The Human Capital Index 2020 UPDATE Human Capital in the Time of COVID-19 The Human Capital Index 2020 UPDATE The Human Capital Index 2020 UPDATE Human Capital in the Time of COVID-19 © 2021 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington, DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 24 23 22 21 This work is a product of the staff of The World Bank with external contributions. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 IGO license (CC BY 3.0 IGO) http:// creativecommons.org/licenses/by/3.0/igo. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution—Please cite the work as follows: World Bank. 2021. The Human Capital Index 2020 Update: Human Capital in the Time of COVID-19. Washington, DC: World Bank. doi:10.1596/978-1-4648-1552-2. License: Creative Commons Attribution CC BY 3.0 IGO. Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to World Bank Publications, The World Bank Group, 1818 H Street NW, Washington, DC 20433, USA; e-mail: pubrights@worldbank.org. ISBN (paper): 978-1-4648-1552-2 ISBN (electronic): 978-1-4648-1647-5 DOI: 10.1596/978-1-4648-1552-2 Cover and interior design: Jihane El Khoury Roederer, World Bank Library of Congress Control Number: 2020951890 Contents Acknowledgments ............................................................................................................................   XV About the Authors ...........................................................................................................................   XVII Abbreviations .....................................................................................................................................   XIX Introduction........................................................................................................................   XXII Overview................................................................................................................................... 4 The HCI 2020 update....................................................................................................................... 5 Income alone does not explain cross-country differences in human capital ................ 5 Most economies achieved human capital gains in the decade before COVID-19........ 6 The extent to which disadvantaged households benefited from human capital gains over time varies across economies......................................................................  7 HCI simulations reveal COVID-19’s large impact on human capital................................ 8 A measure of utilization of human capital highlights significant gender gaps.............. 9 Better measurement enables better policy .............................................................................. 10 Notes...................................................................................................................................................... 11 References...........................................................................................................................................  12 1  The Human Capital Index 2020 Update ................................................................... 14 1.1  The HCI methodology........................................................................................................... 16 1.2  The HCI 2020...........................................................................................................................  18 1.3  HCI 2020: Index components............................................................................................  23 1.4 HCI measures of gender gaps in human capital ..........................................................  30 1.5 Human capital in fragile and conflict-affected contexts............................................. 34 1.6  The HCI 2020 update...........................................................................................................  39 Notes..................................................................................................................................................... 42 References........................................................................................................................................... 43 2 Human Capital Accumulation over Time................................................................... 46 2.1 Human capital accumulation over the past decade...................................................... 47 2.2 Changes in key human capital dimensions over the past decade........................... 52 VIII Cont ents 2.3 A longer-run view of country progress............................................................................ 72 Notes.......................................................................................................................................................76 References........................................................................................................................................... 78 3 Accumulation Interrupted? COVID-19 and Human Capital.................................... 82 3.1 Transmission of the COVID-19 shock to human capital...........................................  83 3.2 The COVID-19 human capital shock: A life-cycle perspective ............................... 87 3.3 Using the HCI to simulate the impact of the pandemic ..........................................  92 Annex 3A.   COVID-19 shock to the under-5 cohorts............................................................. 97 Annex 3B. C   OVID-19 shock to school-age cohorts..............................................................  99 Notes.................................................................................................................................................. 100 References........................................................................................................................................  103 4 Utilizing Human Capital...............................................................................................  108 4.1 Methodology and the basic UHCI measure.................................................................. 110 4.2  The basic UHCI in the data................................................................................................  112 4.3  The full UHCI ........................................................................................................................  114 4.4  The full UHCI in the data.................................................................................................... 117 4.5 Comparing the utilization measures..............................................................................  118 4.6  Disaggregation by region...................................................................................................  120 4.7  Disaggregation by sex..........................................................................................................  120 Notes................................................................................................................................................... 125 References......................................................................................................................................... 126 5 Informing Policies to Protect and Build Human Capital.......................................  128 5.1 Good measurement: Necessity, not luxury................................................................... 129 5.2  Beyond the HCI....................................................................................................................  130 5.3 Building, protecting, and employing human capital in a post-COVID-19 world������������������������������������������������������������������������������������������������������  137 Notes................................................................................................................................................... 139 References......................................................................................................................................... 139 Appendix A  The Human Capital Index: Methodology.............................................  142 Human Capital Index................................................  154 Appendix B  Back-Calculated ­ Appendix C  Human Capital Index Component Data Notes...................................  158 C O N T EN TS IX Boxes 1.1 Learning-adjusted years of schooling...................................................................................... 17 1.2 The Human Capital Index’s aggregation methodology....................................................19 1.3 Limitations of the Human Capital Index.............................................................................. 23 1.4 Measuring Learning Poverty......................................................................................................27 1.5 Schooling for Syrian refugee children in Jordan.................................................................37 1.6 Where did the HCI rankings go?.............................................................................................. 39 2.1 Ensuring comparability across time in the Human Capital Index................................ 48 2.2 Cross-sectoral interventions to address stunting.................................................................57 2.3 Why have expected years of school decreased in Romania?.......................................... 62 2.4 Challenges in test score comparison over time................................................................... 64 2.5 Transforming a low-performing education system into the best school network in Brazil.................................................................................................... 65 2.6 The immediate effects of providing free education in Sierra Leone............................ 71 3.1 Rapid response phone surveys reveal immediate impacts of COVID-19 on the poor............................................................................................................................................. 86 4.1 Deriving the basic Utilization-Adjusted Human Capital Index.....................................111 4.2 Definition of the full Utilization-Adjusted Human Capital Index................................ 116 4.3 Closing gender gaps in human capital outcomes: Where do we go from here?......124 5.1 Innovative data collection in fragile contexts: Examples from West Africa and the Middle East and North Africa................................................................................... 131 5.2 Leveraging national assessments to obtain internationally comparable estimates of education quality..................................................................................................132 5.3 Data quality and freshness in the components of the Human Capital Index..........134 Figures 1.1 The Human Capital Index 2020................................................................................................21 1.2 Concentration of the extreme poor in economies sorted by their Human Capital Index scores.......................................................................................................................21 1.3 Human Capital Index 2020 components, distribution by country income group........................................................................................................................................................ 24 1.4 Decomposition of observed mean HCI differences between selected country income groups............................................................................................................... 25 1.5 Differences between the top and bottom Human Capital Index performers within each country income group......................................................................................... 25 X Cont ents 1.6 Human Capital Index 2020: Index components............................................................ 29 1.7 Sex-disaggregated Human Capital Index and its components...................................31 1.8 Regional and income-group variations in education gaps between boys and girls.............................................................................................................................. 32 1.9 Global variation in gender gaps, Human Capital Index and education components................................................................................................................................ 33 1.10 Human capital and severity of conflict.............................................................................. 36 B1.5.1 Net enrollment rate of Syrian refugees in formal education in Jordan...................37 B1.6.1 Changes in Human Capital Index scores and ranks, 2010 vs. 2020......................... 40 2.1 Changes in the Human Capital Index, circa 2010 vs. circa 2020.............................. 49 2.2 Changes over time.............................................. 51 Human capital and GDP per capita:  2.3 Girl-to-boy ratio, HCI 2010 vs. HCI 2020........................................................................ 52 2.4 Component contribution to Human Capital Index gains, 2010–20....................... 53 2.5 Contribution to changes in the Human Capital Index, by country income group, 2010–20......................................................................................................... 53 2.6 Changes in probability of survival to age 5, circa 2010 vs. circa 2020.....................55 2.7 Changes in fraction of children under 5 not stunted, circa 2010 vs. circa 2020.............................................................................................................. 56 2.8 Changes in adult survival rates, circa 2010 vs. circa 2020........................................... 59 2.9 Changes in expected years of school, circa 2010 vs. circa 2020................................60 2.10 Contribution to change in expected years of school, by country income group, 2010–20..........................................................................................................................61 B2.3.1 Dynamics in enrollment numbers in upper-secondary education, Romania........ 62 B2.3.2 Spending on preprimary and primary education, European Union...................... 63 2.11 Changes in harmonized test scores, circa 2010 vs. circa 2020................................... 65 2.12 Changes in income and Human Capital Index components, 2010–20..................67 2.13 Changes in Human Capital Index components and per capita income, circa 2010 vs. circa 2020, cross-country trajectories..................................................... 69 2.14 Evolution of Human Capital Index components, disaggregated by socioeconomic status............................................................................................................... 70 3.1 Human capital accumulation across the life cycle, key stages and metrics............87 3.2 Learning-adjusted years of schooling lost because of COVID-19 school closures and income shock.................................................................................................... 95 4.1 Employment to population (basic utilization) and Human Capital Index........... 112 C O N T EN TS XI 4.2 Basic UHCI vs. Human Capital Index................................................................................ 113 4.3 Employment to population (basic utilization) and per capita income................... 113 4.4 Basic UHCI and per capita income.................................................................................... 114 4.5 Full utilization rate and per capita income.......................................................................117 4.6 Better employment rates and per capita income.......................................................... 118 4.7 Full UHCI and per capita income....................................................................................... 118 4.8 Basic utilization vs. full utilization...................................................................................... 119 4.9 Basic UHCI vs. full UHCI....................................................................................................... 119 4.10 Regional average UHCI or HCI.......................................................................................... 120 4.11 Employment-to-population ratio (basic utilization) and per capita income ........................................................................................................................................ 121 4.12 Full utilization rate and per capita income...................................................................... 121 4.13 Basic UHCI and per capita income....................................................................................122 4.14 Full UHCI and per capita income.......................................................................................122 4.15 Gender gaps in HCI and UHCI, by region.......................................................................123 A1.1 Components of the Human Capital Index, relative to GDP per capita.................. 147 A2.1 Human Capital Index with uncertainty intervals..........................................................148 B.1 Comparing the 2018 and back-calculated 2018 Human Capital Indexes..............156 C1.1 Comparing original and back-calculated 2018 under-5 mortality rates............... 160 C1.2 Under-5 mortality rates, Human Capital Index 2020, relative to GDP per capita............................................................................................................................................ 161 C1.3 Sex-disaggregated under-5 mortality rates, relative to GDP per capita................. 161 C1.4 Under-5 mortality rates, by income group and region............................................... 162 C2.1 Comparing original and back-calculated 2018 expected years of school..............167 C2.2 Vintage data year for back-calculated 2018 and original 2018, increase of 0.5 year or more in expected years of school..................................................................168 C2.3 Enrollment type for back-calculated 2018 and original 2018, increase of 0.5 year or more in expected years of school................................................................. 169 C2.4 Vintage data year for back-calculated 2018 and original 2018, decrease of 0.5 year or more in expected years of school...................................................................171 C2.5 Enrollment type for back-calculated 2018 and original 2018, decrease of 0.5 year or more in expected years of school.................................................................. 172 C2.6 Expected years of school circa 2020, relative to GDP per capita ............................ 173 C2.7 Sex-disaggregated expected years of school, relative to GDP per capita............... 173 XII Cont ents C2.8 Expected years of school, by income group and region............................................. 174 C3.1 Comparing original and back-calculated 2018 harmonized test scores................ 177 C3.2 Harmonized test scores, Human Capital Index 2020, relative to GDP per capita..................................................................................................................................... 178 C3.3 Sex-disaggregated harmonized test scores, relative to GDP per capita.................179 C3.4 Harmonized test scores, by income group and region................................................179 C4.1 Comparing original and back-calculated 2018 stunting rates...................................182 C4.2 Stunting rates, Human Capital Index 2020, relative to GDP per capita................183 C4.3 Sex-disaggregated stunting rates, relative to GDP per capita....................................184 C4.4 Stunting rates, by income group and region...................................................................184 C5.1 Comparing original and back-calculated 2018 adult mortality rates......................186 Adult mortality rates, Human Capital Index 2020, relative to GDP per C5.2  capita............................................................................................................................................ 187 C5.3 Sex-disaggregated adult mortality rates, relative to GDP per capita.......................188 C5.4 Adult mortality rates, by income group and region.....................................................188 Map B5.3.1 Coverage of live births registration...................................................................................135 Tables 1.1 Human Capital Index 2020, averages by World Bank region..................................... 18 1.2 The Human Capital Index (HCI), 2020.............................................................................. 41 2.1 Regional coverage of the Human Capital Index over-time sample......................... 49 2.2 Changes in Human Capital Index components, 2010–20...........................................54 3.1 Simulated drop in Human Capital Index due to the pandemic’s impacts on children 5 and under.......................................................................................................... 94 3.2 Human Capital Index shock to children currently in school during the pandemic..................................................................................................................................... 96 3.3 Human capital loss of the workforce in 2040.................................................................. 96 3B.1 Mitigation effectiveness, by scenario and income group............................................. 99 3B.2 School productivity................................................................................................................100 C3.1 Source data for economies with different values in original 2018 and back-calculated 2018............................................................................................................... 178 C4.1 Source data for economies with different values in 2018 and back- calculated 2018..........................................................................................................................183 C O N T EN TS XIII C7.1 Data sources for every education level for economies with an absolute change in EYS of at least 0.5, original 2018 and back-calculated 2018...................189 C7.2 Data sources for every level of schooling for economies with a decrease in EYS between 2010 and 2020 ......................................................................................... 193 C8.1 Human Capital Index and components: 2020, 2018 back-calculated, and 2010......................................................................................................................................195 Acknowledgments The Human Capital Index is a collaboration between the Chief Economist offices of the Human Development Practice Group and of the Development Economics Group in the World Bank. The 2020 update was led by Roberta Gatti and Aart Kraay and produced by Paul Corral, Nicola Dehnen, Ritika D’Souza, and Juan Mejalenko. Noam Angrist, Syedah Aroob Iqbal, and Harry Patrinos updated the harmonized test score outcomes. We are grateful to Pablo Ariel Acosta, Rita Kullberg Almeida, D. H. C. Aturupane, Anne Margreth Bakilana, Tekabe Ayalew Belay, Paolo Belli, Livia M. Benavides, Kamel Braham, Fadila Caillaud, Carine Clert, Jorge Coarasa, Gabriel Demombynes, Heba Elgazzar, Sameh El-Saharty, Stefan Emblad, Lire Ersado, Antonio Giuffrida, Inaam Ul Haq, Susanna Hayrapetyan, Samira Ahmed Hillis, Camilla Holmemo, Keiko Inoue, Timothy Johnston, Pierre Joseph Kamano, Olga Khan, Christophe Lemiere, Yasuhiko Matsuda, Muna Meky, Sophie Naudeau, Dorota Agata Nowak, Emre Ozaltin, Aleksandra Posarac, Maria Laura Sanchez Puerta, Hnin Hnin Pyne, Jamele P. Rigolini, Rafael Rofman, Cristina Isabel Panasco Santos, Aparnaa Somanathan, Lars Sondergaard, Michel Welmond, William Wiseman, Ruslan Yemtsov, and Xiaoqing Yu for careful data review. We are also grateful to Luis Eduardo San Martin and Luiza Andrade from the DIME Analytics team for a thorough code review. This report was written by a core team led by Roberta Gatti and including Paul Corral, Nicola Dehnen, Ritika D’Souza, and Juan Mejalenko. Steven Pennings wrote the chapter on human capital utilization. This report benefited from Aart Kraay’s advice and from analysis); Amer Hasan and Fiona Mackintosh analytical inputs by Daniel Halim (gender ­ (case studies narrative); Jigyasa Sharma (fragile contexts); Joao Pedro de Azevedo and Diana Goldemberg (COVID-19 impact on learning-adjusted years of ­ schooling); Dina Abu-Ghaida and Mohamed Audah (schooling in Syria); Alejandro de la Fuente (school- ing in Sierra Leone); Chloé Desjonquères (learning progress in Ceará); Alina Sava and Lars Sondergaard (schooling in Romania); Utz Pape (rapid response phone surveys); Halsey Rogers (challenges in test-score comparison over time); Saskia de Pee, Cecilia Garzón, and Naveed Akbar (nutrition interventions in Pakistan); and Emanuela Galasso, Lisa Saldanha, Meera Shekar, Marie-Chantal Uwanyiligira, and Kavita Watsa (cross-sectoral approaches to combat stunting). We are grateful to Diego Angel-Urdinola, Aneesa Arur, Salman Asim, Anne Margreth Bakilana, Livia Benavides, Catalina Castillo Castro, Carine Clert, Verónica Díaz Hinostroza, Sameh El-Saharty, Karlene Francis, Laura Gregory, Timothy Johnston, Amira Kazem, Flora Kelmendi, Igor Kheyfets, Sophie Naudeau, Jamele Rigolini, Hiroshi Saeki, Maria Laura Sanchez Puerta, Emmanuel Skoufias, Aparnaa Somanathan, Ryoko Tomita, and Inaam Ul Haq for providing country-level insights into changes in human capital outcomes over time. XVI Acknowle dg me nts The team is indebted to David Weil for his overarching guidance. We are grateful to our peer reviewers Shubham Chaudhuri, Rachel Glennerster, William Maloney, and David Weil for their insightful views and to Deon Filmer for his detailed comments on earlier versions of this draft. We thank Kathleen Beegle, Hana Brixi, Emanuela Galasso, Ramesh Govindaraj, Ambar Narayan, Meera Shekar, Sharad Tandon, Tara Vishwanath, and Michael Weber for thoughtful comments and conversations. We are grateful to Alex Irwin and Nora Mara for their outstanding editing touch; to Chloé Desjonquères, Nicola Dehnen, and Mary Fisk for efficiently managing the report’s production process; and to Ruben Conner, Sebastian Insfran, and Andres Yi Chang for their careful read of the report. This Human Capital Index update was developed under the strategic guidance of Mari Pangestu, Annette Dixon, and Mamta Murthi and benefited from the views of Nadir Mohammed and Alberto Rodriguez. September 2020 About the Authors Roberta Gatti, the Task Team Leader for this report, is the Chief Economist of the Human Development Practice Group at the World Bank. In this capacity, she co-leads the conceptualization and release of the Human Capital Index (HCI) and oversees the Service Delivery Indicators surveys initiative. After joining the World Bank as a Young Professional, she has worked in the Development Research Macro Group and in the Social Protection and Jobs units in the Middle East and North Africa and in Europe and Central Asia, based in Washington, DC, and in Bulgaria and Poland. She has also served as the World Bank Global Lead for Labor Policies. Roberta’s research includes theoretical and empirical contributions on labor and household economics, political economy, growth, and social inclusion, and is published in top field journals. She is also the author of numerous World Bank flagship reports on jobs, informal- ity, the Roma inclusion agenda, and the economics of human capital. She has taught at Georgetown University and Johns Hopkins University. An Italian national, Roberta holds a bachelor of arts degree from Università Bocconi and a PhD in economics from Harvard University. Paul Corral is a Senior Economist in the World Bank’s Chief Economist Office for Human Development. He previously worked in the World Bank’s Poverty and Equity Global Practice and was part of the global unit working on just-in-time microsimu- lation models and small area estimation methods and applications. He led the work on small area estimation, which has revamped the institution’s tools used for poverty mapping. He has published peer-reviewed articles on agricultural development and is the author of multiple Stata commands. An Ecuadorian national, he holds a PhD in economics from American University and a master of science degree in agricultural economics from the University of Hohenheim. Nicola Dehnen is a Research Analyst in the World Bank’s Chief Economist Office Human Development. She works on the Human Capital Index (HCI) and the for  Human Capital Project. Her research covers topics in education, health, labor markets, and social protection. Previously, she worked on early learning programs in the World Bank’s Education Global Practice and on social safety nets in the Social Protection, Labor, and Jobs division for West Africa. Before joining the World Bank, Nicola worked at the Inter-American Development Bank, the Leibniz Centre for European Economic Research, and the German Institute for Economic Research. She holds a graduate diploma in economics from the University of Nottingham and a master of science degree in economics from University College London. XVIII A BOU T THE AU T H ORS Ritika D’Souza is a Research Analyst in the World Bank’s Chief Economist Office for Human Development. She works on the Human Capital Index (HCI) and related ana- lytics, including methodologies for the socioeconomic and spatial disaggregation of the HCI. Previously, she worked in the South Asia Chief Economist’s Office, where her research covered the areas of nutrition, education, gender, and jobs. She has also man- aged the field implementation of impact evaluations of agriculture, food security, and nutrition projects in Nepal and Rwanda with the World Bank’s Development Impact Evaluation (DIME) group. She holds a master’s degree in public administration from the School of International and Public Affairs, Columbia University. Juan Mejalenko is a PhD student at the University of Chicago Booth School of Business and has previously worked in the World Bank’s Chief Economist Office for Human Development. Before joining the World Bank, he worked at the Inter-American Development Bank as a Research Analyst in the Office of Strategic Planning and Development Effectiveness, as well as in the Labor Markets division. An Argentinian national, he holds a bachelor’s degree and master of science degree in economics from San Andrés University, Argentina. Steven Pennings is an Economist in the Macroeconomics and Growth Team of the World Bank’s Development Research Group. His research covers a variety of topics in macroeconomics, development, and political economy, including monetary policy, fiscal multipliers and rules, exchange rate pass-through, the determinants of conflict, and the growth contributions of national leaders. He also co-leads the World Bank’s Long-Term Growth Model project. Before joining the World Bank, he worked at the Reserve Bank of Australia and at Save the Children in Vietnam. He has spent time at the International Monetary Fund, the Federal Reserve Board, and the Asian Development Bank. An Australian national, he holds a PhD from New York University and a bachelor of economics degree from the Australian National University. Abbreviations 4Ps Pantawid Pamilyang Pilipino Program (the Philippines) ANER adjusted net enrollment rate BER better employment rate COVID-19 coronavirus disease 2019 DHS Demographic and Health Surveys EGRA Early Grade Reading Assessment EYS expected years of school FCS fragile and conflict-affected situations FQSE Free Quality School Education (program in Sierra Leone) GAML Global Alliance to Monitor Learning GBD Global Burden of Disease (project) GCFF Global Concessional Financing Facility GDP gross domestic product GER gross enrollment rate GNI gross national income GSFP Ghana School Feeding Program HCI Human Capital Index HTS harmonized test score IGME United Nations Interagency Group for Child Mortality Estimation IHME Institute of Health Metrics and Evaluation ILO International Labour Organization JME Joint Child Malnutrition Estimates (database) JOIN Global Jobs Indicators Database (World Bank) LAYS learning-adjusted years of schooling LLECE Latin American Laboratory for Assessment of the Quality of Education MICS Multiple Indicator Cluster Surveys NER net enrollment rate NHIS National Health Insurance Scheme (Ghana) PASEC Program for the Analysis of Education Systems PILNA Pacific Island Learning and Numeracy Assessment PIRLS Progress in International Reading Literacy Study PISA Programme for International Student Assessment PISA-D PISA for Development PPP purchasing power parity RBF results-based financing RRPS rapid response phone survey XX A bbr ev iations The SACMEQ  Southern and Eastern Africa Consortium for Monitoring Educational Quality SDI service delivery indicators SEBJ share of employment in better jobs SNF specialized nutritious food TIMSS Trends in International Mathematics and Science Study TNER total net enrollment rate UHCI Utilization-Adjusted Human Capital Index UIS United Nations Educational, Scientific, and Cultural Organization’s Institute of Statistics UNICEF United Nations Children’s Fund UNPD United Nations Population Division WHO World Health Organization INTRODUCTION T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 1 T he Human Capital Index (HCI) is an international metric that benchmarks key components of human capital across economies. The HCI was launched in 2018 as part of the Human Capital Project, a global effort to accelerate progress toward a world where all children can achieve their full potential. Measuring the human capital that a child born today can expect to attain by her 18th birthday, the HCI highlights how current health and education outcomes shape the productivity of the next generation of workers. In this way, it underscores the importance for governments and societies of investing in the human capital of their citizens. Over the past decade, many economies have made important progress in improving human capital. Today, however, the COVID-19 (coronavirus) pandemic threatens to reverse many of those gains. Urgent action is needed to protect hard-won advances in human capital, particularly among the poor and vul- nerable. Designing the needed interventions, targeting them to achieve the highest effectiveness, and navigating difficult trade-offs in times of reduced fiscal space make investing in better measurement of human capital now more important than ever. Human capital consists of the knowledge, skills, and health that people accumulate over their lives. People’s health and education have undeniable intrinsic value, and human capital also enables people to realize their potential as productive members of society. More human capital is associated with higher earnings for people, higher income for countries, and stronger cohesion in societies. It is a central driver of sustainable growth and poverty reduction. This report accompanies the release of 2020 data on the HCI. Building on momentum from the first edition in 2018, the 2020 issue updates the index using new and expanded data for each of the HCI components through March 2020. As such, the report provides a snapshot of the state of human capital before COVID-19 and a baseline to track its impact. COVID-19 struck at a time when the world was healthier and more educated than ever. Yet data presented in this report reveal that substantial human capital shortfalls and equity gaps existed before the crisis. Worldwide, a child born just before the advent of COVID-19 could expect to achieve on average just 56 percent of her potential productivity as a future worker. Gaps in human capital remain especially deep in low-income economies and those affected by violence, armed conflict, and institutional fragil- ity. Expanded sex-disaggregated data show that girls currently enjoy a slight edge over boys in human capital accumulation in most economies, reflecting in part a female biological advantage early in life. Women continue to be at a substantial disadvantage, however, in many dimensions of human capital that are not captured by the HCI’s components, including participation in economic life. In addition to describing HCI data and methodology, this report documents the evolution of human capital over the last decade. Human capital outcomes progressed in almost all economies by about 4 percent on average during this period, thanks primarily to better health and increased access to schooling. Many economies, however, struggled to improve learning outcomes, because educational quality often failed to keep pace with gains in enrollment. The various dimensions of human capital improved with economic development, and they did so at a surprisingly similar pace across country groups. Progress was only slightly faster in low-income economies, which are further away from the frontier of full health and education. 2 Introd u ct i on The trajectories of individual economies differed considerably, including in how human capital gains were distributed across the socioeconomic spectrum within each economy. In some contexts, the most disadvantaged groups achieved the greatest gains. In others, poorer and richer families benefited equally. Along with economic development, specific policies contributed to some economies’ progress in human capital. Effective policies included expanding the population coverage of health services, notably for maternal and child health; bolstering nutrition and access to sanitation; making school more affordable; and providing financial support to vulnerable families through mechanisms such as cash transfer pro- grams and insurance. Strong gains were more likely in economies that maintained commitment to reform across political cycles and adopted an evidence-based, whole-of-society approach to policy making. These same elements will be essential to protect human capital in the face of the COVID-19 crisis. Although data on COVID-19’s impacts on human capital outcomes are only beginning to emerge, simulations conducted for this report suggest that school closures combined with family hardship are significantly affecting the accumulation of human capital for the current generation of school-age children. These impacts appear comparable in magnitude to the gains that many economies achieved during the previ- ous decade, suggesting that the pandemic may roll back many years’ worth of human capital progress. In parallel, COVID-19’s disruption of health services, losses in income, and worsened nutrition are expected to increase child mortality and stunting, with effects that will be felt for decades to come. The HCI can be a useful tool to track such losses and guide policy to counter them, because the index is based on robust markers for key stages of human capital accumulation in the growth trajectory of a child. But the five components of the HCI do not cover all the important aspects of the accumulation and productive use of human capital. In particular, the index is silent on the opportunities to use accumulated human capital in adulthood through meaningful work. In many economies, a sizable fraction of today’s young people may not be employed when they become adults. Even if they find employment, they may not hold jobs in which they can use their skills and cognitive abilities to increase their productivity. Recognizing the salience of such patterns for how human capital gains are translated into economic progress and shared prosperity, this report analyzes two measures that augment the HCI to account for the utilization of human capital. These measures provide insight on further margins that economies can explore to boost their long-term growth and productivity. Both utilization measures suggest that human capital is particularly underutilized in middle-income economies. A key message is that human capital is also strikingly underutilized for women in many settings: the gender gap in employment rates (a basic measure of utilization) is 20 percentage points on average worldwide, but it exceeds 40 percentage points in South Asia and in the Middle East and North Africa. By bringing salience to the productivity implications of shortfalls in health and education, the HCI has not only clarified the importance of investing in human capital but also highlighted the role that measurement can play in catalyzing consensus for reform. Better measurement enables policy mak- ers to design effective interventions and target support to those who are most in need, which is often where interventions yield the highest payoffs. Investing in better measurement and data use now is a necessity, not a luxury. In the immediate term, it will guide pandemic containment strategies and sup- port for the most affected. In the medium term, better curation and use of administrative, survey, and identification data will be essential to guide policy choices in an environment of limited fiscal space and competing priorities. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 3 Today, hard-won human capital gains in many economies are at risk. But economies can do more than just work to recover the lost ground. Ambitious, evidence-driven policy measures in health, education, and social protection can pave the way for today’s children to surpass the human capital achievements and quality of life of the generations that preceded them. Protecting and extending earlier human capital gains will require, among others, expanding health ser- vice coverage and quality among marginalized communities, boosting learning outcomes together with school enrollments, and supporting vulnerable families with social protection measures adapted to the scale of the COVID-19 crisis. Informed by rigorous measurement, bold policies can drive a resilient recovery from the pandemic and open a future in which rising generations will be able to develop their full potential and tackle the vast challenges that still lie ahead: from ending poverty to preventing armed conflict to controlling climate change. COVID-19 has underscored the shared vulnerability and common responsibility that today link all nations. Fully realizing the creative promise embodied in each child has never been more important. OVERVIEW T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 5 T he Human Capital Index (HCI) measures the human capital that a child born today can expect to attain by her 18th birthday, given the risks of poor health and poor education prevailing in her country.1 The index incorporates measures of different dimensions of human capital: health (child survival, stunting, and adult survival rates) and the quantity and quality of schooling (expected years of school and international test scores). Human capital has intrinsic value that is undeniably important but difficult to quantify, making it a challenge to combine the different components of human capital into a single measure. The HCI uses global estimates of the economic returns to education and health create an integrated index that captures the expected productivity of a child born today as a future to ­ worker, relative to a benchmark—the same for all countries—of complete education and full health. THE HCI 2020 UPDATE The 2020 update of the HCI incorporates the most recent available data to report scores for 174 econ- omies, 17 more than the 2018 edition. The 2020 update uses new and expanded data for each of the HCI components, available as of March 2020. As in the previous issue, data were obtained from official sources and underwent a careful process of review and curation. Given the timing of data collection, this update can serve as a benchmark of the levels of human capital accumulation that existed immediately before the onset of the COVID-19 (coronavirus) pandemic. Globally, the HCI 2020 shows that, before the pandemic struck, a child could expect to attain an aver- age of 56 percent of her potential productivity as a future worker. This global average masks consider- able variation across regions and economies. For instance, a child born in a low-income economy could expect to be 37 percent as productive as if she had full education and full health. For a child born in a high-income economy, this figure is 70 percent. INCOME ALONE DOES NOT EXPLAIN CROSS-COUNTRY DIFFERENCES IN HUMAN CAPITAL What explains these variations in human capital outcomes? Despite a strong correlation between the HCI and gross domestic product per capita, human capital does not always move in lockstep with eco- nomic development. Economies like Burundi, Estonia, the Kyrgyz Republic, Uzbekistan, and Vietnam have human capital outcomes that are higher than predicted by their gross domestic product per cap- ita. Conversely, in a number of economies, human capital is lower than per capita income would sug- gest. Among these are several resource-rich economies in which human capital development has not yet matched the potential that one would anticipate, given these economies’ wealth. Differences in the quantity and quality of schooling account for the largest part of HCI differences across country income groups. Of the 33-percentage-point difference between the scores of the average low- and high-income economy, almost 25 percentage points are accounted for by the differences in ­ learning-adjusted years of schooling, a measure that combines expected years of school with learning as measured by harmonized test scores (that is, test scores that are made comparable across countries). Although education drives HCI differences across country income groups, education’s contribution to gaps within these groups varies. For instance, education accounts for roughly 90 percent of the difference 6 Overview between high and low performers in the high-income economy group but for only 60 percent within the group of low-income economies. In contrast, differences in child survival rates account for less of the difference in HCI scores among high-income economies. The same is true for health, which explains a lower share of differences in the HCI as one moves from low- to high-income groups, because health outcomes tend to be uniformly better as economies get richer.2 Human capital outcomes also vary for girls and boys. A disaggregation of the HCI by sex—now available for 153 of the 174 economies—shows that on average girls have a slight advantage over boys. Girls are not only catching up to but also outperforming boys in expected years of school and learning outcomes in some regions. For example, in the Middle East and North Africa, girls can expect to complete more than half of an additional learning-adjusted year of schooling compared with boys. In Sub-Saharan Africa and in South Asia, however, the reverse is true. Investing in human capital enhances social cohesion and equity while strengthening people’s trust in institutions. Nowhere is this more important than in countries grappling with fragility and conflict. External shocks such as armed conflict and natural disasters have destructive impacts both on countries’ existing human capital stock and on the process of building new human capital. Evidence increasingly suggests that, for armed conflict as well as famine, these negative effects can persist for decades and even across generations, weakening the core of sustainable and equitable economic development. Unfortunately, yet unsurprisingly, the HCI 2020 indicates that, on average, economies affected by fra- gility, conflict, and violence have lower HCI values compared to the rest of the world. In particular, the seven economies with the lowest HCI 2020 scores are all classified by the World Bank as fragile or conflict affected. This situation adds to the urgency of addressing human capital gaps in such settings. Only by preserving and rebuilding human capital can countries durably escape cycles of fragility and underdevelopment. MOST ECONOMIES ACHIEVED HUMAN CAPITAL GAINS IN THE DECADE BEFORE COVID-19 Because this is the first update of the HCI, the 2020 release presents an opportunity to assess the evo- lution of human capital over time, as measured here by the index in the last decade. The HCI is based on outcomes that typically change slowly from year to year. Some of them—such as stunting and edu- cational test scores—are measured infrequently, every three to five years. As a result, changes in the HCI over a short period are small and may simply reflect updates to some components that are measured sporadically. To provide more reliable insights on economies’ human capital trajectories over time, this report focuses on changes in the index over the past decade. To this end, a (circa) 2010 version of the HCI is constructed with data carefully curated to maximize comparability with the 2020 results. In particular, only those economies for which learning scores were measured by the same international assessment program in 2010 and 2020 enter the comparison.3 The resulting sample for the 2010 HCI includes 103 economies.4 T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 7 As measured by the HCI, human capital progressed in the vast majority of economies in this sample. On average, between 2010 and 2020, the HCI improved by 2.6 percentage points, about 4 percent of its average value in 2010.5 One economy in four that experienced a rise in the index recorded gains of more than five percentage points—a substantial achievement. Economies starting from lower levels of human capital improved by larger amounts. Better health (child and adult survival, and reduction of stunting) accounts for about half of the HCI’s changes. Increased enrollments—especially at preprimary secondary school levels—account for the rest. In contrast, progress on learning outcomes has proved and ­ difficult, because test scores failed to keep pace with enrollment gains in many settings. In the human capital dimensions captured by the index, girls and boys made similar progress over time, with only a handful of economies reporting opposite trends. In the 90 economies for which sex-­ disaggregated data are available and comparisons with 2010 are possible, the average gender ratio is similar in 2010 and 2020, at about 1.06 in favor of girls. Around 2010, the HCI was uniformly larger among ­ to​ girls than among boys, with the exception of seven economies. Over the last decade, the girl-­ -boy ratio improved, approaching or surpassing gender parity, in all of these economies. THE EXTENT TO WHICH DISADVANTAGED HOUSEHOLDS BENEFITED FROM HUMAN CAPITAL GAINS OVER TIME VARIES ACROSS ECONOMIES National averages mask differential trends in human capital between richer and poorer households. Using household data from Demographic and Health Surveys and Multiple Indicator Cluster Surveys, it is possible to calculate a version of the HCI disaggregated by socioeconomic status for a number of low- and middle-income economies. There is substantial variability in how gains in human capital ­ outcomes are distributed across the population.6 For instance, Haiti, Malawi, and Senegal all improved their child survival rates over the last decade; however, the gap between rich and poor households in Haiti remained constant but decreased in Malawi and Senegal.7 Similarly, the years of schooling a child could expect in Bangladesh, Burkina Faso, and India increased significantly. But in Burkina Faso the six-year gap in expected years of school between rich and poor households stayed constant over the past 10 years, whereas, in the same period, Bangladesh and India—although starting from different levels—were able to halve the gap between their richest and poorest households. Côte d’Ivoire’s 25-percentage-point gap between stunting rates for rich and for poor households remained unchanged, notwithstanding a ­ significant average reduction in stunting. Conversely, between 2000 and 2016, Uganda was able to nar- row this gap from a difference of 20 percentage points to a difference of 16 percentage points. Addressing such rich–poor gaps in human capital must remain a priority for governments committed to equitable growth, not least because the returns to investment in human capital are often highest for disadvantaged groups, especially for measures that act early in life. Human capital is a central driver of sustainable growth and poverty reduction. Even for governments that recognize the importance of investing in the human capital of their citizens, however, the benefits of designing policy and building institutions that foster human capital accumulation can take years or even decades to fully materialize. This slow process is evidenced in the relatively modest progress mea- sured for the average economy on the HCI over the past decade. Adopting a longer time frame can help identify many forms of government action that can improve human capital. For that purpose, this report 8 Overview incorporates insights from case studies to better understand the trajectories of economies that have made notable improvements in various dimensions of human capital. Sustained political commitment span- ning election cycles, coordination across the many programs and agencies that may influence human capital, and using a robust evidence base to inform policy choices emerge as key elements contributing to successful policies for human capital.8 HCI SIMULATIONS REVEAL COVID-19’S LARGE IMPACT ON HUMAN CAPITAL COVID-19 threatens countries’ hard-won human capital gains. A lesson from past pandemics and crises is that their effects are not only felt by those directly impacted, but often ripple across populations and, in many cases, across generations. This lesson underscores the urgent need to protect and rebuild human capital to foster recovery in the short and longer terms. Setbacks during certain life stages—chiefly early childhood—can have especially damaging and long-­ lasting effects on human capital accumulation. During childhood, the link between parental income and child health is particularly strong (see Almond 2006). In previous crises, poorer nutrition and reduced well-being among pregnant mothers led to permanent losses in their children’s cognitive attainment, as well as to higher chronic disease rates when the children became adults (see Almond and Currie 2011). In this crisis, human capital impacts associated with economic shocks come atop reductions in care linked to service disruptions during the pandemic’s acute phase. As such, the pandemic, even if transitory, may have repercussions for years to come. Children in disadvantaged families will be disproportionately vul- nerable to all these effects, thus deepening existing inequalities. The HCI methodology can be used to quantify some of the potential impacts of COVID-19 on the future human capital of children and youth. For young children—those born during the pandemic or who are currently under the age of five—disruptions to health systems, reduced access to care, and family income losses will materialize as increased child mortality, malnutrition, and stunting. Because stunting and educational outcomes are closely intertwined, the pandemic risks durably setting back learning. According to HCI-based simulations, in low-income economies, young chil- these children’s ­ dren today can expect their human capital to be up to 1 percent lower than it would have been in the absence of COVID-19. At the height of the pandemic, close to 1.6 billion children worldwide were out of school. For school- age children today, the pandemic has meant that formal teaching and learning no longer happen face to face. Because the ability to roll out distance learning differs across—and even within—economies, considerable losses in schooling and learning can be anticipated. The income shocks associated with ­ COVID-19 will also force many children to drop out of school. Putting these effects together suggests that the pandemic could reduce global average learning-adjusted years of schooling by half a year. Translated into HCI units, this loss means a drop of almost 4.5 percent in the HCI of the current cohort of children. For an economy with an HCI of 0.5, this signifies a drop of 0.0225 or 2.25 HCI points, a reduction of the same order of magnitude as the HCI increase that many economies have achieved over the past decade. Without a strong policy response now, the pandemic’s negative human capital effects will likely continue to reduce economies’ productivity and growth prospects for decades. In 20 years, roughly 46 percent of the typical economy’s workforce (people aged 20 to 65 years) will be composed of individuals who T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 9 were either in school or under the age of 5 during the COVID-19 pandemic. The human capital losses of today’s children will translate into a drop of a full HCI point (0.01) for this future workforce. That is, even if the pandemic is brought under control relatively rapidly, the COVID-19 shock could still leave current cohorts of children behind for the rest of their lives. No society can afford to let that happen. A MEASURE OF UTILIZATION OF HUMAN CAPITAL HIGHLIGHTS SIGNIFICANT GENDER GAPS The HCI is based on reasonably directly measured markers for key stages of human capital in the growth trajectory of a child. The five components of the index, however, do not cover all the important aspects of the accumulation and productive use of human capital. When today’s child becomes a worker in the future, in many countries she may not be able to find a job; even if she can, it might not be a job in which she can fully use her skills and cognitive abilities to increase her productivity. In these cases, her human capital can be considered underutilized. Recognizing the importance of this pattern both for individuals and for policy, this report analyzes two simple extensions of the HCI that adjust it for labor market underutilization of human capital. Both Utilization-Adjusted Human Capital Indexes (UHCIs) can be calculated for more than 160 econ- omies. Both have the same simple form—the HCI multiplied by a utilization rate—and represent the long-run income gains if an economy moves to the frontier where human capital is complete utilized.9 Given their different purposes, the UHCIs are meant to complement, not and completely ­ replace, the HCI. The two UHCIs take different approaches to measuring utilization. In the basic UHCI, utilization is measured as the fraction of the working-age population that is employed. Although this measure is simple and intuitive, it cannot capture the fact that a large share of employment in developing countries is in jobs for which workers may not be able to fully use their human capital to increase their ­productivity. The full UHCI adjusts for this shortcoming by introducing the concept of better employment, which represents the types of jobs (for nonagricultural employees and employers) that are common in high-productivity economies. The full utilization rate depends on the fraction of an economy’s working-age population in better employment. Because they have more human capital to underutilize, economies with higher HCI scores also face larger utilization penalties if they show low rates of better employment. Although the different methodologies produce different scores for some individual economies, the basic and full measures yield broadly similar utilization rates across country income groups and regions, and in general. Utilization rates average about 0.6, but they follow U-shaped curves when plotted against per capita income across economies, and are lowest over a wider range of lower-middle-income economies. The analysis of underutilization suggests that moving to a world with complete human capital and com- plete utilization of that human capital could almost triple long-run per capita incomes. Both UHCIs reveal starkly different gender gaps from those calculated using the HCI. Whereas the HCI is roughly equal for boys and girls, with a slight advantage for girls on average, UHCIs are lower for females percentage than males, driven by lower utilization rates. Basic utilization (employment) rates are 20 ­ points lower for women than for men in general, with a gap of more than 40 percentage points in the Middle East and North Africa and in South Asia. Female employment rates follow strongly U-shaped 10 Overview curves when plotted against economies’ levels of income, whereas male employment rates are much flatter, and with less dispersion across economies. The gender gap is also present in the full utilization rate, though it is smaller. These results suggest that, although gender gaps in human capital in childhood and adolescence have closed in the last two decades (especially for education), major challenges remain to translate these gains into opportunities for women. BETTER MEASUREMENT ENABLES BETTER POLICY As the COVID-19 crisis continues to unfold, data and measurement are more vital than ever to shape governments’ immediate response and to guide future policy choices toward (cost-) effective solutions. Better measurement and data use are investments that pay off, a consideration that is particularly import- ant now as countries face dwindling fiscal space and many competing demands. By generating a shared understanding among diverse actors, measurement can shine a light on con- straints that limit progress in human capital. In the same way, effective measurement can facilitate polit- ical consensus based on facts and help muster support for reforms. Measurement also enables policy makers to target support to those who are most in need, which is often where interventions yield the highest payoffs. As policy implementation moves forward, measurement provides feedback to guide course corrections. In the context of a pandemic, governments that use relevant data in real time are better able to monitor the evolution of disease transmission and continuously update containment strategies, while respond- ing to the immediate and long-term effects of the economic crisis on households and communities. At all times, data are especially important in countries affected by fragility or conflict, though measure- ment is far more difficult in these settings. The HCI offers a high-level view of human capital across economies that can help catalyze new conversations with key stakeholders. At the same time, much greater depth in measurement and research is needed to better understand the dynamics of human capital accumulation, including across socioeconomic groups and geography, and how policies can affect it. Some key measurement improvements—such as leveraging phone surveys and making better use of administrative data—can be achieved in the short term. Other improvements will demand a more sustained effort from economies and development partners. These longer-range efforts include rethinking the architecture of country data systems to connect different administrative data sources and fielding surveys to better understand the needs and behavior of teachers and health providers. The COVID-19 crisis threatens gains in human capital that countries have achieved through decades of effort. A renewed, society-wide commitment is needed to protect human capital in the short run and to remediate the looming losses in the longer run. Challenges range from crafting context-sensitive school reopening protocols to deeper reforms that will promote children’s learning at all stages: starting from cognitive stimulation in the early years, then continuing to nurture relevant skills throughout childhood and adolescence. Building blocks for success will include better-prepared teachers, better-managed schools, and incentives that are aligned across the many stakeholders in education reform. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 11 Support to households will be essential not only to buffer income losses but also to sustain the demand side of schooling and health care. Such support can come through cash transfers and through interventions aimed at reconnecting workers to jobs. Strengthening disease surveillance and a renewed commitment to universal health coverage will be critical to build resilient health systems that offer affordable, quality care to all. Investments in water, sanitation, and—increasingly—digitalization are important complements to sustain human capital accumulation. Current deepening inequalities in human capital outcomes make it imperative to target interventions to children from the most disadvantaged families. With fiscal space shrinking as competing priorities multiply, policy makers face hard choices. Proven strategies include engaging the whole of society, identifying cross-sectoral synergies, and using data to select cost-effective interventions and track their effective implementation. These approaches will not make tough policy trade-offs painless. But they will enable leaders to choose the options that have the highest probability of success. Applying these tools, governments can go far toward protecting and rebuilding human capital in the wake of COVID-19. And that is not all. Strong, evidence-driven human capital investments now can do far more than restore what has been lost. Health, education, social protec- tion, and other complementary policies informed by rigorous measurement can take countries’ human capital beyond the levels previously achieved, opening the way to a more prosperous and inclusive future. NOTES 1. The HCI was introduced in World Bank (2018a, 2018b), and the methodology of the HCI is detailed in Kraay (2018). 2. Stunting and adult survival are here considered together for easy comparison. 3. As described in chapter 2, this rule is relaxed for only five economies, which are included in the sam- ple with learning scores from different international assessments (Trends in International Mathemat- ics and Science Study [TIMMS]/Progress in International Reading Literacy Study [PIRLS] in 2010 and the Programme for International Student Assessment [PISA] in 2020). To increase comparability, only scores for secondary schooling are considered for 2010. 4. This sample is, unsurprisingly, skewed toward richer economies for which data tend to be more com- plete and of better quality. 5. Richer economies are closer to the frontier of full schooling and health and would naturally display slower change in their human capital. With the 2010–20 sample skewed toward richer economies, the human capital pace of change is likely underestimated. 6. The analysis of HCI outcomes disaggregated by socioeconomic status is based on D’Souza, Gatti, and Kraay (2019). 7. It is important to note the dramatic increase in child mortality that occurred in Haiti in 2010 in the aftermath of the country’s catastrophic January 2010 earthquake. 8. This approach informs the work of the World Bank’s Human Capital Project (HCP). 9. Specifically, long-run gross domestic product per capita is 1/UHCI times higher in a world with com- plete human capital and complete utilization than under the status quo. This rate is a generalization of the interpretation of the HCI. See Pennings (2020) for details. 12 Overview REFERENCES Almond, D. 2006. “Is the 1918 Influenza Pandemic Over? Long-Term Effects of In Utero Influenza Exposure in the Post-1940 US Population.” Journal of Political Economy 114 (4): 672–712. Almond, D., and J. Currie. 2011. “Killing Me Softly: The Fetal Origins Hypothesis.” Journal of Economic Perspectives 25 (3): 153–72. D’Souza, R., R. Gatti, and A. Kraay. 2019. “A Socioeconomic Disaggregation of the World Bank Human Capital Index.” Policy Research Working Paper 9020, World Bank, Washington, DC. Kraay, A. 2018. “Methodology for a World Bank Human Capital Index.” Policy Research Working Paper 8593, World Bank, Washington, DC. Pennings, S. 2020. “The Utilization-Adjusted Human Capital Index (UHCI).” Policy Research Working Paper 9375, World Bank, Washington, DC. World Bank. 2018a. World Development Report 2018: Learning to Realize Education’s Promise. Washington, DC: World Bank. World Bank. 2018b. “The Human Capital Project.” World Bank, Washington, DC. 1 The Human Capital Index 2020 UPDATE T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 15 A t the organization’s 2018 Annual Meetings, in future worker productivity deriving from gaps the World Bank Group launched the in human capital across countries, the HCI under- Human Capital Project, an unprecedented scores the urgency of improving human capital global effort to support human capital develop- outcomes for children today. ment as a core element of countries’ overall strat- egies to increase productivity and growth. The In response to the call for governments to invest main objective of the project was rapid progress in the human capital of their citizens, 78 econo- toward a world in which all children can achieve mies across the world are now part of the Human their full potential. For that to happen, children Capital Project. These economies have affirmed need to reach school well-nourished and ready building, protecting, and employing human cap- to learn, attain real learning in the classroom, and ital as a national priority and have undertaken enter the job market as healthy, skilled, and pro- difficult reforms, sometimes in very challenging ductive adults. contexts. With a view to maintaining this momen- tum, the 2020 update of the HCI incorporates Central to this effort has been the Human Capital the most recent data to report HCI scores for 174 Index (HCI), a cross-country metric measuring the economies, adding 17 new economies to the index human capital that a child born today can expect relative to the 2018 edition. to attain by her 18th birthday, given the risks of poor health and poor education prevailing in her This update uses new and expanded data for country.1 The HCI brings together measures of dif- each of the HCI components, with a cut-off date ferent dimensions of human capital: health (child of March 2020. Computed using data collected survival, stunting, and adult survival rates) and before COVID-19 (coronavirus) had impact on a the quantity and quality of schooling (expected global scale, the HCI 2020 provides a useful bench- years of school and international test scores). mark to track the evolution of human capital and Using estimates of the economic returns to edu- its key components in the wake of the pandemic. cation and health, the components are combined into an index that captures the expected produc- The next three sections of this chapter outline tivity of a child born today as a future worker, rel- the HCI methodology and describe the main ative to a benchmark of complete education and features of the HCI 2020 and its components. The full health. subsequent sections discuss gender differences across countries and regions, and highlight The HCI ranges from 0 to 1, so that an HCI value the unique human capital challenges that arise of, for instance, 0.5 implies that a child born today in states grappling with fragility, conflict, and will be only half as productive as a future worker violence. The final section provides the HCI 2020 as she would be if she enjoyed complete educa- scores for 174 economies and an explanation of the tion and full health. By benchmarking shortfalls discontinued use of rankings. 16 The H uman   Capital Index 20 20 U pdate 1.1  THE HCI METHODOLOGY next generation, rather than measuring the stock of human capital of the current workforce, which The HCI is designed to highlight how improve- largely reflects policy choices made decades ago, ments in current health and education outcomes when the current workforce was of school age.2 shape the productivity of the next generation of The resulting HCI quantitatively illustrates the workers, assuming that children born today expe- key stages in a child’s human capital trajectory and rience over the next 18 years the same educational their consequences for the productivity of the next opportunities and health risks as children cur- generation of workers, with three components: rently in this age range. Component 1: Survival from birth to school age, The HCI captures key stages of a child’s trajectory measured using under-5 mortality rates. from birth to adulthood. In the poorest countries in the world, there is a significant risk that a child Component 2: Expected years of learning-adjusted will not survive to her fifth birthday. Even if she school, combining information on the quantity does reach school age, there is a further risk that and quality of education. The quantity of educa- she will not start school, let alone complete the tion is measured as the number of years of school full cycle of 14 years of schooling, from preschool a child can expect to obtain by age 18 given the pre- to grade 12, which is the norm in high-income vailing pattern of enrollment rates across grades. countries. The time she does spend in school may The quality of education reflects work undertaken translate unevenly into learning, depending on a at the World Bank to harmonize test scores from variety of factors including the quality of teachers major international student achievement testing and schools that she experiences. When she turns programs (Patrinos and Angrist 2018). These 18, she carries with her the lasting effects of poor two measures are combined into a measure of health and nutrition during childhood that limit learning-adjusted years of schooling as proposed her physical and cognitive abilities as she moves in Filmer et al. (2018) (see box 1.1). into adulthood. Component 3: Health. In the absence of a single Several criteria have guided the design of the broadly accepted, directly measured, and widely HCI. First, the HCI is outcome- rather than available metric, the overall health environment inputs-based, focusing the conversation on what is captured by two proxies: (1) adult survival rates, matters—results. This focus provides incentives for defined as the fraction of 15-year-olds who survive countries not only to invest more but also to invest until age 60, and (2) the rate of stunting for children better in human capital, without concerns that the under age 5. Adult survival rates can be interpreted HCI might be susceptible to gaming. The likeli- as a proxy for the range of fatal and nonfatal health hood that a cross-country benchmarking exercise outcomes that a child born today would experi- can spur policy action is strongly influenced by the ence as an adult, if current conditions prevail into over-time and cross-country coverage of the met- the future. Stunting is broadly accepted as a proxy ric. Aiming for good coverage limits the choice of for the prenatal, infant, and early childhood health components to data that are systematically col- environments, and so summarizes the risks to good lected for a large number of economies over time. health that children born today are likely to expe- Further, for the index to promote change, the rience in their early years—with important conse- components of the HCI should be responsive to quences for health and well-being in adulthood. policy action in the short to medium term. The need to produce such a metric has oriented the The health and education components of human index toward measuring the human capital of the capital have intrinsic value that is undeniably T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 17 Box 1.1: Learning-adjusted years of schooling The knowledge and skills that an individual acquires through schooling form an important part of her human capital. The standard summary measure for education used in aggregate-level contexts—the average number of years of schooling in a population—is an imprecise proxy for education, however, because a given number of years in school leads to much more learning in some settings than in others. As recent research shows, students in different countries who have completed the same number of years of school often have vastly different learning outcomes.a Learning-adjusted years of schooling (LAYS), a measure described in Filmer et al. (2018), addresses this concern by combining information on the quantity and quality of schooling into a single easy-to-understand metric of progress. It is calculated as the product of average years of school and a particular measure of learning relative to a numeraire: LAYSc = Sc × R n   c (B1.1.1) where Sc is a measure of the average years of schooling acquired by a relevant cohort of the pop- ​ , and Rn ulation of country c c c, is a measure of learning for a relevant cohort of students in country ​ relative to a numeraire (or benchmark). For the Human Capital Index, expected years of school, EYS, measures the quantity of education. Harmonized test scores, HTS, from the 2020 update of the Global Dataset on Education Quality, provide information on education quality relative to a benchmark score of 625, which corresponds to the Trends in International Mathematics and Science Study (TIMSS) standard of advanced achievement: HTSc LAYSc = EYSc × (B1.1.2) 625 By adjusting years of school for quality, LAYS reflects the reality that children in some countries learn far less than those in other countries, despite being in school for a similar amount of time. The simplicity and transparency of its construction make LAYS a compelling summary measure of education to use in policy dialogue.b Filmer et al. (2018) also find that LAYS improves upon the standard metric of average years of schooling as a predictor of economic growth. Source: Filmer et al. 2018. a. In Nigeria, for example, 19 percent of young adults who have completed primary education are able to read; by contrast, 80 percent of Tanzanians with the same level of schooling are literate (Kaffenberger and Pritchett 2017, as reproduced in World Bank 2018d). b. Like all aggregate measures, LAYS should be used with caution. Because there are standard errors around test measures, any LAYS measure will also have some error band around it. This means that it is important not to overinterpret small cross-country differences or small changes over time. important but difficult to quantify. This makes it translate them into contributions to worker challenging to combine the different components productivity, relative to a benchmark of complete into a single index. Rather than relying on ad education and full health (see box 1.2).3 The hoc aggregation with arbitrary weights, the HCI resulting index ranges between 0 and 1. A country uses the estimated earnings associated with in which a child born today can expect to achieve an additional unit of health and education to full health (no stunting and 100 percent adult 18 The H uman   Capital Index 20 20 U pdate Table 1.1: Human Capital Index 2020, averages by World Bank region Europe Latin Middle East and America East and Sub- Asia and Central and North North South Saharan Indicator Pacific Asia Caribbean Africa America Asia Africa HCI Component 1: Survival Probability of Survival to Age 5 0.98 0.99 0.98 0.98 0.99 0.96 0.93 HCI Component 2: School Expected Years of School 11.9 13.1 12.1 11.6 13.3 10.8 8.3 Harmonized Test Scores 432 479 405 407 523 374 374 HCI Component 3: Health Survival Rate from Age 15 to 60 0.86 0.90 0.86 0.91 0.91 0.84 0.74 Fraction of Children Under 5 Not 0.76 0.90 0.85 0.82 — 0.69 0.69 Stunted Human Capital Index (HCI) 2020 0.59 0.69 0.56 0.57 0.75 0.48 0.40 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The table reports averages of the index components and the overall Human Capital Index (HCI) by World Bank Group regions. — = not available. survival) and full education potential (14 years upper bounds for the estimates are reported. of high-quality school by age 18) would score a Unlike the HCI 2018 launch, economies’ rankings value of 1. Therefore, a score of 0.70 indicates are not reported, for reasons that are detailed in that the productivity as a future worker of a child box 1.6. born today is 30 percent below what could have been achieved with complete education and full The sobering reality is that, as measured by the HCI health. Because the theoretical underpinnings 2020, worldwide, a child born today would expect of the HCI are in the development accounting to achieve on average only 56 percent of her full literature, the index is linked to real differences productivity as a future worker. And this estimate in how much income a country can generate in does not account for any impact that may have the long run (see box 1.3 for limitations of the resulted from the COVID-19 pandemic. Clearly HCI). If a country has a score of 0.50, then the there is considerable heterogeneity around the gross domestic product (GDP) per worker could 56 percent figure. Importantly, the HCI is lower be twice as high if the country reached the in low-income economies than in high-income benchmark of complete education and full health economies by a substantial margin. In the poor- (see appendix A for a detailed discussion of the est economies in the world, a child born today will HCI methodology). grow up to be only 30 percent as productive as she could be; in the richest economies, the corre- sponding figure is 80 percent or more (see figure 1.2  THE HCI 2020 1.1, which plots the HCI 2020 on the vertical axis against log GDP per capita at purchasing power The HCI 2020 scores for 174 economies are parity on the horizontal axis). Compared to a child reported in the final section of this chapter (see in Europe and Central Asia, a child born in Sub- table 1.2). Economies’ scores are sorted from low- Saharan Africa can expect to be only 58 percent as est to highest. Next to the HCI score, lower and productive (see table 1.1). T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 19 Box 1.2: The Human Capital Index’s aggregation methodology The components of the Human Capital Index (HCI) are combined into a single index by first con- verting them into contributions to productivity relative to a benchmark of complete education and full health. Multiplying these contributions to productivity together gives the overall HCI: HCI = Survival × School × Health (B1.2.1) In the case of survival, the relative productivity interpretation is stark: children who do not survive childhood never become productive adults. As a result, expected productivity as a future worker of a child born today is reduced by a factor equal to the survival rate, relative to the benchmark in which all children survive. 1 – Under-5 Mortality Rate Survival = (B1.2.2) 1 The benchmark of complete high-quality education corresponds to 14 years of school and a har- monized test score of 625. The relative productivity interpretation for education is anchored in the large empirical literature measuring the returns to education at the individual level. A rough consensus from this literature is that an additional year of school raises earnings by about 8 per- cent. The parameter ​ ϕ​= 0.08 measures the returns to an additional year of school and is used to convert differences in learning-adjusted years of school across countries into differences in worker productivity. Harmonized Test Score – 14) School = eϕ (Expected Years of School × 625  (B1.2.3) Compared with a benchmark in which all children obtain a full 14 years of school by age 18, a child who obtains only 10 years of education can expect to be 32 percent less productive as an adult (a gap of 4 years of education, multiplied by 8 percent per year). In the case of health, the relative productivity interpretation is based on the empirical literature measuring the economic returns to better health at the individual level. The key challenge in this literature is the lack of any unique, directly measured summary indicator of the various aspects of health that matter for productivity. This microeconometric literature often uses proxy indica- tors for health, such as adult height, because adult height can be measured directly and reflects the accumulation of shocks to health through childhood and adolescence. A rough consensus drawn from this literature is that an improvement in health associated with a one-centimeter increase in adult height raises productivity by 3.4 percent. Converting this evidence on the returns to one proxy for health (adult height) into the other proxies for health used in the HCI (stunting and adult survival) requires information on the rela- tionships between these different proxies: • For stunting, a direct relationship exists between stunting in childhood and future adult height, because growth deficits in childhood persist to a large extent into adulthood, together with the associated health and cognitive deficits. Available evidence suggests that a reduction in stunting rates of 10 percentage points increases attained adult height by approximately one centimeter, which increases productivity by (10.2 × 0.1 × 3.4) percent, or 3.5 percent. (continued next page) 20 The H uman   Capital Index 20 20 U pdate Box 1.2: The Human Capital Index’s aggregation methodology (Continued) • For adult survival, the empirical evidence suggests that, if overall health improves, both adult height and adult survival rates increase in such a way that adult height rises by 1.9 centimeters for every 10-percentage-point improvement in adult survival. This implies that an improvement in health that leads to an increase in adult survival rates of 10 percentage points is associated with an improvement in worker productivity of (1.9 × 3.4) percent, or 6.5 percent. In the HCI, the estimated contributions of health to worker productivity based on these two alter- native proxies are averaged together, if both are available, and are used individually if only one of the two is available. The contribution of health to productivity is expressed relative to the bench- mark of full health, defined as the absence of stunting, and a 100 percent adult survival rate. Health = e(γASR × (Adult Survival Rate – 1) + γStunting × (Not Stunted Rate – 1))/2 (B1.2.4) For example, compared with a benchmark of no stunting, in a country where the stunting rate is 30 percent, poor health reduces worker productivity by (30 × 0.34) percent, or 10.2 percent. Compared with the benchmark of 100 percent adult survival, poor health reduces worker produc- tivity by (30 × 0.65) percent, or 19.5 percent, in a country where the adult survival rate is 70 per- cent. The average of the two estimates of the effect of health on productivity is used in the HCI. These parameters used to convert the components of the index into their contributions to pro- ϕ​ ductivity (​ = 0.08 for school, γ ​  ​​  ​= 0.65 for adult survival, and γ ​ ASR ​  ​ Stunting​​​= 0.35 for stunting) serve as weights in the construction of the HCI. The weights are chosen to be the same across coun- tries, so that cross-country differences in the HCI reflect only cross-country differences in the component variables. This facilitates the interpretation of the index. This is also a pragmatic choice, because estimating country-specific returns to education and health for all countries included in the HCI is not feasible. Source: Kraay (2018). Despite the high correlation between the HCI and the poor have, the World Bank’s twin goals of pro- GDP per capita, some economies perform signifi- moting shared prosperity and eradicating extreme cantly better than their income levels might suggest. poverty are unlikely to be met without human cap- These economies include Estonia, Kyrgyz Republic, ital improvements. The world’s extreme poor are Vietnam, and West Bank and Gaza. Conversely, in a disproportionately found in economies with the number of economies, human capital is lower than lowest HCI scores; 30 percent of the world’s poor per capita income would suggest. Among them are reside in the 10 economies with the lowest HCI a few resource-rich economies, where human capi- values, although these 10 economies are home to tal has not yet matched the potential that one would only 5 percent of the total global population (fig- envisage given these economies’ development. ure 1.2). In fact, 80 percent of the world’s extreme poor reside in economies with an HCI value under The correlation between poverty and low HCI 0.5. If prosperity is to be shared, growth must be scores is also high. Given that better education and inclusive for those at the bottom of the distri- health translate to improved productivity for peo- bution, and inclusive growth necessitates strong ple, and that human capital is often the only asset investments in human capital. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 21 Figure 1.1: The Human Capital Index 2020 1.0 Japan 0.8 Korea, Rep. Estonia Belarus HCI, circa 2020 Vietnam Luxembourg Qatar Uzbekistan 0.6 Kyrgyz Republic Saudi Arabia West Bank and Gaza Kuwait Panama Botswana Iraq 0.4 Eswatini 0.2 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI) for HCI data and the World Development Indicators and Penn World Tables 9.1 for per capita GDP data. Note: The figure plots country-level HCI on the y-axis and GDP per capita in PPP on the x-axis, in constant 2011 dollars, for most recently available data as of 2019. Per capita GDP data for South Sudan are not available. The dashed line illustrates the fitted regression line between GDP per capita and the HCI 2020. Scatter points above (below) the fitted regression line illustrate economies that perform better (worse) in the HCI than their level of GDP would predict. Economies above the 95th and below the 5th percentile in distance to the regression fitted line are labeled. HCI = Human Capital Index; PPP = purchasing power parity. Figure 1.2: Concentration of the extreme poor in economies sorted by their Human Capital Index scores 1.0 Cumulative share of the world’s poor 0.8 80% of the world’s poor reside in economies with an HCI under 0.5 (US$1.9 2011 PPP) 0.6 0.4 30% of the world’s poor reside in the bottom-10 economies sorted by HCI 0.2 0 0.2 0.4 0.6 0.8 1.0 Share of world population, economies sorted by their HCI 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Poverty values come from Corral et al. (2020) and are calculated before COVID-19. Note: The horizontal axis represents the share of the global population accounted for by the countries sorted by their HCI value. HCI = Human Capital Index; PPP = purchasing power parity 22 The H uman   Capital Index 20 20 U pdate Two elements help explain how different dimen­ 90 percent among high-income to 60 percent sions of human capital contribute to differences among low-income economies. In contrast, in the HCI scores. The first are the weights of the differences in child survival rates account for health and education components of the HCI, less of the difference in HCI scores among high- reflecting the empirical literature on the contribu- income economies, largely because economies tion of health and education to earnings (box 1.2 in this group are close to universal child survival. and appendix A). Second, the components have The same is true for the health component, with different distributions, globally and by country stunting and adult survival taken together for income groupings, according to the World Bank’s easy comparison. Health differences explain a most recent classification. For example, the vari- lower share of HCI differences as one moves ation of child survival is nine times larger among from low- to high-income economies, because low-income than among high-income econo- health outcomes tend to be uniformly better as mies, where child survival is uniformly close to economies get richer.5 These results reflect the 100 percent (figure 1.3). fact that, within the high-income group, values for health and survival components in most A simple decomposition exercise can help account economies are close to the frontier, whereas there for differences in the HCI across country income is still considerable variation in test scores (see the 4 groups. Consider the HCI difference between the box plots in figure 1.3). typical low-income and high-income economy, which is about 0.33 (figure 1.4). Of these 33 HCI Gaps in human capital outcomes between rich points, almost 25 are accounted for by the differ- and poor people within economies can be quite ences in expected years of school (EYS) and har- large. A socioeconomic disaggregation of the HCI, monized test scores. Overall, differences in the constructed using comparable survey data for 50 quality and quantity of schooling account for the low- and middle-income economies, reveals that largest share of index differences across country differences across socioeconomic quintiles within income groups, ranging from 65 to 85 percent. economies account for nearly one-third of the total variation in human capital (D’Souza, Gatti, There is also considerable heterogeneity within and Kraay 2019). Outcomes can also vary across country income groups, and the difference in rural-urban status, as in the case of Romania. HCI between the economy with the lowest HCI Some of that country’s counties have urban areas and the economy with the highest HCI in each with learning outcomes as high as top perform- income group rivals the difference between ers in Europe, whereas some rural areas rank at income groups and, in some cases, exceeds it. For par with economies in the bottom third of the example, the difference in the HCI between the HCI distribution (World Bank 2020a). Some of top and bottom performers among high-income these within-country differences align with eth- economies is roughly 0.38, or 38 HCI points, nic divides. For example, in Vietnam, survey which compares with a difference of 33 points data from 2014 disaggregated by ethnic group between the average HCI values of high- and show that ethnic minorities have an HCI score of low-income economies. Overall, both within and 0.62, compared with a score of 0.75 for the eth- across all groups, education still accounts for the nic majority. At 32 percent, stunting rates are two largest share of the differences observed between times larger among ethnic minorities than among top and bottom performers (figure 1.5); however, the majority. School enrollment also lags among education accounts for a smaller share as one ethnic minorities relative to their majority peers moves down income groups, falling from roughly by 30 percentage points (World Bank 2019b). T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 23 Box 1.3: Limitations of the Human Capital Index Like all cross-country benchmarking exercises, the Human Capital Index (HCI) has limitations. Components of the HCI such as stunting and test scores are measured only infrequently in some economies and not at all in others. Data on test scores come from different international testing programs and need to be converted into common units, and the age of test-takers and the sub- jects covered vary across testing programs. Moreover, test scores may not accurately reflect the quality of the whole education system in an economy, to the extent that test-takers are not repre- sentative of the population of all students. Reliable measures of the quality of tertiary education that are comparable across most economies of the world do not yet exist, despite the impor- tance of higher education for human capital in a rapidly changing world. The data on enrollment rates needed to estimate expected years of school often have many gaps and are reported with significant lags. Socioemotional skills are not explicitly captured. In terms of health, child and adult survival rates are imprecisely estimated in economies where vital registries are incomplete or nonexistent. These limitations have implications not only for the construction of the 2020 update but also for the comparison of the index over time. One objective of the HCI is to call attention to these data shortcomings and to galvanize action to remedy them. Improving data will take time. In the interim and in recognition of these lim- itations, the HCI should be interpreted with caution. The HCI provides rough estimates of how current education and health will shape the productivity of future workers, but it is not a finely graduated measurement that can distinguish small differences between economies. Naturally, because the HCI captures outcomes, it is not a checklist of policy actions, and the proper type and scale of interventions to build human capital will be different in different economies. Although the HCI combines education and health into a single measure, it is too blunt a tool to inform the  cost-effectiveness of policy interventions in these areas, which should instead be assessed through careful cost-benefit analysis and impact assessments of specific programs. Because the HCI uses common estimates of the economic returns to health and education for all economies, it does not capture cross-country differences in how well economies are able to productively deploy the human capital they have. Finally, the HCI is not a measure of welfare, nor is it a summary of the intrinsic values of health and education; rather, it is simply a measure of the contribution of current health and education outcomes to the productivity of future workers. 1.3  HCI 2020: INDEX COMPONENTS Child survival The probability of survival to age 5 is calculated 1.3.1  HCI components and data sources as the complement of the under-5 mortality The components of the HCI are built using pub- rate. The under-5 mortality rate is the probabil- licly available official data, primarily from admin- ity that a child born in a specified year will die istrative sources. The data are subject to a careful before reaching the age of 5 if subject to cur- vetting process with World Bank country teams rent age-specific mortality rates. It is frequently and, at the discretion of country teams, with line expressed as a rate per 1,000 live births, in ministry counterparts. These data and the relevant which case it must be divided by 1,000 to obtain definitions are described in the text that follows the probability of dying before age 5. Under-5 and in more in detail in appendix C. mortality rates are calculated by the United 24 The H uman   Capital Index 20 20 U pdate Figure 1.3: Human Capital Index 2020 components, distribution by country income group A. Probability of survival to age 5 B. Expected years of school 1.00 14 Probability of survival to age 5, circa 2020 Seychelles Panama 12 Expected years of school, Nauru Mauritius Nauru Palau Palau 0.95 Trinidad Panama circa 2020 and 10 Tobago Gabon 8 Botswana 0.90 Iraq Nigeria 6 0.85 4 C. Harmonized test scores D. Fraction of children under 5 not stunted 600 1.0 Fraction of children under 5 not stunted, Harmonized test scores, circa 2020 Vietnam 500 0.8 circa 2020 Marshall Islands 400 0.6 Guatemala Nauru 300 0.4 E. Adult survival rate 1.0 Adult survival rate, circa 2020 0.9 Low-income economies Lower-middle-income economies 0.8 Upper-middle-income economies Namibia High-income economies 0.7 Marshall Islands South Africa Sierra Leone 0.6 Eswatini Central African Republic Lesotho 0.5 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: Each box spans the interquartile range with the upper and lower end of the boxes illustrating the 25th and 75th percentile values. The horizontal lines in the inner boxes represent the median value. Outer horizontal lines show maximum and minimum values excluding outliers. Thinner box plots indicate less dispersion in values. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 25 Figure 1.4: Decomposition of observed mean HCI differences between selected country income groups Upper-middle- vs. high-income economies Lower-middle- vs. high-income economies Low- vs. high-income economies 0 0.1 0.2 0.3 0.4 Contribution to observed di erence in HCI Child survival Expected years of school Harmonized test scores Health Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the contribution to observed HCI differences between country income groups. Figure 1.5: Differences between the top and bottom Human Capital Index performers within each country income group High-income economies Upper-middle-income economies Lower-middle-income economies Low-income economies 0 0.2 0.4 0.6 0.8 1.0 Share of di erence between top and bottom performer explained by each component Child survival Expected years of school Harmonized test scores Health Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the share of the observed HCI differences between selected economies by component. Comparison economies for the high-income group are Singapore and Panama; for the upper-middle-income group, Belarus and Iraq; for lower-middle-income economies, Nigeria and Vietnam; and, for low-income economies, the Central African Republic and Tajikistan. Nations Interagency Group for Child Mortality Expected years of school Estimation (IGME) using mortality as recorded The EYS component of the HCI captures the in household surveys and vital registries. For the number of years of school a child born today can 2020 update of the HCI, under-5 mortality rates expect to obtain by age 18, given the prevailing come from the September 2019 update of the pattern of enrollment rates in her economy. IGME estimates and are available on the IGME Conceptually, EYS is the sum of enrollment rates website. 6 by age from ages 4 to 17. Because age-specific 26 The H uman   Capital Index 20 20 U pdate enrollment rates are neither broadly nor and the Pacific Island Learning and Numeracy systematically available, data on enrollment Assessment (PILNA). It also incorporates Early rates by level of school are used to approximate Grade Reading Assessments (EGRAs) coordinated enrollment rates in different age brackets. by the United States Agency for International Preprimary enrollment rates approximate the Development. The 2020 update of the Global enrollment rates for 4- and 5-year-olds, primary Dataset on Education Quality extends the data- enrollment rates approximate the rates for 6- to base to 184 economies from 2000 to 2019, draw- 11-year-olds, lower-secondary rates approximate ing on a large-scale effort by the World Bank to for 12- to 14-year-olds, and upper-secondary collect global learning data. Updates to the data- rates approximate for 15- to 17-year-olds. Cross- base come from new data from PISA 2018, PISA country definitions in school starting ages and for Development (PISA-D),9 PILNA, and EGRA. the duration of the various levels of school imply The database adds 20 new economies,10 bringing that these rates will only be approximations the percentage of the global school-age popula- of the number of years of school a child can tion represented by the database to 98.7 percent. expect to complete by age 18. Enrollment rates In addition, more recent data points have been for 2020 for each school level and for different added for 94 economies.11 Since the launch of enrollment rate types are obtained from the the HCI in 2018, a complementary measure has United Nations Educational, Scientific and been created to address foundational skills and to Cultural Organization’s Institute for Statistics.7 help economies prioritize their response to HCI These data are then complemented with inputs and learning-adjusted years of schooling (LAYS) from World Bank teams working on specific scores: Learning Poverty represents the share of countries to validate the data and provide more 10-year-olds who cannot read and understand a recent values when available.8 simple text (see box 1.4). The correlation between Learning Poverty and LAYS is high, in the range of Harmonized test scores −0.90. The Learning Poverty measure is available The school quality indicator is based on a large- for 113 of the economies in the HCI 2020. scale effort to harmonize international student achievement tests from several multicountry test- Fraction of children under 5 not stunted ing programs to produce the Global Dataset on The fraction of children under 5 not stunted is Education Quality. A detailed description of the calculated as the complement of the under-5 test score harmonization exercise is provided in stunting rate. The stunting rate is defined as the Patrinos and Angrist (2018), and the HCI draws on share of children under the age of 5 whose height an updated version of this dataset as of January is more than two reference standard deviations 2020. The dataset harmonizes scores from three below the median for their ages. The median and major international testing programs: the Trends standard deviations are set by the World Health in International Mathematics and Science Study Organization (WHO) for normal healthy child (TIMSS), the Progress in International Reading development (World Health Organization 2009). Literacy Study (PIRLS), and the Programme for Child-level stunting prevalence is averaged across International Student Assessment (PISA). It further the relevant 0–5 age range to arrive at an overall includes four major regional testing programs: under-5 stunting rate. The stunting rate is used, the Southern and Eastern Africa Consortium for in addition to the adult survival rate, as a proxy Monitoring Educational Quality (SACMEQ), the for latent health of the population in economies Program for the Analysis of Education Systems where stunting data are available, as discussed in (PASEC), the Latin American Laboratory for the next section. Stunting rates for this edition of Assessment of the Quality of Education (LLECE), the HCI come from the March 2020 update of the T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 27 Box 1.4: Measuring Learning Poverty The World Bank collaborated with the United Nations Educational, Scientific and Cultural Organization Institute for Statistics (UIS) to create a measure of Learning Poverty—the share of 10-year-olds who cannot read and understand a simple text. According to this measure, the World Bank estimates that 53 percent of children in low- and middle-income economies suffer from Learning Poverty. In the poorest economies, the share is often more than 80 percent. Such high levels of Learning Poverty are an early warning sign that the learning-adjusted years of schooling (LAYS) indicator, which measures quantity and quality of education that 18-year-olds have benefited from, will be unacceptably low for that cohort of children. In better-performing systems, virtually all children learn to read with comprehension by age 10. Although it may take decades to build up the high-quality education systems that lead to the highest scores on the LAYS indicator of the Human Capital Index (HCI), teaching children to reach a minimum proficiency in reading requires much less time. Why measure reading? Children need to learn to read so that they can read to learn. Those who do not become proficient in reading by the end of primary school often cannot catch up later, because the curriculum of every school system assumes that secondary school students can learn through reading. Reading is, in other words, a gateway to all types of academic learning. This is not to say that reading is the only skill that matters. Reading proficiency can serve as a proxy or warning indicator for foundational learning in other areas that are also essential, like mathematics and reasoning abilities. Education systems that enable all children to read are likely to succeed in helping them learn other subjects as well. Across economies and schools, the data show that proficiency rates in reading are highly correlated with proficiency in other subjects. How is Learning Poverty calculated? Conceptually similar to the LAYS indicator in the HCI for youth, the Learning Poverty measure combines learning with enrollment, to emphasize the importance of learning for all children and not just those currently in school. The learning com- ponent captures enrolled students who cannot read with comprehension, whereas the participa- tion component corresponds to the out-of-school rate. “Reading with comprehension” is defined here as reaching the global minimum proficiency in literacy. UIS leads the Global Alliance to Monitor Learning (GAML), which agreed to a common definition of minimum proficiency in liter- acy for the purposes of monitoring Sustainable Development Goal 4. With this definition, several cross-national and some national assessments were harmonized by applying GAML’s definition of reading proficiency as a common benchmark. Unlike the HCI, Learning Poverty relies only on assessments targeting children from grades 4 to 6. For each assessment incorporated into the database, the harmonization process looks at the definitions of each level of proficiency for that exam and selects the one that maps most clearly to the GAML definition. The harmonization pro- cess allows much greater coverage of countries than does relying on a single assessment like the Progress in International Reading Literacy Study (PIRLS)—an excellent assessment for mea- suring Learning Poverty, but one in which relatively few low- and middle-income countries partic- ipate. The high correlation between students’ performance on different assessments increased confidence that this harmonization method is valid. Once the share of children below minimum proficiency is calculated, the final step in calculating Learning Poverty is to adjust this share for out-of-school children of primary school age who are considered nonproficient in reading. The HCI, LAYS, and Learning Poverty, each with its own unique mandate and methodology, are synthetic indicators intended to build political commitment and galvanize action. Source: World Bank Education Global Practice, World Bank (2019a). 28 The H uman   Capital Index 20 20 U pdate Joint Malnutrition Estimates ( JME) database (see economies continue to face. Child survival rates UNICEF, WHO, and World Bank Group 2020). also vary significantly by region, with economies This latest release of the database allows an update in the Europe and Central Asia region bundled of stunting rates for 54 economies, and adds stunt- at the top of the distribution and the lowest rates ing rates for Argentina, Bulgaria, and Uzbekistan, in Sub-Saharan Africa, in economies like Chad, which did not have a rate in the previous iteration Nigeria, and Sierra Leone; however, in a number of the HCI. of economies in Sub-Saharan Africa, including Burundi, Malawi, or Rwanda, child survival rates Adult survival rates are significantly higher than those economies’ The adult survival rate is calculated as the com- level of GDP would predict (figure 1.6). plement of the mortality rate for 15- to 60-year- olds. The mortality rate for 15- to 60-year-olds Although internationally comparable stunting is the probability that a 15-year-old in a specified measures are primarily collected in low- and year will die before reaching the age of 60 if sub- middle-income economies, the share of stunted ject to current age-specific mortality rates. The children decreases as economies get richer. But mortality rate is frequently expressed as a rate per income and stunting rates do not always go in 1,000 alive at 15, in which case it must be divided lockstep, including across socioeconomic groups by 1,000 to obtain the probability that a 15-year- within economies (de Onis and Branca 2016). For old will die before age 60. Adult mortality rates for example, in economies such as Burundi, Niger, the 2020 update of the HCI come from the 2019 and Tanzania, the gap in stunting rates between update of the United Nations Population Division the first and the fourth socioeconomic quintiles is 12 (UNPD) World Population Prospects estimates. smaller than the gap between stunting rates in the Because UNPD does not individually report fourth and fifth quintiles (the richest households), adult mortality rates for economies with fewer reflecting the interaction of environmental, eco- than 90,000 inhabitants, UNPD data are supple- nomic, and cultural factors that can contribute to mented with adult mortality rates from the Global slower physical development in children (World Burden of Disease (GBD) project, managed by the Bank 2019b). In economies such as Guatemala, Institute of Health Metrics and Evaluation (IHME). Papua New Guinea, and Timor-Leste, more than Data from this source are used for Dominica and 45 percent of children are stunted. On the other the Marshall Islands. Data for Nauru, Palau, San end of the spectrum are economies like Moldova, Marino, St. Kitts and Nevis, and Tuvalu come from Samoa, Tonga, and West Bank and Gaza, where the WHO. The GBD data for the HCI 2020 come from stunting rate is below 10 percent, and significantly the GBD 2017 update and can be retrieved from lower than their levels of GDP would predict. The 13 the IHME data visualization site. The WHO data second proxy for health—adult survival—is low- are located on the United Nations data platform, est in the Central African Republic, Eswatini, and 14 UNData. Lesotho, where the chances of surviving from age 15 to age 60 are at 60 percent or lower. 1.3.2  Index components across economies All five components of the index increase with Quantity of schooling—as measured by EYS— income, though at a different pace (figure 1.6). increases as economies get richer. High-income Child survival rates range from 0.998 (2 deaths economies are bundled at the top of the distri- per 1,000 live births) in the richest economies to bution, and low-income economies are at the about 0.880 (120 deaths per 1,000 live births) in bottom. In economies like the Kyrgyz Republic, the poorest economies, reflecting the dispropor- Malawi, Nepal, and Zimbabwe, EYS are higher tionate burden of child mortality that low-income than those economies’ levels of GDP would T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 29 Figure 1.6: Human Capital Index 2020: Index components a. Probability of survival to age 5 b. Expected years of school Probability of survival to age 5, circa 2020 1.00 14 Expected years of school, circa 2020 Nicaragua Kyrgyz Republic Kyrgyz Republic Solomon Islands Tuvalu Vanuatu Nepal Rwanda 12 Micronesia, Fed. Sts. Haiti Malawi Kiribati 0.95 Zimbabwe Burundi 10 Malawi Mauritania Angola Congo, Dem. Rep. Côte d’Ivoire Angola Gabon Benin 8 Botswana Mali Guinea 0.90 Iraq Sudan Eswatini Chad Nigeria 6 Mali Liberia 0.85 4 6 8 10 12 6 8 10 12 Log GDP per capita at PPP, circa 2020 Log GDP per capita at PPP, circa 2020 c. Harmonized test scores d. Fraction of children under 5 not stunted Fraction of children under 5 not stunted, circa 2020 600 1.0 St. Lucia North Macedonia Turkey Harmonized test scores, circa 2020 Samoa Bulgaria West Bank and Gaza Argentina Kazakhstan Estonia Poland Vietnam Gambia, The Brunei Darussalam 500 0.8 Haiti Uzbekistan Malaysia Ukraine Kenya Cambodia Qatar Burundi Saudi Arabia 400 0.6 Angola Kuwait Timor−Leste Panama Guatemala Dominican Republic Papua New Guinea South Africa Nigeria Ghana 300 0.4 6 8 10 12 6 8 10 12 Log GDP per capita at PPP, circa 2020 Log GDP per capita at PPP, circa 2020 e. Adult survival rate 1.0 Morocco Nauru Adult survival rate, circa 2020 0.9 West Bank and Gaza Vanuatu Tajikistan Timor−Leste Solomon Islands Nepal 0.8 Namibia 0.7 South Africa Nigeria Côte d’Ivoire Zimbabwe 0.6 Eswatini Central African Republic Lesotho 0.5 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure reports the most recent cross-section of 174 economies for the five HCI components (child survival, expected years of school, harmonized test scores, fraction of children under 5 not stunted, and adult survival), as used to calculate the 2020 HCI. Each panel plots the country-level averages for each component on the y-axis and GDP per capita in PPP on the x-axis. The dashed line illustrates the fitted regression line between GDP per capita and the respective component. Scatter points above (below) the fitted regression line illustrate economies that perform better (worse) in the outcome variable than their level of GDP would predict. Countries above the 95th and below the 5th percentile in distance to the fitted regression line are labeled. 30 The H uman   Capital Index 20 20 U pdate predict, reflecting the progress these econo-  CI MEASURES OF GENDER GAPS IN 1.4  H mies have made in improving access to school- HUMAN CAPITAL figure 1.6). Outliers for which the quantity ing (­ of schooling is about 2.5 to 5.3 years below what Globally, the average HCI is slightly higher for their level of GDP would predict include econ- girls (0.59) than for boys (0.56).16 This pattern can omies such as Iraq, Liberia, and Mali, which are figure 1.7). be observed across all HCI components (­ characterized by different levels of institutional Although the gap between boys and girls has closed fragility and conflict. in these early-life outcomes, boys and girls both remain far from the frontier of complete edu- Quality of schooling—as measured by harmonized cation and full health. The gap in human capital test scores (HTSs)—increases with income, too, compared to full potential far exceeds any gender though seemingly faster than years of education. gap in HCI in most economies. Boys and girls are, The HTS ranges from a score of about 305 in the respectively, 2.6 and 2.5 years of schooling away poorest economies to a score of about 575 in the from completing upper-secondary education. richest economies (figure 1.6). To interpret the units Large shares of boys and girls are stunted—24 and of the HTS, note that 400 corresponds to the bench- 21 percent, respectively. Far too many boys and girls mark of “low proficiency” in TIMSS at the student do not survive beyond their fifth birthday—2.8 and level, whereas 625 corresponds to “advanced profi- 2.4 percent, respectively. Conditional on making it ciency.” Accounting for the level of GDP, economies to age 15, only 83 percent of boys and 89 percent of such as Vietnam, Ukraine, and Uzbekistan, as well girls are expected to survive to age 60. as Cambodia and Kenya, performed particularly well in learning. Vietnam reaches an HTS of 519, The global HCI average, however, masks import- a level similar to economies like the Netherlands, ant regional and income group differences with New Zealand, and Sweden, which are significantly respect to gender (figure 1.8). Although girls still richer.15 Economies for which learning is below surpass boys in the HCI value overall, with lower what their income per capita would predict include stunting and lower child and adult mortality rates high-income economies such as Kuwait, Qatar, in all regions and income groups, advantages for and Saudi Arabia. Their relatively disappointing girls are more prominent in some regions and performance in learning may result in part from a muted in others. For example, the gap in stunting traditional emphasis on investing in school infra- rates between girls and boys is as high as 4.6 per- structure rather than in other factors that are also centage points in Sub-Saharan Africa, with boys necessary to improve educational outcomes. These having a higher stunting rate. factors include governance and accountability, effective monitoring mechanisms, information With regard to EYS, girls are still disadvantaged sharing with parents and students, and school sys- compared to boys in South Asia and Sub-Saharan tems geared toward inclusive learning (Galal et al. Africa, where girls and boys experience 0.45 and 2008). Education systems in these economies may 0.15 years of school disadvantage, respectively also be reacting to the pull from labor markets, (­ figure 1.8). In settings affected by fragility, conflict, where pervasive informality generates low returns and violence, girls on average complete 0.14 years to schooling, and the lure of public employment less schooling than boys. In low-income econo- that puts more emphasis on diplomas than on skills mies, aside from completing less schooling, girls (El-Kogali and Krafft 2020; World Bank 2013). As also have lower HTSs, with a 0.8 percent deficit. a consequence, learning lags behind the progress that economies in this region have achieved in The gender gap in the HCI varies quite widely access to schooling and gender parity. across economies, with a difference in the score T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 31 Figure 1.7: Sex-disaggregated Human Capital Index and its components Human Capital Index Probability of survival to age 5 Expected years of school Harmonized test scores Fraction of children under 5 not stunted Adult survival rate 0 0.2 0.4 0.6 0.8 1.0 Girls-to-boys ratio, circa 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The red vertical line indicates gender parity for each component. Simple averages are computed without population weights. between boys and girls ranging from a low of middle-income economies, girls outperform boys –0.043 in Afghanistan to a high of 0.096 in in enrollment and learning outcomes (Bossavie Lithuania (f­igure  1.9). Overall, girls are outper- and Kanninen 2018). For example, in Guyana, forming boys in 140 of the 153 economies for girls are expected to complete one-fifth of a year which sex-disaggregated data are available. more schooling than boys, with 5 percent higher learning outcomes. This reverse gap in enrollment Gender gaps in EYS and HTSs show similar pat- begins in lower-secondary education and widens terns. The gender gap in EYS favors boys in 46 in upper-secondary, where girls are 11 percent economies (30 percent of all economies with a more likely to be enrolled than boys. sex-disaggregated HCI; figure 1.9). In learning out- comes, boys are favored in 31 economies (20 per- In survival and health outcomes, girls are gener- cent). Although EYS are higher for girls than for ally better off than boys. Girls have higher adult boys in most economies, the magnitude of the survival rates in all of the 153 economies for which resulting gender disparity is larger in those econ- sex disaggregation is available in the HCI 2020. In omies where boys have an advantage over girls all but two economies—India and Tonga—child with respect to schooling. For example, in Kiribati, survival rates are higher for girls than for boys. St. Vincent and the Grenadines, and Tunisia, girls Meanwhile, girls are more likely to be stunted on average complete more than one extra year of than boys in just 5 of 85 economies: Bhutan, Iraq, school compared to boys, whereas, in Angola and Kazakhstan, Moldova, and Tunisia.17 Afghanistan, boys on average complete 2.3 to 2.7 more years of school than girls. The top-five econ- Overall, out of the 13 economies where boys omies where girls outperform boys in learning have a higher HCI score than girls, 8 are in Sub- outcomes are Nauru, Qatar, Oman, Bahrain, and Saharan Africa, 2 in South Asia, 1 in the East Asia Samoa, three of which are in the Middle East and and Pacific region, 1 in Latin America and the North Africa region. Conversely, 6 in 10 econo- Caribbean, and 1 in the Middle East and North mies where boys have higher learning outcomes Africa region. Seven of those economies are than girls are in Sub-Saharan Africa. In high- and low-income, 5 are lower-middle-income, and 1 32 The H uman   Capital Index 20 20 U pdate Figure 1.8: Regional and income-group variations in education gaps between boys and girls a. By income group Low-income economies Lower-middle-income economies Upper-middle-income economies High-income economies 0 0.2 0.4 0.6 0.8 1.0 Girls-to-boys ratio, circa 2020 b. By region East Asia and Pacific Europe and Central Asia Latin America and Caribbean Middle East and North Africa North America South Asia Sub−Saharan Africa 0 0.2 0.4 0.6 0.8 1.0 Girls-to-boys ratio, circa 2020 Expected years of school Harmonized test scores Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). is upper-middle-income. In all 13 economies, in 9 of these 13 economies. In Chad and Guinea, EYS for boys are higher than for girls, ranging this difference reaches more than 14 percent in from a quarter year in Peru to almost three full favor of boys. years in Afghanistan. On average, boys have a 10-percentage-point higher likelihood of com- Human capital accumulation is a complex process. pleting primary education, a 12-percentage-point This complexity is especially clear when looking at higher likelihood of completing lower-secondary the HCI to understand gender gaps. Women, girls, education, and a 13-percentage-point higher like- men, and boys face different challenges at differ- lihood of completing upper-secondary education. ent stages of the life cycle. The HCI focuses on Boys also have better learning outcomes than girls specific life-cycle stages in which girls have slight T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 33 Figure 1.9: Global variation in gender gaps, Human Capital Index and education components a. Human Capital Index 0.1 Disparity favors girls 0.05 Girls-to-boys gap 0 Disparity favors boys −0.05 b. Expected years of school Disparity favors girls 1 0 Girls-to-boys gap −1 −2 Disparity favors boys −3 c. Harmonized test scores 60 Disparity favors girls 40 Girls-to-boys gap 20 0 Disparity favors boys −20 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The x-axis show economies ranked by girls-to-boys gap in the variable in question. The y-axes are on the 0-1 scale for the HCI, years of schooling for EYS, and HTS points for Harmonized Test Scores. 34 The H uman   Capital Index 20 20 U pdate biological advantages over boys in child and adult approach, and effective use of data and measure- survival rates (Crimmins et al. 2019; United Nations ment. These features, however, are not typical of 2011). As with any indicator, the components of the economies that are grappling with fragility, con- index are not perfect proxies of human capital and flict, and violence. By definition, such settings try to balance accuracy and data availability. For are plagued with high levels of institutional and example, the index does not capture gender bias social fragility, often with deteriorating gover- in terms of sex-selective abortions (what might nance capacity. In many cases, these economies 18 be called prebirth survival). Moreover, health is are experiencing prolonged political crises or are proxied by adult mortality rates, but some evi- undergoing a gradual but delicate reform and dence shows that, although women live longer recovery process. Such circumstances complicate than men, they are not necessarily in better health the process of consensus building and resource (Bora and Saikia 2018; Guerra, Alvarado, and mobilization across political cycles and there- Zunzunegui 2008). As a measure of the human fore pose a unique set of challenges in improving capital potential of children today, the index does human capital (World Bank 2020b). not capture gender gaps in human capital among the current population of adults. These caveats The importance of investing in human capital are important backdrops to any analysis of gender extends beyond the gains it promises in labor pro- gaps using the HCI. Finally, the index implicitly ductivity and in ensuring that growth is inclusive assumes that a child born today will be absorbed and sustainable. Human capital is also a cornerstone into the labor market to use her human capital of social cohesion, equity, and trust in institutions potential in terms of income generation, when, (Kim 2018). The seven economies scoring lowest in fact, female labor participation rates, globally, on the HCI in 2020 are all on the World Bank’s are 27 percentage points lower than male labor current annual list of fragile and conflict-affected 19 participation rates. Chapter 4 on human capital situations (FCS).20 Compared with the rest of the utilization delves into this disparity, by proposing world, economies affected by conflict and violence an adjustment of the HCI that captures labor out- are, on average, significantly further away from comes. These outcomes reflect one of many ways reaching the productivity frontier. human capital is utilized to improve well-being and overall economic development. Shocks, such as armed conflict or natural disas- ters, have a lasting impact on human capital. Some Equal access to education and health is far from pathways for this impact are obvious, including the realized. Despite progress, girls continue to face destruction of human potential through combat greater challenges. Child marriage, household deaths and casualties of natural calamities; damage responsibilities, teenage pregnancies, and gender- to critical infrastructure and institutions, such as based violence in schools pose challenges to hospitals and schools; and the loss of skills resulting keeping girls enrolled, especially, but not only, in from mass displacement. But the impacts of these low-income settings. shocks on human capital reach farther. For instance, emerging evidence shows that the destructive impacts of armed conflict on health and education  UMAN CAPITAL IN FRAGILE AND 1.5  H outcomes persist long after the fighting stops— CONFLICT-AFFECTED CONTEXTS extending to future generations not yet born when the conflict occurred (Corral et al. 2020). Human capital accumulation requires a sus- tained political commitment, an adequate and Classic studies of conflict and human capital have timely resource mobilization, a whole-of-society given central attention to health impacts on children T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 35 exposed to conflict settings. The link between vio- Human capital depletion in economies in FCS also lent conflict and a range of negative health outcomes happens through reduced and unequal access to among children has been established causally. For education and through poor learning outcomes example, physical development was stunted in chil- among those who do have access. Refugee and dren who were exposed to the 2002–07 civil conflict internally displaced children embody the losses of in Côte d’Ivoire, and this negative impact increased educational human capital associated with armed with the length of exposure to the conflict (Minoiu conflict. Conflict in the Syrian Arab Republic, for and Shemyakina 2014). The impact of conflict on example, has led to disruptions in education for human capital increases with increasing conflict millions of children, including over a million who severity. Children living in areas of Nigeria that were have been forced to flee to neighboring countries heavily affected by the Boko Haram insurgency had (Sieverding et al. 2018). Jordan hosts one of the lower weight-for-age and weight-for-height z-scores largest populations of Syrian refuges and has made and higher probability of wasting than children liv- concerted efforts to provide access to education 21 ing in less-affected areas. for refugee children. Despite those efforts, Syrian refugee children in Jordan experience delayed The intensity of conflict also determines the extent entry into school and early exit, with enrollment of human capital depletion. For instance, at aggre- rates dropping sharply from around age 12, when gate level, the distance from the HCI frontier (an refugee children come under pressure to work HCI score of 1) increases with the intensity of con- and help support their families (box 1.5; see also flict, even among economies in FCS. Those econo- Tiltnes, Zhang, and Pedersen 2019). mies with high-intensity conflict, defined as having at least 10 conflict deaths per 100,000 people, with a Globally, refugees access education at much lower minimum of 150 casualties, have consistently scored rates than other children do. In 2016, only 61 per- lower on all components of the index, compared cent of refugee children attended primary school, with other FCS and non-FCS economies (figure 1.10). compared to 91 percent of all children. At the sec- ondary level, 23 percent of refugee children were Conflict can have adverse effects on human cap- enrolled, versus 84 percent of eligible young peo- ital across generations. Well-being and health ple worldwide (UNHCR 2017). These shortfalls are outcomes among women in Nepal exposed as especially concerning because the number of ref- children to the country’s post-1996 civil war were ugees and displaced people worldwide has risen significantly worse than for those women and steadily through the past decade and now stands at children who were not exposed to conflict. Not its highest level since World War II. only did the first-generation victims show signif- icant reductions in final adult height, but, when The intergenerational impact of conflict and vio- the conflict-exposed victims had children of lence extends to losses in educational attainment their own, those children also suffered reduced for children not even born when fighting took place. weight-for-height and body mass index z-scores, For example, in utero exposure to the Rwandan on average. Women exposed to the conflict during genocide decreased educational attainment by 0.3 childhood had more children and lived in poorer years and the likelihood of completing primary households as adults. The combination of these school by 8 percent (Bundervoet and Fransen two factors may decrease parents’ ability to invest 2018). The impact on years of schooling was stron- in their children’s human capital during critical ger for females and for individuals exposed to the phases of physical and cognitive development and genocide in the first trimester of gestation. Each therefore may propagate these impacts intergen- additional month of exposure in utero decreased erationally (Phadera 2019). educational attainment by 0.21 years of schooling. 36 The H uman   Capital Index 20 20 U pdate Figure 1.10: Human capital and severity of conflict Human Capital Index Survival rate from age 15−60 Fraction of children under 5 not stunted Probability of survival to age 5 Learning−adjusted years of school Harmonized test scores Expected years of school Lowest Median Highest Average of economies in high−intensity conflict Economy in high−intensity conflict Average of other economies in FCS Other economy in FCS Average of economies not in FCS Economy not in FCS Source: Corral et al. 2020 with updated Human Capital Index (HCI) data for 2020. Note: Economies in high-intensity conflict are defined as having at least 10 conflict deaths per 100,000 people according to the Armed Conflict and Event Data Project (ACLED) and the Uppsala Conflict Data Program (UCDP), while also experiencing a total of more than 250 conflict deaths according to ACLED, or more than 150 conflict deaths according to UCDP. FCS = fragile and conflict-affected situations. Through in utero exposure, conflict-related dis- experiences of been learned, however, from the ­ ruptions of fetal cognitive development may affect various countries. In Afghanistan, for example, fol- children’s subsequent cognitive capacities, educa- lowing the withdrawal of the Taliban in 2001, the tional outcomes, and earning power as adults. Ministry of Public Health had to provide emergency relief services to address the grave health situation How can fragile countries and development partners throughout the country. Yet the health system was confront the losses of human capital driven by con- in ruins after decades of warfare and neglect. As they flict? The best solution is to prevent fragility, con- rolled out emergency health services, including in flict, and violence from engulfing countries in the many areas still subject to conflict, health officials first place. But, when conflict does erupt, effective had to plan for the future, which included rebuilding delivery of health and education services tailored to and sustaining a functional national health system. economies in FCS is vital. Only the preservation and Acknowledging its capacity limitations and with rebuilding of human capital can enable countries to technical assistance from the international commu- durably escape cycles of fragility and violence. nity, the Ministry of Public Health led the creation of an innovative public-private partnership frame- The delivery of health and education services in FCS work for health service delivery in Afghanistan poses daunting challenges, not least because of the (Newbrander, Waldman, and Shepherd-Banigan extreme diversity of FCS contexts. Much has recently 2011). This delivery model has improved key health T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 37 Box 1.5: Schooling for Syrian refugee children in Jordan The government of Jordan has adopted a policy of offering refugee children tuition-free access to the public education system, while also providing accredited schools in refugee camps. As a result, overcrowding has occurred in schools in some locations. Despite these measures, access to school for refugee children is still limited. Only about 152,000 of the estimated 236,000 Syrian refugee children present in Jordan are enrolled (64 percent). Figure B1.5.1 paints a stark picture of the enrollment decline by age among refugees. It shows that enrollment significantly tails off after age 11, more so for boys than for girls. This decline is driven by several factors, including pov- erty (because most families cannot cover the auxiliary costs of education, such as transportation and school materials); early marriage (which is also evident in recent household surveys); and increased opportunity cost of education (because many children start working early to support their families). Reports suggest that bullying at schools and the absence of a safe learning space impede learn- ing for Jordanian boys as well as for Syrian refugees, and the Jordanian government is taking measures to address this issue. In addition, important reforms such as ensuring universal enroll- ment for 5-year-olds in preprimary education apply to all inhabitants of Jordan, including Syrian refugee children. Figure B1.5.1: Net enrollment rate of Syrian refugees in formal education in Jordan 100 Share of children enrolled (%) 80 60 40 20 0 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 Age (years) Girls Boys Source: Krafft et al. 2018. Note: Data include only refugees registered with the United Nations High Commissioner for Refugees. indicators under highly challenging conditions and authorities, donors, and international partners has been recognized as an example for other post- must coordinate their efforts to ensure that health conflict countries (World Bank 2018c). and other essential services for refugees can be sustainably paid for. An important resource to Without adequate health financing, health ser- facilitate such durable support came with the vice delivery simply will not happen. The ongo- 2016 launch of the Global Concessional Financing ing Syrian crisis has underscored that country Facility (GCFF). Led jointly by the World Bank, 38 The H uman   Capital Index 20 20 U pdate the United Nations, and the Islamic Development educational opportunities for refugees. But the Bank, the GCFF is a global platform designed to process remains fraught with challenges related to deliver concessional funding to middle-income system capacities, persistent access barriers, qual- countries that provide a global public good ity, and resources. by hosting large numbers of refugees. GCFF resources enable governments in host countries to Furthermore, the lack of timely, reliable, and offer expanded services to refugees while continu- actionable data and of a robust measurement ing to meet the needs of their own citizens. Early agenda also hinders progress in human capital GCFF concessional loans reduced the acute finan- accumulation in FCS. Although high-quality data cial burden on Lebanon and Jordan, two countries are critical for diagnosing deficiencies and formu- on the front lines of the Syrian refugee response. lating targeted policies and programs to enhance Subsequently, the GCFF has worked to smooth the human capital, such data are not readily available transition from humanitarian assistance to devel- across many economies that are in the midst of, opment by providing medium- and long-term or are recovering from, fragility and conflict. For concessional finance. instance, the HCI score cannot be calculated for some of the economies that are considered FCS Even more than other countries, those in FCS need according to the World Bank 2020 classification, education systems that can promote learning, life often because data informing various HCI com- skills, and social cohesion. Only by securing broad ponents either do not exist or are outdated. Even dissemination of these capacities in the population when the index can be calculated compatibly with through quality education can countries build last- the HCI data inclusion rules (see appendix C), it ing foundations for peace and economic recovery. still might not fully capture the deterioration of But the challenges in delivering equitable, quality human capital that can follow the rapidly changing education are not straightforward. reality of countries in conflict. In the case of the Republic of Yemen, available data for index com- Some of the greatest service delivery challenges in ponents mostly predate the conflict and might not conflict-affected countries involve education for fully represent the effect of the conflict on school- displaced populations and host communities. Over ing or child health. Moreover, comparable data for the years, various flexible learning strategies have refugees and hosts are almost nonexistent in econ- been fielded across different settings. Learning omies afflicted by fragility and conflict. from these global experiences, host countries have consistently moved away from providing refugee Collecting high-quality data requires sustained education in parallel systems that may lack quali- and deliberate efforts. In light of other pressures fied teachers, consistent funding, ability to provide in situations of violence and conflict, measure- diplomas, and quality control. Ethiopia’s Refugee ment is rarely a priority; however, collecting Proclamation, for example, gives refugees access high-quality data is feasible in these settings. For to national schools and gives host children access instance, mobile phone interviews were used for to refugee schools. The Islamic Republic of Iran data collection during the Ebola crisis in Sierra decreed in 2015 that schools accept all Afghan Leone and to inform a response to drought in children regardless of documentation. Turkey Nigeria, Somalia, South Sudan, and the Republic has committed to include all Syrian refugee chil- of Yemen (Hoogeveen and Pape 2020). Likewise, dren in its national education system by 2020 satellite images and machine learning algorithms (UNESCO 2019). The inclusion of refugees in were employed to address the lack of a sampling national education systems dramatically expands frame in the Democratic Republic of Congo and T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 39 in Somalia. When data collection is hampered 1.6  THE HCI 2020 UPDATE by security concerns for enumerators, locally recruited resident enumerators make it possible to Table 1.2 presents the overall HCI 2020 for collect relevant, reliable, and timely evidence that 174 economies. The index ranges from 0 to 1 and is can shed light on the plight of the most vulnerable measured in terms of the productivity of the next populations. These efforts can move the needle on generation of workers relative to the benchmark of addressing persistent data deprivation in FCS. complete education and full health. An economy in which a child born today can expect to achieve com- Protecting and rebuilding human capital in set- plete education and full health will score a value of tings of fragility and conflict are crucial to restore 1 on the index. All of the components of the HCI hope in these countries. They are also critical are measured with some error, and this uncertainty for reaching global poverty goals. Over recent naturally has implications for the precision of the decades, poverty has become steadily more con- overall HCI. To capture this imprecision, the HCI centrated in economies in FCS. Fragility and con- estimates for each economy are accompanied by flict deplete human capital, yet societies must rely upper and lower bounds that reflect the uncertainty heavily on human capital to recover from fragility in the measurement of the components of the HCI. and conflict. This paradox underscores the impor- These bounds are constructed by recalculating the tance of health and education services in FCS HCI using lower- and upper-bound estimates of settings. Delivering these services lays the foun- the components of the HCI and are also reported in dations that will enable countries to emerge from table 1.2.22 This is intended to help move the discus- cycles of violence and return to peace, stability, sion away from small differences in economy ranks and development. Overcoming systemic barriers on the HCI and toward more useful discussions will simultaneously require careful coordination around the level of the HCI and what this implies between humanitarian and development partners for the productivity of future workers (see box 1.6). and a whole-of-society approach. Box 1.6: Where did the HCI rankings go? The 2020 update does not report rankings for the 174 economies with an HCI score. There are four reasons for this change. First, coverage of the index has increased by 17 economies, from 157 economies in the inaugu- ral 2018 HCI to 174 economies in 2020. Therefore, a rank of 37 out of 157, for instance, cannot be compared with a rank of 37 out of 174. Given the change in HCI coverage between 2018 and 2020, simple comparisons of rankings as an indication of an economy’s progress over time are meaningless. Second, even if comparisons were restricted to the set of economies included in both the 2018 and the 2020 versions of the index, rankings artificially inflate small differences in HCI scores. For example, eight economies are clustered between HCI scores of 0.60 and 0.61; if one of those economies at 0.60 improves by just 0.01, it would move up eight places in the ranking. By contrast, only two economies have scores between 0.70 and 0.71; if one of those two economies were to improve its score by 0.01, it would move up only one rank.a (continued next page) 40 The H uman   Capital Index 20 20 U pdate Box 1.6: Where did the HCI rankings go? (Continued) Third, rankings suppress information on the absolute gains and losses economies have made on the HCI. Consider for example the comparison of HCI 2020 and HCI 2010, which is graphed in panel a of figure B1.6.1.b Most economies have improved their human capital out- comes, reflected by the fact that they are above the 45-degree line in the figure. Rankings cannot convey these gains (or losses), because they present only the positions of econo- mies relative to each other, as illustrated in panel b of figure B1.6.1, which plots the same information for 2020 versus 2010 but in rank terms. Even economies that have made gains in human capital accumulation may fall below the 45-degree line simply because of their position relative to other economies. In addition, points in panel b are more spread out than those in panel a, illustrating how ranks artificially magnify small changes. Figure B1.6.1: Changes in Human Capital Index scores and ranks, 2010 vs. 2020 a. Human Capital Index scores, 2010 vs. 2020 b. Human Capital Index rank, 2010 vs. 2020 0.9 100 Macao SAR, China Finland Human Capital Index rank, circa 2020 0.8 Human Capital Index, circa 2020 Macao SAR, China Finland Cyprus Switzerland Cyprus Switzerland Italy 80 0.7 Greece Italy Russian Federation Slovak Republic Greece Albania Russian Federation 0.6 Ecuador Bulgaria 60 Slovak Republic Azerbaijan Romania Albania Bulgaria 0.5 Panama 40 Ecuador Romania Azerbaijan 0.4 Côte d’Ivoire Panama 20 0.3 Côte d’Ivoire 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 0 20 40 60 80 100 Human Capital Index rank, circa 2010 Human Capital Index, circa 2010 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: This figure compares HCI data for 2020 and 2010. The construction of an HCI for 2010 is described in chapter 2. The sample for 2010 does not include any South Asian countries since they are missing learning data for 2010 from a comparable representative international assessment. Fourth, and most important, there is no need to focus on rankings because the index itself is expressed in meaningful units. Because the HCI is measured in terms of the productivity of the next generation of workers relative to the benchmark of complete education and full health, the units of the index have a natural interpretation: a value of 0.50 for an economy means that the productivity as a future worker of a child born in a given year in that economy is only half of what it could be under the benchmark. Rankings place an inordinately large focus on the fact that an economy with an HCI of 0.51 is ahead of an economy with an HCI of 0.50. This interpreta- tion misses the more critical issue, which is that in both economies children born today will grow up with half their human capital potential unfulfilled. This information is vastly more important than whether one economy is “ahead of” another. a This problem is amplified by the fact that the components of the HCI are measured with some error, and this uncertainty naturally has implications for the precision of the overall HCI. To capture this imprecision, the HCI estimates for each economy are accompanied by upper and lower bounds that reflect the uncertainty in the measurement of HCI components. In cases where these intervals overlap for two economies, it indicates that the differences in the HCI estimates for these two economies should not be overinterpreted, because they are small relative to the uncertainty around the value of the index itself. Rankings further amplify these minor differences. b The construction of an HCI for 2010 is described in the following chapter. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 41 Table 1.2: The Human Capital Index (HCI), 2020 Lower Upper Lower Upper Lower Upper Economy Bound Value Bound Economy Bound Value Bound Economy Bound Value Bound Central African Republic 0.26 0.29 0.32 India 0.49 0.49 0.50 Mauritius 0.60 0.62 0.64 Chad 0.28 0.30 0.32 Egypt, Arab Rep. 0.48 0.49 0.51 Uzbekistan 0.60 0.62 0.64 South Sudan 0.27 0.31 0.33 Guyana 0.48 0.50 0.51 Brunei Darussalam 0.62 0.63 0.63 Niger 0.29 0.32 0.33 Panama 0.49 0.50 0.51 Kazakhstan 0.62 0.63 0.63 Mali 0.31 0.32 0.33 Dominican Republic 0.49 0.50 0.52 Costa Rica 0.62 0.63 0.64 Liberia 0.30 0.32 0.33 Morocco 0.49 0.50 0.51 Ukraine 0.62 0.63 0.64 Nigeria 0.33 0.36 0.38 Tajikistan 0.48 0.50 0.53 Seychelles 0.61 0.63 0.66 Mozambique 0.34 0.36 0.38 Nepal 0.49 0.50 0.52 Montenegro 0.62 0.63 0.64 Angola 0.33 0.36 0.39 Micronesia, Fed. Sts. 0.47 0.51 0.53 Albania 0.62 0.63 0.64 Sierra Leone 0.35 0.36 0.38 Nicaragua 0.50 0.51 0.52 Qatar 0.63 0.64 0.64 Congo, Dem. Rep. 0.34 0.37 0.38 Nauru 0.49 0.51 0.53 Turkey 0.64 0.65 0.66 Guinea 0.35 0.37 0.39 Fiji 0.50 0.51 0.52 Chile 0.64 0.65 0.66 Eswatini 0.35 0.37 0.39 Lebanon 0.50 0.52 0.52 Bahrain 0.64 0.65 0.66 Yemen, Rep. 0.35 0.37 0.39 Philippines 0.50 0.52 0.53 China 0.64 0.65 0.67 Sudan 0.36 0.38 0.39 Tunisia 0.51 0.52 0.52 Slovak Republic 0.66 0.66 0.67 Rwanda 0.36 0.38 0.39 Paraguay 0.51 0.53 0.54 United Arab Emirates 0.66 0.67 0.68 Côte d’Ivoire 0.36 0.38 0.40 Tonga 0.51 0.53 0.55 Serbia 0.67 0.68 0.69 Mauritania 0.35 0.38 0.41 St. Vincent and the Grenadines 0.52 0.53 0.54 Russian Federation 0.67 0.68 0.69 Ethiopia 0.37 0.38 0.39 Algeria 0.53 0.53 0.54 Hungary 0.67 0.68 0.69 Burkina Faso 0.36 0.38 0.40 Jamaica 0.52 0.53 0.55 Luxembourg 0.68 0.69 0.69 Uganda 0.37 0.38 0.40 Indonesia 0.53 0.54 0.55 Vietnam 0.67 0.69 0.71 Burundi 0.36 0.39 0.41 Dominica 0.53 0.54 0.56 Greece 0.68 0.69 0.70 Tanzania 0.38 0.39 0.40 El Salvador 0.53 0.55 0.56 Belarus 0.69 0.70 0.71 Madagascar 0.37 0.39 0.41 Kenya 0.53 0.55 0.56 United States 0.69 0.70 0.71 Zambia 0.38 0.40 0.41 Samoa 0.54 0.55 0.56 Lithuania 0.70 0.71 0.72 Cameroon 0.38 0.40 0.42 Brazil 0.55 0.55 0.56 Latvia 0.69 0.71 0.72 Afghanistan 0.39 0.40 0.41 Jordan 0.54 0.55 0.56 Malta 0.70 0.71 0.72 Benin 0.38 0.40 0.42 North Macedonia 0.55 0.56 0.56 Croatia 0.70 0.71 0.72 Lesotho 0.38 0.40 0.42 Kuwait 0.55 0.56 0.57 Italy 0.72 0.73 0.74 Comoros 0.36 0.40 0.43 Grenada 0.55 0.57 0.58 Spain 0.72 0.73 0.73 Pakistan 0.39 0.41 0.42 Kosovo 0.56 0.57 0.57 Israel 0.72 0.73 0.74 Iraq 0.40 0.41 0.41 Georgia 0.56 0.57 0.58 Iceland 0.74 0.75 0.75 Malawi 0.40 0.41 0.43 Saudi Arabia 0.56 0.58 0.59 Austria 0.74 0.75 0.76 Botswana 0.39 0.41 0.43 Azerbaijan 0.56 0.58 0.59 Germany 0.74 0.75 0.76 Congo, Rep. 0.39 0.42 0.44 Armenia 0.57 0.58 0.59 Czech Republic 0.74 0.75 0.76 Solomon Islands 0.41 0.42 0.43 Bosnia and Herzegovina 0.57 0.58 0.59 Poland 0.74 0.75 0.76 Senegal 0.40 0.42 0.43 West Bank and Gaza 0.57 0.58 0.59 Denmark 0.75 0.76 0.76 Gambia, The 0.39 0.42 0.44 Moldova 0.57 0.58 0.59 Cyprus 0.75 0.76 0.76 Marshall Islands 0.40 0.42 0.44 Romania 0.57 0.58 0.60 Switzerland 0.75 0.76 0.77 South Africa 0.41 0.43 0.44 St. Kitts and Nevis 0.57 0.59 0.60 Belgium 0.75 0.76 0.77 Papua New Guinea 0.41 0.43 0.44 Palau 0.57 0.59 0.61 France 0.75 0.76 0.77 Togo 0.41 0.43 0.45 Iran, Islamic Rep. 0.58 0.59 0.60 Portugal 0.76 0.77 0.78 Namibia 0.42 0.45 0.47 Ecuador 0.59 0.59 0.60 Australia 0.76 0.77 0.78 Haiti 0.43 0.45 0.46 Antigua and Barbuda 0.58 0.60 0.61 Norway 0.76 0.77 0.78 Tuvalu 0.43 0.45 0.46 Kyrgyz Republic 0.59 0.60 0.61 Slovenia 0.77 0.77 0.78 Ghana 0.44 0.45 0.46 Sri Lanka 0.59 0.60 0.60 New Zealand 0.77 0.78 0.78 Timor-Leste 0.43 0.45 0.47 Uruguay 0.59 0.60 0.61 Estonia 0.77 0.78 0.79 Vanuatu 0.44 0.45 0.47 Argentina 0.59 0.60 0.61 United Kingdom 0.77 0.78 0.79 Lao PDR 0.44 0.46 0.47 St. Lucia 0.59 0.60 0.62 Netherlands 0.78 0.79 0.80 Gabon 0.43 0.46 0.48 Trinidad and Tobago 0.57 0.60 0.62 Ireland 0.78 0.79 0.80 Guatemala 0.45 0.46 0.47 Colombia 0.59 0.60 0.62 Sweden 0.79 0.80 0.81 Bangladesh 0.46 0.46 0.47 Peru 0.59 0.61 0.62 Macao SAR, China 0.79 0.80 0.80 Zimbabwe 0.44 0.47 0.49 Oman 0.60 0.61 0.62 Finland 0.79 0.80 0.80 Bhutan 0.45 0.48 0.50 Thailand 0.60 0.61 0.62 Canada 0.79 0.80 0.81 Myanmar 0.46 0.48 0.49 Malaysia 0.60 0.61 0.62 Korea, Rep. 0.79 0.80 0.81 Honduras 0.47 0.48 0.49 Mexico 0.60 0.61 0.62 Japan 0.80 0.80 0.81 Cambodia 0.47 0.49 0.51 Bulgaria 0.60 0.61 0.62 Hong Kong SAR, China 0.80 0.81 0.82 Kiribati 0.46 0.49 0.52 Mongolia 0.60 0.61 0.63 Singapore 0.87 0.88 0.89 HCI < 0.40 0.40 ≤ HCI < 0.50 0.50 ≤ HCI < 0.60 0.60 ≤ HCI < 0.70 0.70 ≤ HCI < 0.80 0.80 ≤ HCI Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The Human Capital Index (HCI) ranges between 0 and 1. The index is measured in terms of the productivity of the next generation of workers relative to the benchmark of complete education and full health. An economy in which a child born today can expect to achieve complete education and full health will score a value of 1 on the index. Lower and upper bounds indicate the range of uncertainty around the value of the HCI for each economy. 42 The H uman   Capital Index 20 20 U pdate NOTES 9. PISA-D results are used only for Bhutan. 10. For the 20 new economies included in the 1. The HCI was introduced in World Bank Global Dataset on Education Quality, 8 are (2018a, 2018b), and its methodology is detailed updated using EGRAs, 8 using PILNA, 3 us- in Kraay (2018). ing PISA and PISA-D, and 1 using a national 2. As a result of the criteria for its construction, TIMSS-equivalent assessment. the index measures dimensions of human 11. Of the 94 economies with updated test scores capital that are important, but it does not in- in the Global Dataset on Education Quality, 75 clude all of the important dimensions of hu- use scores from PISA 2018, 7 from PISA-D, 5 man capital. from EGRAs, and 7 from PILNA. 3. The literature has recognized the usefulness 12. See the UNPD website, https://population​ of moving from “a large and eclectic dash- .un.org/wpp/. board” to a single summary metric (Stiglitz, 13. See the IHME website, http://www.healthdata​ Sen, and Fitoussi 2009). Doing so, however, .org/results/data-visualizations. requires a coherent aggregation method, in contrast with “mashup indicators of develop- 14. See https://data.un.org/. ment” that combine different components in 15. Note that Vietnam enters the HCI 2020 with arbitrary ways (Ravallion 2010). The HCI is its 2015 PISA score, because 2018 PISA scores constructed by transforming its components are not reported for the country. Although into contributions to productivity, anchored Vietnam participated in the 2018 round of in microeconometric evidence on the effects PISA using paper-based instruments, the Or- of education and health on worker produc- ganisation for Economic Co-operation and tivity, consistent with the large literature on Development’s country note states that the development accounting (see, for example, international comparability of the country’s Caselli 2005). performance in reading, mathematics, and 4. The decomposition of the group averages is science could not be fully ensured (OECD obtained via a Shapley decomposition. For an 2019). application see Azevedo, Inchauste, and San- 16. This difference is statistically significant at the felice 2013. 5 percent level. 5. Among upper-middle-income economies, the 17. Stunting rates are calculated using survey health component value of the bottom per- data, and differences in average rates between former is higher than that of the top perform- girls and boys may not be statistically signifi- er, and thus it accounts for a negative share of cant. the difference. 18. The number of “missing women” was esti- 6. For more information, see the IGME website, mated to be 126 million in 2010 (Bongaarts http://www.childmortality.org/. and Guilmoto 2015). This term refers to the 7. For Institute for Statistics data, see http://data​ deficit of females relative to males, com- .uis.unesco.org/. See also appendix C for the pared to the figures that would have been description of different enrollment rates: gross, observed had all female fetuses been allowed net, adjusted net, and total net enrollment rates. to be born. 8. For the 2020 update, this review process was 19. Data from ILOSTAT, the International Labour conducted between January and May 2020 Organization’s labor statistics database. Re- in collaboration with the country units of the trieved from World Bank Gender Data Portal, World Bank. https://datatopics.worldbank.org/gender/. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 43 20. These economies are the Central African Re- Caselli, F. 2005. “Accounting for Cross-Country public, Chad, Liberia, Mali, Niger, Nigeria, and Income Differences.” In Handbook of Economic South Sudan. Growth, 1st ed., vol. 1, edited by P. Aghion 21. A z-score is a measure of how many standard and S. Durlauf, 679–741. Elsevier. deviations below or above the population Corral, P., A. Irwin, N. Krishnan, D. G. Mahler, mean a raw score is. See Ekhator-Mobayode and T. Vishwanath. 2020. Fragility and Conflict: and Abebe Asfaw (2019). On the Front Lines of the Fight against Poverty. 22. The upper and lower bounds of the HCI are a Washington, DC: World Bank. tool to highlight to users that the estimated HCI Crimmins, E., H. Shim, Y. S. Zhang, and J. Kim. values for all countries are subject to uncertain- 2019. “Differences between Men and Women ty, reflecting the corresponding uncertainty in in Mortality and the Health Dimensions of the the components. In cases where these intervals Morbidity Process.” Clinical Chemistry 65 (1): 135–45. overlap for two countries, this indicates that de Onis, M., and F. Branca. 2016. “Childhood the differences in the HCI estimates for these Stunting: A Global Perspective.” Maternal & two countries should not be overinterpreted Child Nutrition 12 Suppl 1 (Suppl 1): 12–26. because they are small relative to the uncertain- D’Souza, R., R. Gatti, and A. Kraay. 2019. “A ty around the value of the index itself. Socioeconomic Disaggregation of the World Bank Human Capital Index.” Policy Research Working Paper 9020, World Bank, Washington, REFERENCES DC. Ekhator-Mobayode, U. E., and A. Abebe Asfaw. Azevedo, J. P., G. Inchaust, and V. Sanfelice. 2013. 2019. “The Child Health Effects of Terrorism: “Decomposing the Recent Inequality Decline Evidence from the Boko Haram Insurgency in in Latin America.” Policy Research Working Nigeria.” Applied Economics 51 (6): 624–38. Paper 6715, World Bank, Washington, DC. El-Kogali, S., and C. Krafft. 2020. Expectations and Bongaarts, J., and C. Guilmoto. 2015. “How Aspirations: A New Framework for Education in the Many More Missing Women? Excess Female Middle East and North Africa. Washington, DC: Mortality and Prenatal Sex Selection, 1970– World Bank. 2050.” Population and Development Review 41 (2): Filmer, D., H. Rogers, N. Angrist, and S. Sabarwal. 241–69. 2018. “Learning-Adjusted Years of Schooling Bora, J. K., and N. Saikia. 2018. “Neonatal and (LAYS): Defining a New Macro Measure of Under-Five Mortality Rate in Indian Districts Education.” Policy Research Working Paper with Reference to Sustainable Development 8591, World Bank, Washington, DC. Goal 3: An Analysis of the National Family Galal, A., M. Welmond, M. Carnoy, S. Nellemann, Health Survey of India (NFHS), 2015–2016.” J.  Keller, J. Wahba, and I. Yamasaki. 2008. PLoS ONE 13 (7): e0201125. “The Road Not Traveled: Education Reform Bossavie, L., and O. Kanninen. 2018. “What Explains in the Middle East and North Africa.” MENA the Gender Gap Reversal in Educational Development Report, World Bank, Washington, Attainment?” Policy Research Working Paper DC. 8303, World Bank, Washington, DC. Guerra, R. O., B. E. Alvarado, and M. V. Bundervoet, T., and S. Fransen. 2018. “The Zunzunegui. 2008. “Life Course, Gender and Educational Impact of Shocks in Utero: Ethnic Inequalities in Functional Disability in Evidence from Rwanda.” Economics & Human a Brazilian Urban Elderly Population.” Aging Biology 29 (C): 88–101. Clinical and Experimental Research 20 (1): 53–61. 44 The H uman   Capital Index 20 20 U pdate Hoogeveen, J., and U. Pape, eds. 2020. Data Ravallion, M. 2010. “Mashup Indices of Collection in Fragile States: Innovations from Development.” Policy Research Working Paper Africa  and Beyond. Washington, DC: World 5432, World Bank, Washington, DC. Bank. Sieverding, M., C. Krafft, N. Berri, C. Keo, and Kaffenberger, M., and L. Pritchett. 2017. “More School M. Sharpless. 2018. “Education Interrupted: or More Learning? Evidence from Learning Enrollment, Attainment, and Dropout of Profiles from the Financial Inclusion Insights Syrian Refugees in Jordan.” Working Paper Data.” Background paper, World Development 1261, Economic Research Forum, Giza, Egypt. Report 2018: Learning to Realize Education’s Stiglitz, J., A. Sen, and J.-P. Fitoussi. 2009. “The Promise, World Bank, Washington, DC. Measurement of Economic Performance and Kim, J. 2018. “The Human Capital Gap: Getting Social Progress Revisited: Reflections and Governments to Invest in People.” Foreign Affairs Overview.” Sciences Po, Publications 2009-33, July/August 2018. https://www.foreignaffairs​ Paris Institute of Political Studies. .com/articles/2018-06-14/human-capital-gap. Tiltnes, A., H. Zhang, and J. Pedersen. 2019. The Kraay, A. 2018. “Methodology for a World Bank Living Conditions of Syrian Refugees in Jordan: Human Capital Index.” Policy Research Working Results from the 2017–2018 Survey of Syrian Paper 8593, World Bank, Washington, DC. Refugees Inside and Outside Camps. Oslo: Fafo Krafft, C., M. Sieverding, C. Keo, and C. Salemi. 2018. Institute for Labour and Social Research. “Syrian Refugees in Jordan: Demographics, UNESCO (United Nations Educational, Scientific Livelihoods, Education, and Health.” Working and Cultural Organization). 2019. Global Paper 1184, Economic Research Forum, Giza, Education Monitoring Report 2019: Migration, Egypt. Displacement and Education: Building Bridges, not Minoiu, C., and O. N. Shemyakina. 2014. “Armed Walls. Paris: UNESCO. Conflict, Household Victimization, and Child UNHCR (United Nations High Commissioner Health in Côte d’Ivoire.” Journal of Development for Refugees). 2017. “Left Behind: Refugee Economics 108 (C): 237–55. Education in Crisis.” UNHCR, Geneva. Newbrander, W., R. Waldman, and M. Shepherd- UNICEF (United Nations Children’s Fund), WHO Banigan. 2011. “Rebuilding and Strengthening (World Health Organization), and World Health Systems and Providing Basic Health Bank Group. 2020. Levels and Trends in Child Services in Fragile States.” Disasters 35: 639–60. Mortality: Key Findings of the 2020 Edition. OECD (Organisation for Economic Co-operation UNICEF, WHO, and World Bank Group. and Development). 2019. “Viet Nam Country United Nations. 2011. “Sex Differentials in Note—PISA 2018 Results.” Programme for Childhood Mortality.” Department of International Student Assessment (PISA), OECD. Economic and Social Affairs, Population Patrinos, H. A. and Angrist, N. 2018. “Global Dataset Division, United Nations, New York. on Education Quality: A Review and Update World Bank. 2013. Jobs for Shared Prosperity: Time (2000–2017).” Policy Research Working Paper for Action in the Middle East and North Africa. 8592, World Bank, Washington, DC. Washington, DC: World Bank. Phadera, L. 2019. “Unfortunate Moms and World Bank. 2018a. “The Human Capital Project.” Unfortunate Children: Impact of the Nepali Civil World Bank, Washington, DC. War on Women’s Stature and Intergenerational World Bank. 2018b. “Human Capital: A Project for Health.” Policy Research Working Paper 8927, the World.” Development Committee, World World Bank, Washington, DC. Bank, Washington, DC. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 45 World Bank. 2018c. “Progress in the Face of World Bank. 2020a. Markets and People: Romania Insecurity: Improving Health Outcomes in Country Economic Memorandum. Washington, Afghanistan.” World Bank, Washington, DC. DC: World Bank. World Bank. 2018d. World Development Report World Bank. 2020b. “World Bank Group Strategy 2018: Learning to Realize Education’s Promise. for Fragility, Conflict, and Violence 2020– Washington, DC: World Bank. 2025.” World Bank, Washington, DC. World Bank. 2019a. “Ending Learning Poverty: What World Health Organization. 2009. “The WHO Will It Take?” World Bank, Washington, DC. Multicentre Growth Reference Study World Bank. 2019b. “Insights from Disaggregating (MGRS).” World Health Organization, the Human Capital Index.” World Bank, Geneva. Washington, DC. 2 Human Capital Accumulation OVER TIME T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 47 A s the first update of the Human Capital  UMAN CAPITAL ACCUMULATION 2.1  H Index (HCI), the 2020 release is an oppor- OVER THE PAST DECADE tunity to look at the evolution of human capital outcomes, as measured by the HCI, across To track progress over the past decade, a version economies over time. of the HCI has been calculated for 103 economies using component data from or near 2010. Data Unlike indexes that aggregate laws or regulation, used to populate the 2010 HCI have been care- which can be modified by swift government leg- fully selected to maximize comparability with the islative or regulatory action, the HCI is based 2020 HCI. In particular, only those economies on outcomes that typically change slowly from for which learning scores were measured by the year to year. Some of them—such as stunting same international assessment in 2010 and 2020 and educational test scores—are measured infre- comparison (see box 2.1). Requiring test enter the ­ quently, every three to five years. As a result, scores from the same testing program proved the changes in the HCI over a short period are small main constraint to building large and represen- and might simply reflect updates to components tative coverage, and the resulting 2010 sample is, measured sporadically. In contrast, the that are ­ unsurprisingly, skewed toward richer economies ­ analysis of longer-term trends has a more solid quality that tend to have more complete and better-­ basis, given the scope for smoothing out short- data. For example, the sample does not cover South run idiosyncrasies. Asia, because none of the seven economies in the region with an HCI in 2020 has learning data for capital This chapter examines trends in human ­ 2010 from the same representative international over  time. The first section discusses the test assessment as for 2020. The average 2020 construction of an HCI for 2010 and the evo- ­ HCI is 0.56 for the 174 economies in the overall lution of the HCI between 2010 and 2020. The sample, compared to 0.62 for the 103 economies following section unpacks these dynamics by look- ­ that are part of the comparison-over-time sample ing at changes in the components of the HCI. The (table 2.1). The potential bias is largest in the East final section provides a policy focus. Drawing les- Asia and Pacific region: the average 2020 HCI for sons from case studies, it shows that a longer-run the region is 20 percent higher in economies also perspective on country trajectories can ­ ­ highlight included in the 2010 HCI compared to the overall promising policies, including the role that a 2020 sample (gross ­ domestic product [GDP] per whole-of-government approach, steady political capita is 88 percent higher). commitment, domestic resource mobilization, and evidence-based policies can play in human capital As measured by the HCI, human capital improved progress. in most economies in the last decade. Figure 2.1 48 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Box 2.1: Ensuring comparability across time in the Human Capital Index The 2020 update of the Human Capital Index (HCI) also reports a version of the HCI calcu- lated for 2010, offering an opportunity to track progress on human capital outcomes. The outcome measures that are used to calculate the HCI typically register only small changes from one year to the next. A time frame of 10 years allows the index to track real underlying change in human capital outcomes over a longer period, smoothing out short-run idiosyn- crasies. The HCI for 2010 is calculated for 103 economies for which comparable data are available, and it provides a benchmark year for economies to measure changes over time as well as the pace of their progress. The data used to populate the 2010 HCI are selected to be “near” 2010 and to maximize comparability with 2020. This comparison is straightforward in the case of child survival rates that are updated annually and adult survival rates that are updated every two years.a Although enrollment rates used to calculate the expected years of school are reported annu- ally for some economies, others may have significant gaps in their time series. In the case of gaps in enrollment for 2010, data are imputed using an annualized growth rate derived from available enrollment data for the economy.b In the case of more sporadically reported stunting and test scores data, the surveys and tests used to populate the two time periods are typically selected to be at least five years apart and as close as possible to 2010 and 2020. In the case of test scores, an additional requirement that both data points come from the same testing assessment program ensures comparability over time—with five exceptions to that requirement. For Algeria, harmonized test scores from the Trends in International Mathematics and Science Study (TIMSS) in 2007 are used to populate the 2010 HCI, whereas harmonized test scores based on the Programme for International Student Assessment (PISA) in 2015 are used to populate the 2020 HCI. For North Macedonia and Ukraine, harmonized test scores from TIMSS in 2011 are used to populate the 2010 HCI, whereas harmonized test scores based on PISA in 2018 are used to populate the 2020 HCI. For Morocco and Saudi Arabia an average of test data from the 2011 TIMSS and 2011 Progress in International Reading Literacy Study (PIRLS) are used for the 2010 HCI, and data from the 2018 PISA are used for the 2020 HCI. To maximize comparability with PISA, only secondary-level scores from TIMSS and PIRLS are used to calculate the 2010 harmonized test scores for these five economies. Finally, although child survival, expected years of school, and harmonized test scores are essential to calculating an HCI, the fraction of children not stunted and the adult survival rate both act as proxies for latent health. Consequently, the HCI can be calculated using either one of these proxies if both are not available.c To ensure comparability in HCI scores over time, the same health proxies are used to calculate both the 2010 and 2020 scores. This means that, if data for stunting are unavailable in 2010, they are not used to calculate the HCI for 2020, and vice versa. a Adult survival rates are the complement of mortality rates for 15- through 60-year-olds, reported for five-year periods by the United Nations Population Division. These data are linearly interpolated to produce the annual estimates for economies used to calculate the HCI. See the section on adult survival rates in appendix C for more details. b The methodology to fill in gaps in enrollment data is described in detail in the expected years of school section of appendix C. c See appendix A for a detailed description of how HCI components are aggregated to calculate the final index. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 49 Table 2.1: Regional coverage of the Human Capital Index over-time sample ECONOMIES WITH A 2020 HCI ECONOMIES WITH A 2020 HCI AND A 2010 HCI Real GDP Number of Real GDP Number of REGION HCI 2020 per capita economies HCI 2020 per capita economies East Asia and Pacific 0.59 23,376 31 0.71 43,977 12 Europe and Central Asia 0.69 35,278 48 0.71 39,479 41 Latin America and the Caribbean 0.56 15,572 26 0.58 18,444 13 Middle East and North Africa 0.57 28,437 18 0.60 34202 14 North America 0.75 55,857 2 0.75 55,857 2 South Asia 0.48 6,605 7 — — — Sub-Saharan Africa 0.40 5,125a 42 0.42 6,586 21 Average, total 0.56 21,403a 174 0.62 30,243 103 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI) for HCI data and the World Development Indicators and Penn World Tables 9.1 for per capita GDP data. Note: The table uses real GDP per capita at purchasing power parity, in constant 2011 US dollars, for most recently available data as of 2019. — = not available. Per capita GDP data for South Sudan are not available. a Figure 2.1: Changes in the Human Capital Index, circa 2010 vs. circa 2020 1.0 Macao SAR, China 0.8 Finland Human Capital Index, circa 2020 Cyprus Switzerland Italy Greece Russian Federation Slovak Republic Albania Bulgaria 0.6 Ecuador Azerbaijan Romania Zimbabwe Panama Togo 0.4 Lesotho Côte d’Ivoire 0.2 0.4 0.6 0.8 Human Capital Index, circa 2010 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the 2020 HCI (on the vertical axis) against the 2010 HCI (on the horizontal axis) for 103 economies for which data are available for both 2010 and 2020. The dashed line is a 45-degree line. Points above (below) represent an increase (decrease) in the HCI between 2010 and 2020. 50 Hu m a n Cap ital Acc u mu lat ion ove r T i m e plots HCI 2020 scores against HCI 2010 scores, capital As incomes increase, on average, human ­ likely reflecting underlying secular trends in improves. Panels a and b of figure 2.2 indicate various dimensions of human capital. On average, ­ the direction of change of the HCI from 2010 the HCI increased by 2.6 HCI points (or 0.026) to 2020, denoted, respectively, by the dots and between 2010 and 2020. For economies in which the arrow points. The slopes of the arrows sig- the HCI scores improved—about 80 percent of nal the rate at which rising per capita income is the sample, depicted above the 45-degree line in associated with more human capital. The pace is figure 2.1—scores increased by an average of 3.5 quite uniform across country income groups. In HCI points. One economy in four that experi- low-income economies, however, human capital enced a rise in the index had increases above 5.0 improved slightly more quickly relative to GDP per HCI points. This means that, in those economies, capita. With health accounting for an important ­ the productivity of future workers approached share of improvement in the index, especially in the frontier by 5 percentage points—a substantial low-income economies (see the next section of this progress. Over time, there is convergence in the chapter), a steeper slope in the HCI–GDP relation- ­ HCI. That is, in economies starting at lower val- ship likely reflects global gains in health, such as ues of the HCI in 2010, human capital improved better and less expensive treatments and improved more rapidly than in economies for which the HCI technology, which benefited all economies but was higher to begin with, even after accounting for brought about larger advances in poorer ones. initial GDP per capita.1 ­ Regional and income group averages mask dif- The economies with the largest gains include Macao ferent individual economy trajectories, which are SAR, China; Albania; the Russian Federation; depicted in figure 2.2, panel c. For example, in Azerbaijan; and Côte d’Ivoire, listed in order of Azerbaijan, human capital outcomes increased by the size of their gain. Various factors account for 0.08 (from 0.50 to 0.58), but there was almost no these improvements: improved learning as mea- change in the country’s GDP per capita. By con- sured by higher test scores (Albania and Macao trast, Lithuania experienced only a small increase SAR, China), better health (in the case of Russia, in the HCI despite a significant increase in per specifically improvements in adult survival, mark- capita income. ­ ing a rebound from the drop in life expectancies in the post-Soviet era; see Smith and Nguyen 2013), Looking back over the last decade shows that both and school enrollment (at the preprimary level in girls and boys have made strides in improving Azerbaijan; at the primary level in Côte d’Ivoire human capital. Sex-disaggregated data are avail- and Macao SAR, China; and at the secondary level able for 90 economies in the comparison over in Russia). time sample (figure 2.3). The average gender ratio is similar in circa 2010 and circa 2020, at about Some economies experienced modest declines 1.06 in favor of girls. This stable average, however, in the index. They include the Republic of Korea, conceals considerable differences at the economy Greece, Bulgaria, and Italy, listed in order of the level. Around 2010, in all but seven economies, size of their decline, where the index fell by about the HCI was higher among girls than among boys. 2 or more HCI points. Among the 10 economies Among those seven economies, the girl-to-boy with the largest drops, 8 are European, and only ratio improved in Cameroon, Chad, and Côte 1 is not a high-income economy. These decreases d’Ivoire approaching full gender parity in the last in the HCI can be traced back mainly to drops in decade. These are the countries in the lower left test scores. quadrant of figure 2.3, above the 45-degree line T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 51 Figure 2.2: Human capital and GDP per capita: Changes over time a. Changes by income group b. Changes by regional group 1.0 1.0 0.9 0.9 0.8 0.8 Human Capital Index Human Capital Index NA 0.7 0.7 HIC EAP ECA 0.6 0.6 LAC MENA UMIC 0.5 0.5 LMIC 0.4 0.4 SSA LIC 0.3 0.3 6 8 10 12 6 8 10 12 Log GDP per capita Log GDP per capita c. Changes at the economy level 1.0 Japan 0.8 Human Capital Index Vietnam Serbia Macao SAR, China 0.6 Qatar Oman Ecuador Zimbabwe Azerbaijan Guatemala South Africa 0.4 Togo Burundi Côte d’Ivoire 0.2 6 8 10 12 Log GDP per capita Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: Panel a (panel b) plots the average HCI for income groups (regional groups) using the World Bank Group classification (on the vertical axis) against log real GDP per capita (on the horizontal axis) for 103 economies for which data are available for both 2010 and 2020. The 2010 HCI is denoted by dots, and the HCI 2020 is denoted by arrow points. Panel c plots economy-level data for HCI 2010 and HCI 2020 (on the vertical axis), represented by dots and arrow points, respectively, against log real GDP per capita (on the horizontal axis) for the 103 economies for which data are available for both 2010 and 2020. EAP = East Asia and Pacific; ECA = Europe and Central Asia; HIC = high-income countries; LAC = Latin America and the Caribbean; LIC = low-income countries; LMIC = lower-middle-income countries; MENA = Middle East and North Africa; NA = North America; SSA = Sub-Saharan Africa; UMIC = upper-middle-income countries. and below the horizontal dashed line. Meanwhile, does not capture gaps in other areas of human in Benin, Burkina Faso, and Morocco, girls fully capital development, such as labor force participa- caught up with boys, even surpassing them in the tion. In many countries, women participate in the latter two economies.2 Among the 83 economies labor force at far lower rates than men. This point in which the HCI was higher for girls in 2010, the is taken up further in chapter 4, which discusses ratio in favor of girls widened in 34 economies; an extension of the HCI capturing labor market however, a favorable girl-to-boy ratio in the HCI utilization. 52 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Figure 2.3: Girl-to-boy ratio, HCI 2010 vs. HCI 2020 1.20 Georgia 1.15 Girl-to-boy ratio in HCI, circa 2020 Lithuania Vietnam 1.10 South Africa Burundi Estonia Timor−Leste Kazakhstan 1.05 United Arab Emirates Congo, Rep. Argentina Montenegro Cyprus 1.00 Cameroon Peru Côte d’Ivoire 0.95 Chad 0.90 0.95 1.00 1.05 1.10 1.15 1.20 Girl-to-boy ratio in HCI, circa 2010 Low-income economies Upper-middle-income economies Lower-middle-income economies High-income economies Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the gender ratio (girl to boy) of the 2020 HCI (on the vertical axis) against the 2010 HCI (on the horizontal axis) for 90 economies for which sex-disaggregated data are available for both 2010 and 2020. The light blue dashed line is a 45-degree line; points above (below) represent an increase (decrease) in the HCI gender ratio between 2010 and 2020. The dashed horizontal (vertical) line indicates gender parity in the HCI in 2020 (2010).  HANGES IN KEY HUMAN CAPITAL 2.2  C in stunting and improvements in adult survival. DIMENSIONS OVER THE PAST DECADE Considered together, progress in child survival, stunting, and adult survival accounts for close  omponent contributions to changes 2.2.1  C to half the increase in the HCI; the remainder in the HCI is explained by changes in education—namely, The evolution of the HCI reflects changes in enrollment and, to a limited extent, learning the components of the index. There are consid- (figure 2.4). erable differences in the pace of change across components and in the extent to which they Although economies in every income group contribute to changes in the overall HCI. Similar experienced an increase in their HCI, the fac- to the analysis for the HCI 2020 cross-section, tors that contributed to these improvements a decomposition3 suggests that almost one-third differ across income groups, reflecting both of changes in the HCI over the past decade are economies’ initial conditions and their devel- due to gains in health, as proxied by reductions opment trajectories. Low-income economies T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 53 in the sample experienced considerable gains gains were offset in some economies by declines survival rates (which, on average, rose in child ­ in measured learning. In high-income econo- from 90.6  percent in 2010 to 93.4 percent in mies, which were already closer to the frontier 2020). Low-income economies also registered for most components, increases in the HCI are growth in enrollment rates in preprimary edu- mostly explained by gains in upper-­ secondary cation (from 26.6 to 42.5 percent) and at the enrollment and improvements in health, as percent). These primary level (from 82.3 to 89.6 ­ proxied by adult survival (figure 2.5). Figure 2.4: Component contribution to Human Capital Index gains, 2010–20 1.00 0.90 HCI points contributed 0.80 0.70 0.60 0.50 0.40 0.30 0.20 0.10 0 Child Preprimary Primary Lower- Upper- Harmonized Health survival enrollment enrollment secondary secondary test scores enrollment enrollment Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note that 2 HCI points corresponds to 0.02 in the HCI 0-1 scale. Note: This figure reports a decomposition computed for 103 economies for which data are available for both 2010 and 2020. Figure 2.5: Contribution to changes in the Human Capital Index, by country income group, 2010–20 Child mortality Preprimary enrollment Primary enrollment Lower-secondary enrollment Upper-secondary enrollment Harmonized test scores Health −0.20 −0.15 −0.10 −0.05 0 0.05 0.10 0.15 0.20 0.25 0.30 0.35 0.40 0.45 Share of change explained Low-income economies Upper-middle-income economies Lower-middle-income economies High-income economies Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: This figure reports a decomposition computed for 103 economies for which data are available for both 2010 and 2020. 54 Hu m a n Cap ital Acc u mu lat ion ove r T i m e  hanges in index components over 2.2.2  C births.5 At an average of 3.6 percentage points, time improvements have been most significant among The analysis in this subsection considers the evo- low-­ ­ income economies, which started out with lution of components of the HCI over the last lower rates. In economies such as Angola, Malawi, decade. On average, there has been progress on Niger, Sierra Leone, and Zimbabwe, improve- most components of the HCI, as illustrated in ments in child survival meant between 39 and 58 table 2.2, which looks at the sample of economies fewer deaths per 1,000 live births.6 with an index in both 2010 and 2020. This progress is the result of global improvements Although the HCI comparison between 2010 and in health but also of a combination of greater 2020 is possible for only 103 economies, com- extension of health coverage, better maternal parisons between these two points in time for and childcare, and better sanitation. For example, individual HCI components are possible for a larger ­ Malawi, where child survival rates increased from (and ­ variable) number of economies. The follow- 91 to 95 percent in the last decade, adopted several ing analysis includes all economies for which data evidence-based policies financed by the govern- are available, in order to provide a comprehensive ment and development partners to improve child dimensions of picture of changes in the different ­ health, including the Accelerated Child Survival human capital. The specific trajectories of individ- and Development Strategy, Child Health Strategy, ual components are discussed below. Integrated Management of Childhood Illness, and newborn a road map to accelerate maternal and ­ Child survival survival. These policies and interventions have Progress in child survival over the past decade has led to improved coverage of essential child health been substantial in many economies, improving services and practices across the country, includ- in 136 of the 173 economies for which data are ing immunizations (at 93 percent in 2014), exclu- average, available, as depicted in figure 2.6.4 On ­ sive breastfeeding (from 44 percent in 2000 to the child survival rate rose from 0.96 to 0.97, 70 percent in 2014), prevention of mother-to- which translates to 10 fewer deaths per 1,000 live child HIV transmission, and oral rehydration Table 2.2: Changes in Human Capital Index components, 2010–20 Europe Latin Middle East and America East and Sub- Asia and Central and the North North Saharan Component  Global Pacific Asia Caribbean Africa America Africa Child survival rate (percentage point 0.007 0.004 0.002 0.004 0.003 0.001 0.022 difference) Expected years of school (year 0.437 0.651 0.176 0.351 0.458 0.440 0.862 difference) Harmonized test scores (score −0.110 −3.659 1.002 8.001 0.443 −5.769 −5.106 difference) Fraction of children under 5 not stunted 0.056 0.048 0.048 0.051 0.034 — 0.070 (percentage point difference) Adult survival rate (percentage point 0.030 0.013 0.020 0.013 0.015 0.002 0.082 difference) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: This table reports changes in regional averages (as defined by the World Bank Group regional classification), computed for 103 economies for which data are available for both 2010 and 2020. — = not available. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 55 Figure 2.6: Changes in probability of survival to age 5, circa 2010 vs. circa 2020 1.00 Hong Kong SAR, China Azerbaijan Egypt, Arab Rep. Guatemala Probability of survival to age 5, circa 2020 Zimbabwe Senegal 0.95 Malawi Burundi Burkina Faso Togo Lesotho 0.90 Chad 0.85 0.80 0.85 0.90 0.95 1.00 Probability of survival to age 5, circa 2010 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the probability of survival to age 5, circa 2020 HCI (on the vertical axis), against the probability of survival to age 5, circa 2010 (on the horizontal axis), for 173 economies for which child survival data are available for both 2010 and 2020. The dashed line is a 45-degree line; points above (below) represent an increase (decrease) in the probability of survival to age 5 between 2010 and 2020. Yellow diamonds in the panels indicate economies for which data are available for both 2010 and 2020, but that are not part of the sample used for the HCI analysis of changes over time because they are missing 2010 comparator data for one of the HCI components. The outlier (yellow diamond at far left) is Haiti, where the probability of survival to age 5 was significantly affected by the 2010 earthquake. for diarrhea (up from 48 percent in 2000 to 64 percentage-point (from 56 to 71 percent, a 15-­ percent in 2014), that have in turn contributed to increase), Eswatini (from 60 to 74 percent, a improve child survival rates (CD2015 2015). 14-percentage-point increase), and India (from 52 to 65 percent, a 13-percentage-point increase). Fraction of children under 5 not stunted The fraction of children not stunted declined in Advances in health over time are also reflected in only a small group of countries: Angola (with a decreases in stunting rates for children under 5, decline from 71 to 62 percent), Malaysia (from though declines have been modest, on average. 83 to 79 percent), Niger (from 56 to 52 percent), The fraction of children under 5 not stunted is Papua New Guinea (from 56 to 51 percent), South available for comparison between 2010 and 2020 Africa (from 75 to 73 percent), and Vanuatu (from for 91 economies, of which 42 are in the 2010–20 74 to 71 percent). HCI comparison sample. Across these economies, depicted in figure 2.7, the fraction of children not The overall trend in stunting between 2010 and stunted increased by about 8 percentage points, on 2020 is consistent with its worldwide decline over average. The economies with the largest improve- variety the past decades. Progress resulted from a ­ ments are Côte d’Ivoire (from 61 to 78 percent, an of factors—not only from overall economic increase of 17 percentage points), Sierra Leone development but also from health and nutrition 56 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Figure 2.7: Changes in fraction of children under 5 not stunted, circa 2010 vs. circa 2020 1.0 Paraguay North Macedonia Albania Fraction of children under 5 not stunted, circa 2020 Azerbaijan Congo, Rep. 0.8 Côte d’Ivoire Malaysia Eswatini Zimbabwe Burkina Faso South Africa Indonesia 0.6 Chad Timor-Leste 0.4 0.6 0.8 1.0 Fraction of children under 5 not stunted, circa 2010 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the fraction of children under 5 not stunted, circa 2020 HCI (on the vertical axis), against the fraction of children under 5 not stunted, circa 2010 (on the horizontal axis), for 91 economies for which data are available for both 2010 and 2020. The dashed line is a 45-degree line; points above (below) represent an increase (decrease) in the fraction of children under 5 not stunted between 2010 and 2020. Yellow diamonds in the panels indicate economies for which data are available for both 2010 and 2020, but that are not part of the sample used for the analysis of changes over time because they are missing 2010 comparator data for one of the HCI components. interventions, maternal education and nutrition, Adult survival maternal and newborn care, reductions in fer- Adult survival rates have been improving steadily tility or reduced interpregnancy, and improved over the last decade. In 2010, 82 percent of ­ determinants of sanitation. Given the multiple ­ 15-year-olds were expected to survive to age 60, necessary. stunting, multisectoral solutions are ­ compared with 85 percent in 2020. Figure 2.8 Some examples are described in box 2.2 (see also illustrates the improvement in adult survival rates Bhutta et al. 2020). Of the economies for which over the last 10 years; most economies are above stunting data are available, 25 are classified as the 45-degree line. Economies with the greatest fragile and conflict-affected situations (FCS). improvements include Eswatini, where survival Although  stunting decreased on average in these rates increased—although from an extremely low economies too, improvements in economies in base—by close to 25 percentage points from 35 to FCS were on the order of 3.6 percentage points; 60 percent, and Zimbabwe, where rates increased in nonfragile economies they were on the order of from 47 percent to 65 percent. Although most of 6.1 percentage points. 7 the economies with large improvements in adult T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 57 Box 2.2: Cross-sectoral interventions to address stunting Through its effects on health and cognitive development, undernutrition early in life stunts children’s development and prevents them from reaching their full potential, in school and during adulthood. According to Bhutta et al. (2020), interventions that target nutrition both from within and outside the health sector—through improvements in maternal education and nutrition, maternal and newborn care, reductions in fertility, or extending interpregnancy intervals—can be effective in reducing stunting in a variety of contexts. The following exam- ples illustrate cross-sectoral engagements designed to accelerate stunting reduction. Madagascar. With rates as high as 60 percent in some regions, stunting is one of the most serious impediments to Madagascar’s socioeconomic development. The World Bank, with cofinancing from The Power of Nutrition, is supporting the government of Madagascar’s efforts to reduce stunting through the Multiphase Programmatic Approach to Improve Nutrition Outcomes (World Bank 2018). This intervention aims to reach 75 percent of children in Madagascar over the next 10 years with a high-impact package of services delivered through a strengthened integrated nutrition and health platform. The program evolves on the basis of lessons learned from the field and on scaling up successful and cost-effective interventions. Madagascar’s social safety net programs also play an important role in addressing child malnutrition and development. The Fiavota safety net program in the drought-affected areas of Southern Madagascar had positive impacts on acute malnutrition, and the Human Development Cash Transfer program has had positive impacts on food security as well as on young children’s socio-cognitive development, including language learning and social skills. Rwanda. Over the past two decades, Rwanda has registered strong progress on poverty reduction and human development. Its child stunting rate, however, remains high at 38 percent, particularly among poorer and larger households. The government has been taking evidence-based action to combat stunting and invest in child development across multiple sectors. Social protection has been central to this effort, striking at the nexus between poverty, vulnerability, and child malnutri- tion. Rwanda’s flagship social safety net, the Vision 2020 Umurenge program, has received sus- tained World Bank support over the years, providing over a million poor and vulnerable people with income support and accompanying measures. In recent years, child- and gender-sensitive safety net interventions were introduced in the Vision 2020 Umurenge program and are now being expanded. They include Nutrition-Sensitive Direct Support and a Co-responsibility Cash Transfer, which targets the poorest households with pregnant women, with children under age 2, or with both, incentivizing them to access essential health and nutrition services. Rwanda’s game plan also includes strengthening high-impact health and nutrition interventions on the supply side, as well as agriculture interventions that improve food security and increase dietary diversity, and preprimary level education interventions. Pakistan. Fill the Nutrient Gap, an innovative analysis by the World Food Programme, identifies the bottlenecks that drive malnutrition across the food system, with a special emphasis on the avail- ability, cost, and affordability of a nutritious diet. Using the Cost of the Diet software developed by Save the Children UK, Fill the Nutrient Gap estimates the minimum cost of a nutritious diet using locally available foods. By comparing this cost to household food expenditure data, the proportion of households unable to afford a nutritious diet is estimated. In Punjab, this exercise highlighted that two-thirds of the population could not afford a nutritious diet, with the largest gap for the poor- est 20 percent who are also targeted by the Benazir Income Support Program. The government of (continued next page) 58 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Box 2.2: Cross-sectoral interventions to address stunting (Continued) Pakistan and the World Food Programme jointly evaluated options to complement a cash trans- fer with nutrition-specific interventions, comparing the impact of market-based interventions with a free provision of Specialized Nutritious Foods (SNF), and SNF provision in combination with a fresh food voucher. Locally produced SNF could be an effective way to reduce the nutrient intake gap caused by nonaffordability (World Food Programme 2017, 2019). For instance, research among pregnant and lactating women and children under 2 by Aga Khan University has found effects on some nutritional indicators. On the basis of this finding, the government of Pakistan, together with development partners, designed a nutrition-sensitive conditional cash transfer pro- gram targeting pregnant and lactating women (until 6 months after delivery) and children up to 24 months old. The program included a combination of antenatal care checkups, immunization, growth monitoring and nutrition education, SNF for women and for children, a small cash transfer to encourage the uptake of the services, and a condition of one child per household enrolled at a time to encourage birth spacing. The program will be piloted before a nationwide rollout. The World Bank will support an impact evaluation to determine cost-effectiveness of interventions. Other initiatives are already ongoing, including a nutrition-sensitive conditional cash transfer pro- gram supported by the World Bank in the Federal territories, Punjab Province, and the merged districts of Khyber Pakhtunkhwa province, as is increasing multisectoral collaboration between the federal government and provincial governments to improve nutrition in Pakistan. survival are in Sub-Saharan Africa, survival also increase in institutional deliveries in RBF- substantially in three economies in improved ­ implementation districts, for instance), led to the Eastern Europe and Central Asia: Belarus (from 79 scale-up of RBF implementation across the coun- to 84 percent), Kazakhstan (from 76 to 84 percent), try (World Bank 2016). Maternal mortality also and Russia (from 75 to 80 percent). saw declines through the improved coverage of maternal health services facilitated by urban and Many factors drive these trends. In Zimbabwe, rural voucher schemes providing care to pregnant improvements were fueled by a combination of women (World Bank 2019). Another potential con- increased resources allocated to the health sec- tributing factor is the decrease in HIV/AIDS prev- tor and a progressive focus on results. This focus alence and reduction in HIV/AIDS-related mortal- included the implementation of results-based ity due to the improved coverage of antiretroviral financing (RBF) approaches in health centers and treatment.8 district hospitals, increasing from 2 rural districts in 2011 to 18 rural districts in 2013, and eventually Eswatini also witnessed some of the largest reaching 60 districts. The RBF in Zimbabwe ini- improvements in adult survival rates during the tially focused on reproductive, maternal, newborn, decade. Still, however, it has the second-lowest and child health indicators and later expanded adult survival rate among non-FCS economies to include HIV/AIDS, tuberculosis, malaria, and in the sample, which reflects the high preva- noncommunicable diseases. The early indications lence of HIV/AIDS, the leading cause of deaths of positive performance under RBF, marked by country (CDC 2019). Eswatini contin- in the ­ increased coverage and quality of key maternal experience the highest rate of HIV/AIDS ues to ­ and child health services (a 13-­ percentage-point prevalence globally, affecting 27 percent of 15- to ­ T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 59 Figure 2.8: Changes in adult survival rates, circa 2010 vs. circa 2020 1.0 Iran, Islamic Rep. Estonia Kazakhstan Botswana 0.8 Russian Federation Malawi Congo, Rep. Adult survival rate, circa 2020 Namibia Cameroon South Africa Zimbabwe Côte d’Ivoire 0.6 Eswatini Lesotho 0.4 0.2 0.4 0.6 0.8 1.0 Adult survival rate, circa 2010 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots adult survival rates circa 2020 HCI (on the vertical axis) against adult survival rates circa 2010 (on the horizontal axis) for 169 economies for which adult survival data are available for both 2010 and 2020. The dashed line is a 45-degree line; points above (below) represent an increase (decrease) in adult survival rates between 2010 and 2020. Yellow diamonds in the panels indicate economies for which data are available for both 2010 and 2020, but that are not part of the sample used for the HCI analysis of changes over time because they are missing 2010 comparator data for one of the HCI components. 49-year-olds.9 The  rate of new infections is also economy among this group. In 2020, the adult the highest in the world, with young women ages survival rate for the United States was significantly 15–24 years five times more likely to be infected below the level that would have been predicted on with HIV than their male counterparts.10 Although the basis of income.12 the crisis is far from resolved, the country has made enormous progress in reducing the number Unsurprisingly, child and adult survival improved of AIDS-related deaths, with a 35 percent reduc- together, reflecting a broad improvement in the tion between 2010 and 2018.11 underlying health status of populations. Adult survival rates declined in only a handful of Expected years of school economies, among which Jamaica experienced the Quantity of schooling, as measured by expected largest decline (less than 1 percentage point). The years of school (EYS), increased by about a half year United States, where adult mortality rose from 106 of schooling (0.47 years to be precise) over the past to 110 deaths per 1,000 15-year-olds, is the richest decade in the 119 economies for which schooling 60 Hu m a n Cap ital Acc u mu lat ion ove r T i m e (figure  2.9).13 data are available in 2010 and 2020 ­ Economies that have experienced a significant levels of income These gains materialized across all ­ increase in EYS over the past decade include (figure 2.10). Low-income economies had the larg- Bangladesh; Burkina Faso; Côte d’Ivoire; Macao est improvement, 0.90 years, mostly due to higher SAR, China; and Togo. In Bangladesh, EYS rose from primary edu- enrollment rates in preprimary and ­ 8.2 years in 2010 to 10.2 years in 2020. Although middle-­ cation. In lower-­ income economies, EYS success, the gov- many elements account for this ­ have risen by an average of 0.81 years, and most fertility likely ernment’s sustained effort to reduce ­ of this increase derives from higher enrollment provided incentives to invest more in children’s secondary education. rates in primary and upper-­ schooling. Girls’ participation in secondary school Upper-middle- and high-income economies, was also stimulated by the Bangladesh Female which had the highest EYS values at the start of Stipend Program, which has helped the country to the period, experienced the smallest increases achieve one of its Millennium Development Goals, since 2010. Among high-income economies, gender parity in education (see Gribble and Voss about 50 percent of the rise can be explained 2009; Rob et al. 1987). by an increase in upper-secondary enrollment; among upper-middle-income economies, the Of the 103 economies with an HCI in 2010 and 2020, rise stems from preprimary and upper-secondary 21 exhibit lower EYS in 2020 than in 2010. Among enrollment. these 21 economies, the median economy lost 0.09 Figure 2.9: Changes in expected years of school, circa 2010 vs. circa 2020 14 Macao SAR, China United Arab Emirates Azerbaijan 12 Egypt, Arab Rep. Expected years of school, circa 2020 Jordan Timor−Leste 10 Togo Guatemala Cameroon Gambia, The Côte d’Ivoire 8 Burkina Faso 6 4 6 8 10 12 14 Expected years of school, circa 2010 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots expected years of school circa 2020 HCI (on the vertical axis) against expected years of school, circa 2010 (on the horizontal axis) for 119 economies for which enrollment data are available for both 2010 and 2020. The dashed line is a 45-degree line; points above (below) represent an increase (decrease) in expected years of school between 2010 and 2020. Yellow diamonds in the panels indicate economies for which data are available for both 2010 and 2020, but that are not part of the sample used for the analysis of changes over time because they are missing 2010 comparator data for one of the HCI components. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 61 Figure 2.10: Contribution to change in expected years of school, by country income group, 2010–20 1.5 Contribution to change in expected years of school (years) 1.0 0.5 0 Low-income Lower-middle- Upper-middle- High-income economies income economies income economies economies Preprimary enrollment Lower−secondary enrollment Primary enrollment Upper−secondary enrollment Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: Based on 103 economies with an HCI for 2010 and 2020. Results are the outcome of a Shapley decomposition at the economy level and averaged by income group. years of school. Enrollment rates have declined in Science Study (TIMSS)/Progress in International some of the richer economies, including Bulgaria, Reading Literacy Study (PIRLS) and Programme Luxembourg, Italy, Romania, and Ukraine. In for International Student Assessment (PISA) tests, Romania, between 2010 and 2020, EYS fell by 0.8 respectively (also see box 2.5 for reforms that vastly years, largely driven by decreases in primary and improved learning outcomes in the state of Ceará upper-secondary enrollment (see box 2.3). in Brazil). Meanwhile, the Arab Republic of Egypt and Lebanon saw their harmonized test scores Learning based on TIMSS/PIRLS decline by about 40 points Progress in learning outcomes as measured by har- (from 399 to 356 and 428 to 390, respectively). In monized test scores has been modest over the past Sub-Saharan Africa, test scores in Cameroon, Chad, decade. Despite challenges in comparing test scores and Madagascar dropped significantly between over time (see box 2.4), harmonized test score data the two rounds of the Program for the Analysis of from comparable testing programs are available Education Systems (PASEC). for 103 economies circa 2010 and circa 2020. The average test score from this sample remained vir- Albania witnessed one of the largest improvements figure  2.11); however, tually unchanged, at 452 (­ in learning outcomes, with harmonized test scores substantial underlying this stable average are ­ increasing from 397 (based on PISA 2009) to 434 improvements and declines in different ­ economies (based on PISA 2018). Albania’s PISA score improve- over the past decade. Of these 103 economies, ments coincide with the launch of intensive reform roughly half (49) saw a drop in test scores (appear- government efforts in its education sector. The ­ ing below the 45-degree line in figure 2.11), whereas launched the National Education Strategy in the other half saw small increases. Among econo- 2004, which was the first attempt to develop a mies with improvements in test scores, Ecuador’s long-term road map for the sector. The National harmonized test score based on the Latin American Education Strategy served as a catalyst for a range Laboratory for Assessment of the Quality of of reforms that continued to be implemented Education (LLECE) test went up by 47 points through the Pre-University Education Strategy from 373 to 420, and Cyprus and Qatar recorded launched in 2014. These reforms include improved gains of about 40 points in harmonized test scores teacher recruitment, compensation, and manage- based on Trends in International Mathematics and ment; a revised curriculum for basic and general 62 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Box 2.3: Why have expected years of school decreased in Romania? Three main factors explain why the expected years of school (EYS) in Romania have declined in the past decade (from 12.7 to 11.8 years). First, in the wake of the financial crisis of 2008– 09, Arts and Crafts Schools, which offered a vocational path as part of upper-secondary education, were closed. Enrollment in these schools dropped without a corresponding rise in enrollment in other types of upper-secondary education. (see figure B2.3.1). Although the resident population of school-age children fell by only 7 percent during the decade, net upper-secondary enrollment rates fell by 10 percent (from 86 percent to 77 percent in 2010–18). In short, the young people who would have enrolled in the vocational schools never enrolled in other schools. In 2015, the three-year vocational path was reintroduced, subsequently helping the system to recover. Second, the number of out-of-school children, including those children of primary school age, has continued to increase during the past decade. Specifically, the number of out-of​-school children ages 6–10 more than doubled between 2009 and 2018, from 43,000 to 98,000. The underlying reasons include persistent underfunding of the sector. Government spending on preprimary and primary education is the lowest among European Union countries (see figure B2.3.2). Moreover, Romania still lacks an early warning system to alert authorities about children who are at risk of dropping out. With the help of the European Commission and the World Bank, work is under way to implement such a system (European Commission and World Bank 2018). Figure B2.3.1: Dynamics in enrollment numbers in upper-secondary education, Romania (index, 2009 = 100) 120 100 Index, 2008–09=100 80 60 40 20 0 0 9 12 3 6 4 5 1 7 –1 –1 –0 –1 –1 –1 –1 –1 11– 10 12 09 13 15 16 14 08 20 20 20 20 20 20 20 20 20 High school Vocational education Residents age 14–19 Source: National Institute of Statistics, Romania 2018. (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 63 Box 2.3: Why have expected years of school decreased in Romania? (Continued) Third, in 2012, the government introduced an additional compulsory year of “prepara- tory” schooling for children reaching the age of 6 before the beginning of the school year (Romanian National Education Law No. 1/2011, article 29, paragraph 2). In 2018, six years after the implementation of the new law, some parents were still postponing enrolling their young children in the preparatory school year. These children also added to the count of out-of​-school children in Romania. Figure B2.3.2: Spending on preprimary and primary education, European Union 5.0 4.5 4.0 3.5 Share of GDP (%) 3.0 2.5 2.0 1.5 1.0 0.5 0 M ce ng ce e L lg ia Un ech ith aria d p ia lg ia La nia n tia Ire alta io lo tria er ia Cr tvia bo in en er m Es ium Po lan s rtu ds G gdo ic P urg ed rk nl y Un Aus ry et y ly xe S al G and H ran d ov d r u ag Fi an Bu an ite Re uan Be en av k n l m pa F an Sl olan N C Ita a Swma Ki ub g e De oa he pr n va to re m m l u Ro S Lu Cz an pe ro Eu 2010 2017 Source of figure: “General Government Expenditure by Function,” https://ec.europa.eu/eurostat/data/database. Source: Contributed by Alina Sava and Lars Sondergaard. upper-secondary education focused on compe- (Ministry of Education and Science, Republic of tencies; enhanced transparency and accountability Albania 2005). through reform of the Matura (grade 12 exam), the national student assessment; reduced price and A question that is often part of policy discussions is quality through a reformed improved textbook ­ whether improvements in school access are associ- procurement process; provision of textbook subsi- ated with drops in learning. This sample shows no dies to the poorest households; a stronger focus on clear correlation between changes in years of edu- inclusive education; and expansion of enrollment cation and test scores; however, changes in learning in preprimary and upper-secondary education and in years of education appear to be positively 64 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Box 2.4: Challenges in test score comparison over time Using the Programme for International Student Assessment (PISA) or the Trends in International Mathematics and Science Study (TIMSS) to compare performance of secondary school students at two points in time may be more complicated in a middle- or ­ low-income economy than in a high-income economy. In settings where secondary school completion is far from universal, selection bias can affect the results, because assessments like PISA and TIMSS test only enrolled students. Youth who are still enrolled in school at age 15 (PISA) or in grade 8 (TIMSS) are generally those from better-off, better-educated households or who have higher ability—which is likely to bias test scores upward (Hanushek and Woessmann 2011). This bias causes problems for comparisons not only across countries but also poten- tially over time (Glewwe et al. 2017). If secondary school participation rises significantly between two test rounds, the students who are newly enrolled on the margin will likely score lower on average. This effect will bias downward the change in scores, which would cause PISA or TIMSS to understate the actual system improvement that a constant sample of students would have experienced over the same period. For the average economy over the past decade, this bias affecting test score improvements was probably not very large. On average in middle-income economies, lower-­ secondary completion rates increased only from 76 percent in 2010 to 79 percent in 2018; in low-­income economies, they rose from 36 percent to 41 percent (according to World Development Indicators). Nevertheless, the bias could matter over longer periods of time or for economies that have increased secondary school participation more rapidly. Source: Contributed by Halsey Rogers. middle- and high-income correlated in upper-­ as per capita income rises. For example, in panel economies and (albeit very weakly) negatively a, child survival rates are plotted against log real correlated in lower-middle- and lower-income GDP per capita. A line connects the solid dots economies. Although this evidence is suggestive 14 indicating the country-group average in 2010 to at best, it points to the need to understand more the arrow points indicating the average in 2020. clearly how education systems can be strength- The lines all slope upward, reflecting the pattern ened in poorer countries to achieve high-quality of improved child survival globally. The lines also learning while they expand access. become shorter as they approach the top of the panel, because there is less room for improve-  imensions of human capital and 2.2.3  D ment. The gradient of the lines is also of interest, economic development reflecting the rate at which outcomes improved Much like the overall HCI, changes in individual with changes in per capita GDP. The steep lines, measures of human capital do not happen in a low-­ such as those for ­ income economies and Sub- ­ vacuum and are correlated with changes in income. Saharan Africa, showcase large increases in child Using a similar visualization as in the previous survival rates despite relatively small gains in per section (see figure 2.2), figure 2.12 illustrates the capita GDP. This pattern is likely a reflection of average improvements in the index components improvements in global health, such as better and T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 65 Figure 2.11: Changes in harmonized test scores, circa 2010 vs. circa 2020 600 Macao SAR, China New Zealand Harmonized test scores, circa 2020 Cyprus 500 Seychelles Slovak Republic Greece Albania Ecuador 400 Morocco Cameroon Egypt, Arab Rep. Madagascar South Africa Chad 300 350 400 450 500 550 Harmonized test scores, circa 2010 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots harmonized test scores circa 2020 HCI (on the vertical axis) against harmonized test scores circa 2010 (on the horizontal axis) for 103 economies for which harmonized test score data are available for both 2010 and 2020. The dashed line is a 45-degree line; points above (below) represent an increase (decrease) in harmonized test scores between 2010 and 2020. Yellow diamonds indicate economies for which data are available for both 2010 and 2020, but that are not part of the sample used for the HCI analysis of changes over time because they are missing 2010 comparator data for one of the HCI components. Box 2.5: Transforming a low-performing education system into the best school ­ network in Brazil Ceará is a northeastern state in Brazil that improved its education outcomes much faster than the rest of Brazil, in just over a decade. Home to 9 million people (4 percent of the pop- ulation of Brazil) and with the fifth-lowest gross domestic product per capita in the country, almost all of Ceará’s 184 municipalities had low levels of quality in teaching and had very limited resources, spending about one-third less in per-student education than wealthier Brazilian states such as São Paulo. Among these municipalities is Sobral (with 200,000 inhabitants), which in the late 1990s suffered from a highly fragmented school system, with many poorly maintained small schools, most in rural areas and with multigrade classes. Despite a reorganization of the school network, a 2005 diagnostic found that 40 percent of grade 3 children were not able to read; 32 and (continued next page) 66 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Box 2.5: Transforming a low-performing education system into Brazil’s best school network (Continued) 74 percent of students in primary and lower-secondary schools, respectively, were over grade- appropriate ages; and 21 percent of lower-secondary school students dropped out. Between 2005 and 2015, Sobral managed to achieve remarkable progress in educational outcomes. In 2005, Sobral ranked 1,366th in education among Brazilian municipalities.a A decade later, it ranked first among 5,570 municipalities in the country in both primary and lower-secondary education, achieving learning outcomes comparable to world-class education systems as measured by the Programme for International Student Assessment (PISA). Today, although its per capita gross domestic product amounts to little over half the national average,b Ceará has the lowest rate of learning poverty in Brazil, and Sobral has some of the country’s best primary schools. Education outcomes in both the region and the municipality exceed all expectations, given the socioeconomic context in which students live and learn: Sobral’s student-to-teacher ratio is relatively high, at 28.9, compared with 21.0 in Ceará and 20.3 on average in Brazil as a whole. These points suggest a high efficiency of the education system. Ceará’s approach was driven by a mix of the following elements, whose effectiveness is supported by international evidence: the provision of fiscal and nonmonetary incentives for municipalities to achieve education outcomes; technical assistance to municipal school net- works to enhance teacher effectiveness and achieve age-appropriate learning; the regular use of a robust monitoring and evaluation system, followed by adequate action; and giving municipalities autonomy and accountability to achieve learning. In Ceará, unlike the rest of Brazil, municipalities are responsible for the entirety of the education provided, from prepri- mary to lower-secondary school (Loureiro, di Gropello, and Arias 2020). A key factor enabling Ceará to emerge as one of Brazil’s top performers in education has been the capacity of state political leaders to insulate education from partisan politics. This has contributed to strong, sustained political leadership committed to improving the ­ quality of education. Sobral organized its education policy under four pillars: continuous use of stu- dent assessments, a focused curriculum with a clear learning sequence and prioritization of foundational skills, a pool of well-prepared and motivated teachers, and a system of auton- omous and accountable school management with school principals appointed through a meritocratic technical selection process (Loureiro, di Gropello, and Arias 2020). The municipality’s goal was to achieve the universal completion of lower-secondary ­ ­ education at the right age with appropriate learning. The results obtained show the effectiveness of goal setting and the importance of political leadership for education outcomes. The COVID-19 (coronavirus) pandemic threatens the progress made by Ceará. A recent study shows that two to three weeks of school closures in São Paulo during the previous H1N1 (novel influenza A virus) pandemic resulted in an estimated two months in learning loss. Using this as a proxy for the COVID-19 pandemic, the paper concludes that an esti- mated two to three months’ school closure could induce a learning loss equivalent to a half-semester of a school year in Brazil (World Bank 2020). Ceará’s progress and the pillars that led it there, however, should help the region tackle the tough job that lies ahead once the pandemic subsides. Source: Based on Cruz and Loureiro 2020; World Bank 2020. a According to Brazil’s Basic Education Development Index, IDEB. b Ceará’s per capita gross domestic product was $8,068 in 2019, compared with $10,666 in Sobral and $15,662 in Brazil (all in purchasing power parity US dollars). T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 67 Figure 2.12: Changes in income and Human Capital Index components, 2010–20 a. Probability of survival to age 5 b. Expected years of school 1.00 14 HIC HIC Probability of survival to age 5 0.98 UMIC Expected years of school 12 UMIC 0.96 10 LMIC LMIC 0.94 8 0.92 LIC 0.90 LIC 6 6 8 10 12 6 8 10 12 Log GDP per capita Log GDP per capita c. Harmonized test scores d. Fraction of children under 5 not stunted 1.0 600 Fraction of children under 5 not stunted 550 Harmonized test scores HIC 500 UMIC 0.8 450 UMIC LMIC 400 LIC LMIC 350 300 0.6 LIC 6 8 10 12 6 8 10 12 Log GDP per capita Log GDP per capita e. Adult survival rate 1.0 HIC Adult survival rate UMIC 0.8 LMIC LIC 0.6 6 8 10 12 Log GDP per capita Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: Each panel plots the component average for income groups using the World Bank Group classification (on the vertical axis) against log real GDP per capita (on the horizontal axis) for economies for which data are available for both 2010 and 2020. The 2010 HCI is denoted by dots and the HCI 2020 is denoted by an arrow point. Panel a shows income group averages for the probability of survival to age 5 for 173 economies for which data were available. Panel b shows income group averages for expected years of school for 119 economies for which data were available. Panel c shows income group averages for harmonized test scores for 103 economies for which data were available. Panel d shows income group averages for the fraction of children under 5 not stunted for 91 economies for which data were available. Panel e shows income group averages for adult survival rates for 169 economies for which data were available. HIC = high-income countries; LIC = low-income countries; LMIC = lower-middle-income countries; UMIC = upper-middle-income countries. 68 Hu m a n Cap ital Acc u mu lat ion ove r T i m e cheaper technologies.15 Conversely, flatter slopes in This subsection discusses child survival, enroll- income economies, Europe and Central Asia, high-­ ment, and fraction of children not stunted disag- and North America suggest smaller gains in the gregated by socioeconomic status, using data from outcome relative to increases in per capita GDP. Demographic and Health Surveys and Multiple The lines are also shorter, because these countries Indicator Cluster Surveys for selected economies were already near full child survival in 2010. with large changes in outcomes in the HCI dataset.16 predominantly Because these surveys are fielded ­ The patterns are similar (upward sloping with income ­ in low- and lower-middle-­ economies, the decreasing slopes as income increases) for adult sur- economies. Figure 2.14 examples come from these ­ vival and the absence of stunting across these income reports human capital outcomes over time, groups, though adult survival and child survival rates disaggregated by socioeconomic status against log ­ share the feature of steeper improvements at low GDP per capita. Panel a shows child survival rates, income levels. Learning in low-income economies panel b shows EYS, and panel c shows the frac- dropped marginally with respect to relatively small tion of children under 5 not stunted. Each panel increases in GDP. It stayed virtually unchanged for depicts the average outcomes for each economy middle- and high-income economies. over time, the outcomes for the richest quintile, and the rates in the poorest quintile. Reconstructing this picture at the economy level in figure 2.13 reveals significant heterogeneity, massive In the case of child survival, Haiti made ­ including dramatic improvements in outcomes survival strides between 2000 and 2012, increasing ­ despite little improvement in income (this is the rates from 86 to 91 percent. Between 2000 and case, for example, of survival in Eswatini). No 2015, survival rates in Malawi rose from 80 to 93 economy, however, showed large GDP improve- percent. In Senegal, rates increased from 87 to 94 ment without at least some improvement in some percent between 2005 and 2015. Although each human capital dimension. country showed declines in child mortality, the different. composition of these changes was quite ­  ocioeconomic differences and 2.2.4  S In Malawi and Senegal, the length of the bars, that progress in human capital is, the gap between rich and poor ­ households, Regional and national averages provide important shortened over time because the increase in the insights into development trajectories over time. average child survival rate was driven by improve- They also, however, mask the differential trends in ments in outcomes among the poorest households. human capital across socioeconomic groups within In Haiti, average rates improved, but the size of the economies, particularly between richer and poorer virtually gap between the rich and poor remained ­ households. The HCI relies on component data constant.17 from administrative sources that cannot readily be disaggregated by socioeconomic status. Survey Trends in EYS show similar variation.18 Burkina data—particularly from Demographic and Health Faso raised the EYS by two years, but the gap Surveys and Multiple Indicator Cluster Surveys— between rich and poor households was maintained also measure child survival rates, enrollment rates, at six years. In contrast, Bangladesh increased the and stunting rates disaggregated by quintiles of average EYS and also cut the gap between the socioeconomic status. Although these survey esti- richest and poorest households in half, from four mates are not always directly comparable with to two years, between 2004 and 2016. Azerbaijan administrative data, they can provide insights into improved the EYS by one year, but the gap the rates of change in outcomes for the richest and between rich and poor households rose from 0.5 poorest households within economies. years to 1.0 year. Box 2.6 offers an example from T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 69 Figure 2.13: Changes in Human Capital Index components and per capita income, circa 2010 vs. circa 2020, cross-country trajectories a. Probability of survival to age 5 b. Expected years of school Luxembourg 1.00 14 Lithuania Ukraine Solomon Islands Hong Kong SAR, China Azerbaijan Probability of survival to age 5 12 0.95 Namibia Expected years of school Gabon Zimbabwe Iran, Islamic Rep. Saudi Arabia Lao PDR Macao SAR, China Togo Eswatini 10 Malawi 0.90 Lesotho Namibia Equatorial Guinea Niger 8 Mali Nigeria Botswana Togo 0.85 Chad Senegal Burundi 6 Côte d’Ivoire 0.80 Chad Haiti 4 6 8 10 12 6 8 10 12 Log GDP per capita Log GDP per capita c. Harmonized test scores d. Fraction of children under 5 not stunted 600 1.0 Fraction of children under 5 not stunted Singapore Korea, Rep. Estonia Turkey Harmonized test scores Malaysia 500 Lithuania 0.8 Libya Luxembourg Gambia, The Turkey r Ecuado Bolivia Iran, Islamic Rep. Senegal Eswatini Uganda Indonesia 400 Indonesia 0.6 Central African Republic Eswatini Kuwait Papua New Guinea Uganda Ecuador Malawi India Malawi Gambia, The Burundi Yemen, Rep. 300 0.4 6 8 10 12 6 8 10 12 Log GDP per capita Log GDP per capita e. Adult survival rate 1.0 China Lebanon United States 0.8 Yemen, Rep. Belarus Afghanistan Russian Federation Adult survival rate Burundi Equatorial Guinea Botswana 0.6 Malawi Namibia Zimbabwe 0.4 Lesotho Eswatini 0.2 6 8 10 12 Log GDP per capita Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: Each panel plots the economy-level averages for each component (on the vertical axis) against log real GDP per capita (on the horizontal axis) for economies for which data are available for both 2010 and 2020. The 2010 HCI is denoted by dots and the HCI 2020 is denoted by an arrow point. Panel a shows the probability of survival to age 5 for 173 economies for which data were available. Panel b shows expected years of school for 119 economies for which data were available. Panel c shows harmonized test scores for 103 economies for which data were available. Panel d shows the fraction of children under 5 not stunted for 91 economies for which data were available. Panel e shows adult survival rates for 169 economies for which data were available. 70 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Sierra Leone of how a well-designed intervention as in the cases of Côte d’Ivoire, the Republic of can contribute to improve education outcomes for Congo, and Uganda. In Côte d’Ivoire, the average the most disadvantaged. fraction of children not stunted increased from 72 percent to 80 percent between 2011 and 2016, The fraction of children under 5 not stunted also but the 25-percentage-point gap between rates for increased in most countries in the last decade, rich and poor households remained unchanged. Figure 2.14: Evolution of Human Capital Index components, disaggregated by socioeconomic status a. SES-disaggregated child survival rate b. SES-disaggregated expected years of school 1.00 12 2006 2000 0.95 2014 10 SES−disaggregated child expected years of school 2015 2014 2012 2007 SES−disaggregated 2015 2012 2012 2010 survival rates 2013 8 2004 0.90 2005 2010 2005 2006 2000 6 0.85 2004 2010 4 0.80 2000 2003 2 0.75 6 8 10 6 8 10 Log GDP per capita at PPP Log GDP per capita at PPP Malawi Haiti Senegal Burkina Faso Bangladesh Azerbaijan c. SES-disaggregated fraction of children not stunted 0.9 fraction of children not stunted 2016 2014 0.8 SES−disaggregated 2011 2011 2016 0.7 2005 2011 2006 0.6 2000 0.5 6 8 10 Log GDP per capita at PPP Uganda Côte d’Ivoire Congo, Rep. Source: World Bank calculations based on Demographic and Health Survey/Multiple Indicator Cluster Survey data as reported in Wagstaff et al. 2019 (for child survival rates and fraction of children under 5 not stunted) and Filmer and Pritchett 1998 and subsequent updates (for expected years of school). Note: The figure plots selected Human Capital Index components disaggregated by quintile of socioeconomic status (vertical axis) against log real GDP per capita (horizontal axis). The solid dot indicates the average across quintiles, and the top (bottom) end of the vertical bar indicates the value for the top (bottom) quintile. Colored bars show the spread of components over time. Panel a shows SES-disaggregated child survival rates for Haiti, Malawi, and Senegal. Panel b shows SES-disaggregated expected years of school for Azerbaijan, Bangladesh, and Burkina Faso. Panel c shows SES-disaggregated fraction of children under 5 not stunted for Côte d’Ivoire, the Republic of Congo, and Uganda. PPP = purchasing power parity; SES = socioeconomic status. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 71 Box 2.6: The immediate effects of providing free education in Sierra Leone Although most children in Sierra Leone start school, few successfully complete their sec- ondary education. As a result, Sierra Leone’s learning outcomes are among the lowest in the world, contributing to a significant human capital gap. The reason most often cited for why children drop out of school is not poor quality, however, but cost. Although out-­ of-​ pocket expenditures on education are fairly low, both in absolute terms and as a percent of household expenditure (about 3 percent across income groups), they can still represent a significant barrier for poor families, especially given that school fees are due in September, ­ at the height of the lean season. The government’s flagship program is the Free Quality School Education (FQSE) program, which was launched in September 2018. It provides selected public schools with block grants (calculated on a per-pupil basis) and school materials, such as textbooks, and it mandates that recipient schools not charge fees. The program seeks to reduce out-of-pocket house- hold spending on education (the “free” component in the program’s name) by eliminating or at least reducing school fees. It also seeks to raise the quality of education (the “qual- ity” component) through the provision of textbooks and other measures. Public messaging around the program has also emphasized the importance of enrolling children in school. Data collected in February and March 2019 allow for an assessment of the effects of free school- ing on out-of-pocket household expenditures and enrollment in the first term of the program, because over 4,000 households that had been interviewed for the 2018 Sierra Leone Integrated Household Survey were reinterviewed then. For each child, the specific school attended for the 2017–18 and 2018–19 school years was recorded and linked to the Annual School Census to determine whether the school benefited from the FQSE program in the first term of 2018–19. The main impact of the FQSE program in the first term appears to have been a substantial reduction in out-of-pocket education expenditures by households. Over 90 percent of students at public primary and secondary schools receiving the block grants report that they do not pay school fees, up from about one-third of primary school students and almost no secondary school students in the prior school year. In addition, about two-thirds of students at public schools not yet supported under the program also report that they do not pay school fees. The financial benefits of the program are shared fairly evenly across the welfare distribution, although the poorest 20 percent of households receive the largest benefit as a percentage of total consumption. Administrative data show a large increase in the number of students, but data collected from households reveal no significant change in net or gross enrollment rates. This discrepancy is not unexpected: a young and growing population like Sierra Leone’s will naturally see an increase in the number of school-age students each year, and the way the program is structured gives schools an incentive to maximize their reported enrollment. Increases in secondary school enrollment can only come over time as more students successfully reach this level. There has been a small rise in the percent of 5- to 7-year-olds who start school for the first time; this increase is concentrated among the poorest households. Although the FQSE project has reduced out-of-pocket expenditures, the most keenly felt barrier to education for households, follow-up research will be needed to see whether such reductions will eventually result in higher enrollment rates at the secondary level and higher levels of secondary school completion, and whether the program will be successful in improving the quality of education these students receive. Source: Contributed by Alejandro de la Fuente based on de la Fuente, Foster, and Jacoby (forthcoming). 72 Hu m a n Cap ital Acc u mu lat ion ove r T i m e By contrast, Uganda reduced stunting while also governments that recognize the importance of modestly closing the rich–poor gap, which nar- investing in the human capital of their citizens, rowed from a difference of 20 to 16 percentage however, the process of designing policy and points between 2000 and 2016. In the Republic building institutions that foster human capital of Congo, the rich–poor gap initially increased, accumulation can be complex, with the benefits with the fraction of children not stunted increas- taking years and even decades to materialize. This ing from 71 percent to 78 percent between 2005 is evident in, among others, the relatively mod- and 2011. The country was able to maintain progress in human capital in the past est average ­ momentum, however, in reducing stunting while decade, as measured by the HCI. also reducing the difference between rich and poor households from 24 to 16 percentage points A comprehensive understanding of how coun- between 2011 and 2014. tries can improve their human capital outcomes requires an analysis that adopts a longer time This analysis highlights that countries vary signifi- frame and identifies the many aspects of gov- cantly in the extent to which gains in human capi- ernment intervention that can lead to positive tal outcomes are distributed across the population. change. By allowing a richer appreciation of Addressing these rich–poor gaps in human capital countries’ development trajectories, identifying must remain a priority for governments because, in the policies and institutions that proved critical many cases, the returns to investment in the human to improving outcomes, and documenting the capital of disadvantaged groups, especially early in challenges involved in maintaining momentum, life, are the highest. Related evidence, however, shows a comparative case study approach offers this income economies, that, among low- and middle-­ depth of information. government redistributive policies do, on average, as good a job of reducing human capital inequality This section presents the experiences of four as does increased national income (D’Souza, Gatti, countries that have made notable improvements and Kraay 2019). At the same time, the experiences in their key human capital indicators over roughly of countries like Bangladesh, Senegal, and Uganda the last decade: Ghana, Morocco, the Philippines, show that countries can sometimes decouple chil- and Singapore. The case studies illustrate how dren’s human capital outcomes from the income policies, programs, and processes that the gov- differences among their households. ernments of these countries adopted improved human capital outcomes, documenting three The following section takes an in-depth look at interrelated aspects of the countries’ trajectories: the experiences of a selected set of economies to continuity—sustaining effort over many political understand how concerted government action can cycles; coordination—ensuring that programs and deliver marked improvements in national out- agencies work together; and evidence—building comes linked to human capital over time and also an evidence base to improve and update human reduce rich–poor gaps within countries to achieve capital strategies.20 greater equity. The four countries featured in this section were selected because they have all prioritized invest- A LONGER-RUN VIEW OF COUNTRY 2.3   ments in key dimensions of human capital in PROGRESS recent years. They vary considerably, however, in their levels of development, their choice of pol- Human capital is a central driver of sustain- icies and programs to develop human capital, and able growth and poverty reduction.19 Even for the outcomes they achieved. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 73 With a score of 0.88, the Southeast Asian island has led to remarkable gains in the health of its state of Singapore is one of the top performers citizens. The government has launched efforts on the HCI. It has a population of 5.7 million and to combat child and maternal mortality while a per capita GDP at 2011 purchasing power parity controlling fertility rates through intensive, (PPP) of US$96,477,21 making it the richest of the sustained family planning programs. A diligent four countries studied here. Singapore has built immunization policy has meant that 91 percent a world-class education system with an increas- children are now fully immu- of Moroccan ­ ing emphasis on analytical skills, teamwork, and nized.25 These efforts have improved human creativity. The success of these efforts is evident country, reflected in capital outcomes for the ­ in the increase of mean years of schooling from an HCI score that increased from 0.45 in 2010 4.7 in 1980 to over 11.2 in 2019.22 In the health to 0.50 in 2020.26 sector, Singapore’s life expectancy at birth increased from 67 years in 1965 to 83 years in Finally, Ghana in West Africa has a population of 2017, and infant mortality has been on a down- 28.8 million and a per capita GDP of US$5,194 at ward slope, from 27 deaths per 1,000 live births 2011 PPP, making it the country with the lowest in 1965 to 2 per 1,000 in 2017. 23 Despite this envi- income in this sample. Despite limited fiscal space, able position, the country’s prime minister has Ghana’s commitment to improving human capital stated that “the job is never done,” identifying and its innovative policies have led to marked healthy longevity and early childhood education improvements in the outcomes of its citizens. as areas for improvement (The Straits Times Since the government ­ introduced education 2020). reforms after a major national economic crisis in 1983, primary enrollment rates have increased The Philippines, with a population of 104.9 substantially, for example, from 67 percent to 95 million, is the eighth most populous country in percent between 2000 and 2017. Increasing school Asia (and the most populous country included in enrollments and increased access to education this analysis) and has a per capita GDP at 2011 have led to an influx of students who are more likely PPP of US$8,123. The country’s HCI score of 0.52 disadvantaged ­ to come from ­ families. Despite this means that children born in the country today influx, Ghana’s harmonized test scores have not will only achieve half their potential. The impor- children under the age of 5 declined. Stunting in ­ tance that governments in the 1970s accorded to has fallen significantly, from 22.7 percent in 2011 mass education in the country jump-started an to 17.5 percent in 2017 (UNICEF, WHO, and World expansion in school enrollment, with primary Bank Group 2020). gross enrollment rates at about 100 percent and rates nearing 90 percent at the secondary level The trajectories of policies in these countries in 2017.24 Although access has increased, quality indicate a strong focus on continuity of govern- remains an issue, with 15-year-old Philippine coordination ment support across political cycles, ­ students scoring lower than students in nearly all between sectoral programs and among different other participating economies in the latest round evidence-​ levels and branches of government, and ­ of PISA in 2018. based policies. Although all four countries did policy directions, the not implement all of these ­ Morocco, located in the Maghreb region of Africa, government case studies point to the whole-of-­ has a population of 35.7 million and a per cap- approach as having enormous potential to build ita GDP at 2011 PPP of US$7,641. The country’s human capital in a wide variety of development commitment to human capital development contexts. 74 Hu m a n Cap ital Acc u mu lat ion ove r T i m e  ustaining political commitment to 2.3.1  S distributed among three accounts owned by the human capital development individual: (1) an ordinary account for housing Continuity of commitment and effort over and retirement purposes; (2) a special account successive governments is key to reaching any that is primarily for retirement; and (3) a long-term goals, but especially in growing Medisave account that is used to cover medical human capital, which can take decades and even expenses. The government supplements the generations. Although not all politically stable contributions of low-income earners through a economies were able to maintain a sustained focus workfare scheme and adds to Medisave savings on human capital, ensuring this continuity is of senior citizens. The Central Provident Fund easier if the economy in question enjoys political has also underpinned health care financing stability, as in the cases of Ghana and Singapore, through Medisave and has fostered citizens’ the former characterized by a stable, multiparty responsibility for their own welfare. Thus, policy democracy since 1992. makers have managed to contain the cost of providing the country’s entire population with By contrast, a consistent approach to building affordable, high-quality primary health care by human capital has been harder to achieve in tailoring subsidies to the patient’s age and ability Morocco, where political commitment to education to pay and charging users high copayments across successive governments did not extend to financed from mandatory health savings other policies critical to improving human capital accounts. Regulation and bulk buying of drugs outcomes. In the Philippines, although several have also kept pharmacy costs in check. successive political administrations have adopted and sustained robust strategies to build the human Levels of funding are crucial, but so is using capital of the population, they have not succeeded resources efficiently. The government of in growing sufficiently the capacity and good Singapore has set a high standard in this respect governance needed to implement these efforts on by ensuring that expenditures are tightly man- the ground. aged, including by imposing severe sanctions for corrupt practices. In addition to political commitment, human capital development requires adequate and sustainable Although successive governments in the Philip-​ funding. In particular, domestic resources are pines have enacted human capital development central to achieving development objectives. laws that reflect principles similar to those espoused Economies can enhance the quality and foster the by more successful economies, they have generally legitimacy of tax systems by strengthening the failed to provide adequate financing to ensure operational capacity of tax administrations. Doing effective implementation. The Philippines spends so can be a challenge for developing economies 4.4 percent of its GDP on its health programs and with limited resources, but some economies have 3.5 percent on education programs, compared found innovative ways to finance the necessary with an average of 6.5 percent and 4.5 percent, policies ( Junquera-Varela et al. 2017). respectively, for an average economy at the same income level. This low expenditure has resulted in For example, the Singapore government has understaffed and overcrowded clinics and schools, mobilized domestic resources through the underpaid providers, inadequate infrastructure, Central Provident Fund, which has played a and a lack of administrative and technical capacity, critical role in financing infrastructure, housing, especially at local schools and health facilities. The and other vital investments. Each individual and absence of adequate funding has also hampered his or her employer make monthly contributions efforts to improve governance. Widespread fraud to the Central Provident Fund, which are in the distribution of textbooks, theft of funds T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 75 or supplies, and ghost workers (workers who are Singapore’s public agencies and statutory paid but do not carry out their jobs) in municipal boards, state-of-the-art digital technology, tech- health facilities are all reflected in the country’s savvy administrators, and experienced teachers outcomes. In the PISA 2018 exam, about four- form a robust data-collection infrastructure that fifths of students (81 percent) achieved lower than feeds critical information to policy makers in a minimum level of proficiency in reading, and a real time. Policy makers use these data to assess similarly high percentage of students performed school and student performance, control costs, below the minimum level of proficiency in help managers and teachers make decisions at mathematics. every level, and conduct workforce ­ planning. For example, the Ministry of Education has The lack of adequate financing—resulting in installed an information-gathering mecha- understaffed facilities, underpaid providers, and nism that helps school administrators assess overcrowded clinics and schools—has particularly the strengths and weaknesses of their own insti- affected the country’s low-income households tution and track student performance (using a and more remote regions, which now lag behind Pupil Data Bank). The system has enabled the the rest of the country in terms of access to ministry to keep closer tabs on how individual services. By contrast, Ghana’s innovative funding schools are faring. mechanism—the National Health Insurance Scheme (NHIS)—was designed to expand primary In Ghana, the government used data to effectively care coverage while also reducing inequity in retarget school feeding efforts under the Ghana access to health care by exempting the poor from School Feeding Program (GSFP) after it found that premiums. 27 The NHIS is funded mainly by a the targeted population (the poor) was not being 2.5 percent value added tax on selected goods and reached. Data from national poverty statistics and services, 2.5 percent from the Social Security and a food security and vulnerability analysis were National Insurance Trust (largely paid by formal combined to refine targeting and reduce leakages sector workers), and the payment of premiums. (Drake et al. 2016; World Food Programme 2013). These funds enable the NHIS to provide prenatal After the retargeting exercise was completed, as and postnatal care, maternal health care, of 2013, about 70 to 80 percent of the GFSP was vaccinations, and health and nutrition education, being received by the poorest communities (World all of which may have helped reduce stunting rates Food Programme 2013). In Morocco, by contrast, in Ghana. As a result of the NHIS, the government a paucity of data has stymied improvements to has been able to devote a high percentage of its the country’s Tayssir conditional cash transfer spending to the health budget (10.6 percent as of program. The Audit Office of Morocco explicitly 2013), which has helped bring down the rate of stated in its 2016–17 report that “no quantifiable childhood stunting in Ghana in both absolute and indicators are available to monitor the different relative terms. programs and prepare annual progress and finan- cial reports that enable evaluation of the perfor-  ollecting and using evidence to 2.3.2  C mance of these programs” (as cited in Benkassmi inform policy making and Abdelkhalek 2020, 15–16) Collecting data to inform policy implementation and design is easier in a compact city-state like 2.3.3  “Whole-of-government” approaches: Singapore than in a sprawling island nation like Adopting coordinated, multisectoral the Philippines, but digital technologies are mak- strategies ing it easier for all economies to collect and ana- Multisectoral strategies are most likely to effectively lyze data and to use the resulting evidence when address the complex underlying determinants of making policies and decisions. human capital outcomes. Policies that cut across 76 Hu m a n Cap ital Acc u mu lat ion ove r T i m e sectors and lines of authority can also be especially women’s empowerment. Specifically, it has helped beneficial to economies such as the Philippines reduce short-term poverty and food poverty at that have limited resources and technical and the national scale by up to 1.4 percentage points administrative capacity. In the last 40 years, each—a substantial reduction, given that pre-4Ps successive governments in the Philippines have rates were 26.4 percent for total poverty and 12.5 adopted policies that involved more than one percent for food poverty (Acosta and Velarde 2015). sector, promoted integrated approaches, and encouraged greater participation by stakeholders in Ghana’s progress in decreasing stunting rates has service delivery. In addition, many policies reflect also been due in large part to the multisectoral the fact that factors beyond the social sectors affect approach taken by policy makers (Gelli et al. human  capital development, such as clean air, a 2019). For example, the GSFP links school feeding safe water supply, and the provision of sanitation programs with agriculture development, especially services. smallholder production, thus helping  create new markets for locally grown food (Sumberg and The country has several programs that are Sabates-Wheeler 2011; World Bank 2012). Thus, the organized on multisectoral lines. An example GSFP spans three different sectors—agriculture, is the Pantawid Pamilyang Pilipino Program education, and health.30 Also, initiatives aimed (4Ps), which provides cash to chronically poor at improving water sanitation and hygiene in households living in poor areas and with schools have helped increase access to water and children between 0 and 14 years old. 28 In return, sanitation, which is a proven factor in improving the beneficiary households are required to health and education indicators. undertake certain activities aimed at improving their children’s health and education, such The experiences of the four countries examined as taking them to health centers regularly, here highlight the importance of sustained effort sending them to school, and going to prenatal to improve human capital outcomes across polit- checkups in the case of pregnant women. Thus, ical cycles, sufficient resource mobilization and 4Ps integrates human capital development with effective allocation across programs, data and poverty reduction efforts. The Department of measurement to inform and design, and multisec- Social Welfare and Development was charged toral strategies that address the complex under- with leading the program’s implementation, lying determinants of human capital. These best and worked with the Department of Health, practices are likely to assume an even greater sig- the Department of Education, the Department nificance in the wake of the COVID-19 pandemic, of the Interior and Local Government, and as economies attempt to mitigate its negative the government-owned Land Bank of the effects on human capital outcomes. Philippines. In addition, 4Ps actively involved local service providers (such as school principals NOTES and midwives) in implementation by tasking 1. All the components of the HCI, and the HCI them with verifying that households were fully itself, are bounded above. For example, adult complying with the prerequisite conditions for and child survival rates cannot be larger than the cash transfers.29 100 percent, and the maximum number of learning-adjusted years of schooling between Impact evaluation studies show that the program ages 4 and 17 is fixed at 14. This means that is resulting in improved education and health the absolute size of improvements become outcomes among beneficiaries, including enhanced smaller as countries get closer to the upper food security, community participation, and bound. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 77 2. In Togo, the gender gap in HCI slightly wid- 11. Data from the Joint United Nations Programme ened in favor of boys. Both boys’ and girls’ on HIV/AIDS Eswatini country profile, outcomes improved during this time period. regionscountries​ https://www.unaids.org/en/­ The widening gender gap is driven by different countries​ /­ /swaziland. rates of improvement among boys and girls. 12. Case and Deaton (2020) connect the decrease In expected years of school, girls’ outcomes in life expectancy in the United States to the improved but by slightly less than those of “deaths of despair” phenomenon. boys. Meanwhile, in child survival and stunt- 13. Refer to appendix C for more details on this ing, boys are catching up to girls, closing the calculation and for details on how enrollment gender gaps toward parity. data are imputed when missing. 3. This decomposition is implemented as a 14. Test scores and years of schooling series are Shapley decomposition. For a description negatively correlated within Latin America of the method see Azevedo, Sanfelice, and and the Caribbean (correlation of −0.16), the Nguyen (2012). Middle East and North Africa (−0.28), and 4. Although most economies experienced Saharan Africa (−0.14). Sub-­ declines in child (under-5) mortality, the rates 15. For the role of technology in the progress rose in Grenada, Mauritius, Fiji, Brunei, and in  child survival, see Jamison, Murphy, and Dominica, reported here in ascending order Sandbuc (2016). of the increase. 16. School enrollment data by age disaggregated 5. For more information, see the United Nations by socioeconomic status are based on the lat- Interagency Group for Child Mortality est update to the household wealth and edu- Estimation website, http://www​­ hildmortality​ .c cational attainment dataset first described in .org/. Filmer and Pritchett (1998). The latest version 6. Although most economies experienced mod- of their dataset contains 345 Demographic est changes in child survival rates, a unique and Health Surveys and Multiple Indicator case is Haiti, where the child survival rate Cluster Surveys, with enrollment rates for 99 dropped massively and abruptly to 79 ­ percent countries over 1990–2017. The child (under- (79 of 100 children survive) in 2010 from 5) mortality rates and stunting rates dis- 92  percent in 2009, following a major earth- aggregated by socioeconomic status come quake. Survival rates have since rebounded to from the latest edition of the Health Equity 94 per 100 children. and Financial Protection Indicators database 7. This differential persists even when the initial described in Wagstaff et al. (2019). Both data- level of stunting and GDP per capita are fac- sets calculate the socioeconomic status index tored in. in the same way, using principal component 8. Data from the UNAIDS Zimbabwe country analysis to aggregate responses to questions profile at https://www.unaids.org/en/regions​ on asset ownership and housing characteris- countries/countries/zimbabwe. tics into a household-level ­ socioeconomic sta- 9. Data from the UNAIDS Eswatini country pro- tus index. file at https://www.unaids.org/en/regionscoun​ 17. Notably, the increase in child mortality in tries/countries/swaziland. 2010 in the aftermath of the earthquake in 10. Data from the United Nations Children’s Fund Haiti was massive. HIV/AIDS web page, https://www.unicef.org​ 18. The EYS data used to calculate the HCI rely on eswatini/hivaids. /­ administrative data on preprimary through 78 Hu m a n Cap ital Acc u mu lat ion ove r T i m e upper-secondary enrollment, covering the and National Insurance Trust. Additionally, 4–17 age range for a maximum of 14 years individuals considered too poor to pay are of school. By contrast, Demographic and also exempt from paying the premium. Health Surveys and Multiple Indicator Cluster They include beneficiaries of the Livelihood Surveys collected enrollment data for children Empowerment Against Poverty program. ages 6–17 for a maximum of 12 years of school. 28. Eligible households received between 500 As a result, the EYS reported in the HCI, cal- pesos and 1,400 pesos (US$11.00–US$32.00) culated using administrative data, cannot be per month, depending on the number of eli- compared to the EYS reported in this section, gible children in the household (King 2020). computed using survey data. 29. In 2009, the Department of Social Welfare 19. The analysis in this section is based on four and Development institutionalized the sys- country case studies produced as part of a tem as the National Household Targeting series titled Building Human Capital: Lessons System for Poverty Reduction, and by 2011 it ­ from Country Experiences (see Benkassmi and had shared the database with the Philippine Abdelkhalek 2020; Blunch 2020; King 2020; Health Insurance Corporation, Department Yusuf 2020). of Agriculture, and Department of Health to 20. This approach is based on that used in the help those agencies better target the bene- World Bank’s Human Capital Project (HCP), fits of their own programs (Fernandez and taking a whole-of-government approach. Olfindo 2011). ­ 30. The GSFP is run by the Ghana School Feeding 21. Based on World Bank national accounts data, Program Secretariat under the direct super- and Organisation for Economic Co-operation vision of the Ministry of Local Government and Development National Accounts data files. and Rural Development. Other public part- 22. Data from Statistics Singapore; see https:// directly involved include the Ministry ners ­ www.tablebuilder.singstat.gov.sg/publicfacing​ of Education, the Ministry of Food and /displayChart.action. Agriculture, the Ministry of Health, the 23. Data from the World Development Indicators. Ministry of Women and Children’s Affairs, the Ministry of Finance and Economic Planning, 24. Based on data from the World Bank dataset, and the ­ District Assemblies. https://data.worldbank.org/indicator/SH.STA​ .STNT.ZS?locations=PH-1W-Z4; https://data​ .worldbank.org/indicator/SE.PRM.NENR?​ REFERENCES =PH-1W-Z4. locations​ 25. Data from Enquête Nationale sur la Population Acosta, P., and R. Velarde. 2015. “An Update of et la Santé Familiale (ENPSF), 2011 and 2018. the Philippine Conditional Cash Transfer’s Implementation Performance.” Philippine 26. Note, however, that, as indicated in appendix Social Protection Note No. 8, World Bank, C, comparisons for learning in Morocco, as in Washington, DC. a handful of other economies, refer to differ- Azevedo, J. P., V. Sanfelice, M. and Nguyen. 2012. ent international testing programs. “Shapley Decomposition by Components of 27. Not everybody has to pay the NHIS premium. a Welfare Aggregate.” MPRA Paper No. 85584, Pregnant women are exempt, as are people Munich Personal RePEc Archive. https://mpra​ under 18 years of age, people age 70 and above, .ub.uni-muenchen.de/85584/. and individuals employed in the formal sec- Benkassmi, M., and T. Abdelkhalek. 2020. tor who contribute to the Social Security “Building Human Capital: Lessons from T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 79 Country Experiences—Morocco.” World Bank, Proposed Action Plan for Implementation.” Washington, DC. Administration Agreement SRSS/S2018/063, Bhutta, Z., N. Akseer, E. Keats, T. Vaivada, S. European Commission, Brussels; World Bank, Baker, S. Horton, J. Katz, P. Menon, E. Piwoz, Washington, DC. M. Shekar, C. Victora, and R. Black. 2020. Fernandez, L., and R. Olfindo. 2011. “Overview “How Countries Can Reduce Child Stunting of the Philippines’ Conditional Cash Transfer at Scale: Lessons from Exemplar Countries.” Program: The Pantawid Pamilyang Pilipino American Journal of Clinical Nutrition 112 (2): Program.” Social Protection Discussion 894S–904S. Papers and Notes 62879, World Bank, Blunch, N.-H. 2020. “Building Human Capital: Washington, DC. Lessons from Country Experiences—Ghana.” Filmer, D., and L. H. Pritchett. 1998. “Estimating World Bank, Washington, DC. Wealth Effects without Expenditure Data— Case, A., and A. Deaton. 2020. Deaths of Despair or Tears: With an Application to Educational and the Future of Capitalism. Princeton, NJ: Enrollments in States of India.” Policy Princeton University Press. Research Working Paper 1994, World Bank, CD2015 (Countdown to 2015). 2015. “Malawi: Washington, DC. Understanding Progress on Child Survival.” Gelli, A., E. Aurino, G. Folson, D. Arhinful, Countdown Country Case Study, CD2015, C.  Adamba, I. Osei-Akoto, E. Masset, July. https://www.countdown2015mnch.org​ K.  Watkins, M. Fernandes, L. Drake, L., and d ocuments/CD_Malawi_July2015_2logos​ /­ H. Alderman. 2019. “A School Meals Program _FINAL2.pdf. Implemented at Scale in Ghana Increases CDC (Centers for Disease Control and Prevention). Height-for-Age during Midchildhood in 2019. “CDC in Eswatini Factsheet.” CDC, Atlanta. Girls and in Children from Poor Households: https://www.cdc.gov/globalhealth​ /countries​ A Cluster Randomized Trial.” Journal of /eswatini/pdf/eswatini-factsheet.pdf. Nutrition 149 (8): 1434–42. Cruz, R., and L. Loureiro. 2020. “Achieving World- Glewwe, Paul, Jongwook Lee, Khoa Vu, and Class Education in Adverse Socioeconomic Hai Anh Dang. 2017. “What Explains Conditions: The Case of Sobral in Brazil.” Vietnam’s Exceptional Performance in World Bank, Washington, DC. Education Relative to Other Countries? de la Fuente, A., E. Foster, and H. Jacoby. Analysis of the 2012 PISA Data.” RISE Annual Forthcoming. “The Immediate Effects of the Conference, Center for Global Development, Free Quality Education Program in Sierra Washington, DC. Leone.” World Bank, Washington, DC. Gribble, J., and M. L. Voss. 2009. “Family Planning Drake, L., A. Woolnough, C. Burbano, and D. and Economic Well-Being: New Evidence Bundy. 2016. Global School Feeding Sourcebook: from Bangladesh.” Policy Brief, Population Lessons from 14 Countries. London: Imperial Reference Bureau, Washington, DC, May. College Press. Hanushek, Eric A., and Ludger Woessmann. D’Souza, R., R. Gatti, and A. Kraay. 2019. “A 2011. “Sample Selectivity and the Validity of Socioeconomic Disaggregation of the International Student Achievement Tests in World Bank Human Capital Index.” Policy Economic Research.” Economics Letters 110 (2): Research Working Paper 9020, World Bank, 79–82. Washington, DC. Jamison, D. T., S. M. Murphy, and M. E. Sandbuc. European Commission and World Bank. 2018. 2016. “Why Has Under-5 Mortality Decreased “A Proposed Early Warning Mechanism at Such Different Rates in Different Countries?” for Early School Leaving Prevention and Journal of Health Economics 48 ( July): 16–25. 80 Hu m a n Cap ital Acc u mu lat ion ove r T i m e Junquera-Varela, R. F., M. Verhoeven, G. P. Edition . UNICEF, WHO, and World Bank Shukla, B. Haven, R. Awasthi, and B. Moreno- Group. Dodson. 2017. Strengthening Domestic Resource Wagstaff, A., P. Eozenou, S. Neelsen, and M.-F. Smitz. Mobilization Moving from Theory to Practice in 2019. “Introducing the World Bank’s 2018 Health Low- and Middle-Income Countries. Directions in Equity and Financial Protection Indicators Development. Washington, DC: World Bank. Database.” Lancet Global Health 7 (1): E22–E23. King, E. 2020. “Building Human Capital: Lessons World Bank. 2012. “What Matters Most for School from Country Experiences—Philippines.” Health and School Feeding: A Framework World Bank, Washington, DC. Paper.” SABER Working Paper 3, Systems Loureiro, A., E. di Gropello, and O. Arias. 2020. Approach for Better Education Results, World “There Is No Magic: The Formula for Brazil’s Bank, Washington, DC. Ceará and Sobral Success to Reduce Learning World Bank. 2016. “Rewarding Provider Performance Poverty.” Education for Global Development to Improve Quality and Coverage of Maternal (blog), July 9, 2020. https://blogs.worldbank.org​ and Child Health Outcomes: Zimbabwe Results- e ducation/there-no-magic-formula-­ /­ b razils​ Based Financing Pilot Program.” World Bank, - and-sobral-success-reduce-learning​ -ceara​ Washington, DC. p overty?token=53176c4095d917916aa31ea735 -­ World Bank. 2018. “Reducing Childhood Stunting b5ceaa. with a New Adaptive Approach.” World Bank, Ministry of Education and Science, Republic of Washington, DC, September 28. https:// Albania. 2005. “National Education Strategy . worldbank.org/en/news/immersive​ www​ 2004–2015.” Ministry of Education and - s t o r y/ 2 0 1 8 / 0 9 / 2 8 / re d u c i n g - c h i l d h o o d​ Science, Tirana. https://planipolis.iiep.unesco​ -stunting-with-a-new-adaptive-approach. .org/sites/planipolis/files/ressources/albania​ World Bank. 2019. “Improving Access to Maternal -education-strategy-2004-2015.pdf. Health for Zimbabwe’s Expectant Mothers.” National Institute of Statistics, Romania. 2018. Feature Story, January 10, 2019. World Bank, Statistical Yearbook 2017. Bucharest: INS. Washington, DC. https://www.worldbank. Rob, U., J. F. Phillips, J. Chakraborty, and M. A. org​ /en/news/feature/2019/01/10/improving​ Koenig. 1987. “The Use Effectiveness of the -access-to-maternal-health-for-zimbabwes​ Copper T-200 in Matlab.” International Journal -expectant-mothers. of Gynecology and Obstetrics 25 (4): 315–22. World Bank. 2020. “The Effect of the H1N1 Smith, O., and S. Nguyen. 2013. Getting Better: Pandemic on Learning. What to Expect with Improving Health System Outcomes in Europe and Covid-19?” World Bank, Washington, DC. Central Asia. Washington, DC: World Bank. World Food Programme. 2013. “State of School Sumberg, J., and R. Sabates-Wheeler. 2011. Feeding Worldwide 2013.” World Food “Linking Agricultural Development to School Programme, Rome. Feeding in Sub-Saharan Africa: Theoretical World Food Programme. 2017. “Fill the Nutrient Perspectives.” Food Policy 36 (3): 341–49. Gap: Pakistan. Summary Report.” World Food The Straits Times. 2020. “World Bank Ranks Programme, Rome. Singapore Tops in Human Capital Index.” World Food Programme. 2019. “Cost-Effective Straits Times, October 15, 2020. https://www​ & Sector-Specific Recommendations for .straitstimes.com/singapore/world-bank-ranks​ Improving Nutrition Outcomes in Pakistan: -singapore-tops-in-human-capital-index. Evidence from the Fill the Nutrient Gap UNICEF (United Nations Children’s Fund), Analysis.” World Food Programme, Rome. WHO (World Health Organization), and Yusuf, S. 2020. “How Singapore Does It.” Building World Bank Group. 2020. Levels and Trends Human Capital: Lessons from Country in Child Malnutrition: Key Findings of the 2020 Experiences, World Bank, Washington, DC. 3 Accumulation Interrupted? COVID-19 and HUMAN CAPITAL T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 83 C OVID-19 (coronavirus) is exacting a heavy gaps between countries, because many may not toll in illness and lost lives, and on the have the infrastructure in place for digital connec- economy. Lacking a vaccine or effective tion. Adding to people’s hardship are household pharmaceutical treatment against SARS-CoV-2, income losses due to unemployment and reduced the novel coronavirus responsible for COVID-19, remittances, with effects that might be quite dif- many countries initially resorted to large-scale ferent across developed and developing countries.1 nonpharmaceutical interventions to slow the virus’s spread. Such interventions resulted in Despite tremendous uncertainty still on the over- economywide lockdowns of different levels of all impact of the pandemic on human capital, it restrictiveness. These measures further ampli- is clear that both direct and indirect pathways disruptions that COVID-19 brought to fied the ­ will matter. Those who were most vulnerable to supply chains and global trade, adding to the begin with are likely to be the worst hit, and many already-dramatic economic dimension of the dimensions of inequality are likely to increase. health crisis. A baseline forecast for gross domestic The next two sections of this chapter discuss chan- product (GDP) in 2020 predicts a global drop of 5.2 nels of impact from COVID-19 to human capital percent (World Bank 2020), the worst recession in and their likely effects over people’s full life cycle. eight decades, which is likely to push 100 million The subsequent section discusses how the Human more people into poverty (Mahler et al. 2020b). Capital Index (HCI) can be used to quantify some of the likely impacts of the pandemic on children A lesson from past pandemics and crises is that and youth. their effects not only are felt by those directly impacted, but also often ripple across populations and in many cases across generations. COVID-19  RANSMISSION OF THE COVID-19 3.1  T is no exception. Both the health and economic SHOCK TO HUMAN CAPITAL effects of the disease and its control measures have significant consequences for people’s human cap- 3.1.1  Health system disruptions ital. In many health systems, the fight against the As governments scramble to respond to the imme- pandemic has crowded out other essential health diate consequences of the pandemic, resources are services. At the same time, people’s fear of infec- likely to be diverted from other health efforts that tion has led many to choose not to seek treatment, nonetheless remain critical. In past health emer- possibly derailing years of gains against diseases gencies, substantial negative indirect effects have like tuberculosis, HIV, and malaria. pandemic- resulted from this crowding out of non-­ related health services. For example, in the 2014– Lockdowns translated into school closures with a 15 Ebola outbreak in West Africa, health facility shift to remote learning in some form, which can closures, health worker deaths, and excess demand in many cases worsen learning gaps between chil- placed on the health system led to further loss of dren with a more affluent background and those lives. In Ebola-affected areas, it was reported that who are less well-off. It can also lead to widening maternal and delivery care dropped by more than 84 Accumulation Interr upted? C OVID -1 9 and H uman Capital 80 percent, malaria admissions for children under will depend on the effectiveness of mitigation from percent, and vacci- the age of 5 years fell by 40  ­ remote instruction, closures will likely result in a nation coverage was also considerably reduced slowdown and loss of learning, and an increased (Elston et al. 2017). likelihood of school dropouts, particularly for the most disadvantaged and for girls (see Azevedo Some of these consequences are already appar- et al. 2020).5 ent for COVID-19. The pandemic has interrupted vaccination programs in roughly 68 economies, ­ These human capital losses are not necessarily uni- and some 80 million children under the age of formly distributed across the population. As chil- 1  year will go unvaccinated in low- and middle-­ dren learn from home, social inequalities become income countries as a result (Nelson 2020; World more salient. The closure of schools could widen Health Organization 2020). Supply-chain break- existing gaps in education between children from downs combine with forced mobility restrictions more well-off homes and those who come from under nonpharmaceutical interventions to com- less well-off backgrounds, because poor house- plicate overall access to vaccines (Nelson 2020; holds’ access to technology and infrastructure is World Health Organization 2020). likely to be more limited. Additionally, learning from home requires more inputs from parents, Children and pregnant mothers are not the only and some parents’ limited capacity to guide and ones who will suffer from weakened service deliv- support their children’s learning could exacerbate ery capacities and curtailed access to services. inequalities. During a pandemic, most people are more reluc- tant to seek medical care. During the SARS epi- Along with education, many children receive other demic in Taiwan, China, people’s fear of infection services through their schools. These include meal likely led to sharp drops in demand for access to programs, which tend to benefit poorer children. medical care across the board (see Chang et al. The suspension of school feeding programs could 2004). In the current pandemic, many patients worsen food insecurity and malnutrition. The suffering from other illnesses will be unable to burden of making up the nutritional shortfall now go for routine checkups, because of restricted falls on parents, many of whom are struggling eco- movement and to avoid COVID-19 infection. Such ­ nomically because of the pandemic (Lancker and ­ service interruption will also likely lead to numer- Parolin 2020). ous deaths, many of them avoidable. For exam- ple, in high-burden countries, it is estimated that Income effects, price effects, and food 3.1.3   deaths due to tuberculosis, HIV, and malaria will security increase by 20, 10, and 36 percent, respectively, The emerging literature on containment strate- over the coming five years.2 A lesson is that, when gies highlights the large benefits—in terms of lives determining how to reallocate resources for pan- saved and GDP losses averted—of testing and con- demic response, special attention must be given tact tracing (Acemoglu et al. 2020). Whereas coun- to maintaining coverage of key non-COVID-19 tries such as Iceland and the Republic of Korea health care.3 successfully implemented these strategies early on in the pandemic, most countries resorted to 3.1.2  School closures lockdowns and movement restriction (Hale et al. By the end of April 2020, schools were closed or 2020). Voluntary mobility restrictions combined partly closed in roughly 180 economies, although with government-driven lockdowns generate a schools are now slowly reopening in many juris- significant drop in activity and aggregate demand, dictions.4 Although the impact of school closures leading to a considerable reduction in incomes. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 85 Nevertheless, the largest impacts to the economy from manufacturing for home consumption, are expected to come from reduced consumption which has now increased, shortages can tem- due to people’s avoidance of social interaction for porarily arise and prices increase as a short-run fear of infection (Wren-Lewis 2020). response (Hobbs 2020). Projections show that the resulting economic fall- Concerns about localized food availability may out will be massive and potentially worse than not be unfounded. Because of mobility restric- that of the 2008–09 financial crisis (Wren-Lewis tions, many farmers may experience labor short- 2020). Lockdowns force many nonessential busi- ages, which can reduce yields and further strain nesses to close and will further disrupt supply the supply of staple foods.6 Small-scale farmers chains. Coupled with inherent uncertainty due to may also choose to avoid going to markets to sell the pandemic, these closures and disruptions may their goods, because they fear contagion. Mobility prompt many people to cut back on expenses, restrictions and labor shortages may also prevent which in turn may trigger more businesses to close farmers from transporting their goods to market, and more people to lose their jobs (International which will likely affect the availability of more Monetary Fund 2020). The ensuing economic perishable crops, such as fruits and vegetables. If decline is likely to undo years of gains in the fight these products cannot reach markets in time, they to eradicate extreme poverty. Accordingly, the may simply rot in the fields, because many farm- World Bank has projected an increase in interna- ers lack adequate storage facilities (Tesfaye, Habte, tional extreme poverty for the first time since 1998 and Minten 2020). (Mahler et al. 2020a). Given that many households will experience a fall Closures and decreased economic activity result in in their incomes, they will likely face food insecu- higher unemployment and income losses for many rity. This situation will affect the poorest house- households. Households in countries that rely holds most, because they devote a larger share of on remittances or seasonal migrants for income their incomes to food expenditures. Households report that contributions from these sources have will respond to such events by limiting their food fallen considerably, and many households report intake or relying more on cheaper staple foods, that they expect to lose their remittances alto- reducing dietary diversity and further worsening gether. The fall in household incomes is likely to the nutrition of millions of people.7 Evidence of affect the poor disproportionately, because they such scenarios is already emerging. For example, often experience more fragile labor arrangements in Senegal, 86 percent of respondents to a tele- and, if inadequately covered by safety nets, are phone survey reported a drop in their incomes, likely to fall through the cracks. and more than 33 percent indicated that they restrict their meals four to seven days a week The income shock will probably be exacerbated by (Le  Nestour and Moscoviz 2020). In Nigeria, the initial price shock already observed in many respondents report fear for their health and their countries. The pandemic has created a short-run financial future, with many also experiencing demand shock, with consumers demanding dif- increased prices of major food items and loss of ferent products. As movement restrictions dis- employment (Lain et al. 2020). In Uganda, house- suade people from venturing out in public, many holds on average report a 60 percent reduction activities that would typically happen in markets, in total household incomes and a drop in food restaurants, or other commercial settings end up expenditures of roughly 50 percent per adult taking place at home. Because manufacturing of equivalent. Evidence from Uganda also points goods for restaurants, hotels, and offices differs toward temporary coping mechanisms used by 86 Accumulation Interr upted? C OVID -1 9 and H uman Capital households, many of which increased borrow- probably come through indirect channels. But ing, dipped into their savings, or invested more of indirect does not mean insignificant. Emerging their time in household enterprises (Mahmud and results for a large set of rapid response phone sur- Riley, forthcoming). 8 veys fielded by the World Bank speak to indirect consequences of the pandemic that may perma- Despite the pandemic’s severe direct health nently weaken countries’ human capital for gener- impacts, the largest effects on human capital will ations (see box 3.1). Box 3.1: Rapid response phone surveys reveal immediate impacts of COVID-19 on the poor Although the impacts of the COVID-19 (coronavirus) pandemic are cross-cutting, they are particularly damaging for the poor and vulnerable. Policy makers need timely and relevant information on the impacts of the crisis as well as on the effectiveness of their policy mea- sures to save lives, support livelihoods, and maintain human capital. To track the socioeco- nomic impacts of the pandemic, the World Bank rolled out a rapid response phone survey (RRPS) in more than 100 economies. Traditional face-to-face surveys are hindered by social distancing protocols and mobility restrictions, whereas RRPSs overcome these limitations. RRPSs can be deployed rapidly, implemented at low cost, used to regularly collect longitu- dinal information, and adapted swiftly to changing circumstances. Preliminary results are available for Ethiopia, Kenya, Mongolia, Myanmar, Nigeria, Tajikistan, and Uzbekistan. Severe mobility restrictions imposed to limit the spread of the pandemic have severely disrupted economic activities. Many workers—especially in the services sector but also in ­ agriculture, for example, in Ethiopia and Myanmar—lost employment. In Kenya, the unem- ployment rate tripled, while in Myanmar, Nigeria, Tajikistan, and Uzbekistan more than one in five households lost all employment. Despite some signs of recovery in employment, especially in Ethiopia, Nigeria, Tajikistan, and Uzbekistan, more than half of households report income losses in the remaining economies. Mainly because of the loss of income, food insecurity increased, often substantially. It tripled in Nigeria and doubled in Tajikistan compared to the previous year. In addition, access to education has been severely limited in most economies, particularly for rural and poor households. In all of these economies, schools were closed and replaced with remote learning activities. Although RRPS questions across countries are not strictly comparable, the survey finds that access and utilization of remote learning activities vary widely. During the pandemic, almost all children are engaged in remote learning activities in Uzbekistan, 7 out of 10 children are learning remotely in Kenya and Mongolia, 6 out of 10 in Nigeria, and only 3 out of 10 in Ethiopia and Tajikistan. The types of learning activities also differ, for example, with Kenyan children mainly studying independently, whereas in Uzbekistan almost 9 out of 10 children report using educational television programs. In most economies, children living in rural or poor households are more affected by school closures because those children have more limited access to remote learning. Access to medical services seems less affected, with 10 percent, over 16 percent, and 25 percent of households unable to obtain medical treatment in Mongolia, Uzbekistan, and Kenya, respectively. Source: World Bank Global Poverty team and https://www.worldbank.org/en/topic/poverty/brief/high-frequency​ -monitoring-surveys. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 87 THE COVID-19 HUMAN CAPITAL 3.2   are some typical age-specific markers for human SHOCK: A LIFE-CYCLE PERSPECTIVE capital development, some of which enter as com- ponents into the HCI. The accumulation of human capital is the result of a dynamic process whose dimensions com- 3.2.1  From conception to age 5 plement each other over time. Depending on an During childhood, the link between parental individual’s stage in life, the impact of the pan- income and child health is particularly strong demic on this process may come through different (Almond 2006). For example, reduced nutrition in channels and have a differential impact. Setbacks pregnant mothers could have a substantial impact during certain stages of the life cycle—chiefly lasting on children in utero, including long-­ early childhood—can have especially damaging impacts on chronic health conditions and cog- and long-lasting effects. For example, economic nitive attainment in adulthood (see Almond and hardship can force families to prioritize imme- Currie 2011). The evidence shows that this is the diate consumption needs, forgoing spending on case for children born during a pandemic but also health or education. Because demand for invest- for children born during conflict10 and economic ing in human capital rises with incomes (Bardhan hardship (Rosales-Rueda 2018). For example, chil- and Udry 1999), a fall in incomes could worsen dren who were in utero during the 1918 influenza human capital accumulation for many people, pandemic had lower educational attainment and especially the most disadvantaged.9 Figure 3.1 income during adulthood. The effect was even schematically how some of these shocks depicts ­ more salient among children of infected mothers can affect the process of human capital accumula- (see Almond 2006). Much about the current virus tion over the life cycle. Across the top of the figure remains to be learned. At the moment, the main Figure 3.1: Human capital accumulation across the life cycle, key stages and metrics Stunting School attainment and learning; School unemployment; Years of healthy life Child mortality; worse health Markers attainment and expectancy; increased low birthweight learning morbidity and mortality College In utero Birth 0–5 5–18 18–60 60+ Working life Shock Increased dropout Displaced rates and decline of Morbidity, care at birth learning due to school stress, and closure and lost isolation income Mother’s Possible Unemployment malnutrition due malnutrition and drop in income to lockdown Source: Corral and Gatti 2020. 88 Accumulation Interr upted? C OVID -1 9 and H uman Capital transmission channel affecting the fetus’s human capita is related to a decrease in infant mortality of capital is expected to be through the disruption of 4.5 percent.12 Recent estimates also show that the health care and through lower household income. 11 relationship between income and child mortal- ity is likely higher in low-income countries, sug- Birthweight is often interpreted as a key observable gesting that short-term aggregate income shocks component of a child’s initial endowment (Datar, translate into an increase in child mortality of Kilburn, and Loughran 2010). Children who were 1.3 deaths per 1,000 children among low-income in utero during the 2008–09 recession were born countries, given a 10 percent decrease in per capita with relatively lower birthweight, particularly in GDP.13 families at the bottom of the income distribution (Finch, Thomas, and Beck 2019). This was the case Stunting rates are also likely to increase because for children born in those California regions that of the COVID-19 shock. Common factors related suffered unusually elevated unemployment rates to stunting are maternal nutrition during preg- after the 2008–09 recession (Finch, Thomas, nancy and nutrition during infancy, both of which and Beck 2019). Similarly, in Ecuador during the will likely worsen if families have less disposable 1998 El Niño floods, children who were in utero income (see Galasso and Wagstaff 2019). A fall and especially in the third gestational trimester in aggregate GDP could also lead to weakened were much more likely to be born with low birth- health infrastructure and less funding for nutri- weight, and these children showed substantially tional interventions and services (see Mary 2018). reduced stature five and seven years afterward (see Existing estimates of elasticities suggest that a 10 Rosales-Rueda 2018). These health effects were percent increase in GDP leads to a decrease in attributed to drops in household income follow- stunting that may range from 2.7 to 7.3 percent.14 ing the devastation of El Niño. Similar outcomes Nevertheless, aggregate elasticities may obscure can unfortunately be expected from the COVID- the fact that many of these shocks will dispro- 19 shock. Because low birthweight is associated portionately affect the poor and disadvantaged. with increased likelihood of malnutrition and with Attention must be paid to ensure that these groups developmental delay, COVID-19-induced income have access to any available support mechanisms effects may substantially affect human capital that may mitigate such impacts. attainment for generations to come (see Black, Devereux, and Salvanes 2007; Lahti-Pulkkinen 3.2.2  The school years et al. 2018). With almost all economies having imposed some type of school closure in response to the pan- Child mortality is unfortunately also likely to demic, students in many settings are likely to increase, for two reasons. The first is the disrup- suffer learning shocks. Evidence suggests that tion in maternal and child health services due any interruption in children’s schooling typically to COVID-19. Early simulated values project an worsens learning outcomes. Such interruptions increase of child mortality of up to 45 percent include disruptions caused by epidemics, con- due to health service shortfalls and reductions flict, natural disasters, and even scheduled school in access to food in 118 low-income and middle-­ vacations. US students’ achievement scores income countries (Roberton et al. 2020). Second, appear to decline by about a month’s worth, on economic downturns have been associated with average, during the regular three-month summer significant increases in child mortality, with a break (Cooper et al. 1996).15 more marked increase in lower-income countries. A  meta-analysis of studies for developing coun- Historical experiences illustrate the impacts tries suggests that a 10 percent increase in GDP per of large-scale school closures during a public T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 89 health emergency. Studying the effects of the damage. Cash transfers played an important 1916 polio pandemic on educational attainment ­ mitigating role: four years after the earthquake, in the United States, Meyers and Thomasson households near the fault line were indiscernible, (2017) find that young people ages 14–17 during in welfare terms, from those farther away from the pandemic later showed reduced overall edu- the fault line. Enrollment rates for children resid- cational attainment compared to slightly older ing near the fault line were not affected. Despite peers. Even short-term school closures appeared the apparent return to normalcy, however, test to have lasting impacts on children’s educational scores for children living 10 kilometers away from attainment, though the study finds such effects the fault line were 0.24 standard deviations below only among children who were of legal working those of children residing 40 kilometers away.16 age during the school closures. Many countries have adopted distance learning Increased dropout rates are one factor linking as a means to mitigate learning losses during pro- emergency school closures to future losses in tracted school closures. Remote teaching strate- lifetime educational attainment. In general, as gies include not only online learning but also radio children age, the opportunity cost of staying in and TV programs and text nudges in those coun- school increases, which may make it harder for tries where digital connectivity is limited. These households to justify sending older children back strategies make it less likely that negative effects to school after a forced interruption, especially if of similar magnitude to other interruptions will households are under financial stress. Again, such be replicated; however, the effectiveness of these effects are not restricted to public health emergen- measures has yet to be determined. cies. Among agricultural households in Tanzania, income shocks, even transitory ones, have led to The most recent global projections on the impact increased child labor and reduced school atten- of school closures linked to COVID-19 suggest that, dance (Beegle, Dehejia, and Gatti 2006). using the HCI metric of learning-adjusted years of chapter 1), closures schooling (LAYS; see box 1.1 in ­ Evidence from natural disasters confirms that will result in almost 0.6 years lost. These num- interruptions and trauma in the neurodevelop- bers reflect the loss of schooling that comes from mental process can adversely affect academic per- potential dropouts due to the loss of income,17 as formance. Four years after bushfires in Australia, well as the adjustment in quality due to worsened children from areas that were heavily affected by learning because of inefficient remote teaching the fires performed worse in reading and numer- methods (Azevedo et al. 2020). The lost schooling acy than did their peers from less-affected areas in the face of a mitigation strategy that has medium (Gibbs et al. 2019). The case of the bushfires under- efficiency translates to a yearly loss of over US$872 scores the importance of continued support to in 2011 purchasing power parity US dollars, reach- affected populations, because a longer-term learn- ing a loss of US$16,000 in lifetime earnings in ing divergence was found even though students present value terms at a discount rate of 3 percent did not display any differences in learning out- and assuming a 45-year work life (Azevedo et al. comes immediately after the disaster. 2020).18 As children head back to school, countries with an already overextended education system Further indication of the damage caused by school may grapple with increased demand for public interruptions can be gleaned from outcomes after education, because household income losses have the 2005 earthquake in Pakistan. Areas near the prompted many parents to turn to public schools fault line were devastated, 80 percent of homes rather than private. In June 2020, registration in were destroyed, and schools suffered considerable public schools in the coastal zone of Ecuador, for 90 Accumulation Interr upted? C OVID -1 9 and H uman Capital example, increased by 6.5 percent, bringing some With students in low- and middle-income 120,000 additional students into the public sys- economies less likely to have internet access, tem. This increase occurred despite the govern- between-country inequalities in learning will ment’s offering of a 25 percent subsidy on monthly worsen. Within economies, those at the bottom of private school tuition for parents who had lost jobs the income distribution will also be more affected, (Olsen and Prado 2020). With limited numbers of because they lack access to the necessary materi- qualified teachers available, migration of students als for remote learning. This disparity will again from private to public schools could worsen learn- exacerbate existing inequalities in human capital ing outcomes across many countries. accumulation. Thus, the impacts of school closures extend far Two opposing forces may influence tertiary enroll- beyond initial enrollment drops. For girls, school ment rates. Pandemic-induced high unemploy- closures may also lead to increased exposure to ment rates are likely to reduce the opportunity pregnancy and sexual abuse. In many countries cost of attending college. At the same time, the this outcome could be worsened by policies that recession will affect many households economi- prevent “visibly pregnant girls” from attending cally, and funds for attending college may not be school (Bandiera et al. 2019). Both shorter- and available. After the 2008–09 financial crisis, enroll- longer-term impacts are likely to affect disadvan- ment rates for tertiary education in the United taged families most, further widening inequali- States went up. Because of a substantial decrease ties in learning and human capital accumulation in family incomes, however, students shifted away between socioeconomic groups. from four-year private colleges toward two-year public institutions (Dunbar et al. 2011).19 Finally, a drastic change in the day-to-day lives of children and adolescents is likely to affect their Those who graduate from college now are also mental health. The pandemic may worsen exist- likely to suffer short- to medium-term wage losses. ing mental health issues by provoking or exacer- Evidence from Canada suggests that graduating bating social isolation, economic uncertainty, and during a recession is linked to significant initial fear (Golberstein, Wen, and Miller 2020). A recent earning loss due to less desirable job placements study among Ecuadorian teenagers (ages 14 to 18) but that this penalty fades over some 8 to 10 years found that one in six teenagers reported suffer- (Oreopoulos, von Wachter, and Heisz 2012). ing from depression, and many cited household Nevertheless, the average effect hides substantial finances and social isolation as concerns (Asanov heterogeneity. Recent graduates with the lowest et al. 2020). The use of digital technology, partic- predicted earnings are likely to suffer the largest ularly with voice and video, can ameliorate the losses and often do not recover the lost ground loneliness faced by many teens and children, but after 10 years.20 Starting at a lower-paying job or these technologies are not available to all (Galea, at a less-desirable firm that does not make full use Merchant, and Lurie 2020). of an individual’s existing human capital may well lead to a lag in skill accumulation and result in a  chool-to-work transition and tertiary 3.2.3  S persistent disadvantage. education The pandemic is also disrupting human capital Women who graduate from high school during accumulation for students currently in tertiary the pandemic may choose to respond differ- education. Almost all students currently enrolled ently than their male peers to the shock and in tertiary education are experiencing a new forgo college in the short term. They are also learning modality (Bassett and Arnhold  2020). ­ less likely to join the workforce because of the T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 91 depressed  wages. Evidence from the United For many, the current shock will be the second States suggests that women, but not men, grad- “unprecedented” economic shock of their work- uating from high school are more likely to skip ing lifetime. Workers who have longer tenures in college during recessions because of the lower a company, if dismissed, are likely to face a con- observed returns to education and because the siderable erosion of skills, because many skills cost of more schooling increases (Hershbein they have accumulated may be particular to that 2012). For some, the alternative of child rearing employer. If these workers find employment in may be more attractive in the short term, as was the future, and their new job requires different the case during the 2008–09 global recession. For skills, they are likely to experience a considerable others, disruptions in supply chains may lead to wage penalty (Poletaev and Robinson 2008). Those unintended pregnancies because many women who lose a job during a mass layoff event are likely will lose access to modern contraceptives. 21 to experience large and persistent earning losses, roughly equivalent to 1.7 years of their earnings Finally, the depressed wages and fewer legal before dismissal (Davis and von Wachter 2011). employment options available during a recession mean that crime also becomes more attractive. Evidence also suggests that those who have lost The longer the recession lasts, the more likely that jobs during the pandemic could suffer far more acquired human capital depreciates and crime than lost earnings. One study finds that US work- becomes a worthwhile option. This effect is height- ers who were employed in a firm for at least six ened for those who have lower human capital years and were then dismissed during a recession levels and are less attached to the labor market.22 ­ had higher mortality rates than similar workers who had not been displaced. The estimates sug- 3.2.4  Working life gest an average decrease in life expectancy for the Together with its effects on the economy, the pan- dismissed workers of between 1.0 and 1.5 years, demic has affected labor markets dramatically. likely due to increased chronic stress (Sullivan and According to the International Labour Organization, von Wachter 2009). working hours during the first quarter of 2020 declined by the equivalent of 130 million full-time The pandemic and the nonpharmaceutical inter- jobs. The organization expects that the results will ventions taken to control it are also likely affecting be even worse in the second quarter of 2020, with women more than men. The sectors typically most the number climbing to 305 million full-time jobs affected by lockdowns have high shares of female (International Labour Organization 2020). The employment (Alon et al. 2020). School closures pandemic and lockdown measures are affecting will likely contribute to heavier workloads for workers worldwide but are having particularly dra- many women, mostly because women are likely matic impacts for informal workers. Informal work to be responsible for childcare in the absence of often happens in crowded places, so that lockdown alternatives. These pressures may limit women’s measures—when enforced strictly—make con- paid work (Wenham, Smith, and Morgan 2020). tinuing with these jobs impossible (International Established gender norms are also likely to prevail Labour Organization 2020). Informal workers when a family member falls ill from COVID-19, also often fall through the cracks of social protec- with women in the household expected to care for tion systems, lacking access to unemployment and the sick. Conversely, the current shift to flexible health insurance (Packard et al. 2019). working arrangements could benefit some work- ers, including women, and could promote gender Unemployment stints, even short ones, tend to equality in the labor market in some settings (Alon leave a lasting mark on individuals’ earnings. et al. 2020). 92 Accumulation Interr upted? C OVID -1 9 and H uman Capital Beyond work, interpersonal violence is also on increase human capital resilience. Doing so could the rise, leaving many women more exposed mean not only rethinking policies and services because of the lockdown (van Gelder et al. 2020). for today’s elders but also supporting younger Evidence of this increase has already surfaced. For generations to prepare for a healthy longevity in example, in Argentina, lockdown restrictions were the future. This support will involve stepping up directly linked to a 28 percent increase in calls to prevention of noncommunicable diseases such as the domestic violence hotline. Additionally, also cardiovascular diseases, obesity, and diabetes. in Argentina, women whose partners were also in quarantine were more likely to report an increase in interpersonal violence due to increased expo-  SING THE HCI TO SIMULATE THE 3.3  U sure to the perpetrator (Perez-Vincent et al. 2020).23 IMPACT OF THE PANDEMIC There also is evidence of this effect in India, where domestic violence complaints increased most The HCI is designed to capture the human capi- in regions that implemented a stricter lockdown tal a child born today can expect to attain by age (Ravindran and Shah 2020). 18. Given that the future is uncertain, the best approximation of human capital accumulation 3.2.5  Older adults for a child born today is based on the currently The risk of adverse health effects from COVID-19 observed outcomes of older cohorts. Despite increases significantly with age and ­ comorbidities, uncertainty about how long it will take for the making the elderly especially ­ vulnerable. Residing world to arrive at a post-COVID-19 “new nor- ­ in a long-term care facility also s ubstantially mal” (and what the world will look like then), for increases risk. For example, preliminary analysis the purpose of the long-term outcomes captured of April 2020 COVID-19 exposure data in Italy by the HCI, the pandemic is mostly a transitory indicated that 44 percent of infections during shock. For example, although school closures this period were contracted in nursing homes or affect school-age children now, they are unlikely homes for the disabled (COVID-19 Task Force to affect children who are born today, assuming 2020). In the United States, as of mid-May 2020, that the pandemic has been controlled and school nursing home residents accounted for about is in session by the time those children are ready one-third of COVID-19 fatalities (see CDC 2020; to start school. Yourish et al. 2020). Although such findings are alarming, they probably underestimate actual The disruption to health systems and shocks to infection and case fatality rates among older family income will, however, affect young chil- adults, because there is evidence that, especially dren’s survival and healthy development (stunting) at the beginning of national epidemics, deaths now. In turn, this health outcome will affect their from COVID-19 went unrecorded in many long- learning and schooling. Because all the data for the term care facilities. 2020 HCI were collected before the virus struck, it serves as a pre-COVID-19 baseline, and the HCI An immediate priority for countries fighting construct can be used to simulate the direct and COVID-19 is to protect the elderly and those with indirect impacts of the pandemic on young chil- significant comorbidities. Prevention, control, dren’s human capital.25 Over time, it can be used appropriate staffing, coordination, management, to track the actual changes in human capital out- reporting, communication, and planning are all comes as the pandemic evolves. needed to safeguard older adults living in residen- tial facilities.24 In the longer run, the vulnerabil- The rest of this section discusses an example of a ities revealed by COVID-19 point to the need to simulation of the effects of the pandemic shock T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 93 on the future human capital of young children 1.16 points.27 With about 10 years of schooling under 5 years. Then it uses the HCI to simulate and HTS of 370, the losses due to an increase in how the pandemic—through school closures and stunting could amount to nearly 1 percent of the shock to family income—will affect the future HCI. Add in the likely increase in child mortality human capital of children who are currently in ­ ervice disruptions, and the income due to health s school. drop would further drive down the HCI by an additional 0.10 to 0.47 percent, depending on the 3.3.1  Shock to children under 5 assumptions. Altogether, a decline in income of COVID-19 is seemingly not as directly damag- 10 percent could lead to a decline in the HCI rang- ing to the health of children or pregnant moth- ing from 1.13 to 1.50 percent. previous pandemics have been (Almond ers as ­ and Currie 2011).26 The associated economic Annex 3A reports the methodology of the sim- shock, however, is expected to be harmful for ulation in more detail. For each economy, the the youngest children and children in utero, percentage decline in income due to COVID-19 considerable drops in family income because ­ is estimated as the difference between projec- can translate into food insecurity, in turn lead- tions of per capita GDP growth made in June ing to increased child mortality and stunting. 2020 and the pre-COVID-19 projections made An additional shock is the decrease in coverage in late 2019. The calculations described above of essential health interventions for pregnant are then applied to simulate the likely effects on mothers and young children. This decrease is human capital as measured by the HCI for each due to health services disruptions. These shocks economy. Averaged across all economies, the too will affect child mortality and child health. projected shock would result in an HCI loss of Mapping them into changes in human capital as 0.44 percent. This outcome is worse for low- and measured by the HCI requires estimates of how lower-middle-income economies (losses of 0.73 mortality and stunting change in response to and 0.64 percent, respectively), mostly because shocks to GDP per capita, as well as to reductions the stunting rates are highest for this group (table in health services. 3.1).28 Although the loss may not seem large, it will likely set back children within the affected Because child health (captured by worsened stunt- cohort for years to come, leading to accumu- ing rates) and educational outcomes are closely lated losses. For example, by 1960, the cohort of intertwined, the shock is also expected to affect the adults who were in utero during the 1918 influ- amount of education this cohort of children will enza pandemic had 0.1 fewer years of education attain in the future as well as how much education than those adults born in the year before or after they can retain. Take a reduction in GDP per c ­ apita the pandemic. When comparing the wages of the of 10 percent—a pessimistic scenario. An elastic- same groups in 1960, those who were in utero ity of stunting to income of –0.6 would imply an during the influenza pandemic had wages that increase in stunting of 6 percent. For example, were 2.2 percent lower than those of the neigh- for a country like Bangladesh, where the pre- boring cohorts (Almond 2006). These losses COVID-19 stunting rate was 31 percent, an income build up over time and leave affected cohorts at a shock of this magnitude could increase stunt- considerable disadvantage. ing by 1.85 percentage points. Children who are stunted are less likely to stay in school and learn, 3.3.2  Schooling and learning so this increase in stunting could in turn lead to Through school closures, the human capital of a drop in expected years of school of 0.03  years, the current cohort of school-age children is being and a drop in harmonized test scores (HTSs) of heavily affected by the pandemic: at its peak, nearly 94 Accumulation Interr upted? C OVID -1 9 and H uman Capital Table 3.1: Simulated drop in Human Capital Index due to the pandemic’s impacts on children 5 and under Percentage point difference between June Percent drop 2020 and AM19 GDP per capita growth HCI 2020 in HCI projections World Bank income group       High income 0.707 −0.17 −6.23 Upper-middle income 0.560 −0.42 −7.77 Lower-middle income 0.480 −0.64 −5.36 Low income 0.375 −0.73 −4.34 Global 0.561 −0.44 −6.16 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: Calculations are based on the methodology presented in annex 3A. Projected GDP changes are from the June version of the Global Economic Prospects and the 2019 Macro Poverty Outlook (Annual Meetings of 2019 [AM19]). The simulation assumes a four-month interruption to health system access. The access scenario comes from Roberton et al. (2020) and assumes considerable reductions in the availability of health workers and supply due to pandemic-driven reallocations. The scenario also assumes reduced demand due to fears of infection and movement restrictions as well as economic pressure. The scenario from Roberton et al. (2020) does not include additional child deaths due to wasting. 1.6 billion children worldwide were out of school.29 assumption that the losses due to school closures The simulation framework proposed by Azevedo are not recuperated.32 et al. (2020) quantifies the effects of this shock on the global stock of schooling and learning through Expected years of school are also projected to fall three channels: during closures children (1) lose because the income shock will likely cause many out on opportunities to learn, (2) may forget what children to drop out of school.33 COVID-19 could they have previously learned, and (3)  may drop lead an additional 6.8 million children to drop out out because of household income losses. of school around the world. Sixty percent of these dropouts will be children between 12 and 17 years According to the 2020 HCI, before COVID-19 of age, who are likely to leave school permanently the global average of the expected years of school because of losses in household income. Overall, was 11.2 years, which, when adjusted for learning, the number of out-of-school children is likely to translated into 7.8 LAYS. To simulate the effect increase by 2 percent. of closures on learning, the simulation starts by assuming a value of learning gained in one year of Take, for example, an upper-middle-income schooling. This value is proxied by HTS points per economy like Peru, and assume a drop in GDP per year.30 To determine how much of this learning capita of 10 percent. The drop in GDP and rise in will be lost as a result of closures, the simulation the rate of dropouts will result in a small loss in assumes three scenarios: optimistic, intermediate, expected years of school (0.005).34 School closures and pessimistic, corresponding to three, five, and and the limited capacity of economies to deliver seven months of school closures, respectively. The education during the closures lead to an additional three scenarios also differ on the assumed effec- loss to years of school. Assuming, in the optimis- tiveness of the mitigation measures put in place tic scenario, that schools are closed for 3 months by governments, which vary by income group.31 out of a 10-month school year, without any mit- The three components are used to project HTS igation, students would lose 0.3 years of school. points lost because of school closures under the Assuming instead, in this scenario, a mitigation T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 95 effectiveness of 0.4, only 60 percent of that period Across the globe, the extent of this shortfall will would be lost, leading to a loss of only 0.18 school vary. In high-income economies, where chil- (1.0 – 0.4)). A further dimension of loss years (0.3*  dren were expected to complete 10.3 LAYS comes from the drop in learning. Children in before the  pandemic, the simulations suggest income economies like Peru gain upper-middle-­ that COVID-19 could lower LAYS to 10.1 in the 40 HTS points in a school year. With students optimistic scenario and 9.2 in the pessimistic missing out on 0.18 years of school, they are also scenario. At the other end of the spectrum, chil- losing 7.2 HTS points (40*0.3*(1 – 0.4)). Putting it dren in low-income economies were expected to all together means a loss of 0.27 LAYS. complete 4.3 LAYS before COVID-19. The opti- mistic scenario suggests that this value would Figure 3.2 depicts the combined losses in learn- fall to 4.1 years, whereas the pessimistic scenario ing and expected years of schooling for different foresees a decline to 3.8 years. country income groups. Under the intermediate scenario of a five-month closure, COVID-19 could Putting these losses in LAYS in the context of the lead to a loss of 0.56 years of school, adjusted for HCI implies a 4.5 percent drop in human capital quality. This scenario means that school closures for children of school age, which is on the order due to COVID-19 could bring the average learning of the average gains in human capital made in the that students achieve during their lifetime down past decade (table 3.2). to 7.3 LAYS.35 In the optimistic scenario, the pro- jected loss is 0.25 years of schooling and, in the What is known about the virus itself continues pessimistic scenario, 0.87 years. to evolve, making many behavioral patterns Figure 3.2: Learning-adjusted years of schooling lost because of COVID-19 school closures and income shock High income Upper-middle income Lower-middle income Low income Global –1.2 –1.0 –0.8 –0.6 –0.4 –0.2 0 Learning adjusted years of school lost Pessimistic Intermediate Optimistic Source: Azevedo et al. 2020. Note: Results based on latest available learning-adjusted school years for 174 economies (unweighted average). Coverage of 99 percent of the population ages 4–17 years using the methodology from Azevedo et al. (2020). Projected GDP changes are from the June version of the Global Economic Prospects. School closure length for each scenario: seven months for pessimistic, five months for intermediate, and three months for optimistic. Mitigation effectiveness also differs by scenario and income group. Refer to Azevedo et al. (2020) for full details. 96 Accumulation Interr upted? C OVID -1 9 and H uman Capital Table 3.2: Human Capital Index shock to children currently in school during the pandemic Percent drop in HCI based on GDP per capita Percent drop in HCI if GDP per capita for   projected change as of June 2020 ­ economies dropped by 10 percent all ­ World Bank income group     High income −5.17 −5.34 Upper-middle income −4.71 −5.04 Lower-middle income −4.00 −4.54 Low income −3.07 −3.66 Global −4.45 −4.82 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI) and on Azevedo et al. 2020. Note: The calculation is based on the method presented in annex 3B. Projected GDP changes are from the June 2020 version of the Global Economic Prospects. Table 3.3: Human capital loss of the workforce in 2040 HCI drop based on GDP per capita HCI drop if GDP per capita for all   projected change as of June 2020 ­ ­ conomies dropped by 10 percent e World Bank income group     High income −0.011 −0.012 Upper-middle income −0.009 −0.009 Lower-middle income −0.007 −0.008 Low income −0.005 −0.006 Global −0.0084 −0.0093 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI) and on Azevedo et al. 2020. Note: The calculation is based on the methodology presented in annexes 3A and 3B. Projected GDP changes are from the June 2020 version of the Global Economic Prospects. difficult to predict. For instance, parental con- school or under the age of 5 during the COVID-19 cerns about child and family safety will likely pandemic.36 Assume that the 2020 HCI summa- dominate household decision-making around rizes well the human capital children under the sending children back to schools when they age of 5 could have achieved, and that the 2010 reopen. Hence, any estimates of dropouts that HCI is the best representation of the human cap- consider only the relationship between incomes ital children who are currently in school could and school dropouts are likely to underestimate have achieved. With the HCI losses as calculated the extent to which the pandemic will affect in the earlier sections, the HCI of the workforce children’s schooling and learning. Additionally, in 20 years’ time in the typical economy would these numbers ignore the possibility of remedi- be lower by almost 1 HCI point (0.01) because of ating these losses. COVID-19 today.37 3.3.3  The long-run HCI losses to the cohort As an example, assume that an economy’s HCI for In 20 years, roughly 46 percent of the workforce percent children under 5 is expected to fall by 1  ­ in a typical country (people ages 20 to 65) will and that those children represent 15 percent of be composed of individuals who were either in the workforce in 2040. Also assume that the losses T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 97 due to school closures are 4 percent and that rates is expected to be related to a change in the those school-age children will be 30 percent of years of school completed by affected children. the workforce by 2040. If the economy’s HCI in It is also likely to be related to a change in cog- 2010 was 0.54 and for 2020 is 0.56, then the HCI nition, proxied by HTS. of that economy’s workforce in 2040 will be 0.007 lower than it would have been in the absence of 3. An additional shock is due to the reduced the pandemic.38 access to health services, whether from fear of contagion or from the lockdown measures. Given that children who are currently in school This shock is expected mostly to affect under-5 will be a larger share of the 2040 workforce than mortality. will those currently under 5, and that the losses for the former are larger in high-income econ- In sum, both the fall in access to health services omies, the fall is expected to be largest among and the income shock will lead to an increase in high- and upper-middle-income economies. child mortality and will worsen stunting. Because Those economies are also the ones that have the more children will be stunted when they reach highest levels of HCI and thus are projected to school age, it is also likely that this outcome will lose more (table 3.3). The results shown here are decrease educational outcomes. meant to inspire action and show that, without remediation, an entire generation could be left Income shock behind. Δy / y​ The income shock (​ ) used in the simulations comes from the World Bank Global Economic Prospects (World Bank 2020). The values come  COVID-19 SHOCK TO THE ANNEX 3A.  from the difference between projected GDP UNDER-5 COHORTS per  capita growth for 2020 used in the Macro Poverty Outlook from the World Bank Annual version The starting point for this simulation is a ­ Meetings of 2019 (before COVID-19), and the of  the HCI calculated with stunting only.39 GDP per capita growth projections made in Assuming that changes in stunting, through its June 2020. relationship to height, are sufficient to capture the health component of the index, the relevant HCI Stunting equation in log terms can be written as The effect of the income shock on stunting is  HTS  ∂ Stunting (3A.1) ( ) ln HCI = ln Survival + φ  EYS ×  625 − 14  ∆ lnHCI = −γ Stunting  ∂y ∆y (3A.2) ( + γ Stuning − Stunting Rate ) where γStunting = 10.2 × 0.034 = 0.35​as discussed in ∂ Stunting appendix A. Although a direct value of The changes in HCI for the under-5 cohorts are ∂y is not available, this value is replaced with an elas- driven by the income shock and reduction in ticity from Ruel, Alderman, and the Maternal and health  care access. Consequently, the three path- Child Nutrition Study Group (2013): ways for the pandemic shock are as follows: ∂ Stunting y 1. The income shock affects under-5 mortality rates. = −0.6 (3A.3) ∂y Stunting 2. The income shock also leads to an increase Inserting equation (3A.3) into equation (3A.2) in stunting rates. In turn, a change in stunting yields the following expression for the direct effect 98 Accumulation Interr upted? C OVID -1 9 and H uman Capital of an income-induced increase in stunting on the Inserting equation (3A.8) into equation (3A.7) HCI: yields: ∂lnSurvival ∂U5MR ∆ y ∆lnHCI =  y (3A.9) ∂ Stunting y ∆y ∂U5MR ∂y y ∆lnHCI = −γ Stunting (3A.4) Stunting ∂y Stunting y An additional shock to mortality is assumed to come from the change in access to health services For economies missing stunting data, the average measured in months of disrupted access: rate for its income group is applied. ∂lnSurvival ∂U5MR   ∆lnHCI = ∆ Access (3A.10) ∂U5MR ∂ Access Education Although a direct value of ∂U5MR is not available, The effect of the income shock on education of a ∂ Access child born today is expected to come through the this value is replaced with a monthly access change effect on stunting: to the elasticity of the under-5 mortality rate from ∆lnHCI Roberton et al (2020).41 φ  ∂ EYS ∂ Stunting ∂ HTS ∂ Stunting  ∂U5MR Access = HTS + EYS  ∆y 625   ∂ Stunting ∂y ∂ Stunting ∂y  = 0.136 (3A.11) ∂ Access U5MR (3A.5) This value suggests a monthly relative increase in under-5 mortality of 13.6 percent given a one- ∂ EYS where ϕ = 0.08​and = −1.594 years of edu- month lack of access. Because the values that enter ∂ Stunting ∂ HTS the index are annual, this value is extrapolated to cation,​ and = −0.625 standard deviations.​ ∂ Stunting the year and inserted into equation (3A.9): ∂lnSurvival ∂U 5 MR Access ∆ Access Inserting equation (3A.3) into equation (3A.5) gives ∆lnHCI = U5MR ∂U5MR ∂ Access U5MR Access us the effect of the income shock on education, operating through increased stunting: (3A.12) φ  ∂ EYS ∂ Stunting y Under the baseline scenario, a 3-month change ∆lnHCI = HTS 625   ∂ Stunting ∂y Stunting in the access to care is assumed; thus, in equation ∂ HTS ∂ Stunting y  ΔAccess​ is (3A.12), access is equal to 12 months and ​ + EYS  ∂ Stunting (3A.6) ∂ y Stunting  equal to 3 months. ∆y Stunting y HCI Mortality Putting all these pieces together returns the total The negative income shock increases child mor- HCI change due to the pandemic: tality, with the following effect on the HCI: ∂lnSurvival ∂U 5 MR Access ∆Access ∆ln HCI = U5MR ∂lnsurvival ∂U5MR ∂U5MR ∂ Access U 5 MR Access ∆lnHCI =  ∆y (3A.7) ∂U5MR ∂y ∂lnSurvival ∂U 5 MR y ∆y + U5MR ∂U5MR ∂ y U5MR y ∂lnsurvival −1 where is equal to . Although φ  ∂ EYS ∂ Stunting y ∂U5MR 1 − U5MR +  HTS ∂U5MR 625  ∂ Stunting ∂y Stunting a direct value of is not available, this value ∂y ∂ HTS ∂ Stunting y  is replaced with a semielasticity from Ma et al. + EYS  ∂ Stunting ∂y Stunting  (2020):40 ∆y ∂ Stunting y ∆y ∂U5MR Stunting − γ Stunting Stunting y = −0.013 (3A.8) y ∂y Stunting y ∂y  (3A.13) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 99 Note how the first component, the one related share of out-of-school children is used to get the to access, enters independently from the income new total number of out-of-school children.42 In shock. essence, when household income drops, children move down the welfare quintiles (because the thresholds are maintained). With more children in ANNEX 3B.   COVID-19 SHOCK TO SCHOOL- lower welfare quintiles with higher shares of out- AGE COHORTS of-school children, there will be an overall increase in the share of out-of-school children, because The shock to children who are presently in school the denominator (total number of children in the is derived as in Azevedo et al. (2020), but using specific school-age bracket) stays the same. ­ the data for the 2020 HCI. The shock to chil- dren operates through two channels: the income The first component of the change in LAYS is the years channel, leading to increased dropouts, and the of school lost due to students who drop out due to the school closure channel, leading to loss in learning income shock (D). and in school years. Recall that LAYS at the pre- COVID-19 baseline (0) is When children go to school, they experience in-person learning, assumed to be the most HTS0 LAYS0 = EYS0 × efficient learning mode. With school closures, ­ 625 children will experience different, less efficient, The changes in income, how well governments learning. The length of school closures differs can deliver education while schools are closed, according to different scenarios: three months, and how long schools are closed are all expected to five months, and seven months for the optimistic, decrease LAYS. The number of out-of-school chil- intermediate, and pessimistic scenarios, respec- dren is assumed to increase because of the income tively. The effectiveness of different remote shock. These changes are calculated for each wel- learning strategies deployed, and the scenarios, fare quintile using data from 130 household sur- are linked to the economy’s income group (see veys using the latest available Global Monitoring table 3B.1). Database, separately for children ages 4–11 and 12–17. The shock from the GDP per capita growth The second component of the change in LAYS is the projections is used to arrive at a new welfare value, share of the school year that is lost because of the closure which is achieved by assuming the shock is uni- and the alternative learning modality (S): form across the distribution; thus, the shape of the S = ​ ​ (1−Mitigation) × Closur​ ​ of School Year​ eShare ​ distribution is maintained but is just shifted to the left. The shift of the household welfare of ­ children HTSs are assumed to change over a school year moves children across welfare quintiles, and the p) by a certain amount (​ ​; the amount is dependent quintile thresholds are the same as those from the on the economy’s income group (see table 3B.2). original welfare distribution. Finally, the quintile’s The learning of these children is compromised by Table 3B.1: Mitigation effectiveness, by scenario and income group Lower-middle Upper-middle   Low income income income High income Optimistic 0.2 0.28 0.4 0.6 Intermediate 0.1 0.14 0.2 0.3 Pessimistic 0.05 0.07 0.1 0.15 Source: Azevedo et al. 2020. 100 Accumulation Interr upted? C OVID -1 9 and H uman Capital Table 3B.2: School productivity (HTS points 3. Roberton et al. (2020) suggest that maintain- gained per school year) ing key childbirth interventions like paren- teral administration of uterotonics, antibiotics,   Points anticonvulsants, and clean birth environments High income 50 could lead to 60 percent fewer maternal Upper-middle income 40 deaths. Maintaining coverage of antibiotics for Lower-middle income 30 neonatal sepsis and pneumonia and oral rehy- Low income 20 dration solution for diarrhea would reduce Source: Azevedo et al. 2020. child deaths by 41 percent. These results are Note: HTS = harmonized test score. likely contingent on modeling assumptions. 4. From the United Nations Educational, the closures and the limited effectiveness of the Scientific and Cultural Organization’s COVID- deployed learning modality. 19 web page, “Education: From Disruption to  Recovery,” at https://en.unesco.org/covid19​ The final component to the change in LAYS is the /educationresponse. amount of learning that takes place under the remote 5. Girls’ educational outcomes during a crisis learning scenario (H): tend to fall more than do those of boys. This is particularly the case if parents’ perception H = S × p​ ​ of returns on investments for boys are greater than for girls (Rose 2000). The change in LAYS is then: 6. This effect was observed during the 2014–15 ​ ​​ ΔLAYS = LAY ​S1 ​ ​​​ −LAY ​S0 Ebola outbreak in West Africa (see de la Fuente, HTS0 Jacoby, and Lawin 2019). A similar effect is now ∆LAYS = LAYS1 − EYS0 × seen in India, where nonavailability of migrant 625 labor has interrupted harvesting activities (see ​​​​is equal to LAY ​S1 where ​ Saha and Bhattacharya 2020). ( HTS −H ) 7. Women will often sacrifice their own con- ( LAYS1 = EYS0 − S − D ) 625 0 sumption needs in order to ensure sufficient nutrition for other household members (see Quisumbing et al. 2011). NOTES 8. When surveyed, Ugandan households had yet to resort to selling productive assets to cope 1. Simulations suggest that, in Ireland, 400,000 with the losses in income, perhaps in the hope households may see a drop in their dispos- that the income shortfall will be short-lived. able income of 20 percent or more (Beirne 9. In some cases, the substitution effect (the et al. 2020). In Italy, simulations show that relative change in prices of activities) domi- disposable income losses will be consider- nates the income effect (the drop in purchas- able and more pronounced for the poorest. ing power). For example, Miller and Urdinola Italian households in the poorest quintile are (2010) present evidence of how child health projected to lose 40 percent of their income improved among children of coffee farmers (Figari and Fiorio 2020). in Colombia during a decline in the price of 2. Hogan et al. (2020) find that for HIV the largest coffee. Because the fall in coffee prices made impact is from interruption of antiretroviral time spent farming less valuable, parents therapy, for tuberculosis the impact is due to devoted more time to their children, which reduction of timely diagnosis and treatment, translates into better outcomes for children. and for malaria the impact reflects the inter- Schady (2004) provides evidence that, in Peru, ruption of prevention programs. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 101 children exposed to a crisis in the late 1980s 16. Andrabi, Daniels, and Das (2020) posit that completed on average one additional year of this difference in test scores is equivalent to schooling. 1.5 school grades. To arrive at this value, the 10. For example, Bundervoet and Fransen (2018) authors note that the average 15-year-old has find that children exposed to the Rwandan accumulated 5.6 grades and linearly gains 0.17 genocide while in utero suffered lower educa- standard deviations in performance per grade tional outcomes. The longer the exposure in level in the test the authors use. This result, in utero, the poorer the educational outcomes. the context of harmonized test scores used in 11. Savasi et al. (2020) find that 12 percent of the the HCI, translates to a drop of 24 points. 77 patients in their study (in Italy) had a preterm 17. The simulation by Azevedo et al. (2020) delivery. By contrast, Philip et al. (2020) find implicitly assumes that income effects out- a reduction in preterm births in Ireland, and weigh substitution effects that may arise in a reduction in very low birthweights, falling these cases. Nonetheless, substitution effects from 3.77 cases per 1,000 births to 2.17 cases. may be larger, and enrollment could increase. 12. O’Hare et al. (2013) obtain this estimate Shafiq (2010) presents two cases in which through meta-analysis from a systematic liter- enrollment may increase: (1) falling wages ature search of studies and find a pooled elas- make child labor less attractive, and (2) parents ticity of income on infant mortality of −0.95. place a higher preference on education, per- 13. Ma et al. (2020) find that, in low-income educated workers bear the haps because less-­ countries, a lockdown will potentially lead to brunt of the crisis. 1.17 children’s lives lost per COVID-19 fatal- 18. Values are obtained for 157 economies. Authors ity averted, due to the economic contraction, model different mitigation strategies taken significantly higher than in lower-­ middle- during remote learning, vary the length of middle-income countries (where and upper-­ school closures, and assume children will drop it would stand at 0.48 and 0.06, respectively). out of school because of the income shock. Two factors account for this outcome: the The yearly losses range from US$127 in low-­ younger demographic structure and the income economies to US$1,865 in high-­ income higher estimated elasticity of child mortality economies. to GDP changes in low-income countries. The 19. A similar dynamic was observed in Peru, authors also assume that under-5 mortality is where the opportunity costs of going to school not affected by income shocks in high-income decreased by a considerable amount because countries. wages dropped substantially. Thus, children 14. Mary (2018) suggests that the decrease may exposed to the crisis completed more years of be 2.7 percent, whereas Mary, Shaw, and education (Schady 2004). Paloma (2019) estimate it to be 7.3 percent, and 20. Rothstein (2020) finds evidence that those Ruel, Alderman, and the Maternal and Child who graduated during the 2008–09 finan- percent. Nutrition Study Group (2013) suggest 6 ­ cial crisis had lower wages and employment It is worth noting that these analyses concen- than earlier cohorts. The author shows that trated mostly on low- and middle-income market conditions at the time of labor mar- economies. ket entry matter greatly for cohorts’ employ- 15. More recent research has called this result into ment probabilities. question (see von Hippel and Hamrock 2019 21. Roughly 47 million women in 114 low- and for a more nuanced discussion); however, middle-income economies could lose access to a summer break is not the same as a break contraceptives in the scenario of a six-month during the school year. lockdown or disruptions (UNFPA 2020). 102 Accumulation Interr upted? C OVID -1 9 and H uman Capital 22. Bell, Bindler, and Machin (2017) find that modalities differ by country income. Efficiency cohorts graduating into a recession are 10.2 for lower-, lower-middle-, upper-middle-, and percent more likely to commit criminal activ- high-income economies under the pessimistic ity than cohorts entering the labor market in scenario is 5, 7, 10, and 15 percent, respectively. nonrecession times. The values are doubled in the intermediate 23. The authors of the study also note a consider- scenario and quadrupled under the optimistic able increase (of 57 percent) in the number of scenario. calls related to psychological violence. 32. The outcome results from multiplying the 24. Such facilities include long-term care homes, HTS points per year by the share of the school residential care homes, nursing homes, wel- year that is assumed to be lost, and 1 minus fare homes, and others. the efficiency of the mitigation measure in 25. See appendix A for details on the methodol- place. ogy of the HCI. 33. The authors use household surveys for 130 26. In early 1919, roughly one-third of all new- economies to calculate economy-specific borns had mothers who had been infected dropout income elasticities and welfare using by influenza while pregnant. The 1918 pan- cross-sectional variation by welfare quintiles. demic was disproportionately deadly to those Refer to Azevedo et al. (2020) for details. between 25 and 35 (Almond 2006). 34. See Azevedo et al. (2020) for a detailed explana- 27. These calculations are based on the literature tion on how the income shock is incorporated. review in Galasso and Wagstaff (2019). The 35. Intermediate scenario in Azevedo et al. (2020). authors find that children who are stunted obtain 36. Roughly 34 percent of the workforce will be 1.594 years less education, and score 0.625 stan- composed of individuals whose schooling was dard deviations lower on standardized tests. interrupted by the pandemic, and 12 percent 28. For economies for which stunting data are not of the workforce will be composed of individ- available and that are thus not used in the cal- uals who were under the age of 5 during the culation of the HCI, the income group’s aver- pandemic. age stunting rate is applied to the individual 37. To calculate the HCI loss by economy, the per- economy to simulate the possible losses due to centage change for each of the cohorts (pre- the pandemic. sented in table 3.1 and table 3.2) is applied to 29. This section was contributed by Joao Pedro de the HCI of 2020 for the under-5 cohort and the Azevedo based on Azevedo et al. (2020). The HCI of 2010 for the cohort of 2010 to arrive at results presented in this section use the 2020 a value for HCI that is lost. Economies missing HCI numbers as baseline values. For that rea- an HCI in 2010 were imputed the value of their son, they will be slightly different from those income group. The economy’s projected pop- in the original paper. ulation shares are used to calculate each econ- 30. For high-income economies, the value is omy’s HCI point loss among the workforce. assumed to be 50 points in a year; 40 points The result supposes that on average those who for upper-middle-income economies; 30 are currently between the ages of 18 and 45 will points for lower-middle-income economies; not experience any ill health effects due to the and 20 points for low-income economies. pandemic, and thus in 20 years their human 31. The authors assume that all governments offer capital will be the same. some alternative learning modality. Estimates HCI​  Loss = 0.56 (−  38. ​ 0.15 + 0.54 ​ 0.01) ​ 0.04)​0.3​ (−  . of their effectiveness are informed by exist- ​ 39. Because the parameter γ ​ ​Stunting​​​ embodies the ing multitopic household surveys. Thus, best alternative of the link between stunting to access and effectiveness of the implemented adult height and from adult height to earnings, T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 103 the index can be expressed by relying just on Asanov, I., F. Flores, D. J. Mckenzie, M. Mensmann, stunting as a proxy for health. and M. Schulte. 2020. “Remote-Learning, 40. The elasticities are disaggregated by income Time-Use, and Mental Health of Ecuadorian groups: it is assumed to be 0 for high-income High-School Students during the COVID-19 economies, –0.003 for for upper-middle-­ Quarantine.” Policy Research Working Paper income economies, –0.01 for lower-­ middle- 9252, World Bank, Washington, DC. income economies, and –0.013 for low-­ income Azevedo, J. P., A. Hazan, D. Goldemberg, S. A. Iqbal, economies. Ma et al. (2020) express these values and K. Geven. 2020. “Simulating the Potential as a 1 percent decrease in GDP per capita being Impacts of COVID-19 School Closures on associated with an increase of 0.13 under-5 Schooling and Learning Outcomes: A Set of deaths per 1,000 children (or an increase of Global Estimates.” Policy Research Working 0.013 percentage points in under-5 mortality Paper 9284, World Bank, Washington, DC. rates). Bandiera, O., N. Buehren, M. P. Goldstein, I. Rasul, 41. Roberton et al. (2020) offer three scenarios in and A. Smurra. 2019. “The Economic Lives of which the effect ranges from 8.0 to 34.5 percent. Young Women in the Time of Ebola: Lessons 42. For countries without a household survey, the from an Empowerment Program.” Policy overall change in out-of-school rates for the Research Working Paper 8760, World Bank, country’s income group is used. Washington, DC. Bardhan, P., and C. Udry. 1999. Development REFERENCES Microeconomics. Oxford: Oxford University Press. Acemoglu, D., V. Chernozhukov, I. Werning, and Bassett, M. R., and N. Arnhold. 2020. “COVID- M. D. Whinston. 2020. “Optimal Targeted 19’s Immense Impact on Equity in Tertiary Lockdowns in a Multi-Group SIR Model.” Education.” Education for Global Development NBER Working Paper 27102, National Bureau (blog), April 30, 2020. https://blogs.worldbank​ of Economic Research, Cambridge, MA. .org/education/covid-19s-immense-impact​ Almond, D. 2006. “Is the 1918 Influenza Pandemic -equity-tertiary-education. Over? Long-Term Effects of In Utero Influenza Beegle, K., R. H. Dehejia, and R. Gatti. 2006. “Child Exposure in the Post-1940 US Population.” Labor and Agricultural Shocks.” Journal of Journal of Political Economy 114 (4): 672–712. Development Economics 81 (1): 80–96. Almond, D., and J. Currie. 2011. “Killing Me Softly: Beirne, K., K. Doorley, M. Regan, B. Roantree, The Fetal Origins Hypothesis.” Journal of and D. Tuda. 2020. “The Potential Costs Economic Perspectives 25 (3): 153–72. and Distributional Effect of COVID-19 Alon, T., M. Doepke, J. Olmstead-Rumsey, and Related Unemployment in Ireland.” Budget M. Tertilt. 2020. “The Impact of COVID-19 Perspectives 2021 Paper 1, Economic & Social on Gender Equality.” NBER Working Paper Research Institute, Dublin. 26947, National Bureau of Economic Research, Bell, B., A. Bindler, and S. Machin. 2017. “Crime Cambridge, MA. Scars: Recessions and the Making of Career Andrabi, T., B. Daniels, and J. Das. 2020. “Human Criminals.” The Review of Economics and Statistics Capital Accumulation and Disasters: Evidence 100 (3): 392–404. from the Pakistan Earthquake of 2005.” RISE Black, S. E., P. J. Devereux, and K. G. Salvanes. 2007. Working Paper Series 20/039, Research on “From the Cradle to the Labor Market? The Improving Systems of Education (RISE) Effect of Birth Weight on Adult Outcomes.” The Programme, Oxford. Quarterly Journal of Economics 122 (1): 409–39. 104 Accumulation Interr upted? C OVID -1 9 and H uman Capital Bundervoet, T., and S. Fransen. 2018. “The 2011. “National Postsecondary Enrollment Educational Impact of Shocks in Utero: Trends: Before, during, and after the Great Evidence from Rwanda.” Economics & Human Recession.” Signature Report, National Student Biology 29 (C): 88–101. Clearinghouse. CDC (Centers for Disease Control and Prevention). Elston, J. W. T., C. Cartwright, P. Ndumbi, and 2020. “Coronavirus Disease 2019 (COVID-19).” J. Wright. 2017. “The Health Impact of the https://www.cdc.gov/coronavirus/2019-ncov​ 2014–15 Ebola Outbreak.” Public Health 143 /index.html. (February): 60–70. Chang, H.-J., N. Huang, C.-H. Lee, Y.-J. Hsu, C.-J., Figari, F., and C. V. Fiorio. 2020. “Welfare Hsieh, and Y.-J. Chou. 2004. “The Impact of the Resilience in the Immediate Aftermath of the SARS Epidemic on the Utilization of Medical COVID-19 Outbreak in Italy.” EUROMOD Services: SARS and the Fear of SARS.” American Working Papers EM6/20, EUROMOD at the Journal of Public Health 94 (4): 562–64. Institute for Social and Economic Research, Cooper, H., B. Nye, K. Charlton, J. Lindsay, and University of Essex. S. Greathouse. 1996. “The Effects of Summer Finch, B. K., K. Thomas, and A. N. Beck. 2019. Vacation on Achievement Test Scores: A “The Great Recession and Adverse Birth Narrative and Meta-Analytic Review.” Review of Outcomes: Evidence from California, USA.” Educational Research 66 (3): 227–68. SSM–Population Health 9: 100470. Corral, P., and R. Gatti. 2020. “Accumulation Galasso, E., and A. Wagstaff. 2019. “The Aggregate Interrupted: COVID-19 and Human Capital Income Losses from Childhood Stunting and among the Young.” In COVID-19 in Developing the Returns to a Nutrition Intervention Aimed Economies edited by S. Djankov and U. at Reducing Stunting.” Economics and Human Panizza. London: Centre for Economic Policy Biology 34: 225–38. Research. Galea, S., R. M. Merchant, and N. Lurie. 2020. COVID-19 Task Force (Task Force COVID-19 del “The Mental Health Consequences of COVID- Dipartimento Malattie Infettive e Servizio 19 and Physical Distancing: The Need for di Informatica, Istituto Superiore di Sanità). Prevention and Early Intervention.” JAMA 2020. “Epidemia COVID-19.” Aggiornamento Internal Medicine 180 (6): 817–18. nazionale, April 2020. Gibbs, L., J. Nursey, J. Cook, G. Ireton, N. Alkemade, Datar, A., R. Kilburn, and D. S. Loughran. 2010. M. Roberts, H. C. Gallagher, R. Bryant, K. Block, “Endowments and Parental Investments in R. Molyneaux, and D. Forbes. 2019. “Delayed Infancy and Early Childhood.” Demography 47 Disaster Impacts on Academic Performance of (1): 145–62. Primary School Children.” Child Development Davis, S. J., and T. M. von Wachter. 2011. “Recessions 90 (4): 1402–12. and the Cost of Job Loss.” NBER Working Paper Golberstein, E., H. Wen, and B. F. Miller. 2020. 17638, National Bureau of Economic Research, “Coronavirus Disease 2019 (COVID-19) and Cambridge, MA. Mental Health for Children and Adolescents.” de la Fuente, A., H. G. Jacoby, and K. G. Lawin. JAMA Pediatrics, April 14, 2020. https:// 2019. “Impact of the West African Ebola jamanetwork.com/journals/jamapediatrics​ Epidemic on Agricultural Production and /fullarticle/2764730. Rural Welfare: Evidence from Liberia.” Policy Hale, T., N. Angrist, E. Cameron-Blake, L. Hallas, B. Research Working Paper 8880, World Bank, Kira, S. Majumdar, A. Petherick, T. Phillips, H. Washington, DC. Tatlow, and S. Webster. 2020. “Oxford COVID- Dunbar, A., D. Hossler, D. Shapiro, J. Chen, 19 Government Response Tracker.” Blavatnik S. Martin, V. Torres, D. Zerquera, and M. Ziskin. School of Government, University of Oxford. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 105 Hershbein, B. J. 2012. “Graduating High School Le Nestour, A., and L. Moscoviz. 2020. “Five in a Recession: Work, Education, and Home Findings from a New Phone Survey in Senegal.” Production.” B.E. Journal of Economic Analysis & Center for Global Development (blog), April Policy 12 (1). 24, 2020. https://www.cgdev.org​ /blog/five​ Hobbs, J. E. 2020. “Food Supply Chains during -findings-new-phone-survey-senegal. the COVID-19 Pandemic.” Canadian Journal of Ma, L., G. Shapira, D. de Walque, Q.-T. Do, Agricultural Economics/Revue Canadienne d’agro- J.  Friedman, and A. A. Levchenko. 2020. economie 68 (2): 171–76. “The Intergenerational Dilemma of Hogan, A., B. Jewell, E. Sherrard-Smith, J. Vesga, O. COVID-19 Lockdown Policies.” Unpublished Watson, C. Whittaker, A. Hamlet, J. Smith, K. manuscript. Ainslie, M. Baguelin, S. Bhatt, A. Boonyasiri, N. Mahler, D. G., C. Lakner, A. Castaneda, and Brazeau, L. Cattarino, G. Charles, L. Cooper, H. H. Wu. 2020a. “The Impact of COVID-19 Coupland, G. Cuomo-Dannenburg, A. Dighe, (Coronavirus) on Global Poverty: Why Sub- and T. Hallett. 2020. “Report 19: The Potential Saharan Africa Might Be the Region Hardest Impact of the COVID-19 Epidemic on HIV, Hit.” Data Blog, April 20, 2020. https://blogs​ TB and Malaria in Low- and Middle-Income /opendata/impact-covid-19​ .worldbank.org​ Countries.” Lancet Global Health, July 13, 2020. -coronavirus-global-poverty-why-sub-saharan​ https://www.thelancet.com/journals/langlo​ -africa-might-be-region-hardest. /article/PIIS2214-109X(20)30288-6/fulltext. Mahler, D. G., C. Lakner, A. Castaneda, and H. Wu. International Labour Organization. 2020. “As 2020b. “Updated Estimates of the Impact of Job Losses Escalate, Nearly Half of Global COVID-19 on Global Poverty.” Data Blog, June Workforce at Risk of Losing Livelihoods.” 8, 2020. https://blogs.worldbank.org/opendata​ Press Release, April 29, 2020. https://www.ilo​ /updated-estimates-impact-covid-19-global​ .org/global/about-the-ilo/newsroom/news​ -​poverty. /WCMS_743036/lang--en/index.htm. Mahmud, M., and E. Riley. Forthcoming. “Household International Monetary Fund. 2020. “Global Response to an Extreme Shock: Evidence on the Prospects and Policies.” Chapter 1 in World Immediate Impact of the Covid-19 Lockdown Economic Outlook, April 2020: The Great Lockdown. on Economic Outcomes and Well-being in Rural Washington, DC: International Monetary Fund. Uganda.” World Development. Lahti-Pulkkinen, M., S. Bhattacharya, K. Räikkönen, Mary, S. 2018. “How Much Does Economic Growth C. Osmond, J. E. Norman, and R. M. Reynolds. Contribute to Child Stunting Reductions?” 2018. “Intergenerational Transmission of Birth Economies 6 (4): 55. Weight across Three Generations.” American Mary, S., K. Shaw, and S. G. Paloma. 2019. “Does Journal of Epidemiology 187 (6): 1165–73. the Sectoral Composition of Growth Affect Lain, J., J. Perng, T. Vishwanath, M. A. Azad, Child Stunting Reductions?” Development Policy G. O. Siwatu, A. Palacio-Lopez, K. McGee, Review 37 (2): 225–44. A. Amankwa, and I. Contreras. 2020. “COVID-19 Meyers, K., and M. Thomasson. 2017. “Paralyzed from the Ground Up: What the Crisis Means for by Panic: Measuring the Effect of School Nigerians.” Findings from the Nigeria COVID-19 Closures during the 1916 Polio Pandemic on National Longitudinal Phone Survey Round 1, Educational Attainment.” NBER Working April–May 2020, World Bank, Washington, DC. Paper 23890, National Bureau of Economic Lancker, W. V., and Z. Parolin. 2020. “COVID-19, Research, Cambridge, MA. School Closures, and Child Poverty: A Social Miller, G., and B. P. Urdinola. 2010. “Cyclicality, Crisis in the Making.” Lancet Public Health 5 (5): Mortality, and the Value of Time: The Case of e243–e244. Coffee Price Fluctuations and Child Survival in 106 Accumulation Interr upted? C OVID -1 9 and H uman Capital Colombia.” Journal of Political Economy 118 (1): Food-Price Crisis.” Development in Practice 21 113–55. (4–5): 488–92. Nelson, R. 2020. “COVID-19 Disrupts Vaccine Ravindran, S., and M. Shah. 2020. “Unintended Delivery.” Lancet Infectious Diseases 20 (5): 546. Consequences of Lockdowns: COVID-19 and O’Hare, B., I. Makuta, L. Chiwaula, and N. Bar- the Shadow Pandemic.” NBER Working Paper Zeev. 2013. “Income and Child Mortality in 17638, National Bureau of Economic Research, Developing Countries: A Systematic Review Cambridge, MA. and Meta-analysis.” Journal of the Royal Society of Roberton, T., E. D. Carter, V. B. Chou, A. R. Medicine 106 (10): 408–14. Stegmuller, B. D. Jackson, Y. Tam, T. Sawadogo- Olsen, A. S., and J. Prado. 2020. “COVID-19 Lewis, and N. Walker. 2020. “Early Estimates of y la transición de la educación privada a the Indirect Effects of the COVID-19 Pandemic la pública en Ecuador.” Enfoque Educación on Maternal and Child Mortality in Low-Income (blog), June 30, 2020. https://blogs.iadb.org​ and Middle-Income Countries: A Modelling /covid-19-y-la-transicion-de-la​ /educacion/es​ Study.” Lancet Global Health, May 12, 2020. https:// -privada-a-la-publica-en-ecuador/. -educacion​ www.thelancet.com/­ / langlo​ j ournals​ /article/ Oreopoulos, P., T. von Wachter, and A. Heisz. 2012. PIIS2214-109X(20)30229-1/fulltext. “The Short-and Long-Term Career Effects of Rosales-Rueda, M. 2018. “The Impact of Early Graduating in a Recession.” American Economic Life Shocks on Human Capital Formation: Journal: Applied Economics 4 (1): 1–29. Evidence from El Niño Floods in Ecuador.” Packard, T., U. Gentilini, M. Grosh, P. O’Keefe, Journal of Health Economics 62: 13–44. R. Palacios, D. Robalino, and I. Santos. 2019. Rose, E. 2000. “Gender Bias, Credit Constraints Protecting All: Risk Sharing for a Diverse and and Time Allocation in Rural India.” Economic Diversifying World of Work. Washington, DC: Journal 110 (465): 738–58. World Bank. Rothstein, J. 2020. “The Lost Generation? Labor Perez-Vincent, S., E. Carreras, M. A. Gibbons, T. Market Outcomes for Post Great Recession E. Murphy, and M. Rossi. 2020. “COVID-19 Entrants.” NBER Working Paper 27516, National Lockdowns and Domestic Violence: Evidence Bureau of Economic Research, Cambridge, MA. from Two Studies in Argentina.” Working Ruel, M. T., H. Alderman, and the Maternal and Papers 143. Universidad de San Andres, Child Nutrition Study Group. 2013. “Nutrition- Departamento de Economia. Sensitive Interventions and Programmes: Philip, R. K., H. Purtill, E. Reidy, M. Daly, M., How Can They Help to Accelerate Progress Imcha, D. McGrath, N. H. O’Connell, and C. in Improving Maternal and Child Nutrition?” P. Dunne. 2020. “Reduction in Preterm Births Lancet 382 (9891): 536–51. during the COVID-19 Lockdown in Ireland: Saha, T., and S. Bhattacharya. 2020. “Consequence A  Natural Experiment Allowing Analysis of of Lockdown amid Covid-19 Pandemic on Data from the Prior Two Decades.” MedRxiv, Indian Agriculture.” Food and Scientific Reports June 5, 2020. https://www.medrxiv.org/content​ 1 Special Issue: 47–50. /10.1101/2020.06.03.20121442v1. Savasi, V. M., F. Parisi, L. Patanè, E. Ferrazzi, Poletaev, M., and C. Robinson. 2008. “Human L.  Frigerio, A. Pellegrino, A. Spinillo, S. Capital Specificity: Evidence from the Tateo, M. Ottoboni, P. Veronese, F. Petraglia, Dictionary of Occupational Titles and Displaced P. Vergani, F. Facchinetti, D. Spazzini, and Worker Surveys, 1984–2000.” Journal of Labor I.  Cetin. 2020. “Clinical Findings and Disease Economics 26 (3): 387–420. Severity in Hospitalized Pregnant Women Quisumbing, A., R. Meinzen-Dick, J. Behrman, with Coronavirus Disease 2019 (COVID-19).” and L.  Basset. 2011. “Gender and the Global Obstetrics & Gynecology 136 (2): 252–58. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 107 Schady, N. R. 2004. “Do Macroeconomic Crises the School Years? Measurement Artifacts Always Slow Human Capital Accumulation?” and What We Can Know in Spite of Them.” World Bank Economic Review 18 (2): 131–54. Sociological Science 6: 43–80. Shafiq, M. N. 2010. “The Effect of an Economic Wenham, C., J. Smith, and R. Morgan. 2020. Crisis on Educational Outcomes: An Economic “COVID-19: The Gendered Impacts of the Framework and Review of the Evidence.” Outbreak.” Lancet 395 (10227): 846–48. Current Issues in Comparative Education 12 (2): 5. World Bank. 2020. Global Economic Prospects, June Sullivan, D., and T. von Wachter. 2009. “Job 2020. Washington, DC: World Bank. Displacement and Mortality: An Analysis World Health Organization. 2020. “At Least 80 Using Administrative Data.” Quarterly Journal of Million Children under One at Risk of Diseases Economics 124 (3): 1265–306. Such as Diphtheria, Measles and Polio as COVID- Tesfaye, A., Y. Habte, and B. Minten. 2020. 19 Disrupts Routine Vaccination Efforts, Warn “COVID-19 Is Shifting Consumption and Gavi, WHO and UNICEF.” News Release, Disrupting Dairy Value Chains in Ethiopia.” World Health Organization, May 22, 2020. IFPRI Blog, June 1, 2020. https://www.ifpri​ https://www.who.int/news/item/22-05-2020​ /blog/covid-19-shifting-consumption-and​ .org​ -at-least-80-million-children-under-one-at​ -​disrupting-dairy-value-chains-ethiopia. -risk-of-diseases-such-as-diphtheria-measles​ UNFPA (United Nations Population Fund). 2020. -and-polio-as-covid-19-disrupts-routine​ “Impact of the COVID-19 Pandemic on Family -vaccination-efforts-warn-gavi-who-and- Planning and Ending Gender-based Violence, unicef. Female Genital Mutilation and Child Marriage.” Wren-Lewis, S. 2020. “The Economic Effects of a Interim Technical Note, April 27, UNFPA. Pandemic.” In Economics in the Time of COVID-19, van Gelder, N., A. Peterman, A. Potts, M. O’Donnell, edited by Richard Baldwin and Beatrice Weder di K. Thompson, N. Shah, and S. Oertelt-Prigione. Mauro, 109–112. London: Centre for Economic 2020. “COVID-19: Reducing the Risk of Policy Research. Infection Might Increase the Risk of Intimate Yourish, K., K. K. R. Lai, D. Ivory, and M. Smith. Partner Violence.” EClinicalMedicine 21: 100348. 2020. “One-Third of All U.S. Coronavirus von Hippel, P. T. , and C. Hamrock. 2019. “Do Test Deaths Are Nursing Home Residents or Score Gaps Grow before, during, or between Workers.” New York Times, May 11, 2020. 4 UTILIZING Human Capital T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 109 T he World Bank’s Human Capital Index (HCI) name suggests, adjust the HCI for labor market captures the size of the income gains when underutilization of human capital.1 The UHCIs are today’s better-educated and healthier chil- designed to complement the main HCI, and not dren become tomorrow’s more productive work- to replace it, in part because these two measures ers. Specifically, a child born today can expect to be have different purposes: the HCI is an index of the HCI × 100 percent as productive as a future worker supply of a factor of production (in the future), as she would be if she enjoyed complete educa- whereas the UHCIs are a hybrid between an index tion and full health. But this expectation implicitly of factor supply (capturing investment in human assumes that, when today’s child becomes a future capital) and a utilization index (capturing how worker, she will be able to find a job—which may efficiently that human capital is used in produc- not be the case in economies with low employment tion). Moreover, there are numerous challenges in rates. Moreover, even if today’s child is able to find defining and measuring utilization in a consistent employment in the future, she may not have a job way across diverse economic contexts. As such, in which she can fully use her skills and cognitive the UHCIs should be viewed as a first attempt to abilities to increase her productivity. In these cases, address utilization in a simple way consistently human capital can be considered underutilized, across economies, and should be applied with cau- productivity because it is not being used to increase ­ tion in policy analysis. to the extent it could be. For example, unem- ployed future workers may be underutilizing their Importantly, the HCI and UHCI measure only the human capital, as are those out of the labor force. effect of human capital on labor market earnings Likewise, engineers driving taxis are underutiliz- and future gross domestic product (GDP) per cap- ing their human capital because, even though they ita. But this effect is only one benefit of human are employed, they do not hold jobs in which their capital. In many other domains, human capital education increases their productivity. improves well-being and economic development. Parents with more education have children with In addition, a gender gap—which is not apparent better human capital outcomes, and women with in the human capital dimensions captured by the more human capital are more empowered. Even HCI—emerges and deepens during the working outside the categories of better employment years. In many countries, women face worse jobs (defined later in the chapter), human capital can still and income opportunities compared to men, even increase productivity—for example, smallholder with the same human capital. As such, simply con- efficiently—but farmers might use fertilizer more ­ sidering the HCI by sex may give a partial view in the increase is just less dramatic than for other terms of realizing the potential of human capital incomplete utiliza- employment types. As such, ­ investments. tion should not be interpreted as a lack of gains from human capital investments, but rather as This chapter introduces two Utilization-Adjusted market gains are an indication that private labor ­ Human Capital Indexes (UHCIs) that, as their smaller than they could be. 110 U tilizing H u man Cap ital The two UHCIs take different approaches to mea- For the basic UHCI, this multiplicative form stems suring utilization. In the basic UHCI, utilization is from its connection to economic growth. In the measured as the fraction of the working-age pop- long run, GDP is proportional to the number of ulation that is employed. Although simple and workers (employment) multiplied by the produc- intuitive, this measure cannot capture underuti- tivity boost that each worker gets from her human lization resulting from a mismatch between the capital.2 The basic UHCI inherits this multiplica- skills and cognitive abilities required to do a job, tive form, where the HCI captures the productiv- and the skills and cognitive abilities of the people ity boost from human capital, and the utilization employed to do it. The full UHCI measure adjusts rate captures employment.3 for this mismatch by introducing the concept of better employment, which includes the types of jobs The HCI is derived to measure the effect of human that are common in high-productivity countries. capital on future GDP per capita so that projected HCI​ future per capita GDP will be approximately 1/​ Despite different methodologies, the basic and times higher in a “complete education and full full measures produce broadly similar utilization health” scenario than in a “status quo” scenario rates. Utilization rates are U-shaped in per cap- (Kraay 2018). This definition implicitly assumes ita income across economies, first declining with that utilization rates of human capital—such as income at lower income levels and then rising at employment prospects—are the same in the com- higher income levels. This feature of utilization plete education and full health scenario as in the rates implies that UHCIs are low in the poorest status quo scenario. average, economies, where the HCI is also low on ­ but remain low over a wider range of lower-­ Both UHCI measures are derived in a similar way, middle-income economies where rising HCIs are in keeping with the economic interpretation of the offset by declining utilization. HCI. For the UHCIs, however, utilization rates are now different in the status quo and full human cap- Moreover, both UHCIs reveal starkly different ital scenarios. Specifically, both UHCIs are derived gender gaps from those using the HCI. Girls have as future GDP per capita under the “status quo” a slight advantage over boys in human capital relative to future GDP per capita with “full health, early in life, resulting in a higher HCI for girls on complete education, and complete utilization,” average. But female utilization rates are typically as in equation (4.2). This means that, in the long lower than those for males, resulting in lower run, GDP per capita will be 1/UHCI times higher UHCIs. Although gender gaps in human capital in a world of complete utilization, full health, and in childhood and adolescence (especially in edu- complete education than in the status quo.4 cation) have closed in the last two decades, large challenges remain to realizing these investments UHCI = Future GDP per Capita ( Status Quo ) in terms of income opportunities for women. Future GDP per Capita (Complete Utilization, Full Health, Complete Education) (4.2) 4.1 METHODOLOGY AND THE BASIC UHCI MEASURE For the basic UHCI, the utilization rate is simply the employment rate of the working-age popula- Both the basic and full UHCIs have a simple form, L​ tion. This rate is current employment ​ , relative to the utilization rate multiplied by the HCI: a measure of potential employment under full uti- lization ​L*​​,​the maximum theoretical employment. ​​(4.1)​ UHCI  =  Utilization Rate × HCI​​​ The standard definition of the potential labor force T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 111 is the working-age population, individuals aged future GDP per capita under complete utiliza- 15–64 years. This definition is also adopted for L . * 5 tion and complete human capital is proportional ​ ​is defined as the number of people Employment L to ​ h ​ h*​​ is complete human capital per worker) ​ *L* (​ aged 15–64 who are in paid employment (or are in the denominator of equation (4.2). Because self-employed) to be consistent with the definition the proportionality factors are the same, and so of the potential labor force. cancel out, this expression can be rearranged h*​ HCI = h / ​ as ​ ​multiplied by the basic utilization As mentioned earlier, the basic UHCI takes the ​ til ​(basic)​  =  L / ​ rate U L*. simple multiplicative form in equation (4.1) because, in a standard production function, long- A natural concern is that, in economies with low run GDP per capita is proportional to human cap- basic utilization rates, human capital, as mea- ital per worker ​ h​multiplied by employment L per sured by the HCI, will have less effect on eco- capita (see box 4.1 for a derivation). Future GDP nomic growth. This is not the case, however.6 In per capita under the status quo—in the numera- the framework of the basic UHCI, an increase tor of equation (4.2)—is proportional to ​ hL​ , and in human capital alone has the same effect on Box 4.1: Deriving the basic Utilization-Adjusted Human Capital Index Future gross domestic product (GDP) per capita of the next generation y​in the status quo world is given by a Cobb-Douglas production function: ​ y  =  A ​​ K​​ 1−β​​(hL)​​​  β​ / N​ (B4.1.1) where h represents human capital per worker under current policies, and L represents the num- ber of workers under status quo employment rates. A is total factor productivity, K is the amount of physical capital, and N is the future population. Total factor productivity and the future popula- tion are assumed to grow at the same trend rates in all scenarios. In an alternative world, there is complete human capital per worker, denoted by ​​h​​*​​, and complete employment of potential workers, denoted by ​​L* . Long-run GDP per capita in this   ​​ complete human capital–complete employment world is denoted ​​y  ​​*.​​​​ As in Kraay (2018), the production function can be rearranged in terms of the physical capital-to-output ratio K Y , which is constant in the long run. Then, future GDP per capita under the status quo relative to the complete human capital–complete utilization scenario is given by: ( ) (1− β ) β y A1 β K Y hL N = ( ) y ∗ A1 β K Y (1− β ) β h∗ L∗ N L h (B4.1.2) = * × * L h = Utilization (basic ) × HCI = UHCI (basic ) 112 U tilizing H u man Cap ital long-run economic growth as in the HCI, but which means that UHCI scores will differ from economies can do better by also increasing utiliza- those of the HCI (figure 4.1). Employment rates tion (it is not one or the other).7 average about 0.6, which suggests that the UHCI will, on average, be about 60 percent of the value The main data source for the basic utilization mea- of the HCI (figure 4.2), but with substantial varia- sure is “Employment-to-population by sex and tion across economies. age (%) – Annual,” for youth and adults aged 15–64, from the International Labour Organization (ILO), Employment rates (basic utilization) are using the latest period available.8 The secondary data ­ approximately U-shaped in log per capita income source is the World Bank’s Global Jobs Indicators (figure 4.3).11 High-income economies have the Database ( JOIN), which has employment data based highest utilization rates (about 0.7 for the group as on the same population age group, with a sample a whole), which is unsurprising because it is difficult skewed toward low- and middle-income economies.9 people to have high per capita incomes with few ­ Data are generally taken from the most recent working. Low-income economies have utiliza- source if both are available.10 The median year of the tion rates of about 0.6 on average, though many data is 2017, with 95 percent of economies having income economies—like Burundi (BDI in the low-­ data from the 2010s. The basic utilization measure figures), Madagascar (MDG), and Mozambique is available for 185 economies. The measurement of (MOZ)—also have extremely high employment the full UHCI is discussed later in the chapter. rates of about 0.8. High employment rates among low-income economies are likely because most people are so poor that they need to work out- 4.2  THE BASIC UHCI IN THE DATA side the home to survive. Lower-middle-income economies have the lowest utilization rates, at Basic utilization rates are not strongly correlated about 0.55, mostly because slightly higher incomes with the HCI (correlation coefficient of 0.45), make it feasible for people (especially women) not Figure 4.1: Employment to population (basic utilization) and Human Capital Index 0.9 QAT MDG SLB ISL TZA KHM ZWE VNM 0.8 BDI ARE CHE MAC NZL SWE NLD MOZ THA DEU JPN LBR SYC DNK CZE NOR EST GBR UGA Employment−to−population KWT BHR BLR LTU AUT AUS CAN SGP MLT LVA SVN FIN 0.7 CMR PRY KNAPEROMN RUS USA ratio (basic utilization) HUN PRT MMR PLW URYBGRKAZ CHN CYP IRL HKG NER PAN ROU ECU SVK LUX ISR POL BEN IDN COL COD GMB TLS HND NIC SLV TTO MYS ARG MUS CRI ESP FRA KOR MLI VUT CAF BFA GTM DOM KEN AZE MEX CHL BEL BTN 0.6 GIN ETH COG BGD NRU PHL BRA GEO BRN UKR SRB HRV ITA SLE BWA FJI KGZ ALB GHA FSM JAM LKA MNG CIV HTI TON SAU MNE GRC TGO TUV PAK GUY ARM TUR 0.5 TCD NGA PNG NAM IND MKD RWA LSO SSD SEN GAB TJK BIH MDA AFG SDN ZAF MAR TUN COM LBN SWZ IRN 0.4 MRT MWI LAO KIR EGY WSM UZB IRQ DZA AGO MHL NPL ZMB PSE 0.3 YEM XKX JOR 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Human Capital Index ILO data JOIN data Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 169 economies with available data. Nepal (NPL) and Nicaragua (NIC) are in red. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 113 Figure 4.2: Basic UHCI vs. Human Capital Index 0.8 0.7 ISL JPN SGP CHE MAC SWE NLD 0.6 DEU NZL GBRCAN EST NOR QAT VNM AUS FIN HKG DNK CZE ARE CYP AUT SVN IRL MLT PRT KOR BLR Basic UHCI CHN LTU POL 0.5 SYC BHR RUS LVA USA ISR FRA HUN BEL THA LUX ESP SVK KWT KNA OMNKAZ HRV PER BGR CHL KHM 0.4 PRY URY PLWCOL ECUMYS TTO MUS SRB ITA ZWE ROU CRI BRN MEX GRC IDN ARG UKR MDG SLB PAN AZEKGZ MNGALB HND NIC GEO MNE SLV BRA KEN LKA TZA VUT SAU TLS DOM TUR PHLJAM 0.3 UGA BDI CMR GTM MMR NRU BTN FJI MKD ARM MOZ GMB TON BEN BGD GUY FSM MDA COD LBR BFA GHA COG HTI TGO TUN BIH UZB ETH BWA TUV IND TJK WSM IRN NER SLE GIN PNG GAB MAR LBN 0.2 MLI CAF NGA CIV PAK LSOSEN NAM ZAF LAO EGY NPL DZA PSE RWA AFG COM TCD SDN MWI KIR XKX AGO IRQ SWZ MHL JOR SSD ZMB YEM MRT 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Human Capital Index Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 169 economies with available data. Nepal (NPL) and Nicaragua (NIC) are in red. Dashed line is 45-degree diagonal, and solid line is a fitted model where y = –0.14 + 0.88x. UHCI = Utilization-Adjusted Human Capital Index. Figure 4.3: Employment to population (basic utilization) and per capita income 0.9 QAT MDG SLB ISL TZA KHM ZWE VNM 0.8 BDI MOZ ARE CHE NZL JPN SWE MAC NLD THA DEU DNK Employment−to−population LBR SYC CZE EST BHS GBR NOR BLR KWT CAN AUS UGA LTU BHR AUT FIN SGP LVA MLT ratio (basic utilization) BOL RUS SVN USA 0.7 GNB CMR PRYPER CHN HUN KAZ BRB OMN PLW URY KNA PRT CYP ISR HKGIRL CAF MLI HND IDN ECU BGR POL PAN SVK KOR LUX BFA BEN MMR MYS TTO FRA COL ROU BEL VUT NER ETH KGZ KEN TLS NIC AZE DOM MUS CHL VEN ARG ESP COD GMB BTNSLV BLZ GTM MEXCRI BRN 0.6 COG MNG GEO PHL SRB BRA HRV NRU ITA TGO GIN BGD UKR ALB SUR JAMFJI BWA MDV CUW GHA LKA GRC SLE HTI CIV MNE FSM TUV CPV ARM GUY TONMKD TUR SAU 0.5 TCD PAK NGA PNG LSO IND NAM TJK MDA BIH GAB SSD RWA SEN SDN MAR TUN ZAF LBN COM EGYSWZ LBY IRN 0.4 MWI AFG MRT UZB KIR DZA LAO WSM IRQ NPL AGO MHL ZMB PSE STP 0.3 YEM XKX DJI JOR 0.2 5 6 7 8 9 10 11 12 log GNI per capita (Atlas) ILO data JOIN data Missing HCI data Source: World Bank calculations based on the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 182 economies with available data. Working-age population is 15–64 years of age. GNI = gross national income; HCI = Human Capital Index. to work outside the home (see section 4.7 titled rates of about 0.8, others—including Afghanistan “Disaggregation by Sex” for a discussion). (AFG), Malawi (MWI), and Nepal (NPL)—have employment rates of about 0.4. In part, this dis- Employment rates vary widely among low- parity may reflect the 2013 change in the ILO defi- and  middle-income economies. Whereas many nition of employment to exclude own-use pro- low-income economies have high employment duction workers (mostly subsistence agriculture), 114 U tilizing H u man Cap ital which has been applied in some countries but not double moving to full human capital. Moving to in others. These measurement issues also moti- 12 full human capital and complete utilization of vate using a more specific definition of employ- that human capital, however, results in long-run ment in the full UHCI (discussed in the next GDP per capita that is 3.0 times the status quo in section). Nicaragua, but 5.4 times the status quo in Nepal.13 To understand the implications of differences The basic UHCI is fairly flat over a wide range of log between the UHCI and HCI for long-run eco- income before increasing (figure 4.4). Specifically, nomic growth, consider the example of two econ- the UHCI is almost flat moving from low income omies, Nepal and Nicaragua. These economies (0.23) to lower-middle income (0.26), as higher HCI have similar scores for the HCI (0.5) but very scores are largely offset by lower utilization rates. different employment rates (figure 4.1, red dots). But the UHCI then increases rapidly to upper-­ The employment rate in Nicaragua is about 0.65, middle income (0.32) and high income (0.51), as which is above the median, against that in Nepal, both human capital and utilization rates increase 0.37, which is around the fifth percentile. These together. disparate employment rates mean that the basic UHCI score of Nepal (0.18) is much lower than that of Nicaragua (0.33) (figure 4.2, red dots). As 4.3  THE FULL UHCI mentioned previously, the increase in long-run per capita income moving to full human capital One conceptual issue with the basic utilization is 1/HCI times that in the status quo, and long-run measure (employment rate) is that it assumes per capita income moving to full human capital that all jobs are the same in terms of their abil- and complete utilization is 1/UHCI that in the sta- ity to utilize human capital. In practice, however, tus quo. An HCI score of 0.5 for both economies a large share of employment in developing coun- implies that long-run per capita incomes would tries is in jobs in which workers cannot fully use Figure 4.4: Basic UHCI and per capita income 0.7 SGP ISL JPN SWE MAC NLD 0.6 NZL CAN CHE EST GBR FIN AUS DEU NOR VNM CZE HKGDNK QAT SVN AUTIRL PRT KOR ARE BLR LTU CYP MLT 0.5 POL LVA ISR FRA BEL USA RUS HUN SYC BHR Basic UHCI ESP LUX THA CHN SVK KAZ HRV OMN ITA PER BGR KHM KNA KWT 0.4 COL ECU SRB MYS CHL MUS URY PLW TTO ZWE MEX CRI ROU ARG GRC BRN PRY KGZ SLB UKR MNG ALB IDNAZE GEO MNE MDG KEN SLV LKA BRA TUR PAN TZA NIC MMR HND JAM DOM SAU 0.3 BDI TLS BTN VUT PHL ARM FJI MKD NRU MOZ UGA CMR GTM TON GMB BGD MDA FSM BIH BEN COG GHA UZB GUY BFA HTI LBR ETH TJK IRN TGO IND TUV BWA COD GIN MAR TUN WSM LBN 0.2 NER MLI CIV PAK PNGEGY DZA NAM GAB SLE KIRPSE CAF NPL LSO SEN ZAF AFG RWA NGA LAO MWI COM XKX MRT SDN SWZJORIRQ MHL SSD TCD ZMB AGO YEM 0.1 6 7 8 9 10 11 log GNI per capita (Atlas) Source: World Bank calculations based on the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 182 economies with available data. Working-age population is 15–64 years of age. GNI = gross national income; UHCI = Utilization-Adjusted Human Capital Index. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 115 their human capital. For example, in the poorest The share of employment in better jobs (SEBJ) countries, about half of all workers work on fam- increases from about 20 percent in low-income ily farms or as agricultural laborers, situations countries to 80 percent in high-income countries with low productivity (Merotto, Weber, and Reyes (Merotto, Weber, and Reyes 2018). The main cat- 2018). For the rest, about two-thirds of nonagri- egories excluded from the definition are subsis- cultural workers are self-employed or unpaid in tence own-account/family agriculture, small-scale family businesses. They include many small-scale traders, and landless agricultural laborers, because traders selling household goods or food, with most these employment types are common only in of their time spent waiting for customers. low-income countries—suggesting they are more likely to have lower productivity. By using a more Although there is scope for human capital to specific definition of employment, the full UHCI increase productivity in these jobs, that scope also avoids variation in utilization rates caused by is limited. Filmer and Fox (2014) compare the differences in the definition of employment that income of household enterprise owners of dif- affected the basic UHCI. ­ ountries. ferent education levels in four African c On average, the increase in income due to edu- The definition of better employment is based on cation, although positive, is much less than the way that the work is organized, rather than would be predicted given the number of years whether the job is formal or informal. For exam- of ­schooling.14 Most developing countries suffer ple, nonagricultural employees could be formal or from high rates of mismatch between the level informal.16 Better employment involves work orga- of education required for a job and the education nized in a team consisting of at least an employer of the people doing it—such as the well-known and an employee, and for which employees are paid anecdote of unemployed engineers driving taxis (rather than working out of familial obligation). This (see Battu and Bender 2020 for a survey). 15 The arrangement allows a minimum degree of special- literature often refers to this mismatch as overed- ization and organization, which helps boost produc- ucation, though a more appropriate description tivity and allows for people to use their skills.17 is underutilization, because the lack of jobs and education causes the mismatch. not the level of ­ A second conceptual issue with the basic measure is In some regions, especially the Middle East and that utilization should be relative to potential, which North Africa, underutilization is often associated will depend on how much human capital there with self-­ employment, for example, while waiting is to underutilize. That is, a doctor working as an for a ­ formal sector job (Gatti et al. 2013; Handel, agricultural laborer represents severely underuti- Valerio, and Sanchez Puerta 2016). lized human capital, whereas the human capital of a worker with no education doing the same job is To address this mismatch, the full UHCI introduces closer to being fully utilized. This distinction means a concept of better employment, which is designed to that the utilization scores of countries with higher capture the employment categories in which peo- levels of human capital should be more heavily ple can better use their human capital (subject to penalized by a lack of better employment.18 available data). More specifically, better employ- ment is defined as nonagricultural employees, plus Putting these concerns together suggests that the employers. This definition is not intended as a value full UHCI should depend on the better employ- judgment but rather is based on the types of jobs ment rate (BER)—as a share of the working-age that are relatively rare in low-income countries but population—rather than on the raw employment income countries—­ common in high-­ suggesting rate. But the full utilization rate is not simply the they are associated with higher productivity. BER, because the BER fails to adjust for how much 116 U tilizing H u man Cap ital human capital there is to underutilize if people capital and complete utilization in better employ- are not in jobs where they can fully use the human ment (see Pennings 2020).19 capital. Instead, utilization rates for those without better jobs should depend inversely on the HCI In terms of data, the BER is constructed as the (relative to a natural minimum). The full utili- employment rate (as in the basic utilization zation measure captures both concerns. The full measure) multiplied by the SEBJ. The measure- UHCI is a weighted average of the country’s HCI ment of the SEBJ requires data on the number of (for those in better employment) and the mini- employers, nonagricultural employees, and total mum HCI (for the rest) and is described further in employment. The primary source is the ILO series box 4.2. The full UHCI can also be derived using “Employment by sex, status in employment, and the increase in long-run GDP per capita moving economic activity (thousands),” using the most from the status quo to a world with full human recent year available.20 The secondary source is JOIN. At the time of writing, the public JOIN Box 4.2: Definition of the full Utilization-Adjusted Human Capital Index The full utilization rate is a weighted average of the utilization rates of those in better employ- ment, and the utilization rate of the rest of the working-age population. Workers in better employ- ment (with the better employment rate expressed as BER) are assumed to be as productive as their human capital allows—their human capital is fully utilized (utilization rate of 1). All others, a fraction (1 – BER) of the working-age population, are assumed to be only as productive as raw labor; hence, any excess human capital is underutilized. In the HCI​ , raw labor has productivity of H ​ = 0.2.​This is the productivity of a worker with zero ​ C​Imin​​ years of schooling and the worst possible health outcomes.a In contrast, the potential produc- tivity of a worker in better employment is just HCI​ . Hence, the worker’s productivity relative to ​ C​I​min​​ / HCI​. For example, in an economy with HCI = 0.4, workers potential, or utilization rate, is H without better employment will be half as productive as they could be if they were in better employment (0.2/0.4), so their utilization rate is 0.5. This result means that a shortage of bet- ter employment leads to more severe underutilization in countries with more human capital. The full utilization measure is given by: HCI min Utilization (full measure) = BER × 1 + (1 − BER × ) HCI (B4.2.1) The full UHCI is the full utilization measure multiplied by the HCI, as in equation (4.1) in the main text, meaning that the full UHCI is a weighted average of the HCI (for the share of the population in better employment) and the minimum HCI (for the rest of the working-age population): UHCI (full measure) = BER × HCI + (1 − BER ) × HCI min (B4.2.2) a The minimum HCI score is derived by assuming zero years of schooling, complete stunting, and zero chance o f adults surviving to age 60. HCI min = 1 × e0.08×(0−14) × e(0.65×(0−1)+0.35×(0−1))/2 ≈ 0.2 . See Kraay (2018), equations 9–12. The probability of survival to age 5 is assumed to be 1, because it does not affect the growth calculations. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 117 dataset provides a split by status in employment of the working-age population in better employ- or economic activity, not both, so the SEBJ is cal- ment (figure 4.6), which drives the high utiliza- culated using an unpublished version constructed tion rate. For low-income economies, only about from the underlying microdata. The most recent percent of the working-age population is in better 10 ­ data source is used if both the ILO data and JOIN employment, so the utilization rate for these econ- are available. For many economies, both sources omies is mostly determined by how much human are missing data on the number of agricultural capital there is to underutilize. In the 10 lowest-­ employees. In these cases, the number of agri- income economies, HCI​ ≈ ​ Imin​​ / HCI​ is HC​ 1/3, so ​ cultural employees is interpolated using ILO data about 0.2/0.33 = 0.6 (close to the full utilization rate on total agricultural employment (which is more for those economies in figure 4.5). The full utiliza- widely available). The full utilization measure is income tion rate falls from low-income to middle-­ available for 161 economies. economies, as higher rates of human capital mean that there is more human capital to underutilize (and the BER increases only slightly). 4.4  THE FULL UHCI IN THE DATA The full UHCI also has the same shape in per The full utilization measure has the same U shape capita income as the basic UHCI (and similar in log per capita income as the basic utilization mean values for each income level; figure 4.7). measure, and similar mean values overall (0.62) lowest-income economies, however, the For the ­ and for each income level, though with less disper- UHCI value converges almost exactly to 0.2, with sion (figure 4.5). The U-shaped pattern, however, little variation (as against wide variation in the ­ has quite different causes from those driving the minimum basic UHCI). The reason is that 0.2 is the ­ basic utilization measure. For the full measure, the productivity of HCI score, which is the assumed ­ ­ highest-income economies have about 70 percent raw labor for those without better employment. Figure 4.5: Full utilization rate and per capita income QAT 0.9 ARE KWT ISL DEU MAC 0.8 BHR SWE OMN DNK CHE BLR SYC EST JPNAUT NOR RUS KNA MLT PLW NZL CAN AUS HUNLVA LTU GBRHKG FIN CAF BGR CZE ISR SGP SVN 0.7 TCD KAZ SVKPRTCYPBRN FRANLD TTO SAU IRL LUX ETH BWA HRV ESP BEL POL Full utilization rate ZAF MUS NRU MLI ROU ARG MYS URY PAN NER LBR UKR SWZ PRY PERMNE KOR KHM BRA MOZ COD MKD DOMCHN CHL ITA 0.6 UGA RWA YEM MRT TJK PAK HND BTNAGO GTM SLV NAM GUY PHLMHL FJI TUV JAM SRB MEXCRI SLE CMR COM SLB CIV NIC TUN BIH TZA COG ZMB EGY GRC MDG BFA GMBGIN GHA IDN THA TUR BDI SENBGD GEO DZA ARM AZE GAB AFG BEN COL HTI MMR TLSVNM VUT UZB MNG FSM WSM LKA TON ECU MDA PSE JOR 0.5 TGO NPL ZWE LSO LAO MAR XKX IRN KGZ ALB IND LBN KEN 0.4 0.3 0.2 5 6 7 8 9 10 11 12 log GNI per capita (Atlas) ILO data JOIN data ILO (interpolated agricultural employees) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 161 economies with available data. GNI = gross national income. 118 U tilizing H u man Cap ital Figure 4.6: Better employment rates and per capita income 0.9 QAT 0.8 ARE ISL MAC (share of working-age population) KWT DEU SWE 0.7 BHR EST JPN CHE DNK NOR BLR OMN NZLAUT RUS SYC MLT CAN GBR AUS HKG SGP LVA LTU CZESVN FIN HUN KNA ISR 0.6 Better employment BGR PLW PRT CYP NLDIRL SVK BRB BRN FRA BEL LUX KAZ POL HRV TTO SAU ESP KOR 0.5 MUS MYS ARG UKR ROU URY ITA SURPERMNECHNNRU CHL CUW SRB 0.4 PRY MKD MEXCRI PAN GRC BRA MDV BLZ BWA KHM SLV ZAF DOM BIH TURVEN VNM PHL JAM FJI THA GUY ETH TJK TUNAZE COL 0.3 HND NIC CPV MNG UZBBTN GEO IDNARM EGY LKA DZA GTMNAM ECU MDA TUV WSM BOLPSEJOR SWZALB XKXMHLIRN KGZ DJI FSMTON 0.2 RWA NPL MMR BGD PAK COG SLB GHA MAR GAB GMB UGA COM CMR YEM ZWESTP VUT MRT TLS HTI TZA SEN ZMB KEN CIV IND LAO AGO LBN 0.1 BDI MDGCOD MOZ AFGLBR TGO BEN CAFSLEBFA MLIGIN NER TCD LSO 0 5 6 7 8 9 10 11 12 log GNI per capita (Atlas) ILO data JOIN data ILO (interpolated agricultural employees) Missing HCI data Source: World Bank calculations based on the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 171 economies with available data. GNI = gross national income; HCI = Human Capital Index. Figure 4.7: Full UHCI and per capita income 0.7 SGP MAC SWE JPN ISL 0.6 EST HKGDNK DEU CAN NOR CHE NZLFIN QAT GBR AREAUTAUS SVN IRL NLD CZE PRT CYP BLR MLT ISR FRA LVA LTU BHR KOR BEL RUS 0.5 POL HUN HRVSYC ESP LUX OMNSVK KWT Full UHCI BGR ITA BRN KAZ PLWKNA MUS TTO 0.4 UKR SRB MNE PER CHN CHL GRC MYS ARG URY SAU ROU CRI VNM MEXTUR MKDTHA BRA NRU BIH PRY UZB MNG SLV COL PAN ARM AZE LKA GEO PHL JAM ECU DOM 0.3 TJK KHM KGZ NIC HND BTN IDN MDA TUN PSE ALB DZA WSM XKX GUYFJI IRN EGY JOR TON ZAF BWA GTM FSM TUV NAM NPL ETH MAR MMR BGDSLBGHA MHL GAB KEN GMB UGARWA HTI ZWE COM CMR SEN YEMZMB COG IND LAO VUT SWZ PAK TLS LBN MDG TZAMRT CIV AGO BDI COD MOZ TGO AFG BFA SLE LBR BEN GIN TCDMLI 0.2 NERCAF LSO 0.1 6 7 8 9 10 11 log GNI per capita (Altas) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 161 economies with available data. GNI = gross national income; UHCI = Utilization-Adjusted Human Capital Index. 4.5 COMPARING THE UTILIZATION individual economies (figure 4.8; correlation of only MEASURES 0.6).21 The strongest correlation is for high-income economies, because, in order to generate high per Although the full and basic utilization measures capita incomes, employment rates need to be high have the same U-shaped relationship with per and those people working need to be productive. capita income, they often differ substantially for But, for lower-income economies, the drivers of T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 119 Figure 4.8: Basic utilization vs. full utilization QAT 0.9 ARE KWT ISL MAC 0.8 BHR DEU DNK NOR SWE CHE OMN AUT BLREST JPN SYC PLW KNA RUS MLTAUS CAN NZL BGR HUN LVA HKG ISR LTU FIN SGP GBR CZE BRN SVN 0.7 SAU CAF BEL TTO SVK FRALUX CYP IRL KAZ PRT TCD BWA HRV NLD Full utilization ETH NRU MUS ESP POL ZAF ARG ROU MYSKOR URY LBR SWZ MLI NERPAN PER MNE UKR PRY MKD ITA BRACHLDOM CHN UGA KHM 0.6 YEM AGO MHL MRT TUN TJK RWA NAM GUY PAK TUV FJI PHL SRB MEX CRI JAM BTN GTM SLVCOD HND CMR MOZ TZA SLB COM BIH EGY CIV SLE GRC NIC ZMB TUR COG GIN GMB IDN MDG DZA GAB GHA GEO BFA THA BDI JOR NPL SEN AFG ARM BGD AZE COL WSM UZB FSM LKA MNG BEN VUT MMR VNM TON HTI TLS MAR MDA ZWE 0.5 PSE IRN LSO TGO ECU XKX LAO IND ALB KGZ LBN KEN 0.4 0.3 0.3 0.4 0.5 0.6 0.7 0.8 0.9 Basic utilization High income Upper-middle income Lower-middle income Low income Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 161 economies with available data. Dashed line is 45-degree diagonal; solid line is a fitted model where y = 0.37 + 0.43x. Figure 4.9: Basic UHCI vs. full UHCI 0.7 MAC SGP SWE JPNISL 0.6 HKG DEU DNK QAT NOR EST CANCHE FINGBRNZL AUT AUS ARE IRL SVN NLD CZE ISR FRA CYPPRT BLR MLT BHRBEL LVA LTU KOR RUS POL 0.5 ESP HUN HRVSVKLUXSYC OMN KWT ITA Full UHCI BRN KNA BGR PLW KAZ MUS TTO 0.4 SAU MNEUKR ARG SRB CHL GRC MYS URYPER CHN ROU CRI TUR MEX VNM BIH MKD NRU BRA PRY COL THA UZB PAN SLV MNG ARMJAM PHL LKA GEOAZE 0.3 PSE DZA XKX ZAF TUN WSMTJKIRN MDA GUY FJI DOM IDN NIC KGZ ALB ECU KHM JOR EGY BWA BTN HND TON GTM NAM TUV FSM MHL NPL GAB MARETH GHA BGD MMRKEN SLB SWZ PAKLBN COGGMB VUT TLS ZWE YEM ZMB MRTCOMSEN CIV IND RWA LAO HTI CMR UGA AGO AFG SLE TGO CODBFA MLI GINLBR BEN MOZ BDI TZA MDG 0.2 TCD CAF LSONER 0.1 0.1 0.2 0.3 0.4 0.5 0.6 0.7 Basic UHCI High income Upper-middle income Lower-middle income Low income Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 161 economies with available data. Dashed line is 45-degree diagonal; solid line is a fitted model where y = 0.05 + 0.89x. UHCI = Utilization-Adjusted Human Capital Index. high utilization vary across measures, and the sim- there is little variation in full utilization rates across ilarity of average scores is coincidental. Specifically, low-income economies, because those economies employment rates (basic utilization) are often high have little human capital to underutilize. in low-income economies because people cannot afford not to work, though the rates vary significantly For the UHCI, the scores of individual economies are because of the inconsistent cross-country classifica- very similar in the full and basic measures (figure 4.9; tion of work in subsistence agriculture. In contrast, correlation 0.93). In part, this similarity is because full 120 U tilizing H u man Cap ital Figure 4.10: Regional average UHCI or HCI 0.8 0.69 0.59 0.57 0.55 0.6 0.48 0.47 0.45 0.40 UHCI or HCI 0.38 0.38 0.38 0.35 0.34 0.4 0.32 0.26 0.24 0.24 0.23 0.2 0 a ia an a sia fic ric ric As ci be lA Af Af Pa h rib ra ut n th d nt ra Ca So or an Ce ha N e ia Sa th d d As an an b- d an st Su st pe Ea Ea a ro ic Eu e er dl Am id M tin La HCI Basic UHCI Full UHCI Source: World Bank calculations based on the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Figure reports regional averages. HCI = Human Capital Index; UHCI = Utilization-Adjusted Human Capital Index. and basic UHCI have the HCI as a common compo- Asia and Pacific scores marginally higher, followed nent. It is also because the differences between the by Europe and Central Asia. two utilization measures occur mostly for countries with a low HCI, which mechanically shrinks any dif- ferences in utilization rates when forming the UHCI.22 4.7  DISAGGREGATION BY SEX Many of the trends above are driven by differences 4.6  DISAGGREGATION BY REGION in utilization rates by sex. Although the HCI is roughly equal across sex with a slight advantage for Regions line up similarly according to the UHCI females relative to males, female utilization rates and to the HCI (figure 4.10), though UHCI scores figure 4.11 are typically lower than those for males (­ are lower. Sub-Saharan Africa has the lowest HCI and figure 4.12) leading the UHCI also to be lower (of about 0.40) and also the lowest UHCI (of about for females than males (figure 4.13 and figure 4.14). 0.23). South Asia has a similar UHCI, but a higher Male and female UHCI scores increase proportion- HCI (reflecting slightly lower utilization rates). implying ately, but with a constant gap for females (­ Latin America and the Caribbean and the Middle a larger percentage gap at low UHCI scores). The East and North Africa are next, with HCI scores of gender gap is larger for the basic measure than about 0.56 and UHCI scores of about 0.35, though the full measure. Perhaps surprisingly, when the Middle East and North Africa does relatively women join the labor force, they often move rap- better for the full UHCI than for the basic UHCI, idly into better employment (see Pennings 2020, reflecting higher rates of wage employment. East figure  25A). More generally, female employment T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 121 Figure 4.11: Employment-to-population ratio (basic utilization) and per capita income a. Females b. Males 1.0 1.0 QAT ARE NER MDG TZA KHM SLB OMN BHR KWT ZWE VNM GTM ISL MDG SLB HND JPN CHE ISL NIC PRY NZLNLD MAC 0.8 THA CZEMLT Employment−to−population Employment−to−population BHS BDI TZAKHM ZWE VNM 0.8 UGA BDI MOZ LBR MLI MMR BGD PAK IDN BOL LKA SLVBLZ ECU COL PER FJI DOM MEX CHN KAZMYS PANKNA MUS PLW SAU EST DEU GBR SGP SWE AUS AUT HKG CAN DNKNOR MOZ SWE CHE CMR BLR RUS MDV HUN CRI TTO SVN SYC VENURY KOR USA GMB KGZ ratio (basic utilization) ratio (basic utilization) BLR SYC NZLNLDDNKMAC NOR BFAETH GNB BEN IND BTN PHL POL NRU ROU LVASVK LTU BRBPRTCYP FINIRL LBR LTU EST BHSGBR DEU AZE ARG CHL ISR UGA LVA CAN FIN JPN AUS TLS VUT SUR BGR TUV BRATUR LUX THA RUS BRB CZESVN PRT AUT ISR SGP CAFAFG TJK MNG MAR GEO FSM IRQ JAMIRN ITAFRA ESP BRN BEL NER CMR CYP USA COD HTI CIV KEN COG UKR EGY TUN TON DZAGUY ALB SRB LBN BWA HRV GRC TGO BOL BGR HUN IRL LUX HKG TCD CUW PER KAZ POL KNAMLT FRA GIN LSO SDN 0.6 CAF COD GNBGIN KEN PRY CHN SVK URY PLW BEL QAT 0.6 YEM MRT ARM MKD MNE SLE BEN TLS VUT COG ROU TTO KOR ESP SEN GHA CPV GAB BFA GMB GHA UKR MNG GEO ECU COL AZE SRB HRV PAN CHL CUW BRN SLE RWA COM NGA PSE BIH LBY MLI MMR IDN BWA MYSARG BRA NPL WSM ETH HND BTN ALB DOM VEN ITA ARE TGO UZB NAM NICPNG SLV JAM MNE MUS KWT MWI PNG AGO KIR MHL JOR ZAF CPV SUR CRI NRU SSD MDA SSD HTI KGZCIV NGA MDA PHL NAM BLZ ARM MEX GRC SWZ XKX BHR LAO 0.4 RWA TCD FSMGUY GTM MKD TONFJI MDV 0.4 ZMB DJI BGD LAO LKA ZAF SWZ GAB STP BIH TUV MWI LSO SEN KIR AGO TUR COM UZB MRT STP WSM LBY OMN TJK NPL ZMB MAR MHL 0.2 IND SDN TUN PAK AFG EGY LBN SAU 0.2 DJI DZA IRN XKX PSE JOR IRQ YEM 0 0 5 6 7 8 9 10 11 12 5 6 7 8 9 10 11 12 log GNI per capita (Atlas method) log GNI per capita (Atlas method) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Data for 182 economies. Working-age population is 15–64 years of age. GNI = gross national income. Figure 4.12: Full utilization rate and per capita income a. Females b. Males 1.0 1.0 QAT ARE BHR KWT OMN NRU SAU MAC ISL DEU ISL CHE 0.8 SWE MAC 0.8 KNAMLT DNK NOR SWE HKG AUT RUS HUN EST NZL AUS Full utilization rate DNKNOR Full utilization rate DEU BLR CZE BRN CANSGP EST CHE BGR PLW LVA GBR ISR LTU LVA FIN AUT MYS LTUSVKSVN CYP FIN RUS CAN NZL GBRAUS KAZ NLD LUX BGR ISR ETH BWA ARG PRT KOR FRA HRV BELIRL HUNKNAPRT SVN CZE HKG QAT SGP IRL LBR EGY ZAF ROUPOL PAN TCD KAZ SVK MLT PLW CYP FRA NLD TCD TUN PER MEX PRY TUR URY ESP BRN BEL LUX MLI YEM RWA MRTPAK HNDUKRBTN SLV DZA AGO PHL GTM MKD GUY MHLFJI MNECHN CHL ITA HRV ESPKWT NER CRI NER MLI ETH LBR BWA ZAF MNEROU POL NRU URY ARE COD GMB COG COM BGD CMR SLB NIC JORBIH IDN PSE SRBDOM GRC JAM 0.6 UKR AGO PRY PERDOM ARG PAN BHRKOR 0.6 BDI MDGSLE CIV GHA TZA SEN XKX AZE GEO ARMTUV GAB THA IRN COD GIN SRB CHN MKD JAM MYSCHL ITA GIN MMR TLS BFA BEN AFG HTI VNM UZB FSM VUT WSM TON ECU COL AFGRWA YEM SLE CIV CMR PHL GUY CRI MNG MDG BFA TZA MRT COM SLB HND GTM MHLFJI THA MEX GRC LAO BEN PAK NICMDA COG BTN GEO SLV MNGARM BIH COL TGO KGZ IND MDA ALB BDI GMB TGO SEN GHAVNM VUT AZE IDN GAB HTI MMRBGD UZB LAO TUN FSM OMN WSM TON ECU KGZ TLS EGY ALB TUR SAU IND DZA 0.4 JOR PSE IRN 0.4 0.2 0.2 5 6 7 8 9 10 11 12 5 6 7 8 9 10 11 12 log GNI per capita (Atlas method) log GNI per capita (Atlas method) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: GNI = gross national income. rates are strongly U-shaped in the level of income, Figure 4.15 breaks down the HCI and UHCIs whereas male employment rates are much flatter by gender and region. In almost all regions, the (see Goldin 1995). 23 The largest gaps in utilization female HCI is higher than male HCI (equal for rates across sex (for both measures) are for several South Asia). The opposite is true, however, for the oil/gas producers: Bahrain, Kuwait, Oman, Qatar, UHCI: in almost all regions the female UHCI is and Saudi Arabia. These economies have very lower (for Europe and Central Asia, the full UHCI is high male employment rates—almost all of which similar for males and females). The largest gender is represented by wage employment (perhaps due gaps for the basic UHCI are in the Middle East and workers)—but low or average female to migrant ­ North Africa and in South Asia. In these two regions utilization rates. the female basic UHCI is very low, reflecting low 122 U tilizing H u man Cap ital Figure 4.13: Basic UHCI and per capita income a. Females b. Males 0.7 0.7 SGP ISL NZLNLD ISL MAC CHE SWE 0.6 SWE MAC 0.6 CZE ARE KOR CAN GBR HKG DEUAUS QAT EST NLD FIN SGP NOR CHE EST IRL DNK NOR VNM NZL CAN AUT FIN GBR DNK VNM POL SVN PRTMLT BHR CYP DEUAUS BLR LTU SVN HKG LVACZE PRT AUT 0.5 OMN BEL USA FRA ISR 0.5 ISR IRL BLR CHN HUN LVA CYP THA RUS LTU ESPITA RUS POL FRA COL MEX PER KAZ SVK KWT LUX KOR BEL USA ECU CRICHL KNA MYS TUR MLT HRV SAU Basic UHCI Basic UHCI HUN LUX KGZ PRY SRB BGR ARGURY PLWGRC SVK IDN 0.4 THA BGR KAZ CHN HRV ESP QAT 0.4 NIC HND MNG AZE SLV ALB IRN MNE GTM ROU PAN BRN URY MMR FJI DOM SLBUKR PER SRB KNA PLW ITA BGD IND UKR SLB PHL GEO NRU MDG MNG ALB ECU GEO COL ROU CHL MYS BRN ARE MDG TZA PAK BTN DZA MAR TON ARM JAM MKD BDI TUN FSM BIH ARG GRC 0.3 TZA IDN AZE PRY MNE CRI BHR KWT 0.3 BDI NER AFG GMB BEN HTI CMR TLS UZBVUT PSE GUY EGY TUV KGZ MEX PAN BFAETH WSMIRQ TGO TLS MDA VUT SLVARMJAM COGGHA MDA JOR CMR GHA MMR DOM MLI XKX GAB BWA GMB COG NICHND BTNPHL MKD NRU COD LBR GIN CIV SEN SDN COD LBR BEN BWA TUR YEM MRT COM KIR 0.2 NER SLE GIN BFA HTI ETH PNG UZB FSM TONBIH GUYFJI 0.2 TGO RWA SLETCD NGA LAO PNG MHL ZAF KIR GTM ZAF AGO MLI RWA CIV BGD NGALAO WSMTUV GAB OMN SEN COM TUN 0.1 TCD MRT IND MAR EGY AGO MHLIRN SAU 0.1 PAK SDN DZA AFG XKX PSE JOR IRQ YEM 0 0 6 7 8 9 10 11 6 7 8 9 10 11 log GNI per capita (Atlas method) log GNI per capita (Atlas method) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 148 economies with available data. GNI = gross national income; UHCI = Utilization-Adjusted Human Capital Index. Figure 4.14: Full UHCI and per capita income a. Females b. Males 0.7 0.7 SGP MAC SWE MAC ISL SGP NOR ARE SWE CHE HKG 0.6 EST FIN CAN HKG DEU DNK 0.6 DEU QAT CAN NZL ISL NOR DNK GBRAUS CHE NZL EST AUT GBR AUS LTU SVN BHR LVA IRL AUT CZESVN KOR FINIRL NLD PRT CZE ISR NLD PRTCYP MLT RUS CYP FRA POL FRA BEL BEL OMN ISR POL MLT HUN 0.5 HUN KOR 0.5 BLR RUS LVA LTU SAU ESP KWT LUX HRV SVK ESP QAT LUX HRV ITA SVK Full UHCI Full UHCI BGR KNA BRN KAZ ARE BGR KAZ KNA BRN ITA CHN MYS TUR PLW UKR PLW BHR PERMNE SRB CHL GRC ARG 0.4 SRB MNE URY KWT 0.4 UKR MEXNRU CRI ROU URY CHN CHL GRC ROU ARG VNM VNM PER MYS CRI BIH PRY MKD UZB PSESLV THA DZA TUNAZE IRNCOL PAN MKD MNG JAMCOL THA MEXNRU IDN MNG EGY JOR XKX PHL ARM ECU FJI MDA GEO ARMBIHPRYDOM PAN OMN HND BTN GEOALB JAM DOM GUY 0.3 UZB SLV AZE PHLALB GUY ECU FJI BWA TUR 0.3 KGZ NIC MDA FSM WSM GTM TONZAF KGZ NIC IDNWSMZAF ETH BGD PAK BWA HND TUN SAU MMR GHA MHL GAB TON COG SLB ETH MMR SLB BTNFSMGTM DZA HTI YEM GMB RWA COM MRTCMR TLS SEN IND VUTAGO TUV BGD GHA VUT PSE EGY JORMHLIRNGAB AFG COD TZA CIV LAO MDGSLE GMB RWA HTI TZA SEN COG CMR COM INDLAO TLS BDI MDG TGO LBRMLI BFA BEN GIN TGO CIV NER SLETCD 0.2 BDI NER CODBFA LBR AFGTCD BEN MLI MRT GIN YEM PAK AGO 0.2 0.1 0.1 6 7 8 9 10 11 6 7 8 9 10 11 log GNI per capita (Atlas method) log GNI per capita (Atlas method) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: Based on 138 economies (panel a) or 141 economies (panel b) with available data. GNI = gross national income; UHCI = Utilization- Adjusted Human Capital Index. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 123 Figure 4.15: Gender gaps in HCI and UHCI, by region a. Gaps in Europe and Central Asia, East Asia and Pacific, and Middle East and North Africa 0.8 0.71 0.67 0.60 0.59 0.6 0.57 0.55 0.48 0.47 0.47 UHCI or HCI 0.43 0.44 0.41 0.41 0.39 0.4 0.36 0.33 0.33 0.19 0.2 0 Europe and Central Asia East Asia and Pacific Middle East and North Africa b. Gaps in Latin America and Caribbean, South Asia, and Sub-Saharan Africa 0.6 0.57 0.54 0.45 0.45 0.41 0.39 0.4 0.38 0.35 0.33 UHCI or HCI 0.32 0.29 0.27 0.25 0.24 0.22 0.22 0.20 0.2 0.13 0 Latin America and Caribbean South Asia Sub−Saharan Africa HCI − Male HCI − Female Basic UHCI − Male Basic UHCI − Female Full UHCI − Male Full UHCI − Female Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI), the World Development Indicators, the International Labour Organization, and the Global Jobs Indicators Database. Note: HCI = Human Capital Index; UHCI = Utilization-Adjusted Human Capital Index. 124 U tilizing H u man Cap ital Box 4.3: Closing gender gaps in human capital outcomes: Where do we go from here? The Human Capital Index (HCI) approach implicitly assumes that human capital investments translate into productivity through labor market opportunities. How human capital is utilized in terms of paid work and labor markets, however, varies considerably. In particular, consider- able and well-documented gaps exist in labor market opportunities between men and women. Globally, only 50 percent of women participate in the paid labor force, whereas 80 percent of men do. Across countries, the gender wage gap persists at about 20 percent, on average (International Labour Organization 2018). Women work in lower-paying occupations and jobs. Across the globe, only one in five firms has a female top manager.a Although these outcomes might in part reflect optimizing decisions within the family (for example, see Chioda 2016 for evidence from Latin America), evidence shows that various constraints explain some portion of these gaps, ranging from the lack of childcare and adequate leave policies to social norms that create barriers preventing women from working. These norms include those that put a dis- proportionate responsibility for domestic work and childcare on women, as well as those that result in occupational sex segregation, sexual harassment, and mobility restrictions. Women must also contend with differential constraints in access to finance and markets, a great divide in access to digital technology, and legal/regulatory barriers to start and grow firms.b All these factors result in wasted potential in terms of realizing economic gains from human capital investments in girls. Looking only at the sex-disaggregated HCI misses an important reality concerning gender gaps in how human capital is utilized. For human capital to translate into productivity, humans, who own the capital, need to be employed in work where they can use their human capital. For example, in 2020, boys and girls growing up in Peru have the same HCI score of 0.6. Only 62 percent of women in Peru are employed, however, compared to 78 percent of men, resulting in a basic Utilization-Adjusted Human Capital Index that is 10 percentage points lower for females than for males (Pennings 2020). Economies can act to enable women’s full participation in labor market opportunities. Provision of affordable childcare options, parental leave policies, and flexible work options can accom- modate women’s entry into formal work and help women and men redistribute and balance demands at home and at work (Olivetti and Petrongolo 2017). Safe transport allows women to go to the workplace, and pay transparency can increase women’s power to negotiate equal pay for equal work. Improved access to digital technology for women can unlock potential gains from the digital era. These range from accessing online education to expanded income-generating opportunities through flexible online gig work and e-commerce entrepreneurship (Alatas et al. 2019; Dammert, Galdo, and Galdo 2014; World Bank 2016). Resources need to be mobilized to ensure that women and men have equal access to livelihoods and economic opportunities. Source: Prepared by Daniel Halim. a World Bank Enterprise Surveys data retrieved from World Bank Gender Data Portal, https://datatopics.worldbank​ .org/gender/. b In low- and middle-income countries, only 54 percent of women have access to mobile internet, compared with 74 percent of men (GSMA 2020). employment rates driven by a variety of fac- small gaps in how much human capital there is tors, including social norms. The full UHCI has underutilize.  See box  4.3 for a  discussion of to ­ smaller gender gaps, however, in part because of closing gender gaps. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 125 NOTES 9. For more information on JOIN, see https:// datacatalog.worldbank.org/dataset/global 1. This chapter was prepared by Steven Pennings -jobs-indicators-database. (spennings@worldbank.org) with helpful com- 10. In some cases, the more recent data source of ments from Roberta Gatti, Aart Kraay, Michael employment data is not used if it is missing Weber, Kathleen Beegle, Paul Corral, and data for the full UHCI. David Weil, as well as from other internal and 11. For lower-income countries, the U-shape external reviewers. See Pennings (2020) for an is mostly driven by several outliers with in-depth treatment. A data appendix is available extremely high utilization rates. for download at https://development-data-hub​ 12. In Malawi, a household survey in 2005 -s3-public.s3.amazonaws.com/ddhfiles/144347​ reported employment rates of 0.8, twice the /uhci_dataappendix_sep2020.xlsx. most recent figure (from 2017). Likewise, in 2. See box 4.1 for a derivation. More specifically, Nepal a household survey in 2008 reported this relationship requires a Cobb-Douglas pro- employment rates of 0.84, also about twice the duction function and the assumption that the most recent figure (also from 2017). The differ- capital-to-output ratio is constant in the long ence is likely due to exclusion of own-use pro- run (one of Kaldor’s facts). duction workers in 2017, though this exclusion 3. In the full UHCI, the utilization rate is defined is not well documented. as UHCI/HCI—and so satisfies equation (4.1) 13. For the average low-income economy, long- by construction—but still turns out to have an run incomes in the complete utilization and intuitive interpretation (see section 4.3 titled human capital scenario would be about 4.5 “The Full UHCI” later in the chapter). times those of the status quo (1/0.22 ≈ 4.5), 4. Just like the HCI, the UHCI can also be inter- compared with 2.5 times that with complete preted in terms of productivity: a child born human capital alone (1/0.4 ≈ 2.5). today can expect to be only UHCI × 100 per- 14. On average, those with a complete second- cent as productive as she would be, on average, ary education were earning only 60 percent if she enjoyed complete education and full more than those with no education, which is health, and if her future labor was fully utilized. the equivalent of less than six years with an 5. Naturally, no economies will have employment 8  ­ percent return to education. Omitted vari- rates of 1. But this condition is consistent with the ables such as parental income and ability mean approach in the HCI, in which no economy has that the six years is likely an overstatement. perfect test scores or 14 expected years of school. 15. Another cause of the mismatch can be poor 6. In the full UHCI, discussed later in the chapter, education quality, with the result that those with economies with very low better employment a qualification cannot perform the functions rates will have GDP that is less sensitive to required. In this case, the reason the engineer increases in human capital. But, even in those is driving a taxi is that he or she is not able to economies, improvements in human capital perform the tasks of an engineer because of will still increase growth. poor-quality education. Handel, Valerio, and 7. Technically, this is because the implicit assump- Sanchez Puerta (2016) find that, in 12 low- and tion in the HCI is that basic utilization rates are middle-income countries, the overeducation/ constant across status quo and full human cap- underutilization rate is 36 percent. Overeduca- ital scenarios. A full employment assumption is tion/underutilization rates vary across countries not required. and can depend on how the rates are measured. 8. Data from ILOSTAT, the ILO’s labor statistics 16. The definition of formal employment varies database (accessed December 13, 2019), https:// across countries, but it generally refers to the www.ilo.org/shinyapps/bulkexplorer7/. coverage of the worker with respect to benefits 126 U tilizing H u man Cap ital like unemployment insurance, pensions, sick 22. The one exception is Vietnam, which has a high leave, or annual leave. employment rate, but a low fraction of that is 17. Better employment differs from “decent work” in better jobs. These differences remain prom- (ILO) and “good jobs for development” (World inent in the full UHCI because of Vietnam’s Bank 2012). high HCI score. 18. A final technical issue is that some of the 23. Klasen (2019) shows that the U-shaped pattern increase in GDP in the basic UHCI comes of female employment rates is mostly due to from utilizing people’s time rather than uti- region fixed effects, and not to the develop- lizing their human capital. The full UHCI also ment path for an individual country. addresses this concern (see Pennings 2020). 19. It is important to acknowledge that the defini- tion of better employment and the full UHCI REFERENCES are stylized for simplicity and cross-country data availability. In reality, many people without bet- Alatas, V., A. Banerjee, R. Hanna, B. A. Olken, ter jobs can partially use their human capital to R.  Purnamasari, and M. Wai-Poi. 2019. “Does increase productivity beyond that of raw labor. Elite Capture Matter? Local Elites and Targeted For example, education is positively correlated Welfare Programs in Indonesia.” AEA Papers with high-yield-variety seed choice among and Proceedings 109: 334–39. Indian farmers (Foster and Rosenzweig 2010). Battu, H., and K. Bender. 2020. “Educational Moreover, healthier people may be more pro- Mismatch in Developing Countries: A Review ductive laborers. Assuming that only educational of the Existing Evidence.” In The Economics  of human capital (not health human capital) is Education: A Comprehensive Overview  (2nd  Ed), underutilized outside better employment results edited by S. Bradley and C. Green. Academic ­ apita gross in the same U-shaped pattern in per c Press. national income, but with higher utilization Chioda L. 2016. Work and Family: Latin American rates (and UHCI) for low-income economies and Caribbean Women in Search of a New (not reported). Moreover, there are many exam- Balance. Latin American Development Forum. ples of people without better jobs using their Washington, DC: World Bank. human capital to its full extent, such as self-em- Dammert, A. C., J. C. Galdo, and V. Galdo. 2014. ployed professionals; however, the availability “Preventing Dengue through Mobile Phones: of cross-country employment data limits the Evidence from a Field Experiment in Peru.” amount of nuance possible in this regard. Journal of Health Economics 35: 147–61. 20. Data from ILOSTAT (accessed February 20, Filmer, D., and L. Fox. 2014. Youth Employment in Sub- 2020) (defined using ICSE-93). Saharan Africa. Washington, DC: World Bank. 21. This low correlation is driven by a number of Foster, A., and M. Rosenzweig. 2010. “The economies on the left side of figure 4.8 in the Microeconomics of Technology Adoption.” Middle East and North Africa and elsewhere, Annual Review of Economics 2: 395–424. where a high fraction of total employment is Gatti, R., M. Morgandi, R. Grun, S. Brodmann, classified as better employment (such as wage D.  Angel-Urdinola, J. M. Moreno, D. Marotta, employment), and a number of economies, M. Schiffbauer, and E. Mata Lorenzo. 2013. Jobs often in East Asia and Pacific, with lower rates for Shared Prosperity: Time for Action in the Middle of wage employment on the right side of figure East and North Africa. Washington, DC: World 4.8. Some of these East Asia and Pacific econo- Bank. mies are also penalized in the full measure by Goldin, C. 1995. “The U-Shaped Female Labor having a high HCI that increases the potential Force Function in Economic Development to underutilize human capital. and  Economic History.” In Investment T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 127 in Women’s Human Capital and Economic Kraay, A. 2018. “Methodology for a World Bank Development, edited by T. P. Schultz, 61–90. Human Capital Index.” Policy Research Working University of Chicago Press. Paper 8593, World Bank, Washington, DC. GSMA. 2020. “The Mobile Gender Gap Report Merotto, D., M. Weber, and A. Reyes. 2018. 2020.” GSMA Connected Women, London. “Pathways to Better Jobs in IDA Countries: https://www.gsma.com/mobilefordevelopment​ Findings from Jobs Diagnostics.” Jobs Series /wp-content/uploads/2020/05/GSMA-The​ No. 14, World Bank, Washington, DC. -2020.pdf. -Mobile-Gender-Gap-Report​ Olivetti, C., and B. Petrongolo. 2017. “The Handel, M., A. Valerio, and M. Sanchez Puerta. Economic Consequences of Family Policies: 2016. “Accounting for Mismatch in Low- and Lessons from a Century of Legislation in Middle-Income Countries: Measurement, High-Income Countries.” Journal of Economic Magnitudes, and Explanations.” Directions Perspectives 31 (1): 205–30. in  Development Series. Washington, DC: Pennings, S. 2020. “The Utilization-Adjusted Human World Bank. Capital Index (UHCI).” Policy Research Working International Labour Organization. 2018. “Global Paper 9375, World Bank, Washington, DC. Wage Report 2018/19: What Lies behind World Bank. 2012. “What Matters Most for School Gender Pay Gaps.” Geneva: International Health and School Feeding: A Framework Labour Office. Paper.” SABER Working Paper 3, Systems Klasen, S. 2019. “What Explains Uneven Female Approach for Better Education Results, World Labor Force Participation Levels and Trends Bank, Washington, DC. in Developing Countries?” World Bank Research World Bank. 2016. World Development Report 2016: Observer 34 (2). Digital Dividends. Washington, DC: World Bank. 5 Informing Policies to Protect and Build HUMAN CAPITAL T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 129 T he Human Capital Index (HCI) 2020 Yet fiscal constraints and numerous competing update arrives at a time when policy mak- priorities raise the risk of delays to investments in ers across the world face urgent choices. measurement when, in fact, measurement enables Strategic decisions made now have the power to effective action. protect and strengthen countries’ human capital and, with it, their economic future. Better measurement and transparent informa- tion can be transformational in safeguarding In addition to documenting pre-COVID-19 and strengthening human capital. By generating (coronavirus) changes in human capital across a shared understanding among diverse actors, 174 economies, the HCI 2020 update estab- measurement can shine a light on constraints that lishes a baseline for tracking the pandemic’s limit progress in human capital. Through this pro- effects on human capital. A further task for the cess, effective measurement can facilitate political update has been identifying pathways through consensus based on facts and mobilize support for which COVID-19 can influence human capital reforms. Measurement also enables policy makers outcomes in the short and longer terms. Using to target support to those who are most in need, the HCI methodology to quantify the gaps that which is often where interventions yield the high- will likely emerge in health, skills, and knowl- est payoffs. As policy implementation moves for- edge because of COVID-19, this analysis under- ward, measurement provides feedback to guide scores the urgency of protecting and sustaining course corrections. the recovery of human capital, which will be a cornerstone of countries’ postcrisis recovery and If measurement can improve policy results around future economic growth. human capital in ordinary times, its importance is multiplied during a crisis. Governments that can Good measurement and data are essential to access and use relevant data in real time are better shape well-targeted and cost-effective policies able to act in a coordinated way on multiple fronts. and to design course corrections when needed. In the case of COVID-19, they can monitor the evo- To underscore this point, this report’s final pages lution of disease transmission and continuously map short-term and longer-range agendas for update control strategies while responding to the strengthening the measurement of human capital, immediate and long-term effects of the economic and link these agendas to policy changes necessary crisis on households and communities. Measuring to protect human capital in the wake of COVID-19. how well children are growing, whether they are learning, and how financial stress and insecurity are affecting their development is a necessity, not 5.1 GOOD MEASUREMENT: NECESSITY, a luxury. It is essential to design and target pol- NOT LUXURY icies that can remediate the pandemic’s negative impacts. At a time when demand for government As the COVID-19 crisis continues to unfold, good spending is surging, and fiscal space is limited, data data and measurement are more vital than ever. and their transparent communication are vital to 130 Inform i ng P ol i c i e s to P rot e c t an d B u il d Hum a n Ca pi ta l ensure accountability for how scarce resources are 5.2.1  A short-term measurement agenda used. Because of the dramatic changes in household incomes and service delivery driven by COVID-19, The power of measurement to support transfor- there is an immediate need to measure the pan- mative action in difficult situations extends beyond demic’s welfare impacts. Social distancing, how- public health emergencies. For example, data are ever, is limiting the way in which traditional especially important in countries affected by fra- surveys are collected by enumerators who visit gility or conflict, though measurement is far more families. Phone surveys have helped respond to difficult to carry out in these settings. Insecurity this challenge by reaching households remotely.1 and the lack of robust institutions hinder data col- lection and, in turn, the ability of governments to Phone surveys are relatively inexpensive, an import- take action informed by evidence. Fortunately, ant consideration at a moment when resources are innovative methods have recently enabled some especially scarce and countries face many compet- progress in understanding human capital dynam- ing priorities. Such surveys are well suited for gath- ics in fragile contexts (box 5.1). ering information about behaviors (including access to health services and uptake of remote learning arrangements) or outcomes (such as income and 5.2  BEYOND THE HCI consumption) subject to rapid change. They are likely to return more reliable and informative data The HCI offers a bird’s-eye view of human capital when they build on existing information bases, across economies. By benchmarking the produc- pointing to the importance of triangulating with tivity costs of shortfalls in health and education, existing data collection initiatives. the index has spurred new conversations within governments, bringing discussion on human cap- Facility phone surveys are a complement to house- ital accumulation to the level where decisions hold phone surveys. They can document, for exam- about resource allocation are actually made. This ple, how prepared health facilities are to manage is an important achievement. COVID-19 patients and can identify bottlenecks in the delivery of routine health services, includ- As a measurement tool, however, the HCI has ing immunization and maternal and child health substantial limitations. For example, it does not services. Administrative data can also be used to speak to distributional or geographical differences monitor many aspects of service provision—at a within economies. And although it focuses on what low marginal cost because these data already exist matters—outcomes—it does not chart the specific in most countries. These data could provide valu- pathways that each economy needs to follow to able insights but are often poorly linked, of vary- accelerate progress in human capital. Much greater ing quality, and inaccessible to groups outside of depth in measurement and research is needed to government. Big data can be similarly leveraged to better understand the dynamics of human capi- guide action. For example, data from mobile phone tal accumulation, including across socioeconomic records have been used to monitor mobility (which groups and geography, and how policies can affect is important for modulating disease containment), it. Some key measurement improvements can to nudge behavior, and to improve service delivery, be achieved in the short term (for example, on including delivering educational content. Digital test scores, see box 5.2). Longer-term efforts will technology and data can be harnessed to provide demand a more sustained commitment from social protection benefits more equitably and effi- economies and development partners. ciently, both immediately and in the longer run. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 131 Box 5.1: Innovative data collection in fragile contexts: Examples from West Africa and the Middle East and North Africa Epidemics affect people’s health, and they also disrupt livelihoods and well-being through school closures, workers placed on furlough, restrictions on transportation and gather- ings, and closing of international borders. As such, at the height of the Ebola epidemic in West Africa, in addition to assessing the impact of the disease on people’s health, it was also important to measure and monitor the epidemic’s socioeconomic impact. Given the nature of the epidemic, however, it was impossible and unethical to deploy enumerators to the field for data collection in face-to-face interviews at households and in communities. In 2014, capitalizing on the proliferation of mobile phone networks, and building on the experiences of the mobile phone survey initiative called Listening to Africa, high-frequency mobile phone interventions were designed and implemented to provide rapid monitoring of the socioeconomic impacts of the Ebola crisis in Liberia and Sierra Leone. Two nationally representative surveys, each conducted in Liberia and Sierra Leone when the crisis broke out, were used as the baseline for anchoring estimates in a representative dataset. In Liberia, researchers drew on the country’s Household Income and Expenditure Survey, which had to curtail fieldwork in August 2014. In Sierra Leone, they used the Labor Force Survey, which had completed fieldwork in July 2014. These existing surveys provided a database of phone numbers and household characteristics, which eventually became the sample frame for the phone survey. Data were then collected through call centers, either nationally or internationally, to reach over two thousand respondents in each country. Although phone surveys cannot replace face-to-face household surveys in all contexts, the experience in Liberia and Sierra Leone illustrates substantial benefits of such innovation in specific circumstances and for specific data collection needs, particularly the ability to col- lect timely data in volatile and high-risk environments. Implementing surveys in a rapidly evolving context involves myriad challenges, includ- ing the lack of a relevant and reliable sample frame. For example, excluding displaced populations from national sample frames threatens the representativeness of socioeco- nomic surveys and consequently provides a skewed understanding of the country. As the size of forcibly displaced populations increases globally, it is urgent to devise strategies to include these populations in nationally representative surveys. The sampling proce- dure undertaken for the Syrian Refugee and Host Community Surveys, implemented over 2015–16 in the Kurdistan region of Iraq, in Jordan, and in Lebanon, offers valuable insights on overcoming survey-implementation challenges to obtain representative estimates in challenging contexts. In the absence of updated national sample frames for host communi- ties, and given the lack of comprehensive mapping of forcibly displaced populations, geo- spatial segmenting was used to create enumeration areas where they did not exist. Data collected by humanitarian agencies, including the United Nations High Commissioner for Refugees and the International Organization for Migration, were used to generate sample frames for displaced populations. Source: Based on Hoogeveen and Pape 2020. 132 Inform i ng P ol i c i e s to P rot e c t an d B u il d Hum a n Ca pi ta l Box 5.2: Leveraging national assessments to obtain internationally comparable estimates of education quality The Human Capital Index (HCI) highlights the need for regular and globally comparable measurement of learning to assess the quality of an economy’s education system. Although most data on education quality included in the HCI currently come from assessments designed to be comparable across economies and over time using psychometric methods, those assessments are often infrequent and do not yet cover all economies. Leveraging national learning assessments can help bridge the gaps in learning data. Most economies regularly conduct some form of assessment that can be augmented with short modules of globally benchmarked and validated items to construct globally comparable measures of education quality (Birdsall, Bruns, and Madan 2016; UNESCO 2018). Despite the lack of a comprehensive bank of globally benchmarked items, some items from international assessments can be incorporated into national assessments as linking items. These linking items provide commonality with international assessments, enabling learning outcomes to be placed on a global scale (Kolen and Brennan 2004). For instance, the 2021 National Assessment for Secondary Schools will enable Bangladesh to produce globally comparable learning outcomes. To allow comparison of national education quality on a global scale, the following countries have recently fielded or are planning to include linking items from inter- national assessments in their national assessments. Nigeria. Besides conducting an Early Grade Reading Assessment in 2014 for 4 of its 37 states, Nigeria had only sparse learning data until recently. The HCI 2018 emphasized the need for a nationally representative and internationally comparable assessment of learning outcomes in Nigeria. The Nigerian National Learning Assessment (NLA 2019), supported by the World Bank, is the first nationally representative learning assessment conducted in Nigeria using an internationally recognized methodology. The NLA 2019 measures student learning at grades 4 and 8 in the core subjects of mathematics, English, and science, and includes linking items to allow comparison on an international scale. Once fully harmonized with international assessments, the NLA 2019 will allow for inclusion in a future HCI of a nationally representative and globally comparable learning measure for Nigeria. Sri Lanka. In 2009, the national assessment included linked Trends in International Mathematics and Science Study (TIMSS) items. Subsequent national assessments in Sri Lanka have maintained linking items with TIMSS to allow international comparability. The resulting score is used in the World Bank’s HCI. Uzbekistan. Before 2019, no internationally comparable learning outcomes data were available for Uzbekistan. The launch of the 2018 HCI galvanized the government toward measurement of education quality; in 2019, with World Bank support, the country con- ducted its first nationally representative and internationally comparable assessment (using TIMSS linking items) for grade 5 students in mathematics. That assessment is now part of the country’s 2020 HCI. Relatively few linking items are currently available from international assessments, neces- sitating a cautious approach informed by individual economy contexts: ensuring that the selected linking items align with the economy’s national grade-level curriculum, are translated according to the protocols of the international assessment, are piloted in the economy, and are not too easy or too difficult for the target population; that similar testing conditions are arranged as for international assessments; and that a sufficient number of items is selected to provide reliable internationally comparable estimates of education quality in the economy. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 133 5.2.2 Tackling long-term measurement time, for refugees, displaced persons, and host needs populations, are extremely limited. In addition to solutions that can be deployed rap- idly, economies need strategies to improve the An upcoming World Development Report will focus measurement of human capital in the longer run. on data and provide a comprehensive description of the complex and rapidly evolving measurement The HCI update offered an opportunity for econ- landscape. But it is worth noting here the areas in omies to take stock of their data on human cap- which better-funded and coordinated data collec- ital outcomes. In this process, the World Bank tion and use could improve the understanding of engaged with counterparts in several fruitful human capital accumulation and effective inter- collaborations to improve the quality of data ventions to accelerate it. used to calculate the HCI. For example, thanks to close collaboration with the Ministry of Human One such area concerns the long-term conse- Resource Development in India, it was possible to quences of interventions that have proved suc- significantly improve upon publicly available data cessful in the short run. For example, there is for school enrollment and arrive at a measure of well-established evidence that conditional cash expected years of school (EYS) constructed on the transfers have improved a variety of health and basis of actual age-specific enrollment rates, which education outcomes within a few years of program capture enrollment more precisely.2 inception. Relatively little evidence exists, how- ever, on whether and how the increased time spent Despite improvements, many gaps in the mea- in school thanks to the transfers has led to better surement of internationally comparable key learning outcomes and improved labor market dimensions of human capital persist (box 5.2). opportunities. Similarly, long-term evidence on For example, establishing well-functioning vital the efficacy of some types of interventions often registry systems to record such basic events relies on small pilots that were not followed by as births and deaths is still a work in progress: countrywide scale-up, and questions therefore less than 70 percent of economies record such remain about the generalizability of promising events, and progress to fill these gaps has been findings. slow (box 5.3). The quality of school enroll- ment data, which in the index are based on Administrative data can allow for the tracking of administrative records, is highly variable in critical outcomes over time by, for example, link- low- and middle-income economies, particu- ing educational assessments and hospital records larly at the lower- and upper-secondary levels. to taxation records, social security contributions, Finally, benchmarking learning outcomes inter- or health insurance via unique identifiers. The nationally has been a challenge, both across benefits of user-friendly administrative data sys- economies and especially over time. This diffi- tems are vast, because such systems can inform significantly constrained the coverage culty has ­ policy choices about the design of cost-effective of the long-run analysis of changes in measured interventions, allow regular monitoring of key human capital. These challenges are height- outcomes, and support decision-making in real ened in fragile situations: in some cases, data to time, all at low marginal cost. Within an econ- inform various HCI components simply do not omy, however, administrative data are collected exist; in others, data are too old and likely do not by a variety of ministries and other entities, often sufficiently capture the rapid deterioration of resulting in a patchwork of systems that does not human capital that can occur in fragile contexts. favor integration and optimal use. Taking advan- In addition, comparable data, including over tage of these data requires expertise that is scarce 134 Inform i ng P ol i c i e s to P rot e c t an d B u il d Hum a n Ca pi ta l Box 5.3: Data quality and freshness in the components of the Human Capital Index The Human Capital Index (HCI) has proved a useful tool for policy dialogue, largely because it incorporates human capital outcomes that are easily recognizable, consistently measured across the world, and salient to policy makers. However, even the basic index components suffer from significant data gaps and quality issues. The child and adult survival measures used in the index are based on data on birth and death rates by age group. These data come primarily from national vital registries that are mandated to record vital events like births and deaths. As such, vital statistics are essential to the measurement of demographic indicators like life expectancy and to identifying health priorities for the population. They can also help target health interventions and monitor their progress. Coverage of vital registries varies widely: only 68 percent of economies register at least 90 percent of births (see map B5.3.1), and only 55 percent cover at least 90 percent of deaths.a Birth registration has increased by only 7 percentage points (from 58 percent to 65 percent) in the past decade (UNICEF 2013); in Sub-Saharan Africa, only eight countries have birth registration coverage above 80 percent.b Stunting serves as an indicator for prenatal, infant, and early childhood health environments. The Joint Malnutrition Estimates (JME) database that compiles global stunting data reports data for 152 economies, 33 of which have data that are more than 5 years old.c In 10 econo- mies, the most recent survey is over 10 years old. Gaps also remain in education data. The expected years of school measure is based on enrollment data that national governments provide to the United Nations Educational, Scientific and Cultural Organization Institute for Statistics. Of the 174 economies in the HCI 2020 sample, 22 economies rely on primary enrollment data from 2015 or earlier. Because primary enrollment data are typically the most consistently reported, the issue of data fresh- ness is of even greater concern for other levels of school. Significant gaps also exist in time series data on enrollment rates. Of the 103 economies included in the 2010 HCI sample, 22 economies were missing primary enrollment rates for 2010. Data gaps are more numer- ous at other levels of schooling—over 30 economies were missing secondary-level enroll- ment data for 2010, and 42 economies were missing these data at the preprimary level. Finally, the latest update to the Global Dataset on Education Quality that produces harmo- nized test scores covers 98.7 percent of the school-age population. Of the 174 economies with an HCI, 14 rely for test score data on Early Grade Reading Assessments that are not representative at the national level. Sixty-five economies (roughly 37 percent of the sample) rely on test score data from 2015 or earlier. Significant gaps exist in sex-disaggregated data across HCI components. The JME reports disaggregated stunting data for only 56 percent of the 887 surveys that are part of the database. Whereas sex-disaggregated enrollment rates are reasonably complete at the pri- mary level, they are missing at the lower-secondary level for 29 of the 174 economies in the HCI 2020 sample. Sixteen economies in this sample are also missing disaggregated test score data. As a result of these gaps, 21 of the 174 economies in the 2020 sample do not have sex-disaggregated HCI scores. These gaps in disaggregated data span all regions and income groups.d (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 135 Box 5.3: Data quality and freshness in the components of the Human Capital Index (Continued) The credible and consistent measurement of human capital outcomes is essential to identi- fying priority areas for policy intervention, informing the design of those policies, and track- ing their effectiveness over time. Collection of high-quality data can doubtless be a costly undertaking, but countries can explore more cost-effective ways of monitoring their citi- zens’ health and education outcomes. For instance, instead of bearing the costs of partic- ipating in an international assessment, Uzbekistan incorporated assessment items into its national learning assessment that would allow for linking with the Trends in International Mathematics and Science Study (see box 5.2). Map B5.3.1: Coverage of live births registration 100% 90% 75% 50% 0% No data IBRD 45472 | JANUARY 2021 Source: United Nations Statistics Division. Note: Boundaries and names shown and the designations used in this map do not imply official endorsement or acceptance by the United Nations or the World Bank. a See the United Nations Statistics Division web page, “Coverage of Birth and Death Registration,” https://unstats​ .un.org/unsd/demographic-social/crvs/#coverage. For countries without robust vital registries, the United Nations Interagency Group for Child Mortality Estimation and the United Nations Population Division (that reports on adult mortality) fill data gaps using population censuses, household surveys, and sample registrations combined with model life tables. All these data must then be modeled to produce mortality rates. b “Coverage of Birth and Death Registration” web page. c For more information on the JME database, see https://data.unicef.org/resources/jme-report-2020/. d Of the 21 economies missing a sex-disaggregated HCI score, 13 are from Sub-Saharan Africa; 2 each from East Asia and the Pacific, Latin America and the Caribbean, and South Asia; and 1 each from the Middle East and North Africa and Europe and Central Asia. Of those 21 economies, 3 are high-income, 4 are upper-middle-income, 7 are lower-middle-income, and 6 are low-income. 136 Inform i ng P ol i c i e s to P rot e c t an d B u il d Hum a n Ca pi ta l in many economies. Finally, legitimate privacy behaviors and competencies. Surveys such as concerns also restrict access and can make data those conducted by the Service Delivery Indicators linking incomplete or impossible. (SDI) initiative can help. SDI surveys are nationally representative facility surveys that measure the A related concern involves understanding the pro- quality of services received by average citizens in duction function of health and education outcomes primary health care centers and primary schools.3 from the service delivery perspective. This issue is SDI surveys collect data on critical inputs and essential for designing effective interventions and provider performance and, in the case of schools, systems for quality health care and education. It children’s learning. These types of data allow gov- is even more pressing in a post-COVID-19 world, ernments and service providers to identify gaps in where extensive remediation will be needed to service provision, link financing inputs with health compensate for the losses to human capital caused and education outcomes, and understand the by the shock. Many basic questions remain unan- margins along which social sector spending fails swered. Do students have textbooks? Are health to translate into quality services. SDI surveys are centers stocked with the necessary drugs? important platforms for innovation and research, including measuring the quality of management Beyond assessing fundamental inputs, countries in schools and hospitals. need answers on how to improve the quality of services. They need to understand, for example, The analysis of delivery systems needs to whether teachers actually master the curriculum advance in parallel with a deeper understand- they are teaching and if physicians diagnose dis- ing of how human capital accumulates through eases accurately and treat them appropriately. the life course. For example, evidence points to Selection mechanisms and incentives also mat- the nodal importance of early childhood years ter for the quality of services. For example, pay for lifelong cognitive, physical, and socioemo- for performance has been widely introduced and development. Very few economies, how- tional ­ requires evaluation. How can it best be managed ever, make systematic measurement of skills in and at what level? Private sector financing and early years a p ­ riority. Even when those measures delivery also have the potential to improve ser- are available, the evolution of health status, cog- vice quality. But how can countries make sure that nitive abilities, and noncognitive skills during quality improves while services remain affordable? early childhood is not well understood. Similarly, Rapid advances in information and communica- measuring skills—cognitive and noncognitive— tions technology likewise hold promise to improve among adolescents and adults is still rare in most service delivery. Reliable strategies to make such economies. improvements are not obvious, however, and will differ across country contexts. Additionally, qual- Advancing this long-term measurement agenda ity reflects management capacities and choices. will require purposeful investments. In turn, What management interventions improve ser- funding measurement is a way to increase the effi- vice delivery in cost-effective ways? And how can ciency and impact of future policy action across countries measure the quality of management in multiple domains. By supporting the political the social sectors? economy of reform processes and guiding pol- icy choices toward cost-effective solutions, better Administrative data can answer some of these measurement and data use are investments that questions but cannot provide insights into pay off. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 137 5.3 BUILDING, PROTECTING, AND care may focus on reproductive and child health EMPLOYING HUMAN CAPITAL IN and nutrition; infectious disease control programs A POST-COVID-19 WORLD for HIV, tuberculosis, and malaria; and commu- nity-based health promotion and disease preven- Governments are now working under intense pres- tion. In middle- and higher-income countries, a sure to roll out policies across multiple sectors in focus on improving healthy longevity, addressing response to COVID-19. Measurement is essential to noncommunicable diseases, and linking primary ensure that these policies are strategically designed care practitioners systematically to disease surveil- and well implemented, and that they get results. lance networks will go a long way toward increas- What might effective policy solutions look like in ing resilience. In the face of widening health dis- the domains most important for human capital? A parities, it is essential to ensure that disadvantaged companion paper to the HCI 2020 update (World households and communities have access to qual- Bank 2020b) discusses policy responses to COVID- ity and affordable care. In the past, disruptions to 19 in detail; what follows are some of the broad the health and economic status quo have some- directions these responses may adopt. times enabled countries to introduce bold health system reforms (see McDonnell, Urrutia, and 5.3.1  A data-driven health sector response Samman 2019). In that sense, these difficult times The immediate priority for countries fighting may offer an unexpected opportunity in many COVID-19 remains containment and elimination countries to renew the commitment to universal of the novel coronavirus. Global efforts, such as health coverage (World Bank 2020c). improvements in testing and access to a safe and effective vaccine, will need to accompany local 5.3.2  Preventing losses in learning measures to test, trace, and isolate carriers of Along with more and better investments in health, infection; to support the use of nonpharmaceuti- economies need a broad range of interventions in cal interventions such as masks and social distanc- other sectors to get human capital accumulation ing; and to implement targeted lockdowns when back on track, both in the short and longer terms necessary. Strengthening public health surveil- (World Bank 2020a). Because of school closures and lance capacity will be essential to the timeliness economic hardships, the current generation of stu- and effectiveness of these interventions. Robust dents stands to lose significantly in terms of learning surveillance requires the ability to collect, analyze, and noncognitive skills now, and in terms of earn- and interpret relevant health-related data and use ings later in life. Strategies to remediate schooling these data to plan, implement, and evaluate con- losses will require designing and implementing trol actions. With most pandemics being of zoo- school reopening protocols adapted to the specific- notic origin, closer coordination between health ities of the pandemic. At a minimum, protocols will and the agriculture sector will be instrumental to involve protective equipment and supplies, health prevent future outbreaks, in keeping with a “one screening, and social distancing. Tailored teaching health” approach. and learning resources, especially for disadvantaged children, are urgently needed in many settings to Complementing strong surveillance, it is essen- make up for lost learning (World Bank 2020b). tial to step up health care for COVID-19 patients while maintaining the delivery of core health ser- Deeper reforms will need to follow so that coun- vices. COVID-19 highlights the need to invest in tries can sustain access to schooling and promote primary health care with strong frontline deliv- children’s learning at all stages: from cognitive ery systems. In low- and middle-income coun- stimulation in the early years to nurturing relevant tries, priority measures to strengthen primary skills in childhood and adolescence. Building blocks 138 Inform i ng P ol i c i e s to P rot e c t an d B u il d Hum a n Ca pi ta l for success will include better-prepared teachers, shocks, exactly at a time when they are being hit better-managed schools, and incentives that are the hardest. Risks of gender-based violence can aligned across the many stakeholders in education also be heightened during times of crisis, isola- reform. The efforts that economies have made in tion, and confinement (World Bank 2020d). These providing continuity with remote learning during effects are amplified in fragile settings. the pandemic could carry benefits beyond the cur- rent emergency. Appropriately structured online Deepening inequalities make targeting interven- learning can facilitate the acquisition of those com- tions to the most disadvantaged—and particularly petencies, such as collaboration and higher-order to children in their early years—an imperative, cognitive skills, that are increasingly essential in the to prevent setbacks that are likely to compromise changing world of work (Reimers and Schleicher lifetime health, education, and socioeconomic tra- 2020). To shape resilient education systems, econo- jectories among the most vulnerable. These inter- mies will need to draw lessons from this worldwide ventions should have an explicit gender angle to distance learning experience and expand the infra- help progressively close the gaps now being mag- structure for online and remote learning. nified by COVID-19. 5.3.3 Reinforcing resilience among 5.3.4 Coordinating action across sectors and vulnerable people and communities adopting a whole-of-society approach In the face of sharp declines in income, support COVID-19 has underscored the interdependence to poor and vulnerable households is essential to of multiple sectors that are fundamental for human mitigate COVID-19’s impact and to sustain access capital accumulation. These sectors include health, to services and food security. In the first phase of education, infrastructure, water and sanitation, the pandemic response, the consensus on social information technology, and ­ others. Complex assistance programs has been to cast the net wide, links connect these domains. For example, proper to avoid excluding any of those in need. In the hygiene contributes to limiting diffusion of the virus. medium term, these interventions need to be reas- In turn, reduced transmission is often a precondi- sessed and complemented or replaced by policy tion to reopening of schools. Digital technologies measures geared toward an inclusive and sustain- enable educational continuity when schools cannot able economic recovery with support for employ- physically reopen, but many poor and marginal- ment and livelihoods (including with active labor ized communities lack access to digital tools. These market policies that help match workers to new links point to the need for ambitious investments jobs and upgrade their skills), as well as assistance in many economies to expand access to water, san- to small and microenterprises (World Bank 2020d). itation, and digitalization as key enablers of human In parallel, strengthening social services, including capital accumulation. counseling, will help mitigate impacts on mental health and disruptions in people’s social networks. Connections across sectoral and social boundar- ies emphasize the value of policy approaches that COVID-19 has exacerbated many forms of engage diverse stakeholders. Nurturing a nation’s inequality, notably gender gaps. School closures human capital is everybody’s business. If a child and a reduction in health services can interrupt the accumulates strong human capital during her crit- trajectories of adolescent girls at a critical life junc- ical years of growth and development, it is because ture. With women-owned firms primarily concen- a large network of people and institutions has con- trated in informal or low-paying sectors, the lack tributed to the process. Parents decide what to feed of basic formal social protection deprives women a child, when to take her to the doctor, and whether and their families of buffers against economic and for how long to send her to school. Families T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 139 make these choices within communities that trans- REFERENCES mit norms and that may provide a safety net to households in need. In turn, communities rely on Birdsall, N., B. Bruns, and J. Madan. 2016. services that, in many contexts, are provided largely “Learning Data for Better Policy: A Global by the private sector, including nongovernmental Agenda.” CGD Policy Paper 096, Center for organizations. Finally, governments provide pub- Global Development, Washington, DC. lic goods, address externalities, and ensure equity. Hoogeveen, J., and U. Pape, eds. 2020. Data Wise public policy choices, informed by measure- Collection in Fragile States: Innovations ment, facilitate the shared achievement of human from Africa and Beyond. Washington, DC: capital and make it more than the sum of its parts. World Bank. Kolen, M. J., and R. L. Brennan. 2004. Test Equating, The COVID-19 crisis has put all the links in this net- Scaling, and Linking: Methods and Practices. work under strain, not least the governments them- Second Edition. Springer-Verlag. selves. Under these conditions, progress depends on McDonnell, A., A. Urrutia, and E. Samman. 2019. leadership that recognizes the importance of build- “Reaching Universal Health Coverage: A ing a future in which all children can reach their Political Economy Review of Trends across potential. In the months and years ahead, with lim- 49 Countries.” ODI Working Paper No. 570, ited fiscal space, protecting core spending for human Overseas Development Institute, London. capital will challenge policy makers in many econ- Reimers, F., and A. Schleicher. 2020. “Schooling omies, regardless of their levels of income. Yet, by Disrupted, Schooling Rethought: How the making these investments with a view to the future, COVID-19 Pandemic Is Changing Education.” economies can emerge from the COVID-19 crisis OECD, Paris. prepared to do more than restore the human cap- UNESCO. 2018. “TCG4 Measurement Strategies.” ital that has been lost. Ambitious policies informed TCG4/23, United Nations Educational, by rigorous measurement can take human capital Scientific and Cultural Organization Institute beyond the levels previously achieved, opening the for Statistics, Dubai. way to a more prosperous and inclusive future. UNICEF (United Nations Children’s Fund). 2013. “A Passport to Protection: A Guide to Birth Registration Programming.” UNICEF, New York. NOTES World Bank. 2020a. “The Effect of the H1N1 Pandemic on Learning. What to Expect with 1. Some of the emerging messages from these sur- Covid-19?” World Bank, Washington, DC. veys are summarized in chapter 3 of this report. World Bank. 2020b. “Protecting the Poor and 2. EYS, conceptually, is just the sum of enrollment Vulnerable: Social Response Framework for rates by age from age 4 to age 17. Because age-spe- COVID-19.” Unpublished. Social Protection cific enrollment rates are seldom available, how- and  Jobs Global Practice, World Bank, ever, data on enrollment rates by level of school Washington, DC. are used to approximate enrollment rates for the World Bank. 2020c. “COVID-19 and Health age bracket. In India, enrollment rates provided Systems: Just the Beginning.” Unpublished. and used for EYS calculation are age-specific and World Bank, Washington, DC. thus there is no need to ­ approximate the values. World Bank. 2020d. “Protecting People and Economies: Integrated Policy Responses to 3. For more information on SDI, see https://www​ COVID-19.” COVID-19 Coronavirus Response, .sdindicators.org/. World Bank, Washington, DC. APPENDIXES Appendix A The Human Capital Index: Methodology T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 143 1. COMPONENTS OF THE HUMAN CAPITAL INDEX The Human Capital Index (HCI) measures the human capital that a child born today can expect to attain by age 18, given the risks of poor health and poor education that prevail in the country where she lives. The HCI follows the trajectory from birth to adulthood of a child born today.1 In the poorest countries in the world, there is a sig- nificant risk that the child will not survive to her fifth birthday. Even if she does reach school age, there is a further risk that she will not start school, let alone complete the full cycle of 14 years of school from preschool to grade 12 that is the norm in rich coun- tries. The time she does spend in school may translate unevenly into learning, depend- ing on the quality of the teachers and schools she experiences. When she reaches age 18, she carries with her the lasting effects of poor health and nutrition from her childhood that limit her physical and cognitive abilities as an adult. The HCI quantitatively illustrates the key stages in this trajectory and their conse- quences for the productivity of the next generation of workers, with three components: Component 1: Survival. This component of the index reflects the unfortunate reality that not all children born today will survive until the age when the process of human capital accumulation through formal education begins. It is measured using the under-5 mor- (figure A1.1, panel a), with survival to age 5 as the complement of the under-5 tality rate ­ mortality rate. Component 2: School. This component of the index combines information on the quan- tity and quality of education: •• The quantity of education is measured as the number of  years of school a child can expect to obtain by age 18 given the prevailing pattern of enrollment rates (figure A1.1, panel b). The maximum possible value is 14 years, corresponding to the ­ maximum number of years of school obtained as of her 18th birthday by a child who starts preschool at age 4. In the data, expected years of school (EYS) range from about 4 to close to 14 years. •• The quality of education reflects work at the World Bank to harmonize test scores from major international student achievement testing programs into a measure of harmonized test scores.2 Harmonized test scores are measured in units of the Trends in International Mathematics and Science Study (TIMSS) testing program (figure A1.1, panel c). and range from about 300 to about 600 across economies ­ 144 A ppend ix A: T h e H u m a n Ca p i ta l In d e x : M et ho do lo gy Harmonized test scores are used to convert EYS into learning-adjusted years of school- ing (LAYS). LAYS are obtained by multiplying EYS by the ratio of harmonized test scores to 625, corresponding to the TIMSS benchmark of advanced achievement.3 For example, if EYS in a country is 10 and the average harmonized test score is 400, then the country has 10(400/625) = 6.4 LAYS. The distance between 10 and 6.4 represents a learning gap equivalent to 3.6 years of school. Component 3: Health. There is no single broadly accepted, directly measured, and widely available summary measure of health that can be used in the same way that EYS is used as a standard measure of educational attainment. Instead, two proxies for the overall health environment are used: •• Adult survival rate is measured as the share of 15-year-olds who survive until age 60. This measure of mortality serves as a proxy for the range of nonfatal health out- comes that a child born today would experience as an adult if current conditions prevail into the future. •• Healthy growth among children under age 5 is measured as the fraction of children who are not stunted, that is, as 1 minus the share of children under 5 whose height- for-age is more than two standard deviations below the World Health Organization Child Growth Standards’ median. Stunting serves as an indicator for the prenatal, infant, and early childhood health environments, summarizing the risks to good health that children born today are likely to experience in their early years, with important consequences for health and well-being in adulthood. figure A1.1., panels d and e. Data for all Data on these two health indicators are shown in ­ the components of the HCI 2020 by economy are reported in table C8.1 in appendix C. 1.1  Aggregation methodology The components of the HCI are combined into a single index by first converting them into contributions to productivity.4 Multiplying the component contributions to pro- ductivity gives the overall HCI, which is measured in units of productivity relative to a benchmark corresponding to complete education and full health. In the case of survival, the relative productivity interpretation is stark: children who do not survive childhood never become productive adults. As a result, expected productiv- ity as a future worker of a child born today is reduced by a factor equal to the survival rate, relative to the benchmark where all children survive. In the case of education, the relative productivity interpretation is anchored in the large empirical literature measuring the returns to education at the individual level. A rough consensus from this literature is that an additional year of school raises earn- ings by about 8  percent.5 This evidence can be used to convert differences in LAYS across countries into differences in worker productivity. For example, compared with a benchmark where all children obtain a full 14 years of school by age 18, a child who T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 145 obtains only 10  years of education can expect to be 32  percent less productive as an adult (a gap of 4 years of education, multiplied by 8 percent per year). In the case of health, the relative productivity interpretation is based on the empirical literature measuring the economic returns to better health at the individual level. The key challenge in this literature is that there is no unique directly measured summary indicator of the various aspects of health that matter for productivity. This literature often uses proxy indicators for health, such as adult height (see, for example, Case and Paxson 2008; Horton and Steckel 2011). It does so because adult height can be mea- sured directly and reflects the accumulation of shocks to health through childhood and adolescence. A rough consensus drawn from this literature is that an improvement in health associated with a 1 centimeter increase in adult height raises productivity by 3.4 percent. Converting this evidence on the returns to one proxy for health (adult height) into the other proxies for health used in the HCI (stunting and adult survival) requires informa- tion on the relationships between these different proxies (for details, see Kraay 2018; Weil 2007). For stunting, there is a direct relationship between stunting in childhood and future adult height because growth deficits in childhood persist to a large extent into adult- hood, together with the associated health and cognitive deficits. Available evidence sug- gests that a reduction in stunting rates of 10 percentage points increases attained adult height by approximately 1 centimeter, which increases productivity by (10.2 × 0.1 × 3.4) percent, or 3.5 percent. For adult survival, the empirical evidence suggests that, if overall health improves, both adult height and adult survival rates increase in such a way that adult height rises by 1.9 centimeters for every 10-percentage-point improvement in adult survival. This increase implies that an improvement in health that leads to an increase in adult sur- vival rates of 10 percentage points is associated with an improvement in worker pro- ductivity of 1.9 × 3.4 percent, or 6.5 percent. In the HCI, the estimated contributions of health to worker productivity based on these two alternative proxies are averaged if both are available and are used individ- ually if only one of the two is available. The contribution of health to productivity is expressed relative to the benchmark of full health, defined as the absence of stunt- ing, and a 100 percent adult survival rate. For example, compared with a benchmark of no stunting, poor health reduces worker productivity by 30 × 0.34  percent, or 10.2 percent, in a country where the stunting rate is 30 percent. Similarly, compared with the benchmark of 100 percent adult survival, poor health reduces worker produc- tivity by 30 × 0.65 percent, or 19.5 percent, in a country where the adult survival rate is 70 percent. The average of the two estimates of the effect of health on productivity is used in the HCI. 146 A ppend ix A: T h e H u m a n Ca p i ta l In d e x : M et ho do lo gy The overall HCI is constructed by multiplying the contributions of survival, school, and health to relative productivity, as follows: HCI = Survival × School × Health, (A1.1) with the three components defined as 1 − Under 5 Mortality Rate Survival ≡  (A1.2) 1  Harmonized Test Score  φ  Expected Years of School ×  625 −14    School ≡ e   (A1.3) (γ ASR × ( Adult Survival Rate −1)+γ Stunting × ( Not Stunted Rate −1)) / 2  Health ≡ e (A1.4) The components of the index are expressed here as contributions to productivity rela- tive to the benchmark of complete high-quality education and full health. The param- eter ϕ = 0.08 measures the returns to an additional  year of school. The parameters γASR = 0.65 and γStunting = 0.35 measure the improvements in productivity associated with an improvement in health, using adult survival and stunting as proxies for health. The benchmark of complete high-quality education corresponds to 14 years of school and a harmonized test score of 625. The benchmark of full health corresponds to 100 percent child and adult survival and a stunting rate of 0 percent. These parameters serve as weights in the construction of the HCI. The weights are chosen to be the same across economies, so that cross-country differences in the HCI reflect only cross-country differences in the component variables. This choice facili- tates the interpretation of the index. It is also a pragmatic choice because estimating country-specific returns to education and health for all economies included in the HCI is not feasible. figure  A1.1, panel a, child survival rates range from about 90  percent in As shown in ­ the highest-mortality economies to near 100  percent in the lowest-mortality econo- mies. This range implies a loss of productivity of 10  percent relative to the bench- mark of no mortality. LAYS range from about 3 years to nearly 14 years. This gap in LAYS implies a gap in productivity relative to the benchmark of complete education of eϕ(3−14) = e0.08(−11) = 0.4; that is, the productivity of a future worker in economies with the lowest LAYS is only 40 percent of what it would be under the benchmark of com- plete education. For health, adult survival rates range from 50 to 95 percent, and the share of children not stunted ranges from about 45 percent to 95 percent. Using adult γ survival rates indicates a gap in productivity of e  asr (0.5–1) = e  0.65 (–0.5) = 0.72. Thus, based on adult survival rates as a proxy for health, the productivity of a future worker is only 72 percent of what it would be under the benchmark of full health. Using the share of γ children not stunted leads to a gap in productivity of e  Stunting (0.45–1) = e0.35(–0.55) = 0.82. The productivity of a future worker using the stunting-based proxy for health is therefore only 82 percent of what it would be under the benchmark of full health. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 147 Figure A1.1: Components of the Human Capital Index, relative to GDP per capita a. Child survival b. Harmonized test scores 1 600 Nicaragua Kyrgyz Republic Probability of survival to age 5, circa 2020 Solomon Islands Tuvalu Harmonized test scores, circa 2020 Vanuatu Estonia Rwanda Poland Vietnam Malawi 0.95 500 Burundi Ukraine Uzbekistan Kenya Mauritania Angola Cambodia Côte d’Ivoire Qatar Burundi Mali Benin Saudi Arabia 0.9 400 Guinea Kuwait Panama Chad Nigeria Dominican Republic South Africa Nigeria Ghana 0.85 300 6 8 10 12 6 8 10 12 Log GDP per capita at PPP, circa 2020 Log GDP per capita at PPP, circa 2020 d. Fraction of children under age 5 not stunted c. Expected years of school 1 Fraction of children under 5 not stunted, circa 2020 14 St. Lucia North Macedonia Samoa Turkey West Bank and Gaza Bulgaria Kyrgyz Republic Kazakhstan Expected years of school, circa 2020 Nepal Argentina 12 Micronesia, Fed. Sts. Kiribati Haiti Gambia, The Brunei Darussalam Zimbabwe 0.8 Haiti Malaysia 10 Malawi Congo, Dem. Rep. Angola Gabon Botswana 8 Angola 0.6 Sudan Iraq Timor−Leste Eswatini Guatemala 6 Papua New Guinea Mali Liberia 0.4 4 6 8 10 12 6 8 10 12 Log GDP per capita at PPP, circa 2020 Log GDP per capita at PPP, circa 2020 e. Adult survival 1 Morocco Nauru .9 West Bank and Gaza Adult survival rate, circa 2020 Vanuatu Tajikistan Solomon Islands Nepal Timor−Leste .8 Namibia .7 South Africa Zimbabwe Côte d’Ivoire Nigeria .6 Eswatini Central African Republic Lesotho .5 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital index (HCI). Note: The ­figure reports the most recent cross-section of 174 economies for the five HCI components as used to calculate the 2020 HCI. Each panel plots the country-level averages for each component on the vertical axis and GDP per capita in PPP on the horizontal axis. The dashed line illustrates the fitted regression line between GDP per capita and the respective component. Scatter points above (below) the fitted regression line illustrate economies that perform better (worse) in the outcome variable than their level of GDP would predict. Economies above the 95th and below the 5th percentile in distance to the fitted regression line are labeled. PPP = purchasing power parity. 148 A ppend ix A: T h e H u m a n Ca p i ta l In d e x : M et ho do lo gy 2. THE HCI figure 2.1 in the main text. HCI data The HCI scores for 174 countries are presented in ­ are available at www.worldbank.org/humancapital. The HCI is, on average, higher in rich economies than in poor economies and ranges from about 0.3 to about 0.9. The units of the HCI have the same interpretation as the components measured in terms of relative productivity. Consider an economy such as Morocco, which has an HCI of about 0.5. If current education and health conditions in Morocco persist, a child born today will be only half as productive as she could have been if she enjoyed complete education and full health. All the components of the HCI are measured with some error, and this uncertainty naturally has implications for the precision of the overall HCI. To capture this impre- cision, the HCI estimates for each economy are accompanied by upper and lower bounds that reflect the uncertainty in the measurement of the components of the HCI (figure A2.1). These bounds are constructed by recalculating the HCI using lower- and ­ upper-bound estimates of the HCI components. The resulting uncertainty intervals figure A2.1 as vertical ranges around the value of the HCI for each economy. appear in ­ The upper and lower bounds highlight to users that the estimated HCI values for all economies are subject to uncertainty, reflecting the corresponding uncertainty in the components. When these intervals overlap for two economies, it indicates that the Figure A2.1: Human Capital Index with uncertainty intervals 1 Productivity relative to benchmark 0.8 Finland Macao SAR, China Cyprus Switzerland Italy Greece Russian Federation Slovak Republic Albania 0.6 Bulgaria Ecuador Azerbaijan Panama Zimbabwe Togo 0.4 Lesotho Côte d’Ivoire 0.2 6 8 10 12 Log real GDP per capita Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the HCI (on the vertical axis) against log GDP per capita (on the horizontal axis). The dark blue dots represent country averages and the top (bottom) end of the vertical bars indicate the values for the upper (lower) bounds of the HCI. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 149 differences in the HCI estimates for these two economies should not be overinterpreted because they are small relative to the uncertainty around the value of the index itself. The use of upper and lower bounds is intended to help move the discussion away from small differences in the HCI across economies and toward more useful discussions around the level of the HCI and what it implies for the productivity of future workers. The HCI uses the returns to education and health to convert the education and health indicators into differences in worker productivity across economies. The higher the returns, the larger the resulting worker productivity differences. The size of the returns also influences the relative contributions of education and health to the overall index. For example, if the returns to education are high while the returns to health are low, then cross-country differences in education will account for a larger portion of cross-country differences in the index. Although varying the assumptions about the returns to education and health will affect the relative positions of economies on the index, in practice these changes are small because the health and education indicators are strongly correlated across economies (for more details, see Kraay 2018). 2.1  Connecting the HCI to future growth and income The HCI can be connected to future aggregate income levels and growth following the logic of the development accounting literature. This literature typically adopts a simple Cobb-Douglas form for the aggregate production function, as follows: y  =  Akp kh , α 1−α (A2.1) where y is gross domestic product (GDP) per worker; kp and kh are the stocks of physical and human capital per worker; A is total factor productivity; and α is the output elastic- ity of physical capital. To analyze how changes in human capital may affect income in the long run, it is useful to rewrite the production function as follows: α 1  k  1− α 1− α  p  (A2.2) y=  A kh  y   In this formulation, GDP per worker is proportional to the human capital stock per worker, holding constant the level of total factor productivity and the ratio of physical kp capital to output, . This formulation can be used to answer the following question: y By how much does an increase in human capital raise output per worker, in the long run after taking into account the increase in physical capital that is likely to be induced by the increase in human capital? Equation (A2.2) shows the answer: output per worker increases equiproportionately to human capital per worker, that is, a doubling of human capital per worker will lead to a doubling of output per worker in the long run. Linking this framework to the HCI requires a few additional steps. First, assume that the stock of human capital per worker that enters the production function, kh, is equal to the human capital of the average worker. Second, the human capital of the next gener- ation, as measured in the HCI, and the human capital stock that enters the production 150 A ppend ix A: T h e H u m a n Ca p i ta l In d e x : M et ho do lo gy function need to be linked. This linkage can be done by considering different scenarios. Imagine first a status quo scenario in which the expected  years of learning-adjusted schooling and health as measured in the HCI today persist into the future. Over time, new entrants to the workforce with status quo health and education will replace cur- rent members of the workforce until eventually the entire workforce of the future has the expected years of learning-adjusted schooling and health captured in the current φ sNG + γ z NG human capital index. Let kh , NG = e denote the future human capital stock in this baseline scenario, where sNG represents the number of quality-adjusted years of school of the next generation of workers, and γzNG is shorthand notation for the contribution of the two health indicators to productivity in the HCI in equation (A1.4). Contrast this scenario with one in which the entire future workforce benefits from complete educa- tion and enjoys full health, resulting in a higher human capital stock, kh* = ef s*+γ z*, where s* represents the benchmark of 14  years of high-quality school, and z* represents the benchmark of complete health. Assuming that total factor productivity and the physical capital-to-output ratio are the same in the two scenarios, the eventual steady-state GDP per worker in the two scenar- ios is as follows: y = kh, NG =e ( ) ( φ S NG − S * + γ Z NG − Z * ) (A2.3) * * y kh This expression is the same as the HCI in equations (A1.1)–(A1.4) except for the term corresponding to survival to age 5 (because children who do not survive do not become part of the future workforce), which creates a close link between the HCI and potential future growth. Disregarding the contribution of the survival probability to the HCI, equation (A2.3) shows that an economy with an HCI equal to x could achieve GDP per 1 worker that would be times higher in the future if citizens enjoy complete education x and full health (corresponding to x = 1). For example, an economy such as Morocco with an HCI value of about 0.5 could, in the long run, have future GDP per worker in 1 this scenario of complete education and full health that is = 2 times higher than 0.5 GDP per worker in the status quo scenario. What this means in terms of average annual growth rates depends on how long the long run is. For example, under the assumption that it takes 50 years for these scenarios to materialize, then a doubling of future per capita income relative to the status quo corresponds to roughly 1.4 percentage points of additional growth per year. The calibrated relationship between the HCI and future income described here is sim- ple because it focuses only on steady-state comparisons. In related work, Collin and Weil (2018) elaborate on this relationship by developing a calibrated growth model that traces out the dynamics of adjustment to the steady state. They use this model to trace out trajectories for per capita GDP and for poverty measures for individual countries and global aggregates under alternative assumptions for the future path of human cap- ital. They also calculate the equivalent increase in investment rates in physical capital that would be required to deliver the same increases in output associated with improve- ments in human capital. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 151 NOTES 1. This appendix provides a summary of the methodology for the HCI. For additional details, see Kraay (2018), on which this appendix is based. 2. The methodology for harmonizing test scores is detailed in Altinok, Angrist, and Patrinos (2018) and Patrinos and Angrist (2018). 3. This methodology was introduced by World Bank (2018) and is elaborated on in Angrist et al. (2019). 4. This approach has been used extensively in the development accounting literature (for example, Caselli 2005; Hsieh and Klenow 2010). The approach for health closely follows Weil (2007). Galasso and Wagstaff (2016) apply a similar framework to mea- sure the costs of stunting. 5. The seminal methodology is due to Mincer (1958). See Montenegro and Patrinos (2014) for recent cross-country estimates of the returns to schooling. REFERENCES Altinok, N., N. Angrist, and H. A. Patrinos. 2018. “Global Data Set on Education Quality (1965–2015).” Policy Research Working Paper 8314, World Bank, Washington, DC. Angrist, N., S. Djankov, P. Goldberg, and H. A. Patrinos. 2019. “Measuring Human Capital.” Policy Research Working Paper 8742, World Bank, Washington, DC. Case, A., and C. H. Paxson. 2008. “Stature and Status: Height, Ability, and Labour Market Outcomes.” Journal of Political Economy 116 (3): 499–532. Caselli, F. 2005. “Accounting for Cross-Country Income Differences.” In Handbook of Economic Growth, 1st ed., vol. 1, edited by P. Aghion and S. Durlauf, 679–741. North- Holland: Elsevier. Collin, M., and D. Weil. 2018. “The Effect of Increasing Human Capital Investment on Economic Growth and Poverty: A Simulation Exercise.” Policy Research Working Paper 8590, World Bank, Washington, DC. Galasso, E., and A. Wagstaff. 2016. “The Economic Costs of Stunting and How to Reduce Them.” Policy Research Note, World Bank, Washington, DC. Horton, S., and R. Steckel. 2011. “Global Economic Losses Attributable to Malnutrition 1900–2000 and Projections to 2050.” Assessment Paper: Malnutrition, Copenhagen Consensus on Human Challenges, Tewksbury, MA. Hsieh, C. T., and P. Klenow. 2010. “Development Accounting.” American Economic Journal: Macroeconomics 2 (1): 207–23. Kraay, A. 2018. “Methodology for a World Bank Human Capital Index.” Policy Research Working Paper 8593, World Bank, Washington, DC. Mincer, J. 1958. “Investment in Human Capital and Personal Income Distribution.” Journal of Political Economy 66 (4): 281–302. Montenegro, C. E., and H. A. Patrinos. 2014. “Comparable Estimates of Returns to Schooling around the World.” Policy Research Working Paper 7020, World Bank, Washington, DC. 152 A ppend ix A: T h e H u m a n Ca p i ta l In d e x : M et ho do lo gy Patrinos, H. A., and N. Angrist. 2018. “Global Dataset on Education Quality: A Review and Update (2000–2017).” Policy Research Working Paper 8592, World Bank, Washington, DC. Weil, D. 2007. “Accounting for the Effect of Health on Economic Growth.” Quarterly Journal of Economics 122 (3): 1265–1306. World Bank. 2018. The Human Capital Project. World Bank, Washington, DC. Appendix B Back-Calculated ­Human Capital Index T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 155 T he first iteration of the Human Capital Index (HCI) in 2018 made use of the best and most recently available data as of 2018. It was calculated for 157 economies. As is common with indicators, comprehensive revisions of the source data have been implemented since then. For example, gross domestic product series are revised quite often, as are international poverty numbers when improved harmonization of sur- vey data is implemented (see Atamanov et al. 2019 for an example). Revisions to series incorporate the most recent and accurate data and also ensure temporal comparability. In the case of the HCI, the index makes use of data from different institutions, and most of these institutions release their data annually and revise them periodically, as well as revising them for the past. Such revisions provide the opportunity to reflect these measurement improvements in a newly recalculated (back-calculated, actually) 2018 HCI. As a result, the back-calculated 2018 HCI will differ from the original 2018 HCI. Figure B.1 plots these two 2018 HCI versions. On the y-axis are displayed those econo- mies for which it is now possible to generate a back-calculated 2018 HCI, mainly thanks to newly available harmonized test scores. Consequently, the number of economies with a back-calculated 2018 HCI is 167, 10 more than for the 2018 HCI. All economies are quite close to the 45-degree line. The global average for the back-calculated 2018 index is 0.565 as opposed to 0.567 for the 2018 HCI.1 When looking at individual components, however, the differences between the 2018 HCI and the back-calculated 2018 HCI are starker. This difference is particularly ­ relevant for the expected years of school. With newer vintages of data from the United Nations Educational, Scientific and Cultural Organization’s Institute for Statistics, many of the enrollment rates used in the previous round of the HCI have been updated, so data are added for years that are closer to 2018. Updating these data is further complicated enrollment rate because for some economies a preferred rate is now available (total net ­ enrollment is preferred over adjusted net enrollment rate, which is preferred over net ­ rate, which is preferred over gross enrollment rate).2 For example, in most econo- mies where expected years of school increased by at least half a year, the data come from a year that is closer to 2018, and in many cases there is a move to a preferred enrollment type. 156 ­ U MAN CA PI TA L I N DEX A PPEND IX B: BAC K-CALC U LAT E D H Figure B.1: Comparing the 2018 and back-calculated 2018 Human Capital Indexes 1.0 0.8 Kazakhstan Germany United States Azerbaijan HCI 2018 back−calculated Seychelles 0.6 St. Lucia Palau El Salvador St. Vincent and the Grenadines Samoa India Papua New Guinea Tuvalu Marshall Islands 0.4 Eswatini 0.2 0 0.2 0.4 0.6 0.8 1.0 HCI 2018 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index and the 2018 Human Capital Index. Note: The figure plots the HCI 2018 (on the horizontal axis) against the back-calculated HCI 2018 (on the vertical axis). The black line is a 45-degree line; points above (below) the line represent an increase (decrease) in HCI scores between the HCI 2018 and back-calculated HCI 2018. Economies not present in the 2018 HCI but present in the back-calculated HCI and displayed on the vertical axis are Antigua and Barbuda, Dominica, Federated States of Micronesia, Grenada, St. Kitts and Nevis, St. Lucia, the Marshall Islands, Palau, St. Vincent and the Grenadines, and Samoa. HCI = Human Capital Index. The back-calculated HCI makes use of the most recent data available, which allows for an index that better reflects the human capital that a child born in that year could achieve in the country where she lives. The construction of each component for the 2018 back-calculated HCI is detailed in appendix C, and the back-calculated HCI 2018 scores by economy are reported in table C8.1. NOTES 1. Tuvalu is one of the few outliers. The 2018 HCI makes use of stunting as a proxy for health, and the back-calculated 2018 HCI makes use of adult mortality. Stunting rates used in the 2018 HCI corresponded to a 2007 survey. The back-calculated 2018 HCI uses more recent adult mortality rates from 2012, from the World Health Organization, that were not previously available. 2. For details, see the description of the construction of the expected years of school variable in appendix C. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 157 REFERENCE Atamanov, A., R. Castaneda Aguilar, C. Diaz-Bonilla, D. Jolliffe, C. Lakner, D. Mahler, J. Montes, L. Morena Herrera, D. Newhouse, M. Nguyen, E. Prydz, P. Sangraula, S.  Tandon, and J. Yang. 2019. “September 2019 PovcalNet Update: What’s New.” Global Poverty Monitoring Technical Note 10, World Bank, Washington, DC. Appendix C Human Capital Index Component Data Notes T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 159 1. UNDER-5 MORTALITY RATES The probability of survival to age 5 is calculated as the complement of the under-5 mortality rate. The under-5 mortality rate is the probability that a child born in a speci- fied year will die before reaching the age of 5, if subject to current age-specific mortality rates. It is frequently expressed as a rate per 1,000 live births, in which case it must be divided by 1,000 to obtain the probability of dying before age 5. Under-5 mortality rates are calculated by the United Nations Interagency Group for Child Mortality Estimation (IGME) using mortality as recorded in household surveys and vital registries. The IGME compiles and assesses the quality of all available nation- ally representative data relevant to the estimation of child mortality, including data from vital registration systems, population censuses, household surveys, and sample registration systems. Globally, birth registration coverage remains inadequate, hav- ing increased by only 7 percentage points (from 58 percent to 65 percent) in the past decade (UNICEF 2013). In Sub-Saharan Africa, only eight countries have coverage of percent or more for birth registration.1 80 ­ The IGME assesses data quality, recalculates data inputs, and makes adjustments if needed by applying standard methods. It then fits a statistical model to these data to generate a smooth trend curve that averages over possibly disparate estimates from the different data sources for an economy. Finally, it extrapolates the model to a target year. Data are reported annually and cover 198 economies. The IGME estimates are disaggregated by gender and include uncertainty intervals corresponding to 95 percent confidence intervals. 2020 update Under-5 mortality rates for the 2020 update of the Human Capital Index (HCI) come from the September 2019 update of the IGME estimates, available at the Child Mortality Estimates website (see also UNIGME 2019).2 Data for the back-calculated 2018 HCI come from 2017. Data for the baseline comparator year of 2010 come from 2010. Because under-5 mortality rates are estimated by modeling all available child mortality data from vital registration systems, population censuses, household surveys, and sam- ple registration systems, every new release of data from the IGME updates estimates for all the years in the time series. As a result, data for the same past year might differ slightly across updates. 160 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Values for under-5 mortality rates used to produce the back-calculated HCI 2018 are aligned with but not the same as those used in the previous iteration of the HCI, as illustrated in figure C1.1. Data from the two vintages align along the 45-degree line. The figure highlights the four economies where under-5 mortality rates have changed by more than 10 deaths per 1,000 live births. The largest revisions were for Nigeria (which went from 100 to 122 deaths per 1,000 live births) and Guinea (which went from 86 to 103 deaths per 1,000 live births). Figure C1.2 reports the most recent cross-section of under-5 mortality rates used to calculate the 2020 HCI. Child mortality rates range from about 0.002 (2 per 1,000 live births) in the richest countries to about 0.120 (120 per 1,000 live births) in the poorest countries. Under-5 mortality rates tend to be slightly lower for girls than for boys, as reported in figure C1.3. In the figure, the solid dot indicates the country average, the triangle indicates the average for girls, and the horizontal bar indicates the average for boys. The average under-5 mortality rate for boys was 0.03 (30 deaths per 1,000 live births), compared to 0.025 for girls (25 deaths per 1,000 live births). Figure C1.1: Comparing original and back-calculated 2018 under-5 mortality rates 0.15 Nigeria Under-5 mortality rates, 2018 back-calculated Guinea 0.10 Madagascar 0.05 Lao PDR 0 0.05 0.10 0.15 Under-5 mortality rates, 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the under-5 mortality rates as used in the 2018 HCI (on the horizontal axis) against the under-5 mortality rates used for the back-calculated 2018 HCI (on the vertical axis). Economies where under-5 mortality rates have changed by more than 10 deaths per 1,000 live births between 2018 and back-calculated 2018 are labeled. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 161 Figure C1.2: Under-5 mortality rates, Human Capital Index 2020, relative to GDP per capita Chad 0.10 Under-5 mortality rate, circa 2020 Ethiopia 0.05 Congo, Rep. Senegal India South Africa Indonesia Morocco Seychelles Iran, Islamic Rep. Ecuador Thailand Oman China Bulgaria Hungary Kuwait Switzerland 0 Greece CyprusFinland Ireland –0.05 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots under-5 mortality rates (on the vertical axis) against log GDP per capita at 2011 PPP US dollars (on the horizontal axis). PPP = purchasing power parity. Figure C1.3: Sex-disaggregated under-5 mortality rates, relative to GDP per capita 0.15 Sex-disaggregated under-5 mortality rate, circa 2020 Chad 0.10 Ethiopia 0.05 Congo, Rep. Senegal India South Africa Indonesia Morocco Iran, Islamic Rep. Ecuador Seychelles Thailand Oman China Bulgaria Hungary Kuwait Switzerland GreeceCyprusFinland Ireland 0 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots sex-disaggregated under-5 mortality rates (on the vertical axis) against log GDP per capita (on the horizontal axis). The solid dot indicates the national average, the triangle shows the average value for girls, and the horizontal line shows the average value for boys. PPP = purchasing power parity. 162 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C1.4: Under-5 mortality rates, by income group and region a. Income groups Low income 0.07 Lower-middle income 0.04 Upper-middle income 0.02 High income 0.01 0 0.02 0.04 0.06 0.08 Under-5 mortality rate, circa 2020 b. Regions Sub-Saharan Africa 0.07 South Asia 0.04 East Asia and Pacific 0.02 Latin America and the Caribbean 0.02 Middle East and North Africa 0.02 Europe and Central Asia 0.01 North America 0.01 0 0.02 0.04 0.06 0.08 Under-5 mortality rate, circa 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figures plot regional and income group average values for under-5 mortality rates. Figure C1.4 reports average child mortality rates by income group and by World Bank region. Mortality rates tend to be highest in low-income economies, and regional aver- ages are highest in Sub-Saharan Africa and South Asia, reflecting that poor economies continue to bear a disproportionate burden of child mortality. 2. EXPECTED YEARS OF SCHOOL The expected years of school (EYS) component of the HCI captures the number of years of school a child born today can expect to achieve by age 18, given the prevail- ing pattern of enrollment rates in her country.3 Conceptually, EYS is simply the sum of enrollment rates by age from age 4 to 17. Because age-specific enrollment rates are not broadly or systematically available, more readily available data on enrollment rates by level of school are used to approximate enrollment rates in different age brackets. Preprimary enrollment rates approximate the enrollment rates for 4- and 5-year-olds, primary enrollment rates approximate for 6- to 11-year-olds, lower-secondary rates approximate for 12- to 14-year-olds, and upper-secondary rates approximate for 15- to 17-year-olds. Cross-country definitions in school starting ages and the duration of the T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 163 various levels of school imply that these rates will be only approximations of the num- ber of years of school a child can expect to complete by age 18. Given that the objective is to obtain a close proxy to age-specific enrollment rates, the preferred measure is the total net enrollment rate (TNER). TNER measures the fraction of children in the theoretical age range for a given level of school who are in school at any level. For many countries, the TNER is not readily available for all levels and thus, in many instances, less preferred rates are used. The order of preference for the use of enrollment rates is the following: 1. Total net enrollment rate (TNER) measures the fraction of children in the theoretical age range for a given level of school who are in school at any level. Because there is no level before preprimary, TNER is not available, and ANER is the preferred measure. 2. Adjusted net enrollment rate (ANER) measures the fraction of children in the theo- retical age range for a given level of school who are in that level or the level above. 3. Net enrollment rate (NER) measures the fraction of children in the theoretical age range for a given level of school who are in that level of school. 4. Gross enrollment rate (GER) measures the number of children of any age who are enrolled in a given level as a fraction of the number of children in that age range. The conceptually appropriate enrollment rate to approximate enrollment rates by age brackets is the repetition-adjusted total net enrollment rate. The primary source for enrollment and repetition rates is the United Nations Educational, Scientific, and Cultural Organization’s Institute for Statistics (UIS),4 revised and supplemented with data provided by World Bank country teams that participated in an extensive data review process. When the resulting data on TNERs are incomplete, ANERs, NERs, or GERs are used instead in that order of priority. The same enrollment rate type is used for a given level of education over time. Because EYS is constructed using primarily administrative data on enrollment rates, uncertainty intervals are not available for this component of the HCI. This does not imply that there is no measurement error; instead, the use of administrative data implies that there is no error due to modeling or sampling.5 Consequently, uncertainty in the measurement of EYS is not reflected in the uncertainty intervals of the overall HCI. EYS is calculated as follows: 4 EYS = ∑RateiYi ; i = preprimary , primary , lower−secondary , upper−secondary (C2.1) i where Ratei is the enrollment rate for the preferred enrollment type available for that level, and Yi is the number of years corresponding to each level.6 164 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Enrollment rates for 2020 and 2010 Temporal coverage for enrollment rates is not complete in the UIS public database. Consequently, the first step toward ensuring that the rates used are the most recent and accurate relies on getting inputs from World Bank specialists working on each econ- omy to validate and provide more recent values when available.7 Enrollment rates for 2020 for each school level and for the four enrollment rate types (TNER, ANER, NER, and GER) are obtained from UIS.8 Any inputs from World Bank teams working on specific economies are then added to the corresponding enrollment rates. Existing gaps for 2020 in enrollment rates for each level and economy are filled by setting the 2020 enrollment rate equal to the latest enrollment rate available for that enrollment rate type. This process is henceforth referred to as the carryforward rule. The rule is applied if the latest available enrollment rate is not older than 10 years.9 This process ensures that the HCI of 2020 and the back-calculated HCI for 2018 are calcu- lated compatibly with the first version of the HCI released in 2018. Additionally, enroll- ment rates are adjusted for repetition, when repetition rates are available; otherwise a repetition rate of 0 is assumed. Finally, enrollment rate types are chosen on the basis of the filled series (that is, using the 2020 rates when gaps have been filled) and according to the following order of preference: TNER, ANER, NER, and GER.10 In the current HCI update, an effort has been made to also populate an HCI for 2010 using data circa 2010.11 Because data collection and availability generally improve over time, enrollment rates for 2010 and earlier are less likely to be available than more recent rates. This means that the rule from the first edition of the HCI used to obtain an EYS measure for 2020 and 2018 cannot be applied to obtain rates for 2010, because it is not possible to apply the carryforward rule for all economies for which comparable data over time for other components of the HCI are available circa 2010 and circa 2020. Moreover, to allow that (1) the preferred enrollment type is used for 2020 and (2) the enrollment rate type for a given school level in a given economy is the same over time, different rules are applied to fill in the year 2010 to ensure comparability over time and to maximize economy coverage. The rules used to fill in gaps in enrollment values for 2010 rely on annualized growth rates and are implemented sequentially for each school level (that is, preprimary, pri- secondary, and upper-secondary): mary, lower-­ 1. If a value is available for 2010 that comes from the same enrollment type as the value assigned to 2020, this value is used (provided the 2010 value has not already been used to fill in 2020 using the carryforward rule).12 2. If a 2010 value is not available in the enrollment type used to populate 2020, avail- able data for that enrollment type are used to generate an annualized growth rate (agr) that is then applied to the most recent year before 2010 (Year) for which data are available to generate a value for 2010. Annualized growth rates are calculated between the years before and after 2010 that are closest to 2010 and for which non- missing values of the selected enrollment type are available. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 165 Rate2010 = RateYear (1 + agr)2010−Year (C2.2) 3. If no value is available for the chosen enrollment type before 2010, then annual- ized growth rates are obtained using GER, which are available for most economies. Annualized growth rates using GER are calculated using the same approach as out- lined before, between the years before and after 2010 that are closest to 2010 and for which nonmissing values of GER are available. For example, if TNER is used to populate 2020 and its earliest value is for 2012, then an annualized growth rate is obtained from the GER. The annualized growth rate from the GER is calculated between the first rate available for 2010 or before, and the first rate available for or after 2012, because the TNER is available for 2012. This rate is then applied back- ward to the TNER of 2012 to obtain a value for 2010. Rate2010 = RateYear (1 + agrGER)2010−Year (C2.3) The process described above yields a value for EYS in 2010 for 99 out of 114 eligible economies. For the remaining 15 economies, calculations to populate enrollment for 2010 are on a case-by-case basis in order to populate enrollment rates in 2010. Disaggregation by sex Disaggregation by sex is an important feature of the HCI. Although the rules presented in the previous section are meant to complete the EYS for both sexes, there are still adjustments required to ensure that EYS values for boys and girls are plausible. These adjustments are necessary because, although a certain enrollment type may be available as a combined series, it may lack sex-disaggregated information. In other instances, it may be necessary to adjust the disaggregated series because values for both girls and boys are above (or below) those of the combined enrollment rate. To fill in the sex-disaggregated enrollment rates, the following rules are applied: 1. For every year for which rates for both genders and the aggregate are available, the male-to-female ratio and the population share of males and females are calculated. 2. For years that are missing a sex-disaggregated rate, the shares and ratios calculated in step 1 from the closest year available in the past (but not more than 10 years back) are used to impute missing values. 3. For the remaining years for which the disaggregated enrollment rates for the preferred enrollment type are still missing, the male and female shares and, where ­ available, the male-to-female ratio from GER enrollment rates are used to impute a value. It is still possible that the rules above, when applied, return inconsistent values, and it is necessary to adjust the disaggregated series when the male and female rates are both larger (or smaller) than the aggregate enrollment rate. In those cases, we adjust the 166 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES disaggregated enrollment rate to the value that leaves the aggregate rate Ratemf at the same distance from each of the disaggregated rates. Rate f − Ratem Rate∗f = Ratemf + (C2.4) 2 Ratem − Rate f Rate∗ = Ratemf + (C2.5) m 2 2018 back-calculated EYS Data for the 2020 update of EYS rely on data from UIS, which releases data in September of each year and completes the release in February of the following year. The February 2020 release of enrollment data from UIS is used for the update of EYS, effectively reporting enrollment data up to 2019. The latest data release from UIS is complemented with rates obtained by World Bank staff.13 The updated data provide an opportunity to update EYS values from the 2018 vintage of the HCI to the latest information available to arrive at a back-calculated EYS for 2018. Because the update allows for the calculation of EYS incorporating more recent data or data from a different enrollment type than what was used in the first vin- tage of the HCI, the EYS from the first vintage of the HCI is not comparable with the current vintage of the EYS. Differences in the 2018 value of EYS between the 2018 vintage and the 2020 update may be due to a combination of three factors: 1. Data are updated in UIS, or by World Bank staff. 2. Data on enrollment from a more recent year are now available. 3. Different enrollment types have become available. In some cases, it will be possi- ble to move to a more preferred enrollment type, whereas in others it is necessary to rely on a less preferred enrollment type. The latter may be the case if UIS has removed the series or if the series is too old. The average absolute deviation between the back-calculated EYS and the 2018 vintage is 0.3 year; however, the changes may be substantial for specific countries (see figure C2.1). Although the differences between vintages are considerable, they are mostly due to the fact that the EYS measure generated in this round relies on more preferred rates, newer data, or both. For the 2018 back-calculated EYS, the enrollment data for at least one of the levels for 131 economies come from a more recent year.14 For 85 economies, the enrollment rates for all levels correspond to a more recent year. In 21 economies it is necessary to change to a less preferred series for at least one of the levels. This change occurs mostly when the series has been removed in the update of the source T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 167 data. Conversely, in 20 economies, it is possible to calculate EYS for at least one level with a more preferred type of enrollment rate. Figure C2.2 and figure C2.3 present details for economies in which EYS in the new data vintage has increased by at least half a year. In most economies where EYS increased by at least half a year, there is a move to a data point that is closer to 2018. The exception is Zimbabwe, where all the enrollment rates correspond to the same year and are for the same enrollment type. In the case of Zimbabwe, the dif- ference is explained as being due to the change in the vintage of UIS data. The big- gest change for Zimbabwe is observed for primary, for which the rate increased by almost 10 percentage points. A different case can be observed for Côte d’Ivoire, where every data point comes from a more recent year. In this case, however, the difference is also complicated because the previous EYS was built with rates that did not come from UIS but were drawn from government sources by World Bank staff. In the case of Papua New Guinea, the change is due to two factors. Not only are more recent data used for all levels, but also the data used for all but primary are from a preferred series (figure C2.3). These three countries illustrate the multiple sources for the potential mismatch between the EYS value pro- duced in 2018 and the updated 2018 back-calculated EYS. Figure C2.1: Comparing original and back-calculated 2018 expected years of school 14 Expected years of school, 2018 back-calculated Azerbaijan 12 Kenya Zimbabwe Nicaragua Papua New Guinea South Africa Tuvalu 10 Nigeria Lesotho Bangladesh Madagascar 8 Mauritania Eswatini 6 4 6 8 10 12 14 Expected years of school, 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the expected years of school as used in the 2018 HCI (on the horizontal axis), and the expected years of school used for the back-calculated 2018 HCI (on the vertical axis). Economies where expected years of school changed by 0.75 years or more between 2018 and back-calculated 2018 are labeled. 168 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C2.2: Vintage data year for back-calculated 2018 and original 2018, increase of 0.5 year or more in expected years of school a. Preprimary b. Primary Zimbabwe Zimbabwe West Bank and Gaza West Bank and Gaza Vietnam Vietnam Tonga Tonga Timor-Leste Timor-Leste South Africa South Africa Papua New Guinea Papua New Guinea Pakistan Pakistan Nigeria Nigeria Mauritania Mauritania Madagascar Madagascar Lesotho Lesotho Kenya Kenya India India Dominican Republic Dominican Republic Côte d’Ivoire Côte d’Ivoire Azerbaijan Azerbaijan 2005 2010 2015 2020 2005 2010 2015 2020 c. Lower-secondary d. Upper-secondary Zimbabwe Zimbabwe West Bank and Gaza West Bank and Gaza Vietnam Vietnam Tonga Tonga Timor-Leste Timor-Leste South Africa South Africa Papua New Guinea Papua New Guinea Pakistan Pakistan Nigeria Nigeria Mauritania Mauritania Madagascar Madagascar Lesotho Lesotho Kenya Kenya India India Dominican Republic Dominican Republic Côte d’Ivoire Côte d’Ivoire Azerbaijan Azerbaijan 2005 2010 2015 2020 2005 2010 2015 2020 2018 back-calculated 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The panels plot the year of data used for calculation of expected years of school. Solid dot represents the data used for the back-calculated 2018 HCI, and x indicates the data used for the calculation of the 2018 HCI. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 169 Figure C2.3: Enrollment type for back-calculated 2018 and original 2018, increase of 0.5 year or more in expected years of school a. Preprimary b. Primary Zimbabwe Zimbabwe West Bank and Gaza West Bank and Gaza Vietnam Vietnam Tonga Tonga Timor-Leste Timor-Leste South Africa South Africa Papua New Guinea Papua New Guinea Pakistan Pakistan Nigeria Nigeria Mauritania Mauritania Madagascar Madagascar Lesotho Lesotho Kenya Kenya India India Dominican Republic Dominican Republic Côte d’Ivoire Côte d’Ivoire Azerbaijan Azerbaijan TNER ANER NER GER TNER ANER NER GER c. Lower-secondary d. Upper-secondary Zimbabwe Zimbabwe West Bank and Gaza West Bank and Gaza Vietnam Vietnam Tonga Tonga Timor-Leste Timor-Leste South Africa South Africa Papua New Guinea Papua New Guinea Pakistan Pakistan Nigeria Nigeria Mauritania Mauritania Madagascar Madagascar Lesotho Lesotho Kenya Kenya India India Dominican Republic Dominican Republic Côte d’Ivoire Côte d’Ivoire Azerbaijan Azerbaijan TNER ANER NER GER TNER ANER NER GER 2018 back-calculated 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The panels plot the enrollment type used for calculation of expected years of school. Solid dot represents the data used for the back-calculated 2018 HCI, and x indicates the data used for the calculation of the 2018 HCI. ANER = adjusted net enrollment rate; GER = gross enrollment rate; NER = net enrollment rate; TNER = total net enrollment rate. 170 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C2.4 and figure C2.5 present details for countries in which the back-calculated 2018 EYS has decreased by at least half a year (figure C2.4 reports changes in year and figure C2.5 changes in enrollment types). Only in Tanzania is it necessary to move to an older rate, but it is to a preferable type (GER to TNER), and it is only one year older. In Bangladesh, the change in EYS is mostly driven by changes in prepri- secondary. For preprimary, the back-calculated rate relies on data mary and lower-­ from 2017 versus 2011, although it is for a less preferred rate (GER versus ANER). Meanwhile, for lower-secondary the rate for the back-calculated EYS is for a more recent year but a less preferred rate (ANER versus TNER). In this case, it is necessary to move to a less preferred rate because the TNER series is no longer available in the UIS data vintage for years after 2010. In India, because the latest available TNER series in UIS is for 2013, World Bank staff have sourced more recent data. EYS is now built with age-specific enrollment profiles that make use of information from the updated Unified District Information System for Education (UDISE+) from the Ministry of Human Resource Development, as well as early childhood care and education enrollment from the Ministry of Women and Child Development and Entrepreneurship and population projections from the Ministry of Health and Family Welfare. The resulting EYS for the back-calculated HCI is 10.8 versus 10.2, which was used in the calculation for the 2018 HCI. Figure C2.4 and figure C2.5 present selected evidence comparing 2018 EYS estimates used in the calculation of the 2018 HCI against those used in the 2020 update. For a more detailed look into the differences, table C7.1 presents enrollment data for all the economies where the absolute EYS change between the back-calculated 2018 and the 2018 versions of the index is greater than half a year. 2020 update The 2020 EYS shows a high rank correlation to the EYS from 2018, as well as a strong positive relationship between the 2020 EYS and log gross domestic product per ­ capita (figure C2.6). EYS tends to be slightly higher for girls than for boys, as reported in figure C2.7. In figure C2.7, the solid dot indicates the country average, the triangle indi- cates the average for girls, and the horizontal bar indicates the average for boys. The average EYS for boys was 11.3 compared to 11.4 for girls. Disparity in EYS between girls and boys is lower in richer countries. Figure C2.8 reports average EYS by income group and by World Bank region. EYS tends to be lowest in low-income economies, and regional averages are lowest in Sub- Saharan Africa and South Asia, which suggests that much work remains to be done to close the gap in low-income economies. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 171 Figure C2.4: Vintage data year for back-calculated 2018 and original 2018, decrease of 0.5 year or more in expected years of school a. Preprimary b. Primary Vanuatu Vanuatu Tuvalu Tuvalu Tanzania Tanzania Solomon Islands Solomon Islands Seychelles Seychelles Panama Panama Nicaragua Nicaragua Germany Germany Eswatini Eswatini Bulgaria Bulgaria Bangladesh Bangladesh 2005 2010 2015 2020 2005 2010 2015 2020 c. Lower-secondary d. Upper-secondary Vanuatu Vanuatu Tuvalu Tuvalu Tanzania Tanzania Solomon Islands Solomon Islands Seychelles Seychelles Panama Panama Nicaragua Nicaragua Germany Germany Eswatini Eswatini Bulgaria Bulgaria Bangladesh Bangladesh 2005 2010 2015 2020 2005 2010 2015 2020 2018 back-calculated 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The panels plot the year of data used for calculation of expected years of school. Solid dot represents the data used for the back-calculated 2018 HCI, and x indicates the data used for the calculation of the 2018 HCI. 172 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C2.5: Enrollment type for back-calculated 2018 and original 2018, decrease of 0.5 year or more in expected years of school a. Preprimary b. Primary Vanuatu Vanuatu Tuvalu Tuvalu Tanzania Tanzania Solomon Islands Solomon Islands Seychelles Seychelles Panama Panama Nicaragua Nicaragua Germany Germany Eswatini Eswatini Bulgaria Bulgaria Bangladesh Bangladesh TNER ANER NER GER TNER ANER NER GER c. Lower-secondary d. Upper-secondary Vanuatu Vanuatu Tuvalu Tuvalu Tanzania Tanzania Solomon Islands Solomon Islands Seychelles Seychelles Panama Panama Nicaragua Nicaragua Germany Germany Eswatini Eswatini Bulgaria Bulgaria Bangladesh Bangladesh TNER ANER NER GER TNER ANER NER GER 2018 back-calculated 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The panels plot the enrollment type used for calculation of expected years of school. Solid dot represents the data used for the back-calculated 2018 HCI, and x indicates the data used for the calculation of the 2018 HCI. ANER = adjusted net enrollment rate; GER = gross enrollment rate; NER = net enrollment rate; TNER = total net enrollment rate. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 173 Figure C2.6: Expected years of school circa 2020, relative to GDP per capita 15 Finland Ireland Cyprus Expected years of school, circa 2020 China Greece Switzerland Ecuador Seychelles Hungary Thailand Oman Indonesia Bulgaria Iran, Islamic Rep. Kuwait India Morocco South Africa 10 Congo, Rep. Ethiopia Senegal Chad 5 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots expected years of school (on the vertical axis) against log GDP per capita at 2011 PPP US dollars (on the horizontal axis). PPP = purchasing power parity. Figure C2.7: Sex-disaggregated expected years of school, relative to GDP per capita 14 Ireland Finland Cyprus Greece Switzerland Sex-disaggregated expected years of school, circa 2020 China Seychelles Ecuador Hungary Thailand Oman Indonesia Bulgaria 12 Kuwait Iran, Islamic Rep. India Morocco South Africa 10 Congo, Rep. 8 Ethiopia Senegal 6 Chad 4 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots sex-disaggregated expected years of school. The solid dot indicates the national average, the triangle shows the average value for girls, and the horizontal line shows the average value for boys. PPP = purchasing power parity. 174 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C2.8: Expected years of school, by income group and region a. Income groups Low income 7.6 Lower-middle income 10.4 Upper-middle income 11.8 High income 13.2 0 5 10 15 Expected years of school, circa 2020 b. Regions 8.3 Sub-Saharan Africa 10.8 South Asia 11.6 Middle East and North Africa 11.9 East Asia and Pacific 12. 1 Latin America and the Caribbean Europe and Central Asia 13.1 North America 13.3 0 5 10 15 Expected years of school, circa 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figures plot regional and income group average values for expected years of school. 3. HARMONIZED TEST SCORES The school quality adjustment is based on a large-scale effort to harmonize inter- national student achievement tests from several multicountry testing programs to produce the Global Dataset on Education Quality. A detailed description of the test score harmonization exercise is provided in Patrinos and Angrist (2018), and the HCI draws on an updated version of this dataset as of January 2020.15 The dataset harmonizes scores from three major international testing programs—the Trends in International Mathematics and Science Study (TIMSS) program, the Progress in International Reading Literacy Study (PIRLS), and the Programme for International Student Assessment (PISA)—as well as from four major regional testing programs— the Southern and Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ), the Program for the Analysis of Education Systems (PASEC), the Latin American Laboratory for Assessment of the Quality of Education (LLECE), and the Pacific Island Learning and Numeracy Assessment (PILNA). It also incorporates Early Grade Reading Assessments (EGRAs) coordinated by the United States Agency for International Development. The harmonization methodology relies on the production of an exchange rate between international student achievement tests and their regional counterparts, which can then be used to place tests on a common scale. Test scores are converted into TIMSS units T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 175 as the numeraire, corresponding roughly to a mean of 500 and a standard deviation across students of 100 points. The exchange rate is based on the ratio of average econ- omy scores in each program to the corresponding economy scores in the numeraire testing program for the set of economies participating in both the numeraire and the other testing program. For example, consider the set of economies that participate in both the PISA and the TIMSS assessments. The ratio of average PISA scores to average TIMSS scores for this set of economies provides a conversion factor for PISA into TIMSS scores that can then be used to convert the PISA scores of all economies into TIMSS scores. The exchange rate is calculated pooling all overlapping observations between within-country 2000 and 2017 and is therefore constant over time. This ensures that ­ fluctuations in harmonized test scores over time for a given testing program reflect only changes in the test scores themselves and not changes in the conversion factor between tests.16 The most recent update of the dataset also uses the 2000–17 period to calculate exchange rates, so that the rates between testing programs do not change between the 2018 and 2020 versions of the database. 2020 update The 2020 update of the Global Dataset on Education Quality extends the database to 184 economies, drawing on a large-scale effort by the World Bank to collect learning data globally. Updates to the database come from new data from PISA 2018, PISA for Development (PISA-D), PILNA, and EGRA. The database adds 20 new economies (8 using EGRAs, 8 using PILNA, 3 using PISA and PISA-D, and 1 using a national TIMSS-equivalent assessment). These additions bring the percentage of the global school-age population represented by the database to 98.7 percent. In addition, more recent data points have been added for 94 economies (75 from PISA 2018, 7 from PISA-D, 5 from EGRAs, and 7 from PILNA). In most cases, the tests are designed to be nationally representative. There are, how- ever, some notable cases in which they are not. In the case of China, extrapolations are needed to arrive at nationally representative estimates, because only a small num- ber of relatively affluent regions have participated in PISA assessments. For India, the only internationally comparable assessment is the 2009 PISA. Instead, recent national assessment data and exchange rates with international benchmarks derived from the UIS Global Alliance to Monitor Learning (GAML) process are used to estimate a national harmonized test score (HTS). In a number of countries, EGRAs are not nationally rep- resentative and are identified as EGRANR in the data documentation.17 When economies participate in multiple testing programs, a hierarchy of tests is applied to determine which HTS to use. This hierarchy is based on the strength of the underlying test construction; the number of overlapping economies to produce the exchange rate; and consistency in administration, procedures, and documenta- tion over time. The first HTS choice is an international test like the PISA, TIMSS, or PIRLS. The next-choice HTS is a regional test, like LLECE, SACMEQ, PASEC, and 176 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES PILNA (in that order). Finally, if neither an international nor a regional test is avail- able, an economy is assigned an HTS that comes from an EGRA. The one exception to this rule is the Republic of Yemen, for which TIMSS data from 2007 and 2011 yield implausibly low scores and are replaced with EGRA data from 2011. Uncertainty intervals for HTSs are constructed by bootstrapping. Patrinos and Angrist (2018) take 1,000 random draws from the distribution of subject-grade average test scores for each test in their dataset. They then form exchange rates and calculate HTSs in each bootstrapped sample. The 2.5th and 97.5th percentiles of the distribution of the resulting HTSs across bootstrapped samples constitute the lower and upper bounds of the uncertainty interval for the HTS. Test scores are harmonized by subject and grade and are then averaged across subjects and grades.18 HTSs for the 2020 HCI come from the most recently available test as of 2019, whereas data for the back-calculated 2018 HCI come from the most recent test available as of 2017. Data for the baseline comparator year of 2010 are populated for each economy using the test closest to 2010, typically with a minimum gap of five years between the test used to populate the 2010 and 2020 cross-sections. Some exceptions to this rule include Bahrain, Botswana, the Islamic Republic of Iran, Kuwait, Oman, and South Africa, for which data from the 2011 TIMSS or PIRLS are used to calculate the 2010 HCI, and data from the 2015 TIMSS or PIRLS are used to calculate the 2020 HCI. In addition, data for Timor-Leste come from a 2009 and 2011 EGRA, and data for Vietnam come from a 2012 PISA for the 2010 HCI and a 2015 PISA for the 2020 HCI. In order to ensure the comparability of HTSs across time, we ensure that the 2010 and 2020 cross-sections are populated with scores that come from the same testing program. That is, if an economy has an HTS from a PISA test circa 2020, it must also have scores from another PISA test circa 2010 to be included in the over-time compar- ison. The five exceptions are Algeria, Morocco, North Macedonia, Saudi Arabia, and Ukraine. For Algeria, HTSs from the PIRLS or the TIMSS in 2007 are used to populate the 2010 HCI, and HTSs based on the PISA in 2015 are used to populate the 2020 HCI. For Morocco, North Macedonia, Saudi Arabia, and Ukraine, data from PIRLS or TIMSS in 2011 are used for the 2010 HCI, and data from PISA 2018 are used for the 2020 HCI. To maximize comparability with PISA, only scores from secondary-level schooling are considered for these five economies for the 2010 HCI. Applying these rules yields a sample of 103 economies with test scores in both 2010 and 2020. Test scores used to produce the back-calculated HCI 2018 are similar to those used in the previous iteration of the HCI, as illustrated in figure C3.1. Data from the two vintages align almost perfectly along the 45-degree line because outcomes for these economies come from the same test and the same harmonization methodology. The figure  also highlights the 10 economies for which test scores have changed because a more recent test was made available in the latest version of the database or, as in the case of China and India, because alternate methodologies were used to refine esti- mates of national average learning outcomes (see table C3.1 for details on changes in T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 177 Figure C3.1: Comparing original and back-calculated 2018 harmonized test scores Harmonized test scores, 2018 back-calculated 600 500 El Salvador China 400 India Tonga Gambia, The Tuvalu Haiti Vanuatu Congo, Dem. Rep. Nigeria 300 400 500 600 Harmonized test scores, 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the harmonized test scores as used in the 2018 HCI (on the horizontal axis), and the harmonized test scores used for the back-calculated 2018 HCI (on the vertical axis). Economies where harmonized test scores have changed between 2018 and back-calculated 2018 are labeled. the source of test  data). In the case of El Salvador, a choice guided by consultations with the country team was made to replace the previous test used (TIMSS/PIRLS from 2007) with a 2006 LLECE for reading to enhance comparability to the 2018 EGRA (not representative) used in 2020. Figure C3.2 reports the most recent cross-section of test scores used to calculate the 2020 HCI. HTSs range from about 575 in the richest economies to about 305 in the poorest economies. To interpret these units, note that 400 corresponds to the bench- mark of low proficiency in TIMSS at the student level, and 625 corresponds to advanced proficiency. Test scores tend to be slightly higher for girls than for boys, as reported in figure C3.3. In the figure, the solid dot indicates the country average, the triangle indicates the aver- age for girls, and the horizontal bar indicates the average for boys. Globally, the average HTS for boys was 420, compared with 430 for girls. Figure C3.4 reports average test scores by income group and by World Bank region. Test scores tend to be lowest in low-income economies, and regional averages are low- est in South Asia and Sub-Saharan Africa. 178 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C3.1: Source data for economies with different values in original 2018 and back-calculated 2018 2018 vintage 2020 vintage Economy Test Year Value Test Year Value China PISA/PIRLS (Extrapolated) 2015 456 PISA/PIRLS (Extrapolated) 2015 441 Congo, Dem. EGRANR 2012 318 EGRANR 2015 310 Rep. El Salvador TIMSS/PIRLS 2007 362 LLECE 2006 438 Gambia, The EGRA 2011 338 EGRA 2016 353 Haiti EGRANR 2013 345 EGRA 2016 339 India PISA 2009 355 NAS 2017 399 Malaysia TIMSS 2015 468 TIMSS/PIRLS 2015 468 Nigeria EGRANR 2010 325 EGRANR 2014 309 Tonga EGRA 2014 376 PILNA 2015 370 Tuvalua EGRA 2016 387 EGRA 2016 351 Vanuatu EGRA 2010 356 PILNA 2015 332 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: EGRA = Early Grade Reading Assessment; EGRANR = Non-nationally-representative Early Grade Reading Assessment; LLECE = Latin American Laboratory for Assessment of the Quality of Education; NAS = National Achievement Survey; PILNA = Pacific Island Learning and Numeracy Assessment; PIRLS = Progress in International Reading Literacy Study; PISA = Programme for International Student Assessment; TIMSS = Trends in International Mathematics and Science Study. a. Data for Tuvalu from the 2016 EGRA were revised once student-level data were made available to the harmonized test score team. Figure C3.2: Harmonized test scores, Human Capital Index 2020, relative to GDP per capita 600 Harmonized test scores, circa 2020 Finland Ireland Switzerland 500 Cyprus Hungary Greece Seychelles China Bulgaria Iran, Islamic Rep. Thailand Oman Ecuador Senegal 400 India Indonesia Morocco Kuwait Congo, Rep. Ethiopia South Africa Chad 300 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots harmonized test scores (on the vertical axis) against log GDP per capita at 2011 PPP US dollars (on the horizontal axis). PPP = purchasing power parity. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 179 Figure C3.3: Sex-disaggregated harmonized test scores, relative to GDP per capita 600 Sex-disaggregated harmonized test scores, circa 2020 Finland Ireland Switzerland 500 Cyprus Hungary Greece Seychelles China Bulgaria Iran, Islamic Rep. Thailand Oman Ecuador Senegal 400 India Indonesia Morocco Kuwait Congo, Rep. Ethiopia South Africa Chad 300 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots sex-disaggregated harmonized test scores. The solid dot indicates the national average, the triangle shows the average value for girls, and the horizontal line shows the average value for boys. PPP = purchasing power parity. Figure C3.4: Harmonized test scores, by income group and region a. Income groups Low income 356 Lower-middle income 392 Upper-middle income 411 High income 487 0 100 200 300 400 500 Harmonized test scores, circa 2020 b. Regions South Asia 374 Sub-Saharan Africa 374 Latin America and the Caribbean 405 Middle East and North Africa 407 East Asia and Pacific 432 Europe and Central Asia 479 North America 523 0 100 200 300 400 500 Harmonized test scores, circa 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figures plot regional and income-group average values for harmonized test scores. 180 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES 4. UNDER-5 STUNTING RATES The fraction of children under 5 not stunted is calculated as the complement of the under-5 stunting rate. The stunting rate is defined as the share of children under the age of 5 whose height is more than two reference standard deviations below the refer- ence median for their ages. The reference median and standard deviations are set by the World Health Organization (WHO) for normal healthy child development (World Health Organization 2009). Child-level stunting prevalence is averaged across the rel- evant 0–5 age range to arrive at an overall under-5 stunting rate. The stunting rate is used as a proxy for latent health of the population, in addition to the adult survival rate, in countries for which stunting data are available. Data on stunting rates are taken from the Joint Child Malnutrition Estimates ( JME) database,19 managed by the United Nations Children’s Fund (UNICEF), WHO, and the World Bank (see UNICEF, World Health Organization, and World Bank 2020). The data- base reports the prevalence of stunting, wasting, and underweight, and is populated with estimates from survey data, gray literature, and reports from national authorities, reviewed by the JME interagency team. If required, data are reanalyzed to produce nationally representative estimates for the appropriate age cohort (0–5 years), compa- rable across economies and across time. Surveys presenting anthropometric data for age groups other than 0–59 months or 0–60 months are adjusted using national survey results—gathered as close in time as possible—from the same economy that include the age range 0–59/60 months. National rural estimates are adjusted similarly using another national survey for the same economy as close in time as possible with avail- able national urban and rural data to derive an adjusted national estimate. Historical data that use different growth reference standards are reanalyzed to produce estimates based on WHO standards when raw data are available. If raw data are unavailable, esti- mates are converted to WHO-based prevalence using an algorithm developed by Yang and de Onis (2008). The JME reports stunting rates from surveys and administrative data and is updated twice a year, in March and September. The HCI team supplements stunting data from the JME with data provided by country teams for five countries: Bhutan, Chile, Fiji, Indonesia, and Timor-Leste. It does so primarily to include more recent surveys that have not yet been incorporated in the JME. country-year The March 2020 update of the JME reports data for 152 economies and 887 ­ observations. About 50 percent of the JME data come from the Demographic and Health Surveys (DHS) and Multiple Indicator Cluster Surveys (MICS). Both are nationally rep- resentative household surveys that collect data on measures of population, health, and nutrition.20 About 10 percent of JME data come from country nutrition surveillance programs, whereas the rest of the database is populated using national surveys that collect anthropometric data and measure stunting directly. The JME database reports sex-disaggregated stunting rates for 56 percent of the surveys. It also reports 95 percent confidence intervals around estimates of stunting for about T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 181 40 percent of the observations, primarily those on which the JME team had access to record-level survey data. Absent better alternatives, the HCI team imputes confidence intervals for the remaining observations in the JME database using the fitted values from a regression of the width of the confidence interval on the stunting rate. Surveys from low- and middle-income economies make up 90 percent of the JME database. High-income economies tend to have much lower average stunting rates (the national average for the 13 high-income economies in the JME sample is 6 percent) and are less likely to regularly monitor stunting through frequent surveys. Some high-­ income economies like Kuwait, Oman, and the United States continue frequent mon- itoring of stunting prevalence through national surveys. Inconsistent measurement is of greater concern in middle- or low-income economies where stunting rates continue to be elevated. The most recent survey for 33 economies in the JME database is more than 5 years old, and it is about 10 years old for 10 economies. Conversely, economies like Peru and Senegal elected to field DHS annually. The continuous DHS played a key role in Peru’s national strategy for early childhood development, Crecer, which helped reduce the country’s rate of chronic malnutrition from 28 percent in 2005 to 13 percent in 2016, with an even pace of change among rural and urban children (Marini and Rokx 2017). In the JME data, the average gap between surveys for economies with at least two surveys is 5.6 years, and it is 5.0 years when high-income economies are excluded. 2020 update Stunting rates for the 2020 update of the HCI come from the March 2020 update of the JME database, available at the WHO website.21 Relative to the 2018 edition of the HCI, this latest update to the database allows us to update stunting rates for 54 economies and to add stunting rates for Argentina, Bulgaria, and Uzbekistan, which did not have rates in the previous iteration of the HCI. Stunting rates for the 2020 HCI come from the most recently available survey as of 2019, and data for the back-calculated 2018 HCI come from the most recent survey available as of 2017. Data for the baseline comparator year of 2010 are populated for each economy using the survey closest to 2010 that was fielded between 2005 and 2015. When  populating the 2010 cross-section, we ensure a minimum gap of five years between the survey used to populate the 2010 and 2020 cross-sections. To maximize the overlap among the three cross-sections, we do not rely on stunting rates in the cal- culation of the HCI for high-income economies, even when stunting data are available for some of these economies. This is because stunting rates typically come from surveys that are 5–10 years old for these economies. Further, to ensure consistency across time periods, we use stunting data to calculate the HCI for an economy only if such data are available in both 2010 and 2020. This does not prevent the calculation of an HCI score for high-income economies or those economies missing stunting data in any period; we simply use the adult survival rate as the proxy for latent health in our calculations. Values for stunting rates used to produce the back-calculated 2018 HCI are very similar to those used in the previous iteration of the HCI, as illustrated in figure C4.1, where data from the two vintages align almost perfectly along the 45-degree line. The figure 182 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES highlights eight economies where stunting rates have changed by 3 percentage points or more in the back-calculated 2018 HCI versus the original 2018 HCI. This change is predominantly because the March 2020 update of the JME makes a more recent sur- vey available or, in the case of Sierra Leone, because JME estimates have been updated following a reanalysis of survey data (see table C4.1). Figure C4.2 reports the most recent cross-section of stunting rates used to calculate the 2020 HCI. Stunting ranges from about 2.5 percent in the richest economies in the sample to about 54 percent in the poorest economies. The levels of stunting tend to be slightly lower for girls than for boys, as reported in figure C4.3. In the figure, the solid dot indicates the economy average, the triangle indi- cates the average for girls, and the horizontal bar indicates the average for boys. The average stunting rate is 24 percent for boys, compared with 22 percent for girls. Figure C4.4 reports average stunting rates by income group and by World Bank region. Levels tend to be highest in low-income economies, and regional averages are highest in Sub-Saharan Africa and South Asia. Figure C4.1: Comparing original and back-calculated 2018 stunting rates 0.6 Timor-Leste Stunting rates, 2018 back-calculated 0.4 India Sierra Leone Togo Burkina Faso 0.2 Tajikistan Albania Mongolia 0 0.2 0.4 0.6 Stunting rates, 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots the stunting rates as used in the 2018 HCI (on the horizontal axis), and the stunting rates used for the back-calculated 2018 HCI (on the vertical axis). Economies where stunting rates have changed by 3 percentage points between 2018 and back-calculated 2018 are labeled. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 183 Table C4.1: Source data for economies with different values in 2018 and back-calculated 2018 2018 vintage 2020 vintage Economy Source Year Value Source Year Value Albania DHS 2009 0.23 DHS 2017 0.11 Burkina Faso SMART 2016 0.27 SMART 2017 0.21 India DHS 2015 0.38 NNS 2017 0.35 Mongolia MICS 2013 0.11 NNS 2016 0.07 Sierra Leone MICS 2017 0.26 MICS 2017 0.31 Tajikistan DHS 2012 0.27 DHS 2017 0.18 Timor-Leste Food and Timor-Leste Timor-Leste Nutrition Survey, Final 2013 0.50 Demographic and Health 2016 0.46 Report 2015 Survey 2016 Togo DHS 2014 0.28 MICS 2017 0.24 Source: World Bank calculations based on World Bank 2018 and the 2020 update of the Human Capital Index (HCI). Note: DHS = Demographic and Health Surveys; MICS = Multiple Indicator Cluster Surveys; NNS = National Nutrition Survey; SMART = Standardized Monitoring and Assessment of Relief and Transition. Figure C4.2: Stunting rates, Human Capital Index 2020, relative to GDP per capita 0.6 Stunting rate, circa 2020 0.4 Chad Ethiopia India Indonesia South Africa Ecuador Congo, Rep. 0.2 Senegal Morocco Thailand China Bulgaria 0 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots stunting rates (on the vertical axis) against log GDP per capita at 2011 PPP US dollars (on the horizontal axis). PPP = purchasing power parity. 184 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C4.3: Sex-disaggregated stunting rates, relative to GDP per capita 0.6 Sex-disaggregated stunting rate, circa 2020 0.4 Chad Ethiopia India Indonesia South Africa Ecuador Congo, Rep. 0.2 Senegal Morocco Thailand China Bulgaria 0 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots sex-disaggregated stunting rates. The solid dot indicates the national average, the triangle shows the average value for girls, and the horizontal line shows the average value for boys. PPP = purchasing power parity Figure C4.4: Stunting rates, by income group and region a. Income groups Low income 0.35 Lower-middle income 0.25 Upper-middle income 0.13 0 0.1 0.2 0.3 0.4 Stunting rate, circa 2020 b. Regions Sub-Saharan Africa 0.31 South Asia 0.31 East Asia and Pacific 0.24 Middle East and North Africa 0.18 Latin America and the Caribbean 0.15 Europe and Central Asia 0.10 0 0.1 0.2 0.3 0.4 Stunting rate, circa 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figures plot regional and income-group average values for stunting. High-income-group economies are not included in these figures. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 185 5. ADULT SURVIVAL RATES The adult survival rate is calculated as the complement of the mortality rate for 15- to 60-year-olds. The mortality rate for 15- to 60-year-olds is the probability that a 15-year- old in a specified year will die before reaching the age of 60, if subject to current age-specific mortality rates. It is frequently expressed as a rate per 1,000 alive at 15, in which case it must be divided by 1,000 to obtain the probability that a 15-year-old will die before age 60. Adult mortality rates are estimated on the basis of prevailing patterns of death rates by age and are reported by the United Nations Population Division (UNPD) for five-year periods. The five-year data are interpolated to arrive at annual estimates to calculate the HCI. The measurement of adult survival rates requires data on death rates by age. Although they are readily available in economies with strong vital registries, such data are missing or incomplete in roughly the poorest quarter of economies. In these econ- omies, UNPD estimates death rates by age by linking the limited available age-specific mortality data with model life tables that capture the typical pattern in the distribution of deaths by age. UNPD does not individually report adult mortality rates for economies with fewer than 90,000 inhabitants. For this reason, data from UNPD are supplemented with adult mor- tality rates from the Global Burden of Disease (GBD) project, managed by the Institute of Health Metrics and Evaluation (IHME). Data from this source are used for Dominica and the Marshall Islands. Data for Nauru, Palau, San Marino, St. Kitts and Nevis, and Tuvalu come from WHO. Despite uncertainty on the primary estimates of mortality as well as the process for data modeling, uncertainty intervals are not reported in the UNPD data. Here we use uncertainty intervals reported in the GBD modeling process for adult survival rates.22 The point estimates for adult survival rates in these two datasets are quite similar for most economies. The ratio of the upper (lower) bound to the point estimate of the adult survival rate in the GBD data is applied to the point estimate of the adult survival rate in the UNPD and WHO data to obtain upper (lower) bounds. 2020 update Adult mortality rates for the 2020 update of the HCI come from the 2019 update of the UNPD World Population Prospects estimates, available at the World Population Prospects website.23 The GBD data come from the 2017 update—the most recent available—and can be retrieved from the IHME data visualization site.24 The WHO data ­ are located on the UN Data platform.25 Data for five-year periods from the UNPD are interpolated to arrive at annual estimates. Data from the GBD and WHO are carried for- ward up to 10 years to fill gaps in the series. UNPD adult mortality rates for the 2020 HCI calculated come from the most recent available year, as of 2019, and data for the back-­ 2018 HCI  come from 2017. Data for the comparator year of 2010 come from 2010. 186 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES For economies with data from the GBD, the latest data from 2017 are used to populate the 2020 and back-calculated 2018 rates. For economies with data from WHO, the most recent estimate to populate the 2020 and back-calculated 2018 rates comes from 2012. Because adult mortality rates are estimated by modeling all available data on adult mortality from vital registration systems, population censuses, household surveys, and sample registration systems combined with model life tables, every new release of data from UNPD and GBD updates estimates for all the previous years in the time series. As a result, data for the same year might differ slightly across updates. Values for adult mortality rates used to produce the back-calculated 2018 HCI are simi- lar to those used in the previous iteration of the HCI, as illustrated in figure C5.1, where data from the two vintages align closely along the 45-degree line for most economies. The figure highlights the 10 economies where adult mortality rates have changed by 30 deaths or more per 1,000 15-year-olds. The largest changes were for Angola (which went from 236 to 279 deaths per 1,000 15-year-olds) and Kazakhstan (which went from 203 to 158 deaths per 1,000 15-year-olds). Figure C5.1: Comparing original and back-calculated 2018 adult mortality rates 0.5 Central African Republic 0.4 Adult mortality rates, 2018 back-calculated Côte d’Ivoire 0.3 Angola Uganda Fiji 0.2 Tonga Kazakhstan Mexico Belarus Brunei Darussalam 0.1 0 0.1 0.2 0.3 0.4 0.5 Adult mortality rates, 2018 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots adult rates as used in the 2018 HCI (on the horizontal axis), and the adult mortality rates used for the back-calculated 2018 HCI (on the vertical axis). Economies where adult mortality rates have changed by 30 deaths or more per 1,000 15-year-olds between 2018 and back-calculated 2018 are labeled. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 187 Figure C5.2 reports the most recent cross-section of adult mortality rates used to calcu- late the 2020 HCI. Rates range from about 0.039 (39 deaths per 1,000 15-year-olds) in the richest economies to about 0.477 (477 deaths per 1,000 15-year-olds) in the poorest economies. figure C5.3. Adult mortality rates tend to be lower for women than for men, as reported in ­ In the figure, the solid dot indicates the country average, the triangle indicates the aver- age for women, and the horizontal bar indicates the average for men. The average adult mortality rate for men was 0.183 (183 deaths per 1,000 15-year-olds), compared to 0.120 for women (120 deaths per 1,000 15-year-olds). Figure C5.4 reports average adult mortality rates by income group and by World Bank region. Mortality rates tend to be highest in low-income ecomomies, and regional aver- ages are highest in Sub-Saharan Africa and South Asia, reflecting that poor economies continue to bear a disproportionate burden of adult mortality. Figure C5.2: Adult mortality rates, Human Capital Index 2020, relative to GDP per capita 0.5 0.4 Adult mortality rate, circa 2020 Chad South Africa 0.3 Congo, Rep. Ethiopia 0.2 Senegal India Indonesia Seychelles Bulgaria Ecuador Thailand Hungary 0.1 Oman China Finland Morocco Iran, Islamic Rep. Greece Kuwait Cyprus Ireland Switzerland 0 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots adult mortality rates (on the vertical axis) against log GDP per capita at 2011 PPP US dollars (on the horizontal axis). PPP = purchasing power parity. 188 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Figure C5.3: Sex-disaggregated adult mortality rates, relative to GDP per capita 0.6 Sex-disaggregated adult mortality rate, circa 2020 0.4 Chad South Africa Congo, Rep. 0.2 Ethiopia Senegal India Indonesia Seychelles Bulgaria Ecuador Thailand Hungary Oman China Morocco Iran, Islamic Rep.Greece Finland Kuwait Cyprus Ireland Switzerland 0 6 8 10 12 Log GDP per capita at PPP, circa 2020 East Asia and Pacific Europe and Central Asia Latin America and the Caribbean Middle East and North Africa North America South Asia Sub-Saharan Africa Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figure plots sex-disaggregated adult mortality rates. The solid dot indicates the national average, the triangle is used to show the average value for women, and the horizontal line shows the average value for men. Figure C5.4: Adult mortality rates, by income group and region a. Income groups Low income 0.25 Lower-middle income 0.20 Upper-middle income 0.14 High income 0.08 0 0.05 0.10 0.15 0.20 0.25 Adult mortality rate, circa 2020 b. Regions Sub-Saharan Africa 0.26 South Asia 0.16 Latin America and Caribbean 0.14 East Asia and Pacific 0.14 Europe and Central Asia 0.10 Middle East and North Africa 0.09 North America 0.09 0 0.05 0.10 0.15 0.20 0.25 Adult mortality rate, circa 2020 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: The figures plot regional and income group average values for adult mortality rates. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 189 WORLD BANK–WIDE DATA REVIEW PROCESS AND 6.  QUALITY ASSESSMENT Component data for the HCI 2020 update were subject to extensive World Bank- wide data review to ensure data quality. The review process was conducted between February and July 2020, and was split into two parts. The first part of the process (February to May 2020) focused on the enrollment data used to construct estimates of expected years of school and was conducted with World Bank Program Leaders for Human Development. The second part of the data review process (May to July 2020) focused on the other four index components—child mortality, harmonized test scores, stunting rates, and adult mortality. The enrollment data were validated separately because experience from the first edition of the HCI in 2018 suggested that those data required the most intensive review in terms of time and inputs needed from World Bank country teams, due to extensive gaps in the data as reported by UIS. All component data were reviewed for timeliness and completeness, with gaps filled and revisions made as needed. 7. SUPPLEMENTARY APPENDIX TABLES Table C7.1: Data sources for every education level for economies with an absolute change in EYS of at least 0.5, original 2018 and back-calculated 2018 2018 2018 back-calculated Economy Level EYS Rate Year Source EYS Rate Year Source Preprimary 24.9 2016 UIS (ANER) 61.3 2017 UIS (ANER) Primary 99.1 2016 UIS (TNER) 97.9 2017 UIS (TNER) Azerbaijan Lower-secondary 11.6 93.8 2016 UIS (TNER) 12.4 99.4 2017 UIS (TNER) WB Staff WB Staff Upper-secondary 77.4 2017 77.5 2017 (NER) (NER) Preprimary 59.7 2011 UIS (ANER) 41.7 2017 UIS (GER) WB Staff Primary 92.8 2016 UIS (TNER) 92.5 2017 Bangladesh 11.0 10.2 (NER) Lower-secondary 86.8 2016 UIS (TNER) 69.2 2017 UIS (ANER) Upper-secondary 55.4 2016 UIS (TNER) 57.2 2016 UIS (TNER) Preprimary 95.4 2016 UIS (ANER) 84.0 2017 UIS (ANER) Primary 93.4 2016 UIS (TNER) 88.2 2017 UIS (TNER) Bulgaria 12.9 12.3 Lower-secondary 90.7 2016 UIS (TNER) 87.6 2017 UIS (TNER) Upper-secondary 89.5 2016 UIS (TNER) 90.3 2017 UIS (TNER) WB Staff Preprimary 7.9 2012 4.0 2013 UIS (NER) (ANER) WB Staff WB Staff Primary 72.1 2014 72.1 2014 Congo, (TNER) (TNER) 9.2 8.5 Dem. Rep. WB Staff WB Staff Lower-secondary 81.9 2014 81.9 2014 (TNER) (TNER) WB Staff WB Staff Upper-secondary 74.9 2014 53.0 2014 (TNER) (TNER) (continued next page) 190 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C7.1: Data sources for every education level for economies with an absolute change in EYS of at least 0.5, original 2018 and back-calculated 2018 (Continued) 2018 2018 back-calculated Economy Level EYS Rate Year Source EYS Rate Year Source Preprimary 21.2 2016 UIS (ANER) 22.2 2017 UIS (ANER) WB Staff Primary 60.1 2012 79.0 2017 UIS (TNER) (TNER) Côte d’Ivoire 7.0 WB Staff 7.6 Lower-secondary 61.5 2013 49.6 2017 UIS (TNER) (TNER) WB Staff Upper-secondary 39.0 2013 31.7 2017 UIS (TNER) (TNER) Preprimary 77.1 2016 UIS (ANER) 87.4 2017 UIS (ANER) Dominican Primary 84.9 2016 UIS (TNER) 90.2 2017 UIS (TNER) 11.3 11.9 Republic Lower-secondary 85.5 2016 UIS (TNER) 87.2 2017 UIS (TNER) Upper-secondary 69.8 2016 UIS (TNER) 71.7 2017 UIS (TNER) Preprimary 17.0 2011 UIS (ANER) 18.9 2011 UIS (ANER) WB Staff Primary 64.4 2015 UIS (TNER) 81.4 2017 (NER) Eswatini 8.2 6.4 WB Staff Lower-secondary 75.8 2015 UIS (TNER) 28.3 2017 (NER) WB Staff Upper-secondary 55.9 2015 UIS (TNER) 10.4 2015 (NER) Preprimary 100.0 2015 UIS (GER) 98.8 2017 UIS (ANER) Primary 99.4 2015 UIS (TNER) 99.0 2017 UIS (TNER) Germany 13.9 13.3 Lower-secondary 97.5 2015 UIS (GER) 92.9 2017 UIS (TNER) Upper-secondary 100.0 2015 UIS (GER) 87.6 2017 UIS (TNER) WB Staff Preprimary 12.9 2016 UIS (GER) 13.7 2018 (ANER) WB Staff Primary 97.2 2013 UIS (TNER) 88.7 2018 (TNER) India 10.2 10.8 WB Staff Lower-secondary 84.9 2013 UIS (TNER) 62.2 2018 (TNER) WB Staff Upper-secondary 51.0 2013 UIS (TNER) 30.3 2018 (TNER) Preprimary 36.0 2016 UIS (ANER) 42.4 2016 UIS (ANER) Primary 75.9 2016 UIS (TNER) 88.6 2017 UIS (TNER) Lesotho 8.7 10.0 Lower-secondary 64.3 2016 UIS (TNER) 70.1 2016 UIS (TNER) Upper-secondary 51.3 2016 UIS (TNER) 59.2 2016 UIS (TNER) Preprimary 26.1 2016 UIS (NER) 60.7 2018 UIS (ANER) Primary 100.0 2016 UIS (GER) 72.5 2018 UIS (TNER) Madagascar 7.5 8.4 Lower-secondary 22.8 2016 UIS (ANER) 63.5 2018 UIS (TNER) Upper-secondary 8.7 2016 UIS (ANER) 31.1 2018 UIS (TNER) WB Staff Preprimary 5.5 2008 10.5 2015 UIS (GER) (ANER) Mauritania Primary 6.3 65.5 2016 UIS (TNER) 7.4 76.5 2017 UIS (TNER) Lower-secondary 45.4 2016 UIS (TNER) 52.3 2017 UIS (TNER) Upper-secondary 28.5 2016 UIS (TNER) 32.9 2017 UIS (TNER) (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 191 Table C7.1: Data sources for every education level for economies with an absolute change in EYS of at least 0.5, original 2018 and back-calculated 2018 (Continued) 2018 2018 back-calculated Economy Level EYS Rate Year Source EYS Rate Year Source Preprimary 88.3 2010 UIS (ANER) 55.6 2010 UIS (NER) Primary 90.6 2010 UIS (TNER) 88.7 2010 UIS (TNER) Nicaragua 11.6 10.8 Lower-secondary 82.3 2010 UIS (TNER) 81.8 2010 UIS (TNER) Upper-secondary 63.5 2010 UIS (TNER) 62.4 2010 UIS (TNER) Preprimary 41.8 2010 UIS (GER) 41.8 2010 UIS (GER) Primary 65.9 2010 UIS (TNER) 66.0 2010 UIS (TNER) Nigeria 8.2 7.3 Lower-secondary 52.5 2013 UIS (GER) 45.0 2016 UIS (GER) Upper-secondary 60.3 2013 UIS (GER) 38.6 2016 UIS (GER) WB Staff Preprimary 57.6 2016 UIS (NER) 43.4 2017 (ANER) WB Staff Primary 82.1 2016 UIS (TNER) 77.8 2017 (TNER) Pakistan 8.8 9.3 WB Staff Lower-secondary 53.8 2016 UIS (TNER) 71.4 2017 (TNER) WB Staff Upper-secondary 37.8 2016 UIS (TNER) 55.7 2017 (TNER) Preprimary 78.9 2015 UIS (ANER) 75.6 2017 UIS (ANER) Primary 87.4 2015 UIS (TNER) 84.0 2017 UIS (TNER) Panama 11.3 10.7 Lower-secondary 84.5 2015 UIS (TNER) 84.2 2017 UIS (TNER) Upper-secondary 66.1 2015 UIS (TNER) 55.6 2017 UIS (TNER) Preprimary 98.6 2008 UIS (GER) 71.4 2016 UIS (ANER) Papua New Primary 85.4 2012 UIS (TNER) 84.4 2016 UIS (TNER) 8.2 10.3 Guinea Lower-secondary 15.6 2012 UIS (NER) 77.1 2016 UIS (TNER) Upper-secondary 22.0 2012 UIS (GER) 50.4 2016 UIS (TNER) Preprimary 96.5 2016 UIS (ANER) 97.4 2017 UIS (ANER) Primary 99.8 2016 UIS (TNER) 97.2 2017 UIS (TNER) Seychelles 13.7 13.0 Lower-secondary 92.0 2016 UIS (ANER) 91.8 2017 UIS (ANER) Upper-secondary 99.9 2016 UIS (TNER) 83.7 2017 UIS (TNER) WB Staff Preprimary 65.4 2015 UIS (ANER) 55.7 2017 (ANER) Primary 75.0 2016 UIS (TNER) 92.9 2017 UIS (TNER) Solomon 9.2 8.7 WB Staff Islands Lower-secondary 69.0 2007 UIS (TNER) 37.0 2017 (NER) WB Staff Upper-secondary 44.8 2007 UIS (TNER) 28.4 2017 (NER) Preprimary 21.9 2015 UIS (NER) 14.9 2015 UIS (NER) WB Staff Primary 83.1 2015 UIS (TNER) 92.8 2017 (NER) South Africa 9.3 10.2 Lower-secondary 71.1 2015 UIS (TNER) 72.1 2017 UIS (TNER) WB Staff Upper-secondary 58.4 2015 UIS (TNER) 70.8 2017 (NER) (continued next page) 192 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C7.1: Data sources for every education level for economies with an absolute change in EYS of at least 0.5, original 2018 and back-calculated 2018 (Continued) 2018 2018 back-calculated Economy Level EYS Rate Year Source EYS Rate Year Source Preprimary 45.0 2014 UIS (ANER) 54.7 2017 UIS (ANER) WB Staff Primary 91.5 2017 80.7 2017 UIS (TNER) (TNER) Tanzania 7.8 WB Staff 7.2 Lower-secondary 38.7 2017 27.8 2016 UIS (TNER) (GER) WB Staff Upper-secondary 6.9 2017 14.2 2016 UIS (TNER) (GER) Preprimary 57.3 2016 UIS (ANER) 43.2 2017 UIS (ANER) Primary 68.9 2016 UIS (TNER) 82.4 2017 UIS (TNER) Timor-Leste 9.9 10.6 Lower-secondary 85.9 2016 UIS (TNER) 85.0 2017 UIS (TNER) Upper-secondary 66.6 2016 UIS (TNER) 74.3 2017 UIS (TNER) Preprimary 38.5 2014 UIS (GER) 45.8 2015 UIS (GER) Primary 98.7 2014 UIS (TNER) 98.9 2015 UIS (TNER) Tonga 10.9 11.6 Lower-secondary 96.4 2014 UIS (TNER) 95.1 2015 UIS (TNER) Upper-secondary 43.3 2014 UIS (TNER) 62.0 2015 UIS (TNER) Preprimary 55.9 2015 UIS (NER) 49.7 2015 UIS (NER) Primary 83.6 2015 UIS (TNER) 77.8 2015 UIS (TNER) Vanuatu 10.6 10.1 Lower-secondary 95.3 2015 UIS (TNER) 92.8 2015 UIS (TNER) Upper-secondary 54.9 2015 UIS (TNER) 55.5 2015 UIS (TNER) Preprimary 89.6 2016 UIS (ANER) 99.7 2017 UIS (ANER) WB Staff Primary 96.2 2016 97.9 2017 UIS (TNER) (NER) Vietnam 12.3 WB Staff 12.8 Lower-secondary 89.7 2016 96.7 2017 UIS (NER) (NER) WB Staff WB Staff Upper-secondary 68.1 2016 68.1 2016 (NER) (NER) Preprimary 64.7 2015 UIS (ANER) 64.4 2017 UIS (ANER) West Bank Primary 92.4 2016 UIS (TNER) 97.4 2017 UIS (TNER) 11.4 12.0 and Gaza Lower-secondary 87.7 2016 UIS (TNER) 92.7 2017 UIS (TNER) Upper-secondary 63.5 2016 UIS (TNER) 68.1 2017 UIS (TNER) Preprimary 36.4 2013 UIS (ANER) 40.7 2013 UIS (ANER) Primary 87.9 2013 UIS (TNER) 97.6 2013 UIS (TNER) Zimbabwe 10.0 11.1 Lower-secondary 86.9 2013 UIS (TNER) 93.9 2013 UIS (TNER) Upper-secondary 46.7 2013 UIS (TNER) 52.2 2013 UIS (TNER) Source: World Bank calculations based on World Bank 2018 and the 2020 update of the Human Capital Index. Note: ANER = adjusted net enrollment rate; EYS = expected years of school; GER = gross enrollment rate; NER = net enrollment rate; TNER = total net enrollment rate; UIS = United Nations Educational, Scientific and Cultural Organization’s Institute of Statistics; WB = World Bank. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 193 Table C7.2: Data sources for every level of schooling for economies with a decrease in EYS between 2010 and 2020 2010 2020 Economy Level EYS Rate Year Source EYS Rate Year Source Preprimary 96.9 2010 UIS (ANER) 100.0 2017 UIS (ANER) Primary 96.5 2010 UIS (TNER) 97.0 2017 UIS (TNER) Austria 13.5 13.4 Lower-secondary 98.1 2010 UIS (TNER) 97.5 2017 UIS (TNER) Upper-secondary 93.1 2010 UIS (TNER) 89.4 2017 UIS (TNER) Preprimary 93.2 2010 UIS (ANER) 84.0 2017 UIS (ANER) Primary 99.2 2010 UIS (TNER) 88.2 2017 UIS (TNER) Bulgaria 12.9 12.3 Lower-secondary 88.6 2010 UIS (TNER) 87.6 2017 UIS (TNER) Upper-secondary 81.1 2010 UIS (TNER) 90.3 2017 UIS (TNER) Preprimary 99.1 2010 UIS (ANER) 93.7 2017 UIS (ANER) Primary 98.6 2010 UIS (TNER) 98.9 2017 UIS (TNER) Denmark 13.4 13.4 Lower-secondary 99.2 2010 UIS (TNER) 98.4 2017 UIS (TNER) Upper-secondary 85.3 2010 UIS (TNER) 87.9 2017 UIS (TNER) Preprimary 96.8 2010 UIS (ANER) 98.8 2017 UIS (ANER) Primary 97.2 2010 UIS (TNER) 99.0 2017 UIS (TNER) Germany 13.3 13.3 Lower-secondary 94.5 2010 UIS (TNER) 92.9 2017 UIS (TNER) Upper-secondary 91.4 2010 UIS (TNER) 87.6 2017 UIS (TNER) Preprimary 94.4 2010 UIS (ANER) 92.7 2017 UIS (ANER) Primary 96.4 2010 UIS (TNER) 98.0 2017 UIS (TNER) Greece 13.4 13.3 Lower-secondary 94.9 2010 UIS (TNER) 92.5 2017 UIS (TNER) Upper-secondary 95.4 2010 UIS (TNER) 92.5 2017 UIS (TNER) Preprimary 85.5 2010 UIS (ANER) 85.1 2018 UIS (ANER) Primary 84.1 2011 UIS (TNER*) 81.3 2018 UIS (TNER) Guatemala 10.3 9.7 Lower-secondary 78.9 2010 UIS (TNER) 63.8 2018 UIS (TNER) Upper-secondary 38.3 2010 UIS (TNER) 40.6 2018 UIS (TNER) Preprimary 94.2 2010 UIS (ANER) 87.1 2017 UIS (ANER) Primary 96.5 2010 UIS (TNER) 96.2 2017 UIS (TNER) Hungary 13.0 13.0 Lower-secondary 93.0 2010 UIS (TNER) 94.7 2017 UIS (TNER) Upper-secondary 85.7 2010 UIS (TNER) 86.8 2017 UIS (TNER) Preprimary 99.6 2010 UIS (ANER) 93.9 2017 UIS (ANER) Primary 99.3 2010 UIS (TNER) 97.2 2017 UIS (TNER) Italy 13.6 13.3 Lower-secondary 95.2 2010 UIS (TNER) 95.9 2017 UIS (TNER) Upper-secondary 93.2 2010 UIS (TNER) 88.9 2017 UIS (TNER) WB Staff WB Staff Preprimary 95.6 2010 91.1 2015 (ANER) (ANER) WB Staff WB Staff Primary 99.4 2010 98.8 2015 (TNER) (TNER) Japan 13.7 13.6 WB Staff WB Staff Lower-secondary 99.7 2010 99.9 2015 (TNER) (TNER) WB Staff WB Staff Upper-secondary 94.7 2010 96.4 2015 (TNER) (TNER) (continued next page) 194 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C7.2: Data sources for every level of schooling for economies with a decrease in EYS between 2010 and 2020 (Continued) 2010 2020 Economy Level EYS Rate Year Source EYS Rate Year Source WB Staff WB Staff Preprimary 37.9 2010 36.5 2018 (NER) (NER) WB Staff WB Staff Primary 97.2 2010 92.4 2018 (NER) (NER) Jordan 11.8 11.1 WB Staff WB Staff Lower-secondary 96.2 2010 92.4 2018 (NER) (NER) WB Staff WB Staff Upper-secondary 78.8 2010 70.1 2018 (NER) (NER) WB Staff Preprimary 96.9 2013 95.9 2017 UIS (ANER) (ANER**) Korea, Rep. Primary 13.7 99.8 2010 UIS (TNER) 13.6 97.6 2017 UIS (TNER) Lower-secondary 99.8 2010 UIS (TNER) 94.4 2017 UIS (TNER) Upper-secondary 92.2 2010 UIS (TNER) 99.7 2017 UIS (TNER) Preprimary 91.7 2012 UIS (ANER*) 81.2 2018 UIS (ANER) Primary 98.3 2010 UIS (TNER) 88.4 2018 UIS (TNER) Kuwait 12.7 12.0 Lower-secondary 92.7 2010 UIS (TNER) 92.1 2015 UIS (TNER) Upper-secondary 71.3 2010 UIS (TNER) 78.1 2015 UIS (TNER) Preprimary 95.1 2010 UIS (ANER) 98.2 2017 UIS (ANER) Primary 95.7 2010 UIS (TNER) 95.8 2017 UIS (TNER) Luxembourg 12.8 12.4 Lower-secondary 89.4 2010 UIS (TNER) 84.7 2017 UIS (TNER) Upper-secondary 81.8 2010 UIS (TNER) 72.2 2017 UIS (TNER) Preprimary 92.6 2010 UIS (ANER) 93.3 2018 UIS (ANER) Primary 91.6 2010 UIS (TNER) 91.0 2018 UIS (TNER) Moldova 12.0 11.8 Lower-secondary 87.5 2010 UIS (TNER) 85.0 2018 UIS (TNER) Upper-secondary 66.4 2010 UIS (TNER) 64.5 2018 UIS (TNER) Preprimary 76.8 2010 UIS (ANER) 75.6 2017 UIS (ANER) Primary 91.4 2010 UIS (TNER) 84.0 2017 UIS (TNER) Panama 11.3 10.7 Lower-secondary 83.3 2010 UIS (TNER) 84.2 2017 UIS (TNER) Upper-secondary 60.5 2010 UIS (TNER) 55.6 2017 UIS (TNER) Preprimary 77.9 2013 UIS (ANER*) 92.4 2018 UIS (ANER) Primary 97.5 2010 UIS (TNER) 96.8 2018 UIS (TNER) Qatar 12.9 12.8 Lower-secondary 96.6 2011 UIS (TNER*) 89.6 2018 UIS (TNER) Upper-secondary 86.8 2010 UIS (TNER) 83.0 2010 UIS (TNER) WB Staff WB Staff Preprimary 78.8 2013 83.9 2018 (ANER**) (ANER) WB Staff Primary 95.8 2010 UIS (TNER) 89.5 2018 (TNER) Romania 12.7 11.8 WB Staff Lower-secondary 91.7 2010 UIS (TNER) 84.9 2018 (TNER) WB Staff WB Staff Upper-secondary 86.3 2010 74.4 2017 (TNER) (TNER) (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 195 Table C7.2: Data sources for every level of schooling for economies with a decrease in EYS between 2010 and 2020 (Continued) 2010 2020 Economy Level EYS Rate Year Source EYS Rate Year Source Preprimary 84.7 2010 UIS (ANER) 82.3 2017 UIS (ANER) Slovak Primary 92.0 2010 UIS (TNER) 91.4 2017 UIS (TNER) 12.7 12.6 Republic Lower-secondary 93.2 2010 UIS (TNER) 93.5 2017 UIS (TNER) Upper-secondary 89.1 2010 UIS (TNER) 89.5 2017 UIS (TNER) Preprimary 11.1 2015 14.9 2015 UIS (NER) WB Staff WB Staff Primary 92.3 2010 93.1 2018 (NER) (NER) South Africa 10.2 10.2 Lower-secondary 79.0 2017 UIS (TNER*) 71.2 2017 UIS (TNER) WB Staff WB Staff Upper-secondary 69.7 2010 73.1 2018 (NER) (NER) WB Staff Preprimary 67.1 2013 67.6 2017 UIS (ANER) (ANER**) Turkey Primary 12.1 94.6 2010 UIS (TNER) 12.1 92.6 2017 UIS (TNER) Lower-secondary 96.6 2010 UIS (TNER) 90.4 2017 UIS (TNER) Upper-secondary 74.1 2010 UIS (TNER) 81.5 2017 UIS (TNER) Preprimary 99.1 2010 UIS (GER) 83.9 2013 UIS (GER) Primary 90.7 2010 UIS (TNER) 91.9 2014 UIS (TNER) Ukraine 13.1 12.9 Lower-secondary 94.8 2010 UIS (TNER) 96.3 2014 UIS (TNER) Upper-secondary 94.4 2010 UIS (TNER) 94.1 2014 UIS (TNER) Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Note: ANER = adjusted net enrollment rate; EYS = expected years of school; GER = gross enrollment rate; NER = net enrollment rate; TNER = total enrollment rate; UIS = United Nations Educational, Scientific and Cultural Organization’s Institute of Statistics; WB = World Bank. *interpolated using the same series; **interpolated using GER. 8. HCI AND COMPONENT DATA Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Afghanistan 0.94 8.9 355 5.1 0.79 0.62 0.40 0.39 – Albania 0.99 12.9 434 9.0 0.93 0.89 0.63 0.63 0.54 Algeria 0.98 11.8 374 7.1 0.91 0.88 0.53 0.53 0.53 Angola 0.92 8.1 326 4.2 0.73 0.62 0.36 0.36 – (continued next page) 196 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 (Continued) Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Antigua and 0.99 13.0 407 8.4 0.90 – 0.60 0.58 – Barbuda Argentina 0.99 12.9 408 8.4 0.89 0.92 0.60 0.62 0.59 Armenia 0.99 11.3 443 8.0 0.89 0.91 0.58 0.58 – Australia 1.00 13.6 516 11.2 0.95 – 0.77 0.78 0.75 Austria 1.00 13.4 508 10.9 0.94 – 0.75 0.77 0.74 Azerbaijan 0.98 12.4 416 8.3 0.88 0.82 0.58 0.63 0.50 Bahrain 0.99 12.8 452 9.3 0.93 – 0.65 0.66 0.60 Bangladesh 0.97 10.2 368 6.0 0.87 0.69 0.46 0.46 – Belarus 1.00 13.8 488 10.8 0.85 – 0.70 – – Belgium 1.00 13.5 517 11.2 0.93 – 0.76 0.76 0.75 Benin 0.91 9.2 384 5.7 0.77 – 0.40 0.40 0.37 Bhutan 0.97 10.2 387 6.3 0.81 0.79 0.48 – – Bosnia and 0.99 11.7 416 7.8 0.91 0.91 0.58 0.62 – Herzegovina Botswana 0.96 8.1 391 5.1 0.80 – 0.41 0.41 0.37 Brazil 0.99 11.9 413 7.9 0.86 – 0.55 0.55 0.53 Brunei 0.99 13.2 438 9.2 0.88 0.80 0.63 – – Darussalam Bulgaria 0.99 12.3 441 8.7 0.87 0.93 0.61 0.67 0.64 Burkina Faso 0.92 7.0 404 4.5 0.76 0.75 0.38 0.38 0.32 Burundi 0.94 7.6 423 5.2 0.72 0.46 0.39 0.39 0.34 Cambodia 0.97 9.5 452 6.8 0.84 0.68 0.49 0.49 – Cameroon 0.92 8.7 379 5.3 0.70 0.71 0.40 0.39 0.38 Canada 1.00 13.7 534 11.7 0.94 – 0.80 0.80 0.77 Central African 0.88 4.6 369 2.7 0.59 0.59 0.29 – – Republic Chad 0.88 5.3 333 2.8 0.65 0.60 0.30 0.30 0.29 Chile 0.99 13.0 452 9.4 0.92 – 0.65 0.67 0.63 China 0.99 13.1 441 9.3 0.92 0.92 0.65 0.65 – Colombia 0.99 12.9 419 8.6 0.89 0.87 0.60 0.60 0.58 Comoros 0.93 8.2 392 5.1 0.78 0.69 0.40 0.40 – Congo, Dem. 0.91 9.1 310 4.5 0.75 0.57 0.37 0.36 – Rep. Congo, Rep. 0.95 8.9 371 5.3 0.74 0.79 0.42 0.42 0.41 Costa Rica 0.99 13.1 429 9.0 0.92 – 0.63 0.60 0.60 (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 197 Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 (Continued) Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Croatia 1.00 13.4 488 10.4 0.92 – 0.71 0.73 0.69 Cyprus 1.00 13.6 502 10.9 0.95 – 0.76 0.75 0.69 Czech 1.00 13.6 512 11.1 0.92 – 0.75 0.76 0.73 Republic Côte d’Ivoire 0.92 8.1 373 4.8 0.66 0.78 0.38 0.37 0.30 Denmark 1.00 13.4 518 11.1 0.93 – 0.76 0.77 0.75 Dominica 0.96 12.4 404 8.0 0.86 – 0.54 0.55 – Dominican 0.97 11.9 345 6.6 0.84 0.93 0.50 0.51 – Republic Ecuador 0.99 12.9 420 8.7 0.88 0.76 0.59 0.60 0.53 Egypt, Arab 0.98 11.5 356 6.5 0.86 0.78 0.49 0.49 0.48 Rep. El Salvador 0.99 10.9 436 7.6 0.82 0.86 0.55 0.54 – Estonia 1.00 13.5 543 11.7 0.90 – 0.78 0.77 0.73 Eswatini 0.95 6.4 440 4.5 0.60 0.74 0.37 0.37 0.31 Ethiopia 0.94 7.8 348 4.3 0.79 0.63 0.38 0.38 – Fiji 0.97 11.3 383 7.0 0.78 0.91 0.51 – – Finland 1.00 13.7 534 11.7 0.93 – 0.80 0.81 0.82 France 1.00 13.8 510 11.3 0.93 – 0.76 0.76 0.76 Gabon 0.96 8.3 456 6.0 0.79 0.83 0.46 0.46 – Gambia, The 0.94 9.5 353 5.4 0.75 0.81 0.42 0.40 0.37 Georgia 0.99 12.9 400 8.3 0.85 – 0.57 0.61 0.54 Germany 1.00 13.3 517 11.0 0.93 – 0.75 0.76 0.76 Ghana 0.95 12.1 307 6.0 0.77 0.82 0.45 0.44 – Greece 1.00 13.3 469 10.0 0.93 – 0.69 0.69 0.71 Grenada 0.98 13.1 395 8.3 0.85 – 0.57 0.54 – Guatemala 0.97 9.7 405 6.3 0.85 0.53 0.46 0.46 0.44 Guinea 0.90 7.0 408 4.6 0.76 0.70 0.37 0.37 – Guyana 0.97 12.2 346 6.8 0.77 0.89 0.50 0.49 – Haiti 0.94 11.4 338 6.1 0.78 0.78 0.45 0.44 – Honduras 0.98 9.6 400 6.1 0.86 0.77 0.48 0.48 – Hong Kong 0.99 13.5 549 11.9 0.95 – 0.81 0.82 0.78 SAR, China Hungary 1.00 13.0 495 10.3 0.88 – 0.68 0.71 0.69 Iceland 1.00 13.5 498 10.7 0.95 – 0.75 0.74 0.76 India 0.96 11.1 399 7.1 0.83 0.65 0.49 0.48 – (continued next page) 198 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 (Continued) Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Indonesia 0.98 12.4 395 7.8 0.85 0.72 0.54 0.54 0.50 Iran, Islamic 0.99 11.8 432 8.2 0.93 – 0.59 0.59 0.56 Rep. Iraq 0.97 6.9 363 4.0 0.84 0.87 0.41 0.40 – Ireland 1.00 13.9 521 11.6 0.94 – 0.79 0.81 0.77 Israel 1.00 13.8 481 10.6 0.95 – 0.73 0.76 0.72 Italy 1.00 13.3 493 10.5 0.95 – 0.73 0.75 0.75 Jamaica 0.99 11.4 387 7.1 0.86 0.94 0.53 0.54 – Japan 1.00 13.6 538 11.7 0.95 – 0.80 0.84 0.82 Jordan 0.98 11.1 430 7.7 0.89 – 0.55 0.55 0.56 Kazakhstan 0.99 13.7 416 9.1 0.84 0.92 0.63 0.78 0.59 Kenya 0.96 11.6 455 8.5 0.77 0.74 0.55 0.54 – Kiribati 0.95 11.2 411 7.4 0.81 – 0.49 0.47 – Korea, Rep. 1.00 13.6 537 11.7 0.94 – 0.80 0.83 0.82 Kosovo 0.99 13.2 374 7.9 0.91 – 0.57 0.57 – Kuwait 0.99 12.0 383 7.4 0.94 – 0.56 0.56 0.57 Kyrgyz 0.98 12.9 420 8.7 0.85 0.88 0.60 0.59 – Republic Lao PDR 0.95 10.6 368 6.3 0.82 0.67 0.46 0.46 – Latvia 1.00 13.6 504 11.0 0.84 – 0.71 0.74 0.68 Lebanon 0.99 10.2 390 6.3 0.93 – 0.52 0.52 – Lesotho 0.92 10.0 393 6.3 0.52 0.65 0.40 0.40 0.34 Liberia 0.93 4.2 332 2.2 0.78 0.70 0.32 0.32 – Lithuania 1.00 13.8 496 11.0 0.84 – 0.71 0.73 0.69 Luxembourg 1.00 12.4 493 9.8 0.94 – 0.69 0.69 0.70 Macao SAR, 0.99 12.9 561 11.6 0.96 – 0.80 0.76 0.65 China Madagascar 0.95 8.4 351 4.7 0.80 0.58 0.39 0.39 0.39 Malawi 0.95 9.6 359 5.5 0.74 0.61 0.41 0.41 0.36 Malaysia 0.99 12.5 446 8.9 0.88 0.79 0.61 0.63 0.58 Mali 0.90 5.2 307 2.6 0.75 0.73 0.32 0.32 – Malta 0.99 13.4 474 10.2 0.95 – 0.71 0.71 0.68 Marshall 0.97 9.4 375 5.7 0.70 0.65 0.42 0.40 – Islands Mauritania 0.92 7.7 342 4.2 0.80 0.77 0.38 0.37 – Mauritius 0.98 12.4 473 9.4 0.86 – 0.62 0.62 0.60 (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 199 Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 (Continued) Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Mexico 0.99 12.8 430 8.8 0.86 0.90 0.61 0.61 0.59 Micronesia, 0.97 11.8 380 7.2 0.84 – 0.51 0.47 – Fed. Sts. Moldova 0.98 11.8 439 8.3 0.84 0.94 0.58 0.58 0.56 Mongolia 0.98 13.2 435 9.2 0.80 0.91 0.61 0.62 – Montenegro 1.00 12.8 436 8.9 0.91 0.91 0.63 0.62 0.59 Morocco 0.98 10.4 380 6.3 0.93 0.85 0.50 0.49 0.47 Mozambique 0.93 7.6 368 4.5 0.68 0.58 0.36 0.36 – Myanmar 0.95 10.0 425 6.8 0.80 0.71 0.48 0.47 – Namibia 0.96 9.4 407 6.1 0.71 0.77 0.45 0.45 0.39 Nauru 0.97 11.7 347 6.5 0.93 – 0.51 – – Nepal 0.97 12.3 369 7.2 0.86 0.64 0.50 0.50 – Netherlands 1.00 13.9 520 11.5 0.95 – 0.79 0.80 0.80 New Zealand 0.99 13.7 520 11.4 0.94 – 0.78 0.77 0.78 Nicaragua 0.98 10.8 392 6.7 0.85 0.83 0.51 0.51 – Niger 0.92 5.5 305 2.7 0.77 0.52 0.32 0.32 – Nigeria 0.88 10.2 309 5.0 0.66 0.63 0.36 0.35 – North 0.99 11.0 414 7.3 0.91 0.95 0.56 0.54 0.54 Macedonia Norway 1.00 13.7 514 11.2 0.94 – 0.77 0.77 0.77 Oman 0.99 12.8 424 8.6 0.91 – 0.61 0.61 0.55 Pakistan 0.93 9.4 339 5.1 0.85 0.62 0.41 0.40 – Palau 0.98 11.7 463 8.7 0.87 – 0.59 0.57 – Panama 0.98 10.7 377 6.5 0.89 – 0.50 0.51 0.51 Papua New 0.95 10.3 363 6.0 0.78 0.51 0.43 0.42 – Guinea Paraguay 0.98 11.3 386 7.0 0.86 0.94 0.53 0.53 0.51 Peru 0.99 13.0 415 8.6 0.89 0.88 0.61 0.59 0.55 Philippines 0.97 12.9 362 7.5 0.82 0.70 0.52 0.55 – Poland 1.00 13.4 530 11.4 0.89 – 0.75 0.76 0.70 Portugal 1.00 13.9 509 11.3 0.93 – 0.77 0.78 0.74 Qatar 0.99 12.8 427 8.8 0.96 – 0.64 0.63 0.59 Romania 0.99 11.8 442 8.4 0.88 – 0.58 0.59 0.60 Russian 0.99 13.7 498 10.9 0.80 – 0.68 0.73 0.60 Federation Rwanda 0.96 6.9 358 3.9 0.81 0.62 0.38 0.38 – (continued next page) 200 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 (Continued) Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Samoa 0.98 12.2 370 7.2 0.89 0.95 0.55 0.52 – Saudi Arabia 0.99 12.4 399 7.9 0.92 – 0.58 0.58 0.55 Senegal 0.96 7.3 412 4.8 0.83 0.81 0.42 0.42 0.39 Serbia 0.99 13.3 457 9.8 0.89 0.94 0.68 0.76 0.65 Seychelles 0.99 13.1 463 9.7 0.85 – 0.63 0.63 0.57 Sierra Leone 0.89 9.6 316 4.9 0.63 0.71 0.36 0.35 – Singapore 1.00 13.9 575 12.8 0.95 – 0.88 0.89 0.85 Slovak 0.99 12.6 485 9.8 0.90 – 0.66 0.68 0.68 Republic Slovenia 1.00 13.6 521 11.4 0.93 – 0.77 0.79 0.75 Solomon 0.98 8.3 351 4.7 0.86 0.68 0.42 0.43 – Islands South Africa 0.97 10.2 343 5.6 0.69 0.73 0.43 0.42 0.43 South Sudan 0.90 4.7 336 2.5 0.68 0.69 0.31 0.31 – Spain 1.00 13.0 507 10.5 0.95 – 0.73 0.74 0.71 Sri Lanka 0.99 13.2 400 8.5 0.90 0.83 0.60 0.59 – St. Kitts and 0.99 13.0 409 8.5 0.88 – 0.59 0.57 – Nevis St. Lucia 0.98 12.7 418 8.5 0.87 0.98 0.60 0.59 – St. Vincent and the 0.98 12.3 391 7.7 0.83 – 0.53 0.54 – Grenadines Sudan 0.94 7.1 380 4.3 0.79 0.62 0.38 0.38 – Sweden 1.00 13.9 519 11.6 0.95 – 0.80 0.80 0.76 Switzerland 1.00 13.3 515 10.9 0.95 – 0.76 0.77 0.77 Tajikistan 0.97 10.9 391 6.8 0.87 0.82 0.50 0.54 – Tanzania 0.95 7.2 388 4.5 0.78 0.68 0.39 0.39 – Thailand 0.99 12.7 427 8.7 0.87 0.89 0.61 0.62 0.58 Timor-Leste 0.95 10.6 371 6.3 0.86 0.54 0.45 0.45 0.41 Togo 0.93 9.7 384 6.0 0.74 0.76 0.43 0.42 0.37 Tonga 0.98 11.6 386 7.1 0.83 0.92 0.53 0.52 – Trinidad and 0.98 12.4 458 9.1 0.85 – 0.60 0.60 0.55 Tobago Tunisia 0.98 10.6 384 6.5 0.91 0.92 0.52 0.51 0.53 Turkey 0.99 12.1 478 9.2 0.91 0.94 0.65 0.63 0.63 Tuvalu 0.98 10.8 346 6.0 0.79 – 0.45 0.44 – (continued next page) T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 201 Table C8.1: Human Capital Index and components: 2020, 2018 back-calculated, and 2010 (Continued) Components of HCI 2020 HCI Fraction of Learning- children Probability Expected adjusted Adult under HCI 2018 of survival years of Harmonized years of survival 5 not HCI back- HCI Economy to age 5 school test scores schooling rate stunted 2020 calculated 2010 Uganda 0.95 6.8 397 4.3 0.74 0.71 0.38 0.38 0.34 Ukraine 0.99 12.9 478 9.9 0.81 – 0.63 0.64 0.63 United Arab 0.99 13.5 448 9.6 0.94 – 0.67 0.68 0.62 Emirates United 1.00 13.9 520 11.5 0.93 – 0.78 0.78 0.77 Kingdom United States 0.99 12.9 512 10.6 0.89 – 0.70 0.71 0.69 Uruguay 0.99 12.2 438 8.6 0.89 – 0.60 0.60 0.59 Uzbekistan 0.98 12.0 474 9.1 0.87 0.89 0.62 – – Vanuatu 0.97 10.1 348 5.6 0.87 0.71 0.45 0.44 – Vietnam 0.98 12.9 519 10.7 0.87 0.76 0.69 0.69 0.66 West Bank 0.98 12.2 412 8.0 0.89 0.93 0.58 0.57 – and Gaza Yemen, Rep. 0.95 8.1 321 4.2 0.80 0.54 0.37 0.37 – Zambia 0.94 8.8 358 5.0 0.73 0.65 0.40 0.39 – Zimbabwe 0.95 11.1 396 7.0 0.65 0.77 0.47 0.46 0.41 Source: World Bank calculations based on the 2020 update of the Human Capital Index (HCI). Notes: This table reports the components and overall index scores for the HCI 2020, the back-calculated HCI 2018 and the HCI 2010. The HCI ranges between 0 and 1. The index is measured in terms of the productivity of the next generation of workers relative to the benchmark of complete education and full health. An economy in which a child born today can expect to achieve complete education and full health will score a value of 1 on the index. – indicates missing data. NOTES 1. United Nations Statistics Division web page, “Coverage of Birth and Death Registration,” http://unstats.un.org/unsd/demographic/CRVS/CR_coverage.htm. 2. For more information, see http://www.childmortality.org/. 3. This section borrows heavily from the technical appendix of Kraay (2018). The main source for enrollment data from UIS is administrative data. Data are 4.  collected by UIS on an annual basis from official national statistical authorities. ­ The  data are released in September of every year and include national data for the     school or reference year ending in the previous year. The national data are then updated in February, which completes the UIS publication of educational data for the data col- lection effort of the previous reference year.   5. An important agenda concerns the frequent and substantial discrepancies between household survey–based measures of school enrollment and administrative records. D’Souza, Gatti, and Kraay (2019) briefly discuss these discrepancies. 202 A PPEND IX C : H U MAN CAP ITAL IN D E X C O M PO N EN T DATA N OT ES   6. Yi = 2 for preprimary, Yi = 6 for primary, Yi = 3 for lower-secondary, and Yi = 3 for upper-secondary.    7. For the 2020 update, this process was conducted between January 29 and April 29, resulting in revised enrollment rates for all levels, which are available in individual economy files at https://www.worldbank.org/en/publication/human-capital.    8. See http://data.uis.unesco.org/.    9. The exceptions to this rule are Fiji, Kenya, and Kiribati, for which the most recent data available are from before 2010. 10. Note that one level of schooling may use TNER whereas another may use NER. For a given level of education, however, the same enrollment type is used over time. 11. This effort is made for all economies for which the same test is available in or close to the specified year. 12. Exceptions are Qatar preprimary and primary, for which the same value of 2010 is used. 13. World Bank staff working in each economy obtain these data from local govern- ment sources, for example, the Ministry of Education or National Statistics Office. 14. The flip side is that for 15 economies at least one of the enrollment rates used comes from an older year than was available in 2018’s EYS, mostly because UIS revises the series and in some instances may remove years from the series. 15. For the latest updates on the harmonized test scores, see Angrist et al. (2019). 16. The one exception to this is the 2007 and 2014 PASEC rounds, which were not designed to be intertemporally comparable and in which different overlapping countries were used to construct the exchange rate in the two periods. 17. For the 2020 HCI, 13 economies have an HTS that comes from a nonrepresentative EGRA: Bangladesh, Central African Republic, the Democratic Republic of Congo, Ethiopia, Iraq, Jamaica, Lao PDR, Liberia, Mali, Myanmar, Nigeria, Pakistan, and South Sudan. 18. See Patrinos and Angrist (2018) for further details. 19. See JME (UNICEF-WHO-World Bank Joint Child Malnutrition Estimates) (data- base), 2020 edition, UNICEF, New York, https://data.unicef.org/resources/jme/. 20. The DHS program has fielded over 400 surveys across 90 economies, and over 300 MICS have been carried out in more than 100 economies. 21. See https://www.who.int/publications-detail/jme-2020-edition. 22. See Global Burden of Disease (GBD), database, Institute for Health Metrics and Evaluation (IHME), Seattle, http://www.healthdata.org/gbd. 23. See https://population.un.org/wpp/. 24. See http://www.healthdata.org/results/data-visualizations. 25. See https://data.un.org/. REFERENCES Angrist, N., S. Djankov, P. Goldberg, and H. A. Patrinos. 2019. “Measuring Human Capital.” Policy Research Working Paper 8742, World Bank, Washington, DC. T H E H U MAN CAP ITAL IN DEX 2 0 2 0 UPDAT E: HUM A N CA PI TA L I N T HE T I M E O F C OV I D- 1 9 203 D’Souza, R., R. Gatti, and A. Kraay. 2019. “A Socioeconomic Disaggregation of the World Bank Human Capital Index.” Policy Research Working Paper 9020, World Bank, Washington, DC. Kraay, A. 2018. “Methodology for a World Bank Human Capital Index.” Policy Research Working Paper 8593, World Bank, Washington, DC. Marini, A., and C. Rokx. 2017. “Standing Tall: Peru’s Success in Overcoming Its Stunting Crisis.” World Bank, December 11, 2017. https://www.worldbank.org/en/news​ /video/2017/12/11/standing-tall-perus-success-in-overcoming-its-stunting-crisis. Patrinos, H. A., and N. Angrist. 2018. “Global Dataset on Education Quality: A Review and Update (2000–2017).” Policy Research Working Paper 8592, World Bank, Washington, DC. UNICEF (United Nations Children’s Fund). 2013. “A Passport to Protection: A Guide to Birth Registration Programming.” UNICEF, New York. UNICEF (United Nations Children’s Fund), WHO (World Health Organization), and World Bank. 2020. Levels and Trends in Child Malnutrition: Key Findings from the 2020 Edition. Geneva: World Health Organization. UNIGME (United Nations Interagency Group for Child Mortality Estimation). 2019. “Levels and Trends in Child Mortality, Report 2019.” United Nations Children’s Fund, New York. World Bank. 2018. “The Human Capital Project.” World Bank, Washington, DC. World Health Organization. 2009. “The WHO Multicentre Growth Reference Study (MGRS).” World Health Organization, Geneva. Yang, H., and M. de Onis. 2008. “Algorithms for Converting Estimates of Child Malnutrition Based on the NCHS Reference into Estimates Based on the WHO Child Growth Standards.” BMC Pediatrics 8 (19). ECO-AUDIT Environmental Benefits Statement The World Bank Group is committed to reducing its environmental footprint. In support of this commitment, we leverage electronic publishing options and print- on-demand technology, which is located in regional hubs worldwide. Together, these initiatives enable print runs to be lowered and shipping distances decreased, resulting in reduced paper consumption, chemical use, greenhouse gas emis- sions, and waste. We follow the recommended standards for paper use set by the Green Press Initiative. The majority of our books are printed on Forest Stewardship Council (FSC)–certified paper, with nearly all containing 50–100 percent recycled con- tent. The recycled fiber in our book paper is either unbleached or bleached using totally chlorine-free (TCF), processed chlorine–free (PCF), or enhanced elemen- tal ­ chlorine–free (EECF) processes. More information about the Bank’s environmental philosophy can be found at http://www.worldbank.org/corporateresponsibility. Human capital—the knowledge, skills, and health that people accumulate over their lives—is a central driver of sustainable growth and poverty reduction. More human capital is associated with higher earnings for people, higher income for countries, and stronger cohesion in societies. Much of the hard-won human capital gains in many economies over the past decade is at risk of being eroded by the COVID-19 (coronavirus) pandemic. Urgent action is needed to protect these advances, particularly among the poor and vulnerable. Designing the needed interventions, targeting them to achieve the highest effectiveness, and navigating difficult trade-offs make investing in better measurement of human capital now more important than ever. The Human Capital Index (HCI) is an international metric that benchmarks the key components of human capital across economies. It was launched in 2018 as part of the Human Capital Project, a global effort to accelerate progress toward a world where all children can achieve their full potential. Measuring the human capital that children born today can expect to attain by their 18th birthdays, the HCI highlights how current health and education outcomes shape the productivity of the next generation of workers and underscores the importance of government and societal investments in human capital. The Human Capital Index 2020 Update: Human Capital in the Time of COVID-19 presents the first update of the HCI, using health and education data available as of March 2020. It documents new evidence on trends, examples of successes, and analytical work on the utilization of human capital. The new data— collected before the global onset of COVID-19—can act as a baseline to track its effects on health and education outcomes. The report highlights how better measurement is essential for policy makers to design effective interventions and target support. In the immediate term, investments in better measurement and data use can inform pandemic containment strategies and support for those who are most affected. In the medium term, better curation and use of administrative, survey, and identification data can guide policy choices in an environment of limited fiscal space and competing priorities. In the longer term, the hope is that economies will be able to do more than simply recover ground lost during the current crisis. Ambitious, evidence- driven policy measures in health, education, and social protection can pave the way for today’s children to surpass the human capital achievements and quality of life of the generations that preceded them. ISBN 978-1-4648-1552-2 90000 9 781464 815522 SKU 211552