Transitioning to **No Poverty**

Transitioning to Sustainability Series

MDPI • Basel • Beijing • Wuhan • Barcelona • Belgrade • Manchester • Tianjin • Tokyo • Cluj

EDITORS Isabel Günther Department of Humanities, Social and Political Sciences, ETH Zurich, Switzerland

EDITORIAL OFFICE MDPI St. Alban-Anlage 66 4052 Basel, Switzerland

Rahul Lahoti NADEL - Center for Development and Cooperation, ETH Zurich, Switzerland

For citation purposes, cite each article independently as indicated below:

Author 1, and Author 2. Year. Chapter Title. In *Transitioning to No Poverty*. Edited by Isabel Günther and Rahul Lahoti. Transitioning to Sustainability Series 1. Basel: MDPI, Page Range.

© 2021 by the authors. Chapters in this volume are Open Access and distributed under the Creative Commons Attribution (CC BY 4.0) license, which allows users to download, copy and build upon published articles, as long as the author and publisher are properly credited, which ensures maximum dissemination and a wider impact of our publications. The book taken as a whole is © 2021 MDPI under the terms and conditions of the Creative Commons license CC BY-NC-ND.

**ISBN 978-3-03897-860-2 (Hbk) ISBN 978-3-03897-861-9 (PDF) ISSN: 2624-9324 (Print) ISSN: 2624-9332 (Online) doi:10.3390/books978-3-03897-861-9**

## **In memory of Prof. Stephan Klasen, Ph.D. (October 27, 2020)**

who dedicated his entire life to fight global poverty with science, who contributed to this book in 2019, at a time when the incurable disease amyotrophic lateral sclerosis (ALS) already made writing incredibly difficult for him, who has been a life-long inspiration for the editors of this book, and who has been a good friend to many authors of this book.

## **Contents**



## **About the Editors**

Isabel Günther is Professor of Development Economics at ETH Zurich. She is the academic director of the NADEL Center for Development and Cooperation (www.nadel.ethz.ch) and ETH for Development (www.ethz.eth4d.ch). Through her research and teaching, she aims to help address global inequalities and poverty and to strengthen the collaboration between science, politics, and society. She has conducted research and taught classes in Benin, Burkina Faso, Germany, France, Ghana, Kenya, Switzerland, South Africa, Uganda, and the United States.

Rahul Lahoti is a post-doctoral researcher at ETH, Zürich. He holds a Ph.D. from the University of Göttingen, Germany, and a Master's degree in Public Administration from Columbia University, New York. His research focuses on issues relating to the measurement of poverty, inequality, gender, labor markets and political economy. He has published in several top-rated academic journals including the *Journal of Economic Behavior and Organization, Journal of Economic Inequality, World Development, Feminist Economics*, and *Review of Income and Wealth*.

## **Contributors**

#### ALEJANDRO DE LA FUENTE

Dr., Senior Economist at the Poverty and Equity Global Practice of the World Bank.

#### ANITA BAKU

Dr., Senior Lecturer, Department of Public Administration and Health Services Management of the University of Ghana Business School.

#### EDWARD ASIEDU

Dr., Lecturer, Department of Finance (Development Finance Group) University of Ghana Business School (UGBS), University of Ghana. Affiliate Research Fellow, Chair of Development Economics, University of Passau, Germany.

#### FRITZ BRUGGER

Dr., Senior Scientist, Center for Development and Cooperation (NADEL), Department of Humanities, Social and Political Sciences, ETH Zurich, Switzerland.

#### GÜNTHER FINK

Associate Professor of Epidemiology and Household Economics, University of Basel, Switzerland.

Head of Household Economics and Health Systems Research Unit, Swiss Tropical and Public Health Institute, Switzerland.

#### JAN PRIEBE

Dr., German Institute for Global and Area Studies (GIGA), Georg-August-Universität Göttingen, Germany.

#### JANN LAY

Apl. Prof. Dr., German Institute for Global and Area Studies (GIGA), Georg-August-Universität Göttingen, Germany.

#### JOE HASELL

MSc, DPhil candidate, Department of Social Policy and Intervention, University of Oxford.

#### KANCHANA N. RUWANPURA

Professor of Human Geography at the Institute of Geography, University of Gothenburg, Sweden. Honorary Fellow at the Centre for South Asian Studies, University of Edinburgh, Scotland.

#### KATHLEEN BEEGLE

Dr., Research manager, Development Research Group, World Bank.

#### MARTIN RAVALLION

Dr., Department of Economics, Georgetown University, and Ungku Aziz Centre, University of Malaya.

#### MAX ROSER

Dr., Director, Oxford Martin Programme on Global Development, Oxford Martin School, University of Oxford, UK.

#### MEGAN TODD

MSc, Former graduate of the Institute of Geography, University of Edinburgh, Scotland.

#### RAINER THIELE

Professor Dr., International Development Research Center, Kiel Institute for the World Economy, Germany.

#### SANJAY G. REDDY

Associate Professor of Economics, Department of Economics, The New School for Social Research, USA.

#### STEPHAN KLASEN

Professor of Development Economics, University of Göttingen, Germany.

#### SERVAAS VAN DER BERG

Professor of Economics, Resep (Research on Socio-Economic Policy) Department of Economics, Stellenbosch University, South Africa.

### **Abstracts**

#### **The Fight against Global Poverty: 200 Years of Progress and Still a Very Long Way to Go by Max Roser and Joe Hasell**

Almost one in ten people globally live on less than \$1.90 per day and recent projections suggest we are not on track to achieve the goal of eradicating such extreme poverty by 2030. What the long-run history of global poverty shows clearly, however, is that the continued presence of extreme poverty is far from inevitable. The aim of this chapter is to inform our aspirations for the future of global poverty by summarising what we know about its history. To provide this long-term perspective, we present global poverty estimates based on two methods: household survey-based estimates from the World Bank and national accountsbased estimates derived from historical data on GDP per capita and inequality. The latter method allows us to estimate the trajectory of global poverty over the last two centuries. We compare the methods and data underlying these two approaches to poverty measurement. There are discrepancies between the resulting estimates and sources of uncertainty in each case. However, there are key points of agreement, and the main trends on which they converge are robust to various sources of uncertainty. This evidence shows a substantial decline in global poverty rates over the last two centuries, with particularly fast progress made in recent decades. The extent of the changes we see over the long run should embolden us to reach not only for the eradication of the most extreme forms of poverty but for much more ambitious goals still.

#### **Global Absolute Poverty: The Beginning of the End? by Sanjay G. Reddy**

The first Sustainable Development Goal of "Ending poverty in all its forms everywhere" must be interpreted in light of an understanding of what "poverty in all its forms" means societally. It cannot be reduced to a narrow technical focus on official targets and indicators. There are reasons for concern that the official indicators are unsatisfactory, and that there is a considerable gap between their focus and the societal understanding of poverty. Even if conventional approaches to global poverty estimation are used, considering a range of alternative poverty lines demonstrates that the choice of poverty identification criterion can significantly influence conclusions drawn about how much poverty there is, where it is, how it is evolving over time, and what are the appropriate priorities and policies. Although poverty is expected to be nearly "eliminated" by 2030 in regions other than sub-Saharan Africa at the lowest poverty lines, this is not true at higher poverty lines. The projected regional composition of future poverty is also greatly dependent on the choice of poverty line, because poverty in the other world regions increases markedly at higher lines. These conclusions undermine the widespread presumption that addressing the problem of absolute poverty worldwide requires a singular focus on sub-Saharan Africa. Especially at higher poverty lines, income poverty is a global problem, and sustaining growth throughout the developing world is important for its reduction. Comparison of estimates based on pre- and post- pandemic growth forecasts shows that while the elimination of poverty by 2030 was already unlikely, the global economic contraction due to COVID-19 has made it even more so. For the highest poverty line examined, an additional 500 million people will be poor in 2030 as a result of the growth slowdown due to Covid-19.

#### **SDG 1: The Last 3% by Martin Ravallion**

There is a little-noticed but important difference between the World Bank's original goal for poverty reduction and the subsequent UN Sustainable Development Goal (SDG). While both target the "\$1.90 a day" poverty rate, the Bank's goal was a 3% rate by 2030, while the SDG is to "eradicate" poverty by 2030. Simple linear projections of recorded progress against \$1.90 poverty in the world does suggest that we are on track to attaining the UN's goal. If we can return to the pre-COVID pace of poverty reduction after two or three years, then we should still be roughly on track. However, closer scrutiny of the pre-COVID data leaves one less optimistic. There are a priori reasons why the last few percentage points could be harder to reach with current development policies. Consistent with that hypothesis, the paper documents recent (pre-COVID-19) signs of a levelling-off in progress for the poorest in East Asia—the star performer regionally over the longer term. This is evident in the region's slower progress recently in both lifting the floor—and thus reaching the poorest—and in reducing the poverty rate. This levelling off is also found, on average, for the 18 developing countries that have reduced their poverty rate from over 10% (around the current global rate) to under 3% during the period 1981–2017. Similar to East Asia, progress in reaching the poorest declined once the last 3% had been reached, though some countries did better than others. Overall, the results suggest that returning to "business as usual" post-COVID will not suffice to eradicate extreme poverty.

#### **How Can the International Community Eradicate Poverty and Hunger by 2030? by Stephan Klasen**

In vowing to eradicate poverty and hunger by 2030, the international community has set itself an extremely ambitious goal as the success of this venture depends on effectively addressing a number of challenges in the poorest countries. The classic instruments of development cooperation can contribute little to this agenda. Instead, a much broader agenda must become a political priority of the international community and must involve political, economic, and occasionally, even military engagement.

#### **"Leave No One Behind" in Middle-Income Countries. A Review of Progress and Policies by Jann Lay and Jan Priebe**

We show that "Leave no one behind" (LNOB) can be a meaningful guiding principle for national development policy as well as development cooperation in middle-income economies, as very unequal progress threatens the gains for the poor in countries graduating from low to middle-income status. Our review, measuring the progress of LNOB, clearly illustrates the huge data gaps that remain for key LNOB indicators. Disaggregated indicators or data to compute them are often not (yet) available. Our brief and selective review of LNOB-relevant policies and approaches in middle-income countries shows a very rich foundation for evidencebased policy-making, particularly in education, health and social protection. Some clear messages emerge, for example, a clear call for progressive universal policies in education that emphasize equality in learning achievements. In general, the sectorally interlinked challenges of implementing LNOB demand integrated approaches that combine education, health, and labor market components and pay attention to mainstreaming anti-discrimination efforts.

#### **SDG 1 and Women's Work: Ignoring the Needs of Women and History—The Case of Sri Lanka by Megan Todd and Kanchana N. Ruwanpura**

Our paper evaluates SDG 1 with regard to labour reforms in Sri Lanka, by seeking to understand how women workers in the garment sector may be affected by proposed changes to the laws. Our paper is based on a decade of fieldwork, supplemented by interviews and recent archival work. We first give an account of Sri Lanka's recent history. It offers a necessary context, given that politics and legacies from the ethnic conflict continue to mar current efforts. Critically, it is found that ethnic divisions continue to create tensions, and that these have often been exacerbated by labour conditions. Through investigating the place of women workers in Sri Lanka's apparel sector, including the North and the East of Sri Lanka, we show that labour insecurity remains, and discrimination is rife. Importantly, Sri Lanka has thus far failed to position women's experiences accurately in SDG1 by failing to consider its linkages to SDG5, SDG8 or SDG10. The current inability of policies to alleviate the position of women workers in relation to SDG1 places added importance to recent labour policy reform, which tends to be neglected because of an emphasis on creating pro-market friendly labour conditions.

#### **Social Protection in Ghana—History, Equity- Driven Reforms, Financing and Sustainability by Edward Asiedu and Anita Baku**

Social protection has become very important in development policy, due partly to the widening gap in health, income and opportunities between the rich and the poor. The growing number of national social protection policies and interventions implemented by many developing countries, particularly in sub-Saharan Africa, ties into the emerging consensus around the view that social protection provides an effective response to poverty and vulnerability in developing countries. This chapter examines the history of social protection in Ghana, highlighting the key social protection reforms and interventions to assess how they are aligned with the attainment of societal equity. The historical antecedent of social protection policies and programs in the past three to four decades is therefore provided in order to illustrate a rigorous background for the design and strengthening of social protection in the decade ahead. We also provide a discussion on the financing of social protection policies and their long-term sustainability, emphasizing the role of technology in the sustainable delivery and targeting of social protection programs.

### **Education Access and "Learning Poverty" in Seven Southern African Countries**

#### **by Servaas van der Berg**

Against the backdrop of the shift in emphasis from the MDGs, with the educational focus on access, to the SDGs with the focus on educational outcomes and equity, this chapter discusses some education issues and policy responses in seven southern African countries. These countries—South Africa and its six neighbours—cover a wide economic development range: Mozambique is a lowincome country, Lesotho, Zimbabwe and Eswatini lower-middle income, and Namibia, South Africa and Botswana upper-middle income countries. Examples from these countries show little evidence that the focus in policy debates and practice has shifted to the educational goals formulated and propagated by the international community. This may well also be the case in many other developing countries. The continued focus on broadening access needs to be supplemented with steps to reduce "learning poverty".

#### **Early Childhood Development: Current Status and Gaps by Günther Fink**

We review the literature on early childhood development as well as the current knowledge on developmental gaps between high-, middle- and lowincome countries. While current data on children's early developmental outcomes are limited, the available evidence suggests that early trajectories are comparable among children growing up in home environments providing adequate support and stimulation globally. Large gaps in physical and likely also cognitive early development persist in low- and middle-income countries due to poverty, lack of maternal education and lack of early learning opportunities. These gaps can be reduced by continued efforts to reduce poverty and increase education, as well as targeted government programs to support parents and children during the first few years of children's lives.

#### **Mobilizing Resources for the Poor by Kathleen Beegle and Alejandro de la Fuente**

The SDG agenda to address poverty needs to extend beyond shifting programs and policies. It also requires a careful revisit of a range of fiscal issues especially in countries in Africa that are poor and also resource constrainted. Current levels of public spending in Africa that effectively reach and benefit the poor are not nearly sufficient and often poorly spent. This chapter explores how poverty reduction can be accelerated by mobilizing more resources, domestically and internationally, and by spending more efficiently and with a greater focus on the needs of the poor in terms of both raising their income today and investing in the next generation in Africa. What is the path to tackle these challenges? First, on the revenue side countries need to mobilize more resources domestically. While mobilizing domestic revenues (with VAT expansion currently a favorite vehicle), countries need to make sure the poor are net receivers. Other promising avenues include improving tax compliance, with a larger focus on local large taxpayers, corporate taxes and transfer (mis)pricing (which has a global agenda), as well as excise and property tax collection. Yet, even with improvements in domestic resource mobilization, international development assistance will still be critical in the poorest and most fragile countries, for both direct spending as well as to leverage private capital. Aid makes up more than 8 percent of GDP for half of low-income countries in Africa, but in recent years aid to countries in the region has been declining. Second, spending patterns need to shift towards more pro-poor investments and improve in terms of the levels spent in critical sectors, the instrument/programs for a given investment, and the efficiency of implementation. In levels, spending on "pro-poor" sectors has a mixed track record with some generally reaching international targets (like education) but others falling short for many countries (health, WASH, risk management, and agriculture and rural infrastructurel). The choice of program design matters for given spending— untargeted programs can result in large shares of spending going to non-poor households. One obvious area for attention are the currently high subsidy expenditures (in energy and fertilizer)—often regressive with little impact on poverty. Cash transfers seem more effective and efficient than subsidies where evidence exists, but more is needed to compare their performance relative to public good provision for the poor in agriculture and rural infrastructure, security, risk management, education and health. Agricultural and rural spending should tilt more heavily towards investment in public goods. And finally, there are significant inefficiencies in spending that need to be addressed. The low quality of health and education services is not only explained by low spending levels.

#### **Development Cooperation, Growth and Poverty Reduction: A Survey of the Evidence by Rainer Thiele**

The donor community has taken a prominent role in the implementation of the Millennium Development Goals (MDGs), and is likely to stay strongly involved when it comes to achieving the poverty-oriented targets of the Sustainable Development Goals (SDGs). Against this background, the present paper provides an overview of the empirical evidence regarding the impact of international development cooperation on economic growth, (monetary and non-monetary) poverty and inequality in order to assess whether donors have directly or indirectly contributed to achieving internationally agreed upon poverty reduction targets. The general conclusion is that development cooperation can help achieve growth and poverty reduction in partner countries, even though the effects are likely to be modest. Most confidence can be put into the finding that, in accordance with the MDGs, aid for social infrastructure has contributed to achieving non-monetary goals such as higher school enrollment and lower infant mortality. In contrast, it is inherently difficult to empirically identify income effects of foreign aid at the macro level, which the long-standing and still unresolved debate about the aid–growth relationship illustrates.

#### **A Safety Net for You, a Safety Net for Me? Donor Promotion of Social Protection Schemes Faces Policy Coherence Issues by Fritz Brugger**

Social protection schemes are effective instruments to fight poverty in "normal" times. During crises, they are even more important to prevent people from falling into poverty, as the COVID-19 pandemic has demonstrated. Developing countries have long struggled to generate sufficient tax revenue to fund the social programs and investments needed to protect their populations. In recent years the donor community has increasingly converged on a consensus around the need to boost tax revenue as part of the broader development agenda. At the same time, donors have also promoted and defended international tax standards that benefit their business communities, and which, to a degree, work against the ability of developing countries to obtain a fair share of taxes from multinational corporations operating in their jurisdictions. The rules for taxing multinational enterprises are brokered by the Organization for Economic Co-operation and Development (OECD), the club of industrialized countries where donor governments are heavyweights. This chapter analyzes the reform of the OECD transfer pricing regulation after the financial crisis, which promised to simplify transfer pricing rules in a way that strengthens the position of developing economies towards multinational enterprises. The reform largely failed to deliver on its promise but was a success for those who benefit from the status quo. The political economy behind the reform points to the lack of policy coherence among donor countries and the deep politicization of the seemingly technical topic of international tax policy.

## **Preface: Transitioning to no poverty by 2030**

**Isabel Günther and Rahul Lahoti**

The first sustainable development goal (SDG 1) is "to end poverty in all its forms everywhere and for all", which seems to be non-negotiable for the "world we want", which would provide minimum living standards for all global citizens. However, the question remains: is this goal feasible, especially given the set-backs in the fight against poverty and the unequal access to health resources that we have observed in 2020 and 2021? If yes, how do governments and civil society need to engage and what resources are needed? In this book, we bring together a diverse set of perspectives on SDG 1 from leading scholars around the world. When we first invited scholars in 2019 to reflect on ending poverty by 2030, the world looked very different; or rather, the necessity of global social protection, decent and not only sufficient livings standards and strong international cooperation to fight global poverty became even more apparent—and the pandemic put a spotlight on the lack of these three components. The COVID-19 pandemic and the resultant lockdowns across the world have not only led to millions of lost lives across the world, but have also led to large numbers of people falling once again into extreme poverty.

The exact impact of the pandemic on poverty is already difficult to ascertain in the short term, due to data limitations in tracking the living conditions of the most marginalized communities in the world; and more so in the long term: it depends on the speed of the economic recovery across countries, as well as on the long-term consequences that the pandemic has on the health and education of poor households that might lead to long term poverty traps.

Given the uncertainty, estimates of researchers on the short-term impacts vary, but all agree that many of the gains made in the last twenty years in reducing poverty are likely to be lost. Lakner et al. (2021) estimate that around 120 million people around the world have additionally become extremely poor in 2020. In the absence of the pandemic, poverty was expected to decline by more than 30 million people in the year 2020. Sumner et al. (2020) estimate that, in some regions of the world, we might lose 30 years' worth of gains against poverty reduction. In their worst-case scenario, poverty would increase by about half a billion across the world. Using IFPRI's global model, Vos, Laborde et al. (2021) estimate that over 150 million additional people could have fallen into extreme poverty in 2020. Some initial estimations also indicate large, long-lasting impacts. UNDP (2020) estimates that the number of people living

in extreme poverty will increase by 44–250 million (up from 861 million in No Covid Scenario) in 2030 because of the pandemic. In Chapter 2, Reddy estimates that an additional 500 million individuals will be poor (defined as less than 5 international dollars per day) in 2030 due to COVID-19.

This book starts with a broad perspective on the feasibility of ending extreme poverty by 2030—the likelihood of which has obviously decreased since the start of writing the chapters in 2019 and its publication in 2021. The first three chapters delve into data to present the current status of the progress made against the goals and projections of a reduction in poverty by 2030 across world regions (chapters 1–3). These chapters also discuss several challenges when it comes to measuring extreme poverty, pointing towards our limited understanding of the goal as such. The next six chapters (chapters 4 to 9) discuss the role of policies in ending poverty in all nations, which is now necessary to achieve the goal of no global poverty by 2030. The authors discuss selected essential policies needed to leave no low-income country and no person in middle-income countries behind in more detail, with a particular focus on social protection and investments in new generations, using case studies of countries from across the world. The last three chapters (chapters 10 to 12) discuss ways to mobilize the needed resources to successfully eliminate poverty in the 21st century: achieving SDG 1 requires not only good policies, but also substantive resources to invest in various poverty reduction policies.

#### **Part 1: Is the World on Track to Achieving SDG 1?**

A long list of indicators have been proposed for tracking the progress of SDG 1 of "ending poverty in all its forms everywhere" (United Nations 2017). These indicators refer to those deprived according to the international poverty line, the national poverty lines, poverty in all its dimensions, and those who lack access to basic public services and social protection. In addition, these indicators are meant to be disaggregated and tracked by sex and age. However, there is a lack of data to monitor all these indicators, especially by disaggregated groups. Hence, the international poverty line of 1.90 international dollars per day, which clearly defines the most extreme forms of poverty, is among the most likely indicator to be monitored and used to determine success in meeting SDG 1 by the international community by 2030. In chapters 1–3, we present diverse views on the measurement, history and potential future of global poverty. These chapters offer first-hand insights into both the challenges of understanding global poverty and achieving SDG 1.

In chapter 1, Roser and Hasell document the long-term history of poverty reduction and discuss the learnings from history for future progress in reducing poverty. They first show that from 1981 to 2017, extreme poverty in the world (defined as people living below 1.9 international dollars per day) has declined from 42 percent to 9 percent. They also document the changes using a societal poverty line that combines the absolute poverty line with a relative poverty line, defined based on average incomes in the individual's society. This poverty rate is not only higher, but has also declined at a slower rate: from 45 percent in 1990 to 28 percent in 2017. Similarly, they show that for higher absolute international poverty lines (such as, for example, 5.5 international dollars per day), the pace of decline in poverty has been slower as compared to that of a decline in extreme poverty. To further analyze historical poverty over the last two hundred years, they use average income data from national accounts and historical data on the extent of inequality in each country. These data have, of course, important limitations concerning coverage and quality, but the authors show that even though the exact extent of poverty reduction is difficult to deduce, the broad trends are clearly evident from the data. They find a more or less continuous poverty decline since 1820, with rapid acceleration in the second half of the 20th century. They argue that, based on historical progress made when it comes to poverty, reduction elimination is possible.

In chapter 2, Reddy takes a more critical perspective on how extreme poverty is defined globally and what this means for reaching SDG 1. The author argues that the societal understanding of poverty differs substantially from the technical understanding of how poverty is defined by governmental organizations. A societal understanding of no poverty would mean that everyone is able to meet their basic needs (food, clothing, shelter, etc.), whereas the technical definition is based on the 1.9 international dollars poverty line, which might not be enough to meet the costs to fulfill these necessities in countries that do not belong to the very poorest countries. There is a likelihood that SDG 1 will be met according to the technical view, but not by the societal view of what constitutes poverty. Reddy further estimates poverty using different poverty thresholds and various growth estimates, to determine if the world can achieve the SDG 1 targets. The estimations show that, for the 1.9 dollars international poverty line, poverty would be below 3 percent in almost all regions around the world by 2030, except in Sub-Saharan Africa. In Sub-Saharan Africa, poverty in 2030 is estimated to be between 32 and 53 percent, based on the growth projections that are used. However, the outlook and where the poor are located changes substantially if higher, and according to Reddy, "more realistic poverty" lines are used (2.52, 3.02 and 5.04 international dollars). Reddy estimates that with a

poverty line of 5.04 international dollars (same as the United States Department of Agriculture food poverty line for the United States), about 35 percent of the world would be deemed poor in 2030, with a far larger fraction coming from South Asia.

In chapter 3, Ravallion argues that it is difficult to achieve the SDG 1 goal of "eradicating" poverty, because the current set of development policies do not reach the poorest. The author argues that linear projections currently used to estimate poverty rates in 2030 based on historical trajectories are misleading. The linear projections might indicate that we are on track to achieve SDG 1 goals, but we might fall short because of both slowing growth rates over the next ten years (even before the COVID-19 crisis) and the difficulty in reaching the very poorest, which has already become evident in many countries. The poorest might be hard to reach due to their remote physical location, persistent social exclusion, dynamic poverty traps and deficiencies in state capacity and policy in reaching the very poorest. Ravallion argues that the progress in poverty reduction declines substantially once a country reaches a poverty level close to 3%. Ravallion investigates the experience of 18 middle-income countries across the world that had reduced their poverty rate to 3 percent, and their success in terms of reducing poverty further. Ravallion finds that countries that have been successful in reducing poverty, both in East Asia and other places, witnessed a substantial decline in the pace of poverty rate reduction once they reached 3 percent levels. This experience was not driven by declines in growth rates, but by an inability to reach the poorest in society. The chapter argues that "business as usual" development policies might not be enough to reach the poorest and to achieve SDG 1, and that a change in policy frameworks is needed

#### **Part 2: Policies to End Extreme Poverty**

Building on the analysis of Ravallion, Klasen lays out the required policies for low-income countries, to end widespread poverty in the 21st century (chapter 4). Complementing chapter 4, Lay and Priebe focus instead on the poorest people in middle-income countries and the future policies needed to "leave no one behind" until 2030. Chapters 6–9 discuss two important dimensions of this policy agenda, namely social protection to reach the extreme poor left out of any growth processes, and investments in the future generation, so that as many people as possible can benefit from and contribute to the development of their countries.

In chapter 4, Klasen first argues that the growth-led reduction in poverty that helped achieve the Millennium Development Goal 1 (MDG 1) by 2015, namely reducing the share of people below the extreme international poverty line and who

suffer from hunger by 50%, cannot be relied on any further for achieving SDG 1. MDG 1 was achieved because large populous countries like China and India were successful in substantially reducing poverty, even though many smaller countries did not achieve the MDG target. However, since SDG 1 calls for an elimination of poverty in each country in the world, for the goal to be met, all countries have to be successful, and that will be disproportionately harder. Moreover, the drastic declines in poverty in Asia have been led by a boost in agricultural productivity, industrialization and the development of export-oriented sectors. These three conditions might not be easy to implement across the poorest countries, who are still facing high extreme poverty rates of up to 30% in the 21st century.

Klasen identifies commodity sector-driven growth, conflicts, climate change and lack of fertility declines across different groups of countries as impediments to achieving SDG 1. High commodity prices that have helped several African countries to reduce poverty in recent years are no longer sustainable. The ill effects of a strong commodity sector—lack of structural change, lack of investment in agriculture and industry—severely limit any further progress in poverty reduction in these countries. A large number of poor people also live in conflict-ridden fragile states, and progress in these countries without international political (and military) commitment does not seem possible. The accelerating effects of climate change will further impact progress in poverty reduction in several countries. Parts of Africa have not experienced the decline in fertility rates usually associated with countries as they develop, leading to increased strain on scarce resources. To address these challenges, Klasen calls upon the poorest countries to implement country-specific policies (depending on the major challenge faced) to make progress in poverty reduction and to develop a broad social security net (see also chapters 6–7). He also calls on the international community to significantly expand non-reciprocal trade preferences, fund climate adaptation and provide increased political and military commitments for fragile countries.

In chapter 5, Lay and Priebe discuss the "leave no one behind" agenda and policies that would help middle-income countries to make it a reality. LNOB is a comprehensive principle that emphasizes social, economic and political inclusion that, according to the authors, must go beyond an anti-discrimination and/or an anti-poverty agenda. For policies, this implies a focus on key areas for economic and political participation—hence, education, social protection, and labor market policies. LNOB is an important guiding principle that can help guide national development policy in middle-income countries to eradicate poverty, as very unequal progress may threaten the development gains for poor countries graduating from low- to middle-income status. The authors first note that there are huge data

limitations to identify who has been left behind in particular countries, and in particular related to the poverty status of migrants and people with disabilities. The authors go on to review policies that have worked in achieving LNOB goals in middle-income countries. In the sphere of education conditional cash transfers, policies to improve teacher quality (see also chapter 8), early childcare (see also chapter 9), and affirmative action programs are critical to provide knowledge and skills to all. For social protection policies (see also chapter 7), they conclude that a universal approach to social protection is effective and possible, with special emphasis given to increase coverage for marginalized people working in the informal sector. For labor policies (see also chapter 6), which are key for the poor to benefit from economic growth, Lay and Priebe indicate that large public work programs, minimum wage policies and affirmative action policies for various marginalized groups have been effective in generating decent employment. However, conclusions for labor market policies are less clear than for education and social protection, given that labor markets greatly depend on the macroeconomic developments of countries.

In chapter 6, Ruwanpura and Todd assess SDG 1 from the perspective of working women in Sri Lanka. Extreme poverty rates in Sri Lanka were 14 percent in 1985 and declined to below 1 percent in 2016. The authors outline that, despite this overall progress, women workers and minorities have made uneven progress. Even though women in Sri Lanka are highly educated and have had opportunities in the textile sector, they are concentrated in labor-intensive low-paying jobs and do not have any job security. Moreover, working Tamil minority women in Sri Lanka have to encounter discriminatory practices such as lower wages, derogatory language and abuse. Analyzing the labor law reforms proposed in 2019, the authors contend that though the reforms have several measures that are progressive, but overall they are catered towards attracting foreign investors rather than protecting the interests of the labor force. In particular, the reforms give employers the power to self-monitor and take away oversight from the state and unions. This will impact informal and vulnerable workers the most. Achieving sustainable poverty reduction as laid out in SDG 1, gender inequality and workers' rights need to be recognized and given their due importance—otherwise, women will be left behind. Additionally, the authors argue that the poverty reduction objectives of SDG 1 are intimately linked to other SDGs, and policies need to be more aware of these connections.

In chapter 7, Asiedu and Baku explore the goal of SDG 1 from the perspective of social protection policies in Ghana, a fast-developing country on the African continent. They argue that the rate of poverty reduction for one percentage point of economic growth has declined substantially over time in Ghana, and if the country is to achieve

SDG 1, it cannot rely merely on economic growth, but has to expand its social safety net. Ghana currently spends less than one percent of its GDP on social security schemes (0.63% in 2019), which is substantially lower than even most sub-Saharan African countries (2.16% of GDP). The authors argue that Ghana needs to spend more. The major current social protection schemes include the social grant scheme for vulnerable households, a range of education-focused programs providing free meals and schooling, and labor market interventions like a public works program. The authors also argue for the need for better targeting of these programs to improve the efficiency of public expenditures. The authors recommend the increased use of digital tools, like data obtained from the Taxpayer Identification Number (TIN) to better target benefits of various schemes to the people in need. A positive development in Ghana has been the decrease in reliance on international donors for social security expenditures, with more spending covered through domestic resources, making the social protection programs more sustainable.

In addition to social protection schemes, education is key to achieving SDG 1, as it contributes to economic development to finance any programs to alleviate poverty and, more importantly, provides a more equitable access to any gains from economic development.

In chapter 8, van der Berg discusses education challenges for seven southern African countries, namely South Africa, Mozambique, Lesotho, Zimbabwe, Eswatini and Botswana. The author argues that the MDGs led to a large expansion in enrollment and access to education in these countries and that the SDGs have rightly shifted focus to learning outcomes and equity, but this is not reflected yet in policies of southern African countries. "Learning poverty" is, hence, still widespread with little emphasis on measuring or determining whether children going to school are actually learning. The little that is known about learning levels among these countries indicates that it is very low – the average student in these seven southern African countries is between 3 and 6 years behind students in an average country where Pisa test is conducted. These countries do not regularly participate in international testing and many governments are not interested in measuring and monitoring progress in learning. A shift in focus from expanding and measuring access towards measuring and improving learning outcomes is essential to meet SDG 1.

In chapter 9, and building on van der Berg, Fink reviews the literature on early childhood education. Increasingly, it is being recognized that the early years of a child's life are critical for various life outcomes, including the likelihood of suffering under poverty. Fink first documents the large gaps in the average development of children in low- and middle- income countries as compared to high-income

countries. However, if he compares similarly endowed home environments across the world—with basic needs meet and mothers with completed secondary education—then these gaps disappear. Most of the lower development in early childhood in low- and middle-income countries are, hence, not driven by geographic or climate-related reasons, but linked to poverty and a lack of access to basic public services. A reduction in poverty will help narrow these gaps and a reduction in these gaps will lead to less poverty in the future. Early life interventions have been shown to have high returns, but have not been widely adopted in many countries around the world. Fink recommends home visit programs by trained community agents who provide regular guidance to parents on how to provide stimulating environments for children, as a way to reduce the early childhood gaps. These programs have proven to be effective in several countries and could be adopted at scale in many countries.

#### **Part 3: Resources to End Extreme Poverty**

In addition to targeted policies that differ across countries to achieving SDG 1 (see chapters 5–6), essential policies to increase social protection (see chapters 6–7) or providing better access to education to children across all ages (see chapters 8–9) require substantive resources. There are several alternatives to increasing resources—domestic taxation, revenue from natural resources, international development assistance, corporate taxation including of multi-national corporations, and borrowing from domestic and international markets. The last three chapters in this volume (chapters 10–12) discuss some of the challenges and steps forward in mobilizing resources for achieving SDG 1.

Beegle and de la Fuente, in Chapter 10, delve into the issue of how to mobilize resources and use them more efficiently for the poor across African countries. They first present evidence to show that many African countries still face a huge gap in the resources required to end extreme poverty. The authors argue that low levels of GDP on which to tax and limited ability to borrow from international markets restricts the resources that countries have to effectively address poverty. To increase resources, Beegle and de la Fuente recommend increasing direct tax compliance (such as income taxes) and decreasing reliance on indirect taxes (such as the value added tax), which are often regressive in nature, taxing the rich through property taxes, reducing lost corporate tax revenues due to transfer pricing policies (see also chapter 12), and increasing government revenues from extractive industries. Beegle and de la Fuente further argue that in addition to limited resources, the existing resources are often mistargeted and inefficiently used. In particular, the reallocation of energy

and fertilizer subsidies to other sectors, and improving pro-poor spending within the sectors (towards goods and services used more by the poor, like elementary and secondary education and primary health services), would be two important first steps.

In chapter 11, Thiele analyzes another resource to end global poverty—international aid—and reviews the comprehensive literature on the impact of international aid on poverty and economic growth—including trade and foreign direct investment (FDI) as important drivers of growth. Thiele concludes that the impact of aid on growth is unclear, and that the literature lacks consensus on whether there is an impact or not. International aid leads to an increase in FDIs and exports of middle-income recipient countries, but this impact is missing among low-income countries. Thiele points to an important caveat that, due to inherent difficulties in identifying the causal impact of international aid on macro level development outcomes, the long-standing debate about whether aid can help countries to develop remains unsolved. However, the literature also suggests that foreign aid can support people directly to escape poverty. International aid targeted at improving social infrastructure leads to a substantial impact on reducing non-monetary poverty, such as higher school enrollment or lower child mortality. The open question is how much international aid can contribute to achieving the more ambitious social goals of the SDGs, which require much more context-specific and complex interventions than building social infrastructure.

In chapter 12, Brugger discusses the tensions among high-income countries between supporting low-and middle-income countries to raise sufficient revenues for social protection of the global poor through tax policies and pressure to uphold favorable tax frameworks for global corporations with headquarters in high-income countries. The chapter delves into the international political economy of international corporate taxation. To sustainably fund social security programs, countries need to mobilize resources locally through taxation and rely less on international aid (see also chapter 7). Industrialized countries have pledged to support low- and middle-income countries in developing their tax administration capacity. However, multinational corporations headquartered in high-income countries lobby for favorable tax frameworks, with their governments putting many countries in conflict between the two objectives. The author cites data indicating that the potential for raising domestic revenues in low-income countries by preventing the abusive tax avoidance by multinational corporations far exceeds tax avoidance by local companies. Abusive transfer pricing by multinational corporations allows profits to be transferred from high-taxation countries to lower-taxation countries, and in

turn, avoiding taxation. Even though countries are internationally required to prize transactions within their organization as they would prize other unrelated companies (the arm's length principle), the technical capacity of tax administrations to monitor multinational companies on this arm's length principle is missing in many countries. Brugger documents various efforts to simplify the transfer pricing mechanism and shows that they have faced strong resistance and ultimately failed up to now. Even proposals to set up a more inclusive International Tax Organization (ITO) that has more representation from low- and middle-income countries have stalled. The author argues that if high-income countries are to take their pledge of ending global extreme poverty seriously, they must start thinking about their policies, as much as about policies in low- and middle-income countries.

#### **References**


© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Part 1: Is the World on Track to Achieve SDG 1?**

## **The Fight against Global Poverty: 200 Years of Progress and Still a Very Long Way to Go**

**Max Roser and Joe Hasell**

#### **1. Introduction**

Global poverty is one of the most pressing problems that the world faces today. The poorest in the world are often hungry, without access to basic services such as electricity and safe drinking water, have less access to education, and suffer from much poorer health.

Reflecting its importance, the eradication of extreme poverty by 2030 is the very first of the 169 targets set out in the Sustainable Development Goals (SDGs) by the United Nations. The international poverty line on which this target is based is set to the threshold of living on less than 1.90 international dollars per day. That is a very low threshold, in line with poverty definitions adopted in the world's poorest countries.

Recent projections suggest we are not on track to achieve this goal. Even before the onset of the coronavirus pandemic and the resulting global recession, prevailing rates of economic growth and levels of inequality suggested that around 500 million people—roughly 6% of the world's population—would remain in extreme poverty in 2030 (World Bank 2018, 2020).

Is such extreme poverty inevitable?

The history of global poverty shows us clearly that this is not the case. The aim of this chapter is to summarise what we know about that history, to help inform our aspirations for the future.

The chapter considers two approaches adopted by researchers to estimate the extent of global poverty over time. The first of these are estimates from the World Bank based on household survey data, which cover the period from 1981 onwards.

However, in order to see where we have come from, we must look much further back in time: 30 or even 50 years are not enough. When you only consider how the world has looked during this recent past it is easy to make the mistake of thinking of the world as static—the rich and healthy parts of the world here and the poor and sick regions there—and to falsely conclude that it always was, and will always be, as such. Indeed, this is what polling data suggests that the majority of the public believe to be true.<sup>1</sup>

With a longer perspective, it becomes very clear that the world is not static at all. The countries that are rich today were very poor until just a few generations ago and were in fact worse off than many poor countries today.

To avoid portraying the world in a static way we have to start at least 200 years ago, before the time when living conditions really changed dramatically. To do this, we rely on historical estimates based on data recorded in national accounts and earlier reconstructions of such data made by economic historians.

This evidence shows a substantial decline in poverty rates over the last two centuries, with particularly fast progress made in recent decades. The changes we see over the long run should embolden us to reach not only for the eradication of the most extreme forms of poverty but for much more ambitious goals still.

The chapter is structured as follows: Section 2 first outlines the data and methods used by the World Bank to estimate the evolution of global poverty over time. Considering the range of official poverty lines adopted by richer and poorer countries, it then discusses World Bank estimates for global poverty measured according to multiple poverty lines spanning that range. Section 3 discusses the available historical data on incomes that are needed in order to estimate global poverty trends over the last two hundred years: reconstructions of GDP per capita and data on the extent of inequality. Section 4 presents our long-run global poverty estimates, comparing the trends to those found in the World Bank estimates for recent decades. Section 5 outlines other data and research on the living conditions of people in past centuries as a means of sense-checking the long-run trends in monetary poverty presented in the chapter and setting them in the broader context of human welfare.

<sup>1</sup> A 2016 survey conducted by Glocalities, in partnership with Oxfam, the Bill and Melinda Gates Foundation, and Global Citizen found 87% of people from 24 countries surveyed believe that extreme poverty has either increased or stayed the same over the last 20 years. Overall, 67% of respondents believed that ending global poverty by 2030 was unlikely (Lampert and Papadongonas 2016). Similar results concerning public awareness of extreme poverty trends were found in an 2017 Ipsos MORI poll ("Ipsos MORI" 2017).

#### **2. World Bank Survey-Based Estimates**

To track progress towards the target of eradicating extreme poverty by 2030, the UN relies on World Bank estimates of the share of the world population falling below the international poverty line of \$1.90 per day, shown in Figure 1 below.

**Figure 1.** Number and share of people living in extreme poverty globally, 1981–2017. Source: PovcalNet (World Bank). Note: Extreme poverty is defined as living with per capita household consumption or income below 1.90 international dollars per day (in 2011 PPP prices). International dollars are adjusted for inflation and for price differences across countries. The sharp rise in 1989 reflects a change in survey methodology in China.

The reference to the 'international poverty line' (IPL) here, however, signals not just a particular dollar threshold but also the set of methods adopted by the World Bank in drawing that line and estimating the share of people above or below it.

#### *2.1. The World Bank's Approach to Measuring Global Poverty*

This method was first presented by the World Bank in its *World Development Report 1990: Poverty* (World Bank 1990) which provided estimates of extreme poverty based on a \$1 a day poverty line, expressed in 1985 prices.<sup>2</sup> This line was chosen

<sup>2</sup> As discussed below, the reference to 1985 prices here indicates adjustments to account for both inflation and price differences across countries as observed in 1985. The '\$1 a day' line was initially set at \$1.02 a day based on a sample of national poverty lines, adjusted for price differences across countries, collected by Ravallion et al. (1991). This was revised to \$1.08 upon applying 1993 prices to the same set of poverty lines (Chen and Ravallion 2001, 2007).

so as to measure global poverty by the standards of the world's poorest countries, being representative of the national poverty lines observed in such countries at the time. Following broader adoption in the international development community, this measure became the basis for the first of the eight Millennium Development Goals (MDGs). The goal to halve the rate of extreme poverty between 1990 and 2015 was one of the MDGs that were achieved.<sup>3</sup> Based on an expanded and updated set of national poverty lines, this was revised to \$1.25 at 2005 prices (Ravallion et al. 2009) in line with the official lines observed in the poorest 15 countries.<sup>4</sup> More recently, the IPL was updated to \$1.90 in 2011 prices, a figure obtained by adjusting the same set of 15 national lines for inflation (Ferreira et al. 2016).

The Bank's estimates of the share of the population falling below the IPL are based on national surveys that provide data on households' consumption or income.<sup>5</sup>

These survey data are adjusted to account for price differences across countries and for inflation over time. The resulting figures, as well as the poverty line itself, are expressed in 'international dollars' at a given year's prices. The World Bank's current estimates are based on 2011 prices, such that one international dollar has the same purchasing power as 1 US dollar had in the United States in 2011. In the interest of readability, we do not repeat the full unit of measurement as international dollars in what follows and simply use \$ as an abbreviation throughout.

Since surveys are not conducted every year in every country, in order to estimate the global share of people below the IPL for a given reference year, researchers must rely on the closest available survey data for each country. Data from surveys not conducted in the reference year are 'lined up' using growth rates recorded in the national accounts (Prydz et al. 2019).

<sup>3</sup> Achievement of the MDGs was measured by targets and, out of fourteen targets permitting quantitative assessment, this was one of only five that were achieved. For a collection of data on the achieved and missed MDG targets see https://ourworldindata.org/millennium-development-goals (accessed 25 January 2021).

<sup>4</sup> As ranked by consumption per capita—namely, Malawi, Mali, Ethiopia, Sierra Leone, Niger, Uganda, Gambia, Rwanda, Guinea-Bissau, Tanzania, Tajikistan, Mozambique, Chad, Nepal, and Ghana (Ravallion et al. 2009).

<sup>5</sup> This is largely determined by which type of survey is available in each country. Consumption surveys are used for a majority of countries, and this is particularly true of poorer countries in which most of the world's poor live. But the World Bank's estimates for many countries, most notably many Latin American and Caribbean countries, are based on income surveys. This inconsistency affects the comparability of estimates across countries and over time, discussed later on in the chapter.

#### 2.1.1. Criticism and Alternative Approaches

It is important to remember that there is no concept of poverty that can claim universal agreement. In this regard, measuring the extent of poverty is not like measuring a person's height or weight. Competing normative principles and limitations in the quality of available data leaves room for disagreement as to how poverty should be measured, and this is especially the case at the global level.

One question concerns the level at which the poverty line is to be set. A number of authors view the international poverty line of \$1.90 a day to be too low (Pritchett 2006) or too high (Ravallion 2016b) to capture morally relevant aspects of the global income distribution, or, in any case, to be lacking clear justification (Reddy and Pogge 2009). The inability of a single line to reflect both the depth and breadth of poverty experienced around the world (discussed further in Section 2.3 below) demonstrates the importance of tracking multiple poverty lines, as will be carried out throughout this chapter.

A more general area of contention relates to the aforementioned price adjustments needed to apply *any* fixed poverty line that is constant in terms of purchasing power across countries. The difficulty involved in such an adjustment is the source of significant uncertainty concerning the level of extreme poverty globally and its geographic distribution (Deaton 2010). This is evidenced, for instance, by the substantial revisions to the World Bank's estimates of global poverty that have followed the periodic updates of the price data on which these adjustments are based (Deaton 2010; Dykstra et al. 2014).

Moreover, adjusting for price differences across countries faces the inherent challenge of comparing a diverse set of goods and services, the consumption of which is often specific to particular regions or income levels. Some authors have questioned the validity of the World Bank's approach in adjusting for the price of a basket of goods and services that includes many items only consumed by the non-poor or in rich countries (Reddy and Pogge 2009; Allen 2017). Asali, Reddy and Visaria (2008) and Allen (2017) advocate an alternative approach in which incomes are compared against the local minimum cost of meeting the basic needs of food and shelter, all measured in local currencies and thereby avoiding the need for cross-country price indices altogether.

Within this debate, however, it is important not to take an exaggerated impression of the uncertainty that international price comparisons imply for poverty measurement. Updates to the international price data, although leading to significant revisions of the estimated levels of global poverty, have left our understanding of the key trends in extreme poverty broadly unchanged (Chen and Ravallion 2010;

Deaton 2010). Using household survey data, Deaton and Dupriez (2011) compare purchasing power parity rates (PPPs)—the standard price indices used to compare incomes internationally, including within the World Bank's poverty estimates—with 'poverty-weighted' PPPs that reflect the consumption patterns of households living at or near the poverty line. They find that relative price levels between countries are broadly similar across the two sets of PPPs, implying a limited impact on poverty measures. Furthermore, Moatsos (2021), applying Allen's 'cost of basic needs' approach, finds long-run declines in global poverty broadly similar to those found using the World Bank's methodology (see Section 5.1). In summary, the available evidence concerning the influence of cross-country price adjustments on poverty measures does not undermine the key trends present below—it *reinforces* them.

#### *2.2. Extreme Poverty Since 1981*

Figure 1 shows the global estimates for the number and share of people living below the international poverty line from the World Bank. The estimates begin in 1981, prior to which survey coverage is judged to be too low (Chen and Ravallion 2009). In that year, 42% of the world's population is estimated to have been living on less than \$1.90 per day, roughly 1.9 billion people. The figures show a substantial reduction in extreme poverty in the decades following. By 2017, the latest available year, the World Bank estimates that the share had fallen to 9%—less than one-quarter of its 1981 level. This translates to more than 1 billion fewer people living in extreme poverty, over a period in which the world's population grew by around 3 billion.

Recent decades show us that rapid, substantial reductions in poverty are possible. However, a number of factors point to a future in which progress against extreme poverty is slower.<sup>6</sup>

Firstly, the recession caused by the coronavirus pandemic has likely increased the number of people in extreme poverty. 'Nowcasting' estimates produced by the World Bank suggest that there were more than 100 million more people in extreme poverty in 2020 relative to its expectations for what would have occurred in the absence of the pandemic—'the worst reversal on the path towards the goal of global poverty reduction in at least the last three decades'.<sup>7</sup>

<sup>6</sup> For a discussion of future possibilities for moving beyond very low levels of extreme poverty towards its eradication, see Martin Ravallion's contribution to the present volume, 'SDG1: The Last 3%'.

<sup>7</sup> The World Bank *Poverty and Shared Prosperity 2020* report provides two COVID-19 scenarios that yield 88 million and 115 million people in extreme poverty above the baseline scenario. An update published in January 2021 presents even higher projections of between 119 and 124 million additional

Yet, even before the pandemic, there was evidence of a slowdown in the rate of extreme poverty reduction: we see the lines in Figure 1 flattening from the early 2010s. Projections made by the World Bank and other development research organisations concur that, even if pre-pandemic rates of economic growth and levels of inequality had continued, future progress against extreme poverty would have fallen short of the goal of eradication by 2030.<sup>8</sup>

The reasons for this can be better understood by looking at the regional trends in extreme poverty.

Figure 2 shows that the distribution of the extremely poor across world regions has changed significantly in recent decades. In 1990, more than a billion of the extremely poor lived in China and India alone (Figure 2, left panel). In the decades that followed, those economies grew faster than many of the richest countries in the world, bringing down extreme poverty rates in their regions and across the world as a whole (Figure 2, right panel). As a consequence, the concentration of the world's poorest shifted from East Asia in the 1990s to South Asia in the 2000s and then to sub-Saharan Africa in the 2010s. Sub-Saharan Africa has seen less growth in incomes and poverty rates have, therefore, fallen far slower. The slow decline of the *share* in extreme poverty was offset by population growth, resulting in a slow increase in the number of extremely poor people in sub-Saharan Africa.

Global poverty declined during the last generation because the majority of the poorest people on the planet lived in countries with strong economic growth. This is now different. The majority of the world's poorest today live in economies that have seen little growth in recent decades. A return to the growth trajectories of the time before the pandemic will not be enough to end global extreme poverty—the lack of growth in the economies that are home to the world's poorest populations would imply a future in which hundreds of millions face the prospect of remaining stuck in extreme poverty.

people in extreme poverty (https://blogs.worldbank.org/opendata/updated-estimates-impact-covid-19-global-poverty-looking-back-2020-and-outlook-2021, accessed 25 January 2021). The quote given is taken from an earlier update (https://blogs.worldbank.org/opendata/updated-estimates-impactcovid-19-global-poverty-effect-new-data, accessed 25 January 2021).

<sup>8</sup> See the World Bank projections in its *Poverty and Shared Prosperity* report series (World Bank 2018, 2020). Prior to the coronavirus pandemic, these projections pointed to around 6% of the world's population—roughly 500 million people—living below the international poverty line in 2030. This is similar to the projections made by the Overseas Development Institute (ODI) and the World Poverty Lab jointly with the Brookings Institute, documented by ODI at their blog (https://www.odi.org/blogs/ 10688-new-projections-show-extreme-poverty-falling-not-fast-enough, accessed 25 January 2021).

**Number of people living in extreme poverty**

**Figure 2.** Number and share of people living in extreme poverty by world region. Source: PovcalNet (World Bank). Note: Extreme poverty is defined as living with per capita household consumption or income below 1.90 international dollars per day (in 2011 PPP prices). International dollars are adjusted for inflation and for price differences across countries. The rise in East Asia in 1989 reflects a change in survey methodology in China.

#### *2.3. Poverty at Higher and Lower Thresholds*

#### 2.3.1. Poverty from the Perspective of Richer and Poorer Countries

Whilst the international poverty line has been adopted widely by international organisations, it is important to remember that different individual countries adopt different definitions when assessing the extent of poverty amongst their own citizens.

Comparing across countries, we see that richer countries tend to set substantially higher poverty lines. Figure 3 plots a dataset of national poverty lines collated by Jolliffe and Prydz (2016) against GDP per capita. This chart makes it clear how *extremely low* the international poverty line of \$1.90 is. It denotes a standard of living that falls far beneath the level at which people would be considered poor in rich countries.

**Figure 3.** National poverty lines vs. GDP per capita. Source: Jolliffe and Prydz (2016), World Bank. Note: Both metrics are adjusted for price differences between countries and are measured in international-\$ at 2011 PPP prices. The three horizontal lines mark the three poverty lines adopted by the World Bank (World Bank 2018).

The national poverty lines shown in this figure are set according to both absolute and relative definitions of poverty. Most low- and middle-income countries measure poverty according to an absolute poverty line whose value remains fixed over time.

Most high-income countries use a relative poverty line whose value rises (or falls) in line with the general standard of living in that country. Typically, relative national poverty lines are set at 40%, 50%, or 60% of the median income.

Such a relative concept aims to identify individuals or households whose income is so low relative to the average in their society that they are 'excluded from ordinary living patterns, customs and activities' (Townsend 1979, p. 31). The principle behind setting a poverty line relative to the average income is the idea that as incomes in a society rise, so too does the level of material resources needed in order to participate in 'ordinary' life in that society. On such a definition, poverty can only fall where inequality in the lower half of the distribution is reduced.

Whilst the principles behind these two ways of measuring poverty are very different, Figure 3 shows that the distinction is less stark in practice. The upward-sloping relationship can be observed across all countries, not only the high-income countries that have adopted a relative definition of poverty. This means that absolute poverty lines also tend to be set at a level that is reflective of the standard of living typical for that society, albeit in a less mechanical way. Whilst absolute poverty lines are not pegged to average incomes, they are subject to periodic revision. As countries become richer, they tend to raise the official poverty line. India, China, and Nepal, for instance, have all raised their poverty lines in the last decade as their average incomes have risen (World Bank 2018, p. 74; Chen and Ravallion 2013).

The diversity of definitions of poverty we see across richer and poorer countries raises an important question: Which of these perspectives should be relied on in order to quantify the extent of poverty *globally*? This has been a central concern of poverty researchers since the very first global estimates (Ahluwalia et al. 1979; Ravallion et al. 1991).

One common response has been the use of multiple poverty lines. As is seen in Figure 3, the IPL is set at the level of the poverty lines typical amongst the very poorest countries in the world. The World Bank has adopted two further poverty lines of \$3.20 and \$5.50 in order to monitor global poverty from a perspective more in-line with the definitions adopted in lower- and upper-middle-income countries (World Bank 2018).<sup>9</sup> The extent of recent progress against poverty as measured relative to these higher lines will be assessed in the next section.

<sup>9</sup> These thresholds were the median values found by Jolliffe and Prydz (2016) within their dataset when looking at the national poverty lines adopted in lower-middle- and upper-middle-income countries (defined according to the World Bank's income classification based on the level of GDP per capita).

A second, more recent, response to this question has been the development of a new approach to global poverty measurement in which this relative dimension is brought to the fore. Building on the work of Jolliffe and Prydz (2021), Chen and Ravallion (2013), Atkinson and Bourguignon (2001) and others, the World Bank has adopted an additional 'societal poverty line' (SPL) that combines absolute and relative approaches (World Bank 2018). For the world's poorest countries, the SPL is set at the international poverty line of \$1.90. Above a certain threshold, however, the value of the line begins to rise in proportion to the median level of consumption in each individual country. For every additional \$1 the median level of consumption per day rises, the SPL rises by 50 cents—similar to the way in which national relative poverty lines increase with median income across most high-income countries.

The SPL can be thought of as combining two goals within a single measure of global poverty: firstly, that a minimum absolute level of subsistence is ensured for all, and secondly, that people achieve an acceptable standard of living judged according to the norms of the country in which they live. To be judged non-poor according to this 'societal' measure, a household must fall into neither kind of poverty (Ravallion and Chen 2019; Atkinson and Bourguignon 2001).

According to World Bank estimates for 2017, on top of the roughly 690 million people below the absolute poverty threshold of the IPL, there were a further 1.4 billion people living in relative poverty, bringing the total number of poor under this combined definition to just over 2 billion (World Bank 2020, p. 65). Over time, the global societal poverty rate has fallen but much less rapidly than the extreme poverty rate—from 45% in 1990 to 28% in 2017. Many of the people that managed to leave extreme absolute poverty over this time remained poor measured by a poverty line typical of the income level of their country.

This relative component of the SPL means that two people with the same absolute level of income (above the IPL) may be judged poor in one country and not another, depending on which country they live in. This is brought into particularly sharp relief in aggregating over countries with such widely varying levels of income. The idea that a person living in Liberia that sees their income rise above \$2 and a person living in Norway that sees their income rise above \$32 are both to be considered as having been lifted out of the same concept of poverty may strike some as counter-intuitive or even unethical.<sup>10</sup> Whilst the observation that richer countries

<sup>10</sup> The SPL is calculated as the maximum of either the international poverty line of \$1.90, or else \$1.00 + 0.5 × median consumption. According to survey data provided in Povcalnet (http://iresearch.worldbank. org/PovcalNet/povOnDemand.aspx, accessed 25 January 2021) the median monthly consumption in

tend to adopt higher poverty lines nationally highlights the inherently social nature of poverty, an understanding of the share of people living below higher and lower absolute thresholds remains indispensable.

#### 2.3.2. Global Poverty at Higher and Lower Poverty Lines

The same approach used to monitor the share of people falling below the international poverty line can be relied upon to assess the extent of poverty relative to other fixed poverty lines.

Figure 4 shows the share and the total number of people around the world living below different absolute thresholds. The \$1.90, \$3.20, and \$5.50 lines shown are those adopted by the World Bank to reflect the poverty lines typical of low-, lower-middle-, and upper-middle-income countries, respectively. We have added two higher lines of \$10 and \$30 which broadly cover the range of poverty lines adopted by rich countries, as indicated in Figure 3 above.<sup>11</sup>

We see that globally the share of people below any of the poverty lines was declining up to the latest data in 2017. However, the timing and pace of the decline were very different across the different thresholds.

Whilst the share falling below the international poverty line of \$1.90 decreased fairly steadily at a rate of around 1 percentage point per year since 1981, there was no progress against a poverty line of \$10 per day until around 2000, with roughly only one-quarter of the world population living on more than \$10 a day between 1981 and 2000. The world started to make progress against poverty relative to higher cutoffs only recently, but progress has been fast since then: by 2017, the share living on more than \$10 had increased to more than one-third.

Measured against a \$30 a day line—roughly the level of the poverty lines set in the world's very richest countries—the vast majority of the world population is living in poverty, and the share above the poverty line has increased only slowly over this period.

Liberia in 2016 was \$62.83, or \$2.07 as a daily figure. The SPL is thus 1 + 0.5 × 2.07 = \$2.03. The same source reports Norway's median monthly income in 2017 to have been \$1890, or \$62 per day. The SPL, in this case, is 1 + 0.5 \* 62 = \$ 32.

<sup>11</sup> The figures for each poverty line are the global aggregates as reported by the World Bank's Povcalnet API on 25 January 2021 (see http://iresearch.worldbank.org/PovcalNet/getstarted.aspx).

#### **Share of world population Number of people**

**Figure 4.** The share and number of people globally living below different poverty thresholds. Source: PovcalNet (World Bank). Note: Poverty at each threshold is defined as living with per capita household consumption or income below the indicated level, measured in international-\$ at 2011 PPP prices. International dollars are adjusted for inflation and for price differences across countries.

However, in making use of higher poverty lines, we should not lose sight of what is happening to the very poorest people in the world. Figure 4 also shows the extent of global poverty as measured against a \$1 a day line—far beneath the IPL. In the years running up to 2017, the number of people living below this ultra-low poverty line had stopped falling altogether—with around 170 million stuck in the very deepest poverty.<sup>12</sup> As development economists have emphasised for some time, the very poorest people in the world have seen next to no improvement in their material living conditions in recent decades (Ravallion 2016b) (Lakner and Milanovic

<sup>12</sup> There are some caveats concerning the estimation of very low poverty lines using the World Bank's methodology. For some countries, including China, poverty estimates are derived from fitting an assumed functional form to grouped data, rather than from 'micro-data' concerning individual households. These estimates become less precise in the tails of the distribution. Secondly, since within income surveys a certain proportion of households typically report having zero incomes, this can make comparisons across countries using income and consumption surveys less meaningful when considering very low poverty lines. However, neither issue appears to be of much concern here. Overall, 150 million of the 174 million people estimated to be living on less than \$1 a day in 2017 in the World Bank data were in sub-Saharan African countries. These countries make use of consumption surveys and, as a regional aggregate, saw a \$1 a day poverty rate of 14% (i.e., the line does not fall into the tail of the distribution).

2016). This fact is surely one of the biggest development failures of our time, and yet it is not as widely known as it should be. A big part of the reason why this issue does not receive the attention it deserves is that the international poverty line of \$1.90 is too high for this fact to be seen.

Poverty metrics have several purposes. One is to express a social standard concerning the level of income needed to lead a decent life. Yet, another is to specify a target for progress, such as the Sustainable Development Goal to end extreme poverty by 2030. In seeking to understand the evolution of living standards across the world, however, it is clear that we need to consider multiple poverty lines that make visible important differences in the trends concerning the poor, the extremely poor, and the very poorest.

#### **3. Evidence on the Incomes of the Past**

It is only from the 1980s that the coverage of household surveys is considered to be sufficient for reliable global poverty estimates based on survey data.

Can we know anything about global poverty in earlier decades, or even the distant past?

Thanks to the work of historians, we can. In Section 4, we present estimates of how the extent of poverty globally has changed over the last two hundred years. The estimates are based on a 'national accounts' approach in which data on average incomes available in the national accounts—GDP per capita—are combined with data on the extent of inequality in each country. It is a method that has been used to investigate global poverty trends both for recent decades and for the distant past.<sup>13</sup>

For recent decades, the necessary data on average incomes are available from official national accounts data, while the inequality data are based on the kind of household surveys discussed in the previous section. However, in order to apply this approach to earlier periods, we must rely on the work of economic historians who have produced reconstructions of GDP per capita and estimates of inequality for a range of different countries from available historical sources.

There are, unsurprisingly, important limitations concerning the coverage, comparability, and quality of this historical data. Therefore, the historical poverty estimates provided at the end of this section should be treated as offering a broad

<sup>13</sup> For global poverty estimates for recent decades using the national accounts, see Pinkovskiy and Sala-i-Martin 2016. For historical estimates, see (Bourguignon and Morrisson 2002; Ravallion 2016b).

indication of global trends, rather than very precise estimates for any given point in time.

However, here too, it is important not to overexaggerate the uncertainties. In this section, we discuss the evidence on average incomes and inequality on which our long-run poverty estimates rely, and how their limitations might bias the results. The limitations are real. However, they do not undermine our ability to say a good deal about the broad trends in poverty across the world over the last two hundred years.

#### *3.1. Historical Data on GDP per Capita*

We can learn a lot about the living conditions of people in the past by knowing how average incomes have changed. Thanks to the work of economic historians, who have been able to reconstruct historical estimates of GDP per capita, we have a good idea about the evolution of average incomes for many countries in the world.

How do economic historians estimate incomes in the distant past?

In broad terms, the strategy is to extend the system of national income accounting that countries use today to estimate GDP back to earlier periods. In the absence of complete data collected at the time, researchers have to bring together what evidence they can from historical sources. However, the basic principles are the same. Here, we discuss three key principles on which historical national accounts data are based.

One very important principle to bear in mind is the fundamental identity behind all national accounts: 'Within the methodological framework provided by national income accounting, the estimation of GDP can be approached in three different ways, via income, expenditure and output, all of which ought to yield broadly similar results'(Broadberry et al. 2015, p. xxxii).

For historical estimates, the output approach is often considered the more reliable in practice. Depending on the evidence available, however, information on incomes and expenditure are also used, and all three approaches can provide benchmarks to cross-check the plausibility of estimates.

A second point is that these data relate to real incomes: the figures are adjusted for inflation using available data on the prices of goods and services over time. It is straightforward to compare material prosperity over time relative to goods which remained relatively unchanged over the course of history—economic historians can track the affordability of products such as bread, shirt, beer, nails, meat, books, or candles over time.

This, however, is not easily possible when entirely new products were introduced or when the quality of products and services changed substantially. Many of the

most valuable goods today were not available at all in the past: no king or queen had access to antibiotics, they had no vaccines, no comfortable transport in trains or planes, no electronic devices, no computers, and no light at night.

While modern national accounting practices attempt to take the innovation of new products and changing the quality of existing products into account, there is limited scope to address this in historical accounts. It is important to remember that, no matter how high someone's income might have been in the distant past, some of the goods you might value the most—or would value when you fall ill—were not available at all.<sup>14</sup>

A third key principle is that these estimates of GDP do not just concern the amount of money people had in the past or only the value of goods purchased in the market. This is a common misunderstanding of historical research. Over the last two hundred years, there has been a major shift from people farming for their own consumption to people working for a wage and purchasing goods in the market. Historians of course know about this historical change and take it into account in their analysis of how global prosperity changed. In the important case of subsistence farmers, the value of the food they produce represents both the economic output of the activity and the income received by the farmer. Consumption of that produce then represents a form of expenditure, as it is using up part of the farmer's income.

This issue is not just of importance for historical estimates, but it is also of central relevance today, given the importance that food produced at home, or otherwise received in kind, continues to play in the life of the rural poor, especially in low-income countries. Accordingly, these flows are accounted for in national accounts—both in the official data compiled today and in historical reconstructions.

The extensive work carried out by Broadberry et al. (2015) to produce the historical GDP per capita series for England and the UK, shown in Figure 5, serves as a good illustration. It is difficult to convey the level of detail that is considered in such estimates in a short overview such as this one, but a passage on agricultural output provides some insight.

[The output method] has entailed, first, estimating the amounts of land under different agricultural land uses . . . and, then, deriving valid national trends from spatially weighted farm-specific output information on cropped areas and crop yields and livestock numbers and livestock yields... The

<sup>14</sup> To some extent, the opposite problem also exists, and some goods that were available in the past—such as slaves—are not available today. But this is a much rarer problem.

latter task is further complicated by the need to correct for data biases towards particular regions, periods and classes of producers.

(Broadberry et al. 2015, p. xxxv)

**Figure 5.** GDP per capita in England, 1270–2016. Source: Broadberry et al. (2015) via Bank of England (2017). Note: Data refers to England until 1700 and the UK from then onwards. Adjusted for inflation and measured in British pounds in 2013 prices.

Hundreds of datasets on agricultural outputs are involved in producing these estimates of agricultural production, themselves built upon a substantial body of historical research. To this is added estimates of the output of industry and services in order to yield a measure of aggregate GDP.

There are two key takeaways. First, that historical reconstructions of GDP are the outcome of decades of important academic work. Second, these represent estimates of total production, not just that part of production sold on markets.

The Evolution of Average Incomes over the Long-Run

In order to produce the global poverty estimates presented at the end of this section, we have relied on the Maddison Project's database of historical GDP per capita series for different countries (Bolt and van Zanden 2020).

This database brings together the research efforts of a huge range of country specialists, including the work on England just discussed. In different countries, researchers employ different methods, depending on what historical evidence is

available and is most reliable. In addition to being adjusted for inflation over time, the series in this dataset are adjusted to account for price differences across countries.<sup>15</sup>

Figure 6 shows GDP per capita since 1820 for different world regions and for the world as a whole, as constructed from the Maddison database. Globally, average income per person has increased by roughly a factor of ten over this period. It is worth keeping in mind that this change has occurred while the world population increased fivefold. As we will show below, this rise in the average global income generated a substantial fall in the share of the world population living in poverty over the last two hundred years.

The extent of poverty is not determined solely by average incomes—poverty will be more or less prevalent depending on how equally or unequally incomes are distributed in a country. However, average incomes play a hugely important role, and they set boundaries on what is possible for poverty. If the average income in a country is below the poverty line, so too will be the incomes of the vast majority of people, irrespective of the level of inequality.

This basic but important fact can be appreciated by comparing the regional GDP per capita estimates shown in Figure 6 with the regional distribution of extreme poverty found in the World Bank's estimates based on survey data, as shown in Figure 2. Regions with a high level of GDP per capita have few people in extreme poverty. Regions where GDP per capita was growing rapidly experienced a decline in the number of people in extreme poverty.

Just as the increase in the global average income over the last two hundred years is clear from historians' work, so too is the increase in global inequality. Whilst many Western European countries, as well as the US, Australia, and Canada, experienced rapid economic growth throughout the 19th and 20th centuries, incomes in Asia and Africa stagnated. Exploitative colonialism is one of the institutions to blame for these poor development outcomes (Acemoglu et al. 2001). It was only in the second half of the 20th century that many low-income countries began to see growth rates comparable to, and eventually even higher than, those seen in rich countries. Many

<sup>15</sup> Adjusting for price differences across countries is a difficult task even for recent years (for a discussion, see Deaton and Heston 2010). Moreover, over extended periods of time, inconsistencies can arise between evidence concerning the level of inflation in two countries and comparisons of their price level at two points in time, as the composition of the goods and services produced and consumed in the countries evolves. This is all the more challenging for the distant past given the absence of very detailed price data. Nevertheless, different approaches are available to economic historians to gauge and cross-check relative price levels across countries in the distant past. These methods are discussed in detail in the paper accompanying the 2020 release of Maddison Project dataset (Bolt and van Zanden 2020).

countries, particularly, but not only, in Africa, are still being left behind in terms of economic prosperity as Figure 3 shows. It is in these countries that most of the world's extremely poor populations are to be found today.

**Figure 6.** Regional and world GDP per capita, 1820–2018. Source: Maddison Project Database 2020 (Bolt and van Zanden 2020) Note: GDP per capita adjusted for price changes over time (inflation) and price differences between countries. It is measured in international-\$ in 2011 prices. The 'Western Offshoots' region refers to the US, Canada, Australia, and New Zealand.

#### *3.2. Historical Inequality Data*

GDP per capita is the first relevant metric from which we can learn about living standards in the past; the second one is the level of inequality. In a very unequal country, the majority of people are substantially poorer than indicated by the average income, while in a country with low inequality, the average is much more reflective of the incomes typical across the population.

In order to estimate the extent of poverty from data on GDP per capita, we also need data on inequality.

For recent decades, the data on inequality can be obtained from the kind of household surveys discussed in the previous section. However, for earlier periods, historians must rely on a range of historical sources: 'social tables' that document the average incomes of different social classes; census data; top income shares derived from tax records; evidence on wage levels; in some cases, information regarding the extent of inequality in adult heights.<sup>16</sup>

The inequality data used in the global poverty estimates presented below are taken from a historical dataset produced by van Zanden et al. (2014), which combines estimates based on the range of sources just described. For more recent decades, we rely on the Global Consumption and Income Project (GCIP) dataset which provides estimates of inequality based on household survey data.<sup>17</sup>

Figure 7 plots Gini coefficients from these two data sources for a set of benchmark years along with unweighted and population-weighted averages. Whilst there are clear differences across individual years, we observe no overall trend over the last two centuries: the average across all observations in each period varies between 0.35 and 0.5, with the bulk of observations falling between 0.25 and 0.6.

There are, however, important limitations concerning the comparability and quality of these estimates that make such trends highly uncertain.

Whilst each data point refers to an estimate of the level of income inequality, the measure of welfare in the underlying source varies. This includes incomes assessed before tax, incomes after tax, wage income, and consumption expenditure. Both datasets use statistical models to try to standardise the data in certain ways, but this is inevitably partial and imprecise.<sup>18</sup>

<sup>16</sup> For a discussion of early inequality estimates and the sources of data that these can draw on, see van Zanden et al. (2014 Data Appendix) and Milanovic et al. (2011).

<sup>17</sup> The data are made available at http://gcip.info/ (accessed on 9 September 2021). For a study introducing the dataset and the sources and methods behind it, see Lahoti et al. (2016).

<sup>18</sup> Both data sources adjust expenditure survey data using a statistical model to try to estimate what the level of income inequality would have been. In the case of van Zanden et al. (2014), they also adjust data on net incomes in the same way to be more in line with a measure of gross income inequality. In an online data appendix (https://ourworldindata.org/history-of-poverty-data-appendix (accessed on 9 September 2021)), we provide alternative estimates for global poverty that instead make use of GCIP data where the standardisation is carried out in the opposite direction—towards a consumption basis—and also World Bank data that do not attempt to standardise income and consumption surveys in this way. Whilst there are notable differences in the poverty estimates that these different datasets yield, it does not affect the broad long-run trends that are our focus here.

**Figure 7.** Gini coefficient of income, 1820–2014. Source: van Zanden et al. (2014) and Global Consumption and Income Project (GCIP). Note: The GCIP data shown are survey-year observations that fall within two years of the benchmark year.

There are also many sources of potential bias and uncertainty concerning the individual estimates. As will be discussed in more detail below, this is true even of modern survey data. However, it is all the more true for earlier inequality estimates given the limitations of the underlying data. One particular concern of the historical data is that the value of subsistence farmers' production may not

be properly accounted for in historical sources and early household survey data, implying that earlier estimates of inequality could be overstated.<sup>19</sup>

#### How Sensitive Are Poverty Estimates to Different Assumptions about Inequality?

These limitations mean that there is substantial uncertainty surrounding historical poverty estimates for any given country or any given year. However, for the following reasons, our broad understanding of the changes in global poverty over the last two centuries is not much impacted by this uncertainty.

Firstly, it is important to bear in mind that, although differences between richer and poorer individuals within countries are substantial, they are overall much smaller than the differences we see across countries. Milanovic (2015) shows that around two-thirds of the income differences we see across individuals globally can be predicted just by knowing the country in which they live. It is the very large differences in average incomes we see across richer and poorer countries that contribute the most to overall global inequality today.<sup>20</sup>

Analogous considerations apply over time too. As we saw from the historical data on GDP per capita, the extent of global inequality across countries today is the consequence of substantial economic growth having been achieved in some parts of the world but not in others. The kind of income you receive in your life is greatly determined not just by *where* you were born but *when*. These two factors have a much more decisive influence than the relative position you occupy within your society.

This is not to say that reducing inequality cannot play a vital role in reducing poverty. The forward projections of global poverty prepared for the World Bank by Lakner et al. (2020) find that reductions in inequality compare favourably with assumptions about higher growth rates in their ability to reduce future extreme poverty.<sup>21</sup> Whilst recent trends suggest the world is far from being on track to achieve

<sup>19</sup> We thank an anonymous reviewer for drawing our attention to this potential concern. An assessment of this issue could not be made within the scope of this chapter. As we discuss below, even after allowing for a wide error margin on the inequality estimates, the broad long-run trends we focus on are not affected in any substantial way.

<sup>20</sup> In a population-weighted regression of (within-country) income percentiles derived from household survey data, Milanovic (2015) finds that 66% of the variation can be explained by country controls alone (Milanovic 2015, Table 2). Using an inequality measure that allows for a decomposition of global inequality—the extent inequality across all the world's citizens—into a within-country and between-country components, he finds that in 2008, 70% of global inequality related to differences between countries. See also Lakner and Milanovic (2016) and Milanovic (2020) for more recent estimates.

<sup>21</sup> 'A 1% annual decline in each country's Gini index is shown to have a bigger impact on global poverty than if each country experiences 1 pp higher annual growth rates than forecast' (Lakner et al. 2020).

the goal of eliminating extreme poverty by 2030, their projections suggest that this goal at least 'becomes more viable by reducing inequalities'.

However, the scale of economic growth over the last two hundred years has been large enough such that the differences in poverty generated by shifting from a high- to a low-inequality setting are comparatively much smaller. To illustrate this point, Figure 8 plots modelled income distributions based on the GDP per capita of China in 1820 and 2017. It shows the distributions under three different inequality scenarios. The middle panel shows the distributions using the estimates of inequality found in the datasets just described. The top panel shows a low-inequality scenario in which a Gini coefficient is 0.25 is assumed in both years. The bottom panel shows a high-inequality scenario in which a Gini coefficient of 0.65 is assumed in both years.<sup>22</sup>

The shaded areas to the left of the \$2 a day line show the share of the population with incomes falling under this threshold. The 14-fold growth in GDP per capita between 1820 and 2017, from \$882 to \$12,734, implies a major decline in poverty measured against a \$2 a day line, whatever we assume about inequality. In all three scenarios in 1820, around half or more of the population in China fell below this threshold.<sup>23</sup> Additionally, in all three scenarios in 2017, the vast majority fell above this threshold.

<sup>22</sup> Here, we assume that incomes follow a lognormal distribution. This is a common assumption made by researchers modelling income distributions. This distribution offers a good approximation of the bulk of the distribution observable in survey data, though it can be less accurate in the tails. See (Cowell 2011, pt. 4.4, for a discussion).

<sup>23</sup> The high-inequality scenario for 1820 is not a plausible one at such a low level of average income: it results in a distribution in which a substantial share of the population falls below a credible level of subsistence. On this subject, see the discussion of Milanovic et al. (2011) on the 'inequality possibility frontier'.

**Daily income**

#### **Annual income**

**Figure 8.** Modelled distribution of income in China in 1820 and 2017 under different inequality scenarios. Source: Estimates of GDP per capita are taken from Maddison Project Database 2020 (Bolt and van Zanden 2020). Estimates of inequality are taken from van Zanden et al. (2014) (1820 value) and Global Consumption and Income Project (GCIP) (2017 value). Note: The incomes shown are adjusted for price changes over time (inflation) and price differences between countries. They are given in international-\$ in 2011 prices. The Gini coefficient used for the 2017 distributions relates to a 2014 consumption survey adjusted by GCIP using a statistical model to bring the estimate more in line with an income welfare concept.

The fact that the 1820 and 2017 distributions overlap so little, even in a very high inequality scenario, shows us that these points are not specific to a particular poverty line. Any poverty line under which a substantial proportion of the 2017 population lived is a poverty line that almost the entire population in 1820 must have lived under. That is true whatever we assume about inequality for either period.

Additionally, since between these two periods China experienced income growth broadly in line with the global average, what is true for China is also true for the world. In an online data appendix, we provide hypothetical estimates of global poverty applying such high- and low-inequality scenarios to all countries.<sup>24</sup> Even allowing for such large margins of error does not substantially affect our understanding of the evolution of global poverty over the last two hundred years.

#### *3.3. Incomplete Coverage in Historical Data*

One additional difficulty in arriving at global poverty estimates is the incomplete coverage of the available historical data. For a number of countries, historical estimates for GDP per capita or inequality are either missing for particular years or else are lacking altogether.

In terms of country observations, the early inequality data are particularly sparse. For the early 19th century, the dataset of van Zanden et al. (2014) includes observations for only around 40 countries. However, this includes many of the most populous countries, such that estimates covering around three-quarters of the world's population are available. To produce our estimates of global poverty, countries with missing data for a particular year are attributed the average Gini observed in the region or, in the case of the successor states of the USSR and Yugoslavia, the average within the bloc. While this will not always be an accurate assumption, it will not have a substantial influence on the resulting global trends: this method is applied to a relatively small share of the world's population and, as we previously discussed, in most cases, the overall trends are robust to widely different assumptions concerning inequality.

The available GDP per capita data are more complete but again with notable gaps. To produce our global poverty estimates, we have interpolated between observations, assuming a constant growth rate, and in a number of cases have

<sup>24</sup> https://ourworldindata.org/history-of-poverty-data-appendix (accessed on 9 September 2021).

extrapolated backwards by applying average growth rates observed within the region (or again, the former bloc).<sup>25</sup>

In the case of sub-Saharan Africa, evidence concerning the level of incomes in the distant past is particularly poor. Since coverage prior to 1950 is especially limited within the Maddison database, our poverty estimates instead make use of the economic growth rates for African countries produced by Prados de la Escosura (2012). These estimates are based on inferring total output per head from available records on international trade, and Prados de la Escosura is very explicit about the uncertain nature of the resulting 'quantitative conjectures'.

Again, it is important to put this uncertainty in context. The available evidence concerning incomes of the past does not suggest that people in Africa in the 19th century were much richer than Europeans at the time. Additionally, what we know about living conditions more broadly supports this. Riley (2005) provides estimates of life expectancy for all world regions and suggests that Africa, with a life expectancy of 26 years in 1770, was the worst-off region in this respect (Riley 2005). Finally, the population of sub-Saharan Africa accounted for around 6% of the world population in the 19th century. Uncertainty concerning the level of incomes in this region in the past can only have a limited impact on the resulting global estimates.

Overall, whilst our knowledge of the incomes of the distant past is very far from complete, the bulk of the world's population over the last two hundred years lived in countries that have been studied extensively by economic historians. Moreover, as we demonstrate in an online data appendix, the fact that incomes today are estimated to be several times larger than those of the past means that the broad long-run trends that are the focus of this chapter are robust to wide margins of error.<sup>26</sup> The historical data, though incomplete, are still sufficient to provide us an overall idea of how poverty has evolved across the world over the last two hundred years.

<sup>25</sup> The procedure is required for around 10–15% of the world's population between 1850 and 1950, though this rises to 36% in 1820. From 1950 coverage increases markedly in the Maddison Project Database, and from this point, it is only former USSR and Yugoslavia member states for which this method must be applied.

<sup>26</sup> https://ourworldindata.org/history-of-poverty-data-appendix (accessed on 9 September 2021).

#### **4. Historical National Accounts-Based Estimates of Global Poverty**

#### *4.1. Comparing Two Approaches to Global Poverty Measurement*

Combining the available historical data on GDP per capita and inequality previously described allows us to estimate the extent of poverty across the world over the past two hundred years. The estimates from this 'national accounts' approach to global poverty measurement are presented and discussed below.

However, before doing so, we discuss the important ways in which this approach differs from the estimates based on household surveys outlined in Section 2. Estimating the extent of global poverty based on household surveys is used by international organisations to measure progress towards the SDG goal of eradicating extreme poverty by 2030. Therefore, it is important that we understand how our historical poverty estimates, achieved via a different set of methods, relate to these more familiar poverty estimates.

#### 4.1.1. How Do the Two Approaches Differ?

The key difference between the household survey- and national accounts-based approaches relates to the different average incomes to which the poverty estimates are anchored: whether the average reported by the surveyed households or a national accounts aggregate such as GDP per capita. The averages reported in survey data are typically lower—in some cases, much lower—than the national accounts aggregates. Poverty estimates produced using the national accounts approach accordingly result, therefore, in substantially lower poverty estimates. In many cases, national account aggregates have grown at a faster rate than the survey data averages, and where this occurs, it results in an increasing divergence between the two sets of poverty estimates over time. Pinkovskiy and Sala-i-Martin (2016), for instance, calculate global poverty rates according to the two approaches and find that survey-based estimates are four to five times higher than national accounts estimates and fell less rapidly between 1992 and 2010.

There are several reasons for the discrepancy between national accounts and survey means (see Deaton 2005 for a comprehensive discussion).

Firstly, there are conceptual differences in what is being measured in each case. GDP includes many items that are typically not measured in household income surveys, such as an imputed rental value of owner-occupied housing, the retained earnings of firms, and taxes on production, such as VAT. The gap is even larger when GDP is compared to surveys of household consumption—the latter concept excluding both investment expenditure and government expenditure on public services such

as education and health. Other aggregates beyond GDP are available in the national accounts that are more comparable to the concepts applied in household income and consumption surveys. However, important differences still remain even here. For example, in addition to imputed rents, imputations for the value of certain financial services, such as bank accounts, are included in aggregate household consumption measured in national accounts, with no equivalent for these items recorded in the survey data. In many countries, the consumption of nonprofit institutions serving households (NPISH) is included as part of household consumption within national accounts but not within household surveys.

On top of these conceptual differences is a range of mismeasurement problems that affect both sets of data. Whilst in principle, national accounts aggregates should include the value of unreported economic activity in the informal or secondary economy—including food grown for households' own consumption—in practice, compilers of national accounts face particular difficulties in making such an assessment. Estimates of total agricultural output are often derived by multiplying acres under cultivation by a measure of agricultural productivity—a process that can offer a distorted view where out-of-date assumptions concerning these are applied (Deaton 2005). As Ravallion (2003) explains, incomplete measurement of non-exchanged output or that of informal employment can not only affect estimates of the level of total output, but also the trends: "As an economy develops, the household-based production activities that are not measured in the [national accounts] become "formalised," imparting an upward bias to measured NAS growth rates of output".

Since these activities are thought to be better captured by survey data, this may contribute to the growing discrepancy observed between survey and national accounts means. Survey data are, however, also subject to a number of different sources of measurement error. Although there are potential problems along the income distribution, much of the concern relates to how well incomes or consumption at the top of the distribution are captured. There is evidence, for instance, suggesting that richer people are less likely to respond to surveys and that this may bias downwards both the level and rate of growth of average incomes or consumption reported in survey data (Deaton 2005; Korinek et al. 2006). There is also the problem of the considerable heterogeneity in the survey methods applied across countries and years which can, in some instances, have a very significant impact. Whilst this is unlikely to contribute to the overall divergence in trends with national accounts data, it is a source of substantial 'noise' in the resulting poverty estimates (Karshenas 2003).

The fact that both sets of data suffer from known measurement problems has resulted in some disagreement among poverty researchers as to which of these approaches, or what combination of them, offers the most reliable picture about the evolution of global poverty (Pinkovskiy and Sala-i-Martin 2016; Chen and Ravallion 2010; Karshenas 2003).

Pinkovskiy and Sala-i-Martin (2016) point to the fact that nighttime lights, as viewed in satellite images, are much more closely correlated with GDP per capita than with survey means. They argue that this provides independent evidence that national accounts offer a more accurate picture of the true evolution of average incomes and that, consequently, poverty estimates should be anchored more closely to these means. Even if national accounts data do offer a truer picture of the average level of income or consumption, the fact that much of the concern about mismeasurement in survey data relates to the upper end of the distribution makes many researchers sceptical of the uncritical use of national accounts means for the purposes of poverty measurement (Atkinson 2019, pp. 139–43; Chen and Ravallion 2010; Deaton 2005). Since these measurement errors in the survey data are likely to affect not only estimates of the mean but also of the extent of inequality, in making use of the latter but rejecting the former, there is arguably some inconsistency in the national accounts approach. Korinek et al. (2006) provide empirical evidence for the United States that highlights this point. By comparing survey response rates across geographic areas, they are able to make estimates of the relationship between a household's income and their likelihood of participating in an income survey. Correcting for this differential nonresponse 'appreciably increases mean income and inequality, but has only a small impact on poverty' (Korinek et al. 2006).

Our use of the national accounts method to produce the global poverty estimates provided below is driven by our objective of arriving at a broad understanding of poverty trends over the very long run, rather than any assessment of the relative merits of this method for measuring global poverty today. However, amidst this debate, it is important not to overexaggerate the uncertainties involved and lose sight of the key points on which both approaches agree: there have been substantial reductions in the share of the world's population living in poverty in recent decades across a wide range of different poverty lines.

In addition to helping to pinpoint their respective flaws, the conjunction of the two different approaches increases the confidence of poverty researchers that the share of people below a wide range of poverty lines has indeed substantially decreased.

4.1.2. A Comparison of Recent Trends: Poverty Estimates Based on National Accounts vs. Poverty Estimates Based on Survey Data

With this in mind, before presenting our historical estimates of global poverty over the last two hundred years, we first investigate how the estimates based on national accounts compare to the household survey-based estimates made by the World Bank for recent decades in which both are available.

In making a comparison of trends across the two approaches, it is important to bear in mind that incomes have risen at different rates at different points in the global distribution (Lakner and Milanovic 2016). We noted this earlier when examining the different poverty lines used by the World Bank: the share living below the international poverty line fell faster than the share below higher poverty lines. However, this means that given a difference in the *level* of poverty estimated by the two different approaches, part of the difference in the *trends* we observe is due to a given dollar value poverty line tracking the evolution of a different part of the global distribution in each case.

In order to separate this factor from the concerns of a growing divergence between mean incomes in household survey and national accounts data, we compare the two sets of poverty estimates in two different ways.

Figure 9 shows a comparison of the global poverty rates according to four different poverty lines—\$1.90, \$5.50, \$10, and \$30. Figure 10 shows the share of the population falling below an income corresponding to the level that marked the bottom quarter, the median, and the top quarter of the distribution in 1980 (in the case of the World Bank data, the earliest year, 1981, is used as the reference). In the World Bank estimates, these thresholds were \$1.19, \$2.40, and \$10.40 a day. Within our national accounts-based estimates, these thresholds lie at \$3.05, \$6.48, and \$23.77, respectively—between two and three times higher.

**Figure 9.** The share of world population below different poverty thresholds, according to survey- and national accounts-based estimates. Source: Survey-based estimates from PovcalNet (World Bank); authors' own national accounts-based estimates based on GDP per capita data from Maddison Project Database 2020 (Bolt and van Zanden 2020) and data on income inequality from Global Consumption and Income Project (GCIP).

**Figure 10.** The share of world population below the quartile thresholds in 1980/81, according to survey- and national accounts-based estimates. Source: Survey-based estimates from PovcalNet (World Bank); authors' own national accounts-based estimates based on GDP per capita data from Maddison Project Database 2020 (Bolt and van Zanden 2020) and data on income inequality from Global Consumption and Income Project (GCIP).

We see in Figure 9 that the headcount ratios estimated using the national accounts approach are indeed considerably lower than the World Bank estimates, in line with the discussion in the previous section. Global poverty measured against a \$1.90 a day line following the national accounts method was less than a third of the level of the survey-based estimates in 1980/81 (red lines). This gap subsequently narrowed considerably, although, from 1990, the two estimates fell in proportional terms at roughly the same rate—both falling by slightly more than half until 2017. The size of the gap between the two sets of estimates, and how it changes over time, is somewhat different at each poverty line.

However, a notable observation from Figure 9 is that where the trend lines pertaining to different poverty thresholds happen to fall close to one other—such as the survey estimates relating to the \$1.90 line (solid red) and the national accounts estimates for \$5.50 (dotted purple), or likewise the estimates for \$10 (solid blue) and \$30 (dotted yellow)—the trends move much more in step. Whilst poverty lines defined in terms of a given dollar value yield very different estimates across the two sets of data, poverty lines defined in terms of their position within the global distribution result in much more agreement.

This observation is confirmed by Figure 10 which makes this comparison more directly. The share of people falling below the income level that marked the bottom quartile, the median, and the top quartile in 1980/81 evolved broadly similarly across both sets of data. For instance, the two approaches disagree about the global median level of income in 1980/81: it was \$2.40 according to survey data and \$6.50 when anchoring incomes to GDP per capita. However, both sets of estimates agree that the share of people falling below that level of income fell from 50% to around 20% in 2017. The estimates for the bottom and top quartile thresholds do not move as closely as at the median, but nor are the trends all that dissimilar. Interestingly, they diverge in different directions.

Overall, Figure 10 shows that the trends in global poverty according to the two methods are not in fact as different as one might suspect. It suggests that whilst there may be a good deal of uncertainty in the level of global poverty at any one point in time, our understanding of the key changes seen across the bulk of the global distribution in recent decades is not dependent on the choice of method.

#### *4.2. National Accounts-Based Estimates of Global and Regional Poverty Since 1820*

Figure 11 shows historical estimates of the share and number of people globally living at different income thresholds—\$2, \$5, \$10, and \$20—based on the national accounts method and data sources just outlined. All figures are expressed in 2011 international-\$, so as to adjust for inflation over time and price differences across countries. A more detailed description of all the data and methods used to produce these estimates can be found on our website.<sup>27</sup>

**Share of world population Number of people**

**Figure 11.** The share and number of people globally living below different poverty thresholds, national accounts estimates 1820–2017. Source: Authors' own estimates based on GDP per capita data from Maddison Project Database 2020 (Bolt and van Zanden 2020) and data on income inequality from van Zanden et al. (2014) and the Global Consumption and Income Project (GCIP).

What these estimates allow us to see is that declines in the share of the world's population living in poverty were not limited to the recent decades for which extensive household survey data are available. We see a more or less continuous decline in the share of the world's population below each poverty line that accelerated in the second half of the 20th century (Figure 11, left panel). As the global population rose from around 1 billion in 1820 to 6 billion in 2000, the number of people living in poverty was rising. This is true for all but the lowest threshold of \$2 a day, below which the number of people stayed largely constant until very recently (Figure 11, right panel).

As in the World Bank estimates based on survey data (Figure 4), we see a further acceleration in the decline in poverty rates from around the turn of the new

<sup>27</sup> https://OurWorldInData.org/history-of-poverty-data-appendix (accessed on 9 September 2021).

millennium. From this point, the number of people living below each poverty line began to decrease.

It is not the case that the number of poor people declined everywhere, however. Figure 12 shows the poverty rate measured relative to a \$5 a day line for each region. Figure 13 shows the share of the world population falling below this threshold, where each region's contribution is shown separately.

We see from Figure 12 that, relative to this income threshold, a majority lived in poverty across all parts of the world in 1820. That is true for today's rich countries, although we see that poverty was less prevalent in Western Europe, along with its 'offshoots' (the US, Canada, Australia, and New Zealand), than in other world regions.

**Figure 12.** Share of the population living below \$5 a day, national accounts estimates 1820–2017. Source: Authors' own estimates based on GDP per capita data from Maddison Project Database 2020 (Bolt and van Zanden 2020) and data on income inequality from van Zanden et al. (2014) and the Global Consumption and Income Project (GCIP).

#### **Share of world population living below \$5 a day**

**Figure 13.** Share of the world population living below \$5 a day by region, national accounts estimates 1820–2017. Source: Authors' own estimates based on GDP per capita data from Maddison Project Database 2020 (Bolt and van Zanden 2020) and data on income inequality from van Zanden et al. (2014) and the Global Consumption and Income Project (GCIP).

Globally, almost 90% of the population lived under this threshold in 1820, as we can see from Figure 13. We see that the subsequent fall in the global poverty rates is owed primarily to the decreasing contributions to this total made by Western Europe, Eastern Europe and Central Asia, China, and India. In the case of Western Europe and EECA, this was due to falling poverty rates. However, in the case of China and India, poverty rates remained high. Up until the mid-20th century, their reduced contribution to the global poverty rate was due to their share in the world population declining—from more than half in 1820 to around one-third in 1950.

Until the mid-20th century, global poverty fell because the poverty rate in some regions was falling and because the population grew faster in those same regions than in the regions where incomes were stagnant. However, the growth in incomes in India and China from the mid-20th century onwards meant that, for the first time in history, progress was being made against poverty across most of the world. From this point, the decline in global poverty accelerated.

The important exception to this trend, however, is sub-Saharan Africa. Poverty rates remain high in sub-Saharan Africa following limited economic growth in the late 20th century, coupled with persistently high levels of inequality in many countries in the region. As we discussed regarding the World Bank estimates based on household survey data, global extreme poverty is becoming increasingly concentrated in sub-Saharan Africa. What this long-run view makes very clear, however, is that the low and stagnating incomes endured by a large share of the population in many sub-Saharan African countries should not in any way be accepted as inevitable. Persistently high levels of extreme poverty were once the rule but are now the exception.

#### *4.3. The History of Extreme Poverty in the Last Two Centuries: Combining Survey and National Accounts Estimates*

In this chapter, we have presented evidence on the history of poverty based on two different methods. We have referred to these as survey and national accounts estimates, reflecting the different average incomes to which the estimates are anchored: in the first case, to household survey data, and in the second case, to average incomes observed in national accounts or historical reconstructions of this.

On the one hand, we have pointed to many sources of uncertainty in estimating global poverty. For recent decades, this is underlined by the differences in the poverty estimates arrived at by these two methods. There are known sources of mismeasurement in both approaches and questions remain as to how best to combine all the evidence available from national accounts and survey data. Estimates for the distant past, relying on reconstructions from historical sources, are no doubt more uncertain still.

On the other hand, we have argued that this uncertainty must not be overstated. For recent decades, both approaches show substantial falls in poverty across a wide range of poverty lines. The scale of the changes seen in average incomes in many parts of the world over the last two hundred years gives us a clear indication that global poverty also fell substantially over this longer timeframe, even allowing for significant uncertainty regarding the historical data. Whilst estimates for any given point in time are highly uncertain, the available evidence consistently points to a range of trajectories that is plausible and a range of trajectories that is not.

Within this range of plausible trajectories, however, is it possible to construct a single time series for the evolution of global poverty over the long term?

Martin Ravallion's seminal book on the history and measurement of poverty features a chart that shows one possible approach to provide such a long-term

perspective (Ravallion 2016a, Figure 2.1). He presents the long-run poverty estimates of Bourguignon and Morrisson (2002), which are based on the historical national accounts method discussed above, and brings them together with data based on household surveys for the recent period. Bourguignon and Morrisson's estimates were made in relation to a poverty line set specifically so as to result in the same global extreme poverty rate as that found in survey-based estimates for an overlapping year.<sup>28</sup> In this way, the researchers sought to account for the 'gap' in the poverty estimates resulting from the two different sets of methods, as discussed in Section 4.1 above. Ravallion (2016a) extends this historical series for the global extreme poverty forward using the survey-based estimates published in Chen and Ravallion (2010).

That chart is reproduced here in Figure 14, using recent World Bank estimates of the share of the population living below the updated international poverty line of \$1.90 per day from 1981 onwards. Prior to this, we use our own historical poverty estimates that were presented above. These apply very similar methods as those used by Bourguignon and Morrisson (2002) but make use of more recently published historical evidence on both the extent of inequality and the level of average income. As in Bourguignon and Morrisson's original study, to produce the historical estimates of extreme poverty shown in Figure 14, we set a poverty line that results in estimates that align with the survey-based data in the overlapping year.<sup>29</sup>

Supporting this method of combining the two sets of data is the observation made above in relation to Figure 10. There, we saw that estimates of poverty measured relative to a line anchored to a certain point in the global distribution showed similar reductions over time under both approaches. Indeed, we have shown that this was particularly true at the global median, which the extreme poverty line used in Figure 14 fell close to in 1980/81.

The series in Figure 14—as with Ravallion's original chart and any reconstruction of the history of poverty—suffers from the many uncertainties associated with the available evidence. However, the available historical evidence is clear with respect to the broad features of this trajectory: poverty levels were very high in the past, and the share of the world population living in poverty declined significantly.

<sup>28</sup> Namely, the estimates of poverty measured using the '\$1 a day' line found in Chen and Ravallion (2001). This relates to a \$1.08 poverty line measured in 1993 prices with which the authors updated the \$1 a day line given in 1985 prices.

<sup>29</sup> World Bank estimates of the extreme poverty rate in 1981 and 1982 were 42.4% and 42.1% respectively. We set the extreme poverty line in the National Accounts data at \$5.20 (in 2011 prices). This yields a global poverty rate of 43.0% in 1980, roughly continuing the trend linearly.

Moreover, as we discuss in the following final section, this broad trend concerning *monetary* poverty is also corroborated by historical evidence concerning a range of *non-monetary* metrics.

**Figure 14.** Share of the world population living in extreme poverty, 1820–2017. Source: 1820–1980: Authors' calculations based on Maddison Project Database 2020 (Bolt and van Zanden 2020), van Zanden et al. (2014), and Global Consumption and Income Project (GCIP); 1981–2017: PovcalNet (World Bank). Note: This series is based on the methods employed in the long-run global poverty estimates that appeared in Bourguignon and Morrisson (2002) and Ravallion (2016a). It uses a more recently published set of historical data and, for the period from 1981 onwards, more recent World Bank estimates for the share falling below the updated international poverty line of \$1.90 a day.

#### **5. Other Evidence on Historical Poverty**

#### *5.1. Other Long-Run Estimates of Monetary Poverty*

Figure 15 compares our estimates of the share of the world population living in extreme poverty shown above against three other existing sets of estimates.

Our estimates are shown in red. In green are the original estimates of Bourguignon and Morrisson (2002) whose methodology we have largely emulated in producing our historical estimates. In blue is a series produced by Moatsos (2021) that also follows a similar methodology to that described in the present chapter.<sup>30</sup>

**Figure 15.** The share of the world population living in extreme poverty, according to four sets of estimates. Source: Bourguignon and Morrisson (2002), Moatsos (2021), and the authors' own estimates.

In pink is a second series from Moatsos (2021) that is instead based on the 'cost of basic needs' approach suggested by Allen (2017). Within this approach, incomes are measured against a poverty line that varies across countries according to the local minimum cost of meeting dietary and other basic needs.

One advantage of the approach is that it gives us an understanding of the history of global living standards in terms of a more readily interpretable definition of poverty: as recently as 1950 the majority of people in the world 'could not afford

<sup>30</sup> The approaches are similar in terms of the use of PPP-adjusted incomes and the use of historical national accounts and inequality data to extend estimates of the global income distribution into the past. See Moatsos (2021) for a full description of the methodology.

a tiny space to live, food that would not induce malnutrition, and some minimum heating capacity' (Moatsos 2021, p. 195).

The cost-of-basic-needs estimates are also significant in providing an alternative methodology with which to benchmark poverty estimates, such as those of the World Bank and our own historical estimates presented here, that use incomes adjusted for price differences across countries and expressed in common units of purchasing power. This is one aspect of the World Bank's approach that has been the subject of debate in particular, and which the cost-of-basic-needs approach avoids (see Section 2.1.1 above). Overall, the long-run trends across all four series are remarkably similar. The fact that different researchers using different methodologies and data sources reach similar conclusions concerning the history of extreme poverty greatly increases our confidence in the overall trends.

#### *5.2. Long-Run Evidence from Non-Monetary Metrics*

The poorest people in the world today have worse living conditions more broadly. They live shorter lives, lack access to basic services, and a higher share suffers from hunger and malnutrition. To assess how plausible, or implausible, the historical reconstructions of poverty are, we should therefore look at the historical evidence for the living conditions in the past. Does the historical evidence on non-monetary metrics such as mortality and malnutrition match the reconstructions of high levels of monetary poverty?

#### 5.2.1. Mortality at a Young Age

Figure 16, based on Volk and Atkinson (2013), shows the share of children who died before they reached the end of puberty.<sup>31</sup> These data, covering the last 2400 years, relate to a range of different locations from around the world. What is striking about the historical estimates is how very similar the mortality rates for children were across this wide range of 23 historical cultures. Whether in Ancient Rome, Ancient Greece, the pre-Columbian Americas, Medieval Japan or Medieval England,

<sup>31</sup> In modern global health statistics, child mortality is defined as the share of children who die before the age of five. The historical research does not provide data for this age cutoff. A cutoff at the end of puberty has the advantage that it captures mortality over the entire course of childhood. To compare the historical estimates with modern global health data, we relied on data from the United Nations Inter-agency Group for Child Mortality Estimation (IGME) which publishes the mortality rate up to the age of 15 for countries around the world.

the European Renaissance, or Imperial China—no matter when and where a child was born, almost one in two children did not survive.

Volk and Atkinson also bring together mortality data from 20 different hunter-gatherer societies from very different locations to give an indication of the youth mortality rate in the type of society that humans in prehistoric times lived in.<sup>32</sup> Again they find very similar mortality rates with an average death rate until the end of puberty of 48.8%, almost exactly the same as the historical sample over the last three millennia.

The high mortality of children in all world regions is plausible when we consider the evidence on humanity's population growth. We know that population growth was close to zero, while fertility rates were high. The fertility rate, the average number of children per woman in the reproductive age bracket, was high—an average of 6 or more children per woman was certainly not rare (Roser 2014). A fertility rate of 4 children per woman would imply a doubling of the population size each generation; a rate of 6 children per woman would imply a tripling from one generation to the next. However, instead, the population barely increased: historical reconstructions suggest that between 10,000 BCE and 1700, the world population grew by only about 0.04% annually (Roser et al. 2013). A high number of births without a rapid increase of the population can only be explained by one sad reality: a high share of children died before they could have children themselves. The historical evidence that almost half of all children died certainly does not seem consistent with notions that poverty levels were low in the past.

Equally important for the plausibility of the historical poverty reconstructions is the fact that both population growth and declining mortality levels coincide with the decline in poverty that the national accounts data suggest, at both the global and country level. Globally, the chart shows that the global death rate of children younger than 15 declined from close to 50% to below 5% over the course of the last century. And today, populations in places with high levels of poverty still suffer from

<sup>32</sup> To study mortality at a young age in prehistoric societies, the researchers need to mostly rely on evidence from modern hunter-gatherers. Here, one needs to be cautious of how reflective modern hunter-gatherer societies are of the past. This is because recent hunter-gatherers might have been in exchange with surrounding societies and 'often currently live in marginalised territories', as the authors state. Both of these could matter for mortality levels.To account for this, Volk and Atkinson have attempted to only include hunter-gatherers that are best representative for the living conditions in the past; they limit their sample 'only to those populations that had not been significantly influenced by contact with modern resources that could directly influence mortality rates, such as education, food, medicine, birth control, and/or sanitation'. The one study on mortality rates of paleolithic hunter-gatherers finds a higher youth mortality rate: 56% did not survive to puberty.

youth mortality rates of up to 15%, while in places with the lowest poverty rates, up to 99.7% of children now survive their first 15 years of life.

**Figure 16.** Youth mortality over the last 2400 years. Source: The mortality estimates for historical societies are from a large number of independent studies collected in Volk and Atkinson (2013). Data for 2017 are from the United Nations Inter-agency Group for Child Mortality Estimation (IGME). Global estimates for the 20th century are based on under-five mortality from Our World in Data. Note: Youth mortality measures the share who died as infants or children before reaching the end of puberty (approximately around the age of 15).

#### 5.2.2. Human Height

A second set of evidence that allows us to study whether the reconstructions of monetary poverty over the last few generations are plausible can be found in the mortal remains of people around the world.

An individual person's height is largely determined by their genetic background, but the average height of an entire population is almost entirely determined by their living conditions, particularly the nutrition and health at a young age (Baten and Blum 2014). This allows historians to reconstruct people's living conditions by relying on the average human height in a population as a proxy measure. When there are no records of population height over time, the average height can be reconstructed from bones.

Figure 17 shows how the height of adult men has changed over the course of the last century (estimates of women's heights are published by the same source and show a very similar change).<sup>33</sup> The differences in the height of men born in 1996 correlate closely with levels of monetary poverty today, the height of men born in the richest parts of the world being the tallest, while those born in the poorest countries being among the shortest.

**Figure 17.** Height of adult men, 1896–1996. Source: NCD Risk Factor Collaboration (NCD-RisC), Our World in Data. Note: These data relate to the height of adult men by year of birth. Poor nutrition and illness in childhood limit human growth. As a consequence, the average height of a population is strongly correlated with living standards in a population.

Furthermore, importantly for the question at hand, it is also the case that the historical changes in men's height match closely with the historical reconstructions of monetary poverty presented before. These data show large changes in men's height over time: the global average increased by 9 cm from 1.62 m to 1.71 m, and in the regions and countries that made the fastest progress against poor living conditions, it can be well over 10 cm. The average man in Europe and Central Asia in the

<sup>33</sup> The data are published by NCD Risk Factor Collaboration (NCD-RisC).

late 19th century was smaller than the average man in sub-Saharan Africa today.<sup>34</sup> Additionally, as is the case with poverty reduction, the smallest improvements over the last century are documented for men in sub-Saharan Africa and South Asia. Large increases, on the other hand, are documented for people in North America, Europe, Central Asia, and Latin America and the Caribbean.

The data shown here goes back to the late 19th century when differences in economic prosperity had already emerged. In line with these differences, men in the least poor regions of the world were already the tallest. Reconstructions of the economic history of Europe suggest very poor economic living conditions during the centuries preceding the Industrial Revolution. Long-run reconstructions of human height in Europe by Koepke and Baten (2005) corroborate this long-run perspective.

#### 5.2.3. Escaping the Malthusian Economy

Lastly, it is economic theory and its empirical support that suggests that the very high levels of poverty that we reported earlier do indeed reflect the living conditions in the past.

The mechanism that prevented progress against poverty and hunger in the past is referred to by economic historians as the 'Malthusian trap'. When the large majority of a society suffers from poverty and hunger, only an increase in production can raise living standards and reduce poverty. However, in the past, such productivity increases occurred only very rarely, and whenever they did occur, they only led to a brief increase in living standards because it ultimately caused an increasing size of the population which left everyone as poor as they were before. Due to this basic mechanism, higher productivity did not result in lower levels of poverty but in a larger number of people.

Ashraf and Galor (2011) develop a formal model of the Malthusian economy theoretically and study the evidence for it empirically. If the economic living standards of people in the pre-growth economy were in fact determined by the Malthusian trap, then we would expect to see a positive correlation between the level of productivity in a region and the density of the population in this area. Figure 18 is taken from their publication and confirms the theoretical prediction for the pre-growth economies in the year 1500.

<sup>34</sup> On the question why men in the world's poorest region, sub-Saharan Africa, are slightly taller than men in South Asia, see Bozzoli et al. (2009).

#### **The partial e ect\* of land productivity on population density in 1500 CE**

(logarthmic axis)

**The partial e ect\* of land productivity on income per capita in 1500 CE**

(logarthmic axis)

The colors represent continents: Africa Asia Europe America Oceania

**Figure 18.** The partial effect of land productivity on population density and income per capita in 1500 CE. Source: Ashraf and Galor (2011). Note: The figure depicts the partial regression line for the effect of land productivity while controlling for the timing of the transition from hunting and gathering to agriculture, and the influence of absolute latitude, access to waterways, and continental fixed effects. The x- and y-axes plot the residuals obtained by regressing population density and income per capita, respectively, on these covariates.

All data in this visualisation are reported in the current borders of the world. On the x-axis of both charts, you find the same metric—the productivity of the agricultural land as measured by the quality of the soil and the climate. In the chart on the left we see that those world regions with the most productive land had the highest population density.

On the chart on the right, we see that the higher productivity of the land did not result in higher living standards. The agricultural sector in Spain, India, or Morocco was much more productive than in Finland, Egypt, and Norway, but the people in these countries were not better off—they were merely more numerous. The more productive regions were the more populous regions, and the people in these areas had to share with so many that everyone remained at dismal levels of prosperity.

In the long history before modern economic growth, higher productivity led to larger, but not richer, populations. This mechanism ensured that poverty levels were high everywhere.

#### **6. Conclusions**

Poor material living conditions were such a persistent and pervasive reality that, for much of human history, it was unimaginable that it could ever be different. Poverty did not change, and so, it was easy to believe that poverty was unchangeable. The Reverend Thomas Malthus wrote about the living conditions in his native England 'It has appeared that from the inevitable laws of our nature, some human beings must suffer from want. These are the unhappy persons who, in the great lottery of life, have drawn a blank'.<sup>35</sup>

When Malthus wrote these words in 1789, he was right about the past, but he turned out to be wrong about the world's reality after his death: In the two centuries since his death, many countries broke out of the stagnation of the past, achieved economic growth, and reduced poverty. The reconstructions of poverty presented here make clear that it is *not* an inevitable law of nature that humans must suffer from want. It is not only possible to reduce poverty, but it is a reality.

During the long past when humanity did not make any substantial progress in reducing poverty, there was no one bold enough to even imagine that it could be different in the future. Not only the reality of poverty reduction over the last 200 years stands in sharp contrast to the centuries and millennia preceding it, but with the reduction of poverty, the thinking about poverty has changed as well. Today, poverty is widely considered a social bad, a problem to be solved rather than a reality that needs to be accepted. Ravallion (2013) documents that before the modern reduction of poverty, poverty was considered a social good—'necessary and even desirable for a country's economic success'. That this idea is so repellent for us today makes clear how dramatic the change in perspective has been.

The reality of poverty reduction changed our view on poverty, and this matters substantially for our aspirations for the future. The same evidence that we presented here shows us how far the world has come, and *how* far we still have to go. That fewer than 10% of the world live in extreme poverty is the outcome of unprecedented progress and one of the most atrocious problems that the world faces today. It is the progress that we have made that makes the reality of extreme poverty so appalling—if Malthus was right that poverty was inevitable, we would need to accept the suffering that poverty causes; it is the decline in poverty documented here that makes it clear to us that the world can change and that economic growth and reductions of inequality can reduce poverty.

<sup>35</sup> Thomas Malthus (1798)—An Essay on the Principle of Population. Chapter X, paragraph 29, lines 12–15.

As we emphasised before, the international poverty line is a very low poverty line—the name 'extreme poverty' is apt. As long as there are extremely poor people, it is ethically right that the world considers a poverty line that focuses our attention on the very poorest, but our aspirations should of course not be limited to ending only extreme poverty. According to the PovcalNet data, 85% of the world lives on less than \$30 a day—the world is a very long way away from ending poverty relative to these higher poverty cutoffs. The fact that the global mean income (at \$16 per day according to the 2017 PovcalNet data) is only about half of this poverty line shows just how much the world economy needs to grow to bring an end to poverty into reach.

Even after two centuries of unprecedented progress, an extremely large number of people around the world still suffer from poverty and extreme poverty. What we have learned is that poverty is not inevitable; now it is on us to continue and accelerate the progress the world has made. The history of global poverty reduction has only just begun.

**Conflicts of Interest:** The authors declare no conflict of interest.

#### **References**


© 2021 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).

## **Global Absolute Poverty: The Beginning of the End?**

**Sanjay G. Reddy**

#### **1. Introduction**

#### *The Idea of Ending Poverty, Rhetoric and Reality*

The first Sustainable Development Goal of "Ending poverty in all its forms everywhere" reflects an admirable collective aspiration.<sup>1</sup> The idea that the poor need not always be with us<sup>2</sup> is a revolutionary idea, and arguably a modern one<sup>3</sup> .

If global development goals such as SDG 1 are not meant to be taken literally but rather to provide a guide to action and a horizon for aspiration (see, e.g., Reddy and Kvangraven 2015), neither the use of demanding words such as "ending" and "all", nor the adequacy of their specific definition in terms of targets and indicators, would lead to excessive preoccupation. If goals are meant to provide a concrete objective for policy-making, or a reference for enabling those who frame and implement policies to be held accountable, then the details of their definition may matter a great deal.

The understanding of the first Millennium Development Goal (to "Eradicate Extreme Poverty and Hunger") showed a gap between rhetoric and implementation, because "eradication" was ultimately interpreted with exceedingly modestly, as an intention to halve, between 1990 and 2015, the proportion of people in the developing

<sup>1</sup> I would like to thank Rahul Lahoti for undertaking the calculations, based on the Global Consumption and Income Project, which gave rise to the alternative forecasts reported later in this paper, and for useful suggestions.

<sup>2</sup> Jesus of Nazareth was said to have said (Matthew 26, p. 11): "The poor you will always have with you, *but you will not always have me*". There is no reason to interpret this famous remark as an injunction to fatalism, even if it involved the idea that the eradication of poverty was not a proximate prospect.

<sup>3</sup> Classical political economists such as Smith, Malthus, and Ricardo, because of their notion that wages were determined by the cost of a modest subsistence, were skeptical of the prospects for economic growth leading to a necessary improvement in living standards of ordinary workers, including in particular the eradication of poverty. The attitude of Smith to China is illustrative. He views it both as a most prosperous country and one in which beggars and distress abound. Smith states (in *The Wealth of Nations*, Book 1, Chapter VIII) that "China has been long one of the richest, that is, one of the most fertile, best cultivated, most industrious, and most populous countries in the world", but also that "The poverty of the lower ranks of people in China far surpasses that of the most beggarly nations in Europe".

world living on less than \$ 1.25 (2005 PPP) a day (Pogge 2004; United Nations 2015).<sup>4</sup> Ultimately, the declaration that the first Millennium Development Goal (MDG) had been achieved also turned crucially on an interpretation of the halving of poverty as applying to the global total headcount rather than to regional or national totals. Presumably, the most favorable case for the SDGs would be that they can do both, providing a framework for motivating and directing action and meaningful and well-defined statistical objectives.

This paper examines the likelihood that income poverty will be "ended" by 2030 as demanded by the first Sustainable Development Goal. It is demonstrated that this is unlikely, with the extent of remaining poverty and the regional distribution of poverty depending greatly on the assumptions made. It is also shown that the global economic downturn brought about by policies against COVID-19 has led to a significant setback to the goal. Conceptual issues in estimation, poverty projections, and implications for the attainment of SDG 1 are discussed.

#### **2. Relationship between Goals, Targets, and Indicators: Internal vs. External Views**

As already noted, a basic question when approaching an exercise of a societal nature such as the SDGs is that of the relationship between their public face—the understandings of them in broad social and political contexts—and their technical face—the understandings of them relevant for operational applicability in administrative contexts. Does the technical understanding of SDG 1 correspond to the societal understanding?

An interesting feature of SDG 1 targets and indicators is that they are plural (see United Nations 2019a). The very idea of eliminating poverty in all its forms involves an implicit recognition that any single measure of poverty—which must fail to capture all the forms of poverty that there are—cannot suffice. This recognition is echoed in the fact that diverse targets and indicators were chosen for SDG 1, with indicators referring, for instance, both to "the international poverty line" and to "the national poverty line", to "poverty in all its dimensions", and to the population covered by "social protection floors/systems", having "access to basic services", having "rights to land", affected by "disasters", and that live in localities or countries that adopt and implement "disaster risk reduction strategies". Moreover, each of

<sup>4</sup> The poverty lines referred to are in "international dollars", a unit for assessing purchasing power parity (PPP) that is set notionally equivalent to one US dollar in the United States.

these are required to be disaggregated by various sub-categories, such as age, sex, employment status, geographical location (urban/rural), children, disability status, pregnancy status, whether an individual is a work-injury victim, etc. The idea of poverty adopted for technical purposes appears to involve a somewhat haphazard collection of concepts and is less clear, than the umbrella concept of eliminating poverty in all its forms adopted for public purposes.

How should an analyst or an advocate for poverty reduction respond to this situation? The internal and the external view of the matter may be distinguished. The internal view holds that the technical definition fully determines the meaning of the SDGs. The external view holds that the adequacy of the technical definition of the SDGs must be assessed in light of their broader societal role and responsibility. In the external view, the meaning of the phrase, "eliminating poverty in all its forms" must be examined in light of a broader field of references; accustomed ordinary language uses of terms such as "eliminate" and "poverty" or ambient social and political understandings (as revealed, for instance, by the spirit of political documents such as the Agenda 2030). From the external point of view, although the officially adopted list of targets and indicators (see United Nations (2019a)) provides a relevant, and perhaps even a privileged, reference point, it cannot be viewed as the last word on the subject of whether the goal of "eliminating poverty in all its forms" is adequately being met.

The SDGs ultimately gain their credibility and their authority from their endorsement by political authorities and their acceptance by a wide range of actors; therefore, it seems that the external view demands due attention. Targets and indicators should not become objects of obsession. They must be subject to ongoing scrutiny to assess their individual and joint adequacy for achieving the objective of ultimate interest, "eliminating poverty in all its forms".

#### **3. Slips between Cup and Lip: Questions of Measurement**

There is a wide and well-developed body of literature on the appeal and adequacy of individual poverty measures, which cannot be treated comprehensively here. Many of the questions raised in this literature are relevant to determining the suitability of the chosen SDG indicators. These can guide the application of the external view, since the officially accepted SDG targets and indicators may be inadequate to monitoring whether poverty "in all of its forms" is on course to being "eliminated".

Amartya Sen has noted (see, e.g., Sen 1981) that descriptions of the extent of poverty can be seen as decomposable into two component exercises, viz. identification (e.g., determining who is poor, in what ways and to what degree) and aggregation (e.g., determining the quantity, severity and distribution of poverty in a population). Both exercises can be approached in multiple ways, and there can be reasonable disagreement over the alternative ways of specifying them:

#### (1) Identification

Unidimensionality vs. Multidimensionality:

Should poverty be conceived primarily in terms of inadequate command over material resources (e.g., in the form of income or consumption) or in terms of the presence of deprivations of diverse sorts, whether of means (e.g., access to schooling) or attainments (e.g., years of schooling completed)? In either case, what is the underlying conceptual framework used to determine whether there is inadequacy or deprivation and to guide the selection of indicators?

Adequacy of Thresholds:

In any given dimension (e.g., income or consumption) what is the appropriate threshold to be used in determining adequacy? Specifically, how should a threshold be specified in order for it to have a *meaningful* interpretation as being adequate for poverty avoidance? How should they be defined so as to have a *common meaning* at different points in space and in time? It is not only the setting of a threshold for any one context, but also its translation across contexts to ensure a consistent interpretation that requires reference to a common meaning (see Pogge and Reddy 2010; Reddy 2004, 2007, 2008, 2013, 2020; Reddy and Lahoti 2016; Reddy and Pogge 2006; and Reddy et al. 2008).

#### (2) Aggregation

How should the overall extent of poverty in a society be summarized? For instance, is the number of poor persons, the proportion of poor persons, the typical severity of poverty or a composite measure most suitable? Moreover, is the performance of society in relation to the goal to be judged on the basis of a global aggregate or performance in each region or country? If the latter, what importance is to be given to each region when assessing overall progress?

(3) SDG Targets and Indicators in Light of These Questions

In practice, SDG targets and indicators raise very serious issues. For example, what United Nations (2019a) refers to as the "international poverty line" (the World Bank's \$ 1.90 2011 PPP poverty line, which it has deemed equivalent to its own previous \$ 1.25 2005 PPP poverty line) has come in for serious criticism ["Indicator

1.1.1 Proportion of population below the international poverty line, by sex, age, employment status and geographical location (urban/rural)"]. It has been argued, for instance, that this line lacks meaning in terms of the real requirements for achieving human well-being. This results both from the absence of sound conceptual and empirical underpinnings, and from distortions generated by the price indices used to attempt to maintain purchasing power over space and time (see previously cited writings, and Reddy and Lahoti (2016) for detailed criticism of the\$ 1.90 2011 PPP line and the claim of equivalence to the prior line).

National poverty lines are not necessarily better, because they correspond to many distinct methodologies, often poorly conceived or executed, and subject to political influence ["Indicator 1.2.1 Proportion of population living below the national poverty line, by sex and age"]. Although such measures may be validated by governments, they may not capture poverty in a sense that can be rationally justified and widely accepted. The debates about the adequacy of national poverty lines in many countries testify to this difficulty. Even if these lines have discernible purchasing power interpretations (which they often do not) these are not common across countries. Additionally, many countries, including even otherwise advanced countries, simply do not have official poverty lines (see, e.g., Reddy 2007, 2013; Subramanian 2012, etc.).

In the case of international poverty lines, and very often also in the case of national poverty lines, the focus has been on a stringent "absolutist" concept of poverty, whereas poverty "in all its forms" implies a more expansive concern. For instance, whereas according to the international poverty line, poverty is almost non-existent in most advanced countries, it is frequently present according to national poverty lines, and even prevalent when unofficial poverty lines and rights-based assessments of conditions of the poor are employed.<sup>5</sup>

Similarly, efforts to assess multidimensional poverty using a single composite index, although well-intentioned, may capture but also miss a great deal ["Indicator 1.2.2 Proportion of men, women and children of all ages living in poverty in all its dimensions according to national definitions"]. Such an effort necessarily involves choices about what information to include, or not (depending, in part, on the availability of data), where to place thresholds of adequacy in each dimension, and how to aggregate across dimensions, including, in particular, how to treat correlations

<sup>5</sup> See, for instance, the evaluations of the UN Special Rapporteur on Extreme Poverty and Human Rights, available on https://www.ohchr.org/EN/Issues/Poverty/Pages/CountryVisits.aspx (accessed on 1 August 2021).

between dimensions which reflect intensive concentrations of poverty. As a result, aggregate measures of multidimensional poverty may fail to take adequate note of specific deprivations, or of intense concentrations of multiple deprivations, if these do not greatly influence averages.

The chosen SDG targets and indicators can offer only a partial picture of the extent of poverty in 'all of its forms'. The credibility of the measures which have already been chosen or are likely to play roles as SDG 1 indicators is undermined by various weaknesses. As such, those who adopt the 'external view' of the appropriate relationship between the public and the technical faces of the SDGs cannot, therefore, accept as definitive the picture of global poverty that is presented by such measures alone. This having been said, the measures which are most readily available and are most prominently circulated are also likely to continue to provide public reference points for SDG monitoring. For this reason, we employ in what follows conventional measures (in particular, the \$ 1.90 2011 PPP poverty line and some variants of them within the World Bank's favored "money metric" international poverty line approach) despite our conviction that better poverty monitoring methods are possible and deserve significant additional investment. The use of alternative lines within the money metric approach offers one way of recognizing the uncertainties involved, although a limited one. We have argued extensively elsewhere against the existing measures of global income poverty, and also made a case for an international project to develop credible alternative measures based on the cost of achieving income-dependent human capabilities. Therefore, we shall not make this case again here.<sup>6</sup>

#### **4. Poverty Projections to 2030**

What is the likely evolution of poverty to 2030? We draw on the survey data of the Global Consumption and Income Project (GCIP) (see Lahoti et al. 2016) and consider alternative poverty lines and growth scenarios. Using the data of the GCIP, we were able roughly to replicate the poverty estimates of the World Bank for the SDG initial year (2015), and prior years, although with small discrepancies in estimates

<sup>6</sup> In addition to the readings cited elsewhere in this paper, see also the debate of the author with World Bank economist Francisco Ferreira on the credibility of its existing global poverty estimates, conducted in early 2019: https://www.worldbank.org/en/news/video/2019/03/05/smackdown-debatehow-credible-are-the-world-banks-global-poverty-estimates-how-can-they-be-improved (accessed on 10 February 2020).

for individual regions and countries.<sup>7</sup> In the baseline scenario we considered, we employed projected real income growth rates for individual countries from an available source (the U.S. Department of Agriculture International Macroeconomic Data Set<sup>8</sup> ) which provides publicly available forecasts to 2030, unlike other prominent forecasting sources (notably the IMF World Economic Outlook Database<sup>9</sup> ). We considered both USDA estimates from immediately before the pandemic (January 2020) and more recent ones (January 2021) revised as a result of the pandemic in order to gauge the effect of COVID-19 on global poverty projections.<sup>10</sup>

These are summarized by region and for the world in Table 1. It may be observed that growth projections for the decade fell for all regions as a result of the pandemic, with the world as a whole expected to have annual per capita growth rates that are almost half a percentage point lower than previously expected. In South Asia, the Middle East and North Africa, and Latin America and the Caribbean, more than one percentage point of annual per capita growth is expected to be lost.


**Table 1.** USDA projected compound annual per capita income growth rate between 2020 and 2030.

In the analyses we report on below, we also consider 'low' and high' forecasts, which are based on greater and lesser per capita real income growth rates than those projected by the USDA (one percentage point higher or lower than the baseline growth rate, respectively). Projections can vary greatly depending on the source and

<sup>7</sup> This reflects the presence of some differences between the sources and underlying assumptions of the two databases.

<sup>8</sup> See https://www.ers.usda.gov/data-products/international-macroeconomic-data-set/ (accessed on 25 January 2020).

<sup>9</sup> See https://www.imf.org/external/pubs/ft/weo/2019/02/weodata/index.aspx (accessed on 25 January 2020).

<sup>10</sup> These were downloaded in January 2020 and in March 2021, respectively.

its assumptions, and this gives reason to consider different possibilities. For instance, the USDA estimates of growth rates in sub-Saharan Africa are considerably lower than those of the IMF—a difference which is potentially consequential, due to the presence of high poverty rates in the region. The IMF projected (prior to the onset of the pandemic) an average annual per capita income growth rate through 2024 of 3.89% per annum, whereas the USDA estimate through 2030 was 1.33% per annum.

Population projections were drawn from the same source and used to calculate expected per capita real income growth rates. These alternative growth rates were used to project the initial year per capita real consumption levels for percentiles of national populations, and to arrive at estimated future levels for these same percentiles. Regional consumption levels at each percentile were determined by aggregating national information using the methods described by Lahoti et al. (2016). These were then compared to the real (\$ 2011 PPP) poverty line used (also expressed in terms of real per capita consumption levels) to estimate alternative poverty headcounts and headcount ratios for individual countries, major regions, and for the world as a whole.

The poverty lines chosen were (all in\$ 2011 PPP) \$ 1.90, \$ 2.52, \$ 3.10 and \$ 5.04. The first of these is the "absolute" poverty line, which has been claimed by the World Bank to be equivalent to its previous \$ 1.25 (2005 PPP) IPL (accepted as an SDG indicator by the United Nations). The \$ 3.10 line is the higher poverty line applied by the World Bank (for reasons that are unclear, because limited conceptual justification has been offered for it). The \$ 5.04 line is that which was deemed necessary for meeting basic nutritional requirements in the United States in 2011, according to the Thrifty Food Plan of the USDA (see the discussion in Reddy and Lahoti (2016) of why this should, in principle, provide some guidance as to the minimum cost of basic human requirements elsewhere too, if the PPPs used are taken at face value as preserving purchasing power over relevant commodities). The \$ 2.52 line is half of this basic nutritional standard for the United States, providing a more stringent alternative. In neither case is any allowance made for non-nutritional capabilities.

The current poverty headcount ratios in 2020 in each world region for the various poverty lines used are shown in Table 2.

The baseline pre-pandemic forecast generated the same estimate of the \$ 1.90 2011 PPP global poverty headcount ratio for 2030 as United Nations (2019b), namely, six percent (compared to eleven percent in 2020), as can be seen in Table 3(a). The projected 2030 poverty headcount ratio in sub-Saharan Africa is 36%, considerably greater than that for any other world region. The projected 2030 headcount ratio was expected to be between zero and three percent in every other region. Adopting more

favorable assumptions leads to lower poverty headcount ratios. Growth rates that are two percent higher (closer to IMF estimates) lead to the projected headcount ratio for sub-Saharan Africa being lowered to 27%, and the world headcount ratio falling by one percentage point, to 5%.


**Table 2.** Poverty headcount ratios (%) for different poverty lines for 2020. Source: own estimates based on Global Consumption and Income Project data.

**Table 3.** (**a**) Poverty headcount ratio estimates (%) for \$ 1.90 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount ratio estimates (%) for \$ 1.90 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.



**Table 3.** *Cont.*

As shown in Table 3(b), for the baseline scenario and the \$ 1.90 poverty line, the lower growth estimates as a result of the pandemic lead to a higher expected 2030 level of the poverty headcount ratio in sub-Saharan Africa (42%; six percentage points higher than under the pre-pandemic scenario) and in all regions other than East Asia. The expected 2030 world poverty headcount ratio is raised by two percentage points, or one-third of the pre-pandemic projection, to 8% of the global population. The expected world poverty headcount ratio is also raised in all other scenarios. In the most unfavorable case corresponding to growth rates two percentage points lower than in the baseline estimate, it rises a full three percentage points (to 11% of the global population), with the majority of the population in sub-Saharan Africa (53%) expected to remain in poverty even in 2030.

As can be seen from Table 4(a), under the baseline pre-pandemic growth estimate, the total number of poor persons expected to remain worldwide in 2030 is 515 million people, with the total varying between 385 million and 696 million depending on the growth scenario. The vast majority of these are projected to be in sub-Saharan Africa under all of the scenarios. As shown in Table 4(b), the revised growth estimates as a result of the pandemic lead to much higher estimates of the number of poor, ranging from 470 million to 954 million, with 659 million projected in the baseline scenario.

**Table 4.** (**a**) Poverty headcount estimates for \$ 1.90 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount estimates for \$ 1.90 IPL in 2030 using post-COVID-19 growth estimates.


Number of Poor (in millions)


**(b)**

We also considered alternative poverty lines, recognizing that the \$ 1.90 2011 PPP IPL may be inadequate for specific countries and regions, or globally. For each of these, we once again considered alternative global growth scenarios (the baseline aggregate GDP growth scenario plus or minus one or two percentage points per annum). The estimated poverty headcount ratios and headcounts for regions and for the world are reported for distinct poverty lines and, in each case, for pre- and post-pandemic growth estimates, in Tables 3–10.

The pattern that the majority of the poor remaining in 2030 are expected to be in sub-Saharan Africa does not change when the distinct growth scenarios are applied uniformly across regions (although the specific proportions do, with South Asia becoming a major contributor to the poverty total at the higher poverty lines and under the more unfavorable global growth scenarios. For the pre-pandemic growth estimates, even if the most favorable growth scenario for sub-Saharan Africa (baseline plus two percentage points) is compared with the most unfavorable growth scenario for South Asia (baseline minus two percentage points) and the highest poverty line is considered (\$ 5.04 2011 PPP; see Table 10a), a greater number of poor are expected to be in sub-Saharan Africa (977 million) as compared to South Asia (913 million). The number of poor people in the world in this scenario, even if the other world regions experience strong growth, is more than two billion persons. Considering the various "pure" scenarios, including the baseline and those which raise or lower growth rates uniformly across all regions, leads to the conclusion that at least 2.7 billion people will remain in poverty in all of these scenarios. Even if we consider the lower \$ 2.52 2011 PPP and \$ 3.10 2011 PPP poverty lines, we find that, in all scenarios, at least half a billion people will remain in poverty in 2030. These are hardly circumstances in which poverty will have been 'eliminated'. Considering the less optimistic post-pandemic growth estimates only accentuates this conclusion.

**Table 5.** (**a**) Poverty headcount ratio estimates for \$ 2.52 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount ratio estimates for \$ 2.52 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.


**Table 6.** (**a**) Poverty headcount estimates for \$ 2.52 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount estimates for \$ 2.52 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.




**Table 7.** (**a**) Poverty headcount ratio estimates for \$ 3.10 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount ratio estimates for \$ 3.10 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.


**Table 8.** (**a**) Poverty headcount estimates for \$ 3.10 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount estimates for \$ 3.10 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.




**Table 9.** (**a**) Poverty headcount ratio estimates for \$ 5.04 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount ratio estimates for \$ 5.04 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.


North Africa 27 22 32 18 37 South Asia 47 40 55 33 62

Africa 83 80 86 76 88 World 35 31 39 27 43

Sub-Saharan

**Table 10.** (**a**) Poverty headcount estimates for \$ 5.04 IPL in 2030 using pre-COVID-19 growth estimates; (**b**) poverty headcount estimates for \$ 5.04 IPL in 2030 using post-COVID-19 growth estimates. Source: own estimates based on Global Consumption and Income Project data.


The alternative growth scenarios considered in the tables involve the application of the "same" poverty line in different world regions. However, there is a question as to whether or not the poverty lines involved are in fact the same in a meaningful sense, as a result of deficiencies in current PPPs as constant price indices for the cost of poverty avoidance. In the presence of these problems, poverty lines which, for different world regions or countries, capture the cost of poverty avoidance in terms of purchasing power over commodities necessary to avoid poverty locally according to a common criterion may correspond to *di*ff*erent* nominal PPP dollar amounts. It cannot be known what these discrepancies are without full-fledged studies leading to the construction of suitable country-specific poverty lines reflecting a common understanding of what poverty avoidance demands. The \$ 1.90 2011 PPP poverty line is highly conservative for developed countries (such as the 'base country' for PPP price indices, the United States, where it is clearly inadequate to avoid poverty even according to absolutist standards such as those offered by the Thrifty Food Plan). There is therefore reason to think that more realistic poverty lines, when expressed in 2011 PPP dollars, would be higher for at least some countries. It is likely that more realistic poverty lines would be attained through adjustments that vary across countries and regions. Any appearance that different poverty lines (in \$ 2011 PPP units) are being applied in different regions as a result of such modifications would be only an optical illusion, reflecting the need to correct for systematic mismeasurement of the PPPs being used at present, when applied to poverty lines, in order to ensure that they correctly measure the same thing everywhere.

#### **5. Poverty Reducing the Impact of Economic Growth, and Implications**

Where is the poverty-reducing impact of economic growth the greatest? The relationship between a change in the growth rate and the incremental reduction in poverty in each world region at each poverty line defines a "semi-elasticity". We report in Table 11 the impact of the growth rate being one percentage point less (baseline minus one) or one percentage point more (baseline plus one) at different poverty lines, and for different world regions. It can be seen that the impact of a one percentage point increase or decrease in the growth rate (from the baseline level) is greatest in terms of both headcount ratio and headcount in sub-Saharan Africa, at lower poverty lines. This changes, however, as the poverty line is raised, with South Asia becoming the world region where a change in the growth rate has the largest impact on both the headcount ratio and the total headcount of poverty. At the highest poverty line studied, a one percentage point change makes a difference in the headcount of 142 million in South Asia compared to 43 million in sub-Saharan Africa

(for post-pandemic growth estimates). Growth is poverty-reducing everywhere. A one percentage point increase in global growth makes a difference of between 105 and 321 million poor worldwide, depending on the poverty line chosen, for post-pandemic growth estimates. Growth benefitting sub-Saharan Africa and South Asia has a greater impact. The region where the impact is greatest depends on the poverty line.

At higher poverty lines, additional growth has a sizable impact on poverty in all regions, with its impact on poverty in East Asia rising considerably. At the highest poverty line, the impact of additional growth on the number of poor in East Asia surpasses that on the number of poor in sub-Saharan Africa (although the impact is less than in South Asia). These conclusions qualify the widespread presumption that addressing the problem of absolute poverty worldwide requires a singular focus on sub-Saharan Africa. Especially at higher poverty lines, income poverty is a global problem, and sustaining growth throughout the developing world is important for its reduction.

**Table 11.** (**a**) Headcount semi-elasticities of growth based on pre-COVID-19 growth estimates for different poverty lines (2011 PPP); (**b**) headcount semi-elasticities of growth based on post-COVID-19 growth estimates for different poverty lines (2011 PPP). Source: own estimates based on Global Consumption and Income Project data.






**Table 11.** *Cont.*

#### **6. Conclusions**

Sustained economic growth in developing countries—especially the poorest—is required for global income poverty reduction. The likelihood of achieving the first Sustainable Development Goal of "ending poverty" has diminished, as a result of the economic setbacks experienced in the wake of COVID-19. Our picture of the likely extent of worldwide progress by 2030, and of where remaining poverty is likely to be concentrated, are both greatly dependent on specific assumptions, such as the

poverty line used. Even in the most favorable scenarios, the world will reduce but not "end" poverty. Under less favorable ones, mass poverty is likely to remain a significant concern.

**Funding:** This research received no external funding.

**Conflicts of Interest:** The author declares no conflict of interest.

#### **References**


Subramanian, Sreenivasan. 2012. *The Poverty Line*. New Delhi: Oxford University Press.


© 2021 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (http://creativecommons.org/licenses/by/4.0/).
