of the same 15 national poverty lines that yielded the $1.25 line in 2005 PPPs (Ferreira et al., 2015).

The international poverty line at PPP is converted to local currencies in 2011 price and is then converted to the prices prevailing at the time of the relevant household survey using the best available Consumer Price Index (CPI) (see details about CPI in “what is new” section). (Equivalently, the survey data on household consumption or income for the survey year are expressed in the prices of the ICP base year, and then converted to PPP $’s.) Then the poverty rate is calculated from that survey. All inter-temporal comparisons are real, as assessed using the country-specific CPI. Interpolation/extrapolation methods are used to line up the survey-based estimates with these reference years from 1981 to 2013.

Note that global poverty rates are based on the international poverty line of $1.9 /a day in 2011 PPP at 2011 prices and cannot be directly compared with national level poverty rates, which are derived using country specific poverty lines estimated in local currencies.

To compare the number of poor people across countries and compute regional aggregates, country estimates must be “lined up” first to a common reference year, interpolating for countries in which survey data are not available in the reference year but are available either before, after, or both. The more survey data are available (that is, the more data for different years), the more accurate the interpolation.

The process requires adjusting the mean income or expenditure observed in the survey year by a growth factor to infer the unobserved level in the reference year. Thus, two assumptions are required to implement this process: distribution-neutral growth and a real rate of growth between the survey and reference year.

Distribution-neutral growth implies that income or expenditure levels are adjusted for growth assuming that the underlying relative distribution of income or expenditure observed in survey years remains unchanged. Under this assumption, it is straightforward to interpolate the poverty estimate in a given reference year implied by a given rate of growth in income or expenditure. Rates of change in real consumption per capita should be based on the change in real consumption measured by comparing country survey data across different years. In practice, however, survey data in most countries are not available on an annual basis. Therefore, the change in private consumption per capita as measured from the national accounts is used instead. While, there can be no guarantee that the survey-based measure of income or consumption change at exactly the same rate as private consumption in the national accounts, this appears to be the best available option.

When the reference year falls between two survey years, an estimate of mean consumption at the reference year is constructed by extrapolating the means obtained from the surveys forward and backward to the reference year. The second step is to compute the headcount poverty rate at the reference year after normalizing the distributions observed in the two survey years by the reference year mean. This yields two estimates of the headcount poverty rates in the reference year. The final reported poverty headcount rate for the reference years is the linear interpolation of the two. When data from only one survey year are available, the reference year mean is based on the survey mean by applying the growth rate in private consumption per capita from the national accounts. The reference year poverty estimate is then based on this mean and on the distribution observed in the one survey year. The better data coverage is in terms of number and frequency of available surveys, the more accurate this lining-up process is and the more reliable the regional estimates will be.

The aggregate headcount index for a region is the population-weighted mean of the headcount indices across the countries in that region. The number of poor in each region is the product of the region’s headcount index and total regional population. (See point 6 in the next section.) This assumes that the poverty rate for a country without a household survey is the regional average. The steep rise in food prices in 2008 has been taken account by re-weighting the CPI whenever possible to accord with the food share in a neighborhood of the poverty line. We have done this for as many countries as these data are available, and where food price increases much faster than the national inflation rate.

Take China, for example. The average food spending is about 37% of total spending in urban areas and 43% in rural. But for the poor these figures are about 48% in urban and 68% in rural. After re-weighting the CPI, the inflation rate for the poor is about 32% between 2005 and 2008, instead of 13% using the unadjusted CPI. If we used the latter then we would overestimate the extent of poverty reduction by 20+ million people.

The underlying data

  1. Data sources: The distributional data used here are drawn from nationally representative household surveys, which are conducted by national statistical offices or by private agencies under the supervision of government or international agencies and obtained from government statistical offices and World Bank Group country departments.
  2. Price indices: the best available CPI for the country with documentation is used. A spatial price index (SPI) has only been used when the base is national average and consistent over time with documentation.
  3. Micro data (household level data) vs. grouped data: recently access to micro data has improved dramatically. Primary distributions in PovcalNet are calculated from household level data directly. Few of them are from the country’s statistical office.
  4. The data base in PovcalNet is primarily a set of grouped tabulations in a harmonized data structure. Parametric Lorenz Curves are used for the estimation purposes when they are expected to provide close estimates to the micro data. However, when the parametric models are unlikely to work well (notably when using incomes with a bunching up at zero) PovcalNet estimates directly from the micro data.
  5. The same PPPs are used to convert the international lines to local currency units (LCUs). Three countries were treated differently, China, India and Indonesia. In all three we used separate urban and rural distributions. For China, the ICP 2011 survey was with larger coverage than 2005 ICP which was confined to 11 cities. ICP 2011 rounds for China is national however it is hugely urban biased and the cost-of-living is lower for the poor in rural areas (Chen and Ravallion, 2010) still holds. We treat the ICP PPP as an urban PPP for China and use the ratio of urban to rural national poverty lines to derive the corresponding rural poverty line in local currency units (Chen and Ravallion 2010). For India, the ICP included rural areas, but they were underrepresented. We derived urban and rural poverty lines consistent with both the urban-rural differential in the national poverty lines and the relevant features of the design of the ICP samples for India; further details can be found in Ravallion (2008c). For Indonesia, we converted the international poverty line to LCUs using the official consumption PPP from the 2011 ICP. We then unpack that poverty line to derive implicit urban and rural lines that are consistent with the ratio of the national urban-to-rural lines for Indonesia.
  6. The regional and global population may differ from the current WDI, because PovcalNet uses the classification of countries based on whether they are IBRD eligible or recent graduated When the user adds up the populations of the countries in a region in PovcalNet, the sum is less than the regional total since the regional total in PovcalNet also includes countries without a household survey or missing the PPP or other data.
  7. The per capita income/consumption used in PovcalNet is household income/consumption expenditure dividing by the household size.

Data availability
The PovcalNet has income/consumption distributional data from about 1200 household surveys spanning 1979-2015 and 138 developing countries. More than 2 million randomly sampled households were interviewed in these surveys, representing 96 percent of the population of developing countries. Not all these surveys are comparable in design and sampling methods. Non representative surveys, though useful for some purposes, are excluded from the calculation of international poverty rates.

The World Bank produced its first global poverty estimates for developing countries for World Development Report 1990 using household survey data for 22 countries. Since 1979 there has been considerable expansion in the number of countries that field such surveys, and the number of data sets within two years of any given year rose dramatically, from 26 between 1979 and 1984 to 256 between 2000 and 2004, then 366 from 2005 to 2010, and 143 from 2011 to 2015.

Data coverage is improving in all regions, but the Middle East and North Africa and Sub-Saharan Africa continue to lag. The database, maintained by PovcalNet team in the World Bank’s Development Research Group, is updated several times a year as new survey data become available, and a major reassessment of progress against poverty used to make about every three years until 2008, from 2010 onwards such major updates is assessed every year. A complete overview of data availability by year and country is available at: PovcalNet.

Poverty measures are available for 138 countries while inequality measures are available for 159 countries.

International poverty estimates are available for IBRD eligible countries (includes recent graduated) only. Some high-income economies also report poverty indicators, but the $1.9 day poverty line is not relevant. The estimation of the global poverty headcount assumes that nobody lives below the $1.90 a day in high income countries. Although there are a number of people with household incomes below $1.90 per person in rich countries, estimated per capita consumption is above this threshold for nearly everyone. Countries of this type can't not be used in aggregation. Please note the poverty gap and poverty gap square may be unusually large due to the negative income.

The lag between the reference year and actual production of data series depends on the availability and reliability of the household survey for each country. Lag between the latest available year for aggregate estimates and the actual production year is about 3 years.

Data limitations
No data are ideal. International comparisons of poverty estimates entail both conceptual and practical problems that should be understood by users.

An important step in the process of compiling global poverty estimates is the conversion of the $1.90 a day international poverty line into respective national currency units. PPP exchange rates, such as those from the International Comparison Program or the Penn World Tables, are used because they take into account the local prices of goods and services not traded internationally. Although PPP rates were designed for comparing aggregates from national accounts, they were not intended for making international poverty comparisons. PPPs are based on prices of goods and services that may not be representative of the consumption baskets of the poor, so they may not fully reflect the relative price level faced by very poor consumers. As a result, there is no certainty that an international poverty line measures the same degree of need or deprivation across countries. Similarly, the poverty line may need to be adjusted for different locations (such as urban and rural areas) within the country, if prices or access to goods and services differs. However, for most of countries, this information is not available.

Discrepancy between the national accounts and household surveys also make the poverty estimates more difficult. There is no reason why these sources would agree closely on consumption, as they are not strictly measuring the same thing. But large discrepancies are still of concern, as they may reflect measurement errors.

There are also problems with comparability of surveys, both over time and across countries. Household survey questionnaires can differ widely, and similar surveys may not be strictly comparable because of differences in quality. These problems are diminishing as survey methods improve and become more standardized, but achieving strict comparability is still impossible. Under-reporting of income and selective compliances are other sources of measurement errors, these problems are unlikely to be distribution neutral.

Francisco H. G. Ferreira, Shaohua Chen, Andrew Dabalen, Yuri Dikhanov, Nada Hamadeh, Dean Jolliffe, Ambar Narayan, Espen Beer Prydz, Ana Revenga, Prem Sangraula, Umar Serajuddin and Nobuo Yoshida, 2016, " Global Count of the Extreme Poor in 2012:Data Issues, Methodology and Initial Results”,The Journal of Economic Inequality”, Vol. 14(2), 141-172, 2016.

Chen, Shaohua and Martin Ravallion, 2010, "China is Poorer than we Thought, but no Less Successful in the Fight Against Poverty," in Debates on the Measurement of Poverty, Sudhir Anand, Paul Segal, and Joseph Stiglitz, ed. (Oxford, UK: Oxford University Press).

Shaohua Chen and Martin Ravallion, 2011, "The Developing World is Poorer than we Thought, but no less Successful in the Fight against Poverty," Quarterly Journal of Economics, Vol. 125, Issue 4, pp. 1577-1625.

Martin Ravallion, 2008, "A Global Perspective on Poverty in India," Economic and Political Weekly October 25, Vol.43, No. 43, pp.31-37.

Martin Ravallion, Shaohua Chen, and Prem Sangraula, 2009, "Dollar a Day Revisited," World Bank Economic Review, Vol. 23 , pp. 163-184.

Martin Ravallion, Gaurav Datt, and Dominique van de Walle, 1991, "Quantifying Absolute Poverty in the Developing World," Review of Income and Wealth Vol. 37, pp. 345-361.