Policy Research Working Paper 11137 Global Poverty Revisited Using 2021 PPPs and New Data on Consumption Elizabeth Foster Dean Jolliffe Gabriel Lara Ibarra Christoph Lakner Samuel Tetteh-Baah Development Data Group Development Research Group & Poverty and Equity Global Department June 2025 Policy Research Working Paper 11137 Abstract Recent improvements in survey methodologies have low-income countries, increases by around 40 percent to increased measured consumption in many low- and $3.00 when the more recent national poverty lines as well lower-middle-income countries that now collect a more as the 2021 purchasing power parities are incorporated. comprehensive measure of household consumption. The net impact of the changes in international prices, the Faced with such methodological changes, countries have poverty line, and new survey data (including new data for frequently revised upward their national poverty lines to India) is an increase in global extreme poverty by some 125 make them appropriate for the new measures of consump- million people in 2022, and a significant shift of poverty tion. This in turn affects the World Bank’s global poverty away from South Asia and toward Sub-Saharan Africa. The lines when they are periodically revised. The international changes at higher poverty lines, which are more relevant to poverty line, which is based on the typical poverty line in middle-income countries, are mixed. This paper is a product of the Development Data Group, Development Research Group, Development Economics and the Poverty and Equity Global Department. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://www.worldbank.org/prwp. The authors may be contacted at efoster1@worldbank.org; djolliffe@worldbank.org; glaraibarra@worldbank.org; clakner@worldbank.org; stettehbaah@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Global Poverty Revisited Using 2021 PPPs and New Data on Consumption Elizabeth Foster ⓡ Dean Jolliffe ⓡ Gabriel Lara Ibarra ⓡ Christoph Lakner ⓡ Samuel Tetteh-Baah * JEL classification: I32 - Measurement and Analysis of Poverty Keywords: Global Poverty, Purchasing Power Parities, Poverty Lines, Consumption Surveys, International Poverty Line * The author ordering was constructed through American Economic Association’s randomization tool (confirmation code: 3qWkFBbuER0p). Corresponding author: clakner@worldbank.org. All authors are with the World Bank. The authors are thankful for helpful comments and guidance from François Bourguignon, Andrew Dabalen, Francisco Ferreira, Deon Filmer, Haishan Fu, Indermit Gill, Aart Kraay, Peter Lanjouw, Luis Felipe Lopez-Calva, Nora Lustig, Branko Milanovic, Umar Serajuddin, and Nishant Yonzan. Special thanks also go to Marko Rissanen and Mizuki Yamanaka for their help with the PPPs, and to Daniel Mahler for his advice regarding various aspects of the paper. Many thanks are also due to World Bank poverty economists and members of the World Bank’s Global Poverty Working Group (GPWG) for many helpful discussions. The authors gratefully acknowledge financial support from the UK government through the Data and Evidence for Tackling Extreme Poverty (DEEP) Research Programme. 1. Introduction The global poverty measures produced by the World Bank use purchasing power parities (PPPs) to account for differences in price levels across the world. These PPPs are periodically revised in light of new data on relative living costs. This paper presents new poverty lines for monitoring global poverty using the most recent release of 2021 PPP data. In addition to updated PPPs, new survey data and new information on countries’ own assessments of poverty are also incorporated. The inclusion of new surveys of consumption expenditure from low- and lower-middle-income countries has important implications for the estimation of the new lines. Due to improvements in survey methodologies and questionnaire design, these surveys collect significantly more consumption, thus increasing measured mean consumption (Beegle et al., 2012; Caeyers et al., 2012; Mahler et al., 2024; World Bank, 2024a), not dissimilar to the rebasing of gross domestic product (GDP). 2 The estimates presented in this paper are impacted by these changes in two ways. First, these new surveys significantly increase measured mean consumption in several countries, thus reducing global poverty (when assessed against a poverty line fixed in real terms). The most important case pertains to new survey data for India, which is included in the global poverty estimates with this update. India has changed its method of collecting consumption data from a unified recall of purchases over the last month to a questionnaire that varies the recall period depending on the frequency of purchase. This change alone reduces the extreme poverty rate in India from 23 to 16 percent in 2011/12 (using 2017 PPPs), the last year when both measures are available (also see discussion in World Bank, 2024a). Similar improvements in survey methodology occurred in Sub-Saharan Africa, for example, as part of a harmonized household survey program benefiting 13 West and Central African countries (Castaneda et al., 2024, 2022). Second, upward revisions of national poverty lines in low-income countries, which are a consequence of the same survey revisions, are causing an upward revision of the international poverty line, counteracting the first downward impact. The World Bank’s measure of extreme poverty has always been based on how some of the poorest countries in the world define poverty. It is important to recognize that countries’ poverty lines are inextricably linked to how they measure consumption. Notably, as many low-income West African countries more accurately measured actual food consumption and other essentials in recent years, they also followed the common practice to reconstruct their national poverty line, following a cost-of-basic needs approach (Ravallion, 1998). These collective revisions led to an upward revision in the international poverty line. As a result of the revisions to their national poverty lines, in these countries, national poverty rates only increased modestly or even decreased when the new survey methodologies were introduced, in contrast to large and consistent decreases in the international poverty measures which use an absolute standard fixed in real terms (Mahler et al., 2024). 2 Nigeria’s economy increased in size by 89% when it rebased its GDP in 2014 (Angrist et al., 2021; Mahler et al., 2024). 2 This paper estimates the international poverty line (IPL) – the World Bank’s threshold of monetary extreme poverty – as the median national poverty line of low-income countries following the same methodology applied to earlier sets of PPPs (Jolliffe and Prydz, 2016; Jolliffe ⓡ al., 2024). This line is estimated at $3.00 per person per day in 2021 PPPs, an upward revision from $2.15 with the 2017 PPPs. Most of this upward revision is explained by revisions in the underlying national poverty lines rather than a change in prices (which reflect both general inflation between 2017 and 2021, and changes in relative prices across countries). This reflects significant improvements in the quality of consumption data, through improved survey methods, as already mentioned, but also more timely information on the cost of basic needs in the poorest countries of the world. For example, the majority of the national poverty lines used in the new IPL are from 2020 or more recent. A consistent methodology is used to update the other global poverty lines that are more relevant in middle-income countries: The poverty line typical of lower-middle-income countries is updated from $3.65 to $4.20, and from $6.85 to $8.30 in the case of upper-middle-income countries. In 2022, the estimate of global extreme poverty is revised upwards by 1.5 percentage points to 10.5 percent (an upward revision of about 125 million people). This overall effect is the combination of a downward revision from new survey data from India and an upward revision due to the new IPL. With the exception of South Asia, all regions see an upward revision in extreme poverty, especially Sub-Saharan Africa. As a result, the concentration of extreme poverty in Sub-Saharan Africa is becoming even more pronounced. This revision is relatively small compared to some prior updates. Most notably, the update of the dollar-a-day poverty line to $1.25 with the 2005 PPPs revised upwards the count of poor people by 400 million (Chen and Ravallion, 2010). These earlier estimates suffered from methodological changes to the PPPs as well as limited data on national poverty lines, both of which have been addressed since then. 3 This paper makes the case that it is important to use the most recent data on prices (i.e., PPPs), costs of basic needs (i.e., national poverty lines), and well-being (i.e., surveys) across countries to measure poverty. It would be inconsistent to use consumption aggregates from the most recent surveys without updating the international poverty line as one assesses poverty going forward. While the new international poverty line and the new surveys in many countries are 3 First, the World Bank’s line had limited statistical support, relying on the poverty lines of only 15 poor countries, excluding India, because it was richer, but became poorer because of the adoption of the 2005 PPPs (Deaton, 2010). Second, the methodology the ICP followed to produce the 2005 PPPs was problematic; some researchers have argued that the process of linking PPPs collected across countries using a limited “ring” list of items and countries, resulted in PPPs that were biased upwards, and that were driven by more internationally consumed items, which tend to be more expensive in poor countries and do not reflect the consumption patterns of poor people (Deaton, 2010). Over the years, both methodological issues have been resolved. The World Bank has improved the statistical support for its poverty line, by using a more complete list of low-income countries (Jolliffe ⓡ al., 2024), and the ICP has improved its methodology, stabilizing movements in PPPs between ICP rounds (Deaton and Schreyer, 2022). For example, since the 2011 ICP round, a “global core list” of items priced from all countries is used instead of the “ring” list, and an indicator variable that captures the “importance” of items priced in respective countries is incorporated into the estimation algorithm for the final PPPs. 3 closer to best practice, an important question is what to do with the earlier surveys, whose methods produced considerably lower measured consumption and other countries, that are still far from best practice. Without an adjustment, the historical data is likely to overstate poverty and thus the decline in extreme poverty from the 1990s to recent years will be overstated. While a comprehensive adjustment for all countries was not conducted, the paper includes an adjustment to the historical series for India to make it more comparable to the most recent survey data and new international poverty line (Alfani et al., 2025). In a more complete analysis including all countries with survey data, substantial differences in historical poverty series are observed at the country level that offset each other at the global level (Mahler et al., 2025). This paper proceeds as follows. Section 2 presents the data, and Section 3 estimates the global poverty lines and investigates the drivers of the changes in the poverty lines. Both default and alternative methods of setting the poverty lines are presented for robustness. Section 4 discusses changes in national poverty lines for low-income countries. Section 5 presents new, more comparable estimates of global poverty and by region over time. Section 6 concludes. 2. Data 2.1. Purchasing power parities (PPPs) PPPs are currency conversion factors that account for differences in price levels around the world. This paper analyzes the 2021 PPPs published by the International Comparison Program (ICP) in 2024, as well as relevant PPPs from earlier rounds (World Bank, 2024b). These include the 2017 PPPs and the revised 2011 PPPs, which were published in 2020. For most countries, PPPs are based on periodic price collections, which provide up-to-date information on price levels around the world. For countries that do not participate in the price collection, the ICP estimates PPPs from a cross-country regression, or if available, uses the benchmark data from a previous ICP cycle to extrapolate. Of the 172 countries that are included in our database of household surveys, 154 are covered by a price collection, which is an improvement from the previous 2017 PPPs that included 152 countries in this category. 4 As with previous PPP updates (Ferreira et al., 2016; Jolliffe ⓡ al., 2024), we assess whether the 2021 PPPs result in any extreme price changes when compared to the 2017 PPPs. Based on this assessment, we make exceptions for four countries (the Arab Republic of Egypt, Guinea, São Tomé and Príncipe, and Sudan). Instead of the PPPs published by the ICP, we use the geometric average of the official PPP (based on a price collection) and a 4 Six countries (Guatemala, Kosovo, Lebanon, South Sudan, Syrian Arab Republic, Uzbekistan) participated in data collection and compilation in the 2021 cycle, but not in 2017, while four countries (Barbados, Haiti, the Islamic Republic of Iran, Myanmar) dropped out to become non-benchmark countries. For the countries that participated in 2021 but not in 2017, relatively large PPP movements are to be expected since there is a change in methodology. For more details on the specific methodological changes by country, see Appendix A (discussion around Table A2). For the four new non-benchmark countries, the 2021 PPPs are based on extrapolations of the revised 2017 PPPs. 4 regression-based PPP. Our methodology for determining the exceptions follows what was done for the 2017 PPPs, when exceptions were made for twice as many countries (Jolliffe ⓡ al., 2024). 5 A comprehensive review of the 2021 PPPs, as well as a table of the final PPPs used is available in Appendix A. 6 As in previous PPP rounds, an important criterion for identifying countries with extreme price changes is the delta ratio. The delta ratio (as defined by Ferreira et al., 2016) is a function of PPPs and Consumer Price Indexes (CPIs) between 2017 and 2021, and can be also interpreted as the factor that converts welfare in 2017 PPPs to welfare expressed in 2021 PPPs. 7 The PPPs use the United States as the numeraire, so its PPP conversion factor is always 1. Thus, the delta ratio for the US is its rate of inflation between 2017 and 2021, which is 11%. For all other countries, the delta ratio represents the change in the value of the domestic currency between 2017 and 2021 in PPP terms, relative to the value of the US dollar (Figure 1). Countries with delta ratios above 1.11 have an appreciation of their currency in PPP terms (i.e., one dollar will buy more goods and services in the domestic economy in 2021 than in 2017). Though conceptually different indices, CPIs and PPPs are expected to display similar trends over time. 8 That nearly all countries lie within two standard deviations of the US delta ratio is evidence of the stability of the ICP between the 2017 and 2021 rounds, as well as the reliability of the 2021 PPPs for global poverty monitoring (Figure 1). Large deviations of a country’s delta ratio from the US delta ratio reflect inconsistencies in the movement of CPI and PPP data. When using the officially published 2021 PPPs, the four exception countries (Egypt, Guinea, São Tomé and Príncipe, and Sudan) have delta ratios that are clearly outliers and more than two standard deviations from the US delta (Figure 1). 9 The variance of the delta ratio declines somewhat with GDP per capita (Figure 1). The average delta ratio is 1.13, meaning that nominal PPP-denominated welfare increases by 13% between 2017 and 2021. Grouping by income classification shows a small income gradient with low-income countries having an average delta of 1.21 compared with 1.11 for high-income countries (Table C1). When accounting for population size, the average delta ratio is slightly larger (1.15) and 5 Exceptions were made for 8 countries in the 2017 cycle (Egypt, Guinea, Nigeria, Belize, São Tomé and Príncipe, Sudan, Iraq, and Trinidad and Tobago). There were 6 exception countries in the 2011 cycle (Egypt, the Lao People’s Democratic Republic, the Republic of Yemen, the Syrian Arab Republic, Myanmar, and Iraq). 6 Additional analysis is done, including with metadata on PPPs and CPIs to investigate possible issues with the PPPs or CPIs, akin to the methods used for assessing the 2017 PPPs in Jolliffe ⓡ al. (2024). For some countries, especially Pacific Island nations, the ICP has not published a PPP in either the 2017 or 2021 cycle. Many of these countries participated in a price collection in the 2011 round. For these countries, the revised 2011 PPPs have been extrapolated to 2021. 7 It is a reformulation of the ratio of actual to extrapolated PPPs (Deaton and Aten, 2017). 8 However, differences in scope, basket composition and methodology, make it impossible that full convergence is achieved and some level of discrepancies are expected (for more details, see Jolliffe ⓡ al., 2024). At a fundamental level, CPIs are temporal price indices, while PPPs measure spatial price differences at a point in time. 9 It is important to note that Figure 1 assesses the official 2021 PPP against the 2017 PPP used by the World Bank for global poverty monitoring. All four outlier countries were exceptions previously, so the 2017 PPP in Figure 1 for these countries is not the official 2017 PPP. For example, Egypt has a delta ratio of 1.14 when the official 2017 and 2021 PPPs are compared (see Appendix A). 5 there is a slightly larger income gradient (Figure C1, Table C1), which is related to a relatively large delta ratio value of 1.21 for India. As we will see further below, this explains why the 2021 PPPs imply a slightly pro-poor change in the global welfare distribution and a reduction of global inequality. Figure 1: Real Changes in PPP-Adjusted Dollars between 2017 PPPs and 2021 PPPs (Delta Ratio) Note: The figure plots the change in PPP-based welfare between the 2017 and 2021 cycles against GDP per capita. For the average country, the delta ratio is 1.13, meaning that nominal PPP-denominated welfare increases by 13% between 2017 and 2021. For the US, the delta ratio is 1.11, which by definition equals US inflation between 2017 and 2021. The total number of observations is 148, including only countries which have price-collection-based PPPs in both 2017 and 2021 ICP cycles, and only countries with survey data in the Poverty and Inequality Platform. See Appendix A.1 for the definition of the delta ratio. The graph uses the 2017 PPP used in PIP (Jolliffe ⓡ al., 2024), which might be different from the official 2017 PPP. 2.2. Household income and consumption surveys The paper uses more than 2,400 income and consumption surveys included in the World Bank’s Poverty and Inequality Platform (PIP), collected across 172 countries in the world, since 1981 and 6 covering more than 97% of the world’s population (World Bank, 2025). 10 The surveys capture household income or consumption per capita per day in local currencies, which are deflated to 2021 prices using national CPIs, and then the 2021 PPPs are applied to express welfare in PPP- adjusted USD. To maximize global coverage, the database needs to make two assumptions. First, we mix income and consumption surveys. Income surveys dominate in Latin America, Europe and Central Asia, and high-income countries, while in the rest of the world, we mostly rely on consumption surveys. Second, since not all countries collect annual surveys, we create a balanced panel by interpolating and extrapolating the mean consumption or income of the existing surveys using growth rates in national accounts and assuming no further distributional changes. The June 2025 PIP update, which we rely on largely in this paper, includes new data from 74 country-years, revisions to previously published data and updates to auxiliary data from the World Development Indicators (e.g., population data that are used as weights for country-level poverty estimates when estimating global poverty) (Alfani et al., 2025). Due to the unavailability of recent survey data (in 2020 or later) for several countries, including Nigeria, less than half of the population in Sub-Saharan Africa is covered by survey data in recent years, so the estimates for that region are subject to additional uncertainty. For all other regions, as well as the world as a whole, more than half of the population is covered by a recent survey. The most consequential change on the survey data side with this update is the incorporation of India’s Household Consumption Expenditure Survey 2022/23 in PIP. Besides the new survey round, the existing 2011/12 survey has been revised from a Uniform Reference Period (URP) to a Modified Mixed Reference Period (MMRP) method for creating the household welfare aggregate (World Bank, forthcoming). 11 The URP method follows a survey instrument that asks how much households spent on all food and non-food items in the past month, whereas the MMRP method uses a shorter recall period for food and frequently consumed non-food items and longer recall periods for items that are purchased more infrequently. Studies have shown that methods like the MMRP tend to capture more measured consumption (World Bank, 2024a). For example, the shift to the MMRP aggregate reduces extreme poverty (as measured using 2017 PPPs and a $2.15 line) from 22.9 to 16.2 percent in 2011/12. As noted earlier, these kinds of breaks in consumption methodology are not uncommon. Given India’s share of the global population, its methodological changes matter for the global poverty 10 We primarily use the September 2024 and June 2025 vintages of PIP. These survey data are compiled and harmonized by World Bank teams and form the basis of country dialogue on poverty and inequality trends. The collection of microdata is known as the Global Monitoring Database. For advanced economies, these surveys are supplemented with information from the Luxembourg Income Study (LIS). World Bank (2025) has more information on the PIP methodologies. 11 The 2022/23 survey only collects the MMRP, while the 2011/12 round collects both aggregates. Until this update, PIP had used the URP method for 2011/12 and the entire historic series. For comparability, the 2011/12 survey round now uses the MMRP, which significantly increases measured consumption and reduces poverty in India in 2011/12. In addition, the spatial and temporal deflation has been improved. Before 2011/12, only the URP aggregate is available; to avoid a large artificial reduction in poverty, the entire historic series for India has been adjusted using the relationship between the two aggregates in 2011/12 (Alfani et al., 2025). 7 trends reported in this paper, such that extra efforts have been made to adjust the entire historic series to the new survey methodology (Alfani et al., 2025). Mahler et al. (2025) adjust all non- comparable historic series for all countries in the PIP database to improve comparability within countries, and show that the implications on the aggregate numbers are relatively small. 2.3. National poverty lines The poverty lines used by the World Bank for measuring global poverty have always relied on national poverty lines that reflect countries’ own definitions of poverty. Specifically, the international poverty line that captures extreme poverty has been based on the national poverty lines of the poorest countries in the world. This paper builds on a database of national poverty rates maintained at the World Bank and updated from Jolliffe ⓡ al. (2024). For low- and middle-income countries, national statistical offices (NSO) and official poverty reports are usually the main sources of data on national poverty rates sanctioned by governments. These data are compiled in data repositories at the World Bank, such as the Poverty and Equity Briefs Hub, PIP and the World Development Indicators (WDI). For high-income countries, national poverty rates are usually obtained from the European Statistical Office (Eurostat), and the Organization for Economic Co-operation and Development (OECD). A few lines were obtained directly from NSOs (see Appendix B). National poverty rates are more readily available than national poverty lines, since the former are more prominent in policy discussions. 12 The national poverty rates are converted to harmonized national poverty lines by querying the inverse cumulative distribution functions in PIP (which is expressed in PPP terms). In other words, the harmonized national poverty line is the level of consumption or income that yields the national poverty rate in the welfare distribution in PIP. This approach was originally proposed by Jolliffe and Prydz (2016) and also followed for the 2017 PPPs (Jolliffe ⓡ al., 2024). Countries make different assumptions when setting their poverty lines. For example, some countries may express poverty lines in per capita terms while others use per adult-equivalent terms. By querying the (inverse) PIP distributions, this approach ensures that the harmonized national poverty lines are all expressed in daily per capita 2021 PPP dollars. A total of 1,747 poverty lines form the basis of the global poverty lines. Around 200 national poverty rates cannot be matched to a distribution in PIP, so we are unable to construct a harmonized national poverty line for those observations. See Appendix B for further details. 12 We refer here to poverty estimates that are based on a methodology that has been officially adopted by a government, or to estimates that have been cited in documents published by the government (for instance, National Statistical Offices annual reports, National Development Plans, etc.). 8 3. Setting poverty lines 3.1. Main approach The main specification uses one harmonized poverty line per country. The line for each country that is closest in time to 2021, the PPP benchmark year, is selected. The selected lines are categorized by World Bank income group, using the income classification in the year the poverty line was calculated. The median values of the harmonized poverty lines by income group are chosen as the global poverty lines, rounded to the nearest $0.10. For example, the international poverty line of $3.00 is defined as the median harmonized poverty line of low-income countries (Table 1 and Figure 2). The corresponding lines using lower- and upper-middle-income countries are $4.20 and $8.30, respectively. Table 1: Summary of poverty lines 2017 ICP cycle 2021 ICP cycle Change (%) Income group Year Median Mean Obs Year Median Mean Obs Median Mean Low-income 2014 2.15 2.42 28 2020 3.04 3.00 23 41 24 Lower-middle 2015.5 3.63 3.95 54 2019 4.20 4.71 53 16 19 Upper-middle 2017 6.85 7.06 37 2021 8.29 8.64 45 21 22 High-income 2017 24.36 23.36 38 2021 29.43 27.49 42 21 18 Total obs 157 163 Note: Median and mean values are expressed in daily per capita PPP dollars. The column for Year indicates the median survey year. The results for the 2017 ICP cycle are reproduced from Jolliffe ⓡ al. (2024) for comparison. The table shows the overall changes between ICP cycles, which reflect changes in PPPs, national poverty lines, countries’ income classification, etc. Compared to the 2017 PPPs, the new lines are derived from a broader sample of countries and more recent data (Table 1). In total, the national poverty lines now cover 163 countries compared with 157 in the previous round. The timeliness of data improved especially for low-income countries where the median national poverty line is based on data from one year before the PPP benchmark year, compared with three years previously. Put differently, more than 50 percent of countries had a survey in 2020 or later. 13 The more recent poverty lines from low-income countries, particularly from West Africa, are significantly higher than the earlier lines (see Figure C2 in Appendix C). As a result, the value of the international poverty line increases by around 40 percent, significantly more than the other lines, and more than the average increase in PPP- denominated welfare (mean delta of 1.13) or US inflation (1.11). The timeliness of data matters for the global poverty lines set by the World Bank, especially for the international poverty line. A new national poverty line becomes available when new survey data is released and a national poverty line has been calculated based on information from that 13 See Figure C2 in the Appendix for the national poverty lines underlying the old and new international poverty line. 9 survey. 14 Most low- and lower-middle-income countries lack annual household surveys, which implies that their national poverty lines tend to be more outdated. To show this impact, we rederive the World Bank’s poverty lines for the 2017 PPPs using the current database of national poverty lines and surveys (Table C2 in Appendix C). The results suggest global poverty lines of $2.50 (compared with $2.15), $3.80 (compared with $3.65) and $7.00 (compared with $6.85). This means that if more timely survey data had been available in 2022 at the time the international poverty line was being updated with the 2017 PPPs, the line would have been $2.50, which is 16 percent higher than the currently used value of $2.15. Furthermore, in this scenario, the increase in the international poverty line when switching to the 2021 PPPs would be considerably smaller (an increase of 22 percent, similar to the other lines). Figure 2: Graphical representation of global poverty lines Note: This figure plots national poverty lines expressed in daily per capita 2021 PPP dollars against daily per capita GDP in 2015 constant US dollars of the year with the majority of survey data collection. Each marker represents one of the 163 national poverty lines underpinning the results presented in Table 1. The horizontal lines indicate the median poverty line of the corresponding income group, rounded to the nearest 10 cents. Both axes are on a log scale. The income classifications are based on gross national income (GNI) per capita, which implies that there is not a one-to- one correspondence between GDP per capita and income classifications. 14 In some countries, new data are released, but the national poverty line has not been updated. For example, for India and the Syrian Arab Republic, we include new survey data in this update, but the national poverty rates are still from the previous survey. 10 This comparison of data vintages already shows the important impact of revisions in the underlying national poverty lines on the extreme poverty line, something that the paper explores further below through decomposition exercises. Overall, more timely data would reflect higher lines, even for the same PPP benchmark year. This means that the poverty line for lower-middle- income countries, currently set at $4.20 in 2021 prices, which is based on lines that are less recent than the other lines, would likely be higher if more timely data were available for lower- middle-income countries. 3.2. Alternative approaches Alternative approaches of setting global poverty lines are explored to investigate the robustness of the lines to different assumptions and methods. In the 2017 PPP update, the international poverty line was stable at the median, offering robustness to choosing different country groupings. For example, the median poverty line would have remained at $2.15 when choosing any of the countries ranked between the 11th and 41st poorest. In this update, the distribution is rather bimodal, with some low-income countries clustering at a lower poverty line of around $2.00 or a higher poverty line of around $3.40 (Figures C2 and C3.b in Appendix C). Furthermore, the cumulative distribution function of the distribution of national poverty lines among LICs is flatter now at the median, suggesting that a shift in any one country could change the line by a larger amount than before. Given this new distribution of poverty lines, we check for robustness by using the Harrell-Davis quantile estimator in addition to the simple median in the baseline. The Harrell-Davis quantile estimator is the weighted average of all the order statistics (Harrell and Davis, 1982). It smooths the quantiles and is thus more robust to large jumps in the quantiles. The Harrell-Davis median for the international poverty line in 2021 PPP dollars is $3.01, which is within a few cents of the simple median (Figure C3.b in Appendix C). We re-estimate the global poverty lines, using both the simple and Harrell-Davis median, for a wide range of different samples (full results are included in Table C3 in Appendix C) that can be generally grouped in two sets. Both sets of results show that the baseline results have statistical and methodological support, especially when restricting the sample to recent years. In the first set of robustness checks, we use one line per country as in the baseline, but restrict the samples based on the PPPs. For example, we include only benchmark PPPs, where the PPP is based on a price survey. We replace the exceptional PPPs (see Section 2) with the official PPPs. We also drop countries whose lines are older than 5, 10 or 20 years from 2021. The global poverty lines are robust to these variations. Across the various options in this first set of robustness checks, the extreme poverty line ranges between $2.92 and $3.04, considering both types of medians. The corresponding ranges for the other two lines are [$4.11, $4.31] and [$8.14, $8.46]. The baseline values of $3.00, $4.20 and $8.30 fall centrally within these ranges. 11 In the second set of robustness checks, we use all 1,747 lines instead of only one line per country. An important concern is that the number of lines differs drastically across countries, reflecting largely the availability of household surveys. For example, Costa Rica has a poverty line for every year in the period 1989-2023, whereas Angola has only two poverty lines for 2008 and 2018. To ensure even representation across countries, we weight the lines such that they sum to 1 for each country. We consider two types of weights: i) giving equal weight to all lines for a country, and ii) a triangular weight that gives a greater weight to lines closer to 2021. However, in either specification, an issue remains that more recent lines for a country with many lines would get a lower weight than an old line for a country that only has one line. Therefore, we also obtain results by additionally restricting the sample by age of the line. For the low- and upper-middle-income countries, the pooling approach tends to produce lower lines than the baseline (the ranges are [$2.66, $3.02] and [$7.47, $8.10]), while for the lower- middle-income countries it is centered around the baseline values (the range is [$4.12, $4.52]). When restricting the sample to pooling over the last 5 years, we return to the baseline values for the low-income and the lower-middle-income countries. For the extreme poverty line, this confirms the key finding in the paper, that there was an important increase in the national poverty lines of the poorest countries between the last 5 and 10 years. 15 The result for lower-middle- income countries is somewhat mixed since there is quite a lot of movement in income classification for this group. 16 In the case of the upper-middle-income countries, even using only the last 5 years gives a line of $8.10 (using the preferred triangular weight), which is lower than the baseline. A potential explanation is that countries in this group use relative lines that increase with a country’s average income. Therefore, incorporating lines that are even 5 years older gives a lower poverty line. 17 For reference, we have also updated the 2017 PPP values with US inflation and the delta ratio for each income group. The latter would give lines of $2.59, $4.06 and $7.69, which are substantially lower than the baseline, especially for the low-income (lower by 14%) and upper-middle-income countries (lower by 7%). This also suggests that factors other than price changes between 2017 and 2021 affect the value of these lines. The next sub-section discusses these factors in more detail. 15 This effect is not visible in the first set of robustness checks when we restrict to 5, 10 or 20 years (rows 6-8 in Table C3). The reason is that this robustness check still uses only one line per country, the closest to 2021. In the pooled sample, multiple lines are used for each country. 16 The lines are classified according to the country’s income classification in the year of the line. For example, the historic sample of lower-middle-income countries (LMIC) would include lines for countries that are now upper-middle- income countries (UMIC). At that time these countries were towards the richer end of lower-middle-income countries and thus also had typically higher lines. 17 Furthermore, the data availability around 2021 is not symmetric. For example, this update adds several surveys from 2023, but more recent data are rare. 12 3.3. Decomposing changes in the global poverty lines There are four sources of changes in the global poverty lines between the 2017 and 2021 ICP rounds: (i) price changes from changes in PPPs and CPIs; (ii) the underlying national poverty lines; (iii) the income classification of countries; and (iv) the number of countries with available data. Some of these factors are jointly determined; for example, as a few countries graduate to a higher income status, the composition of countries and poverty lines in each income group changes. The impacts of these factors will be evaluated sequentially following the same approach used for the analysis of the 2017 PPPs (Jolliffe ⓡ al., 2024). These kinds of decompositions are path-dependent; other approaches would yield slightly different results. Table 2 presents the sequential changes in the global poverty lines and the relative shares of different sources in the second panel. We start by replicating the 2017 PPP exercise using the corresponding national poverty lines, the 2017 PPPs, income classifications and sets of countries. When updating PPPs, while holding constant everything else, the $2.15 international poverty line increases to $2.46. Next, the international poverty line increases to $3.42 when we update the national poverty lines from the circa 2017 lines (which were available when the 2017 PPPs were adopted) to the circa 2021 lines selected in the current update. This is the most important source of change in the international poverty line. This impact of the new lines is much higher among low-income countries than other groups, even high-income countries where poverty is measured using relative lines. Table 2: Sources of changes in global poverty lines Low-income Lower-middle Upper-middle High-income Total Description Median Obs Median Obs Median Obs Median Obs obs Lines in 2017 PPPs 2.15 28 3.63 54 6.85 37 24.36 38 157 Update PPPs 2.46 28 3.99 54 7.68 37 26.66 38 157 Update poverty lines 3.42 28 4.23 54 8.29 37 30.82 38 157 Update income group 3.04 23 4.20 53 8.20 42 30.62 39 157 Add new countries 3.04 23 4.20 53 8.29 45 29.43 42 163 Total change ($) 0.89 0.57 1.44 5.07 Contributions PPP impact (%) 34 63 57 45 Line impact (%) 107 42 43 82 Income group (%) -42 -6 -6 -4 New countries (%) 0 0 6 -24 Total change (%) 100 100 100 100 Note: The upper section of the table records the median poverty when accounting for different sources of changes in the global poverty lines, one after the other. The lower section shows the relative contributions of these changes, by income group. The next step in the decomposition is to account for the reclassification of countries across income status over time. For example, the lines for Azerbaijan, Benin, Nepal, Uzbekistan, and Zimbabwe were included in the previous set of 28 low-income countries, but are now classified as middle-income thanks to the availability of survey data concurrent with their latest income 13 classification. 18 When these countries are excluded from the sample of low-income countries, the median poverty line reduces to $3.04. For all income groups, changing income status has a negative impact on the poverty line. Until this point in the decomposition, we only included countries that were in the original 2017 PPP sample, changing sequentially their PPP, their national poverty line and their income group. The final change is the addition of new countries that were not available for the 2017 PPP global lines. Three countries are added to both the upper-middle- and high-income groups. In summary, while the increase in the international poverty line is largely driven by new poverty lines from low-income countries, for the lower-middle-income and upper-middle-income line, both factors matter, with the PPP effect being somewhat larger. Table C4 in Appendix C includes an additional decomposition, which confirms this pattern for the extreme poverty line, while being difficult to interpret for the other lines. Section 4 takes a closer look at why the poverty lines in low-income countries changed so much over this period. It argues that this change does not reflect an increase in their aspirations or the standards by which they define poverty. Rather, the new lines are driven by the adoption of improved survey data and methodologies, which give a better measure of the cost-of-basic-needs in the poorest countries of the world. 3.4. Other poverty lines and parameters In addition to the international poverty line and the higher absolute lines, the World Bank uses other poverty lines and parameters that are expressed in PPP terms. These include the societal poverty line (SPL) (Jolliffe and Prydz, 2021), the Prosperity Gap standard (Kraay et al., 2023), and the bottom censoring threshold (Yonzan et al., forthcoming). In line with earlier methodologies, these parameters are all updated with the mean value of the delta ratio (1.13); that is, the factor that converts income or consumption from 2017 PPP dollars to 2021 PPP dollars (see Table 3). Table 3: Updating other poverty lines and parameters Parameter Value (2017 PPP $) Value (2021 PPP $) Societal poverty line (intercept) 1.15 1.30 Prosperity Gap standard 25 28 Bottom-censoring threshold 0.25 0.28 Note: These parameters are updated from 2017 PPP dollars to 2021 PPP dollars using the mean delta ratio of 1.13. 3.4.1. Societal poverty line : The SPL is a weakly relative poverty line, combining elements of absolute and relative poverty into a country-year-specific poverty line. For the 2017 PPPs, the 18 The income classification refers to the year of the poverty line selected. It is not the case that Azerbaijan was a low- income country in 2017; in fact, it was an upper-middle-income country in both 2017 and 2021. However, in the 2017 PPP exercise, the country’s poverty line was from 2001, when Azerbaijan was low-income, while we now use a line from 2021. 14 SPL is given as: max(2.15, 1.15 + 50% of the median), where the median is country-year-specific (in PPP dollars). The SPL increases with median income above some threshold, while incorporating a floor at the IPL. For updating the SPL to 2021 PPPs, we posit that the slope coefficient should not be impacted by the change in PPPs, since it is not expressed in monetary units (see Jolliffe ⓡ al. (2024) for a similar argument). The intercept term is updated by the average change in incomes, namely the mean delta ratio, which results in a new intercept of $1.30 (see Table 3 above). This results in an SPL of max(3.00, 1.30 + 50% of the median) in 2021 PPP dollars. This formulation for the SPL is confirmed when we rederive it using the same methodology applied with the 2011 and 2017 PPPs. The parameters of the SPL were originally estimated from a cross- country ordinary least squares (OLS) regression of national poverty lines against median income (Jolliffe and Prydz, 2021), similar to Figure 2. The model is specified as follows: = ̂ , � + where represents the harmonized national poverty line per capita per day, is median income or consumption per capita per day, � is the estimated intercept and ̂ is estimated slope coefficient. Following Jolliffe ⓡ al. (2024), the preferred specification includes only countries with absolute national poverty lines, which yields estimates of the slope coefficient and intercept that are not significantly different from 0.5 and 1.30, respectively (see Table C5 in Appendix C). 19 3.4.2. Prosperity Gap m easure : The World Bank’s new measure of shared prosperity, the Prosperity Gap, is the average factor that converts everyone’s income to a prosperity standard. The prosperity standard was set at $25 in 2017 PPP. The prosperity standard was set with two considerations in mind (Kraay et al., 2023): It is approximately the median poverty line of high- income countries ($24.4 in Jolliffe ⓡ al. (2024)) in 2017 PPPs. It is also close to mean per capita income of countries that reach high-income status. Updating the analysis with the 2021 PPPs yields a high-income poverty line of $29.40 (Table 1, Figure 2). High-income countries define poverty using relative poverty lines, which increase in real terms as income increases over time. To keep the prosperity standard fixed in real terms, we have updated it only for price changes across countries between 2017 and 2021, i.e. the mean delta, which results in a prosperity standard of $28 with the 2021 PPPs (Table 3). 3.4.3. Bottom -censoring threshold : The distributions that the World Bank includes in PIP are bottom-censored at $0.25 in 2017 PPPs. Yonzan et al. (forthcoming) show that values below the minimum are too low for anyone to survive and are more likely to represent measurement error. Therefore, any observed value below this level is replaced with $0.25. We update the bottom- censoring threshold to $0.28 with the 2021 PPPs (Table 3), which accounts for price changes as captured by the mean delta ratio. 19 This holds in the unconstrained regression, as well as a version of the regression where we assume that the slope coefficient is 0.5 (see Table C5 in the Appendix). 15 4. Understanding changes in national poverty lines for low- income countries The global poverty lines are often described as absolute poverty lines, so one would expect that the adoption of new PPP data would lead only to nominal price changes. However, as the paper has shown, the changes in the global poverty lines, especially the IPL, are driven by revisions in the underlying national poverty lines. These changes could arise from real changes in how countries define their poverty lines (e.g., expanding the definition of basic needs as economies grow or changing the target number of calories), and/or measurement issues arising from the adoption of improved survey data (e.g., using a recall period that increases measured nonfood consumption, including imputed rent, more comprehensive lists of nonfood items and durable goods). Several pieces of evidence on the surveys and poverty lines used in this update suggest that the global poverty lines are largely driven by the latter, especially in the low-income countries that underpin the IPL. Over the last few decades, countries around the world have improved how they measure consumption. The World Bank has contributed to this effort by providing technical assistance and financing primarily to low- and middle-income countries. The World Bank’s Living Standards Measurement Study (LSMS) program helped refine survey methodologies, for example through methodological experiments that try to identify the most effective methods to collect consumption data (Oseni et al., 2021). Of particular relevance to this paper is the Harmonized Surveys on Household Living Conditions (EHCVM) program, which was launched in 2016 by the West African Economic and Monetary Union (WAEMU) Commission and supported by the World Bank. It includes the eight WAEMU member countries (Benin, Burkina Faso, Côte d’Ivoire, Guinea-Bissau, Mali, Niger, Senegal and Togo), while Guinea and Chad also benefited from technical assistance. The program provided extensive technical assistance and support to NSOs to implement new household surveys, following closely the methodologies recommended by the LSMS program. The first round of the survey program was carried out in 2018/19, followed by a second round in 2021/22. In addition to the 10 countries that were part of the original program, the model was also adopted for new surveys in Central African Republic (2021), Cameroon (2021), and the Republic of Congo (2022, data not yet available). 20 A majority of West and Central African countries now follow this model and a third round (2025/26) is underway with at least 18 countries participating. For most countries, the program led to substantial changes and improvements in the questionnaire used to collect consumption data, as well as the methodology used to construct the total consumption aggregate for each household. Thus, each country also constructed a new national poverty line. In terms of low-income countries, which underpin the IPL, the program 20Burundi also borrowed many aspects of the model for its 2020 survey, although it is not part of the harmonization program. 16 resulted in eight low-income countries with substantially improved measures of household consumption and corresponding new national poverty lines. 21 All the countries constructed new poverty lines via a cost-of-basic needs approach. This means constructing a food basket reflecting the actual dietary habits of the population (or a subset) which provides a target number of calories and calculating its cost. This is the food component of the basket. The impact of more comprehensive surveys on the cost of food component can go either way, as it is the average cost per calorie and the target number of calories that determine its cost. A nonfood component is then added to arrive at the national poverty line, using some variant of the method proposed by Ravallion (1998), which makes use of how households themselves balance food and nonfood needs. If more of households’ actual nonfood consumption is captured (through more comprehensive or better questionnaires), the use of this method means the national poverty line will naturally increase, leaving the national poverty rate roughly unchanged. In this respect, the new lines are the correct ones to use with the new aggregates, and doing so will result in the most accurate understanding of the evolution of poverty (Lanjouw and Lanjouw, 2001). The use of this methodology to set the nonfood component likely led to an increase in the nonfood component and thus the national poverty line for many countries with new surveys that are relevant for this discussion. For the most influential countries constructing new poverty lines, the change is not so much driven by adding completely new categories (e.g., the older surveys and poverty lines already included health expenses, the use values of durable goods and housing costs, including for owners), but rather by having more detailed lists of nonfood items and durable goods, and more appropriate recall periods (in particular, moving away from diaries and “average monthly” recall periods which have been shown to underestimate true consumption). The EHCMV program also harmonized across countries several of the methodological decisions that are involved in defining cost-of-basic-needs poverty lines. For example, this includes the reference population whose specific food consumption patterns are used to set the food basket, the choice of target calorie value, and whether it is expressed in per capita or per adult equivalent terms, and (to some extent) which of the range of values proposed by Ravallion is chosen for the nonfood component. The harmonization resulted in an increase for some countries and a decrease for others, but the net impact of these types of changes on the median national poverty line of low-income countries is likely an increase, due in particular to two countries (Central African Republic and Togo) which moved from a very low calorie per adult equivalent target to the standard 2300 per capita EHCVM target. 22 In addition, Burkina Faso made a methodological 21 These countries include Burkina Faso, Chad, Central African Republic, Guinea, Guinea-Bissau, Mali, Niger, and Togo. Although Guinea is now a lower middle-income country, it was a low-income country in 2018/19 which is the date of the most recent national poverty line. 22 Both Central African Republic and Togo adjusted household consumption by the number of adult (defined as 19 or 20 to 50) male equivalents, using a scale based on calorie requirements by age group and sex. The food basket should thus provide enough calories for an adult male, however values of only 2,100 and 2,400 calories, respectively, were used (the 10th edition of the Recommended Daily Allowances gives a requirement of 2,900 calories per day for a male aged 19 to 50). In their latest poverty lines, both countries move to the standard of 2,300 calories per capita, effectively much higher given the share of women and children in the population. 17 change from using the low bound suggested by Ravallion for the nonfood component to the average of the lower and upper bounds. Finally, even a cost-of-basic-needs poverty line has a relative component. For most countries, the latest poverty lines were constructed 5 to 10 years later than those underpinning the previous international poverty line. 23 Over such periods, these countries could have seen improvements in living standards and thus diets of any given decile of the population, resulting in a higher cost per calorie of the typical diet and thus a higher food poverty line. Similarly, the share of nonfood consumption would have increased. This means that when there are such improvements in living standards, a newly constructed cost-of-basic-needs will be a bit higher than simply updating the previous one for inflation, even without any change to the questionnaire or methodology. The question then becomes whether the countries that underpin the IPL have seen improvements in living standards between the 2017 and 2021 PPP rounds. When there are changes in survey methodology, the growth in mean consumption between the two survey rounds does not provide information on real changes in living standards. Using instead GDP per capita shows that the low- income countries in our sample have not seen a significant improvement in living standards over this period. In fact, the mean GDP per capita declined by almost 9 percent, while median GDP per capita has remained virtually unchanged. 24 Of the 28 low-income countries that formed the basis of the IPL of $2.15 (2017 PPP), five are no longer low-income and thus no longer considered. 25 Of the remaining 23, seven have no new national poverty line available, so the same line is used, updated to 2021 PPP USD. For one of these, South Sudan, a regression-based PPP was used in the 2017 ICP cycle, while the PPP in the 2021 ICP cycle is based on actual price data, so there is a methodological change in the harmonized national poverty line. Haiti, which is the only low-income country outside Sub-Saharan Africa in the sample, saw a change from a benchmark PPP to an extrapolated PPP. Five countries conducted a new survey round in an existing series, producing comparable poverty lines just updated for the increase in the cost of living (sometimes using the official CPI resulting in no change in the harmonized national poverty line as we consider it, or sometimes using an alternative method resulting in a small real change). Eight countries benefited from the EHCVM survey harmonization program and adopted that model, and Burundi adopted many aspects outside of the official program. The final two countries saw changes in their survey methodologies, but they do not follow the EHCVM model or adhere as closely to the LSMS recommendations. 23 And in some cases, even the previous line was simply the result of updating a line constructed years earlier for inflation. In the case of Mali, the poverty line was based on one constructed 20 years previously in 2001. 24 Mean GDP per capita for the country-years underpinning the $2.15 line is $2,436, compared with $2,223 for the $3 line (all figures in 2021 PPPs). The median values are $2,318 and $2,353, respectively. When looking only at the countries that updated their poverty lines, the mean annual growth in GDP per capita between the years of the old and new national poverty line was 1.06 percent in the EHCMV countries and 0.95 percent among the remaining low-income countries. For comparison, lower-middle-income countries that have updated their national poverty line record a mean annual growth rate of 1.24 percent, and the full sample of 101 countries from all income groups has a mean annual growth rate of 1.37 percent. 25 Of these five, two had lines significantly below the median, two significantly above, and one close to the old line ($2.01) so their removal has little impact on the median line. 18 The countries that were part of the EHCVM program saw the largest increases in their national poverty lines (Figure 3). They also cluster around the middle of the distribution in both ICP rounds, so their changes have a direct impact on the median for low-income countries (although the mean also results in an IPL of $3.00 with the 2021 PPPs). The countries with comparable poverty lines across the two ICP cycles line-up closely around the 45-degree line, which means that the poverty lines have evolved roughly in line with price changes (without any increase in real terms). In sum, the 23 countries that were part of the IPL sample are roughly evenly split between increases and decreases, with the exception of the EHCVM countries that show large increases driven by methodological improvements. In addition to the changes in methodology already mentioned, the previous line for Guinea Bissau was somewhat arbitrarily set at $2 per day in 2010, while the new line is constructed via the standard cost-of-basic-needs approach, using actual data on per calorie costs and consumption patterns from the survey. Figure 3: Changes in national poverty lines between 2017 and 2021 PPPs Note: For the 23 low-income countries that were included in the IPL sample in both the 2017 and 2021 PPP exercises, the graph plots the change in their harmonized national poverty line. All national poverty lines are expressed in 2021 PPP dollars, so they can be compared. All countries on the dotted 45-degree line do not have a real change in their poverty lines. All countries lying to the left and above the 45-degree line have a real increase in their poverty line, typically those in the Harmonized Surveys on Household Living Conditions (EHCVM) program. Those countries lying to the right and below the 45-degree line have a real decrease in their poverty line. 19 5. Results 5.1. Overall global poverty trends The adoption of the new global poverty lines (based on the 2021 PPPs and updated national poverty lines) results in revisions to global and regional poverty trends and levels. For each of the three poverty lines, three series are reported; namely, a series from the September 2024 PIP data vintage, using the 2017 PPPs; and two series from the June 2025 PIP data vintage, using both 2017 and 2021 PPPs. The June 2025 PIP data vintage includes new survey data, including the new data for India. The overall changes in poverty estimates—from the 2017-PPP-based estimates in the September 2024 PIP data vintage to the 2021-PPP-based estimates in the June 2025 PIP data vintage—are driven by a combination of different factors; not only switching PPPs, but also the inclusion of new survey data into PIP and revisions to national poverty lines underlying the global poverty lines (more details below). Results are presented for the period from 1990 to 2030, the target date for the SDGs. From 1990 to 2023, the available survey data offer good population coverage. 26 Beyond this period, we also report nowcasts and projections using growth in national accounts. Substantial changes are observed at the IPL compared to the higher lines, especially for historical years (Figure 4). The differences in extreme poverty are the net effect of two main factors: (1) a significant downward revision from the new data for India (shown by the difference across the two vintages using the 2017 PPPs) and (2) an upward revision in the real value of the international poverty line. Given that the latter dominates the former, extreme poverty estimates are revised upwards over the entire period under review. The extreme poverty rate for the world in 1990 is revised upwards by 6 percentage points to 44 percent, whereas in 2022 it is revised upwards by only 1.5 percentage points to 10.5 percent. The global incidence of extreme poverty is thus higher than previously thought, though progress has also been faster. Put differently, between 1990 and 2022, an estimated 1.5 billion people escaped extreme poverty compared to the previously estimated 1.3 billion. The historical revision is primarily driven by East Asia & Pacific (see Figure C4 in Appendix C) and in particular China. 27 26 While survey data are available before 1990, coverage is more limited. The year 2023 (2022) is the most recent year with sufficient population coverage in the June 2025 (September 2024) vintage. Comparisons across the two vintages are done for 2022, which is the most recent year with sufficient coverage in both vintages. The coverage criteria means that at least 50 percent of the global population, as well as 50 percent of the population of low- and lower-middle- income countries, is represented by survey data that are no more than three years old. We also report nowcasts to the current year and projections to 2030 which require additional assumptions. For the September 2024 vintage the nowcasts cover 2023 and 2024; for the June 2025 vintage, the poverty estimates for 2024 and 2025 are nowcasts. 27 Between 1990 and 2012, the growth incidence curve for China is upward sloping. The upward revision in the extreme poverty line classifies additional Chinese people as poor, who had faster growth than the ones previously classified as poor. Therefore, progress is even faster when the higher line is used. 20 Figure 4: Global poverty trends, 1990 – 2030 Note: This figure shows global poverty trends using two vintages of PIP data, namely the old September 2024 vintage and the current June 2025 vintage. Poverty is estimated from both PIP data vintages using the $2.15, $3.65, and $6.85 poverty lines, expressed in 2017 PPP terms. Poverty is also estimated from the current vintage of PIP data using the $3.00, $4.20, and $8.30 poverty lines, expressed in 2021 PPP terms. With the September 2024 PIP data, poverty estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, poverty estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color.The figure also includes projections to 2030 (also indicated in grey color) that rely on national accounts growth, similar to the nowcasts. 21 By 2030, 9 percent of the world’s population is projected to be living in extreme poverty with the 2021 PPPs (Figure 4), compared to the previous projection of 7.3 percent (World Bank, 2024a). This means that the SDG target of eradicating extreme poverty, as well as the World Bank’s goal of 3 percent by 2030, are further out of reach. The 2017 PPPs result in a lower level of extreme poverty in 2030 (7 percent), but it is still more than double the World Bank’s target. Over the last decade, global progress in the reduction of extreme poverty has stalled, as the concentration of extreme poverty has been moving from South Asia to Sub-Saharan Africa, where economic growth is relatively low, and population growth is relatively high (World Bank, 2024a). 28 This slowdown is projected to continue until 2030. The results are mixed at the higher poverty lines. At the poverty line typical of lower-middle- income countries, historical poverty rates are virtually unchanged, while more recent estimates have been revised downwards. In this case, the impact of new surveys, including for India, dominates the impact of switching to a new poverty line. As mentioned in Section 3.1, the revision from $3.65 to $4.20 is roughly in line with the average revision to welfare levels, so PPPs by themselves are not expected to lead to large changes in the poverty rate. 29 At the poverty line typical of upper-middle-income countries, the impact of the upward revision to the poverty line dominates the impact of new survey data. Despite progress in reducing the poverty rate at this line, the number of people in poverty has not changed since 1990. With these revised estimates of global poverty, an estimated number of 3.8 billion of the world’s population would be considered as living in poverty by the standards of upper-middle-income countries in 2022, slightly above 3.7 billion in 1990. With the 2017 PPPs, the corresponding numbers as reported previously were 3.6 billion in 2022, slightly below 3.7 billion in 1990. 5.2. Decomposing changes in absolute poverty in 2022 In 2022, the estimate of global extreme poverty is revised upwards by 1.5 percentage points to 10.5 percent (an upward revision of 125 million people) (Table 4). Holding fixed the 2017 PPPs while adopting the June 2025 vintage, would have yielded an extreme poverty rate of 7.7 percent (a reduction of 101 million people). This reduction is largely due to increased consumption recorded in survey data from India (reducing extreme poverty by 125 million people, see Table C6), slightly offset by increased poverty rates in Sub-Saharan Africa, mainly due to new survey data. 30 The adoption of the 2021 PPPs (together with the associated international poverty line) 28 Similarly, the impact of India’s new data on extreme poverty (i.e., the difference between the two vintages reported for the 2017 PPPs) becomes smaller towards 2030, since the country’s contribution to global extreme poverty is projected to decline. The pattern is somewhat different at the lower-middle-income line. 29 The lower-middle-income poverty line increases by 16 percent, compared with the average delta of 13 percent. The national poverty lines from lower-middle-income countries are on average also older (median year of 2019, see Table 1), which would suggest that the line is somewhat underestimated. 30 In these countries, new surveys replace extrapolations from older surveys that were used in the September 2024 PIP data vintage. The observed revisions reflect these kinds of changes, as well as updates to the auxiliary data such as the national accounts growth rates. Table C6 in the Appendix provides the estimates for the ten countries with the largest changes. The change in the Middle East and North Africa is explained primarily by a revision to growth estimates for Yemen. 22 by themselves would have increased the global count of extreme poor by 226 million, increasing extreme poverty in all regions, especially in those that have large populations close to the extreme poverty line. The net effect is smaller, as new survey data from India partially offsets the increase from the new international poverty line of $3.00. Table 4: Changes in regional and global estimates of extreme poverty, 2022 Poverty rates (%) Changes in millions of poor Region $2.15 (Old) $2.15 $3.00 New PIP data Switching PPPs Total East Asia & Pacific 1.0 0.9 2.5 -1 34 34 Europe & Central Asia 0.5 0.6 1.1 0 2 3 Latin America & Caribbean 3.5 3.6 5.2 1 10 11 Middle East & North Africa 6.1 6.6 8.5 3 8 11 Rest of the World 0.6 0.7 0.7 0 0 1 South Asia 9.7 3.1 7.3 -126 82 -45 Sub-Saharan Africa 37.0 38.2 45.5 22 89 111 World 9.0 7.7 10.5 -101 226 125 Note: This table shows regional and global poverty levels in 2022 using two vintages of PIP data, namely the old September 2024 vintage and the June 2025 vintage. Poverty is estimated from both vintages using the $2.15 international poverty line, expressed in 2017 PPP terms. Poverty is also estimated from the June 2025 vintage using the $3.00 international poverty line, expressed in 2021 PPP terms. The impact of new PIP data is captured by the difference in millions of poor when comparing the old and new PIP vintages at the 2017 PPPs. This includes new surveys incorporated into PIP but also updates to national accounts data, revisions to population data and consumer price indices. The impact of switching PPPs is captured by the difference in millions of poor when comparing poverty at the 2017-PPP-based poverty lines and 2021-PPP-based poverty lines, using the June 2025 PIP data vintage. The total impact is the sum of the two, or alternatively, the difference between September 2024 vintage with the 2017 PPPs and June 2025 vintage with the 2021 PPPs. Table C6 provides this decomposition for the ten countries with the largest changes. The transition to the 2021 PPP-denominated dollars (i.e., the upward revision in global poverty by 226 million people) combines two elements, which can be further decomposed: On the one hand, the update of the PPP conversion factors; on the other hand, the upward revision in the national poverty lines that underpin the international poverty line. Updating the PPPs while holding constant the set of national poverty lines would yield $2.50 (2021 PPP) as the international poverty line (see Table 2). Only incorporating this PPP effect—namely, new information about relative price differences across countries—would imply a downward revision to global extreme poverty by 22 million people (Table 5). Incorporating the latest national poverty lines from low-income countries would increase the global count of extreme poor by about a quarter of a billion. 23 Table 5: Decomposing the change from switching PPPs, 2022 Poverty rates (%) Impact on millions of poor Region $2.15 $2.50 $3.00 New PPPs New lines Total East Asia & Pacific 0.9 1.3 2.5 8 26 34 Europe & Central Asia 0.6 0.7 1.1 1 2 2 Latin America & Caribbean 3.6 3.8 5.2 1 9 10 Middle East & North Africa 6.6 6.4 8.5 -1 9 8 Rest of the World 0.7 0.7 0.7 0 0 0 South Asia 3.1 2.8 7.3 -6 88 82 Sub-Saharan Africa 38.2 36.2 45.5 -25 114 89 World 7.7 7.4 10.5 -22 248 226 Note: This table shows poverty rates estimated from the June 2025 PIP vintage at three poverty lines, namely $2.15 (2017 PPP), $2.50 (2021 PPP), and $3.00 (2021 PPP). $2.50 would be the international poverty line in 2021 PPP terms if only PPPs were updated, holding constant the national poverty lines used in deriving the 2017-PPP-based international poverty line of $2.15. The value of $2.46 is given in Table 2, which is rounded to $2.50. If $2.50 were the new international poverty line, the estimated population in extreme poor would reduce by 22 million globally. At the two higher poverty lines, the total changes in poverty estimates in 2022 are larger than at the IPL (see Appendix tables for detailed results). 31 Measuring poverty at the $4.20 line yields a global poverty rate of 20.1 percent in 2022, which is 2.3 percentage points lower than the global poverty rate with the $3.65 line. This corresponds to 180 million fewer poor people by the standards of lower-middle-income countries, which is largely driven by the inclusion of new survey data from India. The switch to the 2021 PPPs and new line has a smaller impact. 32 On the other hand, the global poverty rate measured at the upper-middle-income line is revised up by 3.1 percentage points to 48.0 percent in 2022 (an increase of around a quarter of a billion people). This change is largely driven by the adoption of the new poverty line of $8.30 with the inclusion of new data playing a smaller role. 33 In summary, the adoption of the new data, new PPPs and new lines leads to a downward revision in global poverty at the lower-middle-income poverty line, and an upward revision at the extreme and upper-middle-income poverty lines. 31 A detailed analysis of these results is presented in Tables C6, C7, C8 and C9. Global and regional poverty rates with the new global poverty lines are provided in Table C10 and the corresponding population living in poverty is provided in Table C11. Figure C5 depicts the distribution of poverty across regions and countries at all three global poverty lines. 32 At this line, switching to the 2021 PPPs decreases poverty by 0.8 percentage points (or 71 million fewer poor people) (Table C8). Further decomposing this difference shows that the reduction from the new PPPs is somewhat larger (209 million) than the increase from the updated lines (139 million), resulting in a net reduction of 71 million (Table C9). The large impact of the new PPPs is driven by India, where poverty would drop by 119 million if the set of national lines were not updated (Table C7). This is because the poverty line would increase by 9.6 percent (from $3.65 to $4.00), while India’s consumption is revised up 21 percent (delta ratio of 1.21). 33 In contrast to the changes at the other two lines, poverty in India as measured by the $6.85 line increases by 52 million people when the new survey data are used (Table C6), which is one-fifth of the overall change. PPPs by themselves (while keeping national poverty lines unchanged) would have reduced global poverty by 67 million (Table C9). Updating to the new set of national poverty lines that underpin $8.30 leads to an upward revision of 229 million (Table C9). 24 5.3. Implications for the regional distribution of global poverty Viewed in isolation, the adoption of the new IPL of $3.00 (2021 PPP) revises up extreme poverty in all regions, including South Asia. However, the new survey data for India offsets this change leading to a net reduction for the region (Table 4). This implies a further redistribution of extreme poverty towards Sub-Saharan Africa, which is already the region where most of the extreme poor live. With the new data and 2021 PPPs, the share of the extreme poor in Sub-Saharan Africa in 2022 is revised up to 67 percent, compared to the 63 percent previously estimated (Figure 5). 34 Figure 5: Share of global poor in Sub-Saharan Africa, 1990 – 2025 Note: This figure shows the share of global poor living in Sub-Saharan Africa using two vintages of PIP data, namely the old September 2024 vintage and the current June 2025 vintage. Poverty is estimated from both PIP data vintages using the $2.15, $3.65, and $6.85 poverty lines, expressed in 2017 PPP terms. Poverty is also estimated from the June 2025 vintage of PIP using the $3.00, $4.20, and $8.30 poverty lines, expressed in 2021 PPP terms. With the September 2024 PIP data, poverty estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, poverty estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color. 34 The figure also shows that only including the new survey data for India, while maintaining an IPL of $2.15 (2017 PPP), would have increased the Sub-Saharan African share to more than three-quarters in 2022. 25 At the lower-middle-income line, Africa’s share is also revised up (from 44 to 48 percent in 2022), also caused by the large reduction of poverty in South Asia (Table C8). 35 By 2025, Sub-Saharan Africa is also expected to account for more than half of the global poor by this standard (Figure 5). At the highest absolute line, the revisions are less pronounced, since Sub-Saharan Africa accounts for a relatively small share at that level. Overall, with the new data, it appears that Sub-Saharan Africa is being left behind further as the world progresses. Much of this poor performance is driven by fragility. A growing share of the global extreme poor live in countries in Sub-Saharan Africa affected by conflict, especially in recent years (Figure C6). While the entire region accounts for around two-thirds of the global extreme poor, the fragile countries by themselves account for 40 percent. In recent years, the majority of the extreme poor has been living in middle-income countries. The redistribution of extreme poverty away from India (a lower-middle-income country) reduces that share, but the upward revision in the IPL counteracts this effect (Figure C7). Overall, the middle- income countries still account for slightly more than half of the extreme poor, but their share is on a downward trend. 5.4. Other indicators The societal poverty line is a relative poverty line with a floor at the extreme poverty line (see Section 3). While poverty rates according to a strongly relative poverty measure are not affected by a change in PPPs, results at the SPL are impacted through its intercept term as well as the floor. 36 With the new data, the societal poverty rate is revised up in the earlier period, but revised down in the last 20 years (Figure C8). In the earlier period, when a greater share of the global population was affected by the absolute section of the SPL, the results are thus closer to the upward revision to extreme poverty. However, in the last decades, the relative part of the SPL has become more important and thus changes in societal poverty rates are strongly influenced by changes in inequality within countries. The revisions in the last decades are driven by the data update rather than a change in PPPs. In particular, inequality in India decreases markedly with the adoption of new survey data; for example, in the 2011/2012 survey the Gini index changes from 35.4 to 28.8 between the September 2024 and June 2025 vintages. The results so far have focused on changes in global poverty at various poverty lines. We can also investigate the impact of adopting new PPPs and survey data on the global distribution, without setting a poverty line. The combined effect of 2021 PPPs and new survey data reduces the global Gini index by 1.1 points to 60.7 in 2022 (Figure C9). The PPP impact by itself results in a downward revision of 0.85 points in 2022, which is a relatively small shift compared to earlier 35 At the lower-middle-income line, the number of poor falls slightly in Sub-Saharan Africa, but the decline in South Asia is so large that Africa’s share still rises (Table C8). 36 Using the Ravallion and Chen (2011) terminology, a strongly relative poverty line is defined as a fraction of mean or median income, such as the 50 percent of median income used by the OECD. 26 changes in PPPs (Table C12). 37 Most of this reduction comes from a reduction in between-country inequality, which accounts for slightly more than two-thirds of global inequality. Almost all of the between-country change is due to the new PPPs, with the new survey data also slightly reducing between-country inequality. Within-country inequality also declined, driven by new survey data. In terms of contributions from individual countries, India is driving most of these changes: The new survey data results in lower inequality and a higher mean, and the 2021 PPPs also increase India’s mean consumption relative to the world’s average (e.g., India’s delta ratio is 1.21, compared with 1.15 for the average person). Like the global Gini index, the Global Prosperity Gap also takes into account both inequality between and within countries, but it also captures improvements in the average income or consumption. Similar to the results for the global Gini, the overall result is a downward revision in the Global Prosperity Gap over the entire period under review (Figure C10). In 2022, for example, everyone’s income in the global distribution will have to increase by a factor of 4.8 on average to attain the prosperity standard of $28 (2021 PPP), which is lower than the value of 5.1 that was previously reported with the 2017 PPPs. This is an encouraging result for the world, which appears better off than previously thought. When looking at the entire distribution, the shift to the 2021 PPPs leads to a pro-poor distributional change, which explains the decline in global inequality. Below the global median, the increase in (nominal) welfare is greater than the population-weighted average price change (Figure C11). In other words, in the bottom part of the global distribution, nominal welfare increases faster than the change in prices, implying a real increase in welfare from the change in PPPs. 6. Conclusion This paper has documented how the global poverty lines used by the World Bank have been updated with the 2021 purchasing power parities (PPPs), including the new international poverty line (IPL) of $3.00 (2021 PPP). The 2021 PPPs reflect the latest available information on how costs of living vary across more than 160 countries in the world. While earlier ICP cycles were often accompanied by changes in methodology, the last three cycles published for 2011, 2017, and 2021 share a similar methodology (World Bank, 2024b). As we have explained in the paper, the World Bank has also defined its poverty lines in the same way over this period, only updating for new data. The IPL of $2.15 (2017 PPP) or $3.00 (2021 PPP) is the median poverty line of low- income countries. Similarly, the median poverty line of lower-middle-income countries is $3.65 (2017 PPP) or $4.20 (2021 PPP), while the median poverty line of upper-middle-income countries is $6.85 (2017 PPP) or $8.30 (2021 PPP). Therefore, the changes observed in global poverty lines, and corresponding revisions to global poverty series, are not driven by methodological changes 37 For example, Lakner and Milanovic (2016) estimate that the shift from 2005 to 2011 PPPs reduced the global Gini by 3.5 points (while holding the data fixed). 27 in the measurement of global poverty, but data updates. Apart from updates to the PPP data, there are two other sources of data updates. First, new survey data have been included in the Poverty and Inequality Platform (PIP) in the June 2025 update (Alfani et al., 2025). Most importantly, the new India survey for 2022/23 has been added, which follows the Modified Mixed Reference Period (MMRP) methodology of recording household consumption. The entire series for India has been revised from the Uniform Reference Period (URP) to the MMRP methodology as much as possible. Under the URP method, households are asked how much they spent on all food and non-food items in the past month, whereas the MMRP method uses a shorter recall period for food and other frequently consumed items, and longer recall periods for items that are purchased less frequently. This switch results in higher measured consumption and lower poverty for a fixed line (World Bank, 2024a). Given its share of the world’s population, the new India series has resulted in significant improvements in measured well-being, not only for India but for the world. At a constant IPL of $2.15 (2017 PPP), the paper reports a drop in extreme poverty globally by 1.3 percentage point to 7.7 percent in 2022, largely driven by 125 million fewer extreme poor people in India. Second, the World Bank’s database of harmonized national poverty lines that form the basis of the global poverty lines has been updated from 1,381 observations (157 countries) to 1,747 observations (163 countries). The new database includes more recent information on national poverty standards across countries for circa 2021. This new information shows that the typical poverty line in low-income countries increased substantially. When compared to circa 2017 poverty lines, the median poverty line in this group of countries increased by around 40 percent, resulting in a marked increase in the IPL. At first sight, this appears surprising given that these countries use absolute poverty lines. Furthermore, over prior updates, such as the move from 2011 to 2017 PPPs, the IPL increased in line with average price movements (Jolliffe ⓡ al., 2024). The upward revision is driven primarily by a group of West African countries (accounting for a third of low-income countries) that have significantly improved their measures of household consumption, such as by expanding lists of nonfood items and adopting more appropriate recall periods, as in the case of India. When countries change their household survey methodology, they also need to redefine their national poverty line. In other words, the upward revisions to the national poverty lines in these countries are a direct result of their improved consumption measures. Furthermore, the national poverty lines that underpin the IPL are timelier than in the prior round. These revisions to national poverty lines in low-income countries lead to an upward revision in the IPL, which is now based on better and more timely information on the cost of basic needs in the poorest countries in the world. The adoption of the new IPL with the new PPPs by itself (i.e., holding the India survey data fixed) would have revised up the population of extremely poor people by almost a quarter of a billion in 2022. The overall change observed in global poverty estimates is the net impact of these various data updates. The paper estimates a more modest increase in the global count of the extreme poor of 125 million in 2022, which is the impact of the adoption of a new IPL in 2021 PPP dollars net of the estimated reduction in poverty in India. The world is poorer than we thought and less likely 28 to eradicate extreme poverty by 2030, as extreme poverty is getting more generally concentrated in Sub-Saharan Africa and fragile countries. However, the world has made somewhat more progress in poverty reduction since 1990. About 1.5 billion people escaped extreme poverty between 1990 and 2022, compared to 1.3 billion previously estimated. Yet, at the higher poverty line of $8.30 (2021 PPP), which is more relevant for measuring poverty in middle income countries that are home to three-quarters of the world’s population, the estimated number of poor people remains more than 3.7 billion globally, the same number as in 1990. In conclusion, adopting the latest and best available data leads to an upward revision in extreme poverty, which stands at 9.9 percent of the world’s population in 2025. While the decline over the last few decades has been somewhat faster, eradicating extreme poverty by 2030 is even further out of reach, since extreme poverty is increasingly concentrated in Sub-Saharan Africa where progress has been slow. 29 7. References Alfani, F., Aron, D.V., Atamanov, A., Castaneda, R.A.A., Diaz-Bonilla, C., Devpura, N.P., Dewina, R., Finn, A., Fujs, T.H.M.J., Gonzalez, M.F., Krishnan, N., Kocchar, N., Kumar, N., Lakner, C., Lara Ibarra, G., Lestani, D., Liniado, J., Lønborg, J.H., Mahler, D.G., Mejía-Mantilla, C., Montalva, V., Moreno, L.L., Nguyen, M.C., Rubiano, E., Sajaia, Z., Sanchez, D.M., Tetteh- Baah, S.K., Mendoza, M.C.V., Wu, H., Yonzan, N., Wambile, A., 2025. June 2025 Update to the Poverty and Inequality Platform (PIP): What’s New (No. 44), Global Poverty Monitoring Technical Note Series. World Bank, Washington, D.C. Angrist, N., Goldberg, P.K., Jolliffe, D., 2021. Why Is Growth in Developing Countries So Hard to Measure? Journal of Economic Perspectives 35, 215–242. https://doi.org/10.1257/jep.35.3.215 Aron, D., Aguilar, R.A.C., Diaz-Bonilla, C., Fujs, T.H.M.J., Rojas, D.G., Hill, R., Jularbal, L., Lakner, C., Ibarra, G.L., Mahler, D.G., Nguyen, M.C., Nursamsu, S., Sabatino, C., Sajaia, Z., Seitz, W., Sjahrir, B.S., Tetteh-Baah, S.K., Mendoza, M.C.V., Winkler, H., Wu, H., Yonzan, N., 2024. September 2024 Update to the Poverty and Inequality Platform (PIP): What’s New (No. 39), Global Poverty Monitoring Technical Note Series. The World Bank. Beegle, K., De Weerdt, J., Friedman, J., Gibson, J., 2012. Methods of household consumption measurement through surveys: Experimental results from Tanzania. Journal of Development Economics, Symposium on Measurement and Survey Design 98, 3–18. https://doi.org/10.1016/j.jdeveco.2011.11.001 Berry, F., Graf, B., Stanger, M., Ylä-Jarkko, M., 2019. Price Statistics Compilation in 196 Economies: The Relevance for Policy Analysis. Caeyers, B., Chalmers, N., De Weerdt, J., 2012. Improving consumption measurement and other survey data through CAPI: Evidence from a randomized experiment. Journal of Development Economics, Symposium on Measurement and Survey Design 98, 19–33. https://doi.org/10.1016/j.jdeveco.2011.12.001 Castaneda, R.A.A., Dewina, R., Diaz-Bonilla, C., Edochie, I.N., Fujs, T.H.M.J., Jolliffe, D.M., Lain, J., Lakner, C., Ibarra, G.L., Mahler, D.G., Meyer, M., Montes, J., Herrera, L.L.M., Mungai, R., Newhouse, D.L., Nguyen, M.C., Castro, D.M.S., Schoch, M., Sousa, L.D., Tetteh-Baah, S.K., Uochi, I., Mendoza, M.C.V., Wu, H., Yonzan, N., Yoshida, N., 2022. April 2022 Update to the Poverty and Inequality Platform (PIP): What’s New (No. 20), Global Poverty Monitoring Technical Note Series. The World Bank. Castaneda, R.A.A., Diaz-Bonilla, C., Fujs, T., Jolliffe, D., Kotikula, A., Lakner, C., Ibarra, G.L., Mahler, D., Montalva Talledo, V., Nguyen, M., Sanchez Castro, D., Suharnoko Sjahrir, B., Tetteh- Baah, S., Uochi, I., Viveros Mendoza, M., Wu, H., Yonzan, N., Wu, H., 2023. March 2023 Update to the Poverty and Inequality Platform (PIP): What’s New (Global Poverty Monitoring Technical Note Series No. 27). The World Bank. Castaneda, R.A.A., Diaz-Bonilla, C., Fujs, T., Lakner, C., Minh, N., Tetteh Baah, S.K., Viveros, M., 2024. March 2024 global poverty update from the World Bank: first estimates of global poverty until 2022 from survey data. URL https://blogs.worldbank.org/en/opendata/march- 2024-global-poverty-update-from-the-world-bank--first-esti (accessed 5.26.22). Chen, S., Ravallion, M., 2010. The developing world is poorer than we thought, but no less successful in the fight against poverty. The Quarterly Journal of Economics 125, 1577–1625. Chen, S., Ravallion, M., 2008. China is poorer than we thought, but no less successful in the fight against poverty, Policy Research Working Paper Series. The World Bank, Washington, DC. Deaton, A., 2010. Price Indexes, Inequality, and the Measurement of World Poverty. American Economic Review 100, 5–34. https://doi.org/10.1257/aer.100.1.5 Deaton, A., Aten, B., 2017. Trying to Understand the PPPs in ICP 2011: Why Are the Results So Different? American Economic Journal: Macroeconomics 9, 243–264. https://doi.org/10.1257/mac.20150153 30 Deaton, A., Schreyer, P., 2022. GDP, Wellbeing, and Health: Thoughts on the 2017 Round of the International Comparison Program. Review of Income and Wealth 68, 1–15. https://doi.org/10.1111/roiw.12520 Ferreira, F.H., Chen, S., Dabalen, A., Dikhanov, Y., Hamadeh, N., Jolliffe, D., Narayan, A., Prydz, E.B., Revenga, A., Sangraula, P., 2016. A global count of the extreme poor in 2012: data issues, methodology and initial results. The Journal of Economic Inequality 14, 141–172. Harrell, F.E., Davis, C.E., 1982. A new distribution-free quantile estimator. Biometrika 69, 635–640. https://doi.org/10.1093/biomet/69.3.635 Jolliffe, D., Prydz, E.B., 2021. Societal Poverty: A Relative and Relevant Measure. The World Bank Economic Review 35, 180–206. https://doi.org/10.1093/wber/lhz018 Jolliffe, D., Prydz, E.B., 2016. Estimating international poverty lines from comparable national thresholds. J Econ Inequal 14, 185–198. https://doi.org/10.1007/s10888-016-9327-5 Jolliffe ⓡ, D., Mahler ⓡ, D.G., Lakner ⓡ, C., Atamanov ⓡ, A., Tetteh-Baah, S.K., 2024. Poverty and Prices: Assessing the Impact of the 2017 PPPs on the International Poverty Line and Global Poverty. The World Bank Economic Review lhae035. https://doi.org/10.1093/wber/lhae035 Kraay, A.C., Lakner, C., Ozler, B., Decerf, B.M.A., Jolliffe, D.M., Sterck, O.C.B., Yonzan, N., 2023. A New Distribution Sensitive Index for Measuring Welfare, Poverty, and Inequality (No. 10470), Policy Research Working Paper Series. The World Bank. Lakner, C., Milanovic, B., 2016. Global income distribution: From the fall of the Berlin Wall to the great recession. The World Bank Economic Review 30, 203–232. https://doi.org/10.1093/wber/lhw014 Lanjouw, J.O., Lanjouw, P., 2001. How to Compare Apples And Oranges: Poverty Measurement Based on Different Definitions of Consumption. Review of Income and Wealth 47, 25–42. https://doi.org/10.1111/1475-4991.00002 Mahler, D.G., Foster, E., Lakner, C., Prinsloo, Z., Mbouendeu, R.T., Tetteh-Baah, S.K., 2025. Constructing Comparable Global Poverty Trends (No. 45), Global Poverty Monitoring Technical Note. World Bank, Washington, D.C. Mahler, D.G., Foster, E., Tetteh-Baah, S., 2024. How Improved Household Surveys Influence National and International Poverty Rates. Washington, DC: World Bank. https://doi.org/10.1596/42225 Oseni, G., Palacios-Lopez, A., Mugera, H.K., Durazo, J., 2021. Capturing What Matters: Essential Guidelines for Designing Household Surveys, Living Standards Measurement Study (LSMS). World Bank, Washington, D.C. Ravallion, M., 1998. Poverty Lines in Theory and Practice, Living Standards Measurement Study Working Paper 133. World Bank, Washington, D.C. Ravallion, M., Chen, S., 2011. Weakly relative poverty. The Review of Economics and Statistics 93, 1251–1261. https://doi.org/10.1162/REST_a_00130 World Bank, 2025. Poverty and Inequality Platform Methodology Handbook. World Bank, 2024a. Poverty, Prosperity, and Planet Report 2024: Pathways out of the Polycrisis. World Bank, Washington D.C. World Bank, 2024b. Purchasing Power Parities and the Size of World Economies: Results from the International Comparison Program 2021. World Bank, Washington, D.C. World Bank, 2020. Purchasing Power Parities and the Size of World Economies. World Bank, Washington, D.C. World Bank, forthcoming. India: Trends in Poverty, 2011-12 to 2022-23. Methodology Note. World Bank, Washington, D.C. Yonzan, N., Nguyen, M.C., Lakner, C., Kraay, A., Jolliffe, D.M., Wu, H., Lara Ibarra, G., forthcoming. Bottom-coding for the measurement of global poverty and inequality. 31 Appendix A. Review of 2021 PPPs and defining exceptions to the official PPPs The criteria for assessing the 2021 PPPs and the decision to make exceptions for four countries follow similar criteria from the 2017 cycle (Jolliffe ⓡ al., 2024). In this exercise, only countries that have household survey data available in PIP are considered. There are three main steps. First, countries are identified whose official 2021 PPPs would lead to large changes compared to existing estimates. For these countries, an alternative PPP imputed from a regression model is estimated in addition to the officially published 2021 PPP. Second, additional analysis provides evidence against or in favor of either version of the 2021 PPPs. This includes analyzing metadata on CPIs and PPPs, as well as non-monetary covariates of poverty, such as the age dependency ratio (ADR) and the multidimensional poverty index (MPI). Third, a decision is made, depending on the overall assessment to use either the official PPP, or the geometric average of the official and imputed PPP. The latter is chosen if there is reasonable evidence against using the official PPP for global poverty monitoring. A1.1. Identifying outlier countries Outlier countries are identified in three ways. The first is according to their delta ratio. The delta ratio is the factor that converts 2017-PPP-based consumption or income into 2021 PPP dollars, for a given survey-year (see more details in the main text). It can be expressed as the ratio of PPPs and CPIs as follows: 2021 2021 2017 = = ∗ 2017 2017 2021 where: 2021 is per capita income or consumption expressed in 2021 PPP dollars 2017 is per capita income or consumption expressed in 2017 PPP dollars 2021 is the consumer price index in 2021 2017 is the consumer price index in 2017 2017 is the 2017 purchasing power parity conversion factor 2021 is the 2021 purchasing power parity conversion factor Countries whose official 2021 PPP yields a delta ratio that is two standard deviations away from US inflation are selected as outliers. Figure A1 shows two versions of the delta ratio, the first using the 2017 PPPs that are used in the Poverty and Inequality Platform (PIP), and the second using the (original) 2017 PPPs published by the International Comparison Program (ICP). Figure 1 in the main text is similar to panel (a), but the former includes only benchmark countries in both 2017 and 2021 ICP cycles. A second approach to identify outliers is based on whether they were treated as exceptions when the 2017 PPPs were adopted. Similar to the current exercise, in a few cases the geometric average 32 of the official PPP and imputed PPPs was used instead of the published 2017 PPPs. This explains the difference between panels a and b in Figure A1. 38 All countries for which an exception was made with the 2017 PPPs are listed in Table A1, as the second type of outlier countries. Figure A1: Real Changes in PPP-Adjusted Dollars between 2017 PPPs and 2021 PPPs (Delta Ratio) (a) Using 2017 PPPs from PIP (b) Using 2017 PPPs from ICP Note: These charts represent the relative change in PPP dollar units when moving from the 2017 PPPs to the 2021 PPPs. Table A1 provides the list of countries considered as outliers. Non-benchmark countries are included. Table A1: Selected countries for which alternative PPPs are considered Delta criterion Special cases from 2017 ICP cycle PPP residual criterion Belize Belize Belarus Guinea Guinea Central African Republic Guatemala Iraq Djibouti Egypt, Arab Rep. Egypt, Arab Rep. Egypt, Arab Rep. St. Lucia Nigeria Honduras Lebanon Trinidad and Tobago Kazakhstan Sudan Sudan Russian Federation São Tomé and Príncipe São Tomé and Príncipe Suriname Panama West Bank and Gaza Syrian Arab Republic Zimbabwe Uzbekistan Iran, Islamic Rep. Timor-Leste Turkmenistan Note: Across the three columns, there are 26 unique countries. Non-benchmark (i.e., countries whose PPPs are estimated from a regression) countries are shown in italics. 38 For example, Egypt was flagged as an outlier when the 2017 PPPs were adopted. Against this baseline, using the official 2021 PPP would lead to a large delta ratio (panel a). The same effect does not exist when the official 2021 PPP is compared against the official 2017 PPP (panel b). 33 The third criterion to label outlier countries uses residuals from the regression model that estimates PPPs for countries that do not have a price survey. The ICP publishes PPPs for two types of countries, namely benchmark and non-benchmark countries. 39 Benchmark countries are those whose PPPs are estimated from actual price data using an index number methodology, while non-benchmark countries are those whose PPPs are predicted from a regression model using the benchmark PPPs, because price data were not collected from these countries. Unlike the delta ratio criterion, which does not indicate whether there are issues with PPPs or CPIs, the residual criterion directly identifies outliers from the PPPs. This approach identifies, among benchmark countries (i.e. those with a PPP based on a price survey), outlier countries two standard deviations from the mean log difference between official and predicted PPPs. The ICP estimates the predicted PPP using a seemingly unrelated regressions (SUR) model specified as follows: − = ∗ ( − ) + where is the price level index of country , calculated as the ratio of the PPP conversion factor to the market exchange rate, and is a vector of explanatory variables: GDP per capita in U.S. dollars (based on market exchange rates), imports as a share of GDP, exports as a share of GDP, age dependency ratio, and dummies for Sub-Saharan Africa, the EU, island economies, and landlocked developing economies. Interaction terms between GDP per capita and the country- group-dummy variables are also included. The ICP runs three separate specifications for GDP, household final consumption including nonprofit institutions serving households (NPISHs), and actual individual consumption (AIC), respectively. The product of the price level index for household final consumption including NPISHs and the market exchange rate is the preferred PPP conversion factor for measuring poverty. The log difference between the published and our predicted PPP is presented in Figure A2. Zimbabwe, for example, has a published PPP that is considerably higher than its predicted PPP. The Arab Republic of Egypt, on the other hand, has a published PPP that is much lower than its predicted PPP. The list of outlier countries by the PPP residual criterion is given in the third column of Table A1. 39Not all non-benchmark countries use a PPP derived from the regression. When available, the ICP also extrapolates benchmark data from a prior ICP cycle. For more details, see the discussion in Section 2.1 and around Table A2. 34 Figure A2: Outlier countries using the PPP residual criterion Note: The analysis includes all countries with benchmark (i.e. price data-based) PPPs, but the chart visualizes only countries with survey data in the Poverty and Inequality Platform. The analysis has uncovered 26 unique countries. Before considering additional evidence against or in favor of the different PPPs in the second step, we exclude several countries from this list. The proposed solution for the outlier countries relies on using information from their imputed PPPs (in combination with the official PPPs published by the ICP). The same SUR model above is used to predict and impute PPPs, except that it now predicts out of sample (i.e., all the outlier countries are excluded from the estimation). If the imputed PPP results in a more extreme delta ratio than the official PPP (i.e., even further outside the 2 SD bound), the proposed solution of using the geometric mean will lead to even larger changes when moving from 2017 to 2021 PPPs. Therefore, we can exclude those countries for which the imputed PPP is even more extreme than the officially published PPP. Also excluded are outlier countries that have become benchmark countries in the 2021 cycle (e.g., Lebanon). For these countries, it is difficult to compare the 2017 and 2021 PPPs, since it involves a change in methodology. 40 All the excluded countries are 40These methodological changes could involve for example a switch from predicted to benchmark PPP, a change from an extrapolated PPP, or a shift from an experimental survey. For each country, the details are a bit different. Guatemala participated in ICP 2011 but not in ICP 2017, and again in ICP 2021. The original 2017 PPP for the country was based on an extrapolation of the ICP 2011, and the revised 2017 PPP (which is not used in PIP) was estimated by interpolating the 2011 and 2021 PPPs. Uzbekistan conducted an experimental survey for household consumption under ICP 2017, which was utilized in the 2017 PPPs estimation. For Kosovo, the ICP incorporated experimental results for the first time under ICP 2021, whereas it was based on model imputation under ICP 2017. Lebanon participated in ICP 2005, but 35 highlighted in Table A2. In summary, the further assessment is done for the remaining countries, including Belize, Egypt, Guinea, Nigeria, Iraq, Sudan, São Tomé and Príncipe, Timor-Leste, and Trinidad and Tobago. Table A2: Countries included for further assessment Delta ratio Delta ratio Country Exclusion criteria (official PPP) (imputed PPP) Belarus 1.07 0.68 Delta ratio more extreme Belize 1.22 1.29 Central African Republic 1.23 1.82 Delta ratio more extreme Djibouti 1.15 1.82 Delta ratio more extreme Egypt, Arab Rep. 1.62 1.08 Guinea 1.60 1.38 Guatemala 1.48 1.59 Non-benchmark in 2017 ICP Honduras 1.08 1.61 Delta ratio more extreme Iran, Islamic Rep. 0.90 0.64 Delta ratio more extreme Iraq 1.09 1.00 Kazakhstan 1.13 0.77 Delta ratio more extreme Lebanon 0.82 0.99 Non-benchmark in 2017 ICP St. Lucia 1.30 1.39 Delta ratio more extreme Nigeria 1.31 1.20 Panama 0.91 0.86 Delta ratio more extreme West Bank and Gaza 1.05 1.67 Delta ratio more extreme Russian Federation 1.16 0.74 Delta ratio more extreme Sudan 1.90 1.37 São Tomé and Príncipe 1.64 1.36 Suriname 1.11 0.68 Delta ratio more extreme Syrian Arab Republic 1.61 2.07 Non-benchmark in 2017 ICP Turkmenistan 1.60 1.61 Delta ratio more extreme Timor-Leste 1.39 1.33 Trinidad and Tobago 1.20 1.07 Uzbekistan 0.88 0.87 Experimental 2017 ICP Zimbabwe 1.15 1.87 Delta ratio more extreme Note: Guatemala and Syrian Arab Republic also have more extreme delta ratio when using the imputed PPP. Countries highlighted are excluded from the rest of the analysis. Timor-Leste is a regression-based non-benchmark country. the 2011 and 2017 PPPs were based on model imputation. South Sudan and the Syrian Arab Republic joined the ICP for the first time under ICP 2021, replacing model-based imputations used in all previous cycles. 36 A1.2. Additional analysis for outlier countries selected for further assessment CPI metadata: Is the CPI series fit for purpose? The delta ratio combines CPIs and PPPs, so when an extreme delta ratio is observed, it is unclear which of the components is driving the extreme value. The quality of the CPIs is assessed with the information published in Berry et al. (2019), which is the latest available cross-country information, although several years old by now. For the CPI series to be fit for purpose, expenditure weights coverage is expected to be national, the reference year for the CPI weights should not be too old (defined as not older than 2010), and CPI expenditures should be based on the internationally recognized Classification of individual consumption by purpose (COICOP). If all these conditions hold, there is little reason to believe that the CPI series has issues, which may suggest that the official PPP is driving the outlier result. 41 This is an argument for treating the country as an exceptional PPP. Table A3 shows the results. Table A3: CPI quality indicators CPI fit CPI expenditure CPI weights CPI classification for Country weights coverage reference year system purpos e Egypt, Arab Rep. National 2010 COICOP Yes Guinea Capital City 2002 Other No Sudan National 2007 Other No São Tomé and Príncipe National 2014 COICOP Yes Belize National 2009 COICOP No Iraq National 2012 COICOP Yes Nigeria National 2004 Other No Trinidad and Tobago National 2009 COICOP No Timor-Leste National 2012 COICOP Yes Source: Berry et al. (2019) Note: The latest available metadata are used but they are several years old. Timor-Leste is a regression-based non- benchmark country. PPP metadata: Has there been a significant change in the share of items priced? The ICP’s price collection essentially consists of a long list of items that need to be priced in every country that participates in the price survey. However, not all items can be collected in every country in every round. Figure A3 compares the share of global core list items priced in the 2021 ICP round with the share priced in the 2017 ICP round for all countries and highlighting the 8 benchmark countries that are part of the additional assessment. 42 The global core list of items is 41 As discussed in the main text, there are fundamental differences between what CPIs and PPPs measure, so even with a perfectly measured CPI, the trends in the two indices may not line up perfectly. 42 Timor-Leste is also included in the additional assessment, but it is not a benchmark country in either round, so it cannot be included in Figure A3. 37 a generalized basket of goods and services priced in all participating countries—662 items in 2021 and 651 items in 2017—used for linking PPPs across countries and regions. More countries are above the 45-degree line, suggesting that more items were priced in 2021 than in 2017, including for 5 of 8 benchmark countries selected for additional analysis highlighted in the figure. None of the 8 countries shows a significant reduction in the share of items priced in the 2021 ICP round. A significant change in the share of items priced is defined as more than 2 standard deviations from the mean log difference evaluated over all countries. This suggests that for the 8 countries that are part of the additional assessment, the survey underpinning the 2021 PPPs does not capture less information than the earlier round. Figure A3: Share of items priced in 2017 ICP and 2021 ICP cycles, percent Note: Eight benchmark countries selected for further investigation are highlighted with light blue color in this chart. The chart includes only benchmark countries in both ICP cycles. Dotted line is a 45-degree line. 38 Correlation with non-monetary measures of well-being: Does the imputed PPP bring the monetary poverty rate closer to what would be expected from a poverty covariate? In this part of the analysis, we examine whether the official or imputed PPP is a better fit for the cross-country relationship between monetary poverty and two covariates of monetary poverty – the age dependency ratio (ADR) and the OPHI multidimensional poverty index (MPI). The MPI, in contrast to the World Bank’s multidimensional poverty measure, does not use input variables denominated in PPPs. In our approach, we regress the monetary poverty rate on a poverty covariate — ADR or MPI — with regional dummies and interactions. All outlier countries identified in step 1 are excluded from the regression. Using the estimated parameters, we predict monetary poverty for the outlier countries. Finally, we compute the absolute difference between the predicted poverty rate and (a) the poverty rate with official 2021 PPP and (b) the poverty rate with imputed 2021 PPP. In the tables below, the imputed PPP is preferable when the former difference is greater than the latter difference (see Tables A4 to A7). Such a case would imply that the poverty rate with imputed 2021 PPP is closer to the poverty rate expected from MPI or ADR than the poverty rate with official 2021 PPP. We run these regressions at two different poverty thresholds – the extreme poverty line of $3.00 and the poverty line for upper-middle- income-countries of $8.30. Table A4: Age dependency ratio and poverty at $3.00 (2021 PPP) Poverty, % Poverty, % Log diff. Log diff. Imputed Country Year (official (imputed (official (imputed PPP PPP) PPP) PPP) PPP) preferable Egypt, Arab Rep. 2021 0.4 3.6 0.2 2.6 No Guinea 2018 9.3 14.3 1.5 1.1 Yes Sudan 2014 6.1 16.0 2.1 1.1 Yes São Tomé and Príncipe 2017 9.5 17.1 1.4 0.8 Yes Belize 2018 1.0 0.7 1.4 1.9 No Iraq 2023 0.5 0.7 1.5 1.7 No Nigeria 2018 34.2 39.6 0.3 0.2 Yes Trinidad and Tobago 1992 3.2 5.0 0.7 0.2 Yes Timor-Leste 2014 24.9 28.6 1.4 1.3 Yes Note: Poverty rates are based on the latest surveys in the Poverty and Inequality Platform. These estimates are matched with annual estimates of age dependency ratio in the World Development Indicators (WDI). The series used is: “Age dependency ratio, young (% of working-age population)”, the ratio of younger dependents (i.e., people younger than 15) to the working-age population (i.e., those aged 15-64). 39 Table A5: Age dependency ratio and poverty at $8.30 (2021 PPP) Poverty, % Poverty, % Log diff. Log diff. Imputed Country Year (official (imputed (official (imputed PPP PPP) PPP) PPP) PPP) preferable Egypt, Arab Rep. 2021 39.3 74.2 0.6 1.2 No Guinea 2018 72.3 80.6 0.2 0.1 Yes Sudan 2014 62.4 81.5 0.4 0.1 Yes São Tomé and Príncipe 2017 64.6 74.7 0.2 0.1 Yes Belize 2018 19.7 18.1 0.4 0.5 No Iraq 2023 28.0 35.1 0.5 0.7 No Nigeria 2018 88.7 91.1 0.0 0.0 Yes Trinidad and Tobago 1992 30.5 36.1 0.2 0.1 Yes Timor-Leste 2014 91.8 92.7 0.0 0.0 No Note: Poverty rates are based on the latest surveys in the Poverty and Inequality Platform. These estimates are matched with annual estimates of age dependency ratio in the World Development Indicators (WDI). The series used is: “Age dependency ratio, young (% of working-age population)”, the ratio of younger dependents (i.e., people younger than 15) to the working-age population (i.e., those aged 15-64). Table A6: Multidimensional poverty index and poverty at $3.00 (2021 PPP) Poverty, % Poverty, % Log diff. Log diff. Imputed Country Year (official (imputed (official (imputed PPP PPP) PPP) PPP) PPP) preferable Egypt, Arab Rep. 2014 0.3 4.1 2.1 0.5 Yes Guinea 2018 9.7 14.9 1.7 1.2 Yes Sudan 2014 6.1 16.0 2.0 1.0 Yes São Tomé and Príncipe 2019 9.0 16.1 0.5 0.1 Yes Belize 2015 3.3 2.9 0.2 0.3 No Iraq 2018 0.4 0.6 2.4 1.9 Yes Nigeria 2021 35.3 40.7 0.1 0.3 No Trinidad and Tobago 2022 0.0 0.0 n.a. n.a. No Timor-Leste 2016 23.9 27.8 1.4 1.2 Yes Note: The latest available estimates of the share of population living in multi-dimensional poverty from the OPHI Database are matched with lined-up estimates of monetary poverty in the Poverty and Inequality Platform. Lined-up estimates of monetary poverty are extrapolated or interpolated from survey-based estimates, assuming that changes in household material well-being follows changes in national accounts data, such as GDP per capita and final household consumption expenditure. 40 Table A7: Multidimensional poverty index and poverty at $8.30 (2021 PPP) Poverty, % Poverty, % Log diff. Log diff. Imputed Country Year (official (imputed (official (imputed PPP PPP) PPP) PPP) PPP) preferable Egypt, Arab Rep. 2014 41.0 74.8 0.1 0.5 No Guinea 2018 72.9 81.1 0.3 0.2 Yes Sudan 2014 62.4 81.5 0.4 0.1 Yes São Tomé and Príncipe 2019 63.5 73.1 0.2 0.4 No Belize 2015 22.9 21.2 0.3 0.3 No Iraq 2018 24.7 30.8 0.7 0.5 Yes Nigeria 2021 89.2 91.5 0.2 0.2 No Trinidad and Tobago 2022 5.8 8.0 1.0 0.7 Yes Timor-Leste 2016 91.5 92.6 0.3 0.3 Yes Note: The latest available estimates of the share of population living in multi-dimensional poverty from the OPHI Database are matched with lined-up estimates of monetary poverty in the Poverty and Inequality Platform. Lined-up estimates of monetary poverty are extrapolated or interpolated from survey-based estimates, assuming that changes in household material well-being follows changes in national accounts data, such as GDP per capita and final household consumption expenditure. Summary of the additional assessment Table A8 summarizes the results from the additional assessment on the outlier countries. Countries with at least 3 flags have additional reasons to justify a deviation from using the official 2021 PPP for global poverty monitoring. While there is evidence that improvement over the official PPP is possible through modeling of the PPP, we do not have strong evidence that the imputed PPP should be taken at face value and preferred to a PPP based on an actual price survey. Thus, and following the World Bank’s approach with the 2017 PPPs (Jolliffe ⓡ al. 2024), the average of the official and imputed PPPs, for Egypt, Guinea, Sudan, and São Tomé and Príncipe, is used for global poverty monitoring. Timor-Leste is already using an imputed PPP from the ICP, being a non-benchmark country. The additional analysis concluded that our own imputed PPP may be a better fit. The difference between the two imputed PPPs is relatively small and stems from the underlying sample (i.e., we exclude all the potential outlier countries). Our takeaway from the additional analysis is that Timor-Leste’s PPP imputed by the ICP may be a poor fit for measuring global poverty, but computing the geometric average between two imputed PPPs that differ because of alternative sampling choices is unlikely to be a solution. Therefore, for Timor-Leste, we will continue using a PPP extrapolated from the 2011 ICP, which is currently used by the World Bank for the 2017 ICP round and seems appropriate for the country’s level of development (Castaneda et al., 2023). 41 Table A8: Summary of results from additional assessment Share of Imputed PPP results in lower residual CPI fit items Delta Total Country for priced ADR- ADR- MPI- MPI- ratio is flags purpose has $3.0 $8.3 $3.0 $8.3 outlier declined Egypt, Arab Rep. Yes No No No Yes No Yes 3 Guinea No No Yes Yes Yes Yes Yes 3 Sudan No No Yes Yes Yes Yes Yes 3 São Tomé and Príncipe Yes No Yes Yes Yes No Yes 4 Belize No No No No No No Yes 1 Iraq Yes No No No Yes Yes No 2 Nigeria No No Yes Yes No No No 1 Trinidad and Tobago No No Yes Yes No Yes No 1 Timor-Leste Yes N/A Yes No Yes Yes Yes 4 Note: Each Yes flag counts as 1. At least one Yes under each category (e.g., age dependency ratio) counts as 1. A1.3. Summary of exceptions to the official 2021 PPPs Tables A9 and A10 summarize the 2021 PPPs for the countries for which we deviate from those published by the ICP. For Egypt, Guinea, Sudan, and São Tomé and Príncipe we use the geometric average of the officially published PPP and an imputed PPP, as just described. For several other countries we use PPPs extrapolated from the 2011 round. The revised 2011 PPP (published in 2020) are used (World Bank, 2020), together with domestic and US CPI inflation between 2011 and 2021. The same was done for these countries for the 2017 PPPs (Jolliffe ⓡ et al., 2024 and Castaneda et al., 2023). For China, the World Bank’s poverty measures have long incorporated separate urban and rural PPPs (Chen and Ravallion, 2010, 2008; Ferreira et al., 2016; Jolliffe ⓡ al., 2024). Following the same methods, we update these calculations with the 2021 ICP round. The resulting urban and rural PPPs are shown in Table A10. In previous ICP rounds, the World Bank’s poverty measures for India and Indonesia followed the same approach. However, for these countries a more detailed spatial price adjustment has now been incorporated into their welfare distributions, so the national PPP conversion factor is used (Alfani et al., 2025; Aron et al., 2024). 42 Table A9: Exceptions from official 2021 PPPs Official PPP Imputed Country PPP used Type (ICP) PPP Geometric average of official and Egypt, Arab Rep. 4.11 6.18 5.04 imputed PPP Geometric average of official and Guinea 3221.26 3741.79 3471.78 imputed PPP Geometric average of official and Sudan 101.43 140.17 119.24 imputed PPP Geometric average of official and São Tomé and Príncipe 9.62 11.62 10.57 imputed PPP Kiribati 0.92 Extrapolated PPP Marshall Islands 0.95 Extrapolated PPP Nauru 1.34 Extrapolated PPP Timor-Leste 0.56 Extrapolated PPP Tuvalu 1.29 Extrapolated PPP Venezuela, RB 15517657088 Extrapolated PPP Yemen, Rep. 572.01 Extrapolated PPP Note: The table provides the 2021 PPP conversion factors for the countries which are not included in the ICP 2021 or for which additional assessments suggested that an alternative PPP might be more appropriate. Table A10: Subnational PPPs for China Category 2017 ICP cycle 2021 ICP cycle Ratio of urban to rural poverty line (ω) 1.24 1.23 ICP urban share of outlets (λ) 0.79 0.76 Rural PPP 3.50 3.42 Urban PPP 4.32 4.20 National PPP 4.15 4.01 Urban to rural PPP ratio 1.24 1.23 Note: For China, separate urban and rural PPPs are used instead of the national PPP provided by the ICP. The formulae for the urban and rural PPPs are: = ; = × . See online ×+(1−) appendix of Ferreira et al. 2016) for details. 43 B. Further details on the database of national poverty lines The database of harmonized national poverty lines is built from information across various sources. A succinct description of how this dataset is created follows, with more details available in the paper’s accompanying reproducibility package. All observations in the World Bank’s Poverty and Equity Briefs are first selected and are supplemented with observations from the Poverty and Inequality Platform (PIP) and the World Development Indicators (WDI). The series used from the WDI is “Poverty headcount ratio at national poverty lines (SI.POV.NAHC)”. The same series is available in PIP, but in addition PIP has historical and non-comparable series (where comparability refers to comparability within countries across surveys, not between countries). Missing country-year observations are filled with Eurostat and OECD data, in that order. The series used from EUROSTAT is “At risk of poverty rate (cut-off point: 60% of median equivalized income after social transfers)”, while the series from the OECD is “60% of the national median disposable income”. Additional national poverty rates for Taiwan, China, are obtained from the Ministry of Health and Wealth, Taiwan. Azerbaijan has limited survey data in PIP, mostly historical data. To fill missing data, 18 observations of national poverty lines expressed in domestic currency (manat) and per capita terms obtained from the State Statistical Committee of Azerbaijan are converted into PPP dollars. 43 Table B1 provides additional information and a breakdown of the sources of data on national poverty rates used for this paper. The vintages of the data sources are indicated in the table with the numerical values of the relevant years and months. For example, PEB-2025-04 means pre- processed data from the April 2025 vintage of the PEBs. We start with 1,976 national poverty rates. Of these, 1,727 could be merged into PIP distributions to derive corresponding harmonized national poverty lines. In addition, 18 poverty lines were obtained for Azerbaijan. In sum, there are 1,747 national poverty lines that form the basis of the analysis. Table B1: Sources of national poverty rates Source All Used Vintage Poverty and Equity Briefs 914 837 PEB-2025-04; 2024-10 Poverty and Inequality Platform / WDI 917 774 PIP-2024-09; 2024-07; WDI-2024-07 Eurostat 59 56 EUROSTAT-2024-06 OECD 76 55 OECD-2024-07 Ministry of Health and Wealth, Taiwan 10 7 TWN-2024-09 Sub-total 1976 1729 State Statistical Committee, Azerbaijan 0 18 AZE-2024-07 Total 1976 1747 Note: Of all 1976 observations of national poverty rates obtained from different sources, 1727 could be matched into distributions in PIP to determine corresponding harmonized national poverty lines. 43 The data source is https://www.stat.gov.az/source/budget_households/?lang=en. 44 C. Additional results, tables and figures Figure C1: Real Changes in PPP-Adjusted Dollars between 2017 PPPs and 2021 PPPs (Delta Ratio), accounting for population size Note: The figure replicates Figure 1 in the main text, but in this version country markers are shown proportional to population size and the 30 most populous countries are labelled. It also shows the final delta ratios, including exceptions from official PPPs for several countries, as described in Appendix A. The figure plots the change in PPP-based welfare between the 2017 and 2021 rounds against GDP per capita. For the average country, the delta ratio is 1.13, meaning that nominal PPP-denominated welfare increases by 13% between 2017 and 2021. For the average person in the world, the delta ratio is slightly higher at 1.15. For the US, the delta ratio is 1.11, which by definition equals US inflation between 2017 and 2021. The cross-sectional relationship between the delta ratio and GDP per capita is negative and significant. The sample consists of all 172 countries with survey data in the Poverty and Inequality Platform (PIP), representing more than 97 percent of the world’s population. Table C1: Real Changes in PPP-Adjusted Dollars between 2017 PPPs and 2021 PPPs (Delta Ratio), average by income group Income classification of countries Delta ratio Delta ratio (population weighted) Observations Low-income countries 1.21 1.21 24 Lower-middle-income countries 1.11 1.17 53 Upper-middle-income countries 1.12 1.12 49 High-income countries 1.11 1.11 46 World 1.13 1.15 172 Note: Averages are computed with and without population weights. Population values are for the year 2021. The income classification of countries is based on gross national income per capita data from 2021. The final delta ratios are used, including exceptions from official PPPs for a select number of countries, as described in Appendix A. The sample consists of all 172 countries with survey data in the Poverty and Inequality Platform, representing more than 97 percent of the world’s population. 45 Figure C2: Distribution of national poverty lines for low-income-countries, comparison of 2017 and 2021 PPPs Table C2: Implication of more recent data on global poverty lines 2017 ICP, old 2017 ICP, new 2021 ICP, new Change surveys surveys surveys (%) Income group Year Median Obs Year Median Obs Year Median Obs Old New Low-income 2014 2.15 28 2018 2.50 25 2020 3.04 23 41 22 Lower-middle 2015.5 3.63 54 2017 3.80 54 2019 4.20 53 16 11 Upper-middle 2017 6.85 37 2017 7.01 42 2021 8.29 45 21 18 High-income 2017 24.36 38 2017 23.24 42 2021 29.43 42 21 27 Total obs 157 163 163 Note: Median values are expressed in daily per capita PPP dollars. The column for Year indicates the median survey year. The old surveys cover the surveys that were available at the time when the poverty lines for the 2017 PPPs were defined. The other results for the 2017 PPPs repeat that analysis but use the poverty lines that are currently available. Changes are estimated for the 2021 ICP, relative to the old and new surveys; e.g., for low-income countries, the poverty line increases by 41 percent when comparing the 2021 PPPs to the 2017 PPPs that use the old surveys ($3.04 vs. $2.15), but only by 22 percent when comparing against the 2017 PPPs that use the new surveys ($3.04 vs. $2.50). 46 Figure C3: Illustration of the Harrell-Davis quintile estimator (for low-income countries) a. 2017 ICP cycle b. 2021 ICP cycle Note: For the low-income countries, the figure shows the harmonized national poverty lines ordered from the lowest to the highest (dark blue), as well as the associated Harrell-Davis quantiles (light blue). 47 Table C3: Robustness of global poverty lines to different samples and assumptions No. Sample (or type) Low-income countries Lower-middle- Upper-middle- High-income income income countries HD HD HD HD Total Median Obs Median Obs Median Obs Median Obs Median Median Median Median obs 1 Default 3.04 3.01 23 4.20 4.14 53 8.29 8.14 45 29.43 29.03 42 163 2 Countries with PPPs from ICP (special PPPs used) 3.04 3.01 23 4.27 4.27 44 8.46 8.39 42 29.74 29.43 41 150 3 Benchmark countries (special PPPs used) 2.93 2.92 22 4.28 4.31 43 8.46 8.39 42 29.74 29.43 41 148 4 Countries with PPPs from ICP (ICP PPPs used) 3.04 3.01 23 4.27 4.27 44 8.46 8.39 42 29.74 29.43 41 150 5 Benchmark countries (ICP PPPs used) 2.93 2.92 22 4.28 4.31 43 8.46 8.39 42 29.74 29.43 41 148 6 Within 5 years of 2021 3.04 2.97 19 4.20 4.18 38 8.46 8.33 38 29.74 29.43 41 136 7 Within 10 years of 2021 3.04 3.01 23 4.13 4.11 50 8.37 8.24 43 29.43 29.03 42 158 8 Within 20 years of 2021 3.04 3.01 23 4.20 4.14 53 8.29 8.14 45 29.43 29.03 42 163 9 Pooling lines, all years, equal weighting 2.70 2.69 202 4.51 4.52 438 7.49 7.49 448 26.14 26.15 659 1747 10 Pooling lines, all years, triangular weighting 2.82 2.82 202 4.26 4.25 438 7.53 7.55 448 27.33 27.34 659 1747 11 Pooling lines, within 5 years of 2021, equal weighting 2.92 2.92 29 4.20 4.20 91 8.02 8.00 178 27.95 27.90 247 545 12 Pooling lines, within 5 years of 2021, triangular weig. 3.02 3.01 29 4.20 4.20 91 8.10 8.10 178 28.65 28.67 247 545 13 Pooling lines, within 10 years of 2021, equal weighting 2.71 2.74 55 4.13 4.15 178 7.49 7.49 288 26.18 26.15 415 936 14 Pooling lines, within 10 years of 2021, triangular weig. 2.82 2.82 55 4.13 4.12 178 7.76 7.74 288 27.38 27.39 415 936 15 Pooling lines, within 20 years of 2021, equal weighting 2.66 2.66 142 4.46 4.46 375 7.48 7.47 428 26.56 26.54 601 1546 16 Pooling lines, within 20 years of 2021, triangular weig. 2.82 2.82 142 4.24 4.24 375 7.52 7.52 428 27.38 27.40 601 1546 17 Updating with US delta ratio (i.e. US inflation) 2.38 4.03 7.57 18 Updating with income-group-specific mean delta ratio 2.59 4.06 7.69 Note: Rows 1-16 show median and Harrell-Davis (HD) median values estimated from different samples of national poverty lines. Sample 2 uses special 2021 PPPs for Egypt, Guinea, Sao Tome and Principe, Timor-Leste, and Sudan, but excludes countries for which extrapolated PPPs are used. Sample 4 uses officially published PPPs from the International Comparison Program (ICP) for these five countries. The same applies to samples 3 and 5, respectively, but excludes Timor-Leste, being a non-benchmark country. In sample 4 or 5, only PPPs directly from the ICP are used. In every other sample, special PPPs are used for these five countries. Both equal and triangular weighting when pooling lines ensure that each country has a weight of 1. In the equal weighting scheme, each line has a weight equal to the inverse of the total number of lines for a country. In the triangular weighting scheme, each line has a weight equal to the inverse of the relative distance from the 2021 ICP reference year. The last two rows directly update the 2017-PPP-based poverty lines ($2.15, $3.65, $6.85) to 2021 PPP dollars using the delta ratio. 48 Table C4: Sources of changes in global poverty lines: alternative decomposition Low-income Lower-middle Upper-middle High-income Total Description Median Obs Median Obs Median Obs Median Obs obs Full sample All lines in 2017 PPPs 2.15 28 3.63 54 6.85 37 24.36 38 157 All lines in 2021 PPPs 3.04 23 4.20 53 8.29 45 29.43 42 163 Total change ($) 0.89 0.57 1.44 5.07 Balanced sample Lines in 2017 PPPs 2.16 23 3.49 47 6.54 32 25.43 36 138 Update PPPs 2.52 23 3.94 47 7.19 32 27.15 36 138 Update poverty lines 3.04 23 4.20 47 7.50 32 31.40 36 138 Change ($) 0.88 0.71 0.96 5.97 Contributions PPP impact (%) 41 78 45 34 Line impact (%) 58 45 22 84 Residual impact (%) 1 -23 33 -18 Total change (%) 100 100 100 100 Note: This approach is also path-dependent but uses a balanced sample of countries in each income group. This balanced sample includes countries with the same income group and welfare type in both 2017 and 2021 ICP cycles. The PPP impact is the change in the median when updating the 2017-PPP-based national poverty lines with new PPPs in the balanced sample, as a share of the total change. The line impact is the change in the median when updating the national poverty lines from the previous step (i.e., the original 2017-PPP-based national poverty line expressed in 2021 PPPs) with new 2021-PPP-based national poverty lines in the balanced sample, as a share of the total change. The residual impact includes changes that are not captured by the balanced sample, such as changes in the number of countries in each income group, and changes in the income status of countries (i.e., graduating to a higher income group), and any changes in the welfare type. The residual impact is relatively large in the middle-income groups due to the relative importance of these factors, such as changes in countries’ income groupings. 49 Table C5: OLS regressions of national poverty lines on the median Unconstrained regressions Constrained regressions Variable Full sample Absolute lines Full sample Absolute lines Median 0.56*** 0.47*** 0.50 0.50 (0.008) (0.027) Constant 0.87*** 1.63*** 2.03*** 1.39*** (0.223) (0.276) (0.189) (0.161) Obs 162 114 162 114 Root Mean Squared Error (RMSE) 2.074 1.722 2.408 1.723 p(Median = 0.5) 0.000 0.297 p(Contant = 1.30) 0.057 0.240 0.000 0.574 Note: Standard errors are in parentheses. * p<.10, ** p<.05, *** p<.01. The constrained regressions impose that the coefficient on the median is 0.5. The rationale for this assumption is that switching PPPs should only impact the parameters of the SPL that are expressed in monetary terms, namely the intercept, and not the underlying relationship between median income and harmonized national poverty lines. The full sample includes 48 countries that define poverty using a relative line, mostly high-income countries. These are excluded when the sample is restricted to absolute lines. The lower section of the table indicates the root mean squared error (RMSE) as a measure of goodness- of-fit for the four competing models. p(Median = 0.5) provides the result of the hypothesis test under the null that the coefficient on the median is 0.5, while p(Contant = 1.30) provides the result of the hypothesis test under the null that the constant term is 1.30 (i.e., the resulting value when using the delta ratio to update the intercept term of the societal poverty line). The analysis here is based on the same harmonized national poverty lines and surveys underpinning the global poverty lines ($3.00, $4.20, and $8.30). Azerbaijan lacks more recent survey data in the Poverty and Inequality Platform (PIP) and could not be included in this analysis, as its poverty line is not derived from the Poverty and Inequality Platform and the median value of the distribution is unknown. 50 Figure C4: Regional poverty trends, 1990 – 2025 Note: Also see Fig. 3 in the main text. Poverty is shown using the old September 2024 vintage and the current June 2025 vintage. The region Rest of the World is dropped for presentational purposes; poverty is almost non-existent in this group of high-income countries at all three lines. 51 Table C6: Countries with the largest changes in millions of poor, 2022 Poverty rates (%) Changes in millions of poor Country $2.15 (Old) $2.15 $3.00 New P I P data Sw itching P P P s Total India 12 3 6 -125 48 -77 Pakistan 3 4 14 2 25 27 Ethiopia 17 30 37 17 8 26 Indonesia 2 2 8 0 16 16 Congo, Dem. Rep. 77 77 84 3 7 10 Uganda 40 41 59 0 8 9 Philippines 6 5 13 -1 9 9 Yemen, Rep. 52 54 68 3 6 9 Nigeria 32 31 35 1 8 8 South Africa 20 21 31 1 7 7 Country $3.65 (Old) $3.65 $4.20 New P I P data Sw itching P P P s Total India 42 31 26 -153 -61 -214 Pakistan 35 37 42 8 12 20 Ethiopia 49 69 62 26 -8 18 Nigeria 64 64 56 2 -19 -17 Indonesia 19 19 24 1 13 14 Bangladesh 30 30 24 -1 -9 -10 Sudan 64 64 41 2 -12 -10 Egypt, Arab Rep. 14 13 7 -1 -7 -8 Syrian Arab Republic 67 67 39 0 -6 -6 Myanmar 25 24 35 0 6 5 Country $6.85 (Old) $6.85 $8.30 New P I P data Sw itching P P P s Total India 80 83 83 52 2 54 China 17 17 21 -3 55 52 Indonesia 63 63 70 2 21 24 Pakistan 82 83 87 9 10 19 Iran, Islamic Rep. 22 23 40 0 16 16 Ethiopia 87 94 93 11 -1 10 Mexico 22 22 27 0 7 7 Philippines 67 68 74 0 7 7 Egypt, Arab Rep. 64 66 58 3 -9 -6 Uzbekistan 17 17 32 0 5 5 Note: Countries are arranged in descending order of the total absolute change in millions of poor between the September 2024 (using 2017 PPPs) and June 2025 (using 2021 PPPs). The top ten countries with the largest total changes are shown for each of the three types of poverty lines. The total change is decomposed into the change due to new data (change from September 2024 to June 2025 vintage, while using the 2017 PPPs) and switching PPPs (change from 2017 PPPs to 2021 PPPs, while using the June 2025 vintage). New PIP data refers to all changes in poverty data in the Poverty and Inequality Platform (PIP) between the September 2024 and June 2025 PIP data vintages. This covers all changes in the data: new surveys ingested into PIP, updated national accounts data used for creating an annual series, revisions to population data for weighting countries, and consumer price indices (CPIs). 52 Table C7: Decomposing changes in millions of poor when switching PPPs, 2022 Poverty rates (%) Impact on millions of poor Country $2.15 $2.50 $3.00 New P P P s New lines Total India 3 2 6 -8 56 48 Pakistan 4 6 14 4 21 25 Indonesia 2 3 8 3 12 16 Philippines 5 7 13 2 7 9 Ethiopia 30 25 37 -7 15 8 Uganda 41 47 59 3 5 8 Nigeria 31 25 35 -14 22 8 Congo, Dem. Rep. 77 78 84 2 6 7 South Africa 21 25 31 3 4 7 Yemen, Rep. 54 58 68 2 4 6 Country $3.65 $4.00 $4.20 New P P P s New lines Total India 31 22 26 -119 58 -61 Nigeria 64 53 56 -25 7 -19 Indonesia 19 21 24 4 8 13 Pakistan 37 37 42 1 11 12 Sudan 64 37 41 -13 2 -12 Bangladesh 30 21 24 -15 5 -9 Ethiopia 69 59 62 -13 4 -8 Egypt, Arab Rep. 13 6 7 -8 1 -7 Syrian Arab Republic 67 36 39 -7 1 -6 Myanmar 24 31 35 4 2 6 Country $6.85 $7.70 $8.30 New P P P s New lines Total China 17 17 21 5 50 55 Indonesia 63 66 70 9 12 21 Iran, Islamic Rep. 23 35 40 11 4 16 Pakistan 83 84 87 3 7 10 Egypt, Arab Rep. 66 51 58 -17 8 -9 Mexico 22 24 27 3 5 7 Philippines 68 71 74 3 4 7 Myanmar 73 80 83 4 2 5 Uzbekistan 17 28 32 4 1 5 Nigeria 91 86 89 -10 5 -5 Note: Also see Table 5 in the main text and Table C9 in the Appendix. The table shows extreme poverty rates estimated from the June 2025 vintage at three poverty lines, namely $2.15 (2017 PPP), $2.50 (2021 PPP), and $3.00 (2021 PPP). $2.50 would be the international poverty line in 2021 PPP terms if only PPPs were updated, holding constant the national poverty lines used in deriving the 2017-PPP-based international poverty line of $2.15. The value of $2.46 is given in Table 2, which is rounded to $2.50. If $2.50 were the new international poverty line, the estimated population in extreme poor would reduce by 8 million in India. The table reports similar results for the lines typical of lower- middle- and upper-middle-income countries. The column labeled “Total” corresponds to the column labeled “Switching PPPs” in Table C6; i.e., the PPP impact shown in Table C6 is further decomposed here. It captures the change in millions of poor between using the 2017 PPP and its associated lines vs. the 2021 PPP and its associated lines, while holding constant the June 2025 vintage. Countries are arranged in descending order of this total absolute change. The top ten countries with the largest total changes are shown for each of the three types of poverty lines. 53 Table C8: Changes in regional and global estimates at higher lines, 2022 Poverty rates (%) Changes in millions of poor R egion $3.65 (Old) $3.65 $4.20 New PI P data Sw itching PPP s Total East Asia & Pacific 5.4 5.4 6.7 -1 29 28 Europe & Central Asia 1.7 2.0 2.4 1 2 3 Latin America & Caribbean 8.9 9.1 9.4 1 2 3 Middle East & North Africa 16.2 16.6 15.0 3 -7 -4 Rest of the World 0.8 0.9 0.9 1 0 1 South Asia 38.8 30.8 27.7 -149 -59 -209 Sub-Saharan Africa 64.2 66.1 63.1 35 -37 -2 World 22.4 20.9 20.1 -109 -71 -180 R egion $6.85 (Old) $6.85 $8.30 New PI P data Sw itching PPP s Total East Asia & Pacific 27.4 27.4 31.8 0 95 95 Europe & Central Asia 8.2 8.9 12.0 4 15 19 Latin America & Caribbean 25.2 25.6 28.6 1 19 20 Middle East & North Africa 45.5 46.0 49.8 6 17 23 Rest of the World 1.3 1.4 1.6 1 3 4 South Asia 78.8 81.4 82.1 60 12 72 Sub-Saharan Africa 87.7 88.3 88.4 24 1 25 World 44.9 45.9 48.0 95 162 258 Note: Also see Table 4 in the main text. This table shows regional and global poverty levels in 2022 using two vintages of PIP data, namely the old September 2024 vintage and the current June 2025 vintage. Poverty is estimated from both PIP data vintages using the $3.65 and $6.85 poverty lines, expressed in 2017 PPP terms. Poverty is also estimated from current vintage of PIP data using the $4.20 and $8.30 poverty lines, expressed in 2021 PPP terms. The impact of new PIP data is captured by the difference in millions of poor when comparing the old and new PIP vintages at a common poverty line (i.e., $3.65 or $6.85 in 2017 PPP). The impact of switching PPPs is captured by the difference in millions of poor when comparing poverty at the 2017-PPP-based poverty lines and 2021-PPP-based poverty lines, using June 2025 PIP data vintage. The total impact is the summation of the two. Alternatively, it is the difference in the headline millions of poor published in September 2024 with the 2017 PPPs and the headline millions of poor published in June 2025 with the 2021 PPPs. 54 Table C9: Decomposing changes in regional and global poverty estimates, 2022 Poverty rates (%) Impact on millions of poor Region $3.65 $4.00 $4.20 New P P P s New lines Total East Asia & Pacific 5.4 5.8 6.7 9 20 29 Europe & Central Asia 2.0 2.1 2.4 1 1 2 Latin America & Caribbean 9.1 8.6 9.4 -3 5 2 Middle East & North Africa 16.6 13.8 15.0 -12 5 -7 Rest of the World 0.9 0.9 0.9 0 0 0 South Asia 30.8 23.7 27.7 -136 77 -59 Sub-Saharan Africa 66.1 60.6 63.1 -67 31 -37 World 20.9 18.3 20.1 -209 139 -71 Region $6.85 $7.70 $8.30 New P P P s New lines Total East Asia & Pacific 27.4 28.3 31.8 19 76 95 Europe & Central Asia 8.9 10.2 12.0 7 9 15 Latin America & Caribbean 25.6 25.7 28.6 1 18 19 Middle East & North Africa 46.0 44.7 49.8 -5 22 17 Rest of the World 1.4 1.4 1.6 1 2 3 South Asia 81.4 78.0 82.1 -66 78 12 Sub-Saharan Africa 88.3 86.5 88.4 -22 23 1 World 45.9 45.1 48.0 -67 229 162 Note: This table shows poverty rates estimated from the June 2025 PIP vintage of PIP, using the standard poverty lines, $3.65 and $6.85 in 2017 PPP terms and $4.20 and $8.30 in 2021 PPP terms. In addition, it shows estimates at alternative poverty lines in 2021 PPP units that isolate the impact of PPPs on the new poverty lines. When one keeps old harmonized national poverty lines and converts these lines to 2021 PPP dollars, the resulting poverty lines are $4.00 and $7.70 for lower-middle-income countries and upper-middle-income countries, respectively (see Table 2 in the main text). For example, if $4.00 were the new poverty line typical of lower-middle-income countries, without accounting for new harmonized national poverty lines, the estimated population of the poor would reduce by 209 million globally. 55 Table C10: Poverty rates at new poverty lines for selected years, percent Region and poverty line 1990 2000 2010 2019 2020 2021 2022 2023 2024 2025 2030 $3.00 (2021 PPP) per day East Asia & Pacific 76.9 52.8 21.2 2.7 2.9 2.8 2.5 2.3 2.0 1.9 1.2 Europe & Central Asia 7.1 13.6 6.6 1.1 1.1 1.1 1.1 1.0 0.9 0.9 0.6 Latin America & Caribbean 21.4 18.6 8.7 6.2 5.8 6.5 5.2 4.7 4.6 4.6 3.9 Middle East & North Africa 12.4 7.3 4.0 7.5 8.6 8.6 8.5 8.7 9.0 9.4 10.8 South Asia 52.8 - 30.1 9.7 10.3 8.8 7.3 6.3 5.4 4.8 2.3 Sub-Saharan Africa 60.9 62.3 49.6 44.7 46.2 46.1 45.5 45.2 45.1 44.4 39.9 Eastern & Southern Africa - 62.8 51.9 52.1 53.8 54.0 53.4 53.2 53.5 52.8 47.5 Western & Central Africa 61.4 - 46.3 33.9 35.2 34.5 33.8 33.3 32.8 32.0 28.5 Rest of the World 0.6 0.6 0.7 0.7 0.5 0.5 0.7 0.7 0.7 0.7 0.6 World 43.6 36.3 21.0 10.7 11.2 11.0 10.5 10.2 10.0 9.9 8.9 FCS - - 37.7 35.7 37.6 38.0 37.7 37.9 38.3 37.9 35.3 $4.20 (2021 PPP) per day East Asia & Pacific 88.1 69.4 36.3 8.9 8.1 6.9 6.7 6.1 5.8 5.5 4.0 Europe & Central Asia 13.3 22.9 9.3 2.8 2.9 2.5 2.4 2.1 2.0 1.9 1.4 Latin America & Caribbean 32.1 28.1 15.4 10.9 10.5 11.1 9.4 8.8 8.6 8.5 7.3 Middle East & North Africa 26.4 18.6 11.7 14.6 16.3 15.7 15.0 15.1 15.4 15.6 16.4 South Asia 78.6 - 60.9 32.7 34.0 31.1 27.7 24.7 22.1 19.8 10.9 Sub-Saharan Africa 73.4 75.5 66.3 61.9 63.6 63.6 63.1 62.9 62.9 62.1 57.5 Eastern & Southern Africa - 75.1 67.5 67.7 69.5 69.9 69.5 69.5 69.8 69.0 63.9 Western & Central Africa 74.7 - 64.5 53.6 54.9 54.4 53.8 53.2 52.7 51.8 47.8 Rest of the World 1.0 0.9 0.9 0.9 0.6 0.6 0.9 0.9 0.8 0.8 0.7 World 55.9 51.0 35.9 21.3 21.9 21.0 20.1 19.3 18.6 18.1 15.4 FCS - - 53.8 51.4 53.8 54.5 54.3 54.6 55.0 54.5 51.4 $8.30 (2021 PPP) per day East Asia & Pacific 96.8 91.1 67.5 37.1 37.7 32.0 31.8 30.3 28.9 27.7 22.2 Europe & Central Asia 36.5 53.3 24.3 14.6 13.9 12.0 12.0 10.1 9.6 9.1 6.9 Latin America & Caribbean 58.7 54.3 39.0 30.8 31.1 31.7 28.6 27.3 26.9 26.5 23.5 Middle East & North Africa 67.8 57.6 49.0 49.9 53.0 51.4 49.8 49.0 49.0 48.7 45.3 South Asia 97.1 - 93.1 84.4 85.2 83.8 82.1 80.1 78.1 76.1 62.9 Sub-Saharan Africa 90.3 91.6 88.7 87.6 88.5 88.5 88.4 88.3 88.3 88.0 85.7 Eastern & Southern Africa - 90.5 88.1 88.9 89.8 90.0 89.9 90.1 90.2 90.0 88.0 Western & Central Africa 92.4 - 89.7 85.7 86.5 86.3 86.2 85.8 85.6 85.1 82.4 Rest of the World 2.6 2.0 1.8 1.7 1.5 1.3 1.6 1.5 1.5 1.5 1.3 World 70.6 70.7 60.3 49.6 50.4 48.6 48.0 47.0 46.3 45.5 41.0 FCS - - 79.4 79.2 81.2 81.7 82.0 82.2 82.4 82.2 80.3 Note: Using the June 2025 PIP vintage. The definition of countries in fragile and conflicted-affected situations (FCS) is kept fixed using the World Bank fiscal year 2025 classification for all years. Poverty estimates are not reported for years with insufficient data coverage (that is, less than 50 percent of the regional population). These missing observations are marked (-). However, poverty estimates are presented for the recent years (2019–23) using nowcasting methods, even if there is insufficient data coverage, and are thus grayed out (for example, Sub-Saharan Africa and FCS). Estimates presented for all regions and country groups for 2024, 2025, and 2030 are based on poverty nowcasts or projections and are also grayed out. For methodological details on data coverage and poverty nowcasts and projections, see World Bank (2025). 56 Table C11: Millions of poor at new poverty lines for selected years Region and poverty line 1990 2000 2010 2019 2020 2021 2022 2023 2024 2025 2030 $3.00 (2021 PPP) per day East Asia & Pacific 1231 960 418 57 61 59 54 48 43 40 26 Europe & Central Asia 33 64 32 5 6 5 5 5 5 4 3 Latin America & Caribbean 94 96 51 39 37 42 34 31 30 30 27 Middle East & North Africa 29 21 14 31 36 37 37 39 41 43 53 South Asia 602 - 504 182 195 168 142 122 107 95 48 Sub-Saharan Africa 317 424 444 510 540 553 559 569 583 587 593 Eastern & Southern Africa - 255 275 352 374 385 391 400 412 416 423 Western & Central Africa 129 - 169 157 167 168 168 170 171 171 170 Rest of the World 6 6 7 8 6 5 8 8 8 8 7 World 2312 2234 1470 832 882 869 838 822 817 808 757 FCS - - 299 348 374 387 391 401 415 421 436 $4.20 (2021 PPP) per day East Asia & Pacific 1411 1263 716 187 171 148 143 131 123 118 85 Europe & Central Asia 62 108 45 14 14 12 12 11 10 9 7 Latin America & Caribbean 140 145 90 69 67 71 61 58 57 56 50 Middle East & North Africa 62 54 41 61 69 67 65 67 70 72 81 South Asia 897 - 1018 614 645 595 536 483 436 395 228 Sub-Saharan Africa 382 514 593 706 743 763 776 793 811 821 854 Eastern & Southern Africa - 305 358 458 482 498 508 522 537 544 569 Western & Central Africa 157 - 235 248 261 264 267 271 275 277 285 Rest of the World 9 9 9 10 7 6 10 10 9 9 8 World 2963 3141 2511 1660 1717 1663 1603 1552 1516 1481 1314 FCS - - 427 501 536 556 564 578 595 605 636 $8.30 (2021 PPP) per day East Asia & Pacific 1550 1657 1333 784 801 681 679 648 618 592 476 Europe & Central Asia 170 253 116 73 69 60 59 50 47 45 34 Latin America & Caribbean 256 280 227 196 199 204 185 178 177 175 160 Middle East & North Africa 159 167 173 208 224 221 218 218 222 224 224 South Asia 1108 - 1557 1582 1614 1605 1585 1564 1540 1515 1316 Sub-Saharan Africa 471 624 794 998 1034 1061 1087 1113 1140 1164 1274 Eastern & Southern Africa - 368 467 601 624 642 658 676 694 710 783 Western & Central Africa 194 - 327 397 411 420 429 437 446 454 491 Rest of the World 23 19 18 18 17 14 18 17 17 17 15 World 3739 4354 4219 3859 3959 3846 3832 3789 3762 3732 3500 FCS - - 629 772 809 832 851 871 893 912 993 Note: Using the June 2025 PIP vintage. The definition of countries in fragile and conflicted-affected situations (FCS) is kept fixed using the World Bank fiscal year 2025 classification for all years. Poverty estimates are not reported for years with insufficient data coverage (that is, less than 50 percent of the regional population). These missing observations are marked (-). However, poverty estimates are presented for the recent years (2019–23) using nowcasting methods, even if there is insufficient data coverage, and are thus grayed out (for example, Sub-Saharan Africa and FCS). Estimates presented for all regions and country groups for 2024, 2025, and 2030 are based on poverty nowcasts or projections and are also grayed out. For methodological details on data coverage and poverty nowcasts and projections, see World Bank (2025). 57 Figure C5: Distribution of millions of poor across regions and countries, 2025 a. $3.00 (2021 PPP) b. $4.20 (2021 PPP) 58 c. $8.30 (2021 PPP) Legend Note: The 10 countries with the most poor people are labeled. Using the June 2025 PIP vintage. 59 Figure C6: Share of extreme poor in fragile Sub-Saharan Africa, 1990 – 2025 Note: The figure shows the share of the extreme poor globally that live in fragile countries in Sub-Saharan Africa. The definition of countries in fragile and conflicted-affected situations is kept fixed using the World Bank fiscal year 2025 classification for all years. This figure uses two vintages of PIP data, the old September 2024 vintage and the current June 2025 vintage. Poverty is estimated from both data vintages using the $2.15 poverty line, expressed in 2017 PPP terms. Poverty is also estimated from current vintage of PIP data using the $3.00 poverty lines, expressed in 2021 PPP terms. With the September 2024 PIP data, the estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, the estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color. 60 Figure C7: Share of extreme poor in low- and middle-income countries, 1990 – 2025 Note: The figure shows the share of the global extreme poor that live in low- and middle-income countries. Countries’ income classification changes from year to year, which causes the large fluctuations in this indicator. This figure uses two vintages of PIP data, the old September 2024 vintage and the current June 2025 vintage. Poverty is estimated from both data vintages using the $2.15 poverty line, expressed in 2017 PPP terms. Poverty is also estimated from current vintage of PIP data using the $3.00 poverty lines, expressed in 2021 PPP terms. The figure includes nowcasts. The horizontal line is drawn at 50% to indicate which group accounts for the majority of the global poor. With the old September 2024 PIP data, the estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, the estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color. 61 Figure C8: Global trends in societal poverty, 1990 - 2025 Note: The societal poverty line is given as max(2.15, 1.15 + 0.5*Median) in 2017 PPP dollars or max(3.00, 1.30 + 0.5*Median) in 2021 PPP dollars. This figure uses two vintages of PIP data, the old September 2024 vintage and the current June 2025 vintage. Poverty is estimated from both data vintages using the 2017 PPPs. Poverty is also estimated from current vintage of PIP data using the 2021 PPPs. With the old September 2024 PIP data, poverty estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, poverty estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color. 62 Figure C9: Global inequality trends, 1990 - 2025 Note: This figure uses two vintages of PIP data, the old September 2024 vintage and the current June 2025 vintage. The global Gini is estimated from both data vintages using the 2017 PPPs. The Gini is also estimated from current vintage of PIP data using the 2021 PPPs. With the old September 2024 PIP data, the estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, the estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color. 63 Table C12: Changes in global inequality, 2022 Gini Mean log MLD MLD MLD (between Vintage index deviation (MLD) (within) (between) share), % Sep 2024 (2017 PPP) 61.8 72.8 23.3 49.6 68.1 Jun 2025 (2017 PPP) 61.5 71.5 22.0 49.5 69.3 Jun 2025 (2021 PPP) 60.7 68.8 21.8 47.0 68.3 Changes due to … New PIP data -0.25 -1.35 -1.31 -0.04 1.22 Switching PPPs -0.85 -2.65 -0.16 -2.49 -0.95 as share of total (%) 77 66 11 98 Total change -1.10 -4.00 -1.47 -2.53 0.27 Note: The table shows global inequality in 2022 using the Gini index and the mean log deviation (MLD, x100), for various vintages of the data. The MLD is also decomposed into differences within and between countries. The total change between September 2024 vintage + 2017 PPPs versus June 2025 vintage + 2021 PPPs is decomposed into the change due to new data (change from September 2024 to June 2025 vintage, while using the 2017 PPPs) and switching PPPs (change from 2017 PPPs to 2021 PPPs, while using the June 2025 vintage). New PIP data refers to all changes in poverty data in PIP between the September 2024 and June 2025 PIP data vintages. This covers all changes in the data: New surveys ingested into PIP, updated national accounts data used for creating an annual series, revisions to population data for weighting countries, and consumer price indices (CPIs). The change in PPPs slightly affects within- country inequality because it impacts the within-country spatial deflation in China. 64 Figure C10: Trend in the Global Prosperity Gap, 1990 – 2025 Note: The prosperity standard is $25 (2017 PPP) or $28 (2021 PPP). This figure uses two vintages of PIP data, the old September 2024 vintage and the current June 2025 vintage. The global prosperiy gap is estimated from both data vintages using the 2017 PPPs. It is also estimated from current vintage of PIP data using the 2021 PPPs. With the old September 2024 PIP data, the estimates for 2023 and 2024 are nowcasts, which are indicated in the chart with light grey color. With the June 2025 PIP data, the estimates for 2024 and 2025 are nowcasts, which are indicated in the chart with a darker grey color. 65 Figure C11: Change in income or consumption when moving from 2017 to 2021 PPPs in 2022, by global percentile Note: This chart plots the rate of change in welfare for each percentile of the global welfare distribution in 2022 when moving from the 2017 PPPs to 2021 PPPs, using the June 2025 vintage of PIP data. In other words, it plots the delta ratio for each global percentile. The percentiles are anonymous, i.e., the same percentile could consist of different country-groups under the two PPPs. The mean delta ratio is population weighted. The US delta ratio is US inflation between 2017 and 2021. 66