© 2018 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 20 19 18 17 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. 2018. World Development Report 2018: Learning to Realize Education’s Promise. Washington, DC: World Bank. doi:10.1596/978-1-4648-1096-1. 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 re-use a component of the work, it is your responsibility to determine whether permission is needed for that re-use 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. ISSN, ISBN, e-ISBN, and DOI: Softcover ISSN: 0163-5085 ISBN: 978-1-4648-1096-1 e-ISBN: 978-1-4648-1098-5 DOI: 10.1596/978-1-4648-1096-1 Hardcover ISSN: 0163-5085 ISBN: 978-1-4648-1097-8 DOI: 10.1596/978-1-4648-1097-8 Cover design: Kurt Niedermeier, Niedermeier Design, Seattle, Washington. Interior design: George Kokkinidis, Design Language, Brooklyn, New York, and Kurt Niedermeier, Niedermeier Design, Seattle, Washington. Contents xi Foreword xiii Acknowledgments xvii Abbreviations 1 Overview: Learning to realize education’s promise 4 The three dimensions of the learning crisis 16 How to realize education’s promise: Three policy responses 27 Learning to realize education’s promise 37 Part I: Education’s promise 38 Chapter 1: Schooling, learning, and the promise of education 38 Education as freedom 38 Education improves individual freedoms 41 Education benefits all of society 44 Learning and the promise of education 57 Part II: The learning crisis 58 Chapter 2: The great schooling expansion—and those it has left behind 58 Most children have access to basic education 60 Poverty, gender, ethnicity, disability, and location explain most remaining schooling disparities 63 For poor parents, schooling requires trade-offs 68 Spotlight 1: The biology of learning 71 Chapter 3: The many faces of the learning crisis 71 For too many, learning isn’t happening 78 Poor children learn the least, which hurts them the most 78 What is causing the learning crisis? 88 Spotlight 2: Poverty hinders biological development and undermines learning 91 Chapter 4: To take learning seriously, start by measuring it 91 The learning crisis is often hidden—but measurement makes it visible 92 Measures for learning guide action 93 Measures of learning spur action v 95 Choose learning metrics based on what the country needs 95 Will learning metrics narrow the vision for education? 96 Six tips for effective learning measurement 102 Spotlight 3: The multidimensionality of skills Part III: Innovations and evidence for learning 107  108 Spotlight 4: Learning about learning 112 Chapter 5: There is no learning without prepared, motivated learners 112 Investing in their early years prepares children for school 116 Providing demand-side support can get kids to school, but not necessarily to learn 119 Remedial education can prepare learners for further education and training 131 Chapter 6: Teacher skills and motivation both matter (though many education systems act like they don’t) 131 Most teacher training is ineffective, but some approaches work 133 Helping teachers teach to the level of the student has proven effective 136 Teacher motivation and incentives make a difference, even with few inputs 145 Chapter 7: Everything else should strengthen the teacher-learner interaction 145 Technological interventions increase learning—but only if they enhance the teacher-learner relationship 147 Other inputs bring learners to school—but promote learning only if they target teaching and learning 148 School management and governance are crucial, and involving communities can help overcome incentive problems and information failures—but only if communities have capacity 154 Chapter 8: Build on foundations by linking skills training to jobs 154 Workplace training can help young people develop skills, yet few benefit from it 156 Short-term job training offers opportunities, but most programs fail to deliver 156 TVET can prepare young people for work, but early sorting into TVET can limit career growth 157 Successful job training programs share several features 164 Spotlight 5: Technology is changing the world of work: What does that mean for learning? Part IV: Making the system work for learning 169  at scale 170 Chapter 9: Education systems are misaligned with learning 171 Misalignments and incoherence impede learning 175 Technical complexities make it hard to align education systems with learning 183 Spotlight 6: Spending more or spending better—or both? 189 Chapter 10: Unhealthy politics drives misalignments 189 Unhealthy politics can intensify misalignments in education systems 190 Multiple actors and interests: Pulling the system out of alignment at each step of the policy cycle 195 Trapped in low-accountability, low-learning equilibriums vi | CONTENTS 199 Chapter 11: How to escape low-learning traps 199 Improving information 203 Building coalitions and strengthening incentives 207 Encouraging innovation and agility 211 How can external actors support initiatives to improve learning? Boxes 1.1 40 Schooling as human capital 6.5 137 Would raising teachers’ salaries formation or as a signaling device? increase their motivation? 1.2 44 Education can’t do it alone 6.6 138 One factor undermining teaching: 1.3 48 Comparing attainment across Poor working conditions countries and economies—learning- 7.1 149 Training better school principals in adjusted years of schooling Jamaica 2.1 61 Access denied: The effects of fragility, 9.1 172 It’s all about (education) systems conflict, and violence 3.1 74 Those who can’t read by the end of 9.2 174 Aligning all the ingredients for grade 2 struggle to catch up effective teaching in Shanghai 3.2 75 Gender-based differences in learning 9.3 176 Can private schooling be aligned to depend on the subject learning for all? 3.3 82 Teachers may perceive low effort as 10.1 191 How do teachers’ unions affect being justified learning? 4.1 92 Good measures of learning illuminate 10.2 193 How politics can derail learning in all parts of the education system conflict-affected states 4.2 97 A global learning metric? 11.1 201 Using information to align incentives 5.1 116 Early childhood education prepares with learning in Brazil young children for school 11.2 202 Citizen-led assessments have raised 5.2 118 Communities can leverage the many awareness of the learning crisis in hours spent outside the classroom to South Asia and Sub-Saharan Africa boost learning 11.3 204 Using the legal system to press for 5.3 120 Providing information on children’s change school performance can help parents to motivate their children 11.4 205 Using “labs” to build coalitions for learning 6.1 133 The landscape of in-service teacher training 11.5 205 Reformers in Chile negotiated 6.2 133 What works in preservice teacher changes gradually training? 11.6 208 High-performing schools in the 6.3 135 Reaching learners in their own West Bank and Gaza offer some language learning lessons 6.4 136 Using diagnostic data to deliver 11.7 209 Burundi improved education services better learning in Latin America by iterating and adapting Figures O.1 5 Shortfalls in learning start early O.4 7 Students often learn little from year to year, and early learning deficits are O.2 6 In several countries, the 75th magnified over time percentile of PISA test takers performs below the 25th percentile of the O.5 8 The percentage of primary school OECD average students who pass a minimum proficiency threshold is often low O.3 7 Children from poor households in Africa typically learn much less CONTENTS | vii O.6 9 School completion is higher for richer 2.3 60 National income is correlated with and urban families, but gender gaps the gap between primary and lower are more context-dependent secondary completion rates O.7 10 Why learning doesn’t happen: Four 2.4 61 Lower-income countries are rapidly immediate factors that break down expanding secondary education at a time when much of their population O.8 11 Socioeconomic gaps in cognitive has not yet completed primary school achievement grow with age—even in preschool years 2.5 62 School completion is higher for richer O.9 11 In Africa, teachers are often absent and urban families, but gender gaps from school or from classrooms while are more context-dependent at school 2.6 63 Multiple exclusions: Girls from poor O.10 12 Management capacity is low in households often have the lowest schools in low- and middle-income rates of education attainment countries S1.1 69 Synapse development over the first O.11 13 Technical and political factors divert 20 years of life schools, teachers, and families from a 3.1 72 Most grade 6 students in West and focus on learning Central Africa are not sufficiently O.12 17 Many countries lack information on competent in reading or mathematics learning outcomes 3.2 72 Most grade 6 students in southern O.13 19 Low-performing countries don’t face and East Africa are not sufficiently sharp trade-offs between learning competent in mathematics, and and other education outputs several countries score poorly in reading as well O.14 21 It’s more complicated than it looks: People act in reaction to the choices 3.3 73 Learning outcomes are substantially of others throughout the system lower for poor children in Latin America O.15 26 Coherence and alignment toward learning B3.2.1 75 Girls outperform boys on reading in all countries and economies, 1.1 39 More schooling is systematically but boys typically do better in associated with higher wages mathematics and science 1.2 40 Mortality rates in the United States 3.4 76 Learning outcomes vary greatly are lower for adults with more across countries and economies—in education several countries, the 75th percentile 1.3 43 People with higher education hold of PISA test takers performs below the stronger beliefs about the importance 25th percentile of the OECD average of democracy 3.5 77 Middle-income countries tend to have 1.4 45 Learning varies widely across lower rates of literacy proficiency countries; in 6 of the 10 countries than high-income countries assessed, only half or fewer of primary completers can read 3.6 77 Reading proficiency is low in many parts of the developing world 1.5 46 What matters for growth is learning 3.7 78 Family socioeconomic status 1.6 47 Increasing learning would yield significantly affects students’ average major economic benefits PISA scores B1.3.1 48 There can be a large gap between 3.8 79 The proximate determinants of learning-adjusted and unadjusted learning years of schooling 3.9 79 Socioeconomic gaps in cognitive 2.1 59 School enrollments have shot up in achievement grow with age—even in developing countries preschool years 2.2 59 Most of the world’s population with less than a primary education is in 3.10 81 A lot of official teaching time is lost South Asia, but rates are similar in B3.3.1 82 Teachers’ beliefs about their effort Sub-Saharan Africa and its effects viii | CONTENTS 3.11 82 Staff compensation consumes the 6.2 137 Prospective engineers typically score largest share of resources available higher than prospective teachers on for public education PISA tests 3.12 83 Management capacity is low in 7.1 146 Information and communication schools in low- and middle-income technology has had a mixed impact countries on learning 7.2 148 Schools vary significantly in S2.1 89 Severe deprivation affects brain management quality structure and function from early in life 8.1 155 Few benefit from workplace training, and those who do tend to already S2.2 89 Risk and protective factors affect have better literacy or education developmental trajectories 8.2 157 Most vocational training students 4.1 95 No internationally comparable data enroll during upper secondary school on learning are available for most children outside of high-income S5.1 165 Technology use has increased countries dramatically over the past decade— but remains low in many countries 4.2 96 Low-performing countries don’t face 9.1 171 Technical and political barriers pull sharp trade-offs between learning education systems away from the and other education outputs goal of learning S3.1 103 Cognitive, socioemotional, and 9.2 173 Simple associations between technical skills interact education spending and learning S4.1 108 The number of experimental and are weak quasi-experimental studies of B9.3.1 178 In Bangladesh, there are 11 different interventions to improve learning has kinds of nonstate providers of mushroomed in recent decades presecondary education S4.2 109 It’s more complicated than it looks: S6.1 184 Governments devote a large share of People act in reaction to the choices their budgets to education of others throughout the system S6.2 185 The relationship between changes 5.1 114 Investments in high-quality in public education spending and programs during children’s early student learning is often weak years pay off 10.1 191 Contradictory interests detract from 5.2 115 Intense deprivation can impair brain learning objectives development B10.1.1 192 Teacher unionization varies across countries 5.3 117 Integrated programs through the early years are necessary for proper 10.2 195 Interdependencies characterize the child development relationship between teachers and politicians 5.4 118 What happens when school fees are eliminated? Evidence from eight 11.1 200 Primary school numeracy has countries increased dramatically in England B11.5.1 205 Reading scores have improved in 5.5 119 Not all education systems are Chile equally productive, but even the least productive deliver some learning to 11.2 207 Problem-driven iterative adaptation some learners drives successful reforms 5.6 120 Young people follow different paths 11.3 210 Trends in public education spending in their education in the Philippines track changes in the broader political and economic 5.7 121 Workers with higher literacy context proficiency are more likely to enter white-collar jobs 11.4 212 Most funding for education comes from domestic sources, but 6.1 134 Only a small fraction of learners international finance is important for keeps up with the curriculum low-income countries CONTENTS | ix Map B6.3.1 135 Linguistic diversity around the world Tables O.1 14 Alignment and coherence both matter B9.3.1 176 Private providers account for a O.2 14 Multiple interests govern the actions significant share of school enrollment of education stakeholders S6.1 186 Inequalities in public education 1.1 39 Examples of education’s benefits spending are common 1.2 42 More schooling leads to more voting 11.1 202 Principles for making the most of information and the roles that actors 3.1 80 Few teachers reach minimum can play thresholds of performance on knowledge assessments 11.2 203 Principles for building effective coalitions and the roles that actors 5.1 113 Models of human behavior can can play guide actions to improve learner preparation: Some examples 11.3 209 Principles for encouraging innovation at scale and the roles that 6.1 132 Models of human behavior can guide actors can play actions to improve teaching: Some examples 7.1 146 Models of human behavior can guide actions to improve the effectiveness of school inputs and governance: Some examples x | CONTENTS Foreword Education and learning raise aspirations, set values, and ultimately enrich lives. The coun- try where I was born, the Republic of Korea, is a good example of how education can play these important roles. After the Korean War, the population was largely illiterate and deeply impoverished. The World Bank said that, without constant foreign aid, Korea would find it difficult to provide its people with more than the bare necessities of life. The World Bank considered even the lowest interest rate loans to the country too risky. Korea understood that education was the best way to pull itself out of economic misery, so it focused on overhauling schools and committed itself to educating every child—and educating them well. Coupled with smart, innovative government policies and a vibrant private sector, the focus on education paid off. Today, not only has Korea achieved universal literacy, but its students also perform at the highest levels in international learning assess- ments. It’s a high-income country and a model of successful economic development. Korea is a particularly striking example, but we can see the salutary effects of education in many countries. Delivered well, education—and the human capital it creates—has many benefits for economies, and for societies as a whole. For individuals, education promotes employment, earnings, and health. It raises pride and opens new horizons. For societies, it drives long-term economic growth, reduces poverty, spurs innovation, strengthens institu- tions, and fosters social cohesion. In short, education powerfully advances the World Bank Group’s twin strategic goals: ending extreme poverty and boosting shared prosperity. Given that today’s students will be tomorrow’s citizens, leaders, workers, and parents, a good education is an investment with enduring benefits. But providing education is not enough. What is important, and what generates a real return on investment, is learning and acquiring skills. This is what truly builds human capital. As this year’s World Development Report documents, in many countries and commu- nities learning isn’t happening. Schooling without learning is a terrible waste of precious resources and of human potential. Worse, it is an injustice. Without learning, students will be locked into lives of poverty and exclusion, and the children whom societies fail the most are those most in need of a good education to succeed in life. Learning conditions are almost always much worse for the disadvantaged, and so are learning outcomes. Moreover, far too many children still aren’t even attending school. This is a moral and economic crisis that must be addressed immediately. This year’s Report provides a path to address this economic and moral failure. The detailed analysis in this Report shows that these problems are driven not only by service delivery failings in schools but also by deeper systemic problems. The human capital lost xi because of these shortcomings threatens development and jeopardizes the future of peo- ple and their societies. At the same time, rapid technological change raises the stakes: to compete in the economy of the future, workers need strong basic skills and foundations for adaptability, creativity, and lifelong learning. To realize education’s promise, we need to prioritize learning, not just schooling. This Report argues that achieving learning for all will require three complementary strategies: • First, assess learning to make it a serious goal. Information itself creates incentives for reform, but many countries lack the right metrics to measure learning. • Second, act on evidence to make schools work for learning. Great schools build strong teacher-learner relationships in classrooms. As brain science has advanced and educators have innovated, the knowledge of how students learn most effectively has greatly expanded. But the way many countries, communities, and schools approach education often differs greatly from the most promising, evidence-based approaches. • Third, align actors to make the entire system work for learning. Innovation in classrooms won’t have much impact if technical and political barriers at the sys- tem level prevent a focus on learning at the school level. This is the case in many countries stuck in low-learning traps; extricating them requires focused attention on the deeper causes. The World Bank Group is already incorporating the key findings of this Report into our operations. We will continue to seek new ways to scale up our commitment to education and apply our knowledge to serve those children whose untapped potential is wasted. For example, we are developing more useful measures of learning and its determinants. We are ensuring that evidence guides operational practice to improve learning in areas such as early-years interventions, teacher training, and educational technology. We are making sure that our project analysis and strategic country diagnoses take into account the full range of system-level opportunities and limitations—including political constraints. And we will continue to emphasize operational approaches that allow greater innovation and agility. Underlying these efforts is the World Bank Group’s commitment to ensuring that all of the world’s students have the opportunity to learn. Realizing education’s promise means giving them the chance not only to compete in tomorrow’s economy, but also to improve their communities, build stronger countries, and move closer to a world that is finally free of poverty. Jim Yong Kim President The World Bank Group xii | FOREWORD Acknowledgments This year’s World Development Report (WDR) was prepared by a team led by Deon Filmer and Halsey Rogers. The core team was composed of Samer Al-Samarrai, Magdalena Ben- dini, Tara Béteille, David Evans, Märt Kivine, Shwetlena Sabarwal, and Alexandria Valerio, together with research analysts Malek Abu-Jawdeh, Bradley Larson, Unika Shrestha, and Fei Yuan. Rafael de Hoyos and Sophie Naudeau were members of the extended team. Stephen Commins provided consultations support. Mary Breeding, Ji Liu, Christian Ponce de León, Carla Cristina Solis Uehara, Alies Van Geldermalsen, and Paula Villaseñor served as consul- tants. The production and logistics team for the Report consisted of Brónagh Murphy and Jason Victor. The Report is sponsored by the Development Economics Vice Presidency. Overall guid- ance for preparation of the Report was provided by Paul Romer, Senior Vice President and Chief Economist, and Ana Revenga, Deputy Chief Economist. In the early months of the Report’s preparation, guidance was provided by Kaushik Basu, former Senior Vice President and Chief Economist, and Indermit Gill, former Director for Development Policy. The team is also grateful for comments and guidance from Shantayanan Devarajan, Senior Director for Development Economics. The Education Global Practice and the Human Development Global Practice Group provided consistent support to the Report team. The team is espe- cially grateful for support and guidance provided by Jaime Saavedra, Senior Director, and Luis Benveniste, Director, of the Education Global Practice. The team received guidance from an advisory panel composed of Gordon Brown (who, together with the Chief Economist, cochaired the panel), Michelle Bachelet, Rukmini Banerji, Julia Gillard, Eric Hanushek, Olli-Pekka Heinonen, Ju-Ho Lee, and Serigne Mbaye Thiam. Although the team valued their advice and found it very useful, the views expressed in the Report do not necessarily reflect those of the panel members. The team also benefited at an early stage from consultations on emerging themes with the Chief Economist’s Council of Eminent Persons. Council members providing comments were Montek Singh Ahluwalia, François Bourguignon, Heba Handoussa, Justin Yifu Lin, Ory Okolloh, Pepi Patrón, Amartya Sen, Joseph Stiglitz, Finn Tarp, and Maria Hermínia Tavares de Almeida. Paul Holtz was the principal editor of the Report. Bruce Ross-Larson provided editorial guidance, and Sabra Ledent and Gwenda Larsen copyedited and proofread the Report. Kurt Niedermeier was the principal graphic designer. Alejandra Bustamante and Surekha Mohan provided resource management support for the team. Phillip Hay, Mikael Reventar, Anushka Thewarapperuma, and Roula Yazigi, together with Patricia da Camara and Kavita Watsa, provided guidance and support on communication and dissemination. Special thanks are extended to Mary Fisk, Patricia Katayama, Stephen Pazdan, and the World Bank’s Formal Publishing Program. The team would also like to thank Maria Alyanak, Laverne Cook, Maria del Camino Hurtado, Chorching Goh, Vivian Hon, Elena Chi-Lin Lee, Nancy Tee Lim, David Rosenblatt, and Bintao Wang for their coordinating roles. xiii The team is grateful for generous support for preparation of the Report provided by the Knowledge for Change Program (KCP, a multidonor Trust Fund) and especially from the governments and development agencies of the following KCP donor countries: Finland, France, and Norway. Background and related research, along with dissemination, are being generously supported by the Bill & Melinda Gates Foundation, Early Learning Partnership Trust Fund, LEGO Foundation, and Nordic Trust Fund. Consultation events attended by government officials, researchers, and civil society organizations were held in Bolivia, Brazil, Canada, China, Côte d’Ivoire, Finland, France, Germany, India, Indonesia, Japan, Kenya, Malaysia, Mexico, Senegal, South Africa, Tanzania, Thailand, Turkey, the United Kingdom, and the United States, with participants drawn from many more countries. The team thanks those who took part in these events for their helpful comments and suggestions. Further information on these events is available at http://www .worldbank.org/wdr2018. Interagency consultations were held with the Association for the Development of Education in Africa (ADEA), Global Development Network (GDN), Global Partnership for Education (GPE), International Commission on Financing Global Education Opportunity (Education Commission), International Monetary Fund (IMF), Organisation for Economic Co-operation and Development (OECD), United Nations Children’s Fund (UNICEF), and United Nations Educational, Scientific, and Cultural Organization (UNESCO). Consulta- tions with bilateral development partners included representatives of the governments of Canada, Finland, Japan, the Republic of Korea, Norway, and Sweden, and of Australia’s Department of Foreign Affairs and Trade (DFAT), the French Development Agency (AFD), German Agency for International Cooperation (GIZ GmbH), German Federal Ministry for Economic Cooperation and Development (BMZ), Japan International Cooperation Agency (JICA), U.K. Department for International Development (DFID), and U.S. Agency for Inter­ national Development (USAID). The team also held consultations with the advisory board of KCP. The team is grateful to all those who took part in these events. Civil society organizations (CSOs) represented at consultations included, among others, ActionAid, Bill & Melinda Gates Foundation, Education International, Global Campaign for Education, LEGO Foundation, MasterCard Foundation, ONE Campaign, Oxfam, Save the Children, Teach for All, and World Vision. In addition, a diverse group of CSOs participated in a CSO Forum session held during the 2017 World Bank/IMF Spring Meetings and in an e-forum held in March 2017. The team is grateful to these CSOs for their input and useful engagement. Researchers and academics provided helpful feedback at WDR-oriented sessions at the 2016 Research on Improving Systems of Education (RISE) Conference at Oxford Univer- sity, 2017 meetings of the Allied Social Sciences Associations (ASSA), 2017 meetings of the Society for Research on Education Effectiveness (SREE), 2017 Mexico Conference on Politi- cal Economy of Education, and 2017 meeting of the Systems Approach for Better Education Results (SABER) Advisory Panel. In addition, events dedicated to the WDR were organized by the Aga Khan Foundation and Global Affairs Canada in Ottawa; Brookings Center for Universal Education in Washington, DC; Columbia School of International and Public Affairs and Cornell University in New York; Development Policy Forum of GIZ GmbH, on behalf of BMZ, in Berlin; JICA in Tokyo; Université Félix Houphouët-Boigny in Abidjan; and USAID in Washington, DC. This Report draws on background papers prepared by Violeta Arancibia, Felipe Barrera- Osorio, Tessa Bold, Pierre de Galbert, Louise Fox, Dileni Gunewardena, James Habyarimana, Michael Handel, Anuradha Joshi, Kanishka Kacker, Michelle Kaffenberger, Upaasna Kaul, Elizabeth M. King, Gayle Martin, Eema Masood, Ezequiel Molina, Sebastián Monroy- Taborda, Kate Moriarty, Anna Popova, Lant Pritchett, Christophe Rockmore, Andrew Rosser, María Laura Sánchez Puerta, Priyam Saraf, M. Najeeb Shafiq, Brian Stacy, Jakob Svensson, Namrata Tognatta, Robert Toutkoushian, Michael Trucano, Waly Wane, Tim Williams, and Attiya Zaidi. xiv | ACKNOWLEDGMENTS The team drew on the analysis, research, and literature reviews of researchers and specialists from across the world. In addition, the team would like to thank the following for their feedback and suggestions: Christine Adick, Ben Ansell, Manos Antoninis, Caridad Araujo, David Archer, Belinda Archibong, Monazza Aslam, Girindre Beeharry, Penelope Bender, Peter Bergman, Raquel Bernal, Robert Birch, Tarsald Brautaset, Barbara Bruns, Annika Calov, Michael Clemens, Luis Crouch, Rohen d’Aiglepierre, Rossieli Soares da Silva, Momar Dieng, Rob Doble, Amy Jo Dowd, Margaret Dubeck, Sandra Dworack, Alex Eble, Marcel Fafchamps, John Floreta, Eli Friedman, Akihiro Fushimi, Paul Gertler, Rachel Glennerster, Paul Glewwe, Amber Gove, Oliver Haas, James Habyarimana, Jeffrey Hammer, Michael Handel, Christoph Hansert, Blanca Heredia, Sam Hickey, Veronika Hilber, Arja-Sisko Holappa, Naomi Hossain, Huang Xiaoting, Ali Inam, Dhir Jhingran, Emmanuel Jimenez, Maciej Jubowski, Ravi Kanbur, Cheikh Kane, Jouni Kangasniemi, Devesh Kapur, Vishnu Karki, Nina Kataja, Venita Kaul, Kim Kerr, Elizabeth M. King, Kenneth King, Geeta Kingdon, Eiji Kozuka, Michael Kremer, K. P. Krishnan, Kazuo Kuroda, Elina Lehtomäki, Henry Levin, Brian Levy, Krystelle Lochard, Karen Macours, Lu Mai, Akshay Mangla, M. A. Mannan, Santhosh Mathew, Imran Matin, Jordan Matsudaira, Karthik Muralidharan, Essa Chanie Mussa, Charles Nelson III, Aromie Noe, Munaz Ahmed Noor, Mario Novelli, Mead Over, Jan Pakulski, Benjamin Piper, Lant Pritchett, Ritva Reinikka, Risto Rinne, Jo Ritzen, Francisco Rivera Batiz, John Rogers, Caine Rolleston, Andrew Rosser, David Sahn, Justin Sandefur, Yasuyuki Sawada, Andreas Schleicher, Ben Ross Schneider, Dorothea Schonfeld, Olaf Seim, Abhijeet Singh, David Skinner, William Smith, Prachi Srivastava, Liesbet Steer, R. Subrahmanyam, Sudarno Sumarto, Jan Svejnar, Jakob Svensson, Soubhy Tawil, Valerie Tessio, Auli Toom, Miguel Urqiola, Jouni Välijärvi, Olli Vesterinen, Joseph Wales, Libing Wang, Michael Ward, Kevin Watkins, Mark Wenz, Yang Po, Khair Mohamad Yusof, and Andrew Zeitlin. Team members also drew heavily on their own experiences and interactions with the many dedicated educators, administrators, and policy makers who work in often difficult conditions to provide students with the best educational opportunities possible. A number of World Bank colleagues provided insightful comments, feedback, and collaboration: Junaid Ahmad, Omar Arias, Nina Arnhold, Ana Belver, Hana Brixi, James Brumby, Pedro Cerdan Infantes, Marie-Hélène Cloutier, Aline Coudouel, Amit Dar, Jishnu Das, Amanda Epstein Devercelli, Gregory Elacqua, Emanuela Galasso, Diana Hincapie, Alaka Holla, Peter Holland, Sachiko Kataoka, Stuti Khemani, Igor Kheyfets, Kenneth King, Eva Kloeve, Steve Knack, Xiaoyan Liang, Toby Linden, Oni Lusk-Stover, Francisco Mar­ molejo, Yasuhiko Matsuda, Julie McLaughlin, Muna Meky, Ezequiel Molina, Caitlin Moss, Matiullah Noori, Anna Olefir, Owen Ozier, Andrew Ragatz, Vijayendra Rao, Dan Rogger, Audrey Sacks, María Laura Sánchez Puerta, Indhira Santos, William Seitz, Shabnam Sinha, Lars Sondergaard, Dewi Susanti, Christopher Thomas, Michael Trucano, Adam Wagstaff, and Melanie Walker. The team would also like to thank the World Bank colleagues who helped organize and facilitate consultations and advised on translations: Gabriela Geraldes Bastos, Paolo Belli, Moussa Blimpo, Andreas Blom, Leandro Costa, Oumou Coulibaly, Meaza Zerihun Demissie, Safaa El-Kogali, Tazeen Fasih, Ning Fu, Elena Glinskaya, Marek Hanusch, Pimon Iamsri- pong, Susiana Iskandar, Nalin Jena, Hamoud Abdel Wedoud Kamil, Adriane Landwehr, Dilaka Lathapipat, Khady Fall Lo, Norman Loayza, André Loureiro, Hope Nanshemeza, Mademba Ndiaye, Koichi Omori, Azedine Ouerghi, Tigran Shmis, Taleb Ould Sid’ahmed, Lars Sondergaard, Dewi Susanti, Yasusuke Tsukagoshi, and Michael Woolcock. In addition, the team is grateful to the many World Bank colleagues who provided writ- ten comments during the formal Bankwide review process: Cristian Aedo, Inga Afanasieva, Ahmad Ahsan, Edouard Al Dahdah, Umbreen Arif, Nina Arnhold, Anna Autio, Arup Banerji, Elena Bardasi, Sajitha Bashir, Ana Belver, Raja Bentaouet Kattan, Luis Benveniste, Moussa Blimpo, Erik Bloom, Vica Bogaerts, Susan Caceres, César Calderón, Ted Haoquan Chu, Punam Chuhan-Pole, Fernando Ramirez Cortes, Michael Crawford, Laisa Daza, Bénédicte de la Brière, Gabriel Demombynes, Shanta Devarajan, Sangeeta Dey, Ousmane Diagana, ACKNOWLEDGMENTS | xv Ousmane Dione, Safaa El Tayeb El-Kogali, Marianne Fay, María Marta Ferreyra, Carina Fonseca, Marie Gaarder, Roberta Gatti, Ejaz Syed Ghani, Elena Glinskaya, Markus Goldstein, Melinda Good, David Gould, Sangeeta Goyal, Caren Grown, Keith Hansen, Amer Hasan, Caroline Heider, Katia Herrera, Niels Holm-Nielsen, Dingyong Hou, Elena Ianchovichina, Keiko Inoue, Sandeep Jain, Omer Karasapan, Michel Kerf, Asmeen Khan, Igor Kheyfets, Youssouf Kiendrebeogo, Daniel John Kirkwood, Eva Kloeve, Markus Kostner, Daniel Lederman, Hans Lofgren, Gladys López-Acevedo, Javier Luque, Michael Mahrt, Francisco Mar­ molejo, Kris McDonall, Mahmoud Mohieldin, Lili Mottaghi, Mary Mulusa, Yoko Nakashima, Shiro Nakata, Muthoni Ngatia, Shinsaku Nomura, Dorota Agata Nowak, Michael O’Sullivan, Arunma Oteh, Aris Panou, Georgi Panterov, Suhas Parandekar, Harry Patrinos, Dhushyanth Raju, Martín Rama, Sheila Redzepi, Lea Marie Rouanet, Jaime Saavedra, Hafida Sahraoui, Sajjad Shah, Sudhir Shetty, Mari Shojo, Lars Sondergaard, Nikola Spatafora, Venkatesh Sundararaman, Janssen Teixeira, Jeff Thindwe, Hans Timmer, Yvonne Tsikata, Laura Tuck, Anuja Utz, Julia Valliant, Axel van Trotsenburg, Carlos Vegh, Binh Thanh Vu, Jan Walliser, Jason Weaver, Michel Welmond, Deborah Wetzel, Christina Wood, and Hanspeter Wyss. The team apologizes to any individuals or organizations inadvertently omitted from this list and expresses its gratitude to all who contributed to this Report, including those whose names may not appear here. The team members would also like to thank their families for their support throughout the preparation of this Report. And finally, the team members thank the many children and youth who have inspired them through interactions in classrooms around the world over the years—as well as the many others whose great potential has motivated this Report. The World Development Report 2018 is dedicated to them. xvi | ACKNOWLEDGMENTS Abbreviations A4L Assessment for Learning ASER Annual Status of Education Report BRN Big Results Now in Education (Tanzania) CAMPE Campaign for Popular Education (Bangladesh) CCT conditional cash transfer CSEF Civil Society Education Fund DISE District Information System for Education (India) EGRA Early Grade Reading Assessment GDP gross domestic product GNECC Ghana National Education Campaign Coalition I-BEST Integrated Basic Education and Skills Training ICT information and communication technology IDEB Índice de Desenvolvimento da Educação Básica (Index of Basic Education Development, Brazil) LLECE Latin American Laboratory for Assessment of the Quality of Education MDG Millennium Development Goal MENA Middle East and North Africa NAFTA North American Free Trade Agreement NGO nongovernmental organization OECD Organisation for Economic Co-operation and Development PASEC Programme d’Analyse des Systèmes Éducatifs de la Confemen (Programme for the Analysis of Education Systems) PIAAC Programme for the International Assessment of Adult Competencies PIRLS Progress in International Reading Literacy Study PISA Programme for International Student Assessment PPP purchasing power parity SACMEQ Southern and Eastern Africa Consortium for Monitoring Educational Quality SAR special administrative region SAT Scholastic Aptitude Test SDG Sustainable Development Goal SIMCE Sistema de Medición de la Calidad de la Educación (Education Quality Measurement System, Chile) SNED Sistema Nacional de Evaluación de Desempeño (National Performance Evaluation System, Chile) SNTE Sindicato Nacional de Trabajadores de la Educación (National Union of Educational Workers, Mexico) TERCE Third Regional Comparative and Explanatory Study TIMSS Trends in International Mathematics and Science Study TVET technical and vocational education and training UNESCO United Nations Educational, Scientific, and Cultural Organization UNRWA United Nations Relief and Works Agency USAID U.S. Agency for International Development WHO World Health Organization WIDE World Inequality Database on Education xvii OVERVIEW  earning to realize L education’s promise Learning to realize education’s promise Assess learning Act on evidence Align actors to make it a serious goal to make schools to make the whole work for all learners system work for learning OVERVIEW Learning to realize education’s promise “Education is the most powerful weapon we can use to change the world.” NELSON MANDELA (2003) If your plan is for one year, plant rice. If your plan is for ten years, plant trees. “ If your plan is for one hundred years, educate children.” KUAN CHUNG (7TH CENTURY BC) Schooling is not the same as learning. In Kenya, transaction, reading a doctor’s instructions, or inter- Tanzania, and Uganda, when grade 3 students were preting a campaign promise—let alone building a asked recently to read a sentence such as “The fulfilling career or educating their children. name of the dog is Puppy,” three-quarters did not This learning crisis is a moral crisis. When deliv- understand what it said.1 In rural India, just under ered well, education cures a host of societal ills. three-quarters of students in grade 3 could not solve For individuals, it promotes employment, earnings, a two-digit subtraction such as 46 – 17, and by grade health, and poverty reduction. For societies, it spurs 5 half could still not do so.2 Although the skills of innovation, strengthens institutions, and fosters Brazilian 15-year-olds have improved, at their cur- social cohesion. But these benefits depend largely rent rate of improvement they won’t reach the rich- on learning. Schooling without learning is a wasted country average score in math for 75 years. In reading, opportunity. More than that, it is a great injustice: it will take more than 260 years.3 Within countries, the children whom society is failing most are the ones learning outcomes are almost always much worse who most need a good education to succeed in life. for the disadvantaged. In Uruguay, poor children in Any country can do better if it acts as though learn- grade 6 are assessed as “not competent” in math at ing really matters. That may sound obvious—after five times the rate of wealthy children.4 Moreover, all, what else is education for? Yet even as learning such data are for children and youth lucky enough to goals are receiving greater rhetorical support, in be in school. Some 260 million aren’t even enrolled in practice many features of education systems conspire primary or secondary school.5 against learning. This Report argues that countries These countries are not unique in the challenges can improve by advancing on three fronts: they face. (In fact, they deserve credit for measuring student learning and making the results public.) Assess learning—to make it a serious goal. This •  Worldwide, hundreds of millions of children reach means using well-designed student assessments young adulthood without even the most basic life to gauge the health of education systems (not skills. Even if they attend school, many leave without primarily as tools for administering rewards and the skills for calculating the correct change from a punishments). It also means using the resulting Overview | 3 learning measures to spotlight hidden exclu- The three dimensions of the sions, make choices, and evaluate progress. • Act on evidence—to make schools work for all learners. learning crisis Evidence on how people learn has exploded in Education should equip students with the skills they recent decades, along with an increase in edu- need to lead healthy, productive, meaningful lives. cational innovation. Countries can make much Different countries define skills differently, but better use of this evidence to set priorities for all share some core aspirations, embodied in their their own practice and innovations. curriculums. Students everywhere must learn how • Align actors—to make the whole system work for to interpret many types of written passages—from learning. Countries must recognize that all the medication labels to job offers, from bank statements classroom innovation in the world is unlikely to to great literature. They have to understand how have much impact if, because of technical and numbers work so that they can buy and sell in mar- political barriers, the system as a whole does not kets, set family budgets, interpret loan agreements, support learning. By taking into account these or write engineering software. They require the real-world barriers and mobilizing everyone who higher-order reasoning and creativity that builds on has a stake in learning, countries can support these foundational skills. And they need the socio- innovative educators on the front lines. emotional skills—such as perseverance and the ability to work on teams—that help them acquire and apply When improving learning becomes a priority, the foundational and other skills. great progress is possible. In the early 1950s, the Many countries are not yet achieving these goals. Republic of Korea was a war-torn society held back by First, the learning that one would expect to happen very low literacy levels. By 1995 it had achieved uni- in schools—whether expectations are based on formal versal enrollment in high-quality education through curriculums, the needs of employers, or just common secondary school. Today, its young people perform at sense—is often not occurring. Of even greater con- the highest levels on international learning assess- cern, many countries are failing to provide learning ments. Vietnam surprised the world when the 2012 for all. Individuals already disadvantaged in society— results of the Programme for International Student whether because of poverty, location, ethnicity, gen- Assessment (PISA) showed that its 15-year-olds were der, or disability—learn the least. Thus education performing at the same level as those in Germany— systems can widen social gaps instead of narrowing even though Vietnam was a lower-middle-income them. What drives the learning shortfalls is becoming country. Between 2009 and 2015, Peru achieved some clearer thanks to new analyses spotlighting both the of the fastest growth in overall learning outcomes—an immediate cause—poor service delivery that ampli- improvement attributable to concerted policy action. fies the effects of poverty—and the deeper system- In Liberia, Papua New Guinea, and Tonga, early grade level problems, both technical and political, that allow reading improved substantially within a very short poor-quality schooling to persist. time thanks to focused efforts based on evidence. And recently, Malaysia and Tanzania launched promising Learning outcomes are poor: Low levels, societywide collaborative approaches to systemati- high inequality, slow progress cally improving learning. The recent expansion in education Progress like this requires a clear-eyed diagnosis, is impressive by historical stan- followed by concerted action. Before showing what dards. In many developing coun- can be done to fulfill education’s promise, this over- tries over the last few decades, net view first shines a light on the learning crisis: how enrollment in education has greatly and why many countries are not yet achieving “learn- outpaced the historic performance ing for all.” This may make for disheartening reading, of today’s industrial countries. For Problem but it should not be interpreted as saying that all is example, it took the United States dimension 1: lost—only that too many young people are not getting 40 years—from 1870 to 1910—to Outcomes the education they need. The rest of the overview increase girls’ enrollments from 57 shows how change is possible if systems commit to percent to 88 percent. By contrast, Morocco achieved “all for learning,” drawing on examples of families, a similar increase in just 11 years.6 The number of educators, communities, and systems that have made years of schooling completed by the average adult in real progress. the developing world more than tripled from 1950 to 4 | World Development Report 2018 Figure O.1 Shortfalls in learning start early Percentage of grade 2 students who could not perform simple reading or math tasks, selected countries a. Grade 2 students who could not read b. Grade 2 students who could not a single word of a short text perform two-digit subtraction 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 . a q co ia ia an ep a na da a da na ua q a i i al aw aw bi Ira di di Ira ny er an oc rd ,R ep ag ha ha an an m In In al Lib Ke al nz Jo or Za N G G en ar Ug Ug M M M Ta ic m N Ye Sources: WDR 2018 team, using reading and mathematics data for Kenya and Uganda from Uwezo, Annual Assessment Reports, 2015 (http://www.uwezo .net/); reading and mathematics data for rural India from ASER Centre (2017); reading data for all other countries from U.S. Agency for International Development (USAID), Early Grade Reading Barometer, 2017, accessed May 30, 2017 (http://www.earlygradereadingbarometer.org/); and mathematics data for all other countries from USAID/RTI Early Grade Mathematics Assessment intervention reports, 2012–15 (https://shared.rti.org/sub-topic/early -grade-math-assessment-egma). Data at http://bit.do/WDR2018-Fig_O-1. Note: These data typically pertain to selected regions in the countries and are not necessarily nationally representative. Data for India pertain to rural areas. 2010, from 2.0 to 7.2 years.7 By 2010 the average worker of grade 6 students in Southern and East Africa were in Bangladesh had completed more years of school- able to go beyond the level of simply deciphering ing than the typical worker in France in 1975.8 This words, and less than 40 percent got beyond basic progress means that most enrollment gaps in basic numeracy.14 Among grade 6 students in West and education are closing between high- and low-income Central Africa in 2014, less than 45 percent reached countries. By 2008 the average low-income country the “sufficient” competency level for continuing stud- was enrolling students in primary school at nearly ies in reading or mathematics—for example, the rest the same rate as the average high-income country. could not answer a math problem that required them But schooling is not the same as learning.9 Chil- to divide 130 by 26.15 In rural India in 2016, only half of dren learn very little in many education systems grade 5 students could fluently read text at the level around the world: even after several years in school, of the grade 2 curriculum, which included sentences millions of students lack basic literacy and numeracy (in the local language) such as “It was the month skills. In recent assessments in Ghana and Malawi, of rains” and “There were black clouds in the sky.”16 more than four-fifths of students at the end of grade 2 These severe shortfalls constitute a learning crisis. were unable to read a single familiar word such Although not all developing countries suffer from as the or cat (figure O.1).10 Even in Peru, a middle- such extreme shortfalls, many are far short of the lev- income country, that share was half before the recent els they aspire to. According to leading international reforms.11 When grade 3 students in Nicaragua were assessments of literacy and numeracy—Progress in tested in 2011, only half could correctly solve 5 + 6.12 International Reading Literacy Study (PIRLS) and In urban Pakistan in 2015, only three-fifths of grade 3 Trends in International Mathematics and Science students could correctly perform a subtraction such Study (TIMSS)—the average student in low-income as 54 – 25, and in rural areas only just over two-fifths countries performs worse than 95 percent of the stu- could.13 dents in high-income countries, meaning that student This slow start to learning means that even stu- would be singled out for remedial attention in a class dents who make it to the end of primary school do not in high-income countries.17 Many high-performing master basic competencies. In 2007, the most recent students in middle-income countries—young men year for which data are available, less than 50 percent and women who have risen to the top quarter of Overview | 5 Figure O.2 In several countries, the 75th percentile of PISA test takers performs below the 25th percentile of the OECD average Performance of 25th, 50th, and 75th percentiles in 2015 PISA mathematics assessment, selected countries Percentile 75th 600 50th 550 500 25th Mathematics score 450 400 350 300 ic Ko a Tu o sia YR do il sia an ol ru ta a av m Fe age n nd y a, d p. ng an e ca a z an or tio v ri i n tvi bl a C mb Pe Re a so Ri rd la Ko nla p ,F ge ni ne EC ietn ap pu La Br ra m er Ja Ire Jo o ia Al er Fi de Re V os on re G In C Si D an ed n ic ac ia O in ss M m Ru Do OECD interquartile range Source: WDR 2018 team, using data from Programme for International Student Assessment (PISA) 2015 (OECD 2016). Data at http://bit.do/WDR2018-Fig_O-2. their cohorts—would rank in the bottom quarter in a income countries as well, with disadvantaged stu- wealthier country. In Algeria, the Dominican Republic, dents greatly overrepresented among the low scorers. and Kosovo, the test scores of students at the cutoff Costa Rica and Qatar have the same average score on for the top quarter of students (the 75th percentile of one internationally benchmarked assessment (TIMSS the distribution of PISA test takers) are well below the 2015)—but the gap between the top and bottom quar- cutoff for the bottom quarter of students (25th per- ters of students is 138 points in Qatar, compared with centile) of Organisation for Economic Co-operation 92 points in Costa Rica. The gap between the top and and Development (OECD) countries (figure O.2). Even bottom quarters in the United States is larger than in Costa Rica, a relatively strong performer in educa- the gap in the median scores between Algeria and the tion, performance at the cutoff for the top quarter of United States. students is equal to performance at the cutoff for the Students often learn little from year to year, but bottom quarter in Germany. early learning deficits are magnified over time. Stu- The learning crisis amplifies inequality: it severely dents who stay in school should be rewarded with hobbles the disadvantaged youth who most need the steady progress in learning, whatever disadvantages boost that a good education can offer. For students in they have in the beginning. And yet in Andhra Pradesh, many African countries, the differences by income India, in 2010, low-performing students in grade 5 level are stark (figure O.3). In a recent assessment were no more likely to answer a grade 1 question cor- (Programme d’Analyse des Systèmes Éducatifs de la rectly than those in grade 2. Even the average student Confemen, PASEC, 2014) administered at the end of in grade 5 had about a 50 percent chance of answering the primary cycle, only 5 percent of girls in Camer- a grade 1 question correctly—compared with about 40 oon from the poorest quintile of households had percent in grade 2.19 In South Africa in the late 2000s, learned enough to continue school, compared with 76 the vast majority of students in grade 4 had mastered percent of girls from the richest quintile.18 Learning only the mathematics curriculum from grade 1; most gaps in several other countries—Benin, the Republic of those in grade 9 had mastered only the mathemat- of Congo, and Senegal—were nearly as wide. Large ics items from grade 5.20 In New Delhi, India, in 2015, gaps among learners afflict many high- and middle- the average grade 6 student performed at a grade 3 6 | World Development Report 2018 Figure O.3 Children from poor households in Africa typically learn much less Percentage of grade 6 PASEC test takers in 2014 who scored above (blue) and below (orange) the sufficiency level on reading achievement: poorest and richest quintiles by gender, selected countries 100 75 50 25 Percent 0 25 50 75 100 F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M F M Poor Rich Poor Rich Poor Rich Poor Rich Poor Rich Poor Rich Poor Rich Poor Rich Poor Rich Niger Togo Cameroon Congo, Rep. Benin Côte d’Ivoire Burkina Faso Senegal Burundi Not competent Low competency High competency Source: WDR 2018 team, using data from World Bank (2016b). Data at http://bit.do/WDR2018-Fig_O-3. Note: Socioeconomic quintiles are defined nationally. “Not competent” refers to levels 0–2 in the original coding and is considered below the sufficiency level for school continuation; “low competency” refers to level 3; and “high competency” refers to level 4. F = female; M = male; PASEC = Programme d’Analyse des Systèmes Éducatifs de la Confemen. level in math. Even by grade 9, the average student Figure O.4 Students often learn little from year had reached less than a grade 5 level, and the gap to year, and early learning deficits are magnified between the better and worse performers grew over over time time (figure O.4). In Peru and Vietnam—one of the low- Assessed grade-level performance of students relative to enrolled grade, est and one of the highest performers, respectively, on New Delhi, India (2015) the PISA assessment of 15-year-old students—5-year- a. Mathematics b. Language olds start out with similar math skills, but students 9 9 in Vietnam learn much more for each year of school- ing at the primary and lower secondary levels.21 Although some countries are making progress 8 8 on learning, their progress is typically slow. Even the Grade-level performance Grade-level performance middle-income countries that are catching up to the 7 7 top performers are doing so very slowly. Indonesia has registered significant gains on PISA over the last 6 6 10–15 years. And yet, even assuming it can sustain its 2003–15 rate of improvement, Indonesia won’t reach the OECD average score in mathematics for another 5 5 48 years; in reading, for 73. For other countries, the wait could be even longer: based on current trends, 4 4 it would take Tunisia over 180 years to reach the OECD average for math and Brazil over 260 years 3 3 to reach the OECD average for reading. Moreover, 6 7 8 9 6 7 8 9 these calculations are for countries where learning Enrolled grade Enrolled grade has improved. Across all countries participating in Expected performance 75th percentile multiple rounds of PISA since 2003, the median gain Average assessed performance 25th percentile in the national average score from one round to the Source: WDR 2018 team, using data from Muralidharan, Singh, and Ganimian (2016). Data at http://bit.do next was zero. /WDR2018-Fig_O-4. Overview | 7 Figure O.5 The percentage of primary school students who pass a minimum proficiency threshold is often low Median percentage of students in late primary school who score above a minimum proficiency level on a learning assessment, by income group and region 100 80 60 Percent 40 20 0 Low- Lower-middle- Upper-middle- High- Sub-Saharan Middle East Latin East Asia Europe and income income income income Africa and America and Pacific Central Asia countries countries countries countries North Africa and the Caribbean Mathematics Reading Source: WDR 2018 team, using “A Global Data Set on Education Quality” (2017), made available to the team by Nadir Altinok, Noam Angrist, and Harry Anthony Patrinos. Data at http://bit.do /WDR2018-Fig_O-5. Note: Bars show the unweighted cross-country median within country grouping. Regional averages exclude high-income countries. India and China are among the countries excluded for lack of data. Minimum proficiency in mathematics is benchmarked to the Trends in International Mathematics and Science Study (TIMSS) assessment and in reading to the Progress in International Reading Literacy Study (PIRLS) assessment. Minimum proficiency in mathematics means that students have some basic mathematical knowledge such as adding or subtracting whole numbers, recognizing familiar geometric shapes, and reading simple graphs and tables (Mullis and others 2016). Minimum proficiency in reading means that students can locate and retrieve explicitly stated detail when reading literary texts and can locate and reproduce explicitly stated information from the beginning of informational texts (Mullis and others 2012). Because of this slow progress, more than 60 per- school. In 2016, 61 million children of primary school cent of primary school children in developing coun- age—10 percent of all children in low- and lower- tries still fail to achieve minimum proficiency in learn- middle-income countries—were not in school, along ing, according to one benchmark. No single learning with 202 million children of secondary school age.24 assessment has been administered in all countries, Children in fragile and conflict-affected countries but combining data from learning assessments in 95 accounted for just over a third of these, a dispropor- countries makes it possible to establish a globally com- tionate share. In the Syrian Arab Republic, which parable “minimum proficiency” threshold in math.22 achieved universal primary enrollment in 2000, the Below this threshold, students have not mastered even civil war had driven 1.8 million children out of school basic mathematical skills, whether making simple by 2013.25 Almost all developing countries still have computations with whole numbers, using fractions pockets of children from excluded social groups who or measurements, or interpreting simple bar graphs. do not attend school. Poverty most consistently pre- In high-income countries, nearly all students—99 per- dicts failing to complete schooling, but other charac- cent in Japan, 98 percent in Norway, 91 percent in Aus- teristics such as gender, disability, caste, and ethnicity tralia—achieve this level in primary school.23 But in also frequently contribute to school participation other parts of the world the share is much lower: just shortfalls (figure O.6). 7 percent in Mali, 30 percent in Nicaragua, 34 percent But it’s not just poverty and conflict that keep in the Philippines, and 76 percent in Mexico. In low- children out of school; the learning crisis does, too. income countries, 14 percent of students reach this When poor parents perceive education to be of low level near the end of primary school, and in lower- quality, they are less willing to sacrifice to keep their middle-income countries 37 percent do (figure O.5). children in school—a rational response, given the Even in upper-middle-income countries only 61 per- constraints they face.26 Although parental perceptions cent reach this minimum proficiency. of school quality depend on various factors, from the The ultimate barrier to learning is no schooling at physical condition of schools to teacher punctuality, all—yet hundreds of millions of youth remain out of parents consistently cite student learning outcomes 8 | World Development Report 2018 Figure O.6 School completion is higher for richer and urban families, but gender gaps are more context-dependent Gaps in grade 6 completion rates (percent) for 15- to 19-year-olds, by wealth, location, and gender a. Richest–poorest b. Urban–rural c. Male–female 80 80 80 Percentage point gap between Percentage point gap between Percentage point gap between richest and poorest quintiles 60 60 60 male and female urban and rural 40 40 40 20 20 20 0 0 0 −20 −20 −20 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 Overall grade 6 completion rate (%) Overall grade 6 completion rate (%) Overall grade 6 completion rate (%) Source: WDR 2018 team, using data from Filmer (2016). Data at http://bit.do/WDR2018-Fig_O-6. Note: The data presented are the latest available by country, 2005–14. Each vertical line indicates the size and direction of the gap for a country. as a critical component.27 These outcomes can affect the foundational cognitive skills are essential, and behavior: holding student ability constant, students systems cannot bypass the challenges of developing in the Arab Republic of Egypt who attended poorer- them as they target higher-order skills. performing schools were more likely to drop out.28 Tackling the learning crisis and skills gaps requires Learning shortfalls during the school years even- diagnosing their causes—both their immediate causes tually show up as weak skills in the workforce. Thus at the school level and their deeper systemic drivers. the job skills debate reflects the learning crisis. Work Given all the investments countries have made in skill shortages are often discussed in a way that is education, shortfalls in learning are discouraging. But disconnected from the debate on learning, but the one reason for them is that learning has not always two are parts of the same problem. Because education received the attention it should have. As a result, systems have not prepared workers adequately, many stakeholders lack actionable information about what enter the labor force with inadequate skills. Measur- is going wrong in their schools and in the broader ing adult skills in the workplace is hard, but recent society, and so they cannot craft context-appropriate initiatives have assessed a range of skills in the adult responses to improve learning. Acting effectively populations of numerous countries. They found that requires first understanding how schools are failing even foundational skills such as literacy and numer- learners and how systems are failing schools. acy are often low, let alone the more advanced skills. The problem isn’t just a lack of trained workers; it is Schools are failing learners a lack of readily trainable workers. Accordingly, many Struggling education systems lack workers end up in jobs that require minimal amounts one or more of four key school-level of reading or math.29 Lack of skills reduces job quality, ingredients for learning: prepared earnings, and labor mobility. learners, effective teaching, learning- The skills needed in labor markets are multi­ focused inputs, and the skilled man- dimensional, so systems need to equip students with agement and governance that pulls far more than just reading, writing, and math—but them all together (figure O.7). The students cannot leapfrog these foundational skills. next section looks at why these links Problem break down; here the focus is on how dimension 2: Whether as workers or members of society, peo- Immediate ple also need higher-order cognitive skills such as they break down. causes problem-solving. In addition, they need socioemo- First, children often arrive in tional skills—sometimes called soft or noncognitive school unprepared to learn—if they skills—such as conscientiousness. Finally, they need arrive at all. Malnutrition, illness, low parental technical skills to perform a specific job. That said, investments, and the harsh environments associated Overview | 9 Figure O.7 Why learning doesn’t happen: Second, teachers often lack the skills or motiva- Four immediate factors that break down tion to be effective. Teachers are the most important factor affecting learning in schools. In the United States, students with great teachers advance 1.5 grade levels or more over a single school year, compared te d with just 0.5 grade levels for those with an ineffective iva Un ot pr teacher.34 In developing countries, teacher quality m le ep un rs ar can matter even more than in wealthier countries.35 e n ch d ar s But most education systems do not attract applicants an ed er a l ed te with strong backgrounds. For example, 15-year-old kil students who aspire to be teachers score below the Uns national average on PISA in nearly all countries.36 Beyond that, weak teacher education results in teach- ers lacking subject knowledge and pedagogical skills. LEARNING In 14 Sub-Saharan countries, the average grade 6 teacher performs no better on reading tests than do the highest-performing students from that grade.37 In S c h at Indonesia, 60 percent of the time in a typical mathe- ts o o do a t h hin matics class is spent on lecturing, with limited time te a pu ng ar ct l an e m ni in remaining for practical work or problem-solving.38 c ag l o o n’ a sn e Meanwhile, in many developing countries substantial le em t ’t h g nd a ent S c t do an d le e ar ct a th ing amounts of learning time are lost because classroom nin ch g te a time is spent on other activities or because teachers are absent. Only a third of total instructional time was used in Ethiopia, Ghana, and Guatemala.39 Across seven African countries, one in five teachers was Source: WDR 2018 team. absent from school on the day of an unannounced visit by survey teams, with another fifth absent from the classroom even though they were at school with poverty undermine early childhood learning.30 (figure O.9).40 The problems are even more severe in Severe deprivations—whether in terms of nutri- remote communities, amplifying the disadvantages tion, unhealthy environments, or lack of nurture by already facing rural students. Such diagnostics are caregivers—have long-lasting effects because they not intended to blame teachers. Rather, they call impair infants’ brain development.31 Thirty percent attention to how systems undermine learning by of children under 5 in developing countries are failing to support them. physically stunted, meaning they have low height Third, inputs often fail to reach classrooms or to for their age, typically due to chronic malnutrition.32 affect learning when they do. Public discourse often The poor developmental foundations and lower levels equates problems of education quality with input of preschool skills resulting from deprivation mean gaps. Devoting enough resources to education is cru- many children arrive at school unprepared to benefit cial, and in some countries resources have not kept fully from it (figure O.8).33 So even in a good school, pace with the rapid jumps in enrollment. For several deprived children learn less. Moreover, breaking reasons, however, input shortages explain only a small out of lower learning trajectories becomes harder as part of the learning crisis. First, looking across systems these children age because the brain becomes less and schools, similar levels of resources are often asso- malleable. Thus education systems tend to amplify ciated with vast differences in learning outcomes.41 initial differences. Moreover, many disadvantaged Second, increasing inputs in a given setting often has youth are not in school. Fees and opportunity costs small effects on learning outcomes.42 Part of the rea- are still major financial barriers to schooling, and son is that inputs often fail to make it to the front lines. social dimensions of exclusion—for example, those A decade ago in Sierra Leone, for example, textbooks associated with gender or disability—exacerbate the were distributed to schools, but follow-up inspec- problem. These inequalities in school participation tions found most of them locked away in cupboards, further widen gaps in learning outcomes. unused.43 Similarly, many technological interventions 10 | World Development Report 2018 Figure O.8 Socioeconomic gaps in cognitive achievement grow with age—even in preschool years Percentage of children ages 3–5 who can recognize 10 letters of the alphabet, by wealth quintile, selected countries a. Central African Republic b. Kazakhstan c. Tunisia 100 100 100 80 80 80 60 60 60 Percent Percent Percent 40 40 40 20 20 20 0 0 0 3 3.5 4 4.5 5 3 3.5 4 4.5 5 3 3.5 4 4.5 5 Age (years) Age (years) Age (years) Richest quintile Poorest quintile 95% confidence interval 95% confidence interval Source: WDR 2018 team, using data from Multiple Indicator Cluster Surveys (http://mics.unicef.org/). Data are for 2010 for the Central African Republic, 2010–11 for Kazakhstan, and 2012 for Tunisia. Data at http://bit.do/WDR2018-Fig_O-8. fail before they reach classrooms, and even when they Figure O.9 In Africa, teachers are often do make it to classrooms, they often do not enhance absent from school or from classrooms teaching or learning. In Brazil, a One Laptop Per Child while at school initiative in several states faced years of delays. Then, Percentage of teachers absent from school and from class on even a year after the laptops finally made it to class- the day of an unannounced visit, participating countries rooms, more than 40 percent of teachers reported never or rarely using them in classroom activities.44 60 Fourth, poor management and governance often undermine schooling quality. Although effective 40 school leadership does not raise student learning Percent directly, it does so indirectly by improving teaching quality and ensuring effective use of resources.45 20 Across eight countries that have been studied, a 1.00 standard deviation increase in an index of manage- ment capacity—based on the adoption of 20 man- 0 agement practices—is associated with a 0.23–0.43 13 14 0 12 4 3 11 3 01 01 01 01 20 20 20 20 standard deviation increase in student outcomes. ,2 ,2 ,2 ,2 a, e, a, l, ia ia go ia ga qu nd ny an an er But school management capacity tends to be lowest To ne Ke bi a ig nz nz Ug Se am N Ta Ta in those countries with the lowest income levels, oz M classroom and management capacity is substantially lower in Teachers absent from the school schools than in manufacturing (figure O.10).46 Inef- fective school leadership means school principals Source: Bold and others (2017). Data at http://bit.do/WDR2018-Fig_O-9. are not actively involved in helping teachers solve Note: “Absent from the classroom” combines absences from school with absences from class among teachers who are at school. Data are from the problems, do not provide instructional advice, and World Bank’s Service Delivery Indicators (SDI) surveys (http://www do not set goals that prioritize learning. School gover- .worldbank.org/sdi). nance—particularly the decision-making autonomy of schools, along with the oversight provided by autonomy, and community engagement fails to affect parents and communities—serves as the framework what happens in classrooms.47 for seeking local solutions and being accountable for Because these quality problems are concentrated them. In many settings, schools lack any meaningful among disadvantaged children, they amplify social Overview | 11 Figure O.10 Management capacity typically disadvantage marginalized communities, is low in schools in low- and middle- but also that resources are used less effectively there, income countries exacerbating the problem. Public policy thus has the effect of widening social gaps rather than offering all Distribution of management scores by sector, participating countries children an opportunity to learn. Systems are failing schools Tanzania Viewed from a systems perspec- tive, the low level of learning and India skills should come as no surprise. Technical complexities and polit- ical forces constantly pull edu- Haiti cation systems out of alignment with learning (figure O.11). Problem Italy dimension 3: Deeper causes Technical challenges: Brazil Reorienting toward learning is hard Complex systems and limited management capacity are obstacles to orienting all parts of an education Mexico system toward learning. First, the various parts of the system need to be aligned toward learning. But Germany actors in the system have other goals—some stated, some not. Promoting learning is only one of these, and not necessarily the most important one. At times, Canada these other goals can be harmful, such as when con- struction firms and bureaucrats collude to provide substandard school buildings for their financial gain. United States At other times, these goals may be laudable, such as nurturing shared national values. But if system ele- Sweden ments are aligned toward these other goals, they will sometimes be at cross-purposes with learning. Even when countries want to prioritize learning, United Kingdom they often lack the metrics to do so. Every system assesses student learning in some way, but many 1 2 3 4 5 systems lack the reliable, timely assessments needed Management score to provide feedback on innovations. For example, is Education Manufacturing a new teacher training program actually making teachers more effective? If the system lacks reliable Sources: Bloom and others (2014, 2015); Lemos and Scur (2016), with updates. information on the quality of teaching and the learn- Data at http://bit.do/WDR2018-Fig_O-10. ing of primary students—comparable across time or Note: The underlying distributions for the education data are shown as classrooms—there is no way to answer that question. bars; for both sectors, the smoothed distributions are shown as curves. The indexes are constructed from the nine items that are comparable across To be truly aligned, parts of the education system sectors. Data on manufacturing are not available for Haiti. also have to be coherent with one another. Imagine that a country has set student learning as a top prior- inequalities. In low-income countries, on average, ity and that it has in place reasonable learning metrics. stunting rates among children under 5 are almost It still needs to leap a major technical hurdle, how- three times higher in the poorest quintile than in ever: ensuring that system elements work together. the richest.48 In schools, problems with teacher If a country adopts a new curriculum that increases absenteeism, lack of inputs, and weak management emphasis on active learning and creative thinking, are typically severest in communities that serve the that alone will not change much. Teachers need to poorest students. It’s not just that spending patterns be trained so that they can use more active learning 12 | World Development Report 2018 methods, and they need to care enough to make the Figure O.11 Technical and political factors divert change because teaching the new curriculum may schools, teachers, and families from a focus on be much more demanding than the old rote learning learning methods. Even if teachers are on board with curricu- lum reform, students could weaken its effects if an unreformed examination system creates misaligned incentives. In Korea, the high-stakes exam system for university entrance has weakened efforts to rs Le e ar reorient secondary school learning. The curriculum ch ne a has changed to build students’ creativity and socio- Te rs emotional skills, but many parents still send their chil- dren to private “cram schools” for test preparation.49 The need for coherence makes it risky to borrow system elements from other countries. Education pol- icy makers and other experts often scrutinize systems LEARNING that have better learning outcomes to identify what they could borrow. Indeed, in the 2000s the search for the secret behind Finland’s admirable record of learn- Sch ing with equity led to a swarm of visiting delegations in what the Finns dubbed “PISA tourism.” Finland’s oo ts pu system gives considerable autonomy to its well- lm na in educated teachers, who can tailor their teaching to l a ge h oo the needs of their students. But lower-performing me Sc systems that import Finland’s teacher autonomy nt into their own contexts are likely to be disappointed: if teachers are poorly educated, unmotivated, and loosely managed, giving them even more autonomy Source: WDR 2018 team. will likely make matters worse. South Africa discov- ered this in the 1990s and 2000s when it adopted a curriculum approach that set goals but left implemen- politicians and teachers happy than on promoting tation up to teachers.50 The approach failed because student learning, or they may simply try to protect it proved to be a poor fit for the capacity of teachers their own positions. Some private suppliers of edu- and the resources at their disposal.51 Home-grown, cation services—whether textbooks, construction, or context-specific solutions are important. schooling—may, in the pursuit of profit, advocate pol- Successful systems combine both alignment and icy choices not in the interest of students. Teachers coherence. Alignment means that learning is the goal and other education professionals, even when moti- of the various components of the system. Coherence vated by a sense of mission, also may fight to main- means that the components reinforce each other in tain secure employment and to protect their incomes. achieving whatever goals the system has set for them. None of this is to say that education actors don’t care When systems achieve both, they are much more likely about learning. Rather, especially in poorly managed to promote student learning. Too much misalignment systems, competing interests may loom larger than or incoherence leads to failure to achieve learning, the learning-aligned interests (table O.2). though the system might achieve other goals (table O.1). Misalignments aren’t random. Because of these competing interests, the choice of a particular policy Political challenges: Key players don’t is rarely determined by whether it improves learning. always want to prioritize student learning More often, the choice is made by the more powerful Political challenges compound technical ones. Many actors in the policy arena. Agents are accountable to education actors have different interests, again one another for different reasons, not just learning. beyond learning. Politicians act to preserve their Given these interests, it should come as no surprise positions in power, which may lead them to target that little learning often results. particular groups (geographic, ethnic, or economic) One problem is that activities to promote learning for benefits. Bureaucrats may focus more on keeping are difficult to manage. Teaching and learning in the Overview | 13 Table O.1 Alignment and coherence both matter Are system Coherent? elements . . . Yes No Yes High performance: Systems well organized to Incoherent strivers: Systems incoherently oriented promote learning toward learning Examples: High performers at each level Examples: Countries that borrow learning-oriented (Shanghai [China], Finland, Vietnam) “best practice” elements but do not ensure that the Aligned toward various elements are coherent with each other learning? No Coherent nonlearners: Systems well organized to Failed systems: Systems that are not trying to promote a different goal achieve learning or anything else in a coherent Examples: Totalitarian or authoritarian systems way focused on promoting loyalty to the state or nation Examples: Systems in failed states building (Stalin-era USSR, Suharto-era Indonesia); systems that focus on school attainment rather than learning (many systems) Source: WDR 2018 team. Table O.2 Multiple interests govern the actions of education stakeholders Examples of . . . Stakeholders Learning-aligned interests Competing interests Teachers Student learning, professional ethic Employment, job security, salary, private tuitions Principals Student learning, teacher performance Employment, salary, good relations with staff, favoritism Bureaucrats Well-functioning schools Employment, salary, rent-seeking Politicians Well-functioning schools Electoral gains, rent-seeking, patronage Parents and students Student learning, employment of Family employment, family income, graduates outdoing others Judiciary Meaningful right to education Favoritism, rent-seeking Employers Skilled graduates Low taxes, narrowly defined self-interests Nongovernment schools (religious, Innovative, responsive schooling Profit, religious mission, funding nongovernmental, for-profit) Suppliers of educational inputs (e.g., High-quality, relevant inputs Profit, influence textbooks, information technology, buildings) International donors Student learning Domestic strategic interests, taxpayer support, employment Source: WDR 2018 team. classroom involve significant discretion by teachers, at simple, easily collected enrollment data. Similarly, as well as regular and repeated interactions between school construction, cash transfer programs, teacher students and teachers.52 These characteristics, cou- hiring, and school grant programs intended to pled with a dearth of reliable information on learn- expand access are all highly visible, easily monitored ing, make managing learning more difficult than investments. pursuing other goals.53 For example, improvements The potential beneficiaries of better foundational in access to education can be monitored by looking learning—such as students, parents, and employers— 14 | World Development Report 2018 often lack the organization, information, or short- other parts of the system, such as higher education or term incentive to press for change. Parents are usually lifelong learning. In these areas, too, many countries not organized to participate in debates at the system suffer from a lack of attention to outcomes, wide gaps level, and they may lack knowledge of the potential in opportunity, and systemic barriers to resolving gains from different policies to improve learning.54 these problems. They also may worry about the potential ramifica- tions for their children or themselves of opposing Still, there are reasons for hope interests such as teachers, bureaucrats, or politicians. Even in countries that seem stuck in low-learning Students have even less power—except sometimes in traps, some teachers and schools manage to strengthen higher education, where they can threaten demon- learning. These examples may not be sustainable—and strations—and, like parents, they may be unaware of they are not likely to spread systemwide without how little they are learning until they start looking efforts to reorient the system toward learning—but for work. Finally, the business community, even if it systems willing to learn from these outliers can bene- suffers from a shortage of skilled graduates to hire, fit. On a larger scale, some regions within countries are often fails to advocate for quality education, instead more successful in promoting learning, as are some lobbying for lower taxes and spending. By contrast countries at each income level. to these potential beneficiaries of reform, the poten- These examples reveal that higher-level system tial losers tend to be more aware of what is at stake equilibriums exist. But is it possible for a whole for them and, in many cases, better organized to act system to escape the low-learning trap, moving collectively. to a better one? There are at least two reasons for As a result, many systems are stuck in low- optimism. First, as countries innovate to improve learning traps, characterized by low accountabil- learning, they can draw on more systematic knowl- ity and high inequality. These traps bind together edge than ever available before about what can work key stakeholders through informal contracts that at the micro level—the level of learners, classrooms, prioritize other goals such as civil service employ- and schools. A number of interventions, innovations, ment, corporate profits, or reelection, perpetuating and approaches have resulted in substantial gains the low-accountability equilibrium. In better-run in learning. These promising approaches come in systems, actors such as bureaucrats and teachers many flavors—new pedagogical methods, ways to can devote much of their energy to improving ensure that students and teachers are motivated, outcomes for students. But in low-learning traps approaches to school management, technologies to those same actors lack either the incentives or the enhance teaching learning—and they may not pay support needed to focus on learning. Instead, they off in all contexts, but the fact that it is possible to are constantly pressured to deliver other services improve learning outcomes should give hope. These for more powerful players. As actors juggle multiple interventions can provide substantial improvements objectives, relying on each other in an environment in learning: almost one or two grade-equivalents for of uncertainty, low social trust, and risk aversion, it some students.55 Even though successful interven- is often in the interest of each to maintain the status tions cannot be imported wholesale into new con- quo—even if society, and many of these actors, would texts, countries can use them as starting points for be better off if they could shift to a higher-quality their own innovations. equilibrium. Second, some countries have implemented This diagnosis has concentrated on the shortfalls reforms that have led to sustained systemwide in foundational learning, as will the priorities for improvements in learning. Finland’s major education action discussed in the next section. However, this reform in the 1970s famously improved the equity focus should not be interpreted as a statement that of outcomes while also increasing quality, so that by other areas are unimportant. Education systems and the time of the first PISA in 2000, Finland topped the their enabling environment are broader and more assessment. More recently, Chile, Peru, Poland, and complex than this Report can cover, so our priority the United Kingdom have made serious, sustained here is to highlight what can be done most immedi- commitments to reforming the quality of their ately to strengthen the foundations of learning on education systems. In all these countries, learning which all successful systems are built. But both the has improved over time—not always steadily, but diagnosis and the priorities for action are relevant for enough to show that system-level reforms can pay off. Overview | 15 The education systems in Shanghai (China) and Assess learning—to make it a serious goal Vietnam today—and Korea decades ago—show that “What gets measured gets man- it is possible to perform far better than income levels aged.” “Just weighing the pig would predict, thanks to a sustained focus on learning doesn’t make it fatter.” There is with equity. Brazil and Indonesia have made consid- some truth to both of these say- erable progress, despite the challenges of reforming ings. Lack of measurement makes large, decentralized systems. it hard to know where things are, where they are going, and what actions are making any difference. Policy How to realize education’s Knowing these things can provide response 1: Assess promise: Three policy focus and stimulate action. But learning measurement that is too removed responses from action can lead nowhere. The challenge is strik- Learning outcomes won’t change unless education ing a balance—finding the right measures for the systems take learning seriously and use learning as right purposes and implementing them within an a guide and metric. This idea can be summarized as appropriate accountability framework. “all for learning.”56 As this section explains, a com- mitment to all for learning—and thus to learning for Use measurement to shine a light on all—implies three complementary strategies: learning The first step to improving systemwide learning is • Assess learning—to make it a serious goal. Measure and to put in place good metrics for monitoring whether track learning better; use the results to guide action. programs and policies are delivering learning. Cred- • Act on evidence—to make schools work for all learners. ible, reliable information can shape the incentives Use evidence to guide innovation and practice. facing politicians. Most notably, information on stu- • Align actors—to make the whole system work for learn- dent learning and school performance—if presented ing. Tackle the technical and political barriers to in a way that makes it salient and acceptable—fosters learning at scale. healthier political engagement and better service delivery. Information also helps policy makers man- These three strategies depend on one another. age a complex system. Adopting a learning metric without any credible way Measuring learning can improve equity by to achieve learning goals will simply lead to frustra- revealing hidden exclusions. As emphasized at the tion. School-level innovations without a learning outset of this overview, the learning crisis is not just metric could take schools off course, and without the a problem for the society and economy overall; it is system-level support they could prove ephemeral. also a fundamental source of inequities and widening And system-level commitment to learning without gaps in opportunity. But because reliable information school-level innovation, and without learning mea- on learning is so spotty in many education systems, sures to guide the reforms, is unlikely to amount to especially in primary and lower secondary schools, more than aspirational rhetoric. But together, the the way the system is failing disadvantaged children three strategies can create change for the better. is a hidden exclusion.58 Unlike exclusion from school, The potential payoff is huge. When children have a lack of learning is often invisible, making it impos- growth mindset, meaning they understand their own sible for families and communities to exercise their great learning potential, they learn much more than right to quality education. when they believe they are constrained by a fixed These measures of learning will never be the only intelligence.57 Societies have the same opportunity. By guide for educational progress, nor should they be. adopting a social growth mindset—recognizing the Education systems should have ways of tracking barriers to learning, but also the very real opportu- progress toward any goal they set for themselves and nities to break them down—they can make progress their students—not just learning. Systems should also on learning. One overarching priority should be to track the critical factors that drive learning—such as end the hidden exclusion of low learning. This is not learner preparation, teacher skills, quality of school just the right thing to do; it is also the surest way to management, and the level and equity of financing. improve average learning levels and reap education’s But learning metrics are an essential starting point full rewards for society as a whole. for improving lagging systems. 16 | World Development Report 2018 Figure O.12 Many countries lack information on learning outcomes Percentage of countries with data to monitor progress toward the Sustainable Development Goals for learning by the end of primary or lower secondary school 100 80 60 Percent 40 20 0 Primary Lower Primary Lower Primary Lower Primary Lower Primary Lower secondary secondary secondary secondary secondary Latin America Arab states Sub-Saharan Asia and World and the Africa Pacific Caribbean Mathematics Reading Source: UIS (2016). Data at http://bit.do/WDR2018-Fig_O-12. Note: Regional groupings follow UNESCO definitions. There is too little measurement of learning, progress toward the United Nations’ Sustainable not too much Development Goals found that of the 121 countries A recommendation to start tackling the learning cri- studied, a third lack the data required to report on sis with more and better measurement of learning the levels of reading and mathematics proficiency may seem jarring. Many education debates highlight of children at the end of primary school.62 Even the risks of overtesting or an overemphasis on tests. more lack data for the end of lower secondary school In the United States, two decades of high-stakes test- (figure O.12). Even when countries have these data, ing have led to patterns of behavior consistent with they are often from one-off assessments that do not these concerns.59 Some teachers have been found to allow systematic tracking over time. A lack of good concentrate on test-specific skills instead of untested measurement means that education systems are subjects, and some schools have engaged in strategic often flying blind—and without even agreement on behavior to ensure that only the better-performing the destination. students are tested, such as assigning students to special education that excuses them from testing.60 In Use a range of metrics with one ultimate the extreme, problems have expanded to convictions goal for systemic cheating at the school district level.61 At Different learning metrics have different purposes, the same time, media coverage of education in many but each contributes to learning for all. Teachers low- and middle-income countries (and some high- assess students in classrooms every day—formally income ones) often focuses on high-stakes national or informally—even in poorly resourced, poorly man- examinations that screen candidates for tertiary edu- aged school systems. But using metrics properly to cation—raising concerns about an overemphasis on improve learning systemwide requires a spectrum of testing. types of assessment that, together, allow educators But in many systems the problem is too little and policy makers to use the right combination of focus on learning—not too much. Many countries teaching approaches, programs, and policies. lack information on even basic reading and math Formative assessment by teachers helps guide competencies. An assessment of capacity to monitor instruction and tailor teaching to the needs of Overview | 17 students. Well-prepared, motivated teachers do not raise awareness of how a country is falling short of its need to operate in the dark: they know how to assess peers in building human capital. the learning of students regularly, formally and Two other types of learning metrics measured informally. As the next section discusses, this type of in nonschool settings can be used to strengthen the regular check-in is important because many students quality and equity focus of assessment systems. lag so far behind that they effectively stop learning. Grassroots accountability movements—led by civil Knowing where students are allows teachers to society organizations such as the ASER Centre in India adjust their teaching accordingly and to give students and Uwezo in East Africa—have deployed citizen-led learning opportunities they can handle. Singapore assessments that recruit volunteers to measure the has successfully used this approach—identifying foundational learning of young children in their lagging students in grade 1 using screening tests and communities. These organizations then use their then giving them intensive support to bring them up learning data to advocate for education reform. Some to grade level.63 multipurpose household surveys also collect learning National and subnational learning assessments data, enabling researchers to analyze how learning provide system-level insights that classroom assess- outcomes correlate with income and community ments by teachers cannot. To guide an education variables. Both types of assessments are administered system, policy makers need to understand whether in people’s homes, not schools. As a result, they students are mastering the national curriculum, in don’t suffer from a key weakness of school-based which areas students are stronger or weaker, whether assessments: when marginal students drop out, their certain population groups are lagging behind and absence can improve the average scores on school by how much, and which factors are associated with assessments, thereby creating a perverse incentive better student achievement. There is no effective way for school leaders. But household-based assessments to aggregate the results of classroom-level formative yield learning metrics that reward systems for assessment by teachers into this type of reliable improving both access and quality. This is crucial to system-level information. This is why systems need ensuring that no child is written off. Even for students assessments of representative samples of students who are in school, household-based assessments pro- across wider jurisdictions, such as countries or prov- vide an alternative source of learning data, which can inces. Such assessments can be an especially import- be important in settings where official assessments ant part of tracking systemwide progress because are of questionable quality. they are anchored in a system’s own expectations for itself. And national assessments can provide a check Measurement can be hard on the quality of subnational assessments by flagging Why isn’t there more and better measurement of cases in which trends or levels of student achieve- learning? As with system barriers to learning, bar- ment diverge across the two. In the United States, riers to better measurement are both technical and the National Assessment of Educational Progress has political. From a technical perspective, conducting played this role.64 good assessments is not easy. At the classroom International assessments also provide informa- level, teachers lack the training to assess learning tion that helps improve systems. Globally bench- effectively, especially when assessments try to cap- marked student assessments such as PISA, TIMSS, ture higher-order skills—say, through project-based and PIRLS, as well as regionally benchmarked ones assessment—rather than rote learning. And at the such as PASEC in West and Central Africa and the system level, education ministries lack the capacity Latin American Laboratory for Assessment of the to design valid assessments and implement them in a Quality of Education (LLECE), provide an additional sample of schools. Political factors intrude as well. To perspective on how well students are learning. They paraphrase an old saying, policy makers may decide allow assessment of country performance in a way it is better to avoid testing and be assumed inef- that is comparable across countries, and they pro- fective than to test students and remove all doubt. vide a check on the information that emerges from And even when they do participate in assessments, national assessments. And international assessments governments sometimes decline to release the learn- can be powerful tools politically: because country ing results to the public, as happened with the 1995 leaders are concerned with national productivity and TIMSS in Mexico.65 Finally, if assessments are poorly competitiveness, international benchmarking can designed or inappropriately made into high-stakes 18 | World Development Report 2018 FIGURE O.13 Low-performing countries don’t face sharp tradeo s between learning and other education outp tests, administrators or educators may have an Figure O.13 Low-performing countries don’t incentive to cheat on them, rendering the assessment face sharp trade-offs between learning and results worthless as a guide to policy. other education outputs Measurement doesn’t need to detract from broader education objectives—it can even support them (e.g., creativity, citizenship) A stronger emphasis on measurable learning High-performing country B doesn’t mean that other education outcomes don’t Other outputs matter. Formal education and other opportunities for learning have many goals, only some of which are captured by the usual assessments of literacy, numeracy, and reasoning. Educators also aspire to High-performing help learners develop higher-order cognitive skills, country A including some (like creativity) that are hard to capture through assessments. Success in life also Low-performing country C depends on socioemotional and noncognitive skills— such as persistence, resilience, and teamwork—that a good education helps individuals develop. Education Measured learning systems often have other goals as well: they want to Source: WDR 2018 team. endow students with citizenship skills, encourage civic-minded values, and promote social cohesion. These are widely shared goals of education, and it is by high-performing countries on the education understandable that people will ask whether, espe- frontier. Economists use the concept of the produc- cially in education systems that are already over- tion possibilities frontier to understand how pro- burdened, increasing the emphasis on measurable ducers—or in this case countries—make trade-offs learning will crowd out these other goals. between the production of different goods. This In fact, a focus on learning—and on the educa- idea encapsulates the debates on education policy tional quality that drives it—is more likely to “crowd in OECD countries on the learning frontier (figure in” these other desirable outcomes. Conditions that O.13). For example, in recent years many stakehold- allow children to spend two or three years in school ers in Korea have argued that their high-performing without learning to read a single word, or to reach education system places too much emphasis on test the end of primary school without learning to do scores (called “measured learning” in figure O.13) and two-digit subtraction, are not conducive to reaching not enough on creativity and certain socioemotional the higher goals of education. Schools that cannot skills such as teamwork (“other outputs”). Implicitly, equip youth with relevant job skills usually will not this Korean debate is about whether to try to move prepare them to launch new companies or analyze up and to the left on the frontier—that is, from A great works of literature either. If students cannot toward B. But in the low-learning trap, represented focus because of deprivation, if teachers lack the by “low-performing country C” in the figure, there is pedagogical skills and motivation to engage students, so much slack and such a weak focus on outcomes if materials meant for the classroom never reach it that this OECD-driven debate is not relevant. Coun- because of poor management, and if the system as a try C has an opportunity to improve on both mea- whole is unmoored from the needs of society—well, is sured learning and other education outputs at the it really plausible to believe that students are develop- same time. An experiment in Andhra Pradesh, India, ing higher-order thinking skills like problem-solving that rewarded teachers for gains in measured learn- and creativity? It is more likely that these conditions ing in math and language led to more learning not undermine the quest for higher goals—and that, con- just in those subjects, but also in science and social versely, improving the learning focus would acceler- studies—even though there were no rewards for ate progress toward those goals as well. the latter.66 This outcome makes sense—after all, lit- Paradoxically, lower-performing countries proba- eracy and numeracy are gateways to education more bly do not face the same sharp trade-offs encountered generally. Overview | 19 Act on evidence—to make schools work for four times as much in Israel as it did in Kenya—and it all learners has had no impact in some contexts.72 In the words of Measurement of learning shortfalls two commentators on this literature: “Knowing ‘what doesn’t provide clear guidance on works’ in the sense of the treatment effect on the trial how to remedy them. Fortunately, population is of limited value without understanding there is now a lot of experience the political and institutional environment in which on ways to improve learning out- it is set.”73 comes at the student, classroom, The next section tackles the question of that and school levels. Cognitive neuro­ broader environment, but in the meantime we first Policy science has evolved dramatically address how to use this evidence most effectively. response 2: in the last two decades, providing There are four main considerations. Act on insights on how children learn.67 First, more important than the individual results evidence This work has revealed how import- from individual studies are the principles of how ant the first several years of life are to a child’s brain and why programs work. In economic terms, “princi- development.68 At the same time, schools and systems ples” correspond to models of behavior that can then around the world have innovated in many ways: by help guide broader sets of approaches to addressing deploying novel approaches to pedagogy, using new problems. Three types of models can prove especially technologies to enhance teaching and learning in insightful: straightforward models in which actors classrooms, or increasing the accountability, and some- maximize their welfare subject to the constraints times autonomy, of various actors in the system. The they face; principal-agent models that incorporate number of systematic evaluations of whether these multiple actors with different goals and perhaps dif- interventions have improved learning has increased ferent information; and behavioral models that factor more than 10-fold, from just 19 in 2000 to 299 in 2016.69 in mental models and social norms. Many interventions have succeeded in improving Second, a gap between what the evidence suggests learning outcomes. The learning gains from effec- may be effective and what is done in practice points tive interventions translate into additional years of to a potential entry point for action. Understanding schooling, higher earnings, and lower poverty. For a why gaps open up helps guide how to address them. group of stunted Jamaican children 9–24 months old, For example, when different actors face different a program to improve cognitive and socioemotional information, or some actors lack information, this development led to much better outcomes 20 years suggests drawing from approaches that show how later—lower crime rates, better mental health, and information can be disseminated and used better. earnings that were 25 percent higher than those of Gaps point to which types of principles should drive nonparticipants.70 Programs to improve pedagogy context-specific innovation. have had an impact greater than the equivalent of an Third, evidence tends to accumulate where it extra half a year of business-as-usual schooling and is easiest to generate, not necessarily where action an 8 percent increase in the present discounted value would make the most difference, so policies focused of lifetime earnings.71 So while tackling the learning only on that evidence might be misguided.74 Though crisis is hard, the fact that there are interventions that the scope of the accumulated evidence in education is improve learning suggests ways forward. broad, just because an approach hasn’t been evaluated This evidence base does not allow us to identify doesn’t mean it lacks potential. Context-specific inno- what works in all contexts because there are no vation may mean trying things that have not been global solutions in education. Improving learning in tried elsewhere. a particular setting will never be as simple as taking Fourth, a focus on underlying principles high- a successful program from one country or region and lights that the problem can’t be solved by one decision implementing it elsewhere. Randomized controlled maker simply prescribing an increase in the quantity, trials and other approaches to evaluate impact place or even the quality, of one or more inputs. Many of a premium on carefully isolating the causal impact the inputs in learning are the result of choices made of an intervention. But such approaches may ignore by the various actors—choices made in reaction to important interactions with underlying factors that the actual and anticipated choices of other actors. affect whether an intervention makes a difference— For example, teachers respond to incentives to attend factors that may not be at play when replicating the school and to improve student outcomes, even intervention in a new context. For example, increas- though the nature of the response varies across con- ing class size by 10 students reduced test scores by texts.75 Likewise, students and parents make choices 20 | World Development Report 2018 Figure O.14 It’s more complicated than it looks: People act in reaction to the choices of others throughout the system Civil so ns ciet cia y liti o rg Po an iz at io ns r Pe c to s Le er er ar se ch s/c a te ne a om Te rs Priv mun ities LEARNING Sch B u re ts oo pu y lm a uc an iar in ag o ol dic rat em ent S ch s In te Ju rn at ion o rs al a act c to r s O ther Source: WDR 2018 team. responding to other decisions. In India and Zambia, There are three key entry points to addressing learner government grants to schools led parents to reduce preparation: their own investments in their children’s schooling.76 All things considered, a more complete character- Set children on high-development trajectories through •  ization of the learning framework might be closer early childhood nutrition, stimulation, and care. Three to the one illustrated in figure O.14: learning how approaches stand out from successful experiences. to improve outcomes by intervening at the student, First, target mothers and their babies with health classroom, and school l ­ evels involves illuminating the and nutrition interventions during the first 1,000 various arrows. days to reduce malnutrition and foster physiolog- Putting all this together sheds light on three sets ical development. Second, increase the frequency of promising entry points: prepared learners, effec- and quality of stimulation and opportunities for tive teaching, and school-level interventions that learning at home (starting from birth) to improve actually affect the teaching and learning process. language and motor development, as well as to Each of these priority areas is founded on evidence cultivate early cognitive and socioemotional skills. from multiple contexts showing that it can make a Third, promote day-care centers for very young real difference for learning. children and preschool programs for children 3–6 years old—along with caregiver programs that Prepare children and youth for learning enhance the nurturing and protection of children— Getting learners to school ready and motivated to to improve cognitive and socioemotional skills in learn is a first step to better learning. Without it, other the short run, as well as education and labor market policies and programs will have a minimal effect. outcomes later in life.77 Program quality matters Overview | 21 a lot: center-based programs with poor process For effective teacher training, design it to be individually •  quality (even with relatively good infrastructure, targeted and repeated, with follow-up coaching—often caregiver training, and caregiver-children ratios) around a specific pedagogical technique. This approach can actually worsen developmental outcomes.78 contrasts starkly with much of today’s profes- •  Lower the cost of schooling to get children into school, but sional development for teachers across a range of then use other tools to boost motivation and effort because countries. In the United States, a team of teacher cost-reducing interventions don’t usually lead to learning training experts characterized professional devel- on their own.79 To improve learning, demand-side pro- opment there as “episodic, myopic, and often mean- grams need to increase a student’s effort or capacity ingless.”90 In Sub-Saharan Africa, teacher training is to learn. School-provided meals, for example, have often too short to be effective and too low in quality had positive effects on access—and also on learning to make a difference.91 By contrast, programs in in places where children have limited access to food Africa and South Asia that provided long-term at home.80 Targeted cash transfers have led to more coaching led to sizable learning gains.92 learning when they have incentivized performance To keep learners from falling behind to the point where •  itself81 or were marketed in a way that induces more they cannot catch up, target teaching to the level of the effort, such as in Cambodia.82 Some information student. Over the course of several grades, often interventions have motivated efforts as well.83 only a fraction of learners progress at grade level, •  To make up for the fact that so many youth lack skills with most falling behind and some learning almost when leaving basic education, provide remediation before nothing. This is partly because teachers teach to the further education and training.84 Remediation in school most advanced students in the class, as documented is a first best approach. After school, the more suc- from Australia to Sweden to the United States,93 or cessful programs share two main features. First, because the curriculum is too ambitious but teach- they provide bridging courses in real-life settings, ers are required to teach it.94 Effective strategies to which allows learners with very low foundational target teaching to the level of the student include skills to build these in the workplace.85 Second, accel- using community teachers to provide remedial les- erated, flexible pathways—not sequential courses sons to the lowest performers, reorganizing classes over multiple semesters—are associated with greater by ability, or using technology to adapt lessons to student retention and ultimate certification.86 individual student needs.95 Use pecuniary and nonpecuniary incentives to improve •  Make teaching more effective the motivation of teachers, ensuring that the incentiv- Effective teaching depends on teachers’ skills and ized actions are within teachers’ capacity. Education motivation, and yet many systems do not take systems typically neither reward teachers for per- them seriously. Teacher salaries are the largest forming well nor penalize them for performing single budget item in education systems, consuming poorly. Incentives are most likely to be effective three-quarters of the budget at the primary level in at improving outcomes when there are straight- developing countries. Yet many systems struggle forward actions that teachers can take to improve to attract strong candidates into teaching and to learning—such as increasing attendance when provide a solid foundation of subject or pedagogical absenteeism is the constraint. But incentives do not knowledge before they start teaching. As a result, new need to be high stakes (or financial) to affect behav- teachers often find themselves in classrooms with ior. In Mexico and Punjab, Pakistan, simply provid- little mastery of the content they are to teach.87 Once ing diagnostic information to parents and schools teachers are in place, the professional development about the schools’ relative performance improved they receive is often inconsistent and overly theo- learning outcomes.96 retical. In some countries, the cost of this training is enormous, reaching $2.5 billion a year in the United Focus everything else on teaching and States.88 Moreover, education systems often have few learning effective mechanisms in place to mentor, support, School inputs, management, and governance must and motivate teachers—even though teachers’ skills benefit the learner-teacher relationship if they are do nothing for learning unless teachers choose to to improve learning—but many do not. Debates on apply them in the classroom.89 Fortunately, teachers’ improving education outcomes frequently revolve skills and motivation can be strengthened, leading around increasing inputs, such as textbooks, technol- to greater effort and more learning, with three main ogy, or school infrastructure. But too often the question promising principles emerging: of why these inputs might actually improve learning is 22 | World Development Report 2018 overlooked. The evidence on successful use of inputs it is high), and when a range of stakeholders (not and management suggests three main principles: just parents) are brought together in ways that lead to action. In Indonesia, school grants improved Provide additional  inputs, including new technologies, •  learning when links between the school and the in ways that complement rather than substitute for village council—a center of local authority—were teachers.97 A computer-assisted learning program strengthened.109 in Gujarat, India, improved learning when it added to teaching and learning time, especially for the The most effective systems—in terms of learn- poorest-performing students.98 A Kenyan program ing—are those that have narrowed gaps between that provided public school teachers with tablets evidence and practice. On learner preparation, for to support instruction increased the reading per- example, East Asian countries such as Korea and Sin- formance of their students.99 But simply providing gapore have achieved high levels of children ready to desktop computers to classrooms in Colombia— learn. Stunting rates among preschool-age children where they were not well integrated with the cur- are low, and children are motivated and supported by riculum—had no impact on learning.100 Even more their families. To promote effective teaching, Finland traditional inputs—such as books—often fail to and Singapore attract some of the most highly skilled affect teaching and learning when they aren’t actu- graduates from tertiary education into teaching and ally deployed in classrooms, or if the content is too provide them with effective professional develop- advanced for the students.101 ment opportunities and sustained support. Ensure that new information and communication tech- •  nology is really implementable in the current systems. Align actors—to make the whole system Interventions that incorporate information and work for learning communication technology have some of the big- Working at scale is not just “scaling gest impacts on learning.102 But for every highly up.” The concept of scaling up in effective program—such as a dynamic computer- education implies taking interven- assisted learning program for secondary school tions that have been shown to be students in Delhi that increased math and language effective on a pilot or experimental scores more than the vast majority of other learning scale and replicating them across interventions tested in India or elsewhere103—there hundreds or thousands of schools. are programs such as the One Laptop Per Child pro- However, this approach often fails Policy grams in Peru and Uruguay, which evaluations sug- because the key actors are human response 3: gested had no impact on student reading or math beings, operating with human Align actors ability.104 Technologies ill-adapted to their settings aspirations and limitations in a often fail to reach the classroom or to be used if they politically charged arena. Real-world complications reach it.105 can undermine well-designed programs, especially •  Focus school management and governance reforms when new, systemwide forces come into play. When on improving teacher-learner interaction. Training the Cambodian government tried to scale up early principals in how to improve that interaction—by child development centers and preschools—programs providing feedback to teachers on lesson plans, that had worked in some parts of the country when action plans to improve student performance, implemented by nongovernmental organizations and classroom behavior—has led to a large impact (NGOs)—low demand from parents and low-quality on student learning.106 In countries ranging from services led to no impacts on child development, Brazil and India to Sweden, the United Kingdom, and even slowed it for some.110 When the Kenyan and the United States, the management capacity of government tried to lower student-teacher ratios by school principals significantly and robustly relates hiring contract teachers—an intervention that had to student performance—even after controlling for improved student outcomes when implemented by a variety of student and school characteristics.107 an NGO—the results were negligible because of both Involving communities, parents, and school actors implementation constraints and political economy in ways that promote local oversight and account- factors.111 And when the Indonesian government ability for service delivery can improve outcomes.108 tried to increase teacher capacity by nearly doubling But community monitoring tends to have more the salaries of certified teachers, political pressures impact when it covers things that parents can watered down the certification process and left only easily observe (such as teacher absenteeism when the pay increase in place. The result was much larger Overview | 23 budget outlays on salaries, but no increase in teach- the absence of good information on learning prevents ers’ skills or student learning.112 stakeholders from judging system performance, The lesson, then, is that better interventions at the designing the appropriate policies, and holding poli- school and student levels will sustainably improve ticians and bureaucrats to account. Thus improving learning only if countries tackle the stubborn system- learning metrics is crucial for drawing attention to level technical and political barriers to change. Tech- problems and building the will for action. In Tanzania nical barriers include the complexity of the system, in the early 2010s, poor results on school-leaving exam- the large number of actors, the interdependence of inations—along with well-publicized results from reforms, and the slow pace of change in education citizen-led learning assessments and surveys showing systems. Political barriers include the competing poor service delivery in schools—motivated policy interests of different players and the difficulty of makers to launch ambitious reforms. In Germany, the moving out of a low-quality equilibrium, especially shock of mediocre results on the first PISA in 2000 led in low-trust environments where risks predominate. to reforms that improved both learning and equity. All of these barriers pull actors away from learning, Efforts in this area need to go beyond just measur- as discussed earlier. Systems that surmount these ing learning; they should track its determinants as barriers and align actors toward learning can achieve well. Understanding these determinants can enable remarkable learning outcomes. Shanghai provided reforms to grapple with the deeper causes, if there proof when it topped the 2012 PISA rankings, in part is a systemwide commitment to improving learning. thanks to policies that ensured that every classroom Take the issue of learner preparedness. When indica- had a prepared, supported, and motivated teacher.113 tors reveal that poorer children already lag far behind To shift the system toward learning, technically by the time they start primary school, this finding and politically, reformers can use three sets of tools: can build political will not only to expand preschool education in low-income areas, but also to combat • Information and metrics. Better information and stunting and educate parents about early stimulation metrics can promote learning in two ways: by of children. When indicators show that many teach- catalyzing reforms and by serving as indicators of ers lack a strong command of what their students whether reforms are working to improve learning are meant to learn, this finding can spark efforts to with equity. Thus they can improve both the politi- improve the quality of teacher education.114 cal and technical alignment of the system. Of course, information and metrics can also be • Coalitions and incentives. Good information will have misleading, irrelevant, or politically unsustainable, a payoff only if there is enough support for prior- so they need to be designed and used wisely. Metrics itizing learning. Politics is often the problem, and may fail to capture important dimensions of the politics must be part of the solution. This requires outcomes the education system is trying to promote. forming coalitions to advocate for broad-based For example, the Millennium Development Goal of learning and skills and to rebalance the political universal primary education by 2015 embodied a cru- incentives. cial goal—equitable access—but it did not represent • Innovation and agility. Schools and societies have what many assumed it did: universal acquisition of achieved high levels of equitable learning in a foundational literacy and numeracy, let alone other variety of ways. Figuring out what approaches will life skills. Another risk is of distorting good metrics work in a given context requires innovation and by putting high stakes on them, if potential benefi- adaptation. This means using evidence to identify ciaries can game the indicators. Thus systems will where to start and then using metrics to iterate need different measures for different purposes.115 with feedback loops. Even if they are technically sound, metrics may prove politically unsustainable if they highlight too All of these tools will be most effective when sup- many problems and do not provide any reason for ported by strong implementation capacity within hope. One way to address this problem is to focus not government. on levels of learning, which may be very low, but on progress over time. Information and metrics Better information and measurement—starting with Coalitions and incentives learning metrics—are critical to creating political Mobilizing everyone who has a stake in learning space for innovation and then using that space to has been an important strategy in efforts to improve achieve continuous improvement. As emphasized, learning. Many countries have used wide-ranging 24 | World Development Report 2018 consultations that have tried to bring in all interest Innovation and agility groups to build support for proposed changes in edu- To develop effective learning approaches that fit their cation policy. Malaysia used a “lab” model to bring contexts, education systems need to encourage inno- together coalitions of stakeholders and involve them vation and adaptation. In many education systems, in all stages of reform, from design to implementa- schools and other education institutions regularly tion.116 Mobilizing citizens through regular informa- adapt to changing circumstances. Through these tion and communication campaigns can also be an adaptations, innovative solutions to education chal- important strategy. In Peru, reformers in the govern- lenges often emerge. Exploring the well-performing ment used information on poor learning outcomes parts of any education system can reveal technically and performance of the education system to mobilize and politically feasible approaches to the problems public support for reforms to strengthen teacher systems face in improving learning. For example, in accountability. That information also catalyzed action Misiones province in Argentina high student dropout by the business community, which funded a cam- rates were widespread, but some schools seemed to paign highlighting the importance of quality educa- buck the trend. A closer look at these “positive devi- tion for economic growth. In parts of Peru, parents ants” revealed very different relationships between used this entry point to protest teacher strikes that teachers and parents. When other schools adopted had disrupted schooling.117 Another tool for building the more constructive approach to parent-teacher coalitions is to bundle reforms, so that each actor relations used by the successful schools, their drop- achieves one of its top priorities. For example, a com- out rates fell significantly.122 Burundi, while recover- mitment to modernize vocational training—a reform ing from a civil war, used an adaptive approach to find that could help employers immediately—could buy the right way to get textbooks to schools. It reduced their support for broader education reforms. delivery times from over a year to 60 days—then rep- Where feasible, a negotiated and gradual approach licated that approach in other areas.123 to reform can provide a more promising alternative Incentives are important in determining whether to direct confrontation. When system actors agree to systems innovate and adopt emerging solutions at collaborate and build trust around shared goals, the scale. Systems that are closed, that limit the autonomy chances of successful reform are likely to be higher. of teachers and schools, and that judge performance by In Chile, successive negotiations between the govern- the extent of compliance with rules governing resource ment and the teachers’ union built broad support for a use often provide little room for innovation. By contrast, series of reforms that adjusted the working conditions more open systems that pay more attention to overall of teachers to improve their overall welfare, while outcomes and reward progress in raising outcomes are linking pay and career development more closely to more likely to see greater innovation and the diffusion performance.118 One approach used by several coun- of new approaches across the education system.124 tries has been to compensate actors who might lose To make a difference at the system level, such out from reforms. In other cases, dual-track reforms innovations needs to be packaged with good metrics have been introduced to phase in changes in a way and with system-level coalitions for learning. With- that protects incumbent actors from their effects—for out both, any improvements from innovation are example, in Peru and the District of Columbia in the likely to prove short-lived or limited to local areas. But United States, pay-for-performance schemes were with such support, a virtuous cycle becomes possible initially voluntary.119 as systems follow these steps: Building strong partnerships between schools and their communities is also important for sustaining Set learning as a clearly articulated goal and mea- •  reforms. Where political and bureaucratic incen- sure it. tives for reform are weak, action at the local level Build a coalition for learning that gives the political •  can act as a substitute. In South Africa, the political space for innovation and experimentation. and economic context constrains efforts to improve Innovate and test approaches that seem the most •  education performance. Yet progress was made in promising for the given context, drawing inspira- improving outcomes at the local level through strong tion from the evidence base and focusing on areas partnerships between parents and schools.120 Even that promise the biggest improvements over cur- where broader incentives exist to improve learning, rent practice. community engagement at the local level is import- Use the measure of learning, along with the other •  ant and can complement national or subnational metrics of delivery, as a gauge of whether the change efforts.121 approach is working. Overview | 25 FIGURE O.15 Coherence and alignment towards learning Figure O.15 Coherence and alignment toward learning CiviC l iv il so ians orgS o anic ietciet t ic P oli z at y Oy ion r s gs . Pe rs Le e ar er r ch to s/c ec ne a om Te te s rs mun Priva ities LEARNING Sch Bure iary oo ts a uc pu lm dic na in l rat a Ju ge oo s me h nt Sc In te rn s at i or on al a r act c to r s Ot he Source: WDR 2018 team. Build on what works, and scale back what doesn’t, •  teaching-learning relationship, or because the system to deliver short-term results that strengthen the doesn’t prioritize learning for disadvantaged children long-term resolve of the coalition for learning. and youth. More financing for business as usual will Repeat. •  therefore just lead to the usual outcomes. But where countries seriously tackle the barriers to learning for The payoff to doing what needs to be done is a all, spending on education is a critical investment for system in which the elements are coherent with each development, especially for those countries where other and everything aligns with learning (figure O.15). overall spending is currently low, as recent major Increased financing can support this learning- studies of global education have emphasized.126 More for-all equilibrium, if the various key actors behave children staying in school longer and learning while in ways that show learning matters to them. This is there will undoubtedly require more public financing a big “if” because higher levels of public spending are for education. An injection of financing—either from not associated statistically with higher completion or domestic or international sources—can help coun- even enrollment rates in countries with weak gover- tries escape the low-learning trap, if they are willing nance.125 Ensuring that students learn is even more to take the other necessary steps laid out here. challenging, and so there is little correlation between spending and learning after accounting for national Implications for external actors income. It is easy to see the reason for this because External actors can reinforce these strategies for of the many ways in which financing can leak out— opening the political and technical space for learn- whether because money never reaches the school, ing. In the realm of information and metrics, for or because it pays for inputs that don’t affect the example, international actors can fund participation 26 | World Development Report 2018 in regional learning assessments (such as PASEC But waiting out the learning crisis isn’t a winning in West Africa or LLECE in Latin America) or global strategy. Even though national income and learning learning assessments (such as PISA or TIMSS) to are somewhat correlated at lower levels of develop- spotlight challenges and catalyze domestic efforts ment, higher incomes do not invariably lead to better for reform. External actors can also develop tools learning outcomes. And to the extent that develop- for tracking the proximate determinants of learning ment does bring better learning and skills, it is partly to aid in feedback loops. Domestic financing usu- because development has been accompanied by a ally makes up the bulk of education financing, so a willingness to tackle the political impasses and gov- high-leverage entry point for international actors is ernance challenges that hamper learning. Ultimately, to fund better information that will make domestic then, those challenges are not avoidable. Furthermore, spending more effective. In the realm of innovation there’s no need to wait for learning. At every level of and experimentation, external funders such as the income, there are countries that not only score better World Bank can provide results-based financing that than others on international assessments, but also— gives countries more room to innovate and iterate and more important—show from the quality of their their way to achieving better outcomes. education systems and their policy making that they are committed to learning. Learning to realize The future of work will place a premium on learn- ing. Rapid technological change has led to major education’s promise shifts in the nature of work, leading some to declare By showing that learning really matters to them, this a new era—the Second Machine Age or the Fourth countries can realize education’s full promise. Beyond Industrial Revolution. In the extreme versions of this being a basic human right, education—done right— vision, all but a few jobs could disappear, decreasing improves social outcomes in many spheres of life. the value of skills for most people. But the seismic For individuals and families, education boosts human changes predicted have yet to permeate the high- capital, improves economic opportunities, promotes income countries, let alone the low- and middle- health, and expands the ability to make effective income ones. More important, no matter how the choices. For societies, education expands economic demand for skills changes in the future, people opportunities, promotes social mobility, and makes will require a solid foundation of basic skills and institutions function more effectively. In measuring knowledge. If anything, rapid change will increase these benefits, research has only recently focused on the returns to learning how to learn, which requires the distinction between schooling and learning. But foundational skills that allow individuals to size the evidence confirms the intuition that these benefits up new situations, adapt their thinking, and know often depend on the skills that students acquire, not where to go for information and how to make sense just the number of years in the classroom. Economies of it. with higher skills grow faster than those with school- ing but mediocre skills; higher literacy predicts better *** financial knowledge and better health, beyond the Countries have already made a tremendous start by effects of schooling; and poor children are more likely getting so many children and youth into school. Now to rise in the income distribution when they grow up it’s time to realize education’s promise by accelerat- in communities with better learning outcomes. ing learning. A real education—one that encourages Taking learning seriously won’t be easy. It’s hard learning—is a tool for promoting both shared pros- enough to work through the technical challenges of perity and poverty elimination. That type of educa- figuring out what will promote learning at the level of tion will benefit many: children and families whose the student and school in any context, let alone tackle positive schooling experience restores their faith in the political and technical challenges of working at government and society rather than eroding it; youth scale. Many countries struggling with the learning who have skills employers are seeking; teachers who crisis may be tempted to continue with business as can respond to their professional calling rather than usual. After all, they may reason, development will to political demands; adult workers who have learned eventually improve learning outcomes: as households how to learn, preparing them for unforeseeable eco- escape poverty and schools take advantage of better nomic and social changes; and citizens who have the facilities, more materials, and better-trained teachers, values and reasoning abilities to contribute to civic better learning outcomes should follow. life and social cohesion. Overview | 27 Notes 1. Uwezo (2014). In all countries, the test was administered 32. Black and others (2017). Stunting is defined by the World in English. In Kenya and Tanzania, it was also admin- Health Organization (WHO) as a height-for-age z-score istered in Kiswahili, and the highest score (English or of less than two standard deviations below the median Kiswahili) was used in the assessment of proficiency. of a healthy reference population. English is the language of instruction in Kenya and 33. Paxson and Schady (2007); Schady and others (2015). Uganda. 34. Hanushek (1992); Rockoff (2004). 2. ASER Centre (2017). 35. Bau and Das (2017). 3. WDR 2018 team, using data from the Programme for 36. Bruns and Luque (2015). International Student Assessment (PISA), 2015 (OECD 37. UIS (2006). 2016). 38. Chang and others (2013). 4. WDR 2018 team, using data from the Third Regional 39. Abadzi (2009); EQUIP2 (2010). Comparative and Explanatory Study (TERCE), 2012 40. Bold and others (2017). (UNESCO 2013). 41. Hanushek (1995); Mingat and Tan (1998); Tan and 5. UNESCO (2016). Mingat (1992); Wolf (2004). 6. World Bank (2011). 42. Glewwe and others (2011); Hanushek (1986); Kremer 7. Barro and Lee (2013). (1995). 8. Pritchett (2013). 43. Sabarwal, Evans, and Marshak (2014). 9. Pritchett (2013). 44. Lavinas and Veiga (2013). 10. Gove and Cvelich (2011). 45. Robinson, Lloyd, and Rowe (2008); Waters, Marzano, 11. Crouch (2006). and McNulty (2003). 12. Castillo and others (2011). 46. Bloom and others (2015). Management areas include 13. ASER Pakistan (2015a, 2015b). operations, monitoring, target setting, and people 14. Southern and Eastern Africa Consortium for Monitor- management. ing Educational Quality (SACMEQ) results for grade 6 47. Bruns, Filmer, and Patrinos (2011); Orazem, Glewwe, and students in 15 countries in 2007 (Hungi and others Patrinos (2007); World Bank (2003). 2010). 48. Data extracted from U.S. Agency for International 15. Programme d’Analyse des Systèmes Éducatifs de la Development’s Demographic and Health Survey (DHS) Confemen (PASEC) results for grade 6 students in 10 StatCompiler, http://www.statcompiler.com/en/. francophone countries in 2014 (PASEC 2015). 49. Park (2016). 16. ASER Centre (2017). 50. Todd and Mason (2005). 17. RTI International (2009). 51. Chisholm and Leyendecker (2008). 18. World Bank (2016b). 52. World Bank (2003). 19. Muralidharan and Zieleniak (2013). 53. Andrews, Pritchett, and Woolcock (2017). 20. Spaull and Kotze (2015). 54. Grindle (2004). 21. Singh (2015). 55. Evans and Yuan (2017). 22. Minimum proficiency is defined as one standard devi- 56. The team thanks Kai-Ming Cheng for suggesting this ation below the mean of the harmonized assessment formulation. scores. 57. Dweck (2008). 23. These numbers are based on analysis of the data in 58. Save the Children (2013). “A Global Data Set on Education Quality” (2017), a data 59. Guilfoyle (2006). set made available to the WDR 2018 team by Nadir Alti- 60. Jacob (2005). nok, Noam Angrist, and Harry Anthony Patrinos. These 61. Fausset (2014). averages do not include China or India because of lack 62. UIS (2016). of data. 63. OECD (2011). 24. UNESCO (2016). 64. Jacob (2007). 25. UIS and EFA (2015). 65. Solano-Flores, Contreras-Niño, and Backhoff Escudero 26. Banerjee, Jacob, and Kremer (2000); Hanushek and (2005). Woessmann (2008); Rivkin, Hanushek, and Kain (2005). 66. Muralidharan and Sundararaman (2011). 27. Alderman, Orazem, and Paterno (2001); Andrabi, Das, 67. De Smedt (2014); Insel and Landis (2013); Kuhl (2010). and Khwaja (2008); Farah (1996); Kingdon (1996); 68. Dua and others (2016). Orazem (2000); Tooley and Dixon (2007). 69. Evans and Popova (2016). 28. Hanushek, Lavy, and Hitomi (2008). 70. Gertler and others (2014). 29. STEP surveys (World Bank 2014). 71. Calculations carried out for WDR 2018. See Evans and 30. Lupien and others (2000); McCoy and others (2016); Yuan (2017). Walker and others (2007). 72. Pritchett and Sandefur (2013). 31. Coe and Lubach (2007); Garner and others (2012); 73. Deaton and Cartwright (2016). Nelson (2016). 74. Romer (2015). 28 | World Development Report 2018 75. Duflo, Hanna, and Ryan (2012); Muralidharan and 112. Chang and others (2013); de Ree and others (2015). Sundararaman (2011). 113. Liang, Kidwai, and Zhang (2016). 76. Das and others (2013). 114. For example, in Mozambique, after the World Bank’s 77. The evidence is from countries ranging from the Service Delivery Indicators revealed very low levels of United States to Argentina, Bangladesh, China, and teacher knowledge and very high levels of absentee- Uganda, among others (Berlinski, Galiani, and Gertler ism—results that were picked up by the local media— 2008; Engle and others 2011). the government launched a program (ultimately sup- 78. Berlinski and Schady (2015); Bernal and others (2016); ported through a loan from the World Bank) to address Grantham-McGregor and others (2014). these issues. 79. Baird and others (2014); Fiszbein and Schady (2009); 115. Neal (2013). Morgan, Petrosino, and Fronius (2012). 116. World Bank (2017). 80. Snilstveit and others (2016). 117. Bruns and Luque (2015). 81. Blimpo (2014); Kremer, Miguel, and Thornton (2009). 118. Mizala and Schneider (2014); Wales, Ali, and Nicolai Direct financial incentives have been less successful in (2014). high-income countries (Fryer 2011), although alternate 119. Birnbaum (2010); Bruns and Luque (2015). designs that deliver incentives immediately after the 120. Levy and others (2016). test have worked (Levitt and others 2016). 121. Mansuri and Rao (2013). 82. Barrera-Osorio and Filmer (2013). 122. Green (2016); Pascale, Sternin, and Sternin (2010). 83. Avitabile and de Hoyos (2015); Nguyen (2008). 123. Campos, Randrianarivelo, and Winning (2015). 84. ILO (2015). 124. Andrews, Pritchett, and Woolcock (2013). 85. Bragg (2014). 125. Rajkumar and Swaroop (2008); Suryadarma (2012). 86. Calcagno and Long (2008); Martorell and McFarlin Jr. 126. See, in particular, the report of the Education Commis- (2011); Scott-Clayton and Rodriguez (2014). sion (2016), which emphasizes the important role of 87. Tandon and Fukao (2015); World Bank (2013, 2016a). finance in complementing reforms. 88. Layton (2015). 89. Bruns and Luque (2015); Mulkeen (2010). 90. Darling-Hammond and others (2009). References 91. Lauwerier and Akkari (2015). 92. Banerjee and others (2007); Conn (2017). Abadzi, Helen. 2009. “Instructional Time Loss in Developing 93. Abadzi and Llambiri (2011); Ciaccio (2004); Leder (1987). Countries: Concepts, Measurement, and Implications.” 94. Banerjee and others (2016); Pritchett and Beatty (2015). World Bank Research Observer 24 (2): 267–90. 95. Banerjee and others (2007); Duflo, Dupas, and Kremer Abadzi, Helen, and Stavri Llambiri. 2011. “Selective Teacher (2011); Kiessel and Duflo (2014); Muralidharan, Singh, Attention in Lower-Income Countries: A Phenomenon and Ganimian (2016). Linked to Dropout and Illiteracy?” Prospects 41 (4): 96. Andrabi, Das, and Khwaja (2015); de Hoyos, Garcia- 491–506. Moreno, and Patrinos (2017). Alderman, Harold, Peter F. Orazem, and Elizabeth M. Paterno. 97. Snilstveit and others (2016). 2001. “School Quality, School Cost, and the Public/Private 98. Linden (2008). School Choices of Low-Income Households in Pakistan.” 99. Piper and others (2015). Journal of Human Resources 36 (2): 304–26. 100. Barrera-Osorio and Linden (2009). Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2008. 101. Glewwe, Kremer, and Moulin (2009); Sabarwal, Evans, “A Dime a Day: The Possibilities and Limits of Private and Marshak (2014). Schooling in Pakistan.” Comparative Education Review 52 102. McEwan (2015). (3): 329–55. 103. Muralidharan, Singh, and Ganimian (2016). —— ——. 2015. “Report Cards: The Impact of Providing School 104. Cristia and others (2012); De Melo, Machado, and and Child Test Scores on Educational Markets.” Policy Miranda (2014). For Uruguay, the evaluation covers Research Working Paper 7226, World Bank, Washing- math and reading impacts in the early years of the ton, DC. program, when its main objective was to provide Andrews, Matt, Lant Pritchett, and Michael Woolcock. 2013. equipment and connectivity for schools; the program “Escaping Capability Traps through Problem Driven Iter- evolved since then to add ICT training for teachers and ative Adaptation (PDIA).” World Development 51: 234–44. adaptive educational technology, and new evaluations —— ——. 2017. Building State Capability: Evidence, Analysis, Action. are expected to be published in late 2017. New York: Oxford University Press. 105. Lavinas and Veiga (2013). ASER Centre. 2017. Annual Status of Education Report (Rural) 106. Fryer (2017). 2016. New Delhi: ASER Centre. http://img.asercentre.org 107. Bloom and others (2015). /docs/Publications/ASER%20Reports/ASER%202016 108. Bruns, Filmer, and Patrinos (2011). /aser_2016.pdf. 109. Pradhan and others (2014). ASER Pakistan. 2015a. “Annual Status of Education Report: 110. Bouguen and others (2013). ASER Pakistan 2015 National (Rural).” Lahore, Pakistan: 111. Bold and others (2013). South Asian Forum for Education Development. Overview | 29 ASER Pakistan. 2015b. “Annual Status of Education Report: .washingtonpost.com/wp-dyn/content/article/2010/09/10 ASER Pakistan 2015 National (Urban).” Lahore, Pakistan: /AR2010091006604.html. South Asian Forum for Education Development. Black, Maureen M., Susan P. Walker, Lia C. H. Fernald, Avitabile, Ciro, and Rafael E. de Hoyos. 2015. “The Heteroge- Christopher T. Andersen, Ann M. DiGirolamo, Chunling neous Effect of Information on Student Performance: Lu, Dana C. McCoy, et al. 2017. “Early Childhood Develop- Evidence from a Randomized Control Trial in Mexico.” ment Coming of Age: Science through the Life Course.” Policy Research Working Paper 7422, World Bank, Wash- Lancet 389 (10064): 77–90. ington, DC. Blimpo, Moussa P. 2014. “Team Incentives for Education in Baird, Sarah Jane, Francisco H. G. Ferreira, Berk Özler, and Developing Countries: A Randomized Field Experiment Michael Woolcock. 2014. “Conditional, Unconditional in Benin.” American Economic Journal: Applied Economics and Everything in Between: A Systematic Review of the 6 (4): 90–109. Effects of Cash Transfer Programmes on Schooling Out- Bloom, Nicholas, Renata Lemos, Raffaella Sadun, Daniela comes.” Journal of Development Effectiveness 6 (1): 1–43. Scur, and John Van Reenen. 2014. “JEEA-FBBVA Lecture Banerjee, Abhijit Vinayak, Rukmini Banerji, James Berry, 2013: The New Empirical Economics of Management.” Esther Duflo, Harini Kannan, Shobhini Mukherji, Marc Journal of the European Economic Association 12 (4): 835–76. Shotland, et al. 2016. “Mainstreaming an Effective Inter- Bloom, Nicholas, Renata Lemos, Raffaella Sadun, and John vention: Evidence from Randomized Evaluations of Van Reenen. 2015. “Does Management Matter in ‘Teaching at the Right Level’ in India.” NBER Working Schools?” Economic Journal 125 (584): 647–74. Paper 22746, National Bureau of Economic Research, Bold, Tessa, Deon Filmer, Gayle Martin, Ezequiel Molina, Cambridge, MA. Brian Stacy, Christophe Rockmore, Jakob Svensson, Banerjee, Abhijit Vinayak, Shawn Cole, Esther Duflo, and et al. 2017. “What Do Teachers Know and Do? Does It Leigh Linden. 2007. “Remedying Education: Evidence Matter? Evidence from Primary Schools in Africa.” Policy from Two Randomized Experiments in India.” Quarterly Research Working Paper 7956, World Bank, Washing- Journal of Economics 122 (3): 1235–64. ton, DC. Banerjee, Abhijit Vinayak, Suraj Jacob, and Michael Kremer. Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, Alice 2000. “Promoting School Participation in Rural Rajas- Ng’ang’a, and Justin Sandefur. 2013. “Scaling Up What than: Results from Some Prospective Trials.” With Jenny Works: Experimental Evidence on External Validity in Lanjouw and Peter Lanjouw. Working paper, Massachu- Kenyan Education.” Working Paper 321, Center for Global setts Institute of Technology, Cambridge, MA. Development, Washington, DC. Barrera-Osorio, Felipe, and Deon Filmer. 2013. “Incentivizing Bouguen, Adrien, Deon Filmer, Karen Macours, and Sophie Schooling for Learning: Evidence on the Impact of Alter- Naudeau. 2013. “Impact Evaluation of Three Types of native Targeting Approaches.” Policy Research Working Early Childhood Development Interventions in Cambo- Paper 6541, World Bank, Washington, DC. dia.” Policy Research Working Paper 6540, World Bank, Barrera-Osorio, Felipe, and Leigh L. Linden. 2009. “The Use Washington, DC. and Misuse of Computers in Education: Evidence from a Bragg, Debra D. 2014. “Career Pathways in Disparate Indus- Randomized Experiment in Colombia.” Policy Research try Sectors to Serve Underserved Populations.” Paper Working Paper 4836, World Bank, Washington, DC. presented at American Educational Research Association Barro, Robert J., and Jong Wha Lee. 2013. “A New Data Set of conference, Philadelphia, April 5. Educational Attainment in the World, 1950–2010.” Journal Bruns, Barbara, Deon Filmer, and Harry Anthony Patrinos. of Development Economics 104: 184–98. 2011. Making Schools Work: New Evidence on Accountability Bau, Natalie, and Jishnu Das. 2017. “The Misallocation of Pay Reforms. Human Development Perspectives Series. and Productivity in the Public Sector: Evidence from the Washington, DC: World Bank. Labor Market for Teachers.” Policy Research Working Bruns, Barbara, and Javier Luque. 2015. Great Teachers: How to Paper 8050, World Bank, Washington, DC. Raise Student Learning in Latin America and the Caribbean. Berlinski, Samuel, Sebastian Galiani, and Paul J. Gertler. With Soledad De Gregorio, David K. Evans, Marco 2008. “The Effect of Pre-primary Education on Primary Fernández, Martin Moreno, Jessica Rodriguez, Guill- School Performance.” Journal of Public Economics 93 (1–2): ermo Toral, and Noah Yarrow. Latin American Develop- 219–34. ment Forum Series. Washington, DC: World Bank. Berlinski, Samuel, and Norbert R. Schady, eds. 2015. The Calcagno, Juan Carlos, and Bridget Terry Long. 2008. “The Early Years: Child Well-Being and the Role of Public Policy. Impact of Postsecondary Remediation Using a Regres- Development in the Americas Series. Washington, DC: sion Discontinuity Approach: Addressing Endogenous Inter-American Development Bank; New York: Palgrave Sorting and Noncompliance.” NBER Working Paper Macmillan. 14194, National Bureau of Economic Research, Cam- Bernal, Raquel, Orazio Pietro Attanasio, Ximena Peña, and bridge, MA. Marcos Vera-Hernández. 2016. “The Effects of the Tran- Campos, Jose Edgardo, Benjamina Randrianarivelo, and Kay sition from Home-Based Community Nurseries to Child- Winning. 2015. “Escaping the ‘Capability Trap’: Turning Care Centers on Children in Colombia.” Working paper, ‘Small’ Development into ‘Big’ Development.” Interna- Universidad de los Andes, Bogotá, Colombia. tional Public Management Review 16 (1): 99–131. Birnbaum, Michael. 2010. “D.C. Schools Unveil Teacher-Pay Castillo, Melba, Vanesa Castro, José Ramón Laguna, and Bonus Plan.” Washington Post, September 12. http://www Josefina Vijil. 2011. Informe de Resultados: EGMS Nicaragua. 30 | World Development Report 2018 Research Triangle Park, NC: Centro de Investigación y de Ree, Joppe, Karthik Muralidharan, Menno Pradhan, and Acción Educativa Social and RTI International. https:// Halsey Rogers. 2015. “Double for Nothing? Experimental shared.rti.org/content/informe-de-resultados-egma Evidence on the Impact of an Unconditional Teacher -nicaragua. Salary Increase on Student Performance in Indonesia.” Chang, Mae Chu, Sheldon Shaeffer, Samer Al-Samarrai, NBER Working Paper 21806, National Bureau of Eco- Andrew B. Ragatz, Joppe De Ree, and Ritchie Stevenson. nomic Research, Cambridge, MA. 2013. Teacher Reform in Indonesia: The Role of Politics and De Smedt, Bert. 2014. “Advances in the Use of Neuroscience Evidence in Policy Making. Directions in Development: Methods in Research on Learning and Instruction.” Human Development Series. Washington, DC: World Frontline Learning Research 2 (4): 7–14. Bank. Dua, Tarun, Mark Tomlinson, Elizabeth Tablante, Pia Britto, Chisholm, Linda, and Ramon Leyendecker. 2008. “Curricu- Aisha Yousfzai, Bernadette Daelmans, and Gary L. Darm- lum Reform in Post-1990s Sub-Saharan Africa.” Interna- stadt. 2016. “Global Research Priorities to Accelerate tional Journal of Educational Development 28 (2): 195–205. Early Child Development in the Sustainable Develop- Ciaccio, Joseph. 2004. Totally Positive Teaching: A Five-Stage ment Era.” Lancet Global Health 4 (12): e887–e889. Approach to Energizing Students and Teachers. Alexandria, Duflo, Esther, Pascaline Dupas, and Michael Kremer. 2011. VA: ASCD. “Peer Effects, Teacher Incentives, and the Impact of Coe, Christopher L., and Gabrielle R. Lubach. 2007. “Mother- Tracking: Evidence from a Randomized Evaluation in Infant Interactions and the Development of Immunity Kenya.” American Economic Review 101 (5): 1739–74. from Conception through Weaning.” In Psychoneuro­ Duflo, Esther, Rema Hanna, and Stephen P. Ryan. 2012. immunology, edited by Robert Ader, 455–74. Burlington, “Incentives Work: Getting Teachers to Come to School.” MA: Elsevier Academic Press. American Economic Review 102 (4): 1241–78. Conn, Katharine M. 2017. “Identifying Effective Education Dweck, Carol S. 2008. Mindset, the New Psychology of Success: Interventions in Sub-Saharan Africa: A Meta-Analysis How We Can Learn to Fulfill Our Potential. New York: Ballan- of Impact Evaluations.” Review of Educational Research tine Books. (May 26). http://journals.sagepub.com/doi/abs/10.3102 Education Commission. 2016. The Learning Generation: Invest- /0034654317712025. ing in Education for a Changing World. New York: Inter­ national Commission on Financing Global Education Cristia, Julián P., Pablo Ibarrarán, Santiago Cueto, Ana Opportunity. Santiago, and Eugenio Severín. 2012. “Technology and Engle, Patrice L., Lia C. H. Fernald, Harold Alderman, Jere Child Development: Evidence from the One Laptop Per Behrman, Chloe O’Gara, Aisha Yousafzai, Meena Cabral Child Program.” IZA Discussion Paper 6401, Institute for de Mello, et al. 2011. “Strategies for Reducing Inequalities the Study of Labor, Bonn, Germany. and Improving Developmental Outcomes for Young Crouch, Luis. 2006. “Education Sector: Standards, Account- Children in Low-Income and Middle-Income Countries.” ability, and Support.” In A New Social Contract for Peru: An Lancet 378 (9799): 1339–53. Agenda for Improving Education, Health Care, and the Social EQUIP2 (Educational Quality Improvement Program 2). Safety Net, edited by Daniel Cotlear, 71–106. World Bank 2010. “Using Opportunity to Learn and Early Grade Country Study Series. Washington, DC: World Bank. Reading Fluency to Measure School Effectiveness in Darling-Hammond, Linda, Ruth Chung Wei, Alethea Andree, Ethiopia, Guatemala, Honduras, and Nepal.” Working Nikole Richardson, and Stelios Orphanos. 2009. “Profes- paper, Educational Policy, Systems Development, and sional Learning in the Learning Profession: A Status Management, U.S. Agency for International Develop- Report on Teacher Development in the United States and ment, Washington, DC. Abroad.” National Staff Development Council, Dallas. Evans, David K., and Anna Popova. 2016. “What Really Works Das, Jishnu, Stefan Dercon, James Habyarimana, Pramila to Improve Learning in Developing Countries? An Analy- Krishnan, Karthik Muralidharan, and Venkatesh Sunda- sis of Divergent Findings in Systematic Reviews.” World raraman. 2013. “School Inputs, Household Substitution, Bank Research Observer 31 (2): 242–70. and Test Scores.” American Economic Journal: Applied Eco- Evans, David K., and Fei Yuan. 2017. “Economic Returns to nomics 5 (2): 29–57. Interventions That Increase Learning.” Background Deaton, Angus S., and Nancy Cartwright. 2016. “Understand- paper, World Development Report 2018, World Bank, Wash- ing and Misunderstanding Randomized Controlled ington, DC. Trials.” NBER Working Paper 22595, National Bureau of Farah, I. 1996. “Road to Success: Self-Sustaining Primary Economic Research, Cambridge, MA. School Change in Rural Pakistan.” With T. Mehmood, de Hoyos, Rafael E., Vicente A. Garcia-Moreno, and Harry Amna, R. Jaffar, F. Ashams, P. Iqbal, S. Khanam, Z. Shah, Anthony Patrinos. 2017. “The Impact of an Accountability and N. Gul-Mastoi. Institute for Educational Develop- Intervention with Diagnostic Feedback: Evidence from ment, Aga Khan University, Karachi, Pakistan. Mexico.” Economics of Education Review 58: 123–40. Fausset, Richard. 2014. “Trial Opens in Atlanta School Cheat- De Melo, Gioia, Alina Machado, and Alfonso Miranda. 2014. ing Scandal.” New York Times, September 29. https://www “The Impact of a One Laptop Per Child Program on .nytimes.com/2014/09/30/us/racketeering-trial-opens-in Learning: Evidence from Uruguay.” IZA Discussion -altanta-schools-cheating-scandal.html?_r=1. Paper 8489, Institute for the Study of Labor, Bonn, Filmer, Deon. 2016. “Educational Attainment and Enrollment Germany. around the World: An International Database.” World Overview | 31 Bank, Washington, DC. http://go.worldbank.org/3 of Dropout Behavior in Developing Countries.” Journal of GEREWJ0E0. Human Capital 2 (1): 69–105. Fiszbein, Ariel, and Norbert R. Schady. 2009. Conditional Hanushek, Eric A., and Ludger Woessmann. 2008. “The Role Cash Transfers: Reducing Present and Future Poverty. With of Cognitive Skills in Economic Development.” Journal of Francisco H. G. Ferreira, Margaret Grosh, Niall Keleher, Economic Literature 46 (3): 607–68. Pedro Olinto, and Emmanuel Skoufias. World Bank Hungi, Njora, Demus Makuwa, Kenneth Norman Ross, Policy Research Report. Washington, DC: World Bank. Mioko Saito, Stéphanie Dolata, Frank Van Cappelle, Fryer, Roland G., Jr. 2011. “Financial Incentives and Student Laura Paviot, et al. 2010. “SACMEQ III Project Results: Achievement: Evidence from Randomized Trials.” Quar- Pupil Achievement Levels in Reading and Mathematics.” terly Journal of Economics 126 (4): 1755–98. Southern and Eastern Africa Consortium for Monitoring ————. 2017. “Management and Student Achievement: Evi- Educational Quality, Paris. dence from a Randomized Field Experiment.” Working ILO (International Labor Organization). 2015. “Global paper, Harvard University, Cambridge, MA. Employment Trends for Youth 2015: Scaling Up Invest- Garner, Andrew S., Jack P. Shonkoff, Benjamin S. Siegel, ments in Decent Jobs for Youth.” ILO, Geneva. Mary I. Dobbins, Marian F. Earls, Laura McGuinn, John Insel, Thomas R., and Story C. Landis. 2013. “Twenty-Five Pascoe, et al. 2012. “Early Childhood Adversity, Toxic Years of Progress: The View from Nimh and Ninds.” Stress, and the Role of the Pediatrician: Translating Neuron 80 (3): 561–67. Developmental Science into Lifelong Health.” Pediatrics Jacob, Brian A. 2005. “Accountability, Incentives, and Behav- 129 (1): e224-e231. ior: The Impact of High-Stakes Testing in the Chicago Gertler, Paul J., James J. Heckman, Rodrigo Pinto, Arianna Public Schools.” Journal of Public Economics 89 (5): 761–96. Zanolini, Christel Vermeersch, Susan Walker, Susan M. — —— —. 2007. “Test-Based Accountability and Student Chang, et al. 2014. “Labor Market Returns to an Early Achievement: An Investigation of Differential Perfor- Childhood Stimulation Intervention in Jamaica.” Science mance on NAEP and State Assessments.” NBER Working 344 (6187): 998–1001. Paper 12817, National Bureau of Economic Research, Cambridge, MA. Glewwe, Paul W., Eric A. Hanushek, Sarah D. Humpage, and Kiessel, Jessica, and Annie Duflo. 2014. “Cost Effectiveness Renato Ravina. 2011. “School Resources and Educational Report: Teacher Community Assistant Initiative (TCAI).” Outcomes in Developing Countries: A Review of the Lit- IPA Brief (March 26), Innovation for Poverty Action, erature from 1990 to 2010.” NBER Working Paper 17554, New Haven, CT. National Bureau of Economic Research, Cambridge, MA. Kingdon, Geeta. 1996. “The Quality and Efficiency of Private Glewwe, Paul W., Michael Kremer, and Sylvie Moulin. 2009. and Public Education: A Case-Study of Urban India.” “Many Children Left Behind? Textbooks and Test Scores Oxford Bulletin of Economics and Statistics 58 (1): 57–82. in Kenya.” American Economic Journal: Applied Economics Kremer, Michael R. 1995. “Research on Schooling: What We 1 (1): 112–35. Know and What We Don’t, a Comment on Hanushek.” Gove, Amber, and Peter Cvelich. 2011. “Early Reading, Ignit- World Bank Research Observer 10 (2): 247–54. ing Education for All: A Report by the Early Grade Learn- Kremer, Michael R., Edward Miguel, and Rebecca Thornton. ing Community of Practice.” Rev. ed. Research Triangle 2009. “Incentives to Learn.” Review of Economics and Statis- Park, NC: Research Triangle Institute. tics 91 (3): 437–56. Grantham-McGregor, Sally M., Lia C. H. Fernald, Rose M. C. Kuhl, Patricia K. 2010. “Brain Mechanisms in Early Language Kagawa, and Susan Walker. 2014. “Effects of Integrated Acquisition.” Neuron 67 (5): 713–27. Child Development and Nutrition Interventions on Lauwerier, Thibaut, and Abdeljalil Akkari. 2015. “Teachers Child Development and Nutritional Status.” Annals of the and the Quality of Basic Education in Sub-Saharan New York Academy of Sciences 1308 (1): 11–32. Africa.” ERF Working Paper 11, Education Research and Green, Duncan. 2016. How Change Happens. Oxford, U.K.: Foresight, Paris. Oxford University Press. Lavinas, Lena, and Alinne Veiga. 2013. “Brazil’s One Laptop Grindle, Merilee Serrill. 2004. Despite the Odds: The Conten- Per Child Program: Impact Evaluation and Implementa- tious Politics of Education Reform. Princeton, NJ: Princeton tion Assessment.” Cadernos de Pesquisa 43 (149). University Press. Layton, Lyndsey. 2015. “Study: Billions of Dollars in Guilfoyle, Christy. 2006. “NCLB: Is There Life Beyond Test- Annual Teacher Training Is Largely a Waste.” Washington ing?” Educational Leadership 64 (3): 8–13. Post, August 4. https://www.washingtonpost.com/local Hanushek, Eric A. 1986. “The Economics of Schooling: /education/study-billions-of-dollars-in-annual-teacher Production and Efficiency in Public Schools.” Journal of -training-is-largely-a-waste/2015/08/03/c4e1f322-39ff Economic Literature 24 (3): 1141–77. -11e5-9c2d-ed991d848c48_story.html. ————. 1992. “The Trade-Off between Child Quantity and Leder, Gilah C. 1987. “Teacher Student Interaction: A Case Quality.” Journal of Political Economy 100 (1): 84–117. Study.” Educational Studies in Mathematics 18 (3): 255–71. ————. 1995. “Interpreting Recent Research on Schooling in Lemos, Renata, and Daniela Scur. 2016. “Developing Man- Developing Countries.” World Bank Research Observer agement: An Expanded Evaluation Tool for Developing 10 (2): 227–46. Countries.” RISE Working Paper 16/007, Research on Hanushek, Eric A., Victor Lavy, and Kohtaro Hitomi. 2008. Improving Systems of Education, Blavatnik School of “Do Students Care about School Quality? Determinants Government, Oxford University, Oxford, U.K. 32 | World Development Report 2018 Levitt, Steven D., John A. List, Susanne Neckermann, and Chestnut Hill, MA. https://timssandpirls.bc.edu/pirls Sally Sadoff. 2016. “The Behavioralist Goes to School: 2011/international-results-pirls.html. Leveraging Behavioral Economics to Improve Educa- Mullis, I. V. S., M. O. Martin, P. Foy, and M. Hooper. tional Performance.” American Economic Journal: Economic 2016.  “TIMSS 2015 International Results in Mathemat- Policy 8 (4): 183–219. ics.” TIMSS and PIRLS International Study Center, Bos- Levy, Brian, Robert Cameron, Ursula Hoadley, and Vinothan ton College, Chestnut Hill, MA. http://timssandpirls.bc Naidoo. 2016. “The Politics of Governance and Basic Edu- .edu/timss2015/international-results/. cation: A Tale of Two South African Provinces.” Occa- Muralidharan, Karthik, Abhijeet Singh, and Alejandro sional Working Paper 2, Graduate School of Develop- Ganimian. 2016. “Disrupting Education? Experimental ment Policy and Practice, University of Cape Town, Cape Evidence on Technology-Aided Instruction in India.” Town. NBER Working Paper 22923, National Bureau of Eco- Liang, Xiaoyan, Huma Kidwai, and Minxuan Zhang. 2016. nomic Research, Cambridge, MA. How Shanghai Does It: Insights and Lessons from the Muralidharan, Karthik, and Venkatesh Sundararaman. 2011. Highest-Ranking Education System in the World. Directions “Teacher Performance Pay: Experimental Evidence from in Development: Human Development Series. Washing- India.” Journal of Political Economy 119 (1): 39–77. ton, DC: World Bank. Muralidharan, Karthik, and Yendrick Zieleniak. 2013. “Mea- Linden, Leigh L. 2008. “Complement or Substitute? The suring Learning Trajectories in Developing Countries Effect of Technology on Student Achievement in India.” with Longitudinal Data and Item Response Theory.” Edited by Michael Trucano. InfoDev Working Paper 17 Paper presented at Young Lives Conference, Oxford (June), World Bank, Washington, DC. University, Oxford, U.K., July 8–9. Lupien, Sonia J., Suzanne King, Michael J. Meaney, and Neal, Derek. 2013. “The Consequences of Using One Assess- Bruce S. McEwen. 2000. “Child’s Stress Hormone Levels ment System to Pursue Two Objectives.” Journal of Eco- Correlate with Mother’s Socioeconomic Status and nomic Education 44 (4): 339–52. Depressive State.” Biological Psychiatry 48 (10): 976–80. Nelson, Charles A. 2016. “Brain Imaging as a Measure of Mansuri, Ghazala, and Vijayendra Rao. 2013. Localizing Devel- Future Cognitive Outcomes: A Study of Children in Ban- opment: Does Participation Work? Policy Research Report gladesh Exposed to Multiple Levels of Adversity.” Presen- Series. Washington, DC: World Bank. tation, CMU Department of Psychology Colloquium, Martorell, Paco, and Isaac McFarlin Jr. 2011. “Help or Hin- Department of Psychology, College of Humanities and drance? The Effects of College Remediation on Academic Social Sciences, Carnegie Mellon University, Pittsburgh, and Labor Market Outcomes.” Review of Economics and September 29. Statistics 93 (2): 436–54. Nguyen, Trang. 2008. “Information, Role Models, and Per- McCoy, Dana Charles, Evan D. Peet, Majid Ezzati, Goodarz ceived Returns to Education: Experimental Evidence Danaei, Maureen M. Black, Christopher R. Sudfeld, from Madagascar.” Economics Department, Massachu- Wafaie Fawzi, et al. 2016. “Early Childhood Developmental setts Institute of Technology, Cambridge, MA. Status in Low- and Middle-Income Countries: National, OECD (Organisation for Economic Co-operation and Devel- Regional, and Global Prevalence Estimates Using Predic- opment). 2011. Strong Performers and Successful Reformers in tive Modeling.” PLOS Medicine 13 (6): e1002034. Education: Lessons from PISA for the United States. Paris: McEwan, Patrick J. 2015. “Improving Learning in Primary OECD. Schools of Developing Countries: A Meta-Analysis of —— ——. 2016. PISA 2015 Results: Excellence and Equity in Educa- Randomized Experiments.” Review of Educational Research tion. Vol. 1. Paris: OECD. 85 (3): 353–94. Orazem, Peter F. 2000. “The Urban and Rural Fellowship Mingat, Alain, and Jee-Peng Tan. 1998. “The Mechanics of School Experiments in Pakistan: Design, Evaluation, and Progress in Education: Evidence from Cross-Country Sustainability.” Economics of Education Review 22 (3): Data.” Policy Research Working Paper 2015, World Bank, 265–74. Washington, DC. Orazem, Peter F., Paul W. Glewwe, and Harry Patrinos. 2007. Mizala, Alejandra, and Ben Ross Schneider. 2014. “Negotiat- “The Benefits and Costs of Alternative Strategies to ing Education Reform: Teacher Evaluations and Incen- Improve Educational Outcomes.” Department of Eco- tives in Chile (1990–2010).” Governance 27 (1): 87–109. nomics Working Paper 07028, Iowa State University, Morgan, Claire, Anthony Petrosino, and Trevor Fronius. Ames. 2012. “A Systematic Review of the Evidence of the Impact Park, Rufina Kyung Eun. 2016. “Preparing Students for South of Eliminating School User Fees in Low-Income Develop- Korea’s Creative Economy: The Successes and Challenges ing Countries.” Evidence for Policy and Practice Informa- of Educational Reform” [refers to the Republic of Korea]. tion and Co-ordinating Centre, Social Science Research Research Report, Asia Pacific Foundation of Canada, Unit, Institute of Education, University of London. Vancouver. Mulkeen, Aidan G. 2010. Teachers in Anglophone Africa: Issues in Pascale, Richard T., Jerry Sternin, and Monique Sternin. Teacher Supply, Training, and Management. Development 2010. The Power of Positive Deviance: How Unlikely Innovators Practice in Education Series. Washington, DC: World Bank. Solve the World’s Toughest Problems. Boston: Harvard Busi- Mullis, I. V. S., M. O. Martin, P. Foy, and K. T. Drucker. 2012. ness Press. “PIRLS 2011 International Results in Reading.” TIMSS PASEC (Programme d’Analyse des Systèmes Éducatifs de la and PIRLS International Study Center, Boston College, Confemen). 2015. PASEC 2014: Education System Performance Overview | 33 in Francophone Africa, Competencies and Learning Factors in Lopez-Boo, et al. 2015. “Wealth Gradients in Early Child- Primary Education. Dakar, Senegal: PASEC. hood Cognitive Development in Five Latin American Paxson, Christina H., and Norbert R. Schady. 2007. “Cogni- Countries.” Journal of Human Resources 50 (2): 446–63. tive Development among Young Children in Ecuador: Scott-Clayton, Judith, and Olga Rodriguez. 2014. “Develop- The Roles of Wealth, Health, and Parenting.” Journal of ment, Discouragement, or Diversion? New Evidence on Human Resources 42 (1): 49–84. the Effects of College Remediation Policy.” Education Piper, Benjamin, Evelyn Jepkemei, Dunston Kwayumba, and Finance and Policy 10 (1): 4–45. Kennedy Kibukho. 2015. “Kenya’s ICT Policy in Practice: Singh, Abhijeet. 2015. “Learning More with Every Year: The Effectiveness of Tablets and E-readers in Improving School Year Productivity and International Learning Student Outcomes.” FIRE: Forum for International Research Divergence.” CESifo Area Conference on the Economics in Education 2 (1): 3–18. of Education, CESifo Group, Munich, September 11–12. Pradhan, Menno, Daniel Suryadarma, Amanda Beatty, Snilstveit, Birte, Jennifer Stevenson, Radhika Menon, Daniel Maisy Wong, Arya Gaduh, Armida Alisjahbana, and Phillips, Emma Gallagher, Maisie Geleen, Hannah Jobse, Rima Prama Artha. 2014. “Improving Educational Qual- et al. 2016. “The Impact of Education Programmes on ity through Enhancing Community Participation: Learning and School Participation in Low- and Middle- Results from a Randomized Field Experiment in Indone- Income Countries: A Systematic Review Summary sia.” American Economic Journal: Applied Economics 6 (2): Report.” 3ie Systematic Review Summary 7, Interna- 105–26. tional Initiative for Impact Evaluation, London. http:// Pritchett, Lant. 2013. The Rebirth of Education: Schooling Ain’t www.3ieimpact.org/media/filer_public/2016/09/20/srs7 Learning. Washington, DC: Center for Global Develop- -education-report.pdf. ment; Baltimore: Brookings Institution Press. Solano-Flores, Guillermo, Luis Ángel Contreras-Niño, and Pritchett, Lant, and Amanda Beatty. 2015. “Slow Down, You’re Eduardo Backhoff Escudero. 2005. “The Mexican Trans- Going Too Fast: Matching Curricula to Student Skill lation of TIMSS-95: Test Translation Lessons from a Levels.” International Journal of Educational Development 40: Post-mortem Study.” Paper presented at Annual Meeting, 276–88. National Council on Measurement in Education, Mon- Pritchett, Lant, and Justin Sandefur. 2013. “Context Matters treal, April 12–14. for Size: Why External Validity Claims and Development Spaull, Nicholas, and Janeli Kotze. 2015. “Starting Behind and Practice Do Not Mix.” Journal of Globalization and Develop- Staying Behind in South Africa: The Case of Insurmount- ment 4 (2): 161–98. able Learning Deficits in Mathematics.” International Rajkumar, Andrew Sunil, and Vinaya Swaroop. 2008. “Public Journal of Educational Development 41: 13–24. Spending and Outcomes: Does Governance Matter?” Suryadarma, Daniel. 2012. “How Corruption Diminishes the Journal of Development Economics 86 (1): 96–111. Effectiveness of Public Spending on Education in Indo- Rivkin, Steven G., Eric A. Hanushek, and John F. Kain. 2005. nesia.” Bulletin of Indonesian Economic Studies 48 (1): 85–100. “Teachers, Schools, and Academic Achievement.” Econo- Tan, Jee-Peng, and Alain Mingat. 1992. Education in Asia: metrica 73 (2): 417–58. A Comparative Study of Cost and Financing. World Bank Robinson, Viviane M. J., Claire A. Lloyd, and Kenneth J. Regional and Sectoral Studies Series. Washington, DC: Rowe. 2008. “The Impact of Leadership on Student World Bank. Outcomes: An Analysis of the Differential Effects of Tandon, Prateek, and Tsuyoshi Fukao. 2015. Educating the Leadership Types.” Educational Administration Quarterly Next Generation: Improving Teacher Quality in Cambodia. 44 (5): 635–74. Directions in Development: Human Development Rockoff, Jonah E. 2004. “The Impact of Individual Teachers Series. Washington, DC: World Bank. on Student Achievement: Evidence from Panel Data.” Todd, Alexa, and Mark Mason. 2005. “Enhancing Learning in American Economic Review 94 (2): 247–52. South African Schools: Strategies beyond Outcomes- Romer, Paul Michael. 2015. “Botox for Development.” Paul Based Education.” International Journal of Educational Romer’s Blog, September 13. https://paulromer.net/botox Development 25 (3): 221–35. -for-development/. Tooley, James, and Pauline Dixon. 2007. “Private Education RTI International. 2009. “Early Grade Reading Assessment for Low-Income Families: Results from a Global Research Toolkit.” Research Triangle Institute, Research Triangle Project.” In Private Schooling in Less Economically Developed Park, NC. Countries: Asian and African Perspectives, edited by Prachi Sabarwal, Shwetlena, David K. Evans, and Anastasia Mar- Srivastava and Geoffrey Walford, 15–39. Oxford Studies shak. 2014. “The Permanent Input Hypothesis: The Case in Comparative Education Series. Oxford, U.K.: Sympo- of Textbooks and (No) Student Learning in Sierra Leone.” sium Books. Policy Research Working Paper 7021, World Bank, UIS (UNESCO Institute for Statistics). 2006. Teachers and Washington, DC. Educational Quality: Monitoring Global Needs for 2015. Save the Children. 2013. “Ending the Hidden Exclusion: Montreal: UIS. Learning and Equity in Education Post-2015.” Education —— — —. 2016. “Sustainable Development Data Digest: Laying Global Initiative, Save the Children International, the Foundation to Measure Sustainable Development London. Goal 4.” UIS, Montreal. Schady, Norbert R., Jere Behrman, Maria Caridad Araujo, UIS (UNESCO Institute for Statistics) and EFA (Education Rodrigo Azuero, Raquel Bernal, David Bravo, Florencia for All). 2015. “A Growing Number of Children and 34 | World Development Report 2018 Adolescents Are Out of School as Aid Fails to Meet the Us about the Effect of Leadership on Student Achieve- Mark.” Policy Paper 22/Fact Sheet 31, UIS, Montreal; EFA, ment.” McRel Working Paper, McRel International, Paris. Denver. UNESCO (United Nations Educational, Scientific, and Cul- Wolf, Alison. 2004. “Education and Economic Performance: tural Organization). 2013. Third Regional Comparative Simplistic Theories and Their Policy Consequences.” and Explanatory Study (TERCE). UNESCO Regional Oxford Review of Economic Policy 20 (2): 315–33. Bureau for Education in Latin America and the Carib- World Bank. 2003. World Development Report 2004: Making bean, Santiago, Chile. http://www.unesco.org/new/en Services Work for Poor People. Washington, DC: World /santiago/education/education-assessment-llece/third Bank; New York: Oxford University Press. -regional-comparative-and-explanatory-study-terce/. ——— —. 2011. World Development Report 2012: Gender Equality ——— —. 2016. Global Education Monitoring Report 2016, Educa- and Development. Washington, DC: World Bank. tion for People and Planet: Creating Sustainable Futures for All. ——— —. 2013. Service Delivery Indicators (database). World Paris: UNESCO. http://unesdoc.unesco.org/images/0024 Bank, Washington, DC. http://datatopics.worldbank.org /002457/245752e.pdf. /sdi/. USAID (U.S. Agency for International Development). 2017. ——— —. 2014. “STEP Skills Measurement Surveys: Innovative Early Grade Reading Barometer. Washington, DC. http:// Tools for Assessing Skills.” Social Protection and Labor www.earlygradereadingbarometer.org/. Discussion Paper No. 1421. Washington, DC. http:// Uwezo. 2014. “Are Our Children Learning? Literacy and documents.worldbank.org/curated/en/51674146817873 Numeracy across East Africa 2013.” Twaweza, Nairobi. 6065/STEP-skills-measurement-surveys-innovative Wales, Joseph, Ahmed Ali, and Susan Nicolai. 2014. “Improve- -tools-for-assessing-skills. ments in the Quality of Basic Education: Chile’s Experi- ——— —. 2016a. Assessing Basic Education Service Delivery in the ence.” With Francisca Morales and Daniel Contreras. Philippines: The Philippines Public Education Expenditure Case Study Report: Education, Overseas Development Tracking and Quantitative Service Delivery Study. Report Institute, London. AUS6799. Washington, DC: World Bank. Walker, Susan P., Theodore D. Wachs, Julie Meeks Gardner, ——— —. 2016b. “Francophone Africa Results Monitor: Basic Betsy Lozoff, Gail A. Wasserman, Ernesto Pollitt, Julie A. Education (Multiple Countries).” World Bank, Washing- Carter, and the International Child Development Steer- ton, DC. http://documents.worldbank.org/curated/en/doc ing Group. 2007. “Child Development: Risk Factors for search/projects/P156307. Adverse Outcomes in Developing Countries.” Lancet 369 ——— —. 2017. “Driving Performance from the Center: (9556): 145–57. Malaysia’s Experience with Pemandu.” Knowledge and Waters, Tim, Robert J. Marzano, and Brian McNulty. 2003. Research: The Malaysia Development Experience Series, “Balanced Leadership: What 30 Years of Research Tells World Bank, Kuala Lumpur. Overview | 35 PART I Education’s promise 1 Schooling, learning, and the promise of education 1 Schooling, learning, and the promise of education No one has yet realized the wealth of sympathy, the kindness and generosity “ hidden in the soul of a child. The effort of every true education should be to unlock that treasure.” EMMA GOLDMAN In the long run, the best way to reduce inequalities with respect to labor as well as “ to increase the average productivity of the labor force and the overall growth of the economy is surely to invest in education.” THOMAS PIKETTY, CAPITAL IN THE TWENTY-FIRST CENTURY Education is a basic human right, and it is central to the potential to reach them. These benefits are both unlocking human capabilities. It also has tremendous monetary and nonmonetary for individuals, families, instrumental value. Education raises human capital, communities, and society as a whole (table 1.1). productivity, incomes, employability, and economic Most people—whether policy makers or parents— growth. But its benefits go far beyond these mon- already recognize the great value of education.3 etary gains: education also makes people healthier Families around the world make great sacrifices to and gives them more control over their lives. And it keep their children in good schools, and political and generates trust, boosts social capital, and creates insti- opinion leaders consistently rank education among tutions that promote inclusion and shared prosperity. their top development priorities. For that reason, this chapter does not try to review all the evidence on the benefits of education. But before launching into the Education as freedom main theme of this Report—the learning crisis and Since 1948, education has been recognized as a basic what to do about it—it is worth surveying briefly human right, highlighting its role as a safeguard for the many ways in which education can contribute human dignity and a foundation of freedom, justice, to progress, highlighting that these benefits often and peace.1 In the language of Amartya Sen’s capabil- depend on learning, not just schooling.4 ity approach, education increases both an individual’s assets and his or her ability to transform them into well-being—or what has been called the individual’s Education improves “beings and doings” and “capabilities.”2 Education can individual freedoms have corresponding salutary effects on communities and societies. Education improves economic opportunities Education expands freedom through many Education is a powerful tool for raising incomes. channels, both raising aspirations and increasing Education makes workers more productive by giving 38 | World Development Report 2018 Table 1.1 Examples of education’s benefits Individual/family Community/society Monetary Higher probability of employment Higher productivity Greater productivity More rapid economic growth Higher earnings Poverty reduction Reduced poverty Long-run development Nonmonetary Better health Increased social mobility Improved education and health of children/family Better-functioning institutions/service delivery Greater resilience and adaptability Higher levels of civic engagement More engaged citizenship Greater social cohesion Better choices Reduced negative externalities Greater life satisfaction Source: WDR 2018 team. them the skills that allow them to increase their out- less educated workers to lose their jobs, and if they do put.5 Each additional year of schooling typically raises they are more likely to find another job. Educated work- an individual’s earnings by 8–10 percent, with larger ers are more attached to the firms they work for. They increases for women (figure 1.1).6 This is not just are also more effective at acquiring and processing because higher-ability or better-connected people job search information.8 Research in Finland and the (who would earn more regardless of their schooling) United States finds that more schooling makes it eas- receive more education, as proposed by the signaling ier for unemployed people to find reemployment.9 In model of education. “Natural experiments” from a less developed economies with large informal sectors wide variety of countries—such as Honduras, Indo- and underemployment, education is associated with nesia, the Philippines, the United Kingdom, and the greater access to full-time jobs in the formal sector.10 United States—prove that schooling does drive the increased earnings (box 1.1).7 Education leads to longer lives and In well-functioning labor markets, education enables better life choices reduces the likelihood of unemployment. In these Education promotes longer, healthier lives. Around economies, high school graduates are less likely than the world, there are strong links among education, Figure 1.1 More schooling is systematically associated with higher wages Median percentage increase in wages associated with each additional year of schooling, by country group and gender 16 12 Percent 8 4 0 Sub-Saharan Latin America East Asia Middle East South Asia Europe and High-income World Africa and the and Pacific and Central Asia countries Caribbean North Africa Female Male Source: WDR 2018 team, using data from Montenegro and Patrinos (2017). Data at http://bit.do/WDR2018-Fig_1-1. Note: Figure is based on the latest available data, 1992–2012. Regions do not include high-income countries. Schooling, learning, and the promise of education | 39 Box 1.1 Schooling as human capital formation or as a signaling device? Why is education associated with higher earnings? Unlike fall, because the usefulness of the signal component would the human capital model, which posits that education presumably decline with age. Finally, education is an increases a worker’s productivity, the signaling model of expensive screening strategy. education states that individuals acquire education creden- If education worked only as a screening device, individ- tials to signal a high ability to potential employers. Having uals with the same years of schooling should have similar a university degree does signal perseverance, grit, and outcomes regardless of the skills they acquired, which is ability—all valuable skills for the labor market. not the case.a In many countries, individuals with higher But the human capital acquired typically drives the link measured skills have been consistently shown to earn more from schooling to earnings, as different types of evidence than their lower-skilled peers who have the same amount show. First, the returns to an additional year of schooling of schooling.b In Mexico, those high school graduates with for those who drop out without a high school or university higher test scores are substantially less likely to be unem- diploma are as large as for those who complete the degree. ployed three years after leaving school (among those who Second, the wage differentials across education levels rise did not go to university) than their lower-scoring peers.c with age, whereas signaling theory suggests they should Source: WDR 2018 team. a. Layard and Psacharopoulos (1974). b. For example, see the results for Organisation for Economic Co-operation and Development (OECD) countries in Hanushek and others (2015) and Valerio and others (2016). For individual countries such as Ghana, see Glewwe (1991), or for South Africa, see Moll (1998). c. de Hoyos, Estrada, and Vargas (2017). better health outcomes, and longer lives.11 Regardless Figure 1.2 Mortality rates in the United of their race, gender, or income, more-educated indi- States are lower for adults with more viduals in Europe and the United States have a lower education probability of having a chronic health condition.12 Relative odds (log-odds coefficient) of death for groups with In the United States, each additional year of school- different years of education, by age, gender, and race ing is associated with a lower probability of death, especially after high school (figure 1.2). One reason 1.2 is that education makes people less likely to smoke, drink in excess, be overweight, or use illegal drugs.13 0.8 In the United States, education makes people less Odds of death likely to smoke; in Uganda, more-educated individ- 0.4 uals were more responsive to HIV/AIDS information campaigns.14 0 Educated individuals have more control over the life they want to pursue—often called “agency.” Increased –0.4 agency manifests itself as a reduction in risky behavior, higher life satisfaction, and greater happiness. Across –0.8 52 countries at all income levels in 2010–14, only 1 in 0 5 10 15 20 10 university graduates felt that they had little or no Years of education control over their lives. 15 When the United Kingdom Black female, ages 25−64 White female, ages 25−64 and the United States extended compulsory schooling, Black female, ages 65+ White female, ages 65+ people who received more education were less likely to Black male, ages 25−64 White male, ages 25−64 report being unhappy later in life.16 Black male, ages 65+ White male, ages 65+ The positive relationship between education Source: WDR 2018 team, using data from Montez, Hummer, and Hayward and agency is partly mediated by the positive effect (2012). Data at http://bit.do/WDR2018-Fig_1-2. of education on income, but there seems to be an Note: Groups exclude Hispanic population. 40 | World Development Report 2018 independent effect as well: the effects on crime mortality.28 Improvements in women’s education have and fertility, for example, are not contingent only been linked to better health outcomes for their children on income. Schooling reduces most types of crime in many countries, including Brazil, Nepal, Pakistan, committed by adults,17 as well as crime during late and Senegal.29 Parental schooling robustly predicts adolescence.18 Among 16- and 17-year-olds in the higher educational attainment for children, even after United Kingdom, school dropouts are three times controlling for other factors. And children’s ability more likely to commit crimes than those who have to benefit from education is shaped by their parents’ stayed in school, and this gap remains well into their education. In the United States, each additional year early 20s. In Sweden, the United Kingdom, and the of a mother’s schooling increases her children’s math United States, completing high school makes youth test scores by 0.1 standard deviation and significantly less likely to commit crimes, and education is linked reduces behavioral problems.30 In Pakistan, mothers with lower crime rates elsewhere—such as in Mexico, who have one more year of schooling have children where high school dropouts were more caught up in who spend an additional hour a day studying at home.31 the violence of the war on drugs.19 Education’s benefits are especially apparent in As for fertility, education reduces teen pregnancy changing environments. Individuals with stronger and increases the control that women have over the skills can take better advantage of new technologies and size of their families. Schooling reduces teenage adapt to changing work. Indeed, experts on technolog- pregnancy indirectly by increasing girls’ aspirations, ical change have long argued that the more volatile the empowerment, and agency. In Turkey, primary state of technology, the more productive education is.32 school completion induced by a change in compul- Returns to primary schooling in India increased during sory schooling laws—allowing research to isolate the Green Revolution, with the more educated farmers the causal effects—reduced teenage fertility by 0.37 adopting and diffusing new technologies.33 More gen- children per woman.20 School subsidies reduced teen erally, globalization and advances in technology are pregnancy (and in some cases school dropout) in Bra- putting a premium on education and skills—both cog- zil, Colombia, Kenya, Malawi, and Peru.21 More gener- nitive and socioemotional (see spotlight 5). New skills ally, women with more schooling have lower fertility facilitate the adoption of technologies and promote rates. In Brazil, increased schooling among young innovation,34 with general skills enabling individu- women explains 40–80 percent of the decline in als to adapt to the economic changes that occur over the fertility rate that began in the late 1960s.22 When their lifetimes.35 When the North American Free Trade school coverage expanded in Nigeria, each additional Agreement (NAFTA) increased labor productivity in year of female schooling reduced fertility by at least Mexico, the benefits were concentrated among more- 0.26 births per woman.23 One reason may be that edu- skilled workers in the richer northern states.36 In gen- cated women earn more, making it costlier for them eral, returns to education are higher in economically to leave the labor market.24 Education also increases free countries with institutions that allow individuals women’s use of contraception, increases their role in to adjust to shocks and market forces.37 family decisions on fertility, and makes them more aware of the trade-offs in having children.25 Education benefits all of The benefits of education are long-lasting society Education can eliminate poverty in families. The Education builds human capital, which translates incomes of parents and their children are highly into economic growth. If improvements are faster correlated: income inequality persists, and poverty among the disadvantaged, the additional growth will is transmitted from one generation to the next.26 But reduce poverty, reduce inequality, and promote social improving education gives poor children a boost: in the mobility. Through its effect on civic agency—mean- United States, the children of households that moved ing high levels of political engagement, trust, and to a (one standard deviation) better neighborhood tolerance—education can create the building blocks had incomes as adults that were more than 10 percent for more inclusive institutions.38 Greater civic agency higher, in part because the move improved learning.27 can create a political constituency for inclusive insti- Better-educated mothers raise healthier and more tutions, strengthening the social contract between educated children. Women’s education is linked to the state and its citizens. A more engaged citizenry many  health benefits for their children, from higher can also provide political support for the reforms immunization rates to better nutrition to lower needed to realize the promise of education. Schooling, learning, and the promise of education | 41 Education promotes economic growth for all children early on, followed by expansion of At the national level, education underpins growth. high-quality secondary and tertiary opportunities.46 Human capital can boost growth in two ways: first, These cases reinforce the idea that strong founda- by improving the capacity to absorb and adapt new tional skills drive growth early in development, but technology, which will affect short- to medium-term also that as countries approach the global technologi- growth, and, second, by catalyzing the technological cal frontier, they need to invest more in higher educa- advances that drive sustained long-term growth.39 tion and in research and development.47 Widespread basic education may provide a bigger As education coverage expands, poor people typ- boost for countries far from the global technological ically benefit the most at the margin, and so income frontier—a group that includes most low- and middle- inequality should fall.48 A review of more than 60 income countries.40 These countries do not need to studies reveals that greater education coverage is push that frontier out through innovation, but they associated with substantial reductions in the income do need widespread basic education to absorb and gap between households across the income distribu- adapt the technologies that are already available glob- tion. Specifically, going from a primary enrollment ally. In countries close to the technological frontier, rate of 50 to 100 percent is associated with an 8 per- mainly high-income countries, higher levels of educa- centage point increase in the share of income going tion can boost growth through innovation.41 Although to households in the poorest decile.49 data limitations make empirical analysis of this rela- tionship challenging, many influential studies have Education creates the building blocks for concluded that higher levels of education do drive inclusive institutions more rapid growth.42 Growth accounting analyses Education strengthens the political development of also suggest that education can explain a significant nations by promoting the civic engagement of their share of growth—a share that may be even larger if populations.50 People with more education consistently unskilled workers are more productive when there participate more in political activities than those with are more skilled workers in an economy.43 less education: education increases awareness and But this statistical evidence is not the only—or understanding of political issues, fosters the socializa- even the most compelling—evidence on the impact of tion needed for effective political activity, and increases education on growth. Countries that have sustained civic skills.51 Evidence from a variety of settings shows rapid growth over decades have typically shown a that this relationship is causal.52 In the United States, strong public commitment to expanding education, getting more education—for example, as a result of pre- as well as infrastructure and health.44 Although the school programs, high school scholarships, or smaller relationship flows the other way as well—in that class sizes—leads people to vote more often (table 1.2).53 rapid growth allows greater investment in all three Using changes in compulsory school laws to identify sectors—research on the East Asian miracle coun- the causal impact of education confirms these findings tries in particular flags education and human capital for the United Kingdom and the United States, while as factors in their rapid growth.45 Countries such as using access to community college or changes in child the Republic of Korea reaped the benefits of their labor laws does so for the United States.54 In Benin, “progressive universalism” approach to education, receiving more education made people more politically in which they ensured high-quality basic education active over their lifetimes. In Nigeria, too, educational Table 1.2 More schooling leads to more voting Percent Graduated from high school Voted Program Control Treatment Control Treatment Perry Preschool experiment 44 65 13 18 “I Have a Dream” scholarships 62 79 32 42 STAR Experiment 85 90 42 47 Source: Sondheimer and Green (2010). Note: The Perry Preschool experiment was an intensive effort to enroll children from low-income families in preschool in Ypsilanti, Michigan. The “I Have a Dream” scholarships were high school scholarships targeted to fifth-grade students who qualified (because of their family’s poverty status) for free or reduced-price lunch in Lafayette, Colorado. The STAR Experiment assigned some students in kindergarten through grade 3 in Tennessee to smaller class sizes. The measure of voting differs across the studies, but corresponds to a time between 2000 and 2004 when the participants would have already graduated from high school. 42 | World Development Report 2018 Figure 1.3 People with higher education hold stronger beliefs about the importance of democracy Percentage of population that believes it is “absolutely important to live in a democracy,” by country and level of education a. Low- and middle-income countries b. High-income countries 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 People with tertiary education People with primary education Source: WDR 2018 team, using data from World Values Survey (World Values Survey Association 2015). Data at http://bit.do/WDR2018-Fig_1-3. expansion substantially increased the civic and politi- teaching styles that encourage teamwork rather than cal engagement of its beneficiaries decades later.55 a more top-down pedagogy appear to promote social As with the other effects of education, context capital: students are more likely to believe in the matters in how education affects political views importance of civic life and the value of cooperation.60 and engagement. In an indicator of perceptions of Education makes institutions work better and one common mechanism for political participation, improves public services. Educated parents are better surveys in 30 developing countries show that more- able to leverage decision-making authority at the educated citizens are more likely to believe that living school level. In The Gambia, a school-based manage- in a democracy is important (figure 1.3). But in Kenya, ment program improved student learning—but only although more education caused young women to when there was a high level of literacy in the village.61 have more political knowledge, it also led them to be A more educated population generally demands more more disenchanted and more accepting of political transparent use of public resources, better service violence, perhaps because democratic institutions delivery, and government accountability. Recent were particularly fragile at the time of the research.56 cross-country research identifies citizen complaints Education increases trust, tolerance, and civic as a primary mechanism: educated citizens complain agency. Evidence from member countries of the more, inducing officials to behave better.62 Education Organisation for Economic Co-operation and Devel- also appears to improve dimensions of governance: opment (OECD), as well as from developing countries, countries that had achieved mass education by 1870 indicates that more-educated individuals are more had less corruption in 2010.63 trusting and tolerant of people they know and even Growth built on human capital rather than other of strangers.57 Although such cross-sectional evidence sources (such as natural resources) may lead to fewer cannot prove a causal relationship, historical analysis incentives for conflict, for three main reasons.64 First, suggests a mechanism: the spread of literacy may have because human capital is difficult to appropriate, contributed to a generalized decline in violence after conquest of a well-educated population may be less the Middle Ages, because the ability to read others’ rewarding than seizure of natural resources or even viewpoints promoted empathy.58 Some educational physical capital.65 Second, education raises the oppor- environments appear to promote trust especially tunity cost of fighting: it is easier to recruit people well. Data for 28 countries reveal that the openness of who have poor job prospects.66 Third, as discussed, a classroom climate, or the “degree to which students education can promote tolerance and cooperation, are able to discuss political and social issues in class,” thereby reducing the propensity to turn to violence to is positively linked to trust and tolerance.59 Similarly, resolve conflicts.67 Schooling, learning, and the promise of education | 43 Learning and the promise of the economy and society (box 1.2). Another problem is that if an education system is managed poorly, it can education promote social “bads” instead of social “goods.” First, Education can be a powerful tool for individual and education can deepen cleavages between favored societal empowerment, but its benefits are not auto- and disadvantaged groups. Young people from poor, matic. It is not just that education cannot do it alone, rural, and otherwise disadvantaged households not in that much also has to go right in other sectors of only complete less schooling, but also learn much Box 1.2 Education can’t do it alone Economics, politics, and society shape the returns to educa- education might be high (for those who land public sector tion. Education systems do not function in a vacuum; they jobs), but the impact of education on growth will be low are part of broader economic, political, and social institu- because improved cognitive skills are not used in ways that tions. For example, does a society uphold property rights? will increase productivity the most.i If not, entrepreneurs are unlikely to invest in risky new ven- Discriminatory norms distort the benefits of education. tures, which cuts into job creation and reduces education’s Prevailing norms on ethnic or gender discrimination can returns in the labor market. Are there regulations to pre- strongly mediate the returns to education for these groups. In vent fraud? If not, those with education might find it more many societies, social norms severely restrict women’s access profitable to engage in socially unproductive but financially to economic opportunities.j Two studies found that nearly 90 remunerative activities. Are women restricted from working percent of women in northern India (from the state of Uttar outside the home? If so, the economic returns from educa- Pradesh) and Nigeria (of Hausa ethnicity) felt they needed tion will be unavailable to them. These are all examples of their husband’s permission to work. But norms vary substan- how formal or informal institutions influence education’s tially: in the Ethiopian capital, this share was only 28 percent.k returns. In general, reliable institutions that implement the Such norms do not always operate through open dis- rule of law, reduce corruption, and protect property rights crimination. Labor market segregation along occupational are associated with higher returns to human capital.a and social lines is often covert. Occupational gender seg- Here are several examples of how problems elsewhere regation is a strong feature of many labor markets across in the economy or society reduce education’s returns: the world.l In Organisation for Economic Co-operation and Low demand for educated labor reduces the return to Development (OECD) countries, women dominate the ser- skills. Education’s returns depend on the interplay between vice sector, whereas men are overrepresented in industry.m demand and supply forces in the labor market. If the In addition to horizontal segregation, women also face a demand for educated labor is low relative to supply, then “glass ceiling” or “vertical segregation” because they do not the returns to education will be low or declining.b In urban advance in their careers as fast or as far as men. In OECD China, the returns to education rose from 4 percent a year countries, just a third of managers were women in 2013, with of schooling in 1988 to 10 percent in 2001, with most of the small variations across countries.n Labor market segregation increase attributable to institutional reforms that increased may also exist along socioeconomic lines.o In the 1960s and the demand for skilled labor.c More generally, shifts from 1970s, during a period of rapid economic growth in Chile, planned to market economies have increased the returns education was significant in determining occupational to human capital.d When the investment climate is poor,e attainment for the middle class. For the upper class and the both investment and demand for labor by private firms are very poor, education was less important, and intergenera- lower, reducing the returns to education.f tional status inheritance was much more likely.p In Jamaica, Countries can incentivize the wrong things. Many edu- a country with a rigid class structure, the massive expansion cated youth in parts of the developing world queue for of educational opportunities at the secondary level did little jobs in already large public sectors. In several countries, to increase the permeability of social structure.q political candidates compete in terms of their ability to The very people who are constrained by social norms offer patronage or public employment to their supporters.g may become complicit in perpetuating them. A study of stu- In several North African countries, for example, it was not dents newly admitted to an elite master’s in business admin- uncommon in the past for governments to guarantee public istration (MBA) program in the United States found that employment opportunities for all university graduates, and single women reported lower desired compensation when the public sector remains the employer of a large share they believed their classmates would see their responses. of wage earners.h In such situations, individual returns to No such differences were observed for men or for women (Box continues next page) 44 | World Development Report 2018 Box 1.2 Education can’t do it alone (continued) who were not single, suggesting that single women were 45 percent of employees reported that they helped a reluctant to signal personality traits, such as ambition, that friend or relative get a job with their current employer.u they perceived to be undesirable in the marriage market.r Nearly 60 percent of enterprises surveyed in 14 countries Social norms can operate in much the same way to inhibit in Sub-Saharan Africa report that their most recent position male access to opportunities. Case studies in Australia and was filled through contacts with “family/friends.”v This Jamaica suggest that underachievement among boys is finding applies as well to places where labor markets are linked to notions of education being a “feminized” realm segmented by kinship and socioeconomic class.w Informal that clashes with expectations of “masculine” behavior.s networks can also be particularly important for certain When getting a job depends on informal institutions, subpopulations—for example, among Mexican migrants in education is less useful.t In Kolkata (formerly Calcutta), India, the United States.x Source: WDR 2018 team. a. World Bank (2011). k. World Bank (2011). b. Pritchett (2001). l. Hegewisch and Hartmann (2014). c. Zhang and others (2005). m. OECD Employment Statistics Database, http://stats.oecd.org. d. Nee and Matthews (1996). n. OECD Family Database, http://www.oecd.org/els/family/database e. World Bank (2012). .htm. f. Almeida and Carneiro (2005); Besley and Burgess (2004); Botero and o. First described by Blau and Duncan (1967). others (2004); Djankov and others (2002); Haltiwanger, Scarpetta, p. Farrell and Schiefelbein (1985). and Schweiger (2008); Klapper, Laeven, and Rajan (2004); Micco and q. Strudwick and Foster (1991). Pagés (2007); Petrin and Sivadasan (2006). r. Bursztyn, Fujiwara, and Pallais (2017). g. Cammett (2009); Kao (2012); Lust-Okar (2009); Sakai, Jabar, and s. Jha and Kelleher (2006). Dawod (2001). t. Granovetter (1995). h. Bteddini (2016); Egypt Census, 2006, Egypt Data Portal, Central Agency u. Beaman and Magruder (2012). for Public Mobilizations and Statistics, Cairo, http://egypt.opendatafor v. Filmer and Fox (2014). africa.org/EGSNS2006/egypt-census-2006; Ghafar (2016). w. A ssaad (1997); Barsoum (2004); Brixi, Lust, and Woolcock (2015). i. Pritchett (2001). x. Munshi (2003). j. Chiswick (1988); Goldin and Polachek (1987); McNabb and Psacha- ropoulos (1981); World Bank (2011). less while in school (see part II of this Report). In such Figure 1.4 Learning varies widely across countries; cases, education does little to enhance social mobility. in 6 of the 10 countries assessed, only half or fewer Second, leaders sometimes abuse education systems of primary completers can read for political ends and in ways that reinforce autocracy Literacy rates at successive education levels, selected countries or the social exclusion of certain groups. 100 Finally, schooling is not the same as learning. Edu- cation is an imprecise word, and so it must be clearly defined. Schooling is the time a student spends in 75 classrooms, whereas learning is the outcome—what Percent the student takes away from schooling. This distinc- 50 tion is crucial: around the world, many students learn little (figure 1.4). To be sure, many students learn 25 something, even in settings facing huge challenges. And students enjoy some benefits from education 0 regardless of whether they are learning. When schools No formal Some Completed Some Completed Some or serve as oases of security in violent areas, or when education primary primary secondary secondary completed tertiary participation in schooling keeps adolescent girls from Highest level of educational attainment becoming pregnant, these are real societal benefits. When graduates can use their degrees to open doors Bangladesh Indonesia Rwanda Ghana Kenya Tanzania to employment, that opportunity changes their lives, India Nigeria Uganda even when the degree represents less learning than Pakistan it should. Source: Kaffenberger and Pritchett (2017). Data at http://bit.do/WDR2018-Fig_1-4. Intuitively, many of education’s benefits depend on Note: Literacy is defined as being able to read a three-sentence passage either “fluently without help” or the skills that students develop in school. As workers, “well but with a little help.” Schooling, learning, and the promise of education | 45 people need a range of skills—cognitive, socioemo- after controlling for the years of schooling completed, tional, technical—to be productive and innovative. As years of schooling do not predict growth once test parents, they need literacy to read to their children or scores are taken into account (figure 1.5), or they to interpret medication labels, and they need numer- become only marginally significant.70 In other words, acy to budget for their futures. As citizens, people need what matters is less the years of education completed literacy and numeracy, as well as higher-order rea- than the knowledge that students acquire while in soning abilities, to evaluate politicians’ promises. As school. Simulations show that providing all students community members, they need the sense of agency with basic cognitive skills could massively boost eco- that comes from developing mastery. None of these nomic outcomes, especially in developing countries capabilities flows automatically from simply attend- (figure 1.6).71 This finding suggests that cross-country ing school; all depend on learning while in school. comparisons of the years of schooling completed— Research on the benefits of education has begun especially when used to explain economic phenom- to reflect this distinction between schooling and ena—could be misguided if they do not account for learning. In the past, most empirical research equated the differences in skills acquired during those years education with schooling—whether measured by (box 1.3). school enrollment, number of years of schooling, or At the micro level, too, growing evidence shows degrees acquired—in part because of lack of other that skills acquisition determines how much individ- good measures of education. But as the focus on uals gain from schooling. For example, learning—not learning has grown, some studies have explored the just schooling—matters in how education affects effects of the skills that students acquire. The results earnings. Across 23 OECD countries, as well as in a confirm the intuition: skills matter. number of other countries, simple measures of foun- The channel by which schooling accelerates eco- dational skills such as numeracy and reading profi- nomic growth appears to be through boosting learn- ciency explain hourly earnings over and above the ing and skills.68 Thanks to the growing availability of effect of years of schooling completed.72 These effects large-scale student assessments, it is now possible to extend beyond the labor market. Across 10 low- and explore how learning mediates the relationship from middle-income countries, schooling improved mea- schooling to economic growth.69 While the relation- sures of financial behavior only when it was associ- ship between test scores and growth is strong even ated with increased reading ability.73 When people Figure 1.5 What matters for growth is learning Annual average per capita growth in GDP, 1970–2015, conditional on test scores, years of schooling completed, and initial GDP per capita a. Test scores and growth b. Years of schooling and growth (conditional on initial GDP per capita (conditional on the initial GDP and years of schooling) per capita and test scores) 3 3 Annual GDP per capita growth (%) Annual GDP per capita growth (%) 2 2 1 1 0 0 −1 −1 y = 0.00 + 1.59x y = 0.00 + 0.07x −2 t = 7.39 −2 t = 0.82 R2 = 0.55 R2 = 0.02 −1.5 −1 −.5 0 .5 1 −2 −1 0 1 2 3 Test scores Years of schooling Source: WDR 2018 team, using data on test scores from Hanushek and Woessmann (2012) and data on years of schooling and GDP from the World Bank’s World Development Indicators (database), 2017. Data at http://bit.do/WDR2018-Fig_1-5. 46 | World Development Report 2018 Figure 1.6 Increasing learning would yield major economic benefits Simulated additional GDP between 2015 and 2090 attributable to increased learning (relative to current GDP), by scenario, selected countries 2,500 2,000 Percent of current GDP 1,500 1,000 500 0 Tu o G ey Po ece l m ly Un S rg d in Po tes ov N nd pu y n y er rk Ic ny Fr d Ire ce Sw and Sw Au n er a ze Be and ite Rep um O lic H CD ng c N Au om Ze lia N an d he da Ko Ja s a, n Fi p. d ga nd Ki ubli Re wa De gar e re pa ic an itz stri C an an xe Ita ite pa G ma Re u b an rk a ed ra la et a a Un h gi ex d rtu E bo rla re m el l l al nl St ak or ew st un l M d Lu c C Sl Bring all test takers to a minimum of 400 points on PISA Bring each country average to the Finnish average of 546 points on PISA Source: OECD (2010). Data at http://bit.do/WDR2018-Fig_1-6. Note: PISA = Programme for International Student Assessment. had acquired more schooling but not more literacy— such as “Parents love their children” or “Farming is which was common in these countries—financial hard work.” Yet even these highly imperfect mea- behaviors did not change. Socioemotional skills sures of skills have considerable predictive and matter as well: various measures have been shown explanatory power. If better measures of skills were to significantly predict earnings over and above the available, skills would likely explain even more of the effects of schooling and cognitive skills.74 impacts of education—and the role remaining for the Learning matters for health, too. Numerous stud- simple schooling measure (which typically retains ies have documented the benefits of girls’ schooling predictive power in these analyses) would be further on outcomes such as lower fertility or better child sur- diminished. vival, but these studies do not typically distinguish Finally, learning promotes social mobility. The between learning and schooling. There are excep- research cited earlier on intergenerational social tions, however. In Morocco, research showed that mobility in the United States also investigated which maternal education improved child health through educational mechanisms were responsible. One can- its effects on the ability of mothers to acquire health didate is school quality based on inputs, such as school knowledge.75 Globally, data from 48 developing coun- spending and class size, and these measures did have tries show that learning is responsible for much of some predictive power. But learning outcomes turn these gains. Each additional year of female primary out to be especially important: the test scores of the schooling is associated with roughly six fewer deaths community in which a child lives (adjusted for the per 1,000 live births, but the effect is about two-thirds income of that community) are among the strongest larger in the countries where schooling delivers the predictors of social mobility later in life.77 most learning (compared with the least).76 The literature on the benefits of learning is still Even limited measures of skills explain a lot. The growing, with much more research needed. But both measures used in the studies just noted are often common sense and the emerging research literature narrow, capturing only simple numeracy or reading make it clear that if investigators care about the ben- proficiency. Sometimes, the measures are coarse. For efits of education, they should focus on whether stu- example, the 48-country study of the relationship dents are learning—not just on how well schools are between schooling and health uses as its measure of equipped or even how long students stay in school. literacy whether a woman can read a single sentence Part II of this Report takes up this issue. Schooling, learning, and the promise of education | 47 Box 1.3 Comparing attainment across countries and economies— learning-adjusted years of schooling A given number of years in school leads to much more Two important facts support the credibility of this learning in some economies than in others. Because they analysis: first, the TIMSS score ratios across economies for do not account for these differences, standard compari- grade 4 are similar to those for grade 8; and second, PISA sons of schooling attainment may be misleading. But how scores tend to increase linearly across the grades in which should they be adjusted to make meaningful comparisons? that test is administered. One approach is to draw on measures of student learning What might such an adjustment reveal? An illustration that are standardized across different economies to adjust using TIMSS math scores from 2015 confirms that years of for quality. International assessments such as the Trends schooling are indeed very different from learning-adjusted in International Mathematics and Science Study (TIMSS) years, and this difference varies a lot across economies. or the Programme for International Student Assessment Whereas people ages 25–29 in Hong Kong SAR, China, and (PISA) provide such measures. If one is willing to assume the United States have similar average years of schooling that the average learning trajectory across economies is (14 and 13.5, respectively), the number of learning-adjusted linear—starting at no learning when learners enter school schooling years in the United States is almost two years less and growing at a constant rate to grade 8—then the ratio (figure B1.3.1). And whereas young Singaporeans have only of scores across two economies would reflect the relative 30 percent more schooling than young Jordanians by the learning per year in one economy versus the other. For standard measure, the learning-adjusted measure shows example, if economy A has twice the score of economy B in Singapore outpaces Jordan by 109 percent in effective grade 8, then, on average, a year of schooling in economy A schooling years. may be considered twice as effective. Figure B1.3.1 There can be a large gap between learning-adjusted and unadjusted years of schooling Years of actual and learning-adjusted schooling among young people, ages 25–29, illustrated using TIMSS data 15 12 Number of years 9 6 3 0 Isr d Ba alia Ja ly Ka Jo an Ku p. tsw ain C ana a SA ab ile C . Isl un a ic ry Ire ep. l Ko kh n a, n hu it al ia M ia ew or a a o or d Fe Qa y ud ra r Si ra n ap a So Slov ore h ia Sw frica ai n Un ab Tur d ite Em key Un Kin tes d om es ae R, Rep Sa de ta a Lit wa za rda re sta i A tio Th de n Ze cc ad n, H hin N M alt N lan ng bi n Ita am ga M an s ut en Re at ng r Ch la w p la d ira ay ite gd Bo hr R o str e an St A Au A Ko t, n Ar g yp ia ss Ira on g d H E ite Ru Un Actual years Learning-adjusted years Source: WDR 2018 team, using data from Barro and Lee (2013) and TIMSS 2015 (Mullis and others 2016). Data at http://bit.do/WDR2018-Fig_B1-3-1. Note: Years of schooling in Singapore are the same as learning-adjusted years because Singapore, which scored highest on the Trends in International Mathematics and Science Study (TIMSS) mathematics assessment in 2015, serves as the basis for comparison in this illustration. For the purposes of this illustration, data for years of education in the United Kingdom are adjusted using the TIMSS score for England. Note that for all countries and economies, the size of the adjustment will reflect the scale of the metric used to make it. 48 | World Development Report 2018 Notes 1. United Nations (1948). Article 26 of the Universal Dec- 2. Lam, Sedlacek, and Duryea (2016). 2 laration of Human Rights (1948) states: “Everybody has 23. Osili and Long (2008). the right to education. . . . Education shall be directed 24. Becker, Cinnirella, and Woessmann (2013). to the full development of the human personality and 25. Lavy and Zablotsky (2011). to the strengthening of respect for human rights and 26. Solon (1999). fundamental freedoms. It shall promote understanding, 27. Chetty, Hendren, and Katz (2016). tolerance and friendship among all nations, racial or 28. Schultz (1975); Thomas, Strauss, and Henriques (1990); religious groups, and shall further the activities for the Welch (1970); World Bank (2011). maintenance of peace.” 29. World Bank (2011). 2. Sen (1985, 1999, 2004). 30. Carneiro, Meghir, and Parey (2013). 3. For example, see UNESCO (2016) for a comprehensive 31. Andrabi, Das, and Khwaja (2012). discussion of the role of education in the United Nations’ 32. Nelson and Phelps (1966). Sustainable Development Goals. 33. Foster and Rosenzweig (1996). 4. Heckman and others (2014). 34. Aghion and others (2009). 5. Becker (1964). 35. Hanushek and others (2017). 6. Montenegro and Patrinos (2017). 36. Hanson (2007). 7. Angrist and Krueger (1992); Bedi and Gaston (1999); 37. King, Montenegro, and Orazem (2012). Card (1993); Duflo (2000); Harmon and Walker (1995); 38. Chong and Gradstein (2015); Dahl (1998); Dewey (1916). Maluccio (1998). 39. Romer (1990); Solow (1956). 8. Mincer (1991). 40. Aghion (2009); Madsen (2014). 9. Kettunen (1997); Riddell and Song (2011). 41. Acemoglu, Aghion, and Zilibotti (2006); Aghion (2009); 10. Filmer and Fox (2014). Aghion and others (2009). 11. See Cutler, Lleras-Muney, and Vogl (2008) and Vogl 42. Barro (2001); Cohen and Soto (2007); Glewwe, Maiga, (2012) for a review of the evidence in developed and and Zheng (2014); Krueger and Lindahl (2001); Mankiw, developing countries, respectively. Romer, and Weil (1992). 12. Cutler and Lleras-Muney (2007); Mackenbach (2006). 43. Bosworth and Collins (2003); Jones (2014). 13. Although there is reverse causality—better health leads 44. Commission on Growth and Development (2008). to more education—natural experiments such as the 45. World Bank (1993). introduction of minimum schooling laws or military 46. Education Commission (2016). draft avoidance have identified the positive and signif- 47. Aghion and Howitt (2006). icant causal effects of education on health. 48. Lanjouw and Ravallion (1999); Younger (2003). 14. de Walque (2007a, 2007b). 49. Abdullah, Doucouliagos, and Manning (2015). 15. The World Values Survey 2010–14 (Wave 6) covers 57 50. Dewey (1916); Lipset (1959, 1960). developed and developing economies (World Values 51. Campante and Chor (2012). Survey Association 2015). The survey measures the 52. Chzhen (2013). beliefs, values, and motivations of 90,000 survey respon- 53. Sondheimer and Green (2010). dents selected in nationally representative samples, 54. Dee (2004); Milligan, Moretti, and Oreopolous (2004). while also collecting socioeconomic data from those 55. Larreguy and Marshall (2017); Wantchekon, Klasnja, and respondents. Estimations include average weights and Novta (2015). consolidated categories for analysis (education level and 56. Friedman and others (2011). scaled responses). 57. Borgonovi and Burns (2015); Chzhen (2013). 16. Oreopoulos (2007). 58. Pinker (2011). 17. Lochner (2004); Lochner and Moretti (2004). 59. Campbell (2006). 18. Belfield and others (2006); Cullen, Jacob, and Levitt 60. Algan, Cahuc, and Shleifer (2013). (2006). 61. Blimpo, Evans, and Lahire (2015). 19. Anderson (2014); de Hoyos, Gutiérrez Fierros, and Var- 62. Botero, Ponce, and Shleifer (2013). gas M. (2016); Hjalmarsson, Holmlund, and Lindquist 63. Chong and others (2014). (2015); Machin, Marie, and Vujić (2011). At least two 64. de la Brière and others (2017). possible mechanisms could explain why education 65. Acemoglu and Wolitzky (2011). reduces crime. First, because education increases poten- 66. Collier, Hoeffler, and Rohner (2009). tial earnings, it also drives up the opportunity costs of 67. Davies (2004). crime. Second, more schooling may reduce crime simply 68. Glewwe, Maiga, and Zheng (2014); Hanushek and by reducing the time available to young people to com- Woessmann (2008, 2012). mit a crime. Some U.S. data support this “incapacitation 69. Barro (2001, 2013). effect” (Anderson 2014). 70. Barro (2013). 20. Güneş (2016). 71. Hanushek and Woessmann (2015); OECD (2010). 21. Azevedo and others (2012); Baird and others (2010); 72. Hanushek and others (2015); Valerio and others (2016). Duflo, Dupas, and Kremer (2014). 73. Kaffenberger and Pritchett (2017). Schooling, learning, and the promise of education | 49 74. For OECD countries, see Heckman, Stixrud, and Urzua Baird, Sarah Jane, Ephraim Chirwa, Craig McIntosh, and (2006); Heineck and Anger (2010); Mueller and Plug Berk Özler. 2010. “The Short-Term Impacts of a Schooling (2006). For countries outside of OECD, see Díaz, Arias, Conditional Cash Transfer Program on the Sexual and Tudela (2012); Valerio and others (2016). Behavior of Young Women.” Health Economics 19 (S1): 75. Glewwe (1999). 55–68. 76. Oye, Pritchett, and Sandefur (2016). Barro, Robert J. 2001. “Human Capital and Growth.” American 77. Chetty and others (2014). Economic Review 91 (2): 12–17. —— — —. 2013. “Education and Economic Growth.” Annals of Economics and Finance 14 (2): 301–28. References Barro, Robert J., and Jong Wha Lee. 2013. “A New Data Set of Educational Attainment in the World, 1950–2010.” Journal Abdullah, Abdul, Hristos Doucouliagos, and Elizabeth Man- of Development Economics 104: 184–98. ning. 2015. “Does Education Reduce Income Inequality? Barsoum, Ghada F. 2004. “The Employment Crisis of Female A Meta-Regression Analysis.” Journal of Economic Surveys Graduates in Egypt: An Ethnographic Account.” Cairo 29 (2): 301–16. Papers 25 (3). Cairo: American University in Cairo Press. Acemoglu, Daron, Philippe Aghion, and Fabrizio Zilibotti. Beaman, Lori, and Jeremy Magruder. 2012. “Who Gets the 2006. “Distance to Frontier, Selection, and Economic Job Referral? Evidence from a Social Networks Experi- Growth.” Journal of the European Economic Association 4 (1): ment.” American Economic Review 102 (7): 3574–93. 37–74. Becker, Gary. 1964. Human Capital. New York: Columbia Uni- Acemoglu, Daron, and Alexander Wolitzky. 2011. “The Eco- versity Press. nomics of Labor Coercion.” Econometrica 79 (2): 555–600. Becker, Sashca O., Francesco Cinnirella, and Ludger Woess- Aghion, Philippe. 2009. “Growth and Education.” Working mann. 2013. “Does Women’s Education Affect Fertility? Paper 56, Commission on Growth and Development, Evidence from Pre-demographic Transition Prussia.” World Bank, Washington, DC. European Review of Economic History 17 (1): 24–44. Aghion, Philippe, Leah Boustan, Caroline Hoxby, and Jerome Bedi, A. S., and N. Gaston. 1999. “Using Variation in School- Vandenbussche. 2009. “The Causal Impact of Education ing Availability to Estimate Educational Returns for on Economic Growth: Evidence from U.S.” Working Honduras.” Economics of Education Review 18 (1): 107–16. paper, Harvard University, Cambridge, MA. Belfield, Clive R., Milagros Nores, Steve Barnett, and Law- Aghion, Philippe, and Peter Howitt. 2006. “Joseph Schum- rence Schweinhart. 2006. “The High/Scope Perry Pre- peter Lecture Appropriate Growth Policy: A Unifying school Program Cost-Benefit Analysis Using Data from Framework.” Journal of the European Economic Association the Age-40 Followup.” Journal of Human Resources 41 (1): 4 (2–3): 269–314. 162–90. Algan, Yann, Pierre Cahuc, and Andrei Shleifer. 2013. “Teach- Besley, Timothy J., and Robin S. L. Burgess. 2004. “Can ing Practices and Social Capital.” American Economic Jour- Labour Regulation Hinder Economic Performance? Evi- nal: Applied Economics 5 (3): 189–210. dence from India.” Quarterly Journal of Economics 119 (1): Almeida, Rita, and Pedro Manuel Carneiro. 2005. “Enforce- 91–134. ment of Regulation, Informal Labor, and Firm Perfor- Blau, Peter M., and Otis Dudley Duncan. 1967. The American mance.” IZA Discussion Paper 1759, Institute for the Occupational Structure. New York: John Wiley. Study of Labor, Bonn, Germany. Blimpo, Moussa P., David K. Evans, and Nathalie Lahire. Anderson, D. Mark. 2014. “In School and Out of Trouble? The 2015. “Parental Human Capital and Effective School Minimum Dropout Age and Juvenile Crime.” Review of Management.” Policy Research Working Paper 7238, Economics and Statistics 96 (2): 318–31. World Bank, Washington, DC. Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2012. “What Borgonovi, Francesca, and Tracey Burns. 2015. “The Educa- Did You Do All Day? Maternal Education and Child Out- tional Roots of Trust.” OECD Education Working Paper comes.” Journal of Human Resources 47 (4): 873–912. 119, Organisation for Economic Co-operation and Devel- Angrist, J. D., and A. B. Krueger. 1992. “Estimating the Payoff opment, Paris. to Schooling Using the Vietnam-Era Draft Lottery.” Bosworth, Barry P., and Susan M. Collins. 2003. “The Empir- NBER Working Paper w4067, National Bureau of Eco- ics of Growth: An Update.” Brookings Papers on Economic nomic Research, Cambridge, MA. Activity 2: 113–79. Assaad, Ragui. 1997. “Kinship Ties, Social Networks, and Seg- Botero, Juan Carlos, Simeon Djankov, Rafael La Porta, Flor- mented Labor Markets: Evidence from the Construction encio Lopez de Silanes, and Andrei Shleifer. 2004. “The Sector in Egypt.” Journal of Development Economics 52 (1): Regulation of Labor.” Quarterly Journal of Economics 119 (4): 1–30. 1339–82. Azevedo, João Pedro, Marta Favara, Sarah E. Haddock, Luis F. Botero, Juan Carlos, Alejandro Ponce, and Andrei Shleifer. López-Calva, Miriam Müller, and Elizaveta Perova. 2012. 2013. “Education, Complaints, and Accountability.” Jour- Teenage Pregnancy and Opportunities in Latin America and the nal of Law and Economics 56 (4): 959–96. Caribbean: On Teenage Fertility Decisions, Poverty, and Eco- Brixi, Hana, Ellen Lust, and Michael Woolcock. 2015. Trust, nomic Achievement. Report 83167 v2 rev. Washington, DC: Voice, and Incentives: Learning from Local Success Stories in World Bank. 50 | World Development Report 2018 Service Delivery in the Middle East and North Africa. Wash- Commission on Growth and Development. 2008. The Growth ington, DC: World Bank. Report: Strategies for Sustained Growth and Inclusive Develop- Bteddini, Lida. 2016. “Middle East and North Africa: Public ment. Washington, DC: World Bank. Employment and Governance in MENA.” Report Cullen, Julie Berry, Brian A. Jacob, and Steven Levitt. 2006. ACS18501, World Bank, Washington, DC. “The Effect of School Choice on Participants: Evidence Bursztyn, Leonardo, Thomas Fujiwara, and Amanda Pallais. from Randomized Lotteries.” Econometrica 74 (5): 2017. “  ‘Acting Wife’: Marriage Market Incentives and 1191–1230. Labor Market Investments.” NBER Working Paper 23043, Cutler, David M., and Adriana Lleras-Muney. 2007. “Educa- National Bureau of Economic Research, Cambridge, MA. tion and Health.” Policy Brief 9, National Poverty Center, Cammett, Melani. 2009. “Democracy, Lebanese-Style.” MER: Gerald R. Ford School of Public Policy, University of Middle East Report Online (August 18). http://www.merip Michigan, Ann Arbor. .org/mero/mero081809. Cutler, David M., Adriana Lleras-Muney, and Tom Vogl. Campante, Filipe R., and Davin Chor. 2012. “Schooling, Polit- 2008. “Socioeconomic Status and Health: Dimensions ical Participation, and the Economy.” Review of Economics and Mechanisms.” NBER Working Paper 14333, National and Statistics 94 (4): 841–59. Bureau of Economic Research, Cambridge, MA. Campbell, David E. 2006. “What Is Education’s Impact on Dahl, Robert A. 1998. On Democracy. Yale Nota Bene Series. Civic and Social Engagement?” In Measuring the Effects of New Haven, CT: Yale University Press. Education on Health and Civic Engagement: Proceedings of the Davies, Lynn. 2004. Education and Conflict: Complexity and Copenhagen Symposium, 25–126. Paris: Organisation for Chaos. New York: RoutledgeFalmer. Economic Co-operation and Development. Dee, Thomas S. 2004. “Are There Civic Returns to Educa- Card, D. 1993. “Using Geographic Variation in College tion?” Journal of Public Economics 88 (9–10): 1697–1720. Proximity to Estimate the Return to Schooling.” NBER de Hoyos, Rafael E., Ricardo Estrada, and María José Vargas. Working Paper wp4438, National Bureau of Economic 2017. “Predicting Well-Being through Test Scores.” World Research, Cambridge, MA. Bank, Washington, DC. Carneiro, Pedro, Costas Meghir, and Matthias Parey. 2013. de Hoyos, Rafael E., Carlos Gutiérrez Fierros, and J. Vicente “Maternal Education, Home Environments, and the Vargas M. 2016. “Idle Youth in Mexico: Trapped between Development of Children and Adolescents.” Journal of the the War on Drugs and Economic Crisis.” Policy Research European Economic Association 11 (S1): 123–60. Working Paper 7558, World Bank, Washington, DC. Chetty, Raj, Nathaniel Hendren, and Lawrence F. Katz. 2016. de la Brière, Bénédicte, Deon Filmer, Dena Ringold, Dominic “The Effects of Exposure to Better Neighborhoods on Rohner, Karelle Samuda, and Anastasiya Denisova. 2017. Children: New Evidence from the Moving to Opportunity From Mines and Wells to Well-Built Minds: Turning Sub- Experiment.” American Economic Review 106 (4): 855–902. Saharan Africa’s Natural Resource Wealth into Human Chetty, Raj, Nathaniel Hendren, Patrick Kline, and Emman- Capital. Directions in Development: Human Develop- uel Saez. 2014. “Where Is the Land of Opportunity? The ment Series. Washington, DC: World Bank. Geography of Intergenerational Mobility in the United de Walque, Damien. 2007a. “Does Education Affect Smoking States.” Quarterly Journal of Economics 129 (4): 1553–1623. Behaviors? Evidence Using the Vietnam Draft as an Chiswick, Barry R. 1988. “Differences in Education and Instrument for College Education.” Journal of Health Earnings across Racial and Ethnic Groups: Tastes, Dis- Economics 26 (5): 877–95. crimination, and Investments in Child Quality.” Quarterly —— — —. 2007b. “How Does the Impact of an HIV/AIDS Infor- Journal of Economics 103 (3): 571–97. mation Campaign Vary with Educational Attainment? Chong, Alberto, and Mark Gradstein. 2015. “On Education Evidence from Rural Uganda.” Journal of Development and Democratic Preferences.” Economics and Politics 27 (3): Economics 84 (2): 686–714. 362–88. Dewey, John. 1916. Democracy and Education: An Introduction to Chong, Alberto, Rafael La Porta, Florencia Lopez-de-Silanes, the Philosophy of Education. New York: Macmillan. and Andrei Shleifer. 2014. “Letter Grading Government Díaz, Juan José, Omar Arias, and David Vera Tudela. 2012. Efficiency.” Journal of the European Economic Association “Does Perseverance Pay as Much as Being Smart? The 12 (2): 277–99. Returns to Cognitive and Non-cognitive Skills in Urban Chzhen, Yekaterina. 2013. “Education and Democratisation: Peru.” Working paper, World Bank, Washington, DC. Tolerance of Diversity, Political Engagement, and Under- Djankov, Simeon, Rafael La Porta, Florencio Lopez-de- standing of Democracy.” Background paper, Report 2014/ Silanes, and Andrei Shleifer. 2002. “The Regulation of ED/EFA/MRT/PI/03, United Nations Educational, Scien- Entry.” Quarterly Journal of Economics 117 (1): 1–37. tific, and Cultural Organization, Paris. Duflo, Esther. 2000. “Schooling and Labor Market Conse- Cohen, Daniel, and Marcelo Soto. 2007. “Growth and Human quences of School Construction in Indonesia: Evidence Capital: Good Data, Good Results.” Journal of Economic from an Unusual Policy Experiment.” NBER Working Growth 12 (1): 51–76. Paper wp7860, National Bureau of Economic Research, Collier, Paul, Anke Hoeffler, and Dominic Rohner. 2009. Cambridge, MA. “Beyond Greed and Grievance: Feasibility and Civil War.” Duflo, Esther, Pascaline Dupas, and Michael Kremer. 2014. Oxford Economic Papers 61 (1): 1–27. “Education, HIV, and Early Fertility: Experimental Schooling, learning, and the promise of education | 51 Evidence from Kenya.” NBER Working Paper 20784, World: Evidence from PIAAC.” European Economic Review National Bureau of Economic Research, Cambridge, MA. 73: 103–30. Education Commission. 2016. The Learning Generation: Invest- Hanushek, Eric A., Guido Schwerdt, Ludger Woessmann, ing in Education for a Changing World. New York: Interna- and Lei Zhang. 2017. “General Education, Vocational Edu- tional Commission on Financing Global Education cation, and Labor-Market Outcomes over the Lifecycle.” Opportunity. Journal of Human Resources 52 (1): 48–87. Farrell, Joseph P., and Ernesto Schiefelbein. 1985. “Education Hanushek, Eric A., and Ludger Woessmann. 2008. “The Role and Status Attainment in Chile: A Comparative Chal- of Cognitive Skills in Economic Development.” Journal of lenge to the Wisconsin Model of Status Attainment.” Economic Literature 46 (3): 607–68. Comparative Education Review 29 (4): 490–506. — —— —. 2012. “Do Better Schools Lead to More Growth? Filmer, Deon, and Louise Fox. 2014. Youth Employment in Cognitive Skills, Economic Outcomes, and Causation.” Sub-Saharan Africa. With Karen Brooks, Aparajita Goyal, Journal of Economic Growth 17 (4): 267–321. Taye Mengistae, Patrick Premand, Dena Ringold, Sid- — —— —. 2015. The Knowledge Capital of Nations: Education and dharth Sharma, and Sergiy Zorya. Report ACS8133. Africa the Economics of Growth. CESifo Book Series. Cambridge, Development Forum Series. Washington, DC: Agence MA: MIT Press. Française de Développement and World Bank. http:// Harmon, C., and I. Walker. 1995. “Estimates of the Economic elibrary.worldbank.org/doi/book/10.1596/978-1-4648-0107-5. Return to Schooling for the United Kingdom.” American Foster, Andrew D., and Mark R. Rosenzweig. 1996. “Technical Economic Review 85 (5): 1278–86. Change and Human-Capital Returns and Investments: Heckman, James J., John Eric Humphries, Greg Veramendi, Evidence from the Green Revolution.” American Economic and Sergio S. Urzua. 2014. “Education, Health, and Review 86 (4): 931–53. Wages.” NBER Working Paper 19971, National Bureau of Friedman, Willa, Michael Kremer, Edward Miguel, and Economic Research, Cambridge, MA. Rebecca Thornton. 2011. “Education as Liberation?” Heckman, James J., Jora Stixrud, and Sergio S. Urzua. 2006. NBER Working Paper 16939, National Bureau of Eco- “The Effects of Cognitive and Noncognitive Abilities on nomic Research, Cambridge, MA. Labor Market Outcomes and Social Behavior.” Journal of Ghafar, Adel Abdel. 2016. “Educated but Unemployed: The Labor Economics 24 (3): 411–82. Challenge Facing Egypt’s Youth.” Policy briefing, Brook- Hegewisch, Ariane, and Heidi Hartmann. 2014. “Occupa- ings Doha Center, Doha, Qatar. tional Segregation and the Gender Wage Gap: A Job Half Glewwe, Paul W. 1991. “Schooling, Skills, and the Returns to Done.” Report, Institute for Women’s Policy Research, Government Investment in Education: An Exploration Washington, DC. Using Data from Ghana.” LSMS Working Paper 76, Heineck, Guido, and Silke Anger. 2010. “The Returns to Cog- Living Standards Measurement Study, World Bank, nitive Abilities and Personality Traits in Germany.” Washington, DC. Labour Economics 17 (3): 535–46. ————. 1999. “Why Does Mother’s Schooling Raise Child Hjalmarsson, Randi, Helena Holmlund, and Matthew J. Health in Developing Countries? Evidence from Lindquist. 2015. “The Effect of Education on Criminal Morocco.” Journal of Human Resources 34 (1): 124–59. Convictions and Incarceration: Causal Evidence from Glewwe, Paul W., Eugenie Maiga, and Haochi Zheng. 2014. Micro-Data.” Economic Journal 125 (587): 1290–1326. “The Contribution of Education to Economic Growth: A Jha, Jyotsna, and Fatimah Kelleher. 2006. Boys’ Underachieve- Review of the Evidence, with Special Attention and an ment in Education: An Exploration in Selected Commonwealth Application to Sub-Saharan Africa.” World Development 59: Countries. London: Commonwealth Secretariat. 379–93. Jones, Benjamin F. 2014. “The Human Capital Stock: A Gener- Goldin, Claudia, and Solomon Polachek. 1987. “Residual alized Approach.” American Economic Review 104 (11): Differences by Sex: Perspectives on the Gender Gap in 3752–77. Earnings.” American Economic Review 77 (2): 143–51. Kaffenberger, Michelle, and Lant Pritchett. 2017. “The Impact Granovetter, Mark. 1995. Getting a Job: A Study of Contacts and of Education versus the Impact of Schooling: Schooling, Careers. Chicago: University of Chicago Press. Reading Ability, and Financial Behavior in 10 Countries.” Güneş, Pinar Mine. 2016. “The Impact of Female Education Background paper, World Bank, Washington, DC. on Teenage Fertility: Evidence from Turkey.” B.E. Journal Kao, Kristen. 2012. “Jordan’s Ongoing Election Law Battle.” of Economic Analysis and Policy 16 (1): 259–88. SADA Middle East Analysis, Carnegie Endowment for Haltiwanger, John, Stefano Scarpetta, and Helena Schwei- International Peace, Washington, DC. http://carnegie ger. 2008. “Assessing Job Flows across Countries: The endowment.org/sada/48781. Role of Industry, Firm Size, and Regulations.” NBER Kettunen, Juha. 1997. “Education and Unemployment Dura- Working Paper 13920, National Bureau of Economic tion.” Economics of Education Review 16 (2): 163–70. Research, Cambridge, MA. King, Elizabeth M., Claudio E. Montenegro, and Peter F. Hanson, Gordon H. 2007. “Globalization, Labor Income, and Orazem. 2012. “Economic Freedom, Human Rights, and Poverty in Mexico.” In Globalization and Poverty, edited by the Returns to Human Capital: An Evaluation of the Ann Harrison, 417–56. Chicago: University of Chicago Schultz Hypothesis.” Economic Development and Cultural Press. Change 61 (1): 39–72. Hanushek, Eric A., Guido Schwerdt, Simon Wiederhold, and Klapper, Leora F., Luc Laeven, and Raghuram Rajan. 2004. Ludger Woessmann. 2015. “Returns to Skills around the “Business Environment and Firm Entry: Evidence from 52 | World Development Report 2018 International Data.” NBER Working Paper 10380, McNabb, Robert, and George Psacharopoulos. 1981. “Racial National Bureau of Economic Research, Cambridge, MA. Earnings Differentials in the UK.” Oxford Economic Papers Krueger, Alan B., and Mikael Lindahl. 2001. “Education for 33 (3): 413–25. Growth: Why and for Whom?” Journal of Economic Litera- Micco, Alejandro, and Carmen Pagés. 2007. “The Economic ture 39 (4): 1101–36. Effects of Employment Protection: Evidence from Inter- Lam, David, Guilherme Sedlacek, and Suzanne Duryea. national Industry-Level Data.” Working Paper 592, 2016. “Increases in Women’s Education and Fertility Research Department, Inter-American Development Decline in Brazil.” Anais do VIII Encontro Nacional de Estu- Bank, Washington, DC. dos Populacionais, Vol. 1, 89–118. Belo Horizonte, Brazil: Milligan, Kevin, Enrico Moretti, and Philip Oreopolous. Associação Brasileira de Estudos Populacionais. 2004. “Does Education Improve Citizenship? Evidence Lanjouw, Peter F., and Martin Ravallion. 1999. “Benefit from the United States and the United Kingdom.” Journal Incidence, Public Spending Reforms, and the Timing of of Public Economics 88 (9): 1667–95. Program Capture.” World Bank Economic Review 15 (2): Mincer, Jacob. 1991. “Education and Unemployment.” NBER 257–73. Working Paper 3838, National Bureau of Economic Larreguy, Horacio A., and John Marshall. 2017. “The Effect of Research, Cambridge, MA. Education on Civic and Political Engagement in Moll, Peter G. 1998. “Primary Schooling, Cognitive Skills, Non-consolidated Democracies: Evidence from Nigeria.” and Wages in South Africa.” Economica 65 (258): 263–84. Review of Economics and Statistics. http://www.mitpress Montenegro, Claudio E., and Harry Anthony Patrinos. 2017. journals.org/doi/abs/10.1162/REST_a_00633. “Comparable Estimates of Returns to Schooling around Lavy, Victor, and Alexander Zablotsky. 2011. “Mother’s the World.” Policy Research Working Paper 7020, World Schooling and Fertility under Low Female Labor Force Bank, Washington, DC. Participation: Evidence from a Natural Experiment.” Montez, Jennifer Karas, Robert A. Hummer, and Mark D. NBER Working Paper 16856, National Bureau of Eco- Hayward. 2012. “Educational Attainment and Adult nomic Research, Cambridge, MA. Mortality in the United States: A Systematic Analysis of Layard, Richard, and George Psacharopoulos. 1974. “The Functional Form.” Demography 49 (1): 315–36. Screening Hypothesis and the Returns to Education.” Mueller, Gerrit, and Erik Plug. 2006. “Estimating the Effect Journal of Political Economy 82 (5): 985–98. of Personality on Male and Female Earnings.” ILR Review Lipset, Seymour Martin. 1959. “Some Social Requisites of 60 (1): 3–22. Democracy: Economic Development and Political Legiti- Mullis, I. V. S., M. O. Martin, P. Foy, and M. Hooper. 2016. macy.” American Political Science Review 53 (01): 69–105. “TIMSS 2015 International Results in Mathematics.” —— ——. 1960. Political Man: The Social Basis of Modern Politics. TIMSS and PIRLS International Study Center, Boston Garden City, NY: Doubleday. College, Chestnut Hill, MA. http://timssandpirls.bc.edu Lochner, Lance. 2004. “Education, Work, and Crime: A /timss2015/international-results/. Human Capital Approach.” International Economic Review Munshi, Kaivan. 2003. “Networks in the Modern Economy: 45 (3): 811–43. Mexican Migrants in the US Labor Market.” Quarterly Lochner, Lance, and Enrico Moretti. 2004. “The Effect of Journal of Economics 118 (2): 549–99. Education on Crime: Evidence from Prison Inmates, Nee, Victor, and Rebecca Matthews. 1996. “Market Transi- Arrests, and Self-Reports.” American Economic Review 94 tion and Societal Transformation in Reforming State (1): 155–89. Socialism.” Annual Review of Sociology 22 (August): 401–35. Lust-Okar, Ellen. 2009. “Legislative Elections in Hegemonic Nelson, Richard R., and Edmund S. Phelps. 1966. “Investment Authoritarian Regimes: Competitive Clientelism and in Humans, Technological Diffusion, and Economic Resistance to Democratization.” Democratization by Elec- Growth.” American Economic Review 56 (1/2): 69–75. tions: A New Mode of Transition, edited by Staffan I. Lind- OECD (Organisation for Economic Co-operation and Devel- berg, 226–45. Baltimore: Johns Hopkins University Press. opment). 2010. “The High Cost of Low Educational Per- Machin, Stephen, Olivier Marie, and Sunčica Vujić. 2011. formance: The Long-Run Economic Impact of Improving “The Crime Reducing Effect of Education.” Economic Jour- PISA Outcomes.” Programme for International Student nal 121 (552): 463–84. Assessment, OECD, Paris. Mackenbach, Johan P. 2006. “Health Inequalities: Europe in Oreopoulos, Philip. 2007. “Do Dropouts Drop Out Too Soon? Profile.” Department of Health, London. Wealth, Health, and Happiness from Compulsory Madsen, Jakob B. 2014. “Human Capital and the World Tech- Schooling.” Journal of Public Economics 91 (11): 2213–29. nology Frontier.” Review of Economics and Statistics 96 (4): Osili, Una Okonkwo, and Bridget Terry Long. 2008. “Does 676–92. Female Schooling Reduce Fertility? Evidence from Maluccio, J. 1998. “Endogeneity of Schooling in the Wage Nigeria.” Journal of Development Economics 87 (1): 57–75. Function: Evidence from the Rural Philippines.” Food Oye, Mari, Lant Pritchett, and Justin Sandefur. 2016. “Girls’ Consumption and Nutrition Division Discussion Paper Schooling Is Good, Girls’ Schooling with Learning Is 54, International Food Policy Research Institute, Wash- Better.” Education Commission, Center for Global Devel- ington, DC. opment, Washington, DC. Mankiw, N. Gregory, David Romer, and David N. Weil. 1992. Petrin, Amil, and Jagadeesh Sivadasan. 2006. “Job Security “A Contribution to the Empirics of Economic Growth.” Does Affect Economic Efficiency: Theory, a New Statis- Quarterly Journal of Economics 107 (2): 407–37. tic, and Evidence from Chile.” NBER Working Paper Schooling, learning, and the promise of education | 53 12757, National Bureau of Economic Research, Cam- UNESCO (United Nations Educational, Scientific, and Cul- bridge, MA. tural Organization). 2016. Global Education Monitoring Pinker, Steven. 2011. The Better Angels of Our Nature: The Decline Report 2016, Education for People and Planet: Creating Sustain- of Violence in History and Its Causes. New York: Viking. able Futures for All. Paris: UNESCO. http://unesdoc.unesco Pritchett, Lant. 2001. “Where Has All the Education Gone?” .org/images/0024/002457/245752e.pdf. World Bank Economic Review 15 (3): 367–91. United Nations. 1948. “Universal Declaration of Human Riddell, Craig W., and Xueda Song. 2011. “The Impact Rights.” Document A/RES/3/217 A, New York. of Education on Unemployment Incidence and Re- Valerio, Alexandria, María Laura Sánchez Puerta, Namrata Employment Success: Evidence from the US Labour Raman Tognatta, and Sebastián Monroy-Taborda. 2016. Market.” Labour Economics 18 (4): 453–63. “Are There Skills Payoffs in Low- and Middle-Income Romer, Paul Michael. 1990. “Endogenous Technological Countries? Empirical Evidence Using Step Data.” Policy Change.” Journal of Political Economy 98 (5): 71–102. Research Working Paper 7879, World Bank, Washington, Sakai, Keiko, Faleh Abdul Jabar, and Hosham Dawod. 2001. DC. “Tribalism and the State: Remarks on the Army, Cabinets, Vogl, Tom S. 2012. “Education and Health in Developing and the National Assembly, 1980–1990.” In Tribes and Economies.” Working Paper 1453, Research Program in Power: Nationalism and Ethnicity in the Middle East, edited Development Studies, Woodrow Wilson School of Public by Faleh Abdul Jabar and Hosham Dawod, 136–64. and International Affairs, Princeton University, Prince- London: Saqi Books. ton, NJ. Schultz, Theodore W. 1975. “The Value of the Ability to Deal Wantchekon, Leonard, Marko Klasnja, and Natalija Novta. with Disequilibria.” Journal of Economic Literature 13 (3): 2015. “Education and Human Capital Externalities: Evi- 827–46. dence from Colonial Benin.” Quarterly Journal of Economics Sen, Amartya. 1985. Commodities and Capabilities. Amsterdam: 130 (2): 703–57. North-Holland. Welch, Finis. 1970. “Education in Production.” Journal of ————. 1999. Development as Freedom. New York: Oxford Uni- Political Economy 78 (1): 35–59. versity Press. World Bank. 1993. The East Asian Miracle: Economic Growth and —— — —. 2004. “Capabilities, Lists, and Public Reason: Continu- Public Policy. Policy Research Report Series. Washington, ing the Conversation.” Feminist Economics 10 (3): 77–80. DC: World Bank; New York: Oxford University Press. Solon, Gary. 1999. “Intergenerational Mobility in the Labor —— ——. 2011. World Development Report 2012: Gender Equality Market.” In Handbook of Labor Economics, Vol. 3A, edited and Development. Washington, DC: World Bank. by Orley Ashenfelter and David Card, 1761–800. Hand- —— ——. 2012. World Development Report 2013: Jobs. Washington, books in Economics Series. Amsterdam: Elsevier. DC: World Bank. Solow, Robert M. 1956. “A Contribution to the Theory of ————. 2017. World Development Indicators (database). Economic Growth.” Quarterly Journal of Economics 70 (1): World Bank, Washington, DC. http://data.worldbank.org 65–94. /data-catalog/world-development-indicators. Sondheimer, Rachel Milstein, and Donald P. Green. 2010. World Values Survey Association. 2015. World Values Survey “Using Experiments to Estimate the Effects of Education Wave 6 2010-2014 Official Aggregate V.20150418. King’s on Voter Turnout.” American Journal of Political Science College, Old Aberdeen, U.K. http://www.worldvalues 54 (1): 174–89. survey.org/WVSDocumentationWV6.jsp. Strudwick, Jeremy, and Philip Foster. 1991. “Origins and Des- Younger, Stephen D. 2003. “Benefits on the Margin: Obser­ tinations in Jamaica.” International Journal of Educational vations on Marginal Benefit Incidence.” World Bank Development 11 (2): 149–59. Economic Review 17 (1): 89–106. Thomas, Duncan, John Strauss, and Maria-Helena Hen- Zhang, Junsen, Yaohui Zhao, Albert Park, and Xiaoqing Song. riques. 1990. “Child Survival, Height for Age, and House- 2005. “Economic Returns to Schooling in Urban China, hold Characteristics in Brazil.” Journal of Development 1988 to 2001.” Journal of Comparative Economics 33 (4): Economics 33 (2): 197–234. 730–52. 54 | World Development Report 2018 PART II The learning crisis 2 The great schooling expansion— and those it has left behind 3 The many faces of the learning crisis 4 To take learning seriously, start by measuring it 2 The great schooling expansion—and those it has left behind  n 1945, when Indonesia declared independence, only 5 percent of its people could I read and write. In 2015, 95 percent could (UIS 2016). In Nepal in 1981, only one in five adults were literate. In 2015, nearly two-thirds were (UIS 2016). Over the last 50 years, schooling has expanded dra- The recent expansion in schooling in low- matically in most low- and middle-income countries. income countries is especially remarkable in its In some countries, this expansion has been at his- scope and speed. The years of schooling completed torically unprecedented rates. Another pattern is the by the average adult in the developing world more rapid expansion of postprimary education, though than tripled between 1950 and 2010—from 2.0 to many young people remain excluded from even 7.2 years.4 This rate is historically unprecedented. In primary education. So even in countries with strong Zambia, secondary enrollment increased by nearly schooling expansions, exclusions due to poverty, gen- 75 percentage points between 2000 and 2010, faster der, ethnicity, disability, and location persist. Fragile than the rate experienced by any high-income and postconflict countries also remain glaring excep- country during its fastest phase of secondary tions to the global boom in schooling. expansion.5 It took the United States 40 years—from 1870 to 1910—to increase girls’ enrollments from 57 Most children have access to percent to 88 percent. Morocco achieved a similar increase in just 11 years.6 Accordingly, the enrollment basic education gaps between low- and high-income countries are Schooling has expanded almost universally. In 1970 closing. By 2008 the average low-income country the gross primary enrollment rate was 68 percent in was enrolling students in primary school at nearly Sub-Saharan Africa and 47 percent in South Asia. By the same rate as the average high-income country 2010, that rate was above 100 percent in both regions.1 (figure 2.1). Despite these gains, there is a large stock These numbers reflect the progress made in nearly all of uneducated adults—322 million in South Asia countries regardless of regime type, rate of economic alone (figure 2.2). growth, or quality of governance.2 As a result, most Previously marginalized groups, especially girls, children today enroll in primary school—and every are now much more likely to start primary school. new cohort of young people spends more time in Between 2000 and 2014, the number of out-of-school school than previous ones.3 children fell by about 112 million.7 At the same time, 58 | World Development Report 2018 Figure 2.1 School enrollments have shot up in developing countries Net enrollment rates, by country group (1820–2010) a. Primary school b. Secondary school 100 100 80 80 Net enrollment rate (%) Net enrollment rate (%) 60 60 40 40 20 20 0 0 25 50 75 50 75 20 0 10 25 50 75 50 75 20 0 10 00 25 00 25 0 0 18 18 18 19 19 20 18 18 18 19 19 20 19 19 19 19 East Asia and Pacific Latin America and the Caribbean South Asia Europe and Central Asia Middle East and North Africa Sub-Saharan Africa High-income countries Source: WDR 2018 team, using data from Lee and Lee (2016). Data at http://bit.do/WDR2018-Fig_2-1. the share of girls enrolled in basic education reached Figure 2.2 Most of the world’s population with less a historic high.8 In primary and secondary schools than a primary education is in South Asia, but rates in the developing world, the ratio of girls to boys are similar in Sub-Saharan Africa jumped from 0.84 to 0.96 between 1991 and 2007.9 Stock of educational attainment (ages 15–64), by country group (2010) Indeed, girls outnumber boys in secondary school Percent of world’s population in 38 developing countries (out of 121 for which data 0 20 40 60 80 100 are available).10 Gender parity, however, has yet to be 100 achieved; 62 million girls between the ages of 6 and 15 years are still out of school,11 with the highest con- group’s population 75 Percent of country centrations in West and South Asia and Sub-Saharan Africa.12 Although many girls start primary school, their likelihood of completing it remains low in some 50 countries. By 2014 the primary enrollment rate of girls in low-income countries was at 78 percent, but 25 their completion rate was only 63 percent.13 The strongest schooling expansions have occurred 0 at the primary level, leading to a sharp increase in East Asia and Pacific South Asia High-income countries the demand for secondary education. Secondary Europe and Middle East Sub-Saharan enrollment rates have risen above 50 percent in every Central Asia and North Africa Africa region except parts of Sub-Saharan Africa. But at that Latin America and the Caribbean level there remain big gaps between low- and high-in- come countries, especially for completion. In 2016 the Highest level of education attained: secondary completion rate was 96 percent in high- Postsecondary Secondary Primary Less than primary income Organisation for Economic Cooperation and Source: WDR 2018 team, using data from Lee and Lee (2016). Data at http://bit.do/WDR2018-Fig_2-2. The great schooling expansion—and those it has left behind | 59 Figure 2.3 National income is Poverty, gender, ethnicity, correlated with the gap between primary and lower secondary disability, and location completion rates explain most remaining Distribution of completion rates across countries, by country schooling disparities group and level of schooling Mawut, a refugee from South Sudan, lived in a refugee Percent camp in Kenya for years so he could finish primary 0 20 40 60 80 100 school. “In 2010 my sister [decided to return] to Sudan and I decided not to follow her because I knew if I High-income (OECD) did so, that would be the end of my education. I went High-income (non-OECD) to Kakuma Refugee Camp with my brother, where I completed the two remaining years of my primary Upper-middle-income course. I passed very well despite the problems that Lower-middle-income I had experienced” (Kelland 2016). Low-income Nadya is a 25-year-old from Mashkhail, a remote district in southeastern Afghanistan. “I am the only literate woman in this heavily populated province Europe and Central Asia working outside home and, more important, the only East Asia and Pacific female teacher in this traditional province. . . . When the elders of Mashkhail found out I was literate . . . South Asia [they asked] my husband if I could volunteer teaching Latin America their daughters, mostly those older girls who were not and the Caribbean allowed to be taught by a male teacher” (IRIN 2003). Middle East and North Africa Conflict-affected countries remain a glaring excep­tion Sub-Saharan Africa to the global schooling expansion (box 2.1). The net primary enrollment rate in conflict-affected South Sudan was 41 percent in 2011; the enrollment rate in World neighboring Ethiopia was 78 percent.15 Conflict-af- fected countries are home to more than a third of out- 25th 50th 75th Percentile of-school children.16 Children in these countries are Primary school less likely to complete school—30 percent less likely Lower secondary school for primary, 50 percent less likely for lower second- ary.17 They have higher dropout rates, lower comple- Source: WDR 2018 team, using data from UIS (2016). Data at http://bit.do /WDR2018-Fig_2-3. tion rates, higher gender disparities, lower literacy Note: Geographic regions exclude high-income countries. The data levels, and disproportionately high out-of-school presented are the latest available, by country, for the period 2010–16. numbers.18 Conflict can also erase past gains. The Completion rates include students whose age exceeds the official age group for a particular education level, and so the rate may exceed 100 percent. Syrian Arab Republic had achieved universal primary enrollment in 2000. But in 2013, 1.8 million children Development (OECD) countries, but only 35 percent were out of school due to conflict.19 in low-income countries (figure 2.3).14 Exclusions based on poverty, location, gender, Developing countries are following a very dif- and ethnicity persist (figure 2.5). In 2014 an estimated ferent path to schooling expansion than developed 61 million primary school-age children and 202 mil- countries did. Low-income countries are leapfrogging lion secondary school-age youth—with a dispropor- the progress experienced by high-income countries— tionate share from poor households—were out of with strong expansion in postprimary education school.20 Only about a quarter of the poorest children even while primary education remains inaccessible in low-income countries—compared with three- to many young people (figure 2.4). quarters in the richest—complete primary school.21 60 | World Development Report 2018 These gaps are even larger when disaggregating by Figure 2.4 Lower-income countries are rapidly gender, where the double exclusions from gender and expanding secondary education at a time when poverty mean that only 25 percent of the poorest girls much of their population has not yet completed in low-income countries complete primary school.22 primary school In some contexts, ethnicity can be an important pre- Evolution in the stock of educational attainment (ages 15–64), by present-day dictor of education access. In 2011 only 10 percent of income group (1890–2010) adult Roma had completed secondary education in 100 0 Romania, compared with 58 percent of non-Roma 1890 1890 1890 living nearby.23 Children of indigenous groups in Latin America are more than twice as likely as other 1920 1950 children to be working.24 80 20 1950 Children from the poorest families are less likely to start school. Those who do start school are more likely 1980 60 40 to drop out early, though at varying rates across coun- 1980 1950 tries. In some countries, such as Mali and Pakistan, the Less than primary (%) Primary (%) 1890 effect of poverty on education levels is already visible 40 2010 60 at the start of primary school. In Indonesia and Peru, gaps emerge later (figure 2.6). In nearly every country, parents’ wealth and education attainment are the main 2010 20 1920 80 determinants of their children’s education.25 On aver- 1980 age, in developing countries there is a 32 percentage 1950 point gap between the chances of children in the poor- 2010 0 2010 1980 100 est and richest quintiles completing primary school— 100 80 60 40 20 0 with these wealth-related inequalities increasing in 10 Secondary or above (%) of 25 such countries for which data are available.26 The poorest people are the most affected by any marginal Low-income Upper-middle-income Lower-middle-income High-income increase in or contraction of public spending on edu- cation.27 Not surprisingly, then, making school more Source: WDR 2018 team, using data from Lee and Lee (2016). Data at http://bit.do/WDR2018-Fig_2-4. Box 2.1 Access denied: The effects of fragility, conflict, and violence Children living in the most fragile contexts make up about 20 textbooks. Other manifestations include the singular use of percent of the world’s primary school-age population. Yet a nonindigenous language as part of noninclusive “nation they constitute about 50 percent of those not in school, an building,” the denial of education to marginalized groups, increase from 42 percent in 2008.a Children in fragile states the manipulation of history for political purposes, and the are up to three times more likely to be out of school than use of geography lessons to promote a particular ideolog- those living in nonconflict contexts, and they are far more ical view. likely to drop out of primary school before completion. Even Displaced children face significant obstacles to learning. when fragility, conflict, and violence do not directly disrupt Only one out of every two refugee children has access to access, they can affect learning by changing the pedagogical primary education; a refugee child is five times more likely experience, such as through lack of teachers and resources than the average child to be out of school.b Education for or trauma from violence. Conflict tends to exacerbate exclu- these vulnerable children can provide a sense of normalcy sions based on ethnicity, religion, or gender. and structure, with high returns.c But the challenge of Education systems can exacerbate conflict through, equipping these children with the necessary skills and for example, ethnic, religious, or gender stereotyping in knowledge has often fallen to host governments, some (Box continues next page) The great schooling expansion—and those it has left behind | 61 Box 2.1 Access denied: The effects of fragility, conflict, and violence (continued) of which are already struggling to provide quality educa- cities in one large country, for example, more than half of tion for their own populations. Lebanon, for example, has all students had been subjected to some form of violent increased the size of its public education system by almost punishment at school.e One-quarter of the children who 50 percent since 2011, largely because of the conflict in the were physically punished said they sustained injuries as Syrian Arab Republic. In Lebanon, refugees make up almost a result. Children already discriminated against based on a third of the total enrollment in education.d disability, poverty, caste, class, ethnicity, or sexual orien- In addition to conflict and violence at the societal tation are more likely than their peers to suffer corporal level, school-level violence hinders learning. Physical and punishment. In some contexts, sexual violence in schools psychological violence are common forms of so-called dis- is also an issue—for example, authority figures may abuse cipline, with students in many parts of the world routinely their power by demanding sex in return for better grades or subjected to corporal punishment. Across three major for waiving school fees. Source: Commins (2017). a. UNESCO (2013). b. UNHCR (2016). c. Burde and others (2015). d. World Bank (2016a). e. NCCM and UNICEF (2015). affordable—both in terms of defraying direct costs and track to complete secondary school.30 Even in regions compensating for opportunity costs—increases the where gender parity has been achieved at the primary school participation of children from poorer families.28 level, such as North Africa and West Asia, gender Globally, girls are twice as likely as boys never to disparities in enrollment exist at the lower secondary start school, which results in lower school completion school level and become more pronounced in upper rates.29 In Sub-Saharan Africa, poor rural girls are secondary school.31 seven times less likely than nonpoor urban boys to Gender reinforces other disadvantages. It often complete school; less than 1 in 20 of these girls is on compounds disadvantages related to socioeconomic Figure 2.5 School completion is higher for richer and urban families, but gender gaps are more context-dependent Gaps in grade 6 completion rates (percent) for 15- to 19-year-olds, by wealth, location, and gender a. Richest–poorest b. Urban–rural c. Male–female 80 80 80 Percentage point gap between Percentage point gap between Percentage point gap between richest and poorest quintiles 60 60 60 male and female urban and rural 40 40 40 20 20 20 0 0 0 −20 −20 −20 20 40 60 80 100 20 40 60 80 100 20 40 60 80 100 Overall grade 6 completion rate (%) Overall grade 6 completion rate (%) Overall grade 6 completion rate (%) Source: WDR 2018 team, using data from Filmer (2016). Data at http://bit.do/WDR2018-Fig_2-5. Note: The data presented are the latest available by country, 2005–14. Each vertical line indicates the size and direction of the gap for a country. 62 | World Development Report 2018 Figure 2.6 Multiple exclusions: Girls from poor households often have the lowest rates of education attainment Percentage of youth (ages 15–19) who have completed each grade, by wealth quintile and gender, selected countries (2012) a. Pakistan b. Mali c. Peru d. Indonesia 100 100 100 100 80 80 80 80 60 60 60 60 Percent Percent Percent Percent 40 40 40 40 20 20 20 20 0 0 0 0 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Grade Grade Grade Grade Richest quintile Middle quintile Poorest quintile Girls in the poorest quintile Source: WDR 2018 team, using data from U.S. Agency for International Development’s Demographic and Health Surveys for 2012 (http://www.dhsprogram.com). Data at http://bit.do/WDR2018-Fig_2-6. status, ethnicity, location, religion, sexual orientation, outside it—determines their children’s enrollment, disability, age, and race.32 Across 44 countries, boys grade completion, and learning outcomes.39 In some in the poorer half of the population were almost 75 contexts, this calculus might involve sending just percent more likely to complete grade 5 than girls; some—but not all—children to school. For example, by contrast, in the richer half of the population the only about a quarter of rural households in Burkina boys’ advantage was less than 20 percent.33 Nearly 70 Faso enroll all their children in school.40 Cutting percent of all girls who were not enrolled in primary the cost of schooling, therefore, significantly raises school in 2006 came from socially excluded groups.34 school participation by children from poorer fami- Children with disabilities face substantial obsta- lies.41 The removal of direct costs to schooling through cles to education—and substantially lower participa- universal primary education in Uganda increased tion in school.35 In Burkina Faso, having a disability primary enrollment by over 60 percent and lowered increases the probability of a child never attending cost-related dropouts by over 33 percentage points.42 school by about two times.36 Even in countries with In Malawi, free primary education increased enroll- high overall primary school enrollments, children ment by half, favoring girls and poor people.43 with disabilities are still significantly less likely to For some poor households, distance to the nearest attend school. In Moldova, 97 percent of children school is a predictor of school participation, espe- without disabilities between the ages of 7 and 15 years cially where social norms or safety concerns make it are enrolled in primary school, whereas only 58 per- difficult for children—particularly girls—to travel far cent of children with disabilities are in school.37 At the from home.44 In Indonesia, each school built for every same time, quality education for children with dis- 1,000 children increased education by an average abilities has significant economic and social returns. of 0.12 years.45 But school availability matters most Across 12 developing countries, each additional year when starting from a point of low availability, and of schooling for people with a disability decreased school construction by itself can only do so much.46 their probability of being in the poorest two quintiles Perceived returns, whether in the labor market by between 2 and 5 percentage points.38 or in realms such as the marriage “market,” often determine how willing poor parents are to send their For poor parents, schooling children to school.47 Thus the demand for education is likely to be lower if parents underestimate the requires trade-offs returns to education.48 Parents might also misunder- Millions of poor parents make difficult choices about stand how the returns to education vary by level. If whether to educate their children. This cost-benefit they believe the returns from secondary education assessment—where costs include both the direct cost are significantly higher than the returns from pri- of school and the opportunity cost of a child’s time mary, it might make more sense to focus on sending The great schooling expansion—and those it has left behind | 63 their brightest child to secondary school rather than families in the Arab Republic of Egypt could differ- sending all their children to primary school.49 In the entiate schools by their quality of education—and face of extreme poverty and perceived low returns students attending lower-quality schools were more to schooling, poor people might restrict their overall likely to drop out.56 Indeed, parents seem willing aspirations for education.50 to bypass lower-quality public schools in favor of Parents’ perceptions about whether their children higher-quality, more remote public ones, or better- are learning affect their decisions about whether to quality private ones with higher fees.57 continue schooling. In most low-income countries, students who have to repeat grades or who exceed *** the average age of their classmates by several years Worldwide, parents and students have incredible are more likely to drop out before completing primary faith in the power of education. People everywhere school.51 In the Philippines, a child’s perceived educa- know that education can transform lives, including tional ability is a key determinant of whether parents the lives of their children. The rapid growth in school- choose to keep him or her in school, or in the work- ing in poor rural areas around the world is indicative place instead.52 Similarly, adolescents in Burkina Faso of this demand.58 Almost all parents say they want are far more likely to be enrolled when they score their children to complete school—even parents who high on an intelligence test, but much less likely if did not attend school themselves.59 The last 50 years their sibling scores higher.53 have generated high hopes of strong returns to educa- When parents perceive the education available to tion and a great drive toward universal school enroll- be of low quality, it also affects their choices about ment. But much remains to be done. Achieving this schooling.54 Although parental perceptions of school promise means addressing gaps in school participa- quality depend on a variety of factors—from the phys- tion and ensuring that education leads to learning. As ical condition of the school to teacher punctuality— chapter 3 shows, the great schooling expansion has student learning outcomes are a critical aspect.55 not translated into commensurate gains in learning. Holding student ability and achievement constant, Attention must now shift to ensuring learning for all. Notes 1. UIS (2016). Gross enrollment includes students whose 3. World Bank (2014). 2 age exceeds the official age group for a particular educa- 24. López-Calva and Patrinos (2015). tion level, and so the rate may exceed 100 percent. 25. Alderman, Orazem, and Paterno (2001); Bailey and 2. Pritchett (2013). Dynarski (2011); Lincove (2015). 3. United Nations (2015). 26. Education Commission (2016). 4. Barro and Lee (2013). 27. Lanjouw and Ravallion (1999). 5. WDR 2018 team calculations using data from Lee and 28. Kremer and Holla (2009); Orazem and King (2008). Lee (2016). 29. Education Commission (2016). 6. World Bank (2011b). 30. Education Commission (2016). 7. UNESCO (2016). 31. UNESCO (2016). 8. UNESCO (2015). 32. Kabeer (2015); Lewis and Lockheed (2006). 9. World Bank (2011a). 33. Filmer (2005). 10. World Bank (2017). Data used are the latest available 34. Lockheed (2010). between 2010 and 2015. 35. World Bank (2007). 11. World Bank (2016b). 36. Kobiané and Bougma (2009). 12. UNESCO (2015). 37. Mete (2008). 13. UIS (2016). 38. Filmer (2008). 14. UIS (2016). 39. Becker (2009); Glewwe (2002); Hanushek and Woess- 15. World Bank (2016c). mann (2008). 16. UNESCO (2016). 40. Akresh and others (2012). 17. Education Commission (2016). 41. Kremer and Holla (2009); Orazem and King (2008). 18. UNESCO (2011). 42. Deininger (2003). 19. UIS and UNESCO (2015). 43. Bentaouet-Kattan and Burnett (2004). 20. UNESCO (2016). 44. Burde and Linden (2012). 21. UIS (2016). 45. Duflo (2001). 22. UIS (2016). 46. Filmer (2007). 64 | World Development Report 2018 47. Behrman, Rosenzweig, and Taubman (1994); Jensen the Family and in the Marriage Market: The Twins (2010); Nguyen (2008). Experiment.” Journal of Political Economy 102 (6): 1131–74. 48. Banerjee and Duflo (2011); Murnane and Ganimian Bentaouet-Kattan, Raja, and Nicholas Burnett. 2004. “User (2014). Fees in Primary Education.” Education for All Working 49. Banerjee and Duflo (2011). Paper, World Bank, Washington, DC. 50. Dalton, Ghosal, and Mani (2016); Genicot and Ray Burde, Dana, Ozen Guven, Jo Kelcey, Heddy Lahmann, and (2014). Khaled Al-Abbadi. 2015. “What Works to Promote Chil- 51. Glick and Sahn (2010); UNESCO and UNICEF (2015). dren’s Educational Access, Quality of Learning, and Well- 52. Bacolod and Ranjan (2008). Being in Crisis-Affected Contexts.” Education Rigorous 53. Akresh and others (2012). Literature Review, U.K. Department for International 54. Banerjee, Jacob, and Kremer (2000); Rivkin, Hanushek, Development, London. and Kain (2005). Burde, Dana, and Leigh L. Linden. 2012. “The Effect of 55. Alderman, Orazem, and Paterno (2001); Andrabi, Das, Village-Based Schools: Evidence from a Randomized and Khwaja (2008); Tooley and Dixon (2007). Controlled Trial in Afghanistan.” NBER Working Paper 56. Hanushek, Lavy, and Hitomi (2006). 18039, National Bureau of Economic Research, Cam- 57. Andrabi, Das, and Khwaja (2008); He and Giuliano bridge, MA. (2017). Commins, Stephen. 2017. “Fragility, Conflict and Violence.” 58. Tooley and Dixon (2006). Background paper, WDR 2018, World Bank, Washington, 59. Mukerji and Walton (2013). DC. Dalton, Patricio S., Sayantan Ghosal, and Anandi Mani. 2016. “Poverty and Aspirations Failure.” Economic Journal 126 References (590): 165–88. Akresh, Richard, Emilie Bagby, Damien de Walque, and Deininger, Klaus. 2003. “Does Cost of Schooling Affect Harounan Kazianga. 2012. “Child Ability and Household Enrollment by the Poor? Universal Primary Education in Human Capital Investment Decisions in Burkina Faso.” Uganda.” Economics of Education Review 22 (3): 291–305. Economic Development and Cultural Change 61 (1): 157–86. Duflo, Esther. 2001. “Schooling and Labor Market Conse- Alderman, Harold, Peter F. Orazem, and Elizabeth M. Paterno. quences of School Construction in Indonesia: Evidence 2001. “School Quality, School Cost, and the Public/Private from an Unusual Policy Experiment.” American Economic School Choices of Low-Income Households in Pakistan.” Review 91 (4): 795–813. Journal of Human Resources 36 (2): 304–26. Education Commission. 2016. The Learning Generation: Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2008. Investing in Education for a Changing World. New York: “A Dime a Day: The Possibilities and Limits of Private International Commission on Financing Global Educa- Schooling in Pakistan.” Comparative Education Review tion Opportunity. 52 (3): 329–55. Filmer, Deon. 2005. “Gender and Wealth Disparities in Bacolod, Marigee P., and Priya Ranjan. 2008. “Why Children Schooling: Evidence from 44 Countries.” International Work, Attend School, or Stay Idle: The Roles of Ability Journal of Education Research 43 (6): 351–69. and Household Wealth.” Economic Development and Cul- —— — —. 2007. “If You Build It, Will They Come? School tural Change 56 (4): 791–828. Availability and School Enrolment in 21 Poor Countries.” Bailey, Martha J., and Susan M. Dynarski. 2011. “Inequality Journal of Development Studies 43 (5): 901–28. in Postsecondary Education.” In Whither Opportunity? —— — —. 2008. “Disability, Poverty, and Schooling in Develop- Rising Inequality, Schools, and Children’s Life Chances, edited ing Countries: Results from 14 Household Surveys.” by Greg J. Duncan and Richard J. Murnane, 117–32. World Bank Economic Review 22 (1): 141–63. Chicago: Spencer Foundation; New York: Russell Sage —— — —. 2016. “Education Attainment and Enrollment around Foundation. the World: An International Database.” World Bank, Banerjee, Abhijit Vinayak, and Esther Duflo. 2011. Poor Washington, DC. http://go.worldbank.org/3GEREWJ0E0. Economics: A Radical Rethinking of the Way to Fight Global Genicot, Garance, and Debraj Ray. 2014. “Aspirations and Poverty. Philadelphia: Public Affairs. Inequality.” NBER Working Paper 19976, National Bureau Banerjee, Abhijit Vinayak, Suraj Jacob, and Michael of Economic Research, Cambridge, MA. Kremer. 2000. “Promoting School Participation in Rural Glewwe, Paul W. 2002. “Schools and Skills in Developing Rajasthan: Results from Some Prospective Trials.” Countries: Education Policies and Socioeconomic Out- With Jenny Lanjouw and Peter F. Lanjouw. Working comes.” Journal of Economic Literature 40 (2): 436–82. paper, Massachusetts Institute of Technology, Cam- Glick, Peter, and David E. Sahn. 2010. “Early Academic Perfor- bridge, MA. mance, Grade Repetition, and School Attainment in Barro, Robert J., and Jong-Wha Lee. 2013. “A New Data Set of Senegal: A Panel Data Analysis.” World Bank Economic Educational Attainment in the World, 1950–2010.” Journal Review 24 (1): 93–120. of Development Economics 104: 184–98. Hanushek, Eric A., Victor Lavy, and Kohtaro Hitomi. 2006. Becker, Gary. 2009. A Treatise on the Family (Enlarged Edition). “Do Students Care about School Quality? Determinants Cambridge, MA: Harvard University Press. of Dropout Behavior in Developing Countries.” NBER Behrman, Jere R., Mark R. Rosenzweig, and Paul Taubman. Working Paper 12737, National Bureau of Economic 1994. “Endowments and the Allocation of Schooling in Research, Cambridge, MA. The great schooling expansion—and those it has left behind | 65 Hanushek, Eric A., and Ludger Woessmann. 2008. “The Role the Need to Turn India’s Right to Education Act Upside of Cognitive Skills in Economic Development.” Journal of Down.” In India Infrastructure Report 2012: Private Sector Economic Literature 46 (3): 607–68. in Education, edited by IDFC Foundation, 109–26. New He, Sylvia Y., and Genevieve Giuliano. 2017. “School Choice: Delhi: Routledge. Understanding the Trade-Off between Travel Distance Murnane, Richard J., and Alejandro Ganimian. 2014. and School Quality.” Transportation. DOI 10.1007/s11116 “Improving Educational Outcomes in Developing Coun- -017-9773-3. tries: Lessons from Rigorous Evaluations.” NBER Work- IRIN. 2003. “Interview with Nadya, Rural Teacher.” Inter- ing Paper 20284, National Bureau of Economic Research, view: Human Rights (October 13), IRIN, Geneva. http:// Cambridge, MA. www.irinnews.org/report/20764/afghanistan-interview NCCM (National Council for Childhood and Motherhood) -nadya-rural-teacher. and UNICEF (United Nations Children’s Fund). 2015. Jensen, Robert. 2010. “The (Perceived) Returns to Education Violence against Children in Egypt: A Quantitative Survey and and the Demand for Schooling.” Quarterly Journal of Qualitative Study in Cairo, Alexandria and Assiut. Cairo: Economics 125 (2): 515–48. NCCM and UNICEF. Kabeer, Naila. 2015. “Tracking the Gender Politics of the Nguyen, Trang. 2008. “Information, Role Models, and Per- Millennium Development Goals: Struggles for Interpre- ceived Returns to Education: Experimental Evidence tive Power in the International Development Agenda.” from Madagascar.” MIT working paper, Massachusetts Third World Quarterly 36 (2): 377–95. Institute of Technology, Cambridge, MA. Kelland, Zoe. 2016. “Education: 4 Children Explain How Edu- Orazem, Peter F., and Elizabeth M. King. 2008. “Schooling in cation Has Changed Their Lives.” Global Citizen (February Developing Countries: The Roles of Supply, Demand, 25), Global Poverty Project, New York. https://www and Government Policy.” In Handbook of Development .globalcitizen.org/en/content/4-children-explain-how Economics, Vol. 4, edited by T. Paul Schultz and John A. -education-has-changed-their/. Strauss, 3475–559. Handbooks in Economics Series 9. Kobiané, Jean-François, and Moussa Bougma. 2009. Burkina Amsterdam: North-Holland. Faso, RGPH 2006, Rapport d’analyse du thème IV: Instruction, Pritchett, Lant. 2013. The Rebirth of Education: Schooling Ain’t Alphabétisation et Scolarisation. Ouagadougou, Burkina Learning. Washington, DC: Center for Global Develop- Faso: Institut National de la Statistique et de la ment; Baltimore: Brookings Institution Press. Démographie. Rivkin, Steven G., Eric A. Hanushek, and John F. Kain. 2005. Kremer, Michael R., and Alaka Holla. 2009. “Improving “Teachers, Schools, and Academic Achievement.” Econo- Education in the Developing World: What Have We metrica 73 (2): 417–58. Learned from Randomized Evaluations?” Annual Review Tooley, James, and Pauline Dixon. 2006. “ ‘De Facto’ Privatisa- of Economics 1: 513–45. tion of Education and the Poor: Implications of a Study Lanjouw, Peter F., and Martin Ravallion. 1999. “Benefit Inci- from Sub-Saharan Africa and India.” Compare 36 (4): dence, Public Spending Reforms, and the Timing of Pro- 443–62. gram Capture.” World Bank Economic Review 13 (2): 257–73. —— — —. 2007. “Private Education for Low-Income Families: Lee, Jong-Wha, and Hanol Lee. 2016. “Human Capital in the Results from a Global Research Project.” In Private Long Run.” Journal of Development Economics 122 (Septem- Schooling in Less Economically Developed Countries: Asian ber): 147–69. and African Perspectives, edited by Prachi Srivastava and Lewis, Maureen A., and Marlaine E. Lockheed. 2006. Inexcus- Geoffrey Walford, 15–39. Oxford Studies in Comparative able Absence: Why 60 Million Girls Still Aren’t in School and Education Series. Oxford, U.K.: Symposium Books. What to Do about It. Washington, DC: Center for Global UIS (UNESCO Institute for Statistics). 2016. Education (data- Development. base). Montreal. http://data.uis.unesco.org/. Lincove, Jane Arnold. 2015. “Improving Identification of UIS (UNESCO Institute for Statistics) and UNESCO (United Demand-Side Obstacles to Schooling: Findings from Nations Educational, Scientific, and Cultural Organiza- Revealed and Stated Preference Models in Two SSA tion). 2015. “A Growing Number of Children and Adoles- Countries.” World Development 66 (February): 69–83. cents Are Out of School as Aid Fails to Meet the Mark.” Lockheed, Marlaine. 2010. “Gender and Social Exclusion.” Paris: Policy Paper 22/Fact Sheet 31, UIS, Montreal; UNESCO, United Nations Educational, Scientific, and Cultural Paris. Organization. —— — —. 2016. “Leaving No One Behind: How Far on the Way López-Calva, Luis F., and Harry Anthony Patrinos. 2015. to Universal Primary and Secondary Education?” Policy “Exploring the Differential Impact of Public Interven- Paper 27/Fact Sheet 37, UIS, Montreal; UNESCO, Paris. tions on Indigenous People: Lessons from Mexico’s Con- UNESCO (United Nations Educational, Scientific, and Cul- ditional Cash Transfer Program.” Journal of Human Devel- tural Organization). 2011. EFA Global Monitoring Report opment and Capabilities 16 (3): 452–67. 2011, The Hidden Crisis: Armed Conflict and Education. Paris: Mete, Cem, ed. 2008. Economic Implications of Chronic Illness UNESCO. and Disability in Eastern Europe and the Former Soviet Union. —— — —. 2013. Children Still Battling to Go to School. Paris: Report 42851 rev. Washington, DC: World Bank. UNESCO. Mukerji, Shobhini, and Michael Walton. 2013. “Learning —— — —. 2015. EFA Global Monitoring Report 2015, Education for the Right Lessons: Measurement, Experimentation and All 2000–2015: Achievements and Challenges. Paris: UNESCO. 66 | World Development Report 2018 ——— —. 2016. Global Education Monitoring Report 2016, Educa- ——— —. 2011b. World Development Report 2012: Gender Equality tion for People and Planet: Creating Sustainable Futures for All. and Development. Washington, DC: World Bank. Paris: UNESCO. ——— —. 2014. Diagnostics and Policy Advice for Supporting Roma UNESCO (United Nations Educational, Scientific, and Cul- Inclusion in Romania. Report 89621. Washington, DC: tural Organization) and UNICEF (United Nations Chil- World Bank. dren’s Fund). 2015. “Fixing the Broken Promise of Educa- ——— —. 2015. World Development Report 2015: Mind, Society, and tion for All: Findings from the Global Initiative on Behavior. Washington, DC: World Bank. Out-of-School Children.” UNESCO Institute for Statis- ——— —. 2016a. “Lebanon: Support to Reaching All Children tics, Montreal. with Education (Race 2) Program-for-Results.” Program UNHCR (United Nations High Commissioner for Refugees). Appraisal Document, Report 108014-LB, International 2016. “Missing Out: Refugee Education in Crisis.” Development Association, World Bank, Washington, Geneva. http://www.unhcr.org/57d9d01d0. DC. United Nations. 2015. The Millennium Development Goals ——— —. 2016b. “Reaching Girls, Transforming Lives.” Snap- Report. New York: United Nations. shot, Education Global Practice, World Bank, Washing- World Bank. 2007. “People with Disabilities from India: ton, DC. From Commitments to Outcomes.” Report 41585, World ——— —. 2016c. World Development Indicators (database). Bank, Washington, DC. World Bank, Washington, DC. http://data.worldbank ——— —. 2011a. “Learning for All: Investing in People’s Knowl- .org/data-catalog/world-development-indicators. edge and Skills to Promote Development; World Bank ——— —. 2017. EdStats (database). Washington, DC. http:// Group Education Strategy 2020.” World Bank, Washing- datatopics.worldbank.org/education/. ton, DC. The great schooling expansion—and those it has left behind | 67 SPOTLIGHT 1 The biology of learning Research has dramatically expanded our understand- Although different parts of the brain have dif- ing of how the brain works—and therefore how peo- ferent sensitive periods, their development is inter­ ple learn. The brain is very malleable: it adapts to its dependent. Neural circuits (series of synapses) form surroundings. This phenomenon is called neuroplasti- sequentially and cumulatively: simpler networks city, which means the ability to learn is dictated not develop first, more complex ones later. Just as with only by genetic endowments, but also by how genes the construction of a house, the robustness of pro- interact with experiences and environmental inputs. gressively more complex brain structures depends Genes govern when specific brain circuits are formed, on the robustness of foundational ones. For example, but experiences can turn those genes on or off, as well the development of increasingly complex skills and as determine which neural connections—synapses— functions builds on circuits formed earlier: linguis- survive over the life cycle. Environmental inputs such tic development relies on visual and auditory func- as caregivers’ or teachers’ stimulation, or nutrition, or tions that are dependent on neural circuits lower violence shape the architecture of the brain from the in the hierarchy, which are most malleable earlier formative years on.1 in life; neural circuits that support higher cognitive The brain is malleable throughout life, even if functions, most malleable until as late as adoles- most brain development is completed by late adoles- cence, build on sensorial stimuli as well as linguistic cence or early adulthood. The fastest synaptic growth development.4 Moreover, physical, sensory-motor, (thus malleability) occurs between the prenatal period cognitive, and socioemotional development are inter- and age 3 (1 million new neural connections a second), dependent, constituting a web of dynamic links that with growth then gradually slowing.2 Because differ- ultimately determine a person’s ability to thrive. For ent parts of the brain develop at different times, and example, higher levels of health promote learning; because neuroplasticity is highest during develop- emotional security fosters child exploration, which mental stages, not all areas of the brain are equally leads to learning; and higher self-regulation reduces malleable at the same time (figure S1.1). The periods health risks.5 of greatest plasticity, or “sensitive periods,” whose A range of enriching experiences leads to more length varies widely by brain region, are character- complex synapses, but cumulative exposure to risk ized by an initial stage in which the brain develops far factors (such as neglect or violence) either eliminates more synapses than it needs. That stage is followed by synapses associated with healthy brain development, synaptic pruning in which, to maximize efficiency in or it consolidates those associated with unhealthy brain functionality, the neural connections used more development. Experiences affect the architecture of often grow more permanent, while those used less the brain in part because of the hormonal response are discarded to reach optimal levels of synapses (that they trigger. Hormones such as dopamine (trig- is, the adult level of synapses in figure S1.1).3 Because gered when the brain encounters novelty) stimulate most sensitive periods take place early in life, a 3-year- information absorption,6 whereas hormones such as old has, in total, far more brain synapses (about 1 qua- cortisol (associated with stress as well as negative drillion) than an adult (100–500 trillion). emotions) can shut off learning.7 68 | World Development Report 2018 FIGURE S1.1 Figure S1.1 Synapse development over the first 20 years of life Year 1 Birth Number of connections Number of connections (synapses) in the brain (synapses) in the brain at birth at age 6 months Experience-dependent synapse formation Peak Synapse levels stabilize in adulthood 0 -.5 0 1 5 10 15 20 Age (years) Sensory pathways Language Higher cognitive function Sources: Parker (2015); Thompson and Nelson (2001). Adapted with permission from Lawson Parker/National Geographic Creative; further permission required for reuse. Synapse drawings based on Golgi stain preparations (1939–1967) by J. L. Conel. Note: The figure is a representation of synapse development for selected brain functions over the life course. Not drawn to scale. The available insights on brain development have Interventions to improve learning and skills implications for investments in learning and skill should place a greater emphasis on the areas of the formation. Because brain malleability is much greater brain that are the most malleable over the life course. earlier in life and brain development is sequential Children’s brains are most efficient at incorporating and cumulative, establishing sound foundations can new information through exploration, play, and inter- lead to a virtuous cycle of skill acquisition. More- actions with caring adults or peers. Because of this over, investment in experiences and environmental receptivity, preschool programs should concentrate inputs that foster learning at the very earliest stages on building foundational skills through developmen- increases the impact of investments at later stages: tally appropriate program structures that emphasize skills beget skills.8 Weak foundations, by contrast, result play and interaction.10 Although foundational cogni- in the accumulation of learning gaps, as well as higher tive skills become less malleable after age 10, some risks of poor biological development that hamper skill areas associated with socioemotional development formation—with repercussions over the life cycle remain highly malleable through early adulthood. (see chapter 5). Yet the optimal periods for cultivat- Accordingly, interventions that aim to improve the ing higher-order cognitive and socioemotional skills school-to-work transition, as well as social inclusion occur throughout childhood, adolescence, and early for youth with weak foundational skills, may prove adulthood.9 Furthermore, the brain’s ability to adapt most effective when they emphasize socioemotional to its environment, learn, and acquire new skills skills.11 continues throughout life (that is, the experience- Teaching strategies can deeply influence how dependent synapse formation in figure S1.1). Thus students approach challenges in and out of school. investments in environmental inputs are needed well Because the brain thrives when exposed to novelty, beyond early childhood to sustain learning along with incorporating enriching opportunities for learning skills development. along with exploration may lead to better learning The biology of learning | 69 outcomes. Finally, intense stress or sustained nega- from flourishing in the classroom (see spotlight 2 on tive emotions—such as those associated with crises or the effects of poverty). Consequently, programs that acute deprivation, where multiple stressors coexist— increase the availability of protective factors to shel- interfere with the brain’s ability to learn, retain, and ter children from stress (such as nurturing care from use information. Extended exposure to stressors is at least one meaningful relationship that teaches chil- toxic for biological systems, particularly for develop- dren how to cope) can improve not only schooling, ing brains, and it may impede disadvantaged children but also overall life outcomes. Notes 1. Knudsen (2004). Circuit Explain Behavioral Changes in Reinforcement 2. Shonkoff and Phillips (2000). Learning.” Frontiers in Behavioral Neuroscience 5 (15). 3. Knudsen and others (2006). Kautz, Tim, James J. Heckman, Ron Diris, Bas Ter Weel, and 4. Center on the Developing Child (2009). Lex Borghans. 2014. “Fostering and Measuring Skills: 5. Heckman (2007); Knudsen and others (2006). Improving Cognitive and Non-cognitive Skills to Pro- 6. Hong and Hikosaka (2011). mote Lifetime Success.” NBER Working Paper 20749, 7. McEwen and Gianaros (2010). National Bureau of Economic Research, Cambridge, MA. 8. Cunha and Heckman (2007); Cunha and others (2006). Knudsen, Eric I. 2004. “Sensitive Periods in the Development 9. Guerra, Modecki, and Cunningham (2014). of the Brain and Behavior.” Journal of Cognitive Neuro­ 10. Whitebread and Bingham (2011). science 16 (8): 1412–25. 11. Kautz and others (2014). Knudsen, Eric I., James J. Heckman, Judy L. Cameron, and Jack P. Shankoff. 2006. “Economic, Neurobiological, and Behavioral Perspectives on Building America’s Future References Workforce.” Proceedings of the National Academy of Sciences 103 (27): 10155–62. Center on the Developing Child. 2009. “In Brief: The Science McEwen, Bruce S., and Peter J. Gianaros. 2010. “Central Role of Early Childhood Development.” Center on the Devel- of the Brain in Stress and Adaptation: Links to Socioeco- oping Child, Harvard University, Cambridge, MA. http:// nomic Status, Health, and Disease.” Annals of the New York developingchild.harvard.edu/resources/inbrief-science Academy of Sciences 1186 (1): 190–222. -of-ecd. Parker, Lawson. 2015. “Neural Network.” Graphic in Yudhijit Cunha, Flavio, and James J. Heckman. 2007. “The Technology Bhattacharjee, “Baby Brains: The First Year,” National Geo- of Skill Formation.” American Economic Review 97 (2): graphic, January. http://ngm.nationalgeographic.com/2015 31–47. /01/baby-brains/bhattacharjee-text. Cunha, Flavio, James J. Heckman, Lance Lochner, and Dimi- Shonkoff, Jack P., and Deborah A. Phillips, eds. 2000. From triy V. Masterov. 2006. “Interpreting the Evidence on Life Neurons to Neighborhoods: The Science of Early Childhood Cycle Skill Formation.” In Handbook of the Economics of Development. Washington, DC: National Academies Education, Vol. 1, edited by Eric A. Hanushek and Finis Press. Welch, 697–812. Handbooks in Economics Series 26. Thompson, Ross A., and Charles A. Nelson. 2001. “Develop- Amsterdam: North-Holland. mental Science and the Media: Early Brain Develop- Guerra, Nancy, Kathryn Modecki, and Wendy Cunningham. ment.” American Psychologist 56 (1): 5–15. 2014. “Developing Social-Emotional Skills for the Labor Whitebread, David, and Sue Bingham. 2011. “School Readi- Market: The Practice Model.” Policy Research Working ness: A Critical Review of Perspectives and Evidence.” Paper 7123, World Bank, Washington, DC. TACTYC Occasional Paper 2, Association for the Profes- Heckman, James J. 2007. “The Economics, Technology, and sional Development of Early Years Educators, University Neuroscience of Human Capital Formation.” Proceedings of Cambridge, Cambridge, U.K. of the National Academy of Sciences 104 (33): 13250–55. Hong, Simon, and Okihide Hikosaka. 2011. “Dopamine- Mediated Learning and Switching in Cortico-Striatal 70 | World Development Report 2018 The many faces of the learning crisis 3  abia Nura, a 16-year-old girl from Kano in northern Nigeria, goes to school despite R ever-present threats from Boko Haram. She is determined to become a doctor (Smith 2014). But 37 million African children will learn so little in school that they will not be much better off than kids who never attend school (van Fleet 2012). The global schooling expansion hides another statis- read a single word by the end of grade 2.2 In Guyana tic: for millions, schooling is not producing enough in 2008 (when the most recent data were collected), learning. Learning outcomes in basic education are so that share was 29 percent at the start of grade 3.3 low, in so many contexts, that the developing world A similar picture emerges for numeracy. In rural India is facing a learning crisis. In many low-income coun- in 2016, less than 28 percent of students in grade 3 tries, learning levels are low in an absolute sense, could master double-digit subtraction.4 Emerging while in many middle-income countries average data on student achievement show that, for millions, learning levels remain far behind those in high- schooling is producing little learning in crucial early income countries. The learning crisis disproportion- grades.5 ately affects children from poor households: they are Millions complete primary education without far more likely to leave school without acquiring basic acquiring the basic competencies needed for further skills like literacy or numeracy. Ultimately, the learn- learning. According to a 2014 regional assessment, ing crisis translates into severe shortcomings in the among grade 6 students in West and Central Africa, skills of the workforce. nearly 58 percent are not sufficiently competent in reading or mathematics to continue schooling (figure 3.1).6 Similarly, the most recently available regional For too many, learning isn’t assessment of grade 6 students in southern and East happening Africa (from 2007) shows that 37 percent are not com- petent in reading, and more than 60 percent are not In South Africa, 27 percent of 12-year-olds competent in mathematics (figure 3.2).7 This lack of were enrolled in grade 6 but were functionally basic competency is systematically lower for students illiterate; in Zambia that share was 44 percent from poorer families. In Honduras, half of the grade 6 (Hungi and others 2010). students from the poorest quintile scored at the low- est reading competency level in a 2013 regional Latin Globally, 125 million children are not acquiring func- American assessment; only 7 percent of those from tional literacy or numeracy, even after spending at the richest quintile did (figure 3.3). least four years in school.1 In Malawi and Zambia Low-performing education systems are failing in 2012, more than 89 percent of students could not to meet their own curriculum standards (box 3.1). The many faces of the learning crisis | 71 Figure 3.1 Most grade 6 students in West and Central Africa are not sufficiently competent in reading or mathematics Competency levels from PASEC (2014), by subject, participating countries a. Reading b. Mathematics Niger Niger Chad Chad Togo Côte d’Ivoire Congo, Rep. Congo, Rep. All PASEC countries Cameroon Côte d’Ivoire All PASEC countries Cameroon Benin Benin Togo Burundi Senegal Burkina Faso Burkina Faso Senegal Burundi 100 50 0 50 100 100 50 0 50 100 Percent Percent Not competent Low competency High competency Source: WDR 2018 team, using data from Programme d’Analyse des Systèmes Éducatifs de la Confemen (PASEC 2015) and World Bank’s World Development Indicators (World Bank 2016c). Data at http://bit.do/WDR2018-Fig_3-1. Note: For the PASEC reading exam, “not competent” refers to levels 0–2 in the original coding, “low competency” to level 3, and “high competency” to level 4. For the PASEC mathematics exam, “not competent” refers to levels 0–1 in the original coding, “low competency” to level 2, and “high competency” to level 3. Figure 3.2 Most grade 6 students in southern and East Africa are not sufficiently competent in mathematics, and several countries score poorly in reading as well Competency levels from SACMEQ (2007), by subject, participating countries a. Reading b. Mathematics Malawi Zambia Zambia Malawi Lesotho Namibia South Africa Lesotho Uganda Uganda Mozambique Mozambique Namibia South Africa Zimbabwe All SACMEQ countries All SACMEQ countries Zimbabwe Botswana Botswana Seychelles Swaziland Mauritius Tanzania Kenya Seychelles Tanzania Kenya Swaziland Mauritius 100 50 0 50 100 100 50 0 50 100 Percent Percent Not competent Low competency High competency Source: WDR 2018 team, using data from Southern and Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ 2007) and World Bank’s World Development Indicators (World Bank 2016c). Data at http://bit.do/WDR2018-Fig_3-2. Note: “Not competent” refers to levels 1–3 in the original SACMEQ coding, “low competency” to level 4, and “high competency” to levels 5–8. 72 | World Development Report 2018 Figure 3.3 Learning outcomes are substantially lower for poor children in Latin America Competency levels for TERCE (2013) grade 6, by subject, for students in poorest and richest socioeconomic quintiles, participating countries a. Reading b. Mathematics Poor Poor Paraguay Dominican Republic Rich Rich Poor Poor Dominican Republic Paraguay Rich Rich Poor Poor Honduras Nicaragua Rich Rich Poor Poor Panama Panama Rich Rich Poor Poor Nicaragua Honduras Rich Rich Poor Poor Peru Guatemala Rich Rich Ecuador Poor Ecuador Poor Rich Rich Guatemala Poor Peru Poor Rich Rich Argentina Poor Colombia Poor Rich Rich Brazil Poor Argentina Poor Rich Rich Mexico Poor Brazil Poor Rich Rich Uruguay Poor Costa Rica Poor Rich Rich Colombia Poor Uruguay Poor Rich Rich Costa Rica Poor Mexico Poor Rich Rich Chile Poor Chile Poor Rich Rich 100 50 0 50 100 100 50 0 50 100 Percent Percent Not competent Low competency High competency Source: WDR 2018 team, using data from Third Regional Comparative and Explanatory Study (TERCE), 2013 (UNESCO 2013). Data at http://bit.do /WDR2018-Fig_3-3. Note: Socioeconomic quintiles are defined nationally. “Not competent” refers to level 1 in the original coding, “low competency” to levels 2–3, and “high competency” to level 4. While test scores on international assessments may These low learning levels are not an inevitable be worryingly low, similar patterns emerge when by-product of rapidly expanding education. Starting assessing students against national standards. In in the 1950s, the Republic of Korea focused on ensur- urban Pakistan in 2015, only three-fifths of grade 3 ing quality primary education for the vast majority of students could correctly perform a subtraction like its population before shifting to a similar emphasis 54 – 25; in rural areas only two-fifths could.8 Across 51 on secondary and ultimately higher education—with countries, only about half of women who completed excellent learning results. The success of this strategy grade 6 (but no higher) could read a single sentence.9 shows it is possible to ensure quality education even It is hard to imagine that these women are reaping while rapidly expanding schooling. The key ingredi- the full potential economic or social returns from ent is a persistent emphasis on the needs of the poor their years of schooling. and disadvantaged.10 The more recent experience The many faces of the learning crisis | 73 Box 3.1 Those who can’t read by the end of grade 2 struggle to catch up Illiteracy at the end of grade 2 has long-term consequences Second, schools do not offer struggling students a chance for two reasons. First, learning is cumulative. Education to catch up. In many contexts, the pace of classroom instruc- systems around the world expect students to acquire foun- tion is determined by the need to cover an overly ambitious dational skills such as reading by grades 1 or 2. By grade 3, curriculum rather than by the pace of student learning.c This students need to read to access their curriculum. Students means teachers have no choice but to ignore students who are who master these foundational skills early are at an falling behind. In India and Kenya, for example, the curriculum advantage: skills from early grades are strongly positively has been designed for the elite.d Teachers and textbooks focus associated with later school performance (see spotlight on advanced topics that are of little use in helping struggling 1). a Children who cannot read by grade 3 fall behind and students.e These students then fall even further behind— struggle to catch up, perhaps irreparably.b eventually so far that no learning whatsoever takes place.f Source: WDR 2018 team. a. Glick and Sahn (2010). b. Muralidharan and Zieleniak (2013). c. Pritchett and Beatty (2012). d. Banerjee and Duflo (2012); Glewwe, Kremer, and Moulin (2009). e. Pritchett and Beatty (2012). f. Pritchett and Beatty (2012). of Vietnam reinforces that lesson; it, too, has main- standards of quality has steadily increased over the tained education quality during rapid expansion by last 25 years.16 ensuring that disadvantaged students receive rela- Low student achievement in some middle-income tively equitable access to quality schooling.11 countries relative to their economic competitors Even in middle-income countries, millions of signals a failure to live up to their own expectations. students are lagging behind. In Brazil, internation- According to the leading international assessments ally comparable assessments reveal that more than of literacy and numeracy, the average student in three-quarters of youth are reaching the age of 15 low-income countries performs worse than 95 percent without being able to perform at the lowest level of of the students in Organisation for Economic Co-oper- competence on the Programme for International ation and Development (OECD) countries—meaning Student Assessment (PISA) tests.12 Similarly, a third of that student would be singled out for remedial atten- students in Paraguay have only a basic grasp of read- tion in a class in a wealthier country.17 In Colombia, ing skills (“reading for meaning”) by grade 6.13 These Indonesia, and Peru, student performance at the 75th students are therefore ill-equipped to participate percentile on the PISA math test is barely above that in their economy and society.14 These numbers also at the 25th percentile of the OECD average. In Algeria, show interesting gender-based differences (box 3.2). the Dominican Republic, Kosovo, and Tunisia, it is However, some countries are doing better. below the 25th percentile of the OECD average (figure Albania, Peru, and Portugal have made impressive 3.4). The disparity between the average PISA score progress in improving average student achievement for Latin American countries and OECD countries is relative to countries with similar incomes.15 Latvia equivalent to over two full years of math education. outperforms several countries in eastern Europe; Based on its rate of progress in average PISA scores Vietnam is a positive outlier in Southeast Asia. from 2003 to 2015, it would take Tunisia over 180 years Although it is not always possible to clearly isolate to reach the OECD average in math. This slow rate the factors responsible for systemwide improve- of improvement is especially problematic for middle- ments in student learning, a policy focus on educa- income countries trying to position themselves as tion quality appears to be important. For example, a important players in the global economic landscape. major component of Vietnam’s strong performance The mapping between schooling and workforce has been a convergence in school quality within the skills varies dramatically across countries. For exam- country. The share of schools that meet the national ple, the working-age population in Colombia reaches 74 | World Development Report 2018 Box 3.2 Gender-based differences in learning depend on the subject Data from internationally benchmarked tests show that tested.a Whether on UNESCO’s Third Regional Comparative boys lag behind girls on test averages. In all but 6 of the 72 and Explanatory Study (TERCE), the Scholastic Aptitude countries and economies participating in the Programme Test (SAT) in the United States, or an array of national for International Student Assessment (PISA), 15-year-old assessments, girls outperform boys on total test averages boys are more likely than girls of the same age to be low in most countries and economies.b achievers on the composite average of the three subjects Figure B3.2.1 Girls outperform boys on reading in all countries and economies, but boys typically do better in mathematics and science Distribution across countries and economies of gap between mean score for girls and mean score for boys Boys marginally outperform girls Girls outperform boys in in mathematics reading by a substantial and science Density of gender gaps margin in every country and economy −30 −20 −10 0 10 20 30 40 50 60 70 Gender gap (girls-boys) Mathematics Science Reading Source: WDR 2018 team, using data from Programme for International Student Assessment (PISA) collected in 2015 (OECD 2016a). Data at http://bit.do /WDR2018-Fig_B3-2-1. Note: Distribution based on data from 72 countries and economies that participated in PISA 2015. This higher average performance by girls masks important grade 2, boys score slightly better in mathematics and worse variations across subjects (figure B3.2.1). Girls consistently in reading starting in grade 3. This subject-specific gender score higher in reading and writing; boys tend to perform bet- gap continues to grow through secondary school.d ter on mathematics and science in most countries and econo- But the mathematics and science gap in favor of mies where these tests—and others, such as the Programme boys may be shrinking. Results from the 2015 Trends in d’Analyse des Systèmes Éducatifs de la Confemen (PASEC) International Mathematics and Science Study (TIMSS) and Southern and Eastern Africa Consortium for Monitoring assessment were much more mixed than for previous Educational Quality (SACMEQ)—are administered.c years: in about half the countries and economies tested, Whereas girls and boys perform at equal levels in there were no statistically significant differences in gender both mathematics and reading from kindergarten through performance in these subjects.e Source: WDR 2018 team. a. OECD (2015). b. TERCE: UNESCO (2016); SAT: Fryer and Levitt (2010); national assessments: Bharadwaj and others (2015); Cornwell, Mustard, and Van Parys (2013); Uwezo (2014, 2015). c. Dickerson, McIntosh, and Valente (2015). d. Fryer and Levitt (2010); Singh (2016); UNESCO (2016). e. Mullis, Martin, and Loveless (2016). The many faces of the learning crisis | 75 Figure 3.4 Learning outcomes vary greatly across countries and economies— in several countries, the 75th percentile of PISA test takers performs below the 25th percentile of the OECD average Performance of 25th, 50th, and 75th percentiles in 2015 PISA assessment, participating non-OECD economies and selected OECD economies a. Reading b. Mathematics Percentile Percentile 25th 50th 75th 25th 50th 75th OECD average OECD average Singapore Singapore Finland Hong Kong SAR, China Hong Kong SAR, China Taiwan, China Korea, Rep. Macao SAR, China Macao SAR, China China (B-S-J-G) Poland Korea, Rep. Taiwan, China Finland China (B-S-J-G) Poland Russian Federation Russian Federation Latvia Vietnam Croatia Malta Vietnam Latvia Lithuania Lithuania Chile Croatia Malta Kazakhstan Cyprus Malaysia Bulgaria Romania Uruguay Bulgaria Malaysia Cyprus Romania United Arab Emirates United Arab Emirates Chile Turkey Moldova Trinidad and Tobago Turkey Argentina Montenegro Montenegro Uruguay Costa Rica Trinidad and Tobago Colombia Albania Kazakhstan Thailand Mexico Mexico Moldova Argentina Jordan Georgia Thailand Costa Rica Albania Qatar Brazil Lebanon Georgia Colombia Qatar Peru Peru Jordan Indonesia Indonesia Tunisia Brazil Dominican Republic Macedonia, FYR Macedonia, FYR Tunisia Kosovo Kosovo Algeria Algeria Lebanon Dominican Republic 300 400 500 600 300 400 500 600 Score Score Non-OECD OECD OECD interquartile range Source: WDR 2018 team, using data from Programme for International Student Assessment (PISA) collected in 2015 (OECD 2016a). Data at http://bit.do /WDR2018-Fig_3-4. Note: PISA 2015 defines baseline levels of proficiency at a score of 407 for reading and 420 for mathematics. China (B-S-J-G) = China (Beijing-Shanghai- Jiangsu-Guangdong). basic literacy proficiency by the lower secondary percent of Ghana’s working-age population and over 60 level, whereas the population of Bolivia needs six percent of Kenya’s have just level 1 literacy or below— more years to attain even close to the same profi- that is, their literacy proficiency is limited to under- ciency. Similarly, among 18- to 37-year-olds in Nigeria, standing basic texts, but they are not able to integrate, only 19 percent of primary completers can read; in evaluate, or interpret information from a variety of text Tanzania, 80 percent can.18 materials (figure 3.5).19 This contrasts with the average In some countries, large proportions of “educated” for high-income countries, where only 15 percent of the working adults are effectively low-skilled. Nearly 80 working-age population is at level 1 or below. Individuals 76 | World Development Report 2018 Figure 3.5 Middle-income countries underpin the development of higher-order cognitive, tend to have lower rates of literacy technical, and specialized skills. This skills deficit proficiency than high-income countries limits opportunities for further education or training (HICs) because the capacity to make up for lost skills shrinks over time: second-chance adult education programs Percentage of working-age population relative to minimal level of foundational literacy (2011–14) have limited success, and on-the-job training usually favors workers with more education and skills.21 The Middle-income consequences are dead-end jobs with relatively flat Ghana Kenya lifetime income growth for students leaving school Bolivia with poor foundational skills, a situation that will Turkey Colombia only get worse as technology affects the demand for Vietnam skills (see spotlight 5). Georgia Serbia Estimates based on 41 countries where skill mea- Armenia sures are available suggest that, globally, more than 2.1 Ukraine billion of 4.6 billion working-age adults (ages 15–64) Russian Federation lack crucial foundational skills.22 Among younger High-income adults (ages 15–24), the number is 418 million. While Chile Israel these skills gaps exist in all countries, their magnitude Italy is greater in developing countries (figure 3.6), with Spain Greece an estimated 92 million 15- to 24-year-olds affected Singapore in East Asia and Pacific, 120 million in South Asia, Slovenia France and 47 million in Latin America and the Caribbean. Poland The implications, already profound, will be felt more United States Germany acutely as jobs continue to shift from physical to Ireland more cognitive or socioemotional tasks. Progress in United Kingdom Canada Denmark Median, HICs Figure 3.6 Reading proficiency is low in many parts of 75 50 25 0 25 50 75 100 the developing world Percent Estimated population ages 15–24, by country group and level of reading Literacy proficiency: Low Medium–High proficiency Source: WDR 2018 team, using data from Programme for the International Assessment of Adult Competencies (PIAAC) collected between 2011 and Percent of world’s population 2014 (OECD 2016b, 2016c) and STEP Skills Measurement Program, 2011–14 0 20 40 60 80 100 (http://microdata.worldbank.org/index.php/catalog/step/about). Data at http://bit.do/WDR2018-Fig_3-5. 100 Note: Data are the latest available by country. PIAAC is representative at the national level for adults, ages 16–65. STEP is representative for urban 75 group’s population Percent of country populations, ages 15–64. Low proficiency is defined as level 1 and below on the assessments and indicates limited understanding of basic texts. Medium to high proficiency is defined as level 2 and above and indicates the ability to integrate, evaluate, and interpret information from a variety of text materials. 50 with low literacy proficiency are poorly prepared for 25 the labor market, further education, and on-the-job training. In rapidly modernizing labor markets, most 0 high-quality jobs—and even job training—require read- East Asia and Pacific South Asia High-income countries ing competency beyond minimum proficiency.20 Europe and Middle East Sub-Saharan Low skills continue to undermine career oppor- Central Asia and North Africa Africa tunities—and earnings—long after students leave Latin America school. Gaps in foundational skills affect not only and the Caribbean the starting points of new workers entering the labor Reading proficiency: Medium–High Low market but also their growth trajectories. Good foun- Source: WDR 2018 team, using data from Larson and Valerio (2017). Data at http://bit.do/WDR2018-Fig_3-6. dational skills are essential for further skills accu- Note: Model predicts proportion of working-age adults scoring at level 1 or below on the PIAAC-STEP mulation. Worldwide, many students leave school scale based on 41 countries; then projects to world population. PIAAC = Programme for the International without mastering the key cognitive skills that Assessment of Adult Competencies; STEP = STEP Skills Measurement Program. The many faces of the learning crisis | 77 meeting global development goals will be limited as evidence suggests not. In Pakistan, rich-poor learning long as the dimensions of this problem, its origins, gaps are smaller than learning gaps between children and its implications remain unrecognized. from good and bad schools. In tests of English lan- guage, the difference in learning between a high-per- forming and a low-performing public school is 24 Poor children learn the least, times the difference between children from poor and which hurts them the most nonpoor backgrounds, after controlling for observed Learning deficits are largest for poor people. In nearly child-level differences.29 Analysis of the 2009 PISA all countries, students’ family backgrounds—includ- found that “the best performing school systems [in ing parental education, socioeconomic status, and Canada; Finland; Hong Kong SAR, China; Japan; the conditions at home (such as access to books)—remain Republic of Korea; and Shanghai, China] manage to the largest predictors of learning outcomes (figure provide high-quality education to all students” rather 3.7).23 In France, the difference in science performance than only to students from privileged groups.30 on the 2015 PISA between the richest and the poorest students was 115 points.24 In Hungary, this difference What is causing the learning was 202 points.25 A 100-point difference in PISA scores is roughly equivalent to three years of schooling.26 crisis? The learning gap between rich and poor students A simple framework can be used to organize the prox- grows as students move to higher grades. In South imate (or immediate) determinants of the learning Africa, children in grade 3 from the poorest households crisis.31 Proximate determinants are those most are three years’ worth of learning behind children from directly linked to learning outcomes and are them- the richest households. This gap grows to four years’ selves the result of deeper determinants. The frame- worth of learning by grade 9.27 In Andhra Pradesh, work identifies four proximate determinants: learner India, testing the same set of students each year reveals preparation, teacher skills and motivation, the that this gap increases every year after grade 2.28 availability of relevant inputs, and the school man- Are learning gaps between rich and poor students agement and governance that bring these together simply a matter of household characteristics? Recent (figure 3.8). This approach provides a simple tool for Figure 3.7 Family socioeconomic status significantly affects students’ average PISA scores Distribution of scores on PISA 2015 across 69 countries (pooled) for students from the bottom and top quintiles of socioeconomic status, by subject a. Reading b. Mathematics c. Science 120 points = 137 points = 143 points = ~3.6 years ~4.1 years ~4.3 years Density of scores Density of scores Density of scores 200 300 400 500 600 700 200 300 400 500 600 700 200 300 400 500 600 700 Score Score Score Low socioeconomic status High socioeconomic status Source: WDR 2018 team, using data from Programme for International Student Assessment (PISA) collected in 2015 (OECD 2016a). Data at http://bit.do /WDR2018-Fig_3-7. Note: A year of education is assumed to equal roughly 33 points on the PISA exam in this analysis, and the gap is calculated as the difference between modal averages of the top and bottom quintiles for each subject. 78 | World Development Report 2018 FIGURE 3.7 systematically integrating a wide range of actors and Figure 3.8 The proximate determinants of learning factors in determining the learning process. Children do not arrive ready to learn Children from disadvantaged backgrounds tend to exhibit learning deficits years before they start school; rs Le these deficits leave them ill-prepared for the demands e ar of formal education. Acquiring foundational skills in ch ne a Te early childhood is essential for learning, and robust rs early childhood development can launch children on higher learning trajectories (see spotlight 1). But for children from disadvantaged backgrounds, adversities begin to accumulate before they are born. Chronic malnutrition, illness, the cumulative effects LEARNING of material deprivation, low parental support, and the unpredictable, chaotic, or violent environments that Sch can be associated with poverty all undermine early childhood development learning (see spotlight 2).32 oo ts Steep socioeconomic gradients in cognitive, lin- pu guistic, and early literacy development help determine lm na in a o l school outcomes. Language and cognitive gaps are evi- ge me ho dent before a child’s first birthday.33 In a wide range of nt Sc countries, from the Democratic Republic of Congo to the United States,34 children from poor households lag behind their more affluent peers by age 3, with gaps widening as children age (figure 3.9). Gaps in early Source: WDR 2018 team. Figure 3.9 Socioeconomic gaps in cognitive achievement grow with age—even in preschool years Percentage of children (ages 3–5) who can recognize 10 letters of the alphabet, by wealth quintile, selected countries a. Central African Republic b. Kazakhstan c. Tunisia 100 100 100 80 80 80 60 60 60 Percent Percent Percent 40 40 40 20 20 20 0 0 0 3 3.5 4 4.5 5 3 3.5 4 4.5 5 3 3.5 4 4.5 5 Age (years) Age (years) Age (years) Richest quintile Poorest quintile 95% confidence interval 95% confidence interval Source: WDR 2018 team, using data from Multiple Indicator Cluster Surveys (http://mics.unicef.org). Data are for 2010 for the Central African Republic, 2010–11 for Kazakhstan, and 2012 for Tunisia. Data at http://bit.do/WDR2018-Fig_3-9. The many faces of the learning crisis | 79 language and cognitive abilities are very alarming levels or more over a single school year, whereas because they are important predictors of performance those with a poor teacher advance just 0.5 grade throughout school and into early adulthood.35 levels.40 Across kindergartens in Ecuador, differences Poor foundations are evident in other crucial in learning outcomes for language, math, and execu- determinants of school performance such as socio- tive function are strongly associated with differences emotional and executive functions. Socioemotional in teacher behaviors and practices.41 No other school- skills include teamwork, motivation, and confidence, level factor has an impact nearly this large on student while executive functions (which rely on both socio- achievement.42 emotional and cognitive skills) include planning, But high-quality teachers are in short supply organizing, implementing, and multitasking, among in low-income countries.43 Less than 25 percent of others.36 The evidence on these developmental Sub-Saharan Africans currently complete secondary dimensions is more limited because of measure- education.44 Thus there are simply not enough qual- ment difficulties. Still, gaps in working memory and ified candidates to meet the growing demand for sustained attention (executive functions) for poorer teachers. In some countries, teacher training colleges children are evident starting at 6 months of age and have had to lower their entry requirements to ensure through the preschool years.37 Even in Madagascar, a an adequate supply of teachers. These colleges also very poor country, wealth gradients are apparent after push trainees through the programs in two years or accounting for maternal education and household less to try to fill the need for primary school teachers.45 inputs—gaps that widen with age.38 There are wealth In Latin America, there is evidence that candidates gradients of socioemotional development as well. entering the teaching profession are academically One in every three children between the ages of 3 and weaker than the pool of higher education students. 4 in a range of countries fails to meet basic milestones Fifteen-year-olds who identified themselves as inter- in socioemotional development, such as the ability to ested in a teaching career had much lower PISA math control aggressive behaviors, avoid distractions, and scores than students interested in engineering in get along with peers.39 every country in the region, and they scored below Because learning is cumulative and skills beget the national average in nearly all countries.46 Teachers skills (see spotlight 1), the cognitive and socioemo- may also not always have the necessary pedagogical tional developmental gaps that emerge at young ages skills; classroom observations in six countries in worsen over time. So do learning gaps: poor develop- Sub-Saharan Africa found that few public primary mental foundations and lower preschool skills mean school teachers are able to assess children’s abilities disadvantaged children arrive at school late and unpre- and evaluate students’ progress, and few engage in the pared to benefit fully from learning opportunities. As practices typically associated with good teaching.47 these children get older, it becomes harder and harder As a result, teachers often do not have sufficient for them to break out of lower learning trajectories. mastery of concepts they are expected to teach. In several Sub-Saharan countries, the average teacher Teachers often lack the needed skills and does not perform much better on reading tests than motivation the highest-performing grade 6 students.48 Across six Teachers are the most important determinant of stu- countries in the region, 40 percent of primary school dent learning. Estimates suggest that in the United teachers are not as knowledgeable as their students States, students with great teachers advance 1.5 grade should be (table 3.1).49 In Bihar, India, only 10.5 percent Table 3.1 Few teachers reach minimum thresholds of performance on knowledge assessments Percentage of teachers who score at least 80 percent on a test of grade 4 material Kenya Mozambique Nigeria Tanzania Togo Uganda Subject Average (2012) (2014) (2013)a (2014) (2013) (2013) Equivalent to student language curriculum 61 66 77 24 41 54 90 Equivalent to student mathematics curriculum 56 82 26 31 62 24 55 Source: Bold and others (2017). a. Data based on four states in Nigeria. 80 | World Development Report 2018 Figure 3.10 A lot of official teaching time is lost Percentage of time officially allocated to schooling that a teacher is scheduled to teach, is present in the classroom, and is actually teaching 100 80 60 Percent 40 20 0 p. . ep co n sia a ia ag o a ia a a e na R co il r na B l qu no Re g di bi ny nd ca ga bu z PD er an oc ,R ha To ni m ra bo m ) Ke bi ba as a ig ne b nz Tu or Za G en Ug o am ra N am Le La Se M Ta m A ad oz C Ye t, M M yp er Eg (P Percentage of time teacher is scheduled to teach present in the classroom teaching Sources: WDR 2018 team, using data from Abadzi (2009): Brazil (Pernambuco state), Ghana, Morocco, and Tunisia; Benveniste, Marshall, and Araujo (2008): Cambodia; Benveniste, Marshall, and Santibañez (2007): Lao People’s Democratic Republic; Millot and Lane (2002): Arab Republic of Egypt, Lebanon, and Republic of Yemen; World Bank (2016a): Madagascar; World Bank (2016b): Zambia; World Bank’s Service Delivery Indicators, 2012–13 (http://www.worldbank .org/sdi): Kenya, Mozambique, Nigeria, Senegal, Tanzania, Togo, and Uganda. Data at http://bit.do/WDR2018-Fig_3-10. Note: For Brazil, Cambodia, Ghana, Lao PDR, Senegal, Tanzania, and Tunisia, data include public schools. For all other countries, data include both public and private schools. of tested public school teachers are able to solve a This problem is particularly concerning because three-digit by one-digit division problem and show the bulk of national education budgets goes to teacher the steps correctly.50 salaries. In Latin America and the Caribbean, teacher Many developing countries suffer significant salaries absorb nearly 4 percent of the regional losses of instructional time (figure 3.10). Unan- gross domestic product (GDP).55 Staff compensation nounced visits to primary schools in six countries accounts for 80 percent of public spending on edu- found that in public schools, on average, about one cation in some countries (figure 3.11). If one in five teacher in five was absent on a typical school day.51 government primary school teachers is absent from Even when teachers are present in school, they may school, developing countries are wasting consider- not be teaching. In seven Sub-Saharan countries, able resources. students receive only about two and a half hours of According to recent data on 1,300 villages in teaching a day—less than half the scheduled time.52 India, nearly 24 percent of teachers were absent Teacher absenteeism and low time on task when in during unannounced visits, at an associated fiscal class—combined with other factors such as informal cost of US$1.5 billion a year.56 Reducing absenteeism school closures or student absenteeism—mean that in these schools would be over 10 times more cost- only about one-third of the total instructional time effective at increasing student-teacher contact time is used in Ethiopia, Ghana, and Guatemala.53 Even than hiring additional teachers. in middle-income countries in Latin America, about 20 percent of potential instructional time is lost—the School management skills are low equivalent of one less day of instruction a week.54 The effective management of schools relies on capac- There are many reasons for this loss of instructional ity and autonomy for decision making at the school time, including poor training and other demands on level, which are often lacking. Higher management teachers, and some teachers may perceive it as justi- quality57 and school leadership are associated with fied (box 3.3). But whatever the cause, lost teaching better education outcomes.58 Yet in many developing time reduces student learning. countries effective school management is missing The many faces of the learning crisis | 81 Box 3.3 Teachers may perceive low effort as being justified Teachers across a variety of countries justify certain types of service delivery gaps. Consider teacher perceptions of two basic aspects of teacher performance, based on teacher surveys administered in 2017 (figure B3.3.1). Figure B3.3.1 Teachers’ beliefs about their effort and its effects a. Teachers often consider their b. Teachers often feel they have little absence acceptable impact on poor children’s learning 100 100 Percent of teachers who agree with Percent of teachers who agree with the statements below the figure the statements below the figure 80 80 60 60 40 40 20 20 0 0 a an ar da sia n n ar da a an sia ar ar l l sta sta ga ga in in ib ib nm nm ist ist an an ne ne nt nt ne ne ki ki nz nz jik jik ge ge Ug Ug do do ya ya Pa Pa Se Se Za Za Ta Ta Ar Ar In In M M It is acceptable for teachers to be absent if they There is little teachers can do if parents complete the assigned curriculum do not have the necessary education leave students with work to do have too many personal or financial problems do something useful for the community Note: Responses are not mutually exclusive. Data are from public schools, except for Senegal (private and public schools) and Uganda (private schools). Source: Sabarwal and Abu-Jawdeh (2017). Data at http://bit.do/WDR2018-Fig_B3-3-1. Figure 3.11 Staff compensation consumes the largest share of resources available for public education Staff compensation as percentage of total expenditure on public education, by country and income group 100 80 60 Percent 40 20 0 Low-income Lower-middle-income Upper-middle-income High-income Source: WDR 2018 team, using data from UNESCO Institute for Statistics (UIS 2017). Data at http://bit.do/WDR2018-Fig_3-11. Note: Figure includes all countries with populations of over 500,000 for which spending data are available at the primary or secondary levels. Latest data available. 82 | World Development Report 2018 Figure 3.12 Management capacity (figure 3.12). Moreover, lack of autonomy prevents is low in schools in low- and middle- head teachers or school management committees income countries from improving service delivery.59 Even when the req- uisite autonomy exists, it may not be enough. Schools Distribution of management scores by sector, participating countries may choose not to exercise the provided authority or may lack the will and capacity to do so.60 For exam- ple, a survey in Uganda found that only 57 percent of Tanzania school management committee members reported having read their committee’s handbook.61 In Uttar India Pradesh, India, a quarter of village education commit- tee members surveyed did not even know they were members.62 Haiti School inputs have not kept pace In many developing countries, the expansion of Italy inputs has not kept pace with the explosion in enroll- ments. Governments have built classrooms and Brazil recruited teachers at unprecedented levels. But these efforts may not have kept up with rising enrollments, leading to a decline in per capita input availability. In Mexico Malawi between 2008 and 2015, as the gross enroll- ment rate in primary schools increased from 131 to Germany 146 percent, the average number of students per class increased from 85 to 126.63 Uganda introduced univer- sal primary education in 1997. The 68 percent increase Canada in primary school enrollment that followed increased the student-teacher ratio from 38:1 in 1996 to 80:1 in 1997 and the student-to-classroom ratio from 68:1 in United States 1996 to 105:1 in 1997.64 *** Sweden The learning crisis is real, but too often education systems operate as if it is not. Many policy makers United Kingdom do not realize how low learning levels are. Others do not acknowledge them or simply equate low learning 1 2 3 4 5 with low resources. Still, there are reasons for opti- Management score mism. First, learning is increasingly in the spotlight. Education Manufacturing Second, learning metrics are generating irrefutable Sources: Bloom and others (2014, 2015); Lemos and Scur (2016), with evidence of the learning crisis, thereby creating updates. Data at http://bit.do/WDR2018-Fig_3-12. pressure for action (see chapter 4).65 Third, promising Note: The underlying distributions for the education data are shown as new insights on how to tackle the crisis are becoming bars; for both sectors, the smoothed distributions are shown as curves. The indexes are constructed from the nine items that are comparable available (see parts III and IV of this Report). across sectors. Data on manufacturing are not available for Haiti. Notes 1. UNESCO (2014). 7. Southern and Eastern Africa Consortium for Monitor- 2. RTI International (2015). ing Educational Quality (SACMEQ) results for grade 6 3. Gove and Cvelich (2011). students in 15 countries in 2007 (Hungi 2010). 4. ASER Centre (2017). 8. ASER Pakistan (2015a, 2015b). 5. Muralidharan and Zieleniak (2013); Pritchett (2013). 9. Pritchett and Sandefur (2017). 6. Programme d’Analyse des Systèmes Éducatifs de la 10. Lee and Hong (2016). Confemen (PASEC) results for grade 6 students in 10 11. Dang and Glewwe (2017). francophone countries in 2015 (PASEC 2015). 12. Filmer, Hasan, and Pritchett (2006). The many faces of the learning crisis | 83 13. UNESCO (2015). 2. Bold and others (2017). 5 14. Filmer, Hasan, and Pritchett (2006). 53. Informal school closures may stem from strikes, inclem- 15. OECD (2016a). ent weather, or ad hoc holidays. For Ethiopia and Guate- 16. Dang and Glewwe (2017). mala see EQUIP2 (2010); for Ghana, see Abadzi (2009). 17. Crouch and Gove (2011), which is based on the Progress 54. Bruns and Luque (2015). in International Reading Literacy Study (PIRLS) and 55. Bruns and Luque (2015). Trends in International Mathematics and Science Study 56. Muralidharan and others (2017). (TIMSS). 57. Bloom and others (2015); Fryer (2017). 18. Kaffenberger and Pritchett (2017). 58. Robinson, Lloyd, and Rowe (2008). 19. The literacy skills proficiency construct covers a range of 59. Bruns, Filmer, and Patrinos (2011); Orazem, Glewwe, and skills from the decoding of written words and sentences Patrinos (2007). to the comprehension, interpretation, and evaluation of 60. King, Özler, and Rawlings (1999). complex texts. It takes into account workplace, personal, 61. Najjumba, Habyarimana, and Bunjo (2013). society, and community. To facilitate the interpretation 62. Banerjee and others (2010). of individual scores, described proficiency scales are 63. MoEST (2008, 2015); World Bank (2016c). Gross enroll- available to articulate the requisite skills and knowledge ment includes students whose age exceeds the official needed to perform specific tasks along a 500-point scale. age group for a particular education level, so the rate The tasks move progressively in complexity from level 0 may exceed 100 percent. to level 5. See ETS (2014) and OECD (2016c). 64. Bentaouet-Kattan (2006). 20. Desjardins and Rubenson (2011); OECD (2016b). 65. The World Bank’s 2020 sector strategy and the U.K. 21. di Gropello (2011); Fouarge, Schils, and de Grip (2013); Department for International Development’s 2010 strat- Heckman (2000); O’Connell and Jungblut (2008); Wind- egy are learning for all; the U.S. Agency for International isch (2015). Development’s strategy is opportunity through learn- 22. WDR 2018 team estimates using literacy proficiency as a ing; and AusAID is also adopting learning goals. proxy for foundational skills. 23. Bruns and Luque (2015); Filmer and Pritchett (1999). 24. OECD (2016a). References 25. OECD (2016a). Abadzi, Helen. 2009. “Instructional Time Loss in Developing 26. OECD (2016a). Countries: Concepts, Measurement, and Implications.” 27. Spaull and Kotze (2015). World Bank Research Observer 24 (2): 267–90. 28. Muralidharan and Zieleniak (2013). Araujo, María Caridad, Pedro Carneiro, Yyannú Cruz- 29. Das, Pandey, and Zajonc (2006). Aguayo, and Norbert R. Schady. 2016. “Teacher Quality 30. OECD (2010). and Learning Outcomes in Kindergarten.” Quarterly 31. Hanushek (1979). Journal of Economics 131 (3): 1415–53. 32. Lupien and others (2000); McCoy and others (2016); ASER Centre. 2017. Annual Status of Education Report (Rural). Walker and others (2007). New Delhi: ASER Centre. 33. Rubio-Codina and others (2015). ASER Pakistan. 2015a. “Annual Status of Education Report: 34. Countries where these outcomes have been evaluated ASER Pakistan 2015 National (Rural).” Lahore, Pakistan: include Cambodia, Chile, the Democratic Republic of South Asian Forum for Education Development. Congo, Ethiopia, India, Madagascar, Mozambique, Nige- —— ——. 2015b. “Annual Status of Education Report: ASER ria, Sierra Leone, Togo, the United States, and Vietnam. Pakistan 2015 National (Urban).” Lahore, Pakistan: South 35. Fernald, Marchman, and Weisleder (2013). Asian Forum for Education Development. 36. Galasso, Weber, and Fernald (2017); McCoy and others Banerjee, Abhijit Vinayak, Rukmini Banerji, Esther Duflo, (2016). Rachel Glennerster, and Stuti Khemani. 2010. “Pitfalls of 37. Fernald and others (2012); Lipina and others (2005); Participatory Programs: Evidence from a Randomized Noble, Norman, and Farah (2005). Evaluation in Education in India.” American Economic 38. Galasso, Weber, and Fernald (2017). Journal: Economic Policy 2 (1): 1–30. 39. McCoy and others (2016). Banerjee, Abhijit Vinayak, and Esther Duflo. 2012. Poor 40. Hanushek (1992); Rockoff (2004). Economics: A Radical Rethinking of the Way to Fight Global 41. Araujo and others (2016). Poverty. New York: PublicAffairs. 42. Bruns and Luque (2015). Bentaouet-Kattan, Raja. 2006. “Implementation of Free 43. UIS (2006). Basic Education Policy.” Education Working Paper 7 44. UNESCO Institute for Statistics, 2016 (UIS 2017). (December), World Bank, Washington, DC. 45. Mulkeen (2010). Benveniste, Luis, Jeffrey Marshall, and M. Caridad Araujo. 46. Bruns and Luque (2015). 2008. “Teaching in Cambodia.” Human Development 47. Bold and others (2017). Sector, East Asia and the Pacific Region, World Bank, 48. UIS (2006). Washington, DC, and Ministry of Education, Youth and 49. Bold and others (2017). Sport, Phnom Penh, Cambodia. 50. Sinha, Banerji, and Wadhwa (2016). Benveniste, Luis, Jeffrey Marshall, and Lucrecia Santibañez. 51. Chaudhury and others (2006). 2007. “Teaching in the Lao PDR.” Human Development 84 | World Development Report 2018 Sector, East Asia and the Pacific Region, World Bank, Mathematics in Africa.” Economics of Education Review 46: Washington, DC, and Ministry of Education, Vientiane, 1–22. Lao People’s Democratic Republic. di Gropello, Emanuela. 2011. Skills for the Labor Market in Indo- Bharadwaj, Prashant, Giacomo De Giorgi, David R. Hansen, nesia: Trends in Demand, Gaps, and Supply. With Aurelien and Christopher Neilson. 2015. “The Gender Gap in Kruse and Prateek Tandon. Directions in Development: Mathematics: Evidence from a Middle-Income Country.” Human Development Series. Washington, DC: World Staff Report 721 (March 2), Federal Reserve Bank of New Bank. York, New York. EQUIP2 (Educational Quality Improvement Program 2). Bloom, Nicholas, Renata Lemos, Raffaella Sadun, Daniela 2010. “Using Opportunity to Learn and Early Grade Scur, and John Van Reenen. 2014. “JEEA-FBBVA Lecture Reading Fluency to Measure School Effectiveness in 2013: The New Empirical Economics of Management.” Ethiopia, Guatemala, Honduras, and Nepal.” Working Journal of the European Economic Association 12 (4): 835–76. Paper, Educational Policy, Systems Development, and Bloom, Nicholas, Renata Lemos, Raffaella Sadun, and John Management, U.S. Agency for International Develop- Van Reenen. 2015. “Does Management Matter in ment, Washington, DC. Schools?” Economic Journal 125 (584): 647–74. ETS (Educational Testing Service). 2014. “A Guide to Under- Bold, Tessa, Deon Filmer, Gayle Martin, Ezequiel Molina, standing the Literacy Assessment of the STEP Skills Brian Stacy, Christophe Rockmore, Jakob Svensson, et al. Measurement Surveys.” ETS, Princeton, NJ. 2017. “What Do Teachers Know and Do? Does It Matter? Fernald, Anne, Virginia A. Marchman, and Adriana Weis- Evidence from Primary Schools in Africa.” Policy leder. 2013. “SES Differences in Language Processing Research Working Paper 7956, World Bank, Washing- Skill and Vocabulary Are Evident at 18 Months.” Develop- ton, DC. mental Science 16 (2): 234–48. Bruns, Barbara, Deon Filmer, and Harry Anthony Patrinos. Fernald, Lia C. H., Patricia Kariger, Melissa Hidrobo, and Paul 2011. Making Schools Work: New Evidence on Accountability J. Gertler. 2012. “Socioeconomic Gradients in Child Devel- Reforms. Human Development Perspectives Series. opment in Very Young Children: Evidence from India, Washington, DC: World Bank. Indonesia, Peru, and Senegal.” Proceedings of the National Bruns, Barbara, and Javier Luque. 2015. Great Teachers: How to Academy of Sciences 109 (Supplement 2): 17273–80. Raise Student Learning in Latin America and the Caribbean. Filmer, Deon, Amer Hasan, and Lant Pritchett. 2006. “A Mil- With Soledad De Gregorio, David K. Evans, Marco lennium Learning Goal: Measuring Real Progress in Fernández, Martin Moreno, Jessica Rodriguez, Guill- Education.” Working Paper 97 (August), Center for Global ermo Toral, and Noah Yarrow. Latin American Develop- Development, Washington, DC. ment Forum Series. Washington, DC: World Bank. Filmer, Deon, and Lant Pritchett. 1999. “The Effect of House- Chaudhury, Nazmul, Jeffrey Hammer, Michael R. Kremer, hold Wealth on Educational Attainment: Evidence from Karthik Muralidharan, and F. Halsey Rogers. 2006. 35 Countries.” Population and Development Review 25 (1): “Missing in Action: Teacher and Health Worker Absence 85–120. in Developing Countries.” Journal of Economic Perspectives Fouarge, Didier, Trudie Schils, and Andries de Grip. 2013. 20 (1): 91–116. Cornwell, Christopher, David B. Mustard, and Jessica Van “Why Do Low-Educated Workers Invest Less in Further Parys. 2013. “Noncognitive Skills and the Gender Dispar- Training?” Applied Economics 45 (18): 2587–601. ities in Test Scores and Teacher Assessments: Evidence Fryer, Roland G., Jr. 2017. “Management and Student from Primary School.” Journal of Human Resources 48 (1): Achievement: Evidence from a Randomized Field Exper- 236–64. iment.” NBER Working Paper 23437, National Bureau of Crouch, Luis, and Amber K. Gove. 2011. “Leaps or One Step at Economic Research, Cambridge, MA. a Time: Skirting or Helping Engage the Debate? The Case Fryer, Roland G., Jr., and Steven D. Levitt. 2010. “An Empirical of Reading.” In Policy Debates in Comparative, International, Analysis of the Gender Gap in Mathematics.” American and Development Education, edited by John N. Hawkins Economic Journal: Applied Economics 2 (2): 210–40. and W. James Jacob, 155–74. New York: Springer. Galasso, Emanuela, Ann Weber, and Lia C. H. Fernald. 2017. Dang, Hai-Anh H., and Paul W. Glewwe. 2017. “Well Begun, “Dynamics of Child Development: Analysis of a Longitu- but Aiming Higher: A Review of Vietnam’s Education dinal Cohort in a Very Low Income Country.” Policy Trends in the Past 20 Years and Emerging Challenges.” Research Working Paper 7973, World Bank, Washington, Policy Research Working Paper 8112, World Bank, Wash- DC. ington, DC. Glewwe, Paul W., Michael R. Kremer, and Sylvie Moulin. Das, Jishnu, Priyanka Pandey, and Tristan Zajonc. 2006. 2009. “Many Children Left Behind? Textbooks and Test “Learning Levels and Gaps in Pakistan.” Policy Research Scores in Kenya.” American Economic Journal: Applied Working Paper 4067, World Bank, Washington, DC. Economics 1 (1): 112–35. Desjardins, Richard, and Kjell Rubenson. 2011. “An Analysis Glick, Peter, and David E. Sahn. 2010. “Early Academic Perfor- of Skill Mismatch Using Direct Measures of Skills.” mance, Grade Repetition, and School Attainment in OECD Education Working Paper 63, Organisation for Senegal: A Panel Data Analysis.” World Bank Economic Economic Co-operation and Development, Paris. Review 24 (1): 93–120. Dickerson, Andy, Steven McIntosh, and Christine Valente. Gove, Amber, and Peter Cvelich. 2011. “Early Reading, Ignit- 2015. “Do the Maths: An Analysis of the Gender Gap in ing Education for All: A Report by the Early Grade The many faces of the learning crisis | 85 Learning Community of Practice.” Rev. ed. Research Tri- —— — —. 2015. “Education Management Information System.” angle Institute, Research Triangle Park, NC. Report, MoEST, Lilongwe, Malawi. Hanushek, Eric A. 1979. “Conceptual and Empirical Issues in Mulkeen, Aidan G. 2010. Teachers in Anglophone Africa: Issues the Estimation of Educational Production Functions.” in Teacher Supply, Training, and Management. Development Journal of Human Resources 14 (3): 351–88. Practice in Education Series. Washington, DC: World ————. 1992. “The Trade-Off between Child Quantity and Bank. Quality.” Journal of Political Economy 100 (1): 84–117. Mullis, Ina V. S., Michael O. Martin, and Tom Loveless. 2016. Heckman, James J. 2000. “Policies to Foster Human Capital.” “20 Years of TIMSS: International Trends in Mathemat- Research in Economics 54 (1): 3–56. ics and Science Achievement, Curriculum, and Instruc- Hungi, Njora. 2010. “What Are the Levels and Trends in tion.” International Association for the Evaluation of Grade Repetition?” SACMEQ Policy Issues 5 (September), Educational Achievement, TIMSS and PIRLS Interna- Southern and Eastern Africa Consortium for Monitoring tional Study Center, Lynch School of Education, Boston Educational Quality, Paris. College, Chestnut Hill, MA. Hungi, Njora, Demus Makuwa, Kenneth Ross, Mioko Saito, Muralidharan, Karthik, Jishnu Das, Alaka Holla, and Aakash Stephanie Dolata, Frank van Cappelle, Laura Paviot, et al. Mohpal. 2017. “The Fiscal Cost of Weak Governance: Evi- 2010. “SACMEQ III Project Results: Pupil Achievement dence from Teacher Absence in India.” Journal of Public Levels in Reading and Mathematics.” Working Docu- Economics 145: 116–35. ment Number 1, Southern and Eastern Africa Consor- Muralidharan, Karthik, and Yendrick Zieleniak. 2013. “Mea- tium for Monitoring Educational Quality, Paris. suring Learning Trajectories in Developing Countries Kaffenberger, Michelle, and Lant Pritchett. 2017. “More with Longitudinal Data and Item Response Theory.” School or More Learning? Evidence from Learning Pro- Paper presented at the Young Lives Conference, Oxford files from the Financial Inclusion Insights Data.” Back- University, Oxford, U.K., July 8–9. ground paper, World Bank, Washington, DC. Najjumba, Innocent Mulindwa, James Habyarimana, and King, Elizabeth M., Berk Özler, and Laura B. Rawlings. 1999. Charles Lwanga Bunjo. 2013. School-Based Management: “Nicaragua’s School Autonomy Reform: Fact or Fiction?” Policy and Functionality. Vol. 3 of Improving Learning in Working Paper 19, Impact Evaluation of Education Uganda. World Bank Study Series. Washington, DC: Reforms, World Bank, Washington, DC. World Bank. Larson, Bradley, and Alexandria Valerio. 2017. “Estimating Noble, Kimberly G., M. Frank Norman, and Martha J. Farah. the Stock of Skills around the World: A Technical Note.” 2005. “Neurocognitive Correlates of Socioeconomic Sta- World Bank, Washington, DC. tus in Kindergarten Children.” Developmental Science 8 (1): Lee, Ju-Ho, and Song-Chang Hong. 2016. “Accumulating 74–87. Human Capital for Sustainable Development in Korea.” O’Connell, Philip J., and Jean-Marie Jungblut. 2008. “What Paper presented at the Korea Development Institute’s Do We Know about Training at Work?” In Skill Formation: International Conference on More and Better Invest- Interdisciplinary and Cross-National Perspectives, edited by ment in Global Education, Seoul, June 14. Karl Ulrich Mayer and Heike Solga, 109–25. New York: Lemos, Renata, and Daniela Scur. 2016. “Developing Man- Cambridge University Press. agement: An Expanded Evaluation Tool for Developing OECD (Organisation for Economic Co-operation and Devel- Countries.” RISE Working Paper 16/007, Research on opment). 2010. PISA 2009 Results, What Students Know and Improving Systems of Education, Blavatnik School of Can Do: Student Performance in Reading, Mathematics, and Government, Oxford University, Oxford, U.K. Science. Vol. 1. Paris: OECD. Lipina, Sebastián J., María I. Martelli, Beatriz Vuelta, and —— — —. 2015. The ABC of Gender Equality in Education: Aptitude, Jorge A. Colombo. 2005. “Performance on the A-Not-B Behaviour, Confidence. Paris: OECD. Task of Argentinean Infants from Unsatisfied and Satis- —— — —. 2016a. PISA 2015 Results: Excellence and Equity in Educa- fied Basic Needs Homes.” Interamerican Journal of Psychol- tion. Vol. 1. Paris: OECD. ogy 39 (1): 46–60. —— — —. 2016b. Skills Matter: Further Results from the Survey of Lupien, Sonia J., Suzanne King, Michael J. Meaney, and Adult Skills. OECD Skills Studies Series. Paris: OECD. Bruce S. McEwen. 2000. “Child’s Stress Hormone Levels —— — —. 2016c. The Survey of Adult Skills: Reader’s Companion. Correlate with Mother’s Socioeconomic Status and 2d ed. OECD Skills Studies Series. Paris: OECD. Depressive State.” Biological Psychiatry 48 (10): 976–80. Orazem, Peter F., Paul W. Glewwe, and Harry Anthony McCoy, Dana Charles, Evan D. Peet, Majid Ezzati, Goodarz Patrinos. 2007. “The Benefits and Costs of Alternative Danaei, Maureen M. Black, Christopher R. Sudfeld, Strategies to Improve Educational Outcomes.” Depart- Wafaie Fawzi, et al. 2016. “Early Childhood Developmental ment of Economics Working Paper 07028 (November), Status in Low- and Middle-Income Countries: National, Iowa State University, Ames. Regional, and Global Prevalence Estimates Using Pre­ PASEC (Programme d’Analyse des Systèmes Éducatifs de la dictive Modeling.” PLOS Medicine 13 (6): e1002034. Confemen). 2015. PASEC 2014: Education System Performance Millot, Benoît, and Julia Lane. 2002. “The Efficient Use of in Francophone Africa, Competencies and Learning Factors in Time in Education.” Education Economics 10 (2): 209–28. Primary Education. Dakar, Senegal: PASEC. MoEST (Malawi, Ministry of Education, Science, and Tech- Pritchett, Lant. 2013. The Rebirth of Education: Schooling Ain’t nology). 2008. “Education Management Information Learning. Washington, DC: Center for Global Develop- System.” Report, MoEST, Lilongwe, Malawi. ment; Baltimore: Brookings Institution Press. 86 | World Development Report 2018 Pritchett, Lant, and Amanda Beatty. 2012. “The Negative Con- UIS (UNESCO Institute for Statistics). 2006. Teachers and sequences of Overambitious Curricula in Developing Educational Quality: Monitoring Global Needs for 2015. Countries.” CGD Working Paper 293, Center for Global Montreal: UIS. Development, Washington, DC. ——— —. 2017. Education (database). Montreal. http://data.uis Pritchett, Lant, and Justin Sandefur. 2017. “Girls’ Schooling .unesco.org/. and Women’s Literacy: Schooling Targets Alone Won’t UNESCO (United Nations Educational, Scientific, and Cul- Reach Learning Goals.” CGD Policy Paper 104, Center for tural Organization). 2013. Third Regional Comparative Global Development, Washington, DC. and Explanatory Study (TERCE). UNESCO Regional Robinson, Viviane M. J., Claire A. Lloyd, and Kenneth J. Bureau for Education in Latin America and the Carib- Rowe. 2008. “The Impact of Leadership on Student Out- bean, Santiago, Chile. http://www.unesco.org/new/en comes: An Analysis of the Differential Effects of Leader- /santiago/education/education-assessment-llece/third ship Types.” Educational Administration Quarterly 44 (5): -regional-comparative-and-explanatory-study-terce/. 635–74. ——— —. 2014. EFA Global Monitoring Report 2013/4, Teaching and Rockoff, Jonah E. 2004. “The Impact of Individual Teachers Learning: Achieving Quality for All. Paris: UNESCO. on Student Achievement: Evidence from Panel Data.” ——— —. 2015. Informe de resultados, TERCE: Logros de aprendizaje. American Economic Review 94 (2): 247–52. Paris: UNESCO. Santiago, Chile: UNESCO Regional RTI International. 2015. Status of Early Grade Reading in Sub- Office for Education in Latin America and the Caribbean; Saharan Africa. Washington, DC: U.S. Agency for Interna- Paris: UNESCO. tional Development. ——— —. 2016. “Gender Inequality in Learning Achievement in Rubio-Codina, Marta, Orazio Attanasio, Costas Meghir, Primary Education: What Can TERCE Tell Us?” Third Natalia Varela, and Sally Grantham-McGregor. 2015. Regional Comparative and Explanatory Study, UNESCO “The Socioeconomic Gradient of Child Development: Regional Office for Education in Latin America and the Cross-Sectional Evidence from Children 6–42 Months in Caribbean, Santiago, Chile; UNESCO, Paris. Bogota.” Journal of Human Resources 50 (2): 464–83. Uwezo. 2014. “Are Our Children Learning? Literacy and Sabarwal, Shwetlena, and Malek Abu-Jawdeh. 2017. “Under- Numeracy in Kenya 2014.” Twaweza East Africa, Nairobi. standing Teacher Effort: Insights from Cross-Country ——— —. 2015. “Are Our Children Learning? Five Stories on the Data on Teacher Perceptions.” Background paper, World State of Education in Uganda in 2015 and Beyond.” Bank, Washington, DC. Twaweza East Africa, Kampala, Uganda. SACMEQ (Southern and Eastern Africa Consortium for van Fleet, Justin W. 2012. “Africa’s Education Crisis: In School Monitoring Educational Quality). Various years. Univer- but Not Learning.” Up Front (blog), September 17. https:// sity of Botswana, Gaborone. http://www.sacmeq.org/. www.brookings.edu/blog/up-front/2012/09/17/africas Singh, Abhijeet. 2016. “Starting Together, Growing Apart: -education-crisis-in-school-but-not-learning/. Gender Gaps in Learning from Preschool to Adult- Walker, Susan P., Theodore D. Wachs, Julie Meeks Gardner, hood in Four Developing Countries.” Paper presented Betsy Lozoff, Gail A. Wasserman, Ernesto Pollitt, Julie A. at the Association for Public Policy Analysis and Carter, et al. 2007. “Child Development: Risk Factors for Management’s International Conference, “Inequalities: Adverse Outcomes in Developing Countries.” Lancet 369 Addressing the Growing Challenge for Policymakers (9556): 145–57. Worldwide,” London School of Economics, London, Windisch, Hendrickje Catriona. 2015. “Adults with Low June 13–14. Literacy and Numeracy Skills: A Literature Review on Sinha, Shabnam, Rukmini Banerji, and Wilima Wadhwa. Policy Intervention.” OECD Education Working Paper 2016. Teacher Performance in Bihar, India: Implications for 123, Organisation for Economic Co-operation and Devel- Education. Directions in Development: Human Develop- opment, Paris. ment Series. Washington, DC: World Bank. World Bank. 2016a. “Africa Education Service Delivery in Mada- Smith, David. 2014. “Nigerian Schoolchildren Defiant in City gascar: Results of 2016 Service Delivery Indicator Survey.” That Defied Boko Haram.” Guardian (May 17). https:// Washington, DC: World Bank. www.theguardian.com/world/2014/may/18/nigeria-kano ——— —. 2016b. “Education Sector Public Expenditure Tracking and -schoolchildren-boko-haram. Service Delivery Survey in Zambia.” Washington, DC: World Spaull, Nicholas, and Janeli Kotze. 2015. “Starting Behind Bank. and Staying Behind in South Africa: The Case of Insur- ——— —. 2016c. World Development Indicators (database). mountable Learning Deficits in Mathematics.” Interna- World Bank, Washington, DC. http://data.worldbank tional Journal of Educational Development 41: 13–24. .org/data-catalog/world-development-indicators. The many faces of the learning crisis | 87 SPOTLIGHT 2 Poverty hinders biological development and undermines learning Life outcomes are hugely influenced by a child’s systems direct resources toward survival rather than development during the early years. Biological sys- promoting growth—physical or mental. For example, tems develop sequentially and cumulatively, so what one in four children worldwide is stunted due to happens early in life lays the foundation for future chronic malnutrition.3 Stunting between gestation and development. Between the time of gestation and a a child’s second birthday is associated with late school child’s sixth birthday, the brain matures faster than enrollment, lower cognition, poorer executive func- at any other time of life. This period is also when the tion, and less school attainment.4 Some catch-up is brain, along with its supporting systems, is most mal- possible after a child’s second birthday, but previously leable. This malleability is a double-edged sword: high stunted bodies remain highly sensitive to disease susceptibility to early environmental influence serves and infection. Children born with low birth weight as both a window of opportunity and a source of vul- (suggesting fetal undernutrition) are at higher risk of nerability, because it means experiences can shape chronic adult diseases such as hypertension, diabetes, how development unfolds.1 The environment children obesity, and coronary artery disease (“metabolic syn- grow up in is a key determinant of their developmen- drome”). The extent to which the associated earlier tal trajectories toward outcomes later in life. cognitive impairment can be reversed is uncertain. Growing up in poverty usually exposes children Exposure to multiple risk factors without the buff- to many risk factors. In poor households, low levels ering support of available, well-informed, responsive of parental education exacerbate material deprivation caregivers can cause toxic stress.5 Stress triggers the by undermining investment choices for children’s flight-or-fight response, an intense physiological development, in terms of how parents use both their reaction that puts the body in a state of alertness to financial resources and their time. Moreover, parents’ deal with potential threats. Continual activation of the limited mental bandwidth, as well as the psycholog- flight-or-fight response in early life endangers develop- ical stress imposed by poverty (including working ing systems, because the brain focuses on addressing many hours in often precarious conditions to make the perceived danger to the detriment of further devel- ends meet), further undermine the time, energy, opment of biological systems not essential for survival. and care they can give their children.2 For the child, Toxic stress in the early years can undermine this often results in poor physical inputs starting in lifelong health, learning, and behavior. Hormones the womb, such as insufficient nutrition or extreme associated with the flight-or-fight response, such as deprivation. It also results in poor social inputs, such cortisol, can inhibit physical growth as well as weaken as insufficient stimulation (not being held, responded immune systems and metabolic regulatory mecha- to, talked to, or played with), neglect, abuse, exposure nisms, all of which permanently increase an individ- to violence, displacement, or maternal depression. ual’s susceptibility to illness.6 Moreover, toxic stress Acute adversity during the early years becomes during the early years can impair the development embedded in children’s bodies. In the face of depri- of neural connections in parts of the brain that are vation, disease, or noxious environments, developing critical for learning—such as those associated with 88 | World Development Report 2018 Figure S2.1 Severe deprivation affects socioemotional outcomes and executive functions, brain structure and function from early the biological foundations of learning.7 in life Severe deprivation, along with the associated Total white and gray matter in infants, by stunting status stress, can impair healthy brain development.8 Neuro- imaging data from studies in Bangladesh, The Gam- 700 bia, Romania, the United Kingdom, and the United p = 0.00004 States reveal differences in brain development (both 600 structural and functional) linked to socioeconomic 500 status. The studies confirm reduced brain connec- Volume (milliliters) tivity as well as smaller brain volumes in areas asso- 400 ciated with language, memory, executive function, p = 0.0001 p = 0.0016 and decision-making skills, on the one hand (figure 300 S2.1),9 and high activation of regions associated with 200 emotional reactivity, on the other.10 Such connectivity p = 0.0090 patterns and associated biological maladaptations are 100 very difficult to reverse. These biologically embedded responses lead 0 Total Gray White Cerebral to worse developmental trajectories and impaired matter matter spinal fluid learning, hurting foundational skills from the earliest Never stunted: n = 18 (9 boys, 9 girls); stages of life. Because early childhood development age 73 days ± 8 days outcomes are interdependent (see spotlight 1), subpar Stunted growth: n = 9 (6 boys, 3 girls); development in any one dimension is likely to affect age 82 days ± 11 days the others. Children with stunted bodies and brains Source: WDR 2018 team, using data from Nelson and others (2017). Data at attempting to compensate for developmental gaps http://bit.do/WDR2018-Fig_S2-1. face daunting odds as they start formal schooling Note: Data obtained from infants 2–3 months old in Dhaka, Bangladesh, using magnetic resonance imaging (MRI). Graph depicts two groups of infants: 18 because of the sequential nature of development, not stunted (not malnourished) and 9 stunted (malnourished). Graph shows coupled with the sharp decrease in brain malleability (from left to right) total amount of brain volume; total amount of gray matter, where most neural computations are performed; total amount of white after a child’s sixth birthday. Investments in early matter, which transmits electrical signals between gray matter and affects childhood development enable the normal, timely brain function and learning (that is, the information pathways of the brain); and cerebral spinal fluid, which protects the brain and spinal cord from injury development of biological systems, shaping children’s and infection and is generally involved in many aspects of brain health. long-term ability to learn (figure S2.2). Well-designed Figure S2.2 Risk and protective factors affect developmental trajectories Behavioral Reduction in risk factors, increase competence in protective factors, or intervention trajectories during a sensitive period Protective factors > risk factors Risk factors > protective factors Brain Infancy Early childhood Adolescence Adulthood function Optimum Recovery Below potential Source: Walker and others (2011). Poverty hinders biological development and undermines learning | 89 early childhood interventions that increase poor children’s normal, timely biological development, children’s access to protective factors (nutrition, stim- thereby strengthening their long-term ability to learn ulation, care, protection from stress) can enable those (see chapter 5). Notes 1. Knudsen (2004). McCoy, Dana Charles, and C. Cybele Raver. 2014. “Household 2. Mullainathan and Shafir (2013). Instability and Self-Regulation among Poor Children.” 3. UNICEF, WHO, and World Bank (2016). Stunting is Journal of Children and Poverty 20 (2): 131–52. defined as a height-for-age z-score of less than two stan- McEwen, Bruce S. 2007. “Physiology and Neurobiology of dard deviations below the median of a healthy reference Stress and Adaptation: Central Role of the Brain.” Physio- population. logical Reviews 87 (3): 873–904. 4. Black and others (2013); Christian and others (2014). Mullainathan, Sendhil, and Eldar Shafir. 2013. Scarcity: Why 5. Center on the Developing Child (2016). Having Too Little Means So Much. New York: Macmillan. 6. McEwen (2007). Nelson, Charles A., Nadine Gaab, Yingying Wang, Swapna 7. Evans and Kim (2013); McCoy and Raver (2014). Kumar, Danielle Sliva, Meaghan Mauer, Alissa Wester- 8. Center on the Developing Child (2016). lund, et al. 2017. “Atypical Brain Development in Bangla- 9. Bright Project (http:/www.globalfnirs.org/the-bright deshi Infants Exposed to Profound Early Adversity.” -project); Nelson and others (2017); Noble and others Paper presented at Society for Research in Child Devel- (2015); Vanderwert and others (2010). opment Biennial Meeting, Austin, TX, April. 10. Pavlakis and others (2015). Noble, Kimberly G., Suzanne M. Houston, Natalie H. Brito, Hauke Bartsch, Eric Kan, Joshua M. Kuperman, Natacha Akshoomoff, et al. 2015. “Family Income, Parental Educa- References tion, and Brain Structure in Children and Adolescents.” Black, Robert E., Cesar G. Victora, Susan P. Walker, Zulfiqar Nature Neuroscience 18 (5): 773–78. A. Bhutta, Parul Christian, Mercedes de Onis, Majid Pavlakis, Alexandra E., Kimberly Noble, Steven G. Pavlakis, Ezzati, et al. 2013. “Maternal and Child Undernutrition Noorjahan Ali, and Yitzchak Frank. 2015. “Brain Imaging and Overweight in Low-Income and Middle-Income and Electrophysiology Biomarkers: Is There a Role in Countries.” Lancet 382 (9890): 427–51. Poverty and Education Outcome Research?” Pediatric Center on the Developing Child. 2016. “From Best Practices Neurology 52 (4): 383–88. to Breakthrough Impacts: A Science-Based Approach to UNICEF (United Nations Children’s Fund), WHO (World Building a More Promising Future for Young Children Health Organization), and World Bank. 2016. “Levels and and Families.” Center on the Developing Child, Harvard Trends in Child Malnutrition: UNICEF/WHO/World University, Cambridge, MA. Bank Group Joint Child Malnutrition Estimates, Key Christian, Parul, Laura E. Murray-Kolb, James M. Tielsch, Findings of the 2016 Edition.” UNICEF, New York; WHO, Joanne Katz, Steven C. LeClerq, and Subarna K. Khatry. Geneva; World Bank, Washington, DC. http://www.who 2014. “Associations between Preterm Birth, Small-for- .int/nutgrowthdb/estimates2015/en/. Gestational Age, and Neonatal Morbidity and Cognitive Vanderwert, Ross E., Peter J. Marshall, Charles A. Nelson III, Function among School-Age Children in Nepal.” BMC Charles H. Zeanah, and Nathan A. Fox. 2010. “Timing of Pediatrics 14 (1): 1–15. Intervention Affects Brain Electrical Activity in Children Evans, Gary W., and Pilyoung Kim. 2013. “Childhood Poverty, Exposed to Severe Psychosocial Neglect.” PLoS One 5 (7): Chronic Stress, Self-Regulation, and Coping.” Child Devel- e11415. opment Perspectives 7 (1): 43–48. Walker, Susan P., Susan M. Chang, Marcos Vera-Hernández, Knudsen, Eric I. 2004. “Sensitive Periods in the Development and Sally M. Grantham-McGregor. 2011. “Early Child- of the Brain and Behavior.” Journal of Cognitive Neuro­ hood Stimulation Benefits Adult Competence and science 16 (8): 1412–25. Reduces Violent Behavior.” Pediatrics 127 (5): 849–57. 90 | World Development Report 2018 To take learning seriously, start by measuring it 4 The results tell us about the fact of the drop in education quality, after releasing “ the 2015 results of the Trends in International Mathematics and Science Study. . . . We cannot ignore what is happening to our education, and we cannot afford the repercussions of not reforming it.” QUEEN RANIA OF JORDAN, FACEBOOK POST, DECEMBER 2016 Why does the learning crisis persist? How can chil- official education management data, it is missing dren attend school for years but remain functionally from the agendas of politicians and bureaucrats. This illiterate? Why don’t the people in education systems is evident in how politicians often talk about edu- fix this? One big reason is that, for many, the learning cation only in terms of inputs—number of schools, crisis is invisible. Education systems have little sys- number of teachers, teacher salaries, school grants— tematic information on who is learning and who is but rarely in terms of actual learning. Lack of data not. As a result, it is impossible to generate an impetus on learning means that governments can ignore or for action—let alone a plan. obscure the poor quality of education, especially for To tackle the crisis, it is necessary—though not disadvantaged groups. enough—to measure learning. But learning metrics Without objective information on learning, par- must facilitate action, be adapted to country needs, ents may be unaware of the poor quality of education. and consist of a range of tools to meet the needs of the This prevents them from demanding better services system, including at the classroom level. from schools and governments. In Kenya, one study found that less than half of the children in grade 4 could pass basic proficiency tests in literacy or The learning crisis is often numeracy, yet more than two-thirds of adults were hidden—but measurement broadly satisfied with the government’s performance in education.1 The realization that learning outcomes makes it visible are poor may come only when children face poor labor “Almost no low-income countries have market prospects, but by then it is too late. If parents standardized (equated over time) national assessment have no real information on how much (or little) their systems to track learning and provide a feedback children are learning, how can they hold schools or mechanism to national education policies and governments accountable? programs” (Birdsall, Bruns, and Madan 2016, 2). Without clear information on what students do not know, how can schools improve instruction? Education systems routinely report on enrollment— Teachers may find it hard to judge to what extent but not on learning. Because learning is missing from students understand what is being taught. This is To take learning seriously, start by measuring it | 91 particularly true in low-income countries, where complements of, not substitutes for, careful, con- teachers face large classrooms that mix students of text-specific analysis to determine how to improve very different abilities. For example, a study from learning.4 Delhi, India, found that the same grade may contain students whose achievement level spans the equiva- lent of five to six grades.2 In such contexts, learning Measures for learning guide measures provide teachers with timely feedback action about which students may need additional support. Testing in Rio de Janeiro, Brazil, happens at More broadly, these measures provide school man- two levels. First, every two years all students in agement with information about which areas need grades 5 and 9 take a national test (the Prova Brasil) attention to improve instruction. If the information designed to assess public education. Second, students is shared with parents or students, it can help them are tested at the end of each two-month curriculum direct their own efforts toward improving learning. block. These tests, given by municipal education Yet concerted action is often derailed by concerns departments, aim to provide quick feedback to about the possible pitfalls of learning metrics. These teachers and principals, allowing schools and the metrics generate much debate on, for example, the broader system to provide more support to struggling outsize impacts of international assessments on students (Elwick and McAleavy 2015). local policy, the limited use of national assessments for improving classroom practice, or the potential Identifying learning gaps in the classroom is the first gaming of high-stakes testing.3 But measurement of step toward resolving them. In environments of low learning is not shorthand for international testing learning, there is often a gap between the level of stu- such as the Programme for International Student dents and the level at which classes are being taught.5 Assessment (PISA) or for the high-stakes account- This might be because teachers are unaware of stu- ability approach implemented through the U.S. No dents’ levels. Fostering a culture of classroom-based Child Left Behind policy. Instead, the term covers a assessments can address this problem. In Singapore, range of assessments, including formative classroom students are given screening tests at the start of grade assessments (box 4.1). Even in this form, measures 1, which helps teachers identify those who require of learning provide information on only some of the additional instruction to learn to read.6 skills students acquire as they develop (see spotlight Learning metrics help highlight where support 3 on multidimensionality of skills). Thus metrics are is most needed. School districts and schools are then Box 4.1 Good measures of learning illuminate all parts of the education system Formative classroom assessments facilitate instruction advanced placements in the education system or job by providing real-time feedback to support teaching and market. Because of their role in determining labor market learning. This feedback allows teachers to identify strug- outcomes, these examinations are high-stakes for students. gling students, thereby enabling them to adjust instruction They significantly affect what is taught and how, and they to meet the learning needs of different students. Classroom are critical for managing the flow of students through the assessments also generate valuable feedback for students system. and parents. International assessments benchmark student perfor- National assessments provide information on the overall mance by evaluating education systems across countries education system by highlighting achievements along with and over time using representative samples of children. challenges, such as inequalities. They are useful for educa- There has also been a steady increase in the use of citizen- tion management, policy, and reform. led assessments. These can be important for fostering National examinations certify student achievement, public awareness, showing what is possible, advocating for with a focus on transparently selecting students for more change, and informing research. Source: WDR 2018 team. 92 | World Development Report 2018 better able to target resources to improve service deliv- heard or acknowledged in a national assessment pro- ery. In Brazil, national assessments have been widely cess, they will likely reject its findings. For example, adopted by states and municipalities to strengthen teachers are more likely to resist quantitative forms school performance.7 Learning metrics have also of evaluation when metrics do not take into account guided big-banner education reforms. In Chile, PISA’s context.17 This is particularly the case for measures of reading framework guided national curriculum learning disseminated as rankings, which are suscep- reform.8 Similarly, findings from the Southern and tible to being taken out of context. In some education Eastern Africa Consortium for Monitoring Educa- systems, such friction is heightened by the use of tional Quality (SACMEQ) I, 1995–99, underpinned a technology, which raises questions about privacy review of Mauritius’s education master plan.9 In some and transparency. Approaches using technology also cases, learning metrics have been instrumental in involve limited social interaction, which is associated making education reform data-driven. In Germany, with less impact.18 lower than expected results—especially for students For measurement to guide action, it must be from poorer backgrounds—on the 2000 PISA led to the actionable. It also needs to be available to stakehold- development of more support for disadvantaged stu- ers. At the design stage, stakeholders have to ask dents, especially those from immigrant backgrounds.10 themselves how learning data will be used. In Chile, For learning metrics to guide action effectively, all students in grades 4 and 8 take the Sistema de they need to be used as a range of tools to serve Medición de la Calidad de la Educación (SIMCE) each different needs, from classroom practice to system year. After the test identifies the 900 schools scoring management. Measures of learning come in various in the lowest 10 percent on the tests in their province, forms, with different measures serving different pur- these schools receive special resources. The data, then, poses for different actors. These range from simple are clearly linked to action. Many assessment systems oral questions posed by a teacher to national assess- measure outcomes too infrequently or too broadly to ments that help policy makers prioritize action (box be of practical use. The most recent publicly available 4.1). In well-functioning systems, these different tools data from the SACMEQ are for 2007. Another con- complement one another to form a coherent whole.11 straint is the lag between when data are collected and Policy makers should rely on a broad range of when they are made available, as well as how data are information instead of any one measure. When a made available. Many ministries produce only hard single metric becomes the sole basis for big policy copies of summary reports, which make them diffi- triggers, the corresponding stakes may become cult to use. dangerously high. A striking example is the U.S. No Child Left Behind policy enacted in 2001. This policy had strong negative repercussions for schools that Measures of learning spur performed poorly on annual statewide standardized action tests. Though the policy led some poorly performing “Shock as 60 [Percent] of Tanzania Students schools to improve, it also generated various unde- Fail National Exam” (East African, 2013) sirable strategic responses by teachers and school administrators.12 These included reclassifying stu- In the United States since 2001, information on dents as requiring special education, exempting different schools’ performance on standardized certain students from testing, reallocating resources tests has notably increased turnout in local to students at the margin of passing, and suspending school board elections (Holbein 2016). low-scoring students near test dates.13 Even in the case of PISA, some studies have suggested that the Measures of learning motivate action through three performance in some places—Argentina, Malaysia, channels:19 Vietnam, Shanghai (China)—could be tied in part to (perhaps inadvertent) “selective samples” that may Participation. Learning outcomes are often far •  exclude some poorly performing schools or students.14 worse than stakeholders realize. In Uganda, nearly Education systems also routinely underuse the three-quarters of parents said they were satisfied information generated by learning metrics—making with the quality of education—yet only a quarter for a lot of measurement that leads to little action.15 of grade 4 students could pass a math test based on Often, findings are simply not communicated in a grade 2 questions.20 By documenting service deliv- timely way to relevant audiences.16 There may also ery shortfalls, learning metrics can motivate par- be credibility issues. If teachers or schools do not feel ents to hold their schools accountable for learning. To take learning seriously, start by measuring it | 93 In such contexts, learning metrics can correct infor- decreased the frequency with which the test is admin- mation failures, which are especially severe for the istered, and delayed the publication of results by two poor. This correction can in turn rebalance the rela- years to obscure the poor performance of students.28 tionship between users and providers. This channel Teachers, too, might resist learning assessments to operates via the direct or short route of accountabil- minimize opportunities for blame.29 In Chile, teacher ity running from parents directly to schools. training institutions have shown resistance to the Choice. Providing parents with hard evidence •  national assessment.30 Assessments are also political about learning outcomes at alternative schools can because they can affect the flow of resources or pres- encourage schools to improve learning by increas- tige in an education system—as in the United States ing competitive pressures. When parents have under the No Child Left Behind policy.31 Underlying objective information about learning outcomes politics can make student assessment systems partic- across schools, they can punish poorly performing ularly hard to reform (see part IV of this Report). schools by “voting with their feet.” Public schools When does measurement mobilize citizens to care about such outcomes because their resources demand accountability for learning? Because of lim- are often tied to the number of students they ited attention, information is often ignored, especially enroll.21 But this channel may also disproportion- if it is complex or provides unwelcome news.32 There- ately penalize schools that serve poor children. fore, for measurement to spur action, information Voice. Learning metrics can facilitate lobbying for •  must be available in an easily digestible way. But this reform by providing information on what needs in itself may not be enough. Learning metrics can gal- fixing. Lack of reliable metrics, by contrast, under- vanize communities to hold their schools accountable mines accountability for results.22 This channel for learning only when collective action problems are operates via the long route of accountability, where resolved.33 A participatory approach—where schools learning metrics may help citizens use the political and communities have a say in what type of “learning process to hold politicians accountable for learning. metrics” are generated at the school level—may be likely to work better here.34 In addition, for citizens That said, the links from measurement to action to be able to act on information, fear of reprisals must are neither automatic nor straightforward. India’s be low. Finally, for citizens to act in behalf of change, citizen-led assessment, the Annual Status of Educa- they must believe that their own individual actions tion Report (ASER), has documented low proficiency can make a difference.35 scores since it was introduced in 2004. However, clear Efforts to benchmark country performance or sustained improvements are not yet visible for the through international or regional assessments have country as a whole.23 At the same time, some Indian in some cases galvanized action because interna- states have shown significant improvements in grade tional comparisons make learning politically salient. 3 reading levels between ASER 2010 and 2016.24 This Release of the Trends in International Mathematics shows that it is not just the information but action and Science Study (TIMSS) or PISA rankings often that matters. For learning to improve, not only do triggers intense media interest, inserting learning learning assessments need to be available, but also into political and economic debates.36 This increase in someone needs to act on them. In fact, an evaluation interest often generates momentum for government of the impact of citizen-led assessments in Kenya action—an effect known as “PISA shock”—thereby finds that for information on learning to spur action, unleashing targeted reforms. About half the countries those who receive the information must understand participating in the PISA assessments under the aegis it, see it as actionable, care about the topic, and believe of the Organisation for Economic Co-operation and that their actions will improve outcomes.25 Development (OECD) have launched reforms because Political pressures may limit the extent to which of the results.37 Learning assessments also spur measures of learning spur positive action. Where action by making learning a tangible goal. Whereas education quality is low, politicians have an incentive the United Nations’ Millennium Development Goals to hide or obscure learning outcomes.26 They may also (MDGs), which inspired efforts by governments and try to evade blame for poor performance by setting donors, focused on enrollment, the current Sustain- low standards, trying to limit year-to-year compara- able Development Goals (SDGs) place greater empha- bility, or restricting access to outcome information.27 sis on learning.38 The success of the SDGs will depend For example, Argentina amended its standardized on countries’ ability to turn rhetoric into action by test so that year-to-year comparisons are not possible, tracking learning. 94 | World Development Report 2018 Choose learning metrics half of the countries surveyed produce data or partic- ipate in any regional or international tests to assess based on what the country mathematics at the end of lower secondary school. needs Just under half assess reading. This means that comparable information about learning is missing When choosing which measures of learning to invest for most children and youth outside of high-income in, policy makers must consider the context. If assess- countries (figure 4.1).40 ment systems are nascent, priority should be given to fostering classroom assessment. Once that piece is in place, countries can develop relatively quick, Will learning metrics narrow sample-based, low-cost national assessments. When classroom and national assessments are established, the vision for education? much can be gained from participating in regional or Putting emphasis on measurable learning does not global assessments that enable performance bench- mean ignoring other outcomes of education, such as marking. The ultimate goal is to build assessment physical, moral, civic, or artistic development. Indeed, systems in which different parts are aligned but serve focusing on learning—and on the educational quality different needs. that drives it—is more likely to crowd in these other Not every student needs to be tested in national desirable outcomes. Conditions that allow children to assessments. Sample-based assessments can accu- spend two or three years in school without learning rately measure a system’s performance. These to read a single word or to reach the end of primary assessments still require capable administrators, school without learning two-digit subtraction are but they are much less expensive than census-based not conducive to reaching the higher goals of edu- assessments. They can also be administered more cation. An experiment in Andhra Pradesh, India, that often. Schools participating in these assessments do rewarded teachers for gains in measured learning in not have to be identified. This helps lower the stakes, math and language led to improved outcomes not just making the assessments less susceptible to perverse in those subjects, but also in science and social stud- responses by teachers or schools. ies—even though there were no rewards for improve- Assessment systems should test students at an age ment in the latter two subjects.41 A study of ninth when effective remedial action remains possible. Of graders in the United States found that behavioral 121 countries in four regions, a third lack any report- factors correlate positively with test scores.42 Another ing data on the reading and mathematics proficiency U.S. study revealed that teachers who improve test levels of children at the end of primary school.39 Only scores also improve broad outcomes into adulthood.43 Figure 4.1 No internationally comparable data on learning are available for most children outside of high-income countries Percentage of children in countries that have reported mathematics and reading scores since 2000 for ASER, EGRA, LLECE, PASEC, PIRLS, PISA, SACMEQ, and TIMSS, by income group a. Grades 2–3 b. Grades 4–6 c. Grades 8–10 100 100 100 80 80 80 60 60 60 Percent Percent Percent 40 40 40 20 20 20 0 0 0 Low- Middle- High- World Low- Middle- High- World Low- Middle- High- World income income income income income income income income income Type of exam: Standardized regional Nonstandardized regional International None (LLECE, PASEC, SACMEQ) (ASER, EGRA) (PIRLS, PISA, TIMSS) Source: WDR 2018 team, using data from Sandefur (2017). Data at http://bit.do/WDR2018-Fig_4-1. Note: ASER = Annual Status of Education Report; EGRA = Early Grade Reading Assessment; LLECE = Latin American Laboratory for Assessment of the Quality of Education; PASEC = Programme d’Analyse des Systèmes Éducatifs de la Confemen; PIRLS = Progress in International Reading Literacy Study; PISA = Programme for International Student Assessment; SACMEQ = Southern and Eastern Africa Consortium for Monitoring Educational Quality; TIMSS = Trends in International Mathematics and Science Study. To take learning seriously, start by measuring it | 95 Learning assessments of key foundational subjects to understand how producers—or in this case, coun- such as language and mathematics are likely to be tries—make trade-offs between production of differ- good proxies for whether an education system is ent goods (figure 4.2). For example, in recent years delivering on its broad promise. many stakeholders in the Republic of Korea have That said, cognitive skills are not the only skills argued that their high-performing education system that matter. Socioemotional skills (sometimes called places too much emphasis on test scores (shown in noncognitive skills) such as grit, self-control, self- figure 4.2 as “measured learning”) and not enough management, effective communication, and pro- on creativity or certain socioemotional skills such as social behavior can be central to not just economic teamwork (“other outputs”). Implicitly, this Korean outcomes but life outcomes more broadly.44 Evidence debate is about whether to try to move up and to from high-income countries suggests that such skills the left on the frontier—that is, from A toward B. strongly affect employment status, work experience, But in the low-learning trap, represented by “low- occupational choice, and wages.45 They also reduce performing country C” in the figure, there is so much risky behaviors such as crime, violence, or drug slack that this OECD-driven debate is not relevant. use.46 For example, a study from the United Kingdom Country C has an opportunity to improve on both found that even after controlling for cognitive skills, measured learning and other education outputs at socioemotional skills were important for predicting the same time. whether individuals stayed in school, obtained a degree, were employed, smoked, or were involved in Six tips for effective learning a crime.47 An understanding of how to measure these skills, along with how to influence them, is growing measurement rapidly. Like cognitive skills, socioemotional skills 48 Tip 1: Measure gaps. The learning crisis will be truly develop early in life but are malleable.49 In fact, socio- salient politically only when vulnerable subpopula- emotional skills help build cognitive skills and vice tions, who are disproportionately likely to suffer from versa, with current skill levels dependent on invest- learning gaps, are adequately covered by national ments made earlier in life (see spotlight 3).50 assessment systems. To ensure that happens, assess- Lower-performing countries probably do not face ments should be deployed in a way that shines a light ming countries don’t face sharp tradeo s between learning and other education outputs the same sharp trade-offs faced by high-performing on all children. Measurement must allow for the countries on the education frontier. Economists use disaggregation of data around important dimensions the concept of the production possibilities frontier such as socioeconomic status, gender, location, or dis- ability status. In particular, groups at risk for social or economic exclusion may need to be oversampled to Figure 4.2 Low-performing countries don’t face ensure adequate representation.51 sharp trade-offs between learning and other Tip 2: Track progress. The use of uniform method- education outputs ologies, approaches, and psychometrics across years is crucial for education systems to discern trends in learning over time and changes in learning gaps across tests. Year-on-year comparisons of learning progress should also be ensured for vulnerable subpopulations. (e.g., creativity, citizenship) High-performing country B Tip 3: Test students when effective action is still possible. Other outputs Returns from student assessments will be maximized if they focus on ensuring that students attain basic skills—literacy, numeracy, critical thinking—early in their schooling. Systems should also consider High-performing country A household-based testing, which would allow assess- ments to cover students not currently in school, Low-performing making the resulting measurement more useful for country C universal learning targets. Household testing would also allow more nuanced understanding of all the Measured learning different influences on a child’s school access and learning outcomes. To that end, standardized learn- Source: WDR 2018 team. ing modules can be included at little additional cost 96 | World Development Report 2018 in surveys conducted both nationally (such as income of understandable results to key stakeholders. Another and consumption surveys) and internationally (such factor is an open, collaborative process for instrument as Living Standards Measurement Study surveys or design. Student assessments developed with the col- Demographic and Health Surveys). laboration of various stakeholders are more likely to Tip 4: Balance the stakes. No single measure should be considered valid and relevant at local levels. be misused or overused. One way to avoid that out- Tip 6: Exploit global public goods on learning. Lever- come is to frame learning measures that guide policy aging international assessments can yield high as low-stakes diagnostic tools—not as one summary returns. For example, there is considerable advantage number that determines sanctions and rewards. to forging common links between international and Again, “learning metrics” should be considered a sys- regional assessments so they can be put on the same tem of tools, each with its own place and purpose.52 scale. This not only increases harmonization between Tip 5: Good design is not enough—facilitate action. international assessments such as PISA and TIMSS, Learning measures should be used explicitly not but also allows ties to national and citizen-led assess- just for tracking progress, but also for policy mak- ments, enabling meaningful global tracking (box 4.2). ing.53 One way to ensure that happens is to devote Researchers have tried to link various assessments resources (including effort) to the timely distribution after the fact, but these attempts have faced severe Box 4.2 A global learning metric? A global learning metric could help bring learning center But most of these problems are surmountable. Global stage, making it more salient. Such a metric would use an advocacy is generating sound technical recommendations internationally comparable scale to consistently track prog- on what a global metric could look like. Although there are ress and identify gaps across contexts. It would enable com- no agreed-on standards of proficiency and no agreed-on parisons across children, households, schools, and locations. tests to ensure that countries’ measures of learning are Beyond its technical dividends, a global metric would comparable to each other and over time, several global motivate action and generate accountability for learning. initiatives—such as the Global Alliance to Monitor Learning, By showing what is possible, it could point to what coun- the Assessment for Learning (A4L) initiative, and the tries should be aspiring to—and create pressure to meet International Commission on Financing Global Education those aspirations. By benchmarking learning gaps among Opportunity—are generating momentum. Other challenges disadvantaged groups, a global metric could also create could be overcome through clear goals and quality thresh- pressures for social mobility within countries. Furthermore, olds. A global metric can succeed only if it is explicitly comparable learning data could increase the effectiveness framed as a complement to national assessment systems— of global research, international partnerships, and global not as a substitute for them. In fact, information from the aid for learning. Such data could also help countries global metric could be used to strengthen the capacity of develop their capacity for analyzing results to drive policy. national systems. To be sure, there are technical and political challenges The political will needed for a global metric might be that would go hand in hand with adopting a global metric. easier to mobilize if the needs of developing countries The first is how to generate a global consensus on the met- are prioritized and the metric’s advantages are clearly ric’s scope. A global metric would require making choices communicated. Estimates suggest that only 3 percent of about approach, target sample, and interpretation, which official development assistance for education is spent on could prove controversial. In addition, challenges would global public goods such as data and research; for health, arise with financing, implementation capacity, and political that share is 20 percent.a Returns from investing more on will. Many developing countries lack infrastructure for data education data could be enormous if they help focus atten- collection, organization, analysis, and mechanisms to pro- tion on ensuring that students attain basic skills in their vide feedback to educators, parents, or communities. These early years. are all necessary ingredients for turning metrics into action. Source: WDR 2018 team. a. Schäferhoff and Burnett (2016). To take learning seriously, start by measuring it | 97 technical challenges.54 Ex ante linking of measure- be effective, “learning metrics” must overcome two ments through common items is likely to prove much important challenges: ensuring that information more technically sound and cost-effective. leads to action, and minimizing the potential per- verse impacts of measurement. Alarm at the rise of *** a “testing” culture has dominated recent discourse. Education systems are unlikely to tackle the learning But in most low-learning contexts there is too little crisis unless it becomes clearly visible. This is possible assessment and, consequently, too little accountabil- only through well-designed measures of learning. To ity for learning in the system. Notes 1. Pritchett, Banerji, and Kenny (2013). countries still lack measures of learning means that 2. Muralidharan, Singh, and Ganimian (2016). comparable information remains missing for many chil- 3. Eggen and Stobart (2014); Sellar and Lingard (2013). dren and youth outside of high-income countries. 4. Carnoy and others (2016). 41. Muralidharan and Sundararaman (2011). 5. Pritchett (2013). 42. Jackson (2016). 6. OECD (2011). 43. Chetty and others (2010). 7. Guimarães de Castro (2012). 44. Durlak and others (2011); Heckman, Pinto, and Savelyev 8. Breakspear (2012). (2013); Murnane and others (2001). 9. Kulpoo (1998). 45. Heckman, Stixrud, and Urzua (2006). 10. Ertl (2006). 46. Durlak, Weissberg, and Pachan (2010). 11. Greaney and Kellaghan (2008). 47. Carneiro, Crawford, and Goodman (2007). 12. Dee and Jacob (2011). 48. Carneiro, Crawford, and Goodman (2007); Heckman, 13. Booher-Jennings (2005); Cullen and Reback (2006); Figlio Pinto, and Savelyev (2013). and Getzler (2006); Jacob (2005); Jennings and Beveridge 49. Heckman, Stixrud, and Urzua (2006). (2009); Neal and Schanzenbach (2010); Reback (2008). 50. Cunha and Heckman (2007, 2008); OECD (2015). 14. Carnoy and others (2016); Glewwe and others (2017); 51. Sandefur (2016). OECD (2016); Xu and Dronkers (2016). 52. Neal (2013). 15. Székely (2011). 53. Guimarães de Castro (2012). 16. Greaney and Kellaghan (2008). 54. Altinok, Diebolt, and Demeulemeester (2014); Altinok 17. Baker and others (2010); Dixon and others (2013). and Murseli (2007); Sandefur (2017). 18. Bellamy and Raab (2005); Meijer (2009). 19. Bruns, Filmer, and Patrinos (2011). 20. Afrobarometer (2015); Uwezo (2014). References 21. World Bank (2003). 22. Pritchett, Banerji, and Kenny (2013). Afrobarometer. 2015. “Uganda, Round 6 Data (2015).” Democ- 23. R4D (2015). racy in Africa Research Unit, Center for Social Science 24. ASER Centre (2016). Research, University of Cape Town, Rondebosch, South 25. Lieberman, Posner, and Tsai (2014). Africa. http://afrobarometer.org/data/uganda-round-6 26. Michener and Ritter (2016); Tanaka (2001). -data-2015. 27. Nicolai and others (2014). Altinok, Nadir, Claude Diebolt, and Jean-Luc Demeule­ 28. Ganimian (2015). meester. 2014. “A New International Database on Educa- 29. Fox (2007); Hood (2010); Worthy (2015). tion Quality: 1965–2010.” Applied Economics 46 (11): 30. Meckes and Carrasco (2006). 1212–47. 31. Benveniste (2002); Peterson and West (2003). Altinok, Nadir, and Hatidje Murseli. 2007. “International 32. Loewenstein, Sunstein, and Golman (2014). Database on Human Capital Quality.” Economics Letters 33. Björkman and Svensson (2010). 96 (2): 237–44. 34. Barr and others (2012); Björkman and Svensson (2010). ASER Centre. 2016. “Annual Status of Education Report 35. Barr and others (2012); Lieberman, Posner, and Tsai 2016.” New Delhi. http://www.asercentre.org/p/289.html. (2014). Baker, Eva L., Paul E. Barton, Linda Darling-Hammond, 36. Breakspear (2012). Edward Haertel, Helen F. Ladd, Robert L. Linn, Diane 37. Figazzolo (2009). Ravitch, et al. 2010. “Problems with the Use of Student 38. Tawil and others (2016). Test Scores to Evaluate Teachers.” EPI Briefing Paper 39. UIS (2016). #278, Economic Policy Institute, Washington, DC. 40. By linking items across assessments—and including Barr, Abigail, Frederick Mugisha, Pieter Serneels, and national assessments—coverage might be increased. Andrew Zeitlin. 2012. “Information and Collective Action But the fact that several large low- and middle-income in Community-Based Monitoring of Schools: Field and 98 | World Development Report 2018 Lab Experimental Evidence from Uganda.” Working Dee, Thomas S., and Brian Jacob. 2011. “The Impact of No paper, Georgetown University, Washington, DC. Child Left Behind on Student Achievement.” Journal of Bellamy, Christine, and Charles Raab. 2005. “Joined-Up Gov- Policy Analysis and Management 30 (3): 418–46. ernment and Privacy in the United Kingdom: Managing Dixon, Ruth, Christiane Arndt, Manuel Mullers, Jarmo Tensions between Data Protection and Social Policy, Part Vakkuri, Kristiina Engblom-Pelkkala, and Christopher II.” Public Administration 83 (2): 393–415. Hood. 2013. “A Lever for Improvement or a Magnet for Benveniste, Luis. 2002. “The Political Structuration of Blame? Press and Political Responses to International Assessment: Negotiating State Power and Legitimacy.” Educational Rankings in Four EU Countries.” Public Comparative Education Review 46 (1): 89–118. Administration 91 (2): 484–505. Birdsall, Nancy, Barbara Bruns, and Janeen Madan. 2016. Durlak, Joseph A., Roger P. Weissberg, Allison B. Dymnicki, “Learning Data for Better Policy: A Global Agenda.” CGD Rebecca D. Taylor, and Kriston B. Schellinger. 2011. “The Policy Paper, Center for Global Development, Wash­ Impact of Enhancing Students’ Social and Emotional ington, DC. http://www.cgdev.org/sites/default/files Learning: A Meta-Analysis of School-Based Universal /learning-data-better-policy.pdf. Interventions.” Child Development 82 (1): 405–32. Björkman, Martina, and Jakob Svensson. 2010. “When Is Durlak, Joseph A., Roger P. Weissberg, and Molly Pachan. Community-Based Monitoring Effective? Evidence from 2010. “A Meta-Analysis of After-School Programs That a Randomized Experiment in Primary Health in Uganda.” Seek to Promote Personal and Social Skills in Children Journal of the European Economic Association 8 (2–3): 571–81. and Adolescents.” American Journal of Community Psychol- Booher-Jennings, Jennifer. 2005. “Below the Bubble: ‘Educa- ogy 45 (3–4): 294–309. tional Triage’ and the Texas Accountability System.” Eggen, Theo J. H. M., and Gordon Stobart, eds. 2014. High- American Educational Research Journal 42 (2): 231–68. Stakes Testing in Education: Value, Fairness, and Consequences. Breakspear, Simon. 2012. “The Policy Impact of PISA: An New York: Routledge. Exploration of the Normative Effects of International Elwick, Alex, and Tony McAleavy. 2015. Interesting Cities: Five Benchmarking in School System Performance.” OECD Approaches to Urban School Reform. Reading, U.K.: CfBT Education Working Paper 71, Organisation for Economic Education Trust. Co-operation and Development, Paris. Ertl, Hubert. 2006. “Educational Standards and the Chang- Bruns, Barbara, Deon Filmer, and Harry Anthony Patrinos. ing Discourse on Education: The Reception and Conse- 2011. Making Schools Work: New Evidence on Accountability quences of the PISA Study in Germany.” Oxford Review of Reforms. Human Development Perspectives Series. Education 32 (5): 619–34. Washington, DC: World Bank. Figazzolo, Laura. 2009. “Impact of PISA 2006 on the Educa- Carneiro, Pedro, Claire Crawford, and Alissa Goodman. 2007. tion Policy Debate.” Working paper, Education Interna- “The Impact of Early Cognitive and Non-cognitive Skills tional, Brussels. on Later Outcomes.” CEE Discussion Paper 0092, Centre Figlio, David N., and Lawrence S. Getzler. 2006. “Accountabil- for the Economics of Education, London School of ity, Ability, and Disability: Gaming the System?” In Economics. Improving School Accountability, edited by Timothy J. Carnoy, Martin, Tatiana Khavenson, Prashant Loyalka, Wil- Gronberg and Dennis W. Jansen, 35–49. Advances in liam H. Schmidt, and Andrey Zakharov. 2016. “Revisiting Applied Microeconomics Series 14. Bingley, U.K.: Emer- the Relationship between International Assessment ald Publishing Limited. Outcomes and Educational Production: Evidence from a Fox, Jonathan. 2007. “The Uncertain Relationship between Longitudinal PISA-TIMSS Sample.” American Educational Transparency and Accountability.” Development in Practice Research Journal 53 (4): 1054–85. 17 (4–5): 663–71. Chetty, Raj, John N. Friedman, Nathaniel Hilger, Emmanuel Ganimian, Alejandro J. 2015. El Termómetro Educativo: Informe Saez, Diane Whitmore Schanzenbach, and Danny Yagan. Sobre el Desempeño de Argentina en los Operativos Nacionales 2010. “How Does Your Kindergarten Classroom Affect de Evaluación (One) 2005–2013. Buenos Aires: Proyecto Edu- Your Earnings? Evidence from Project Star.” NBER Work- car 2050. ing Paper 16381, National Bureau of Economic Research, Glewwe, Paul W., Jongwook Lee, Khoa Vu, and Hai-Anh H. Cambridge, MA. Dang. 2017. “What Explains Vietnam’s Exceptional Per- Cullen, Julie Berry, and Randall Reback. 2006. “Tinkering formance in Education Relative to Other Countries? toward Accolades: School Gaming under a Performance Analysis of the 2012 PISA Data.” Paper presented at the Accountability System.” In Improving School Accountabil- RISE Annual Conference, Center for Global Develop- ity, edited by Timothy J. Gronberg and Dennis W. Jansen, ment, Washington, DC, June 15–16. 1–34. Advances in Applied Microeconomics Series 14. Greaney, Vincent, and Thomas Kellaghan. 2008. Assessing Bingley, U.K.: Emerald Publishing Limited. National Achievement Levels in Education. Vol. 1, National Cunha, Flavio, and James J. Heckman. 2007. “The Technology Assessments of Educational Achievement. Washington, DC: of Skill Formation.” IZA Discussion Paper 2550, Institute World Bank. for the Study of Labor, Bonn, Germany. http://nbn Guimarães de Castro, Maria Helena. 2012. “Developing the -resolving.de/urn:nbn:de:101:1-20080425464. Enabling Context for Student Assessment in Brazil.” —— — —. 2008. “Formulating, Identifying and Estimating the SABER Student Assessment Working Paper 7, Systems Technology of Cognitive and Noncognitive Skill Forma- Approach for Better Education Results, World Bank, tion.” Journal of Human Resources 43 (4): 738–82. Washington, DC. To take learning seriously, start by measuring it | 99 Heckman, James J., Rodrigo Pinto, and Peter Savelyev. 2013. Murnane, Richard J., John B. Willett, M. Jay Braatz, and Yves “Understanding the Mechanisms through Which an Duhaldeborde. 2001. “Do Different Dimensions of Male Influential Early Childhood Program Boosted Adult High School Students’ Skills Predict Labor Market Suc- Outcomes.” American Economic Review 103 (6): 2052–86. cess a Decade Later? Evidence from the NLSY.” Economics Heckman, James J., Jora Stixrud, and Sergio Urzua. 2006. of Education Review 20 (4): 311–20. “The Effects of Cognitive and Noncognitive Abilities on Neal, Derek. 2013. “The Consequences of Using One Assess- Labor Market Outcomes and Social Behavior.” Journal of ment System to Pursue Two Objectives.” Journal of Eco- Labor Economics 24 (3): 411–82. nomic Education 44 (4): 339–52. Holbein, John. 2016. “Left Behind? Citizen Responsiveness to Neal, Derek, and Diane Whitmore Schanzenbach. 2010. “Left Government Performance Information.” American Politi- Behind by Design: Proficiency Counts and Test-Based cal Science Review 110 (2): 353–68. Accountability.” Review of Economics and Statistics 92 (2): Hood, Christopher. 2010. The Blame Game: Spin, Bureaucracy, 263–83. and Self-Preservation in Government. Princeton, NJ: Prince- Nicolai, Susan, Leni Wild, Joseph Wales, Sébastien Hine, and ton University Press. Jakob Engel. 2014. “Unbalanced Progress: What Political Jackson, C. Kirabo. 2016. “What Do Test Scores Miss? The Dynamics Mean for Education Access and Quality.” ODI Importance of Teacher Effects on Non-test Score Out- Development Progress Working Paper 5, Overseas comes.” NBER Working Paper 22226, National Bureau of Development Institute, London. Economic Research, Cambridge, MA. OECD (Organisation for Economic Co-operation and Devel- Jacob, Brian A. 2005. “Accountability, Incentives, and Behav- opment). 2011. Strong Performers and Successful Reformers in ior: The Impact of High-Stakes Testing in the Chicago Education: Lessons from PISA for the United States. Paris: Public Schools.” Journal of Public Economics 89 (5): 761–96. OECD. Jennings, Jennifer L., and Andrew A. Beveridge. 2009. “How —— — —. 2015. Skills for Social Progress: The Power of Social and Does Test Exemption Affect Schools’ and Students’ Aca- Emotional Skills. Paris: OECD. demic Performance?” Educational Evaluation and Policy —— — —. 2016. PISA 2015 Results: Excellence and Equity in Educa- tion. Vol. 1. Paris: OECD. Analysis 31 (2): 153–75. Peterson, Paul E., and Martin R. West. 2003. No Child Left Kulpoo, Dhurumbeer. 1998. “The Quality of Education: Some Behind? The Politics and Practice of School Accountability. Policy Suggestions Based on a Survey of Schools: Mauri- Washington, DC: Brookings Institution Press. tius.” SACMEQ Policy Research Report No. 1, United Pritchett, Lant. 2013. The Rebirth of Education: Schooling Ain’t Nations Educational, Scientific, and Cultural Organiza- Learning. Washington, DC: Center for Global Develop- tion, Port Louis, Mauritius. ment; Baltimore: Brookings Institution Press. Lieberman, Evan S., Daniel N. Posner, and Lily L. Tsai. 2014. Pritchett, Lant, Rukmini Banerji, and Charles Kenny. 2013. “Does Information Lead to More Active Citizenship? “Schooling Is Not Education! Using Assessment to Evidence from an Education Intervention in Rural Change the Politics of Non-learning.” CGD Report, Kenya.” World Development 60: 69–83. Center for Global Development, Washington, DC. Loewenstein, George, Cass R. Sunstein, and Russell Golman. R4D (Results for Development Institute). 2015. “Bringing 2014. “Disclosure: Psychology Changes Everything.” Learning to Light: The Role of Citizen-Led Assessments Annual Review of Economics 6 (1): 391–419. in Shifting the Education Agenda.” R4D, Washington, DC. Meckes, Lorena, and Rafael Carrasco. 2006. “SIMCE: Lessons Reback, Randall. 2008. “Teaching to the Rating: School from the Chilean Experience in National Assessment Accountability and the Distribution of Student Achieve- Systems of Learning Outcomes.” Paper presented at ment.” Journal of Public Economics 92 (5): 1394–1415. World Bank and Inter-American Development Bank’s Sandefur, Justin. 2016. “Internationally Comparable Mathe- Conference, “Lessons from Best Practices in Promoting matics Scores for Fourteen African Countries.” CGD Education for All: Latin America and the Caribbean,” Working Paper 444 (December), Center for Global Devel- Cartagena de Indias, Colombia, October 9–11. opment, Washington, DC. Meijer, Albert. 2009. “Understanding Modern Transpar- —— — —. 2017. “The Case for Global Standardized Testing.” ency.” International Review of Administrative Sciences 75 (2): Views from the Center: Education, Education Reform 255–69. (blog), April 27. https://www.cgdev.org/blog/case-global Michener, Gregory, and Otavio Ritter. 2016. “Comparing -standardized-testing. Resistance to Open Data Performance Measurement: Schäferhoff, Marco, and Nicholas Burnett. 2016. “Rethinking Public Education in Brazil and the UK.” Public Administra- the Financing and Architecture of Global Education.” tion 95 (1): 4–21. Background Paper: The Learning Generation (April 29), Muralidharan, Karthik, Abhijeet Singh, and Alejandro J. International Commission on Financing Global Educa- Ganimian. 2016. “Disrupting Education? Experimental tion Opportunity, New York. Evidence on Technology-Aided Instruction in India.” Sellar, Sam, and Bob Lingard. 2013. “The OECD and Global NBER Working Paper 22923, National Bureau of Eco- Governance in Education.” Journal of Education Policy nomic Research, Cambridge, MA. 28 (5): 710–25. Muralidharan, Karthik, and Venkatesh Sundararaman. 2011. Székely, Miguel. 2011. “Toward Results-Based Social Policy “Teacher Performance Pay: Experimental Evidence from Design and Implementation.” CGD Working Paper 249, India.” Journal of Political Economy 119 (1): 39–77. Center for Global Development, Washington, DC. 100 | World Development Report 2018 Tanaka, Shinichiro. 2001. “Corruption in Education Sector Uwezo. 2014. “Are Our Children Learning? Literacy and Development: A Suggestion for Anticipatory Strategy.” Numeracy across East Africa 2013.” Twaweza, Nairobi. International Journal of Educational Management 15 (4): World Bank. 2003. World Development Report 2004: Making 158–66. Services Work for Poor People. Washington, DC: World Tawil, Sobhi, Margarete Sachs-Israel, Huong Le Thu, and Bank; New York: Oxford University Press. Matthias Eck. 2016. Unpacking Sustainable Development Worthy, Ben. 2015. “The Impact of Open Data in the UK: Goal 4 Education 2030—Guide. Paris: United Nations Edu- Complex, Unpredictable, and Political.” Public Administra- cational, Scientific, and Cultural Organization. tion 93 (3): 788–805. UIS (UNESCO Institute for Statistics). 2016. “Laying the Xu, Duoduo, and Jaap Dronkers. 2016. “Migrant Children Foundation to Measure Sustainable Development Goal in Shanghai: A Research Note on the PISA-Shanghai 4.” Sustainable Development Data Digest, UIS, Montreal. Controversy.” Chinese Sociological Review 48 (3): 271–95. To take learning seriously, start by measuring it | 101 SPOTLIGHT 3 The multidimensionality of skills Having knowledge is not the same as being able to reflect enduring patterns in how individuals respond apply it.1 Having a skill means having the ability to do to various situations. Socioemotional skills are trans- something well. Having a skill requires knowledge, versal skills, meaning they are relevant to a broad but having knowledge does not necessarily imply range of disciplines. They work together with cogni- having skills.2 Knowing how a wind turbine works tive skills, in that success in meeting many workplace does not mean a person has the skill to fix one. and life challenges depends on both types of skills. Technical skills are the acquired knowledge, exper- Skills are multidimensional, tise, and interactions needed by a worker for com- petent performance of the duties associated with a dynamic, and interactive specific job. Technical skills require mastery of the Promoting a breadth of skills means “educating for knowledge, materials, tools, and technologies needed a mastery of a wide range of competencies that will to do a job.8 help mitigate the challenges posed by our changing Cognitive skills and socioemotional skills reinforce world context.”3 This Report uses three broad catego- each other. Individuals with characteristics such as ries of skills (figure S3.1): drive, diligence, perseverance, or good social skills Cognitive skills refers to the “ability to understand are more likely to apply themselves to acquiring complex ideas, to adapt effectively to the environ- cognitive skills, as well as to have positive relation- ment, to learn from experience, to engage in various ships in their lives. Yet cognitive skills are distinct forms of reasoning, to overcome obstacles by taking from socioemotional skills.9 Acquiring an early solid thought.”4 Cognitive skills are needed for learning, base of both is critical because both set the course of personal and professional development, and the lifetime trajectories. Individuals with early advantages development of other types of skills. They can be bro- tend to gain more skills over their lifetimes, and it is ken down into foundational skills—which include basic difficult for others to close widening gaps over time. literacy, numeracy, critical thinking, and problem- solving—and higher-order skills such as more advanced versions of these cognitive skills and others like adap- Skills can be acquired tive learning. Different types of skills can be developed over time, Socioemotional skills are the behaviors, attitudes, depending on an individual’s neurobiological and psy- and values that a person needs to “navigate interper- chological development.10 sonal and social situations effectively,”5 as well as to Most cognitive skills are acquired during child- “deal effectively and ethically with daily tasks and hood, but they can be reinforced through young adult- challenges.”6 Self-awareness, leadership, teamwork, hood. Early childhood is an optimal period to acquire self-control, and motivation are socioemotional skills.7 foundational cognitive skills because they are a pre- Sometimes referred to as noncognitive skills, socioemo- requisite to developing further cognitive and socio- tional skills include so-called personality traits, which emotional skills during later developmental periods.11 102 | World Development Report 2018 FIGURE S3.1 Figure S3.1 Cognitive, socioemotional, and technical skills interact Socioemotional skills ________________________________ • Self-awareness • Self-management • Social awareness • Relationship skills A B Cognitive skills Technical skills ______________________________ ______________________________ • Foundational skills C • Entrepreneurial - general academic • Digital (literacy, numeracy) - general cognitive • Higher-order A: Decision making, communication, grit, self-control B: Problem-solving, organizational skills C: Mid-level technical, high-level technical Source: WDR 2018 team. Higher-order cognitive skills are regularly developed positive identity, or leadership—are better acquired in late adolescence and early adulthood, in parallel in middle childhood and during adolescence.13 Even with technical skills that are relevant for the labor though the neurobiological and psychosocial bases market.12 Given the ages that correspond to optimal are already well established at this stage, socioemo- skills development periods, foundational cognitive tional skills can also be learned well during early skills are usually learned in school and at home. adulthood through new experiences.14 Similarly, socioemotional skills can be acquired Technical skills can be learned at ages and in set- through adulthood, though the optimal period is tings that correspond to the fields of study or jobs that in early childhood, while the best stage to reinforce a person chooses. Thus these skills can be acquired them is early adulthood. But unlike cognitive skills, throughout life, in school and the workplace, as well certain socioemotional skills—such as self-esteem, as through specific training and education.15 The multidimensionality of skills | 103 Notes 1. Schönfeld (2017). Labor Market: The Practice Model.” Policy Research 2. For a discussion of alternative definitions of the term Working Paper 7123, World Bank, Washington, DC. skills, see Green (2011) and Warhurst and others (2017). Handel, Michael J., Alexandria Valerio, and María Laura 3. Winthrop and McGivney (2016, 14). Sánchez Puerta. 2016. Accounting for Mismatch in Low- 4. Neisser and others (1996, 77). and Middle-Income Countries. Directions in Development: 5. Guerra, Modecki, and Cunningham (2014, 5). Human Development Series. Washington, DC: World 6. “Core SEL Competencies,” CASEL, http://www.casel.org Bank. /core-competencies/. Also see Pierre and others (2014) John, Oliver P., and Filip DeFruyt. 2015. “Education and Social and Taylor and others (2017). Progress: Framework for the Longitudinal Study of 7. Duckworth and Yeager (2015); Durlak and others (2011); Social and Emotional Skills in Cities.” Report EDU/CERI John and DeFruyt (2015); Kautz and others (2014); Payton /CD, Organisation for Economic Co-operation and Devel- and others (2008). opment, Paris. 8. Pierre and others (2014). Kautz, Tim, James J. Heckman, Ron Diris, Bas Ter Weel, and 9. Kautz and others (2014). Lex Borghans. 2014. “Fostering and Measuring Skills: 10. Cunha, Heckman, and Schennach (2010); Guerra, Improving Cognitive and Non-cognitive Skills to Pro- Modecki, and Cunningham (2014). mote Lifetime Success.” NBER Working Paper 20749, 11. Cunha and others (2006). National Bureau of Economic Research, Cambridge, MA. 12. Handel, Valerio, and Sánchez Puerta (2016). Neisser, Ulric, Gwyneth Boodoo, Thomas J. Bouchard Jr., 13. Cunningham, Acosta, and Muller (2016). A. Wade Boykin, Nathan Brody, Stephen J. Ceci, Diane F. 14. Sánchez Puerta, Valerio, and Gutiérrez Bernal (2016); Halpern, et al. 1996. “Intelligence: Knowns and Taylor and others (2017). Unknowns.” American Psychologist 51 (2): 77–101. 15. Handel and others (2016). Payton, John, Roger P. Weissberg, Joseph A. Durlak, Allison B. Dymnicki, Rebecca D. Taylor, Kriston B. Schellinger, and Molly Pachan. 2008. “The Positive Impact of Social and References Emotional Learning for Kindergarten to Eighth-Grade Students: Findings from Three Scientific Reviews.” Tech- Cunha, Flavio, James J. Heckman, Lance Lochner, and nical Report, Collaborative for Academic, Social, and Dimitriy V. Masterov. 2006. “Interpreting the Evidence Emotional Learning, Chicago. on Life Cycle Skill Formation.” In Handbook of the Econom- Pierre, Gaëlle, María Laura Sánchez Puerta, Alexandria ics of Education, Vol. 1, edited by Eric A. Hanushek and Valerio, and Tania Rajadel. 2014. “STEP Skills Measure- Finis Welch, 697–812. Handbooks in Economics Series ment Surveys: Innovative Tools for Assessing Skills.” 26. Amsterdam: North-Holland. Social Protection and Labor Discussion Paper 1421, Cunha, Flavio, James J. Heckman, and Susanne M. World Bank, Washington, DC. Schennach. 2010. “Estimating the Technology of Cogni- Sánchez Puerta, María Laura, Alexandria Valerio, and tive and Noncognitive Skill Formation.” Econometrica Marcela Gutiérrez Bernal. 2016. Taking Stock of Programs 78 (3): 883–931. to Develop Socioemotional Skills: A Systematic Review of Cunningham, Wendy V., Pablo Acosta, and Noël Muller. Program Evidence. Directions in Development: Human 2016. Minds and Behaviors at Work: Boosting Socioemotional Development Series. Washington, DC: World Bank. Skills for Latin America’s Workforce. Directions in Develop- Schönfeld, Manuel. 2017. “Work Readiness Assessment Tools ment: Human Development Series. Washington, DC: in Comparison: From Administration to Z-scores.” World World Bank. Bank, Washington, DC. Duckworth, Angela L., and David Scott Yeager. 2015. “Mea- Taylor, R.D., E. Oberle, J. A. Durlak, and R. P. Weissberg. 2017. surement Matters: Assessing Personal Qualities Other “Promoting Positive Youth Development through Than Cognitive Ability for Educational Purposes.” Educa- School-Based Social and Emotional Learning Interven- tional Researcher 44 (4): 237–51. tions: A Meta-Analysis of Follow-Up Effects.” Child Devel- Durlak, Joseph A., Roger P. Weissberg, Allison B. Dymnicki, opment 88 (4): 1156–71. Rebecca D. Taylor, and Kriston B. Schellinger. 2011. “The Warhurst, Chris, Ken Mayhew, David Finegold, and John Impact of Enhancing Students’ Social and Emotional Buchanan, eds. 2017. The Oxford Handbook of Skills and Learning: A Meta-Analysis of School-Based Universal Training. Oxford, U.K.: Oxford University Press. Interventions.” Child Development 82 (1): 405–32. Winthrop, Rebecca, and Eileen McGivney. 2016. “Skills for a Green, Francis. 2011. “What Is Skill? An Inter-Disciplinary Changing World: Advancing Quality Learning for Synthesis.” LLAKES Research Paper 20, Centre for Learn- Vibrant Societies.” Center for Universal Education, ing and Life Chances in Knowledge Economies and Brookings Institution, Washington, DC. Societies, Institute of Education, University of London. Guerra, Nancy, Kathryn Modecki, and Wendy V. Cunning- ham. 2014. “Developing Social-Emotional Skills for the 104 | World Development Report 2018 PART III Innovations and evidence for learning 5 There is no learning without prepared, motivated learners 6 Teacher skills and motivation both matter (though many education systems act like they don’t) 7 Everything else should strengthen the teacher-learner interaction 8 B  uild on foundations by linking skills training to jobs SPOTLIGHT 4 Learning about learning Identifying gaps between evidence and practice helps set priorities for action. As evidence of the learning crisis has grown, so has translates into clearer insights into how to improve understanding of what produces learning. Cognitive learning at the level of the student, the classroom, neuroscience has evolved dramatically, with brain and the school. Beyond the increase in their num- imaging revealing new insights into how children ber, these impact evaluations have also grown more learn.1 Over the last two decades, neuroscience has sophisticated over time, making them more useful for been instrumental to understanding early child brain policy making. They are now more likely to compare development and the crucial nature of the early years.2 multiple interventions, more likely to study a wide Schools in many parts of the world are innovating in range of interventions overall, and more likely to approaches to pedagogy, professional development, study interventions on a large scale. The evaluations and the use of new technologies.3 Governments and show that many of these interventions have sizable nonprofits are trying out innovative programs to impacts. Several pedagogical interventions, for exam- upgrade teachers’ skills on the job.4 ple, deliver learning gains greater than what students At the same time, evidence on which programs would learn in a year of business-as-usual schooling.6 most effectively boost learning is mushrooming. One example of that growth: the number of impact Making better use of evaluations of interventions intended to improve learning outcomes in developing countries rose from evidence 19 in 2000 to 299 by 2016 (figure S4.1).5 This evidence Not all evidence is created equal, but many different kinds of evidence can be credible. Scientific evidence Figure S4.1 The number of experimental and quasi- demonstrates the pathways of brain development and experimental studies of interventions to improve functioning. Social science evidence can effectively learning has mushroomed in recent decades answer the question of what would have happened in the absence of a reform or intervention (often called 299 a counterfactual). Randomized controlled trials or 300 analyses of “natural experiments” are useful tools for No. of cumulative studies determining such a counterfactual. Implementation science and case studies can provide a detailed picture 200 of how an intervention or a phenomenon works. The best evidence of what improves learning draws from a range of methods. 100 Even when an intervention in one education sys- tem has a positive impact, it may not work everywhere. 19 Effects may differ when translating from one location 0 to another or from a pilot study to a large-scale pro- 1980 1986 1992 1998 2004 2010 2016 gram. What works in Peru may not work in Burundi Sources: WDR 2018 team, using data from 3ie (2016) and Evans and Popova (2016b). Data at http://bit because the education systems and societies are differ- .do/WDR2018-Fig_S4-1. ent. A common intervention that has been tested in a Note: The blue segment on the white bars represents the increment from the previous year. range of settings is to reduce class size. But increasing 108 | World Development Report 2018 class size by 10 students reduced test scores by four theory—models of human behavior—to explain why times as much in Israel as it did in Kenya.7 A pilot inter- some proposed solutions work and others do not, as vention may allow for more controlled conditions than well as why the same solution may work in one locale an at-scale intervention. In Kenya, an intervention to or time but not in another. hire contract teachers was effective on a small scale, but when it was implemented at scale through govern- Producing learning is ment systems, salaries were delayed and ultimately the contract teachers were converted to civil servants.8 The complex, but investments scaled-up program no longer resembled the successful that change what happens pilot, and the learning gains failed to materialize. in the classroom are a To make sense of the evidence, policy makers should consider the likely principles behind effective good bet programs rather than fixating on results (or “point Many actors contribute to the learning process, and estimates”) from individual studies.9 For example, pro- they all face their own incentives. The direct inputs grams that provide financial incentives for teachers to the learning process include the choices made by have had mixed effects. Rather than taking a simple learners themselves, as well as by their parents, teach- average of the effects, a nuanced assessment would ers, and other school leaders, interacting with the reveal that these programs tend to work better when available infrastructure and materials. Less immedi- improving quality is relatively simple and within a ate but still important, bureaucrats, politicians, and teacher’s control—for example, when they increase nonstate players make decisions that influence edu- teacher attendance or teaching time while at school.10 cation quality. Understanding these relationships is Viewing evidence through models of human crucial to interpreting evidence. FIGURE O.14 behavior is one way to focus on principles. This Each actor in the learning process reacts to the means examining patterns of results and using mod- others, so changing one element of the process does els to infer why results vary across settings. The first not guarantee more learning. Many of the inputs to step would be a nuanced synthesis, bringing together the learning process are choices made by the actors— the results of a range of studies and examining choices made in reaction to the actual and anticipated empirical patterns. The second step would be using choices of other actors (figure S4.2). Teachers react to Figure S4.2 It’s more complicated than it looks: People act in reaction to the choices of others throughout the system Civil so ciet ians yo li tic rg Po an iz at io ns r Pe c to s Le er er ar se ch s/c a te ne a om Te rs Priv mun ities LEARNING Sch B u re ts oo pu y lm a uc an iar in ag ol ho dic rat em ent Sc s Ju In te rn at ion to rs al a c to r r ac s Othe Source: WDR 2018 team. Learning about learning | 109 changes in school leadership, school directors react to and practice requires good information on what the community demands, and parents react to changes evidence says, as well as what current practice is, it in government policy. In India and Zambia, grants to is likely that many opportunities for improvement schools led parents to reduce their own investments have yet to be discovered. in their children’s schooling.11 In a household with Intuition and common sense are not enough. One few resources, if the government begins providing fundamental lesson from the growing evidence base textbooks, a parent may well reallocate education is that intuition is not always a trustworthy guide. It resources to other needs, such as health. may miss the complexity of motivations and reac- How can we make sense of all of these complex, tions in the real world, as can happen when teacher dynamic relationships? Models of human behavior financial incentives induce cheating rather than illuminate the motives for choices and actions, and more effort.14 Intuition may fail to capture the net they can help guide solutions. Simple optimizing effect of conflicting forces, such as when separating behavior models—in which actors maximize their students by ability allows teachers to target teaching well-being subject to limited budgets and other con- more specifically to students’ level—which should straints—explain why parents reduce their contribu- increase their learning—but also distances them from tions when schools increase theirs. Principal-agent their high-performing peers—which may decrease models that incorporate multiple actors with dif- their learning. ferent objectives explain why teachers may fail to Knowledge about improving learning must teach when not sufficiently motivated or monitored. take both the costs and the benefits of learning inter- Behavioral models also play a role: student learning ventions into account. A computer-assisted learning and educational aspirations can be affected by the intervention in India increased learning more than salience of stereotypes. Economic phenomena such as employing contract teachers in Kenya, but hiring information, market, and coordination failures play a contract teachers was so much cheaper that it deliv- role in these models. The models can also illuminate ered a higher return on investment.15 The evidence why a gap is often observed between evidence on how base on costs is much thinner than that on ben- to improve learning and actual practice. efits, with a tiny fraction of studies examining both.16 But some programs have been evaluated on both effectiveness and cost-effectiveness.17 This evi- Focusing where the gaps dence on costs—adapted to local contexts—should between evidence and qualify policy recommendations.18 practice are largest The gaps between evidence and practice signal promising places to start, rather than the end of Gaps between evidence and actual practice provide learning how to improve learning. Interventions can- entry points for efforts to improve education. These not simply be exported from one country to another. gaps come to light when evidence shows that certain Indeed, at times the effectiveness of an apparently approaches or interventions can improve outcomes, similar intervention can vary even within a country, but the approaches used in practice are different.12 depending on how the program is implemented.19 For example, the accumulated research evidence The cost of implementation will also vary dramat- demonstrates high returns to early investments ically across contexts.20 But this does not mean that in children, yet families and governments in low- evidence from other contexts is without value. On the income environments do not prioritize these invest- contrary, successes in other environments—coupled ments. Evidence shows that certain types of teacher with a careful analysis of why the programs work— professional development deliver much higher provide a starting point. Policy makers can draw on learning gains than others, but outdated training this evidence and experiment in their own policy methods persist.13 Because the gap between evidence environment. Notes 1. De Smedt (2014); Insel and Landis (2013); Kuhl (2010). 6. Evans and Yuan (2017). 2. Dua and others (2016). 7. Pritchett and Sandefur (2013). 3. Chisholm and Leyendecker (2008); Schweisfurth (2011). 8. Bold and others (2016); Duflo, Dupas, and Kremer (2015). 4. Popova, Evans, and Arancibia (2016). 9. Muralidharan (2017). 5. Evans and Popova (2016b). 10. Muralidharan and Sundararaman (2011). 110 | World Development Report 2018 11. Das and others (2013). (TAP) in Chicago: Year One Impact Report, Final Report.” 12. Montagu and Goodman (2016); Pakenham-Walsh (2004). Mathematica Policy Research, Princeton, NJ. 13. Lauwerier and Akkari (2015). Insel, Thomas  R., and Story  C. Landis. 2013. “Twenty-Five 14. Glazerman, McKie, and Carey (2009); Jacob and Levitt Years of Progress: The View from Nimh and Ninds.” (2003). Neuron 80 (3): 561–67. 15. Kremer, Brannen, and Glennerster (2013). Jacob, Brian A., and Steven D. Levitt. 2003. “Rotten Apples: An 16. McEwan (2015). Investigation of the Prevalence and Predictors of Teacher 17. Kremer, Brannen, and Glennerster (2013). Cheating.” Quarterly Journal of Economics 118 (3): 843–78. 18. Evans and Popova (2016a). Jinnai, Yusuke. 2016. “To Introduce or Not to Introduce Mon- 19. Bold and others (2016); Kerwin and Thornton (2015). etary Bonuses: The Cost of Repealing Teacher Incen- 20. Evans and Popova (2016a). tives.” Economics and Management Series EMS-2016-08 (January), IUJ Research Institute, International Univer- sity of Japan, Minamiuonuma, Niigata Prefecture, References Japan. http://www.iuj.ac.jp/research/workingpapers /EMS_2016_08.pdf. 3ie (International Initiative for Impact Evaluation). 2016. Kerwin, Jason T., and Rebecca L. Thornton. 2015. “Making “Impact Evaluation Repository.” 3ie, London. http:// the Grade: Understanding What Works for Teaching www.3ieimpact.org/en/evidence/impact-evaluations Literacy in Rural Uganda.” PSC Research Report 15-842, /impact-evaluation-repository/. Population Studies Center, Institute for Social Research, Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, Alice University of Michigan, Ann Arbor. Ng’ang’a, and Justin Sandefur. 2016. “Experimental Evi- Kremer, Michael R., Conner Brannen, and Rachel Glenner- dence on Scaling Up Education Reforms in Kenya.” Eco- ster. 2013. “The Challenge of Education and Learning in nomic Development and Institutions Project Working the Developing World.” Science 340 (6130): 297–300. Paper, Institute for International Economic Studies, Kuhl, Patricia K. 2010. “Brain Mechanisms in Early Language Stockholm University. Acquisition.” Neuron 67 (5): 713–27. Chisholm, Linda, and Ramon Leyendecker. 2008. “Curricu- Lauwerier, Thibaut, and Abdeljalil Akkari. 2015. “Teachers lum Reform in Post-1990s Sub-Saharan Africa.” Interna- and the Quality of Basic Education in Sub-Saharan tional Journal of Educational Development 28 (2): 195–205. Africa.” ERF Working Paper 11, Education Research and Das, Jishnu, Stefan Dercon, James Habyarimana, Pramila Foresight, Paris. Krishnan, Karthik Muralidharan, and Venkatesh Sunda- McEwan, Patrick J. 2015. “Improving Learning in Primary raraman. 2013. “School Inputs, Household Substitution, Schools of Developing Countries: A Meta-Analysis of and Test Scores.” American Economic Journal: Applied Eco- Randomized Experiments.” Review of Educational Research nomics 5 (2): 29–57. 85 (3): 353–94. De Smedt, Bert. 2014. “Advances in the Use of Neuroscience Montagu, Dominic, and Catherine Goodman. 2016. “Pro- Methods in Research on Learning and Instruction.” hibit, Constrain, Encourage, or Purchase: How Should Frontline Learning Research 2 (4): 7–14. We Engage with the Private Health-Care Sector?” Lancet Dua, Tarun, Mark Tomlinson, Elizabeth Tablante, Pia Britto, 388 (10044): 613–21. Aisha Yousfzai, Bernadette Daelmans, and Gary L. Darm- Muralidharan, Karthik. 2017. “Field Experiments in Educa- stadt. 2016. “Global Research Priorities to Accelerate tion in Developing Countries.” In Handbook of Field Exper- Early Child Development in the Sustainable Develop- iments, edited by Abhijit Vinayak Banerjee and Esther ment Era.” Lancet Global Health 4 (12): e887–e889. Duflo, Vol. 2, 323–88. Handbooks in Economics Series. Duflo, Esther, Pascaline Dupas, and Michael R. Kremer. Amsterdam: North-Holland. 2015. “School Governance, Teacher Incentives, and Pupil- Muralidharan, Karthik, and Venkatesh Sundararaman. 2011. Teacher Ratios: Experimental Evidence from Kenyan “Teacher Performance Pay: Experimental Evidence from Primary Schools.” Journal of Public Economics 123 (March): India.” Journal of Political Economy 119 (1): 39–77. 92–110. Pakenham-Walsh, Neil. 2004. “Learning from One Another Evans, David K., and Anna Popova. 2016a. “Cost- to Bridge the ‘Know-Do Gap.’ ” BMJ 329 (7475): 1189. Effectiveness Analysis in Development: Accounting for Popova, Anna, David K. Evans, and Violeta Arancibia. 2016. Local Costs and Noisy Impacts.” World Development 77: “Training Teachers on the Job: What Works and How 262–76. to Measure It.” Policy Research Working Paper 7834, — —— —. 2016b. “What Really Works to Improve Learning in World Bank, Washington, DC. Developing Countries? An Analysis of Divergent Find- Pritchett, Lant, and Justin Sandefur. 2013. “Context Matters ings in Systematic Reviews.” World Bank Research Observer for Size: Why External Validity Claims and Development 31 (2): 242–70. Practice Do Not Mix.” Journal of Globalization and Develop- Evans, David K., and Fei Yuan. 2017. “Economic Returns to ment 4 (2): 161–98. Interventions That Increase Learning.” Background Schweisfurth, Michele. 2011. “Learner-Centred Education in paper, World Bank, Washington, DC. Developing Country Contexts: From Solution to Prob- Glazerman, Steven, Allison McKie, and Nancy Carey. 2009. lem?” International Journal of Educational Development “An Evaluation of the Teacher Advancement Program 31 (5): 425–32. Learning about learning | 111 5 There is no learning without prepared, motivated learners Strong foundations underpin all learning and skills development. Learning depends on students who are prepared, present, and motivated—but getting students there will often require policy change within and beyond education systems. Schools cannot produce learning without prepared, To address the fact that so many youth leave basic •  present, motivated learners. Around the world, many education lacking skills, recognize that remediation children receive too little investment in nutrition and often needs to be the first step in further education stimulation during their early years, and many lack and training. access to quality early learning opportunities that can prepare them for first grade. The one in four children worldwide who are stunted cannot achieve their Investing in their early years potential in school.1 Nor can the 263 million young prepares children for school people who do not make it to school at all. Among those who do attend, motivation to learn often suffers Children’s early years offer a rare window for societies when the quality of education is low. A poor-quality to make investments in their children with extremely basic education also means that learners who should high returns (figure 5.1). Efforts to improve children’s be gaining advanced skills from tertiary education or lives can significantly increase individual and societal technical training lack the preparation to do so.2 Thus, productivity while reducing inequality.3 Children can- just as the fundamental investments needed for pri- not thrive with stunted bodies and brains, and early mary education must be made before a child enters gaps in learning and skills trap them in lower develop- school, the same is true for skills training. In many mental trajectories from which it becomes increasingly cases, the failure to invest effectively can be under- difficult to escape (spotlight 2). Though children’s bod- stood through models of human behavior, which also ies are resilient, and catch-up after early childhood may point the way to solutions (table 5.1). A synthesis of the be possible when inputs improve, it is extremely dif- evidence in these areas reveals three key principles ficult to reverse the effects of exposure to risk factors for improving learning: in the first few years of a child’s life. Doing so entails costly, high-quality interventions that typically need to •  To set children on high-development trajectories, happen at a sufficiently young age to be effective. foster cognitive and socioemotional development through early child nutrition, care, stimulation, and Recognizing the dangers that poverty learning opportunities. poses to children’s development and •  To get children into school—an essential first step learning to learning—lower school costs and then use other Children need quality environmental inputs to grow tools to boost motivation for learning. in a healthy, timely fashion. Essential physical inputs 112 | World Development Report 2018 Table 5.1 Models of human behavior can guide actions to improve learner preparation: Some examples Models that identify a Approaches that address the Synthesis principle Where this fails mechanism behind this failure modeled mechanism Provide early child Just one in five children Information failure: Stakeholders In Jamaica, a program taught nutrition, care, in low-income countries may not be aware of relative caregivers to provide psychosocial stimulation, and learning attend preschool. One in returns to early investments or how stimulation that improved stunted opportunities. four children worldwide to support early development. children’s developmental scores are stunted. and later life outcomes. Simple optimization with liquidity In Mexico, a conditional cash and credit constraints: Parents are transfer program improved aware but lack the resources to cognitive and motor development. invest. Behavioral (mental bandwidth): In Argentina, Bangladesh, China, Stress of poverty undermines and Uganda, center-based parenting capacity. programs improved children’s outcomes. Lower school costs; 263 million children Simple optimization with liquidity In Cambodia, providing boost motivation and remain out of school. Many and credit constraints: Parents are scholarships to girls dramatically effort. countries still charge fees aware but lack the resources to increased enrollment. for lower secondary school, invest in any or all children. and primary school, while usually tuition-free, still Information failure: Youth and In the Dominican Republic and entails cash outlays in parents may underestimate the Madagascar, providing information many settings. returns to education. on the returns to education improved enrollment and learning. Behavioral (hyperbolic discounting): In Pakistan, reporting child test Youth may recognize the value of scores to parents increased education but plan to invest later enrollment and learning outcomes. (yet “later” never comes). Ensure that, where Many skills training Information failure: Training In U.S. community colleges, needed, remediation is programs assume programs receive imperfect signals improving course placement the first step in further prerequisite skills that about the quality of incoming accuracy and support services education and training. youth do not have. learners. helped increase students’ long-term performance. Simple optimization (on the part of In the United States, bridge training centers): Remedial students programs help learners move past are highly likely to drop out. remediation quickly. Source: WDR 2018 team. include quality pre- and postnatal nutrition, health diseases, and chemically toxic or physically dan- care, and safe physical environments.4 Equally crucial gerous environments affect many poor children are social inputs, including nurturing, protection, and not only after birth, but also in the womb. Exposure stimulation.5 Interactions between children and their to any of these factors during sensitive periods can caregivers—who are often, but not always, their par- inhibit normal biological development (spotlight 1), ents—leave a significant imprint, literally shaping the but poor children often encounter these factors in developing brain.6 Yet poor children’s access to these tandem and over time.7 At the same time, the strains inputs—along with caregivers’ awareness of their associated with poverty can disrupt parents’ deci- importance—is often limited. So are programs that sion making and limit their availability, sensitivity, invest in children’s early development and the policies and responsiveness.8 As a result, poorer children not that guide them. only have fewer resources such as books or toys, but Poor children are more exposed to health shocks also receive less stimulation, direction, and support.9 and less likely to receive stimulation, care, and pro- Poor children are also more likely to experience tection from stress. Nutrient deprivation, infectious neglect and harsher discipline, which disrupts early There is no learning without prepared, motivated learners | 113 FIGURE 5.1 It pays to invest in high‐quality programs during the early years Figure 5.1 Investments in high-quality programs most regions. In Sub-Saharan Africa, on average just during children’s early years pay off 2 percent of the education budget goes to preprimary education.15 In Latin America, the average per capita government spending on children under 5 is a third Investments in of that for children ages 6–11.16 Investments in the the early years early years have increased in developing countries, Brain development but strategies often focus on building preschools, neglecting children who have not yet reached pre- school age. Though preschool can help, foundations across developmental dimensions are set before age 3. Yet this age group typically receives little government Schooling coverage beyond health and nutrition checkups—not enough for healthy overall development. Job training Early exposure to risks associated with poverty may prevent children from realizing the promise Rate of return of education. Intense deprivation can result in poor to investment in human capital developmental outcomes—such as stunted growth or impaired brain development—that are difficult to Preschool School Postschool address (figure 5.2; spotlight 2). Children who have fallen behind in their physical, cognitive, linguistic, or Age socioemotional development are more likely to enter Source: WDR 2018 team, based on Carneiro, Cunha, and Heckman (2003); Martin (2012). grade 1 late, score poorly in school, repeat grades, drop out before they complete primary school, experience poor health throughout their lives, engage in high-risk emotional organization—the keystone of socioemo- behavior (particularly in adolescence), be less produc- tional abilities—and is associated with worse school tive, and have lower earnings.17 The scale of the prob- performance.10 lem is vast: nearly half of children under 5 in develop- Early childhood development programs are ing countries are stunted or live in extreme poverty, insufficient in number and quality to compensate threatening their prospects of benefiting from the for poor children’s disadvantages, especially in the opportunities education can provide.18 developing world. In poor communities, resources that stimulate early development outside the home— Strengthening children’s ability to learn including quality child care, libraries, recreation with well-designed interventions centers, and preschool programs—tend to be limited Effective early childhood interventions can signifi- and low in quality.11 Only half of 3- to 6-year-olds have cantly improve poor children’s ability to learn. In the access to preprimary education. Coverage is strongly United States, at-risk children who participated in associated with income, ranging from 19 percent in well-designed interventions—Perry Preschool, Abece- low-income countries to 86 percent in high-income darian, the Nurse-Family Partnership—benefited well countries, with poorer children enrolled at the lowest beyond their early years: their school performance, rates in every country.12 Children under 3 are widely employment, income, overall welfare, and social underserved, with access to services for this age integration all improved. Such interventions have group especially inequitable and uncoordinated.13 substantial potential in developing countries because Moreover, reliance on poorly compensated child care of their lower baselines. In Jamaica, the Reach Up and workers who receive little to no training, mentoring, Learn program, which promoted early child stimula- or monitoring undermines sustainability, retention, tion, led to lower crime rates, better mental health, and and quality.14 25 percent higher earnings two decades later. There is Governments do not invest enough in young chil- a consensus on what children need: nutrition, care, dren. Insufficient understanding of the high payoffs stimulation, nurturing, and protection. The evidence to early interventions, budget constraints, and the on when to implement programs is in line with bio- challenges of delivering wide-ranging early childhood logical evidence: prevention and early remediation are interventions—health, nutrition, early learning— most cost-effective at specific points in development result in low public investment in young children in because adjustments beyond sensitive periods are 114 | World Development Report 2018 difficult, costly, and usually incomplete. But iden- Figure 5.2 Intense deprivation can impair brain tifying the most effective approaches to improving development poor children’s developmental outcomes has proven Brain structure and wiring by stunting status challenging because of the enormous heterogeneity in interventions as well as contexts. Still, several a. Infant representative of never-stunted growth approaches show promise. Health and nutrition interventions during the first 1,000 days of life (starting at conception) improve children’s development. Programs that increase access to maternal health services improve maternal nutrition through diet, supplements, and fortifica- tion, while reducing child mortality and early health problems.19 In isolation, nutritional interventions for children have only modest effects on height or stunt- ing.20 But when combined with improved sanitation, along with access to child health services, nutritional interventions can yield significant benefits.21 Breast- feeding and micronutrient supplements are associ- ated with better health and greater cognitive ability, leading to better educational outcomes in developing countries.22 Deworming, iodine supplements, and immunizations have also led to major improvements b. Infant representative of stunted growth in children’s ability to learn.23 Programs that build caregivers’ capacity to sup- port healthy development can substantially improve children’s outcomes. Interventions include coaching caregivers at home on positive discipline, as well as promoting increased frequency of quality interven- tions through nurturing, protection, and stimulating activities (storytelling, singing, playing with house- hold objects). Such interventions have been delivered in diverse ways, including home visits, community meetings, and health checkups.24 The most effective programs have systematic training and curriculums, as well as opportunities for caregivers to practice and receive feedback.25 An emerging generation of programs is offering parents incentives through positive reinforcements, with indirect “nudges” when providing information is insufficient or when beliefs Source: Nelson and others (2017). © Nadine Gaab and Charles A. Nelson. Used with the permission of Charles A. Nelson; further permission required for reuse. or norms are detrimental.26 Note: The images illustrate two infants, 2–3 months old. The growth of one infant was stunted (panel b); Programs that provide caregivers with cash or the growth of the other infant was not (panel a). The images were obtained in Dhaka, Bangladesh, using psychosocial support complement interventions magnetic resonance imaging (MRI). The left side of each panel shows the left side of the head. Each gold line represents a fiber tract—the long, thin fibers (axons) in the brain that transmit information to different to improve parenting. Cash transfer programs can neurons, muscles, and glands. It is apparent how much denser and more elaborate the connections are address acute material deprivation in households in the nonstunted infant. The colored images on the right side of each panel illustrate the same principles (neural connections) from a different orientation—a cross-section of the brain, from front to back. and improve developmental outcomes, particularly when provided alongside—or conditional on—prena- tal care and child services. For example, conditional child outcomes beyond the direct effects of the trans- cash transfer (CCT) programs in Ecuador, Mexico, fers.28 Also important, transfer programs can alleviate and Nicaragua have reduced stunting, improved parental time and psychological constraints. In addi- cognitive development, and promoted better par- tion, interventions delivered by supervised, nonspe- enting practices.27 In Mexico, parenting support cialist health or community workers to address acute programs integrated with CCT programs improved maternal stress, depression, and anxiety have led to There is no learning without prepared, motivated learners | 115 better cognitive development, more physical growth, factors that cannot be adequately addressed by any less diarrhea, and higher immunization rates.29 single intervention. Multifactor programs capture the Center-based care can promote foundational complex, complementary nature of early childhood skills. In countries from Ethiopia to the United States, development and exploit complementarities (figure high-quality, center-based programs have shown sub- 5.3).34 To be effective, interventions must be delivered stantial benefits in developing children’s language, during specific stages of development.35 Packaging cognitive, motor, and socioemotional skills.30 By con- interventions to address sequential or related devel- trast, attending a low-quality, center-based program opmental goals can increase effectiveness, especially can be worse than attending none at all.31 The quality if intervention packages incorporate benefits for of child-caregiver interactions is a key determinant caregivers as well. Integrated intervention packages of such programs’ impacts, as Indonesia and Mozam- can build on existing platforms such as community- bique demonstrated with effective center-based based strategies or social safety nets, though the preschool programs for children ages 3 to 6. These pro- effectiveness of any specific strategy will depend on grams included minimal infrastructure investments contextual factors.36 Quality should not be diluted in but improved children’s cognitive abilities thanks to the effort to increase investments in the early years— their interactions with well-trained caregivers (box say, by relying on volunteers or unqualified workers 5.1).32 Delivering quality, center-based interventions for to deliver services, which is common.37 children under 3 is harder because they require costlier structural investments (such as lower child-to-staff ratios). Consequently, programs to build parenting Providing demand-side capacity might be most cost-effective for children support can get kids to under 3 in resource-constrained environments or to school, but not necessarily reach marginalized populations.33 to learn Bringing it all together School is a key input to at-scale learning. Despite Integrating programs can lead to better development major gains in access, many children still don’t attend outcomes. Poor children are exposed to multiple risk school. Even though school is not the only place that Box 5.1 Early childhood education prepares young children for school Preschool programs targeting children ages 3–6 can foster play, and interaction with others.c Key elements of pro- foundational skills and boost children’s ability to learn. grams that have led to strong preschool outcomes include Children who attend preschool have higher attendance and curriculums that foster crucial pre-academic abilities (emo- better achievement in primary school. Moreover, they are tional security, curiosity, language, self-regulation) through less likely to repeat, drop out, or need remedial or special play; professional development plus coaching that enable education, all of which benefit not only students but also teachers to effectively implement relevant curriculums; education systems because efficiency is increased.a Across and positive, engaging classrooms that promote children’s countries at all income levels, the most disadvantaged innate drive to learn.d For early child education gains to be children benefit most from quality early child education sustained, the content, budget, and capacity of providers programs.b But early child education programs are not all of preschool programs should be integrated into formal equally effective; overly academic and structured programs education systems. In addition, the quality of subsequent for children under 5 may undermine their cognitive and learning environments in primary school is an import- socioemotional skills, as well as their motivation to learn, ant determinant of the long-term effects of preschool because young children learn best through exploration, programs.e Source: WDR 2018 team. a. Klees (2017). b. Britto and others (2016). c. Whitebread, Kuvalja, and O’Connor (2015). d. Phillips and others (2017). e. Johnson and Jackson (2017). 116 | World Development Report 2018 Figure 5.3 Integrated programs through the early years are necessary for proper child development Key interventions for young children and their families Pregnancy Birth 1 year 2 years 3 years 4 years 5 years 6 years 1 Family support package Parental support for vulnerable families: planning for family size and spacing; maternal education; education about early stimulation, growth, and development; parental leave and adequate child care; prevention and treatment of parental depression; social assistance transfer programs; child protection regulatory frameworks Health, nutrition, and sanitation for families: access to health care; access to safe water; adequate sanitation; hygiene/handwashing; micronutrient supplementation and fortification 2 4 Child health and development package Immunizations; deworming; prevention and treatment of acute Pregnancy malnutrition; complementary feeding and adequate, nutritious, package 3 and safe diet; therapeutic zinc supplementation for diarrhea Antenatal Birth care; iron package and folic acid; counseling Attended on adequate delivery; 5 Preschool package diets exclusive AB breastfeeding; C Preschool education programs (early childhood and preprimary); continuity birth to quality primary schools registration Source: Denboba and others (2014). children learn (box 5.2), most parents want their chil- school supplies, learning materials, transportation— dren to go to school. Moreover, most children want to are in addition to formal fees.41 These costs of school- go. In a survey of Indian mothers with an average of ing widen the gaps in school participation separating less than three years of education, 94 percent hoped poorer children from their wealthier peers. their children would complete at least grade 10.38 In High aspirations for schooling among children Kenya, among parents with no education at all, more and their parents explain why initiatives that ease than half wanted a university education for their constraints to schooling for households—so-called children.39 demand-side interventions—have been so effective Significant costs—both formal fees and a wide at getting children to school. In many countries, the array of other expenses—prevent children, especially elimination of school fees has raised enrollments, the most vulnerable, from learning. Nearly 90 percent suggesting that parents simply did not have the of the world’s low-income countries proclaim free resources to pay the fees (figure 5.4).42 The inter- primary education. But for lower secondary educa- ventions, which have sought to reduce other costs tion, more than 40 percent of the countries charge associated with school, have consistently improved fees, along with 10 percent of middle-income coun- access in the form of enrollment as well as atten- tries.40 In Africa, almost half the expenditures that dance.43 Nonmerit scholarships—which reduce fees households incur to send their children to school—for on a smaller scale—have increased enrollment at the There is no learning without prepared, motivated learners | 117 Box 5.2 Communities can leverage the many hours spent outside the classroom to boost learning Much learning happens outside the classroom, including implementing read-a-thons (in which all the books that from tutoring and at-home programs. Across Africa and children read during a specific period are recorded), and Asia, the Literacy Boost program has implemented com- providing mini-libraries. Children who participate in such munity reading activities to leverage the many hours that activities have better reading outcomes. In Rwandese learners spend outside school. These include pairing strug- communities, implementing Literacy Boost led to better gling readers with stronger readers (“reading buddies”), reading skills and school advancement.a Source: WDR 2018 team. a. Dowd and others (2017); Friedlander and Goldenberg (2016). Figure 5.4 What happens when school primary level in Kenya and at the secondary level fees are eliminated? Evidence from in Ghana.44 The flip side of reducing school fees is eight countries increasing household income, which cash transfer programs do. These programs have increased both Gross enrollment in years before and after elimination of school fees, selected countries primary and secondary enrollments.45 Information interventions are particularly prom- 160 ising because they cost little.46 In some cases, demand for education remains low because students and their families underestimate the returns to education. In 140 the Dominican Republic and Madagascar, simply providing information on the returns to education led to improved educational outcomes, though a similar Gross enrollment (%) intervention in rural China had no impact.47 In India, 120 providing job recruiting services for women in their 20s increased school enrollment for teenage girls. Gender leadership quotas in Indian villages elimi- 100 nated the gender gap in educational attainment.48 Though interventions that reduce the cost of schooling are highly effective at increasing school participation for most children, especially at young 80 ages, some children do require additional incentives to attend school. In some countries, parents give priority to sending to school their children with the highest cognitive ability or higher perceived—not nec- 60 essarily actual—returns to schooling (such as boys).49 −10 −5 0 5 10 In Burkina Faso, beginning in 2008, some families Years before and after received unconditional cash transfers, while others elimination of school fees received cash transfers conditional on children’s Malawi Lesotho school enrollment. Boys and children who scored bet- Cambodia Zambia ter on tests were equally likely to be enrolled in school Uganda Cameroon under both schemes, but transfers with conditions Kenya Tanzania were significantly more beneficial for girls and chil- Source: WDR 2018 team, using data from World Bank (2017); year of policy change from Bentaouet Kattan (2006). Data at http://bit.do/WDR2018 dren who started out at lower levels of learning.50 This -Fig_5-4. finding suggests that the most vulnerable children Note: Vertical line indicates last year with fees. Gross enrollment rates include students whose age exceeds the official age group for a particular may need more than simple cost reductions to guar- education level, and so the rate may exceed 100 percent. antee enrollment in school. 118 | World Development Report 2018 Demand-side interventions can improve learning relevant, quality education that reaches them at their when programs increase either capacity to learn or current level of learning. In Kenya, students who student effort. Targeted cash transfers have led to more drop out of school say their inability to perform well, learning when framed to induce more effort, as have rather than costs or parental pressures, caused them some information interventions.51 Even in low-quality to leave.53 Some systems seek to further motivate stu- education systems, students learn more in school dents with merit-based scholarships or prizes. Such than out of it: there is a learning crisis, but the positive incentives can improve effort as students strive to relationship between schooling and literacy persists qualify—whether for a direct financial prize, such as (figure 5.5). When individuals with similar literacy in Benin and Mexico, or a scholarship for girls, such and numeracy levels are compared, those with more as in Kenya.54 Direct financial incentives have been schooling have higher earnings, most likely because of less successful in high-income countries, though other benefits of schooling, including improved socio- alternate designs that deliver incentives immediately emotional skills such as discipline.52 Getting learners after tests have raised test scores.55 Providing caregiv- into school is beneficial in its own right. ers with information about learner performance can In addition to getting to school, learners must also have a large impact, helping caregivers to trans- be motivated. One way to increase motivation is to late motivation into action (box 5.3). But in general, ensure that learners’ skills are rewarded, whether a positive overall educational experience is likely the by a labor market that offers high returns or by a backbone of student motivation. higher education system that admits students based on merit rather than connections. Perhaps the most immediate way to motivate students is to provide Remedial education can prepare learners for further Figure 5.5 Not all education systems education and training are equally productive, but even the Many young people leave formal education with least productive deliver some learning weak foundational skills, and thus they are unpre- to some learners pared for further education and training. Globally, Percentage of women ages 25–34, by highest grade of every 100 students entering primary education, completed, who can read all of a single sentence in their 61 complete lower secondary education, and just chosen language, selected countries 35 complete upper secondary (figure 5.6).56 About a 100 third of youth leave school between lower and upper secondary. This problem is especially pronounced in several developing countries, where sizable shares of 80 15- to 24-year-olds score below the minimum level of literacy proficiency—23 percent in Chile, 29 percent in urban Bolivia, 34 percent in urban Ghana.57 Improv- 60 ing foundational skills early can alter workers’ labor Percent market trajectories. Employed adults ages 15–64 who score at level 258 or above in literacy proficiency have 40 significantly higher probabilities of holding high- skill, better-paid white-collar jobs (figure 5.7).59 Youth vary greatly in skills and maturity, putting 20 them on a range of different pathways. Some young school leavers enroll in second-chance programs 0 seeking to obtain formal education equivalency diplo- mas so they can gain access to further education or 0 2 4 6 training.60 Others pursue remedial coursework to Highest grade completed fulfill admission requirements for postsecondary Rwanda Ethiopia Bangladesh education or training institutions.61 Another group— Peru Average Nigeria usually those with the most serious skills gaps—goes Source: Oye, Pritchett, and Sandefur (2016). Data at http://bit.do/WDR2018 into unstable, low-wage, low-productivity jobs, while -Fig_5-5. some youth remain out of both school and the labor Note: The average is calculated across 51 countries. force.62 It is difficult to reach all these young people. There is no learning without prepared, motivated learners | 119 Box 5.3 Providing information on children’s school performance can help parents to motivate their children Most parents want their children to succeed in school. to behave poorly in school, had better grades, and were Promising interventions in several countries show that more likely to move up to the next grade. After receiving providing parents with information about their children’s the messages, parents expressed a willingness to pay for performance can lead to better educational outcomes. In the service, suggesting that they saw real value to it.d the United States, text messages sent to parents when But simply providing information to parents is no guaran- secondary school students missed assignments led not tee of success: a program in Kenya that provided parents only to more assignment completion but also to higher test with information on their children’s literacy levels and scores.a Sending letters to parents about student absences suggested strategies to improve them did not lead to also reduced absenteeism.b In Malawi, providing parents change.e The programs that have been effective have pro- with information about their children’s academic ability vided parents with regular updates on the inputs to learn- enabled them to buy the appropriate books for their chil- ing—attendance and performance on individual assign- dren.c In Chile, low-income families received text messages ments—rather than just on learning levels. Such information each week detailing their child’s attendance record along interventions can be automated, making them extremely with a monthly message on behavior and test performance. cost-effective because they leverage the intrinsic motiva- Students whose parents received the texts were less likely tion of families. Source: WDR 2018 team. a. Bergman (2015). b. Rogers and Feller (2016). c. Dizon-Ross (2016). d. Berlinski and others (2016). e. Lieberman, Posner, and Tsai (2014). Figure 5.6 Young people follow different paths in their education Completion and attrition rates (percent), by cohort and region a. Of 100 students who enter b. But given disparate attrition rates across regions, primary around the world . . . upper secondary attainment varies greatly 90 complete primary 61 complete 75 lower secondary 35 complete 50 upper secondary 25 Percent 0 10 leave before completing 25 29 leave before 50 completing 26 leave before 75 completing Europe and East Asia Middle East Latin America South Sub-Saharan Central Asia and Pacific and and the Asia Africa North Africa Caribbean Completed Left before completing Left previously Source: WDR 2018 team, using data from UIS (2017); UNESCO (2015); WIDE (2017). Data at http://bit.do/WDR2018-Fig_5-6. Note: Estimates are for circa 2010. 120 | World Development Report 2018 Figure 5.7 Workers with higher literacy • Second-chance programs offer early school leavers, proficiency are more likely to enter many of whom are low-skilled, an opportunity to white-collar jobs reengage with education and training. • Remedial coursework at the onset of postsecondary Marginal probability of entering high-skill white-collar jobs relative to blue-collar jobs when scoring at level 2 or above education and training increases young people’s in literacy proficiency, for all workers in urban areas of chances of completing their programs of study. participating countries (2011–14) Remedial prevention programs can help Serbia low-performing students and keep them Ukraine in school Bolivia Remedial prevention programs can help at-risk youth who are in the formal education system to prepare Georgia for rigorous academic work in further education or Colombia training.66 Three remedial prevention approaches show promise.67 The first offers support to primary Vietnam and secondary students willing to stay in school and Armenia master foundational skills. Programs in India and Kenya Mexico City that offer additional instruction for dis- advantaged students have shown positive impacts on Ghana foundational skills (especially in India).68 The second −10 0 10 20 30 approach offers students early assessments of their Marginal probability (%) academic standing, along with extra instruction to Estimate and 95% confidence interval: improve performance. A statewide early assessment Significant Not significant program in California that supports academically Source: WDR 2018 team, using data from World Bank’s STEP Skills Measure- at-risk students shows declining needs for remedi- ment Program (http://microdata.worldbank.org/index.php/catalog/step /about). Data at http://bit.do/WDR2018-Fig_5-7. ation at later stages of education and training.69 The third approach gives secondary school students the option of registering concurrently in postsecondary Motivating them to join second-chance or remedial courses. Participants in such programs in the United programs is not easy, especially if they have been States are less likely to require remediation and more out of the education system for some time. Many likely to persist in tertiary education and improve are uncertain about the benefits of remedial courses, academic outcomes.70 and returning to school settings can stir up negative feelings. In Uganda, early school leavers said they suf- Second-chance programs offer a way to fered from diminished self-worth, limited life oppor- return to education and obtain training tunities, and social exclusion associated with early Second-chance programs give youth who have departure from formal education.63 dropped out of school a path to reengage in nontra- Remedial education interventions can work—if ditional learning environments, obtain secondary they reach the right people using the right approach.64 education equivalency qualifications, and enter job Effective remedial education interventions meet training.71 These programs offer a learning experience young people where they are, helping them transition that signals a level of achievement to participants, into careers. Remedial programs are more likely to their families, and employers. In Australia and the support students’ interests when they are short, rele- United States, early school leavers are encouraged vant to students’ lives, delivered by experienced teach- to enroll in programs that provide an equivalent to ers, and part of a long-term plan for career growth.65 an upper secondary diploma.72 Though equivalency Most evidence to date comes from programs in programs can improve employment, wages, and high-income countries, with three main types of other education indicators (relative to outcomes for interventions standing out as promising: individuals with no credentials), such impacts are often smaller than those for individuals holding •  Remedial prevention programs support academ- traditional educational credentials.73 Across second- ically weak students by strengthening their foun- chance interventions, socioemotional skills play an dational skills and encouraging them to complete a important role in student success—with skills formal education. such as the ability to work toward long-term goals There is no learning without prepared, motivated learners | 121 sometimes mattering more than the equivalency leads students to drop out.82 New accelerated remedial certificate itself. models addressing this problem include fast-track The demand for second-chance programs is high courses, self-paced modularized courses, and efforts and the evidence is promising, but keeping youth to mainstream students directly into postsecondary engaged in further education and training requires courses while providing additional instructional an integrated policy approach. In Sub-Saharan Africa, support. In the U.S. state of Indiana, a study of two there is a demand for programs to reengage early fast-track programs found participants achieve better school leavers, especially in low-income or conflict course pass rates and fewer course withdrawals than regions.74 But in practice, programs tend to be small, students in longer remedial programs.83 Similarly, and few operate within a policy framework that inte- evidence on self-paced modularized and mainstream- grates them into the formal education and training ing programs indicates that participants have higher systems.75 For low-income students, who usually postsecondary pass rates in math, complete more make up a disproportionate number of early school rigorous course requirements, and attempt tertiary leavers around the world, second-chance programs courses at higher rates than nonparticipants.84 like the Open Basic Programme in India can provide Contextualized instruction improves the effec- important pathways to educational opportunities.76 In tiveness of remedial education interventions, because India, Indonesia, the Philippines, and Thailand, equiv- learners benefit most when they engage, interpret, and alency programs for early school leavers improve stu- generate meaning from instructional content relevant dents’ self-development, especially when programs to their background.85 These models are designed are aligned with the formal education system.77 Sim- to reinforce foundational skills, while emphasizing ilarly, second-chance programs in Latin America and learners’ career aspirations.86 New approaches include the Caribbean yield better results when they take into contextualized vocational learning. An example that account the multidimensional needs of young people, blends foundational skills upgrading with occupa- connect students to pathways for further education tional training is the I-BEST (Integrated Basic Educa- and training opportunities, and provide support  to tion and Skills Training) program in the U.S. state of help participants return to productive adulthood.78 Washington. An evaluation of the program finds that participation has positive effects on student learning, Postsecondary remedial education including course credit accumulation, persistence in programs can help youth succeed in their tertiary education, and earning of occupational cer- programs of study tificates.87 Learning community approaches, which Many students enrolling in postsecondary education emphasize multisubject instruction, project-based and training are not prepared for the rigor of their work, and learner social interactions, also are showing programs of study. In Chile and Mexico, several promising results. In the United States, participation postsecondary institutions offer remedial support in these programs has a significantly positive relation- to academically underprepared students, but impact ship with a number of factors associated with student evaluations of such interventions are rare.79 In the success, such as level of course engagement, student United States, participation in postsecondary reme- and faculty interactions, or continuation to advanced dial education is widespread, often at great cost to courses.88 individuals and institutions.80 About 42 percent of Intensive student support can provide an institu- incoming students in two-year institutions and 20 tional safety net for at-risk youth. New approaches percent in four-year institutions enroll in remedial showing promising results include intensive tutor- courses at an annual cost of $1–$7 billion, depend- ing with supplemental instruction, intensive advis- ing on how the estimates are calculated. Due to this ing, and student success courses. Intensive tutoring high cost, U.S. institutions have been experimenting programs range from providing general academic with new approaches. There are three main types counseling and tutoring to offering special skills of remedial models that show promise: accelerated training.89 Evaluations of programs offering sus- remediation, contextualized instruction, and inten- tained tutoring show improvements in course com- sive student support.81 pletion and academic standing.90 Intensive personal- Accelerated remediation models reduce the ized advising services help students navigate course time students spend on remedial coursework. Con- selection and develop career plans. These services can ventional remedial education programs are often help students take advantage of other forms of sup- designed as a series of sequential courses that can port; beneficiaries are also more likely to complete take multiple semesters to complete, which often their remedial coursework and stay on in school after 122 | World Development Report 2018 program completion.91 Student success courses are Finally, recent developments in self-directed technol- usually stand-alone, credit-bearing courses for new ogy models are opening new opportunities for youth students that emphasize the development of study to work independently to meet their learning needs skills. Experimental evidence from the United States and upgrade their skills, but this remains a new area shows promising results on participants’ number of for remedial education research, and evidence on credits earned, classes passed, and class standing.92 their impacts is still sparse.93 Notes 1. UNICEF, WHO, and World Bank (2016). 4. Attanasio and others (2014); Denboba and others (2014). 3 2. Hungi (2010). 35. Britto and others (2016). 3. Cunha and others (2006). 36. Richter and others (2016). 4. Black and others (2008); Horton, Alderman, and Rivera 37. Devercelli, Sayre, and Denboba (2016). (2008); Thompson and Nelson (2001). 38. Serneels and Dercon (2014). 5. Coe and Lubach (2007); Garner and others (2012). 39. Oketch, Mutisya, and Sagwe (2012). 6. Center on the Developing Child (2016). 40. World Policy Analysis Center (various years). 7. Walker and others (2007). 41. Foko, Tiyab, and Husson (2012). 8. Bendini (2015). 42. Al-Samarrai and Zaman (2007); Bold, Kimenyi, and San- 9. Black and others (2017). defur (2013); Deininger (2003); Grogan (2009); Lucas and 10. Bradley and Corwyn (2005); McCoy and Raver (2014); Mbiti (2012); Nishimura, Yamano, and Sasaoka (2008). Shonkoff and others (2010). 43. Morgan, Petrosino, and Fronius (2012); Zuilkowski, 11. Farah and others (2006); McLoyd (1998). Jukes, and Dubeck (2016). 12. These numbers likely overestimate global preschool cov- 44. Duflo, Dupas, and Kremer (2017); Kremer, Miguel, and erage because many low-income countries do not report Thornton (2009). access data (Save the Children 2017). 45. Filmer and Schady (2008); Fiszbein and Schady (2009). 13. Black and others (2017). 46. J-PAL (2013). 14. Devercelli, Sayre, and Denboba (2016). 47. Avitabile and de Hoyos (2015); Jensen (2010); Loyalka and 15. ACPF (2011). others (2013); Nguyen (2008). 16. Berlinski and Schady (2015). 48. Beaman and others (2012); Jensen (2012). 17. Naudeau and others (2011). 49. Akresh and others (2012); Garg and Morduch (1998); 18. Black and others (2017). This is likely an underestimate Parish and Willis (1993). of the true number of young children who are at risk of 50. Akresh, de Walque, and Kazianga (2013). not reaching their developmental potential, given the 51. Avitabile and de Hoyos (2015); Barrera-Osorio and Filmer multiple risk factors associated with poverty. (2013); Nguyen (2008). 19. Bhutta and others (2013); Britto and others (2016). 52. Valerio and others (2016). 20. Galasso and Wagstaff (2016). 53. Zuilkowski, Jukes, and Dubeck (2016). 21. Galasso and Wagstaff (2016); Skoufias (2016). 54. Benin: Blimpo (2014); Kenya: Kremer, Miguel, and 22. Eilander and others (2010); Horta, Loret de Mola, and Thornton (2009); Mexico: Behrman and others (2015). Victora (2015). 55. Fryer (2011); Levitt and others (2016). 23. Galasso and Wagstaff (2016). 56. The primary completion cohort is approximated by 24. Almond and Currie (2011); Baker-Henningham and estimating the net intake rate to last grade in primary, López Bóo (2010). and the lower secondary completion rate is modeled as 25. Aboud and Yousafzai (2015); Britto and others (2016). a function of three components: primary completion 26. For a review, see World Bank (2015). rate, effective primary-to-lower-secondary transition 27. Britto and others (2016); World Bank (2015). rate, and net intake rate to last grade in lower secondary. 28. Denboba and others (2014). The regional upper secondary completion rates are esti- 29. Rahman and others (2013). mates from UNESCO’s World Inequality Database on 30. Berlinski, Galiani, and Gertler (2008); Engle and others Education (WIDE 2017), and the global upper secondary (2011); Favara and others (2017); García and others (2016); completion rate estimate is based on UNESCO’s 2015 Rao and others (2014). projection (UNESCO 2015). 31. Bouguen and others (2013); Rosero and Oosterbeek 57. OECD (2016); Roseth, Valerio, and Gutiérrez (2016). (2011). 58. Low proficiency is defined as level 1 and below on 32. Martinez, Naudeau, and Pereira (2012); Nakajima and OECD’s Programme for the International Assessment others (2016). of Adult Competencies (PIAAC) and the World Bank’s 33. However, center-based care can have the important added Skills Measurement Program (STEP) literacy assess- benefit of increasing labor force participation or further ments, and indicates limited understanding of basic skills acquisition among parents, especially mothers. texts. Medium to high proficiency is defined as level 2 There is no learning without prepared, motivated learners | 123 and above and indicates the ability to integrate, evaluate, 6. California Basic Skills Initiative (2009). 8 and interpret information from a variety and complexity 87. Jenkins, Zeidenberg, and Kienzl (2009). of text materials. 88. Engstrom and Tinto (2008); Visher and others (2010); 59. Estimates are based on marginal effect of literacy skills Zhao and Kuh (2004). at or above level 2 on the predicted probabilities of entry 89. Zachry Rutschow and Schneider (2011). into high-skill white-collar and blue-collar and other 90. Scrivener and others (2008); Zachry (2008). jobs (base outcome). The full specification includes 91. Bahr (2008); Visher, Butcher, and Cerna (2010). background control variables such as sex, age, education 92. Scrivener and others (2008); Scrivener, Sommo, and attainment, and proxies for family endowment. Collado (2009). 60. Zachry and Schneider (2010). 93. Zachry Rutschow and Schneider (2011). 61. Almeida, Johnson, and Steinberg (2006); NCES (2004). 62. de Hoyos, Rogers, and Székely (2016). Referencesa 63. Black, Polidano, and Tseng (2012); Tukundane and others Aboud, Frances, and Aisha Yousafzai. 2015. “Global Health (2014); Windisch (2015). and Development in Early Childhood.” Annual Review of 64. The terms “remedial education” and “developmental Psychology 66: 433–57. education” are often used interchangeably to describe ACPF (African Child Policy Forum). 2011. The African Report on programs aimed at supporting low-performing students Child Wellbeing: Budgeting for Children. Addis Ababa: ACPF. to enter and complete postsecondary and training pro- http://resourcecentre.savethechildren.se/sites/default grams. The Report uses the term “remedial education” /files/documents/3764.pdf. because it is a more widely recognized concept in low- Akresh, Richard, Emilie Bagby, Damien de Walque, and and middle-income countries. (See Bailey and others Harounan Kazianga. 2012. “Child Ability and Household 2010; Bailey, Bashford, and others 2016; Long and Boat- Human Capital Investment Decisions in Burkina Faso.” man 2013.) Economic Development and Cultural Change 61 (1): 157–86. 65. Post (2016). Akresh, Richard, Damien de Walque, and Harounan 66. “At-risk” students are defined here as having a higher Kazianga. 2013. “Cash Transfers and Child Schooling: propensity of dropping out of formal education or of not Evidence from a Randomized Evaluation of the Role of completing post-basic education and training programs. Conditionality.” Policy Research Working Paper 6340, 67. The typology used is from a systematic review of World Bank, Washington, DC. remedial (developmental) education by Rutschow and Almeida, Cheryl, Cassius Johnson, and Adria Steinberg. 2006. Crary-Ross (2014); Tukundane and others (2015); Wilson “Making Good on a Promise: What Policymakers Can Do and Tanner-Smith (2013); and Zachry Rutschow and to Support the Educational Persistence of Dropouts.” Schneider (2011). Double the Numbers Series (April), Jobs for the Future, 68. Gutiérrez and Rodrigo (2014); Lakshminarayana and Boston. others (2013). Almond, Douglas, and Janet Currie. 2011. “Human Capital 69. Howell, Kurlaender, and Grodsky (2010). Development before Age Five.” In Handbook of Labor 70. Karp and others (2008). Economics, Vol. 4, Part B, edited by Orley Ashenfelter and 71. Jepsen, Mueser, and Troske (2012). David Card, 1315–1486. Amsterdam: North-Holland. 72. De Witte and others (2013). Al-Samarrai, Samer, and Hassan Zaman. 2007. “Abolishing 73. Tyler and Lofstrom (2009). School Fees in Malawi: The Impact on Education Access 74. Inoue and others (2015). and Equity 1.” Education Economics 15 (3): 359–75. 75. Tukundane and others (2015). Attanasio, Orazio P., Camila Fernández, Emla O. A. Fitz­ 76. UNESCO (2010). simons, Sally M. Grantham-McGregor, Costas Meghir, 77. UNESCO (2010). and Marta Rubio-Codina. 2014. “Using the Infrastructure 78. Cunningham and others (2008). of a Conditional Cash Transfer Program to Deliver a Scal- 79. Cabrera (2013); Figueroa and others (2015); Micin and able Integrated Early Child Development Program in others (2015). Colombia: Cluster Randomized Controlled Trial.” BMJ 80. Bailey (2009); Clotfelter and others (2015); Scott-Clayton 349 (September 29): g5785. and Rodriguez (2014). Avitabile, Ciro, and Rafael E. de Hoyos. 2015. “The Hetero­ 81. Bailey, Jaggars, and Scott-Clayton (2013); Clotfelter geneous Effect of Information on Student Performance: and others (2015); Moss, Kelcey, and Showers (2014); Evidence from a Randomized Control Trial in Mexico.” Scott-Clayton and Rodriguez (2014). Policy Research Working Paper 7422, World Bank, Wash- 82. Bailey (2009); Scott-Clayton and Rodriguez (2014). ington, DC. 83. Brown and Ternes (2009). Bahr, Peter Riley. 2008. “Cooling Out in the Community College: What Is the Effect of Academic Advising on 84. Epper and Baker (2009); “Tennessee Board of Regents: Students’ Chances of Success?” Research in Higher Educa- Developmental Studies Redesign Initiative, Jackson tion 49 (8): 704–32. State Community College,” National Center for Aca- Bailey, Thomas R. 2009. “Challenge and Opportunity: demic Transformation, Saratoga Springs, NY. http:// Rethinking the Role and Function of Developmental www.thencat.org/States/TN/Abstracts/JSCC%20Algebra _Abstract.htm#FinalRpt. a.  References to titles of publications that include Taiwan refer 85. CSS (2007). to Taiwan, China. 124 | World Development Report 2018 Education in Community College.” New Directions for Berlinski, Samuel, and Norbert R. Schady, eds. 2015. The Community Colleges 2009 (145): 11–30. Early Years: Child Well-Being and the Role of Public Policy. Bailey, Thomas, Dong Wook Jeong, and Sung-Woo Cho. 2010. Development in the Americas Series. Washington, DC: “Referral, Enrollment, and Completion in Developmen- Inter-American Development Bank; New York: Palgrave tal Education Sequences in Community Colleges.” Eco- Macmillan. nomics of Education Review 29 (2): 255–70. Bhutta, Zulfiqar A., Jai K. Das, Arjumand Rizvi, Michelle F. Bailey, Thomas, Joanne Bashford, Angela Boatman, John Gaffey, Neff Walker, Susan Horton, Patrick Webb, et al. Squires, and Michael Weiss. 2016. Strategies for Postsecond- 2013. “Evidence-Based Interventions for Improvement of ary Students in Developmental Education: A Practice Guide Maternal and Child Nutrition: What Can Be Done and at for College and University Administrators, Advisors, and Fac- What Cost?” Lancet 382 (9890): 452–77. ulty. Washington, DC: U.S. Department of Education, Black, David, Cain Polidano, and Yi-Ping Tseng. 2012. “The National Center for Education Evaluation and Regional Re-engagement in Education of Early School Leavers.” Assistance, What Works Clearinghouse, Institute of Economic Papers 31 (2): 202–15. Education Sciences. Available at http://ies.ed.gov/ncee Black, Maureen M., Susan P. Walker, Lia C. H. Fernald, Chris- /wwc/Docs/PracticeGuide/wwc_dev_ed_112916.pdf. topher T. Andersen, Ann M. DiGirolamo, Chunling Lu, Bailey, Thomas R., Shanna Smith Jaggars, and Judith Dana Charles McCoy, et al. 2017. “Early Childhood Devel- Scott-Clayton. 2013. “Commentary: Characterizing the opment Coming of Age: Science through the Life Course.” Effectiveness of Developmental Education: A Response Lancet 389 (10064): 77–90. to Recent Criticism.” Journal of Developmental Education Black, Robert E., Lindsay H. Allen, Zulfiqar A. Bhutta, Laura 36 (3): 18–22, 24–25. E. Caulfield, Mercedes de Onis, Majid Ezzati, Colin Baker-Henningham, Helen, and Florencia López Bóo. 2010. Mathers, et al. 2008. “Maternal and Child Undernutri- “Early Childhood Stimulation Interventions in Develop- tion: Global and Regional Exposures and Health Conse- ing Countries: A Comprehensive Literature Review.” IDB quences.” Lancet 371 (9608): 243–60. Working Paper 213, Inter-American Development Bank, Blimpo, Moussa P. 2014. “Team Incentives for Education in Washington, DC. Developing Countries: A Randomized Field Experiment Barrera-Osorio, Felipe, and Deon Filmer. 2013. “Incentivizing in Benin.” American Economic Journal: Applied Economics Schooling for Learning: Evidence on the Impact of Alter- 6 (4): 90–109. native Targeting Approaches.” Policy Research Working Bold, Tessa, Mwangi S. Kimenyi, and Justin Sandefur. 2013. Paper 6541, World Bank, Washington, DC. “Public and Private Provision of Education in Kenya.” Beaman, Lori, Esther Duflo, Rohini Pande, and Petia Journal of African Economies 22 (supplement 2): ii39–ii56. Topalova. 2012. “Female Leadership Raises Aspirations Bouguen, Adrien, Deon Filmer, Karen Macours, and Sophie and Educational Attainment for Girls: A Policy Experi- Naudeau. 2013. “Impact Evaluation of Three Types of ment in India.” Science 335 (6068): 582–86. Early Childhood Development Interventions in Cambo- Behrman, Jere R., Susan W. Parker, Petra E. Todd, and Ken- dia.” Policy Research Working Paper 6540, World Bank, neth I. Wolpin. 2015. “Aligning Learning Incentives of Washington, DC. Students and Teachers: Results from a Social Experiment in Mexican High Schools.” Journal of Political Economy 123 Bradley, Robert H., and Robert F. Corwyn. 2005. “Caring for (2): 325–64. Children around the World: A View from Home.” Interna- Bendini, Maria Magdalena. 2015. “The Effect of Stress on tional Journal of Behavioral Development 29 (6): 468–78. Developmental Trajectories: Empirical Evidence from Britto, Pia Rebello, Stephen J. Lye, Kerrie Proulx, Aisha K. Peru.” Dissertation, University of Maryland, College Yousafzai, Stephen G. Matthews, Tyler Vaivada, Rafael Park, MD. Perez-Escamilla, et al. 2016. “Nurturing Care: Promoting Bentaouet Kattan, Raja. 2006. “Implementation of Free Basic Early Childhood Development.” Lancet 389 (10064): Education Policy.” Education Working Paper 7, World 91–102. Bank, Washington, DC. Brown, R., and R. Ternes. 2009. “Final Report to the Lilly Bergman, Peter. 2015. “Parent-Child Information Frictions Endowment Grant: Grant for Targeted and Accelerated and Human Capital Investment: Evidence from a Field Remediation.” Ivy Tech Community College, Indianapo- Experiment.” CESifo Working Paper 5391, Center for lis, IN. Economic Studies and Ifo Institute, Munich. Cabrera, Gabriela. 2013. “Programas de Apoyo a Transiciones Berlinski, Samuel, Matias Busso, Taryn Dinkelman, and Académicas del Sistema Escolarizado en la UNAM.” Claudia Martinez. 2016. “Reducing Parent-School Infor- Paper presented at Tercera Conferencia Latinoamericana mation Gaps and Improving Education Outcomes: sobre el Abandono en la Educación Superior, Mexico Evidence from High Frequency Text Messaging in City, November 13–15. Chile.” Working paper, Abdul Latif Jameel Poverty California Basic Skills Initiative. 2009. “Contextualized Action Lab, Massachusetts Institute of Technology, Teaching and Learning, a Faculty Primer: A Review of Cambridge, MA. Literature and Faculty Practices with Implications for Berlinski, Samuel, Sebastian Galiani, and Paul J. Gertler. California Community College Practitioners.” Research 2008. “The Effect of Pre-primary Education on Primary and Planning Group, Academic Senate for California School Performance.” Journal of Public Economics 93 (1–2): Community Colleges, and Bay Area Workforce Funding 219–34. Collaborative, San Rafael, CA. There is no learning without prepared, motivated learners | 125 Carneiro, Pedro, Flavio Cunha, and James J. Heckman. 2003. Dowd, Amy Jo, Elliott Friedlander, Christine Jonason, Jane “Interpreting the Evidence of Family Influence on Child Leer, Lisa Zook Sorensen, Jarrett Guajardo, Nikhit D’Sa, Development.” Paper presented at Federal Reserve Bank et al. 2017. “Lifewide Learning for Early Reading Develop- of Minneapolis and McKnight Foundation’s conference, ment.” New Directions for Child and Adolescent Development “Economics of Early Childhood Development: Lessons 155: 31–49. for Economic Policy,” Minneapolis, October 17. Duflo, Esther, Pascaline Dupas, and Michael R. Kremer. Center on the Developing Child. 2016. “From Best Practices 2017. “The Impact of Free Secondary Education: Experi- to Breakthrough Impacts: A Science-Based Approach to mental Evidence from Ghana.” Paper presented at Building a More Promising Future for Young Children Ghana Education Evidence Summit 2017, “Towards and Families.” Center on the Developing Child, Harvard Quality Education in Ghana: Using Evidence to Achieve University, Cambridge, MA. Better Learning Outcomes,” Accra, Ghana, March 28. Clotfelter, Charles T., Helen F. Ladd, Clara Muschkin, and Eilander, Ans, Tarun Gera, Harshpal S. Sachdev, Catherine Jacob L. Vigdor. 2015. “Developmental Education in Transler, Henk C. M. van der Knaap, Frans J. Kok, and North Carolina Community Colleges.” Educational Evalua- Saskia J. M. Osendarp. 2010. “Multiple Micronutrient tion and Policy Analysis 37 (3): 354–75. Supplementation for Improving Cognitive Performance Coe, Christopher L., and Gabrielle R. Lubach. 2007. “Mother- in Children: Systematic Review of Randomized Con- Infant Interactions and the Development of Immunity trolled Trials.” American Journal of Clinical Nutrition 91 (1): from Conception through Weaning.” In Psychoneuro­ 115–30. immunology, edited by Robert Ader, 455–74. Burlington, Engle, Patrice L., Lia C. H. Fernald, Harold Alderman, Jere R. MA: Elsevier Academic Press. Behrman, Chloe O’Gara, Aisha Yousafzai, Meena Cabral CSS (Center for Student Success). 2007. Basic Skills as a Foun- de Mello, et al. 2011. “Strategies for Reducing Inequalities dation for Student Success in California Community Colleges. and Improving Developmental Outcomes for Young San Rafael, CA: CSS, Research and Planning Group. Children in Low-Income and Middle-Income Countries.” Cunha, Flavio, James J. Heckman, Lance J. Lochner, and Dim- Lancet 378 (9799): 1339–53. itriy V. Masterov. 2006. “Interpreting the Evidence on Engstrom, Cathy McHugh, and Vincent Tinto. 2008. “Learning Life Cycle Skill Formation.” In Handbook of the Economics Better Together: The Impact of Learning Communities on of Education, Vol. 1, edited by Eric A. Hanushek and Finis the Persistence of Low-Income Students.” Opportunity Mat- Welch, 697–812. Handbooks in Economics Series 26. ters 1: 5–21. Amsterdam: North-Holland. Epper, Rhonda M., and Elaine D. Baker. 2009. “Technology Cunningham, Wendy V., Linda McGinnis, Rodrigo García Solutions for Developmental Math: An Overview of Verdú, Cornelia Tesliuc, and Dorte Verner. 2008. Youth at Current and Emerging Practices.” Journal of Developmental Risk in Latin America and the Caribbean: Understanding the Education 26 (2): 4–23. Causes, Realizing the Potential. Directions in Development: Farah, Martha J., David M. Shera, Jessica H. Savage, Laura Human Development Series. Washington, DC: World Betancourt, Joan M. Giannetta, Nancy L. Brodsky, Elsa K. Bank. Malmud, et al. 2006. “Childhood Poverty: Specific Associ- de Hoyos, Rafael E., Halsey Rogers, and Miguel Székely. 2016. ations with Neurocognitive Development.” Brain Research “Out of School and Out of Work: Risk and Opportunities 1110 (1): 166–74. for Latin America’s Ninis.” World Bank, Washington, DC. Favara, Marta, Martin Woodhead, Juan Francisco Castro, Deininger, Klaus. 2003. “Does Cost of Schooling Affect Grace Chang, and Patricia Espinoza. 2017. “Pre-school Enrollment by the Poor? Universal Primary Education in Education and Skills Development in Peru, Vietnam, Uganda.” Economics of Education Review 22 (3): 291–305. Ethiopia, and India: Evidence from Young Lives.” World Denboba, Amina D., Rebecca K. Sayre, Quentin T. Wodon, Bank, Washington, DC. Leslie K. Elder, Laura B. Rawlings, and Joan Lombardi. Figueroa, Lorna, Bernardita Maillard, Nelson Veliz, Samara 2014. “Stepping Up Early Childhood Development: Toledo, and Máximo González. 2015. “La Experiencia de Investing in Young Children for High Returns.” October, los Programas Propedéuticos y su Articulación con la Children’s Investment Fund Foundation and World Escuela.” Paper presented at Quinta Conferencia Latino­ Bank, Washington, DC. americana sobre el Abandono en la Educación Superior, Devercelli, Amanda E., Rebecca K. Sayre, and Amina D. Talca, Chile, November 11–13. Denboba. 2016. “What Do We Know about Early Child- Filmer, Deon, and Norbert R. Schady. 2008. “Getting Girls hood Development Policies in Low and Middle Income into School: Evidence from a Scholarship Program in Countries?” SABER-ECD Brief Note 1, World Bank, Wash- Cambodia.” Economic Development and Cultural Change ington, DC. 56 (3): 581–617. De Witte, Kristof, Sofie Cabus, Geert Thyssen, Wim Groot, Fiszbein, Ariel, and Norbert R. Schady. 2009. Conditional Cash and Henriëtte Maassen van den Brink. 2013. “A Critical Transfers: Reducing Present and Future Poverty. With Fran- Review of the Literature on School Dropout.” Educational cisco H. G. Ferreira, Margaret E. Grosh, Niall Keleher, Research Review 10: 13–28. Pedro Olinto, and Emmanuel Skoufias. World Bank Pol- Dizon-Ross, Rebecca. 2016. “Parents’ Beliefs and Children’s icy Research Report. Washington, DC: World Bank. Education: Experimental Evidence from Malawi.” Work- Foko, Borel, Beifith Kouak Tiyab, and Guillaume Husson. ing paper, Booth School of Business, University of 2012. “Household Education Spending: An Analytical Chicago. and Comparative Perspective for 15 African Countries.” 126 | World Development Report 2018 Working paper, Pôle de Dakar for Education Sector State Community and Technical College System’s Inte- Analysis, Regional Bureau for Education in Africa, grated Basic Education and Skills Training Program: United Nations Educational, Scientific, and Cultural Findings from a Multivariate Analysis.” CCRC Working Organization, Dakar, Senegal. Paper 16, Community College Research Center, Teachers Friedlander, Elliott, and Claude Goldenberg, eds. 2016. Liter- College, Columbia University, New York. acy Boost in Rwanda: Impact Evaluation of a Two Year Ran- Jensen, Robert T. 2010. “The (Perceived) Returns to Educa- domized Control Trial. Stanford, CA: Stanford University. tion and the Demand for Schooling.” Quarterly Journal of Fryer, Roland G. 2011. “Financial Incentives and Student Economics 125 (2): 515–48. Achievement: Evidence from Randomized Trials.” Quar- — —— —. 2012. “Do Labor Market Opportunities Affect Young terly Journal of Economics 126 (4): 1755–98. Women’s Work and Family Decisions? Experimental Galasso, Emanuela, and Adam Wagstaff. 2016. “The Eco- Evidence from India.” Quarterly Journal of Economics nomic Costs of Stunting and How to Reduce Them.” 127 (2): 753–92. With Sophie Naudeau and Meera Shekar. Policy Jepsen, Christopher, Peter R. Mueser, and Kenneth R. Troske. Research Note 5, World Bank, Washington, DC. 2012. “Labor-Market Returns to the GED Using Regres- García, Jorge Luis, James J. Heckman, Duncan Ermini Leaf, sion Discontinuity Analysis.” IZA Discussion Paper 6758, and María José Prados. 2016. “The Life-Cycle Benefits of Institute for the Study of Labor, Bonn, Germany. an Influential Early Childhood Program.” NBER Work- Johnson, Rucker C., and C. Kirabo Jackson. 2017. “Reducing ing Paper 22993, National Bureau of Economic Research, Inequality through Dynamic Complementarity: Evi- Cambridge, MA. dence from Head Start and Public School Spending.” Garg, Ashish, and Jonathan Morduch. 1998. “Sibling Rivalry NBER Working Paper 23489, National Bureau of Eco- and the Gender Gap: Evidence from Child Health Out- nomic Research, Cambridge, MA. comes in Ghana.” Journal of Population Economics 11 (4): J-PAL (Abdul Latif Jameel Poverty Action Lab). 2013. “Inform- 471–93. ing Future Choices.” J-PAL Policy Briefcase, J-PAL, Massa- Garner, Andrew S., Jack P. Shonkoff, Benjamin S. Siegel, chusetts Institute of Technology, Cambridge, MA. Mary I. Dobbins, Marian F. Earls, Laura McGuinn, John Karp, Melinda Jane Mechur, Juan Carlos Calcagno, Kather- Pascoe, et al. 2012. “Early Childhood Adversity, Toxic ine Lee Hughes, Dong Wook Jeong, and Thomas R. Bai- Stress, and the Role of the Pediatrician: Translating ley. 2008. “Dual Enrollment Students in Florida and New Developmental Science into Lifelong Health.” Pediatrics York City: Postsecondary Outcomes.” CCRC Brief 37, 129 (1): e224–e231. Community College Research Center, Teachers College, Grogan, Louise. 2009. “Universal Primary Education and Columbia University, New York. School Entry in Uganda.” Journal of African Economies Klees, Steven J. 2017. “Will We Achieve Education for All and 18 (2): 183–211. the Education Sustainable Development Goal?” Compara- Gutiérrez, Emilio, and Rodmiro Rodrigo. 2014. “Closing the tive Education Review 61 (2): 425–40. Achievement Gap in Mathematics: Evidence from a Kremer, Michael R., Edward Miguel, and Rebecca L. Thorn- Remedial Program in Mexico City.” Latin American ton. 2009. “Incentives to Learn.” Review of Economics and Economic Review 23 (14): 1–30. Statistics 91 (3): 437–56. Horta, Bernardo L., Christian Loret de Mola, and Cesar G. Lakshminarayana, Rashmi, Alex Eble, Preetha Bhakta, Chris Victora. 2015. “Breastfeeding and Intelligence: A System- Frost, Peter Boone, Diana Elbourne, and Vera Mann. atic Review and Meta-Analysis.” Acta Paediatrica 104 2013. “The Support to Rural India’s Public Education (S467): 14–19. System (STRIPES) Trial: A Cluster Randomised Con- Horton, Sue, Harold Alderman, and Juan A. Rivera. 2008. trolled Trial of Supplementary Teaching, Learning Mate- “The Challenge of Hunger and Malnutrition.” Copen­ rial and Material Support.” PLoS ONE 8 (7): e65775. hagen Consensus 2008 Challenge Paper, Copenhagen Levitt, Steven D., John A. List, Susanne Neckermann, and Consensus Center, Tewksbury, MA. Sally Sadoff. 2016. “The Behavioralist Goes to School: Howell, Jessica S., Michal Kurlaender, and Eric Grodsky. Leveraging Behavioral Economics to Improve Educa- 2010. “Postsecondary Preparation and Remediation: tional Performance.” American Economic Journal: Economic Examining the Effect of the Early Assessment Program Policy 8 (4): 183–219. at California State University.” Journal of Policy Analysis Lieberman, Evan S., Daniel N. Posner, and Lily L. Tsai. 2014. and Management 29 (4): 726–48. “Does Information Lead to More Active Citizenship? Hungi, Njora. 2010. “What Are the Levels and Trends in Evidence from an Education Intervention in Rural Grade Repetition?” SACMEQ Policy Issues 5, Southern Kenya.” World Development 60: 69–83. and Eastern Africa Consortium for Monitoring Educa- Long, Bridget T., and Angela Boatman. 2013. “The Role of tional Quality, Paris. Remedial and Developmental Courses in Access and Inoue, Keiko, Emanuela di Gropello, Yesim Sayin Taylor, and Persistence.” In The State of College Access and Completion: James Gresham. 2015. Out-of-School Youth in Sub-Saharan Improving College Success for Students from Underrepresented Africa: A Policy Perspective. Directions in Development: Groups, edited by Laura W. Perna and Anthony P. Jones, Human Development Series. Washington, DC: World 77–95. New York: Routledge. Bank. Loyalka, Prashant, Chengfang Liu, Yingquan Song, Hong- Jenkins, Davis, Matthew Zeidenberg, and Gregory S. Kienzl. mei Yi, Xiaoting Huang, Jianguo Wei, Linxiu Zhang, et al. 2009. “Educational Outcomes of I-BEST, Washington 2013. “Can Information and Counseling Help Students There is no learning without prepared, motivated learners | 127 from Poor Rural Areas Go to High School? Evidence from Nguyen, Trang. 2008. “Information, Role Models, and Per- China.” Journal of Comparative Economics 41 (4): 1012–25. ceived Returns to Education: Experimental Evidence Lucas, Adrienne M., and Isaac M. Mbiti. 2012. “Access, Sort- from Madagascar.” MIT working paper, Massachusetts ing, and Achievement: The Short-Run Effects of Free Institute of Technology, Cambridge, MA. Primary Education in Kenya.” American Economic Journal: Nishimura, Mikiko, Takashi Yamano, and Yuichi Sasaoka. Applied Economics 4 (4): 226–53. 2008. “Impacts of the Universal Primary Education Pol- Martin, Paul. 2012. “Responsabilidad Social Corporativa y icy on Educational Attainment and Private Costs in Rural Primera Infancia.” Paper presented at Ministry of Devel- Uganda.” International Journal of Educational Development opment and Social Inclusion’s Semana de la Inclusión, 28 (2): 161–75. Lima, October 21–24. OECD (Organisation for Economic Co-operation and Devel- Martinez, Sebastian, Sophie Naudeau, and Vitor Pereira. opment). 2016. Skills Matter: Further Results from the Survey 2012. “The Promise of Preschool in Africa: A Randomized of Adult Skills. OECD Skills Studies Series. Paris: OECD. Impact Evaluation of Early Childhood Development in Oketch, Moses, Maurice Mutisya, and Jackline Sagwe. 2012. Rural Mozambique.” Save the Children, Fairfield, CT, “Parental Aspirations for Their Children’s Educational February 14. World Bank, Washington, DC. http://site Attainment and the Realisation of Universal Primary resources.worldbank.org/INTAFRICA/Resources/The Education (UPE) in Kenya: Evidence from Slum and _Promise_of_Preschool_in_Africa_ECD_REPORT.pdf. Non-slum Residences.” International Journal of Educational McCoy, Dana Charles, and C. Cybele Raver. 2014. “Household Development 32 (6): 764–72. Instability and Self-Regulation among Poor Children.” Oye, Mari, Lant Pritchett, and Justin Sandefur. 2016. “Girls’ Journal of Children and Poverty 20 (2): 131–52. Schooling Is Good, Girls’ Schooling with Learning Is McLoyd, Vonnie C. 1998. “Socioeconomic Disadvantage and Better.” Education Commission, Center for Global Devel- Child Development.” American Psychologist 53 (2): 185–204. opment, Washington, DC. Parish, William L., and Robert J. Willis. 1993. “Daughters, Micin, Sonia, Natalia Farías, Beatriz Carreño, and Sergio Education, and Family Budgets: Taiwan Experiences” Urzúa. 2015. “Beca Nivelación Académica: La Experiencia [refers to Taiwan, China]. Journal of Human Resources 28 de una Política Pública Aplicada en una Universidad (4): 863–98. Chilena.” Calidad en la Educación 42: 189–208. Phillips, Deborah A., Mark W. Lipsey, Kenneth A. Dodge, Ron Morgan, Claire, Anthony Petrosino, and Trevor Fronius. Haskins, Daphna Bassok, Margaret R. Burchinal, Greg J. 2012. “A Systematic Review of the Evidence of the Impact Duncan, et al. 2017. “Puzzling It Out: The Current State of of Eliminating School User Fees in Low-Income Develop- Scientific Knowledge on Pre-kindergarten Effects, a Con- ing Countries.” Evidence for Policy and Practice Informa- sensus Statement.” Center for Child and Family Policy, tion and Co-ordinating Centre, Social Science Research Duke University, Durham, NC; Brookings Institution, Unit, Institute of Education, University of London, Washington, DC. London. Post, David. 2016. “Adult Literacy Benefits? New Opportuni- Moss, Brian G., Ben Kelcey, and Nancy Showers. 2014. “Does ties for Research into Sustainable Development.” Interna- Classroom Composition Matter? College Classrooms as tional Review of Education 62 (6): 751–70. Moderators of Developmental Education Effectiveness.” Rahman, Atif, Jane Fisher, Peter Bower, Stanley Luchters, Community College Review 42 (3): 201–20. Thach Tran, M. Taghi Yasamy, Shekhar Saxena, et al. Nakajima, Nozomi, Amer Hasan, Haeil Jung, Sally Anne 2013. “Interventions for Common Perinatal Mental Dis- Brinkman, Menno Prasad Pradhan, and Angela Kinnell. orders in Women in Low- and Middle-Income Countries: 2016. “Investing in School Readiness: An Analysis of the A Systematic Review and Meta-Analysis.” Bulletin of the Cost-Effectiveness of Early Childhood Education Path- World Health Organization 91 (8): 593–601. ways in Rural Indonesia.” Policy Research Working Rao, Nirmala, Jin Sun, Jessie M. S. Wong, Brendan Weekes, Paper 7832, World Bank, Washington, DC. Patrick Ip, Sheldon Shaeffer, Mary Young, et al. 2014. Naudeau, Sophie, Naoko Kataoka, Alexandria Valerio, “Early Childhood Development and Cognitive Develop- Michelle J. Neuman, and Leslie Kennedy Elder. 2011. ment in Developing Countries: Education Rigorous Investing in Young Children: An Early Childhood Development Literature Review.” Faculty of Education, University of Guide for Policy Dialogue and Project Preparation. Directions Hong Kong, Hong Kong SAR, China. in Development: Human Development Series. Washing- Richter, Linda M., Bernadette Daelmans, Joan Lombardi, Jody ton, DC: World Bank. Heymann, Florencia López Bóo, Jere R. Behrman, Chun- NCES (National Center for Education Statistics). 2004. ling Lu, et al. 2016. “Investing in the Foundation of Sus- “Educational Attainment of High School Dropouts 8 tainable Development: Pathways to Scale Up for Early Years Later.” Issue Brief NCES 2005–026, NCES, Institute Childhood Development.” Lancet 389 (10064): 103–18. of Education Sciences, U.S. Department of Education, Rogers, Todd, and Avi Feller. 2016. “Intervening through Washington, DC. Influential Third Parties: Reducing Student Absences at Nelson, Charles A., Nadine Gaab, Yingying Wang, Swapna Scale via Parents.” Working paper, Harvard University, Kumar, Danielle Sliva, Meaghan Mauer, Alissa Wester- Cambridge, MA. lund, et al. 2017. “Atypical Brain Development in Bangla- Rosero, José, and Hessel Oosterbeek. 2011. “Trade-Offs deshi Infants Exposed to Profound Early Adversity.” Pre- between Different Early Childhood Interventions: sented at conference of Society for Research in Child Evidence from Ecuador.” Tinbergen Institute Discus- Development, Austin, TX, April. sion Paper TI 2011–102/3, Faculty of Economics and 128 | World Development Report 2018 Business, University of Amsterdam; Tinbergen Institute, UNESCO (United Nations Educational, Scientific, and Cul- Amsterdam. tural Organization). 2010. “Achieving EFA through Equiv- Roseth, Viviana V., Alexandria Valerio, and Marcela Gutiérrez. alency Programmes in Asia-Pacific: A Regional Overview 2016. Education, Skills, and Labor Market Outcomes: Results with Highlights from India, Indonesia, Thailand, and from Large-Scale Adult Skills Surveys in Urban Areas in 12 the Philippines.” Asia and Pacific Regional Bureau for Countries. STEP Skills Measurement Series. Washington, Education, UNESCO, Bangkok. DC: World Bank. —— ——. 2015. “How Long Will It Take to Achieve Universal Rutschow, Elizabeth Zachry, and Shane Crary-Ross. 2014. Primary and Secondary Education?” Technical back- “Beyond the GED: Promising Models for Moving High ground note for the Framework for Action on the post- School Dropouts to College.” MDRC, New York. 2015 education agenda. http://en.unesco.org/gem-report Save the Children. 2017. “Windows into Early Learning and /how-long-will-it-take-achieve-universal-primary-and Development: Cross Country IDELA Findings Fueling -secondary-education. Progress on ECD Access, Quality, and Equity.” Save the UNICEF (United Nations Children’s Fund), WHO (World Children International, London. Health Organization), and World Bank. 2016. “Levels and Scott-Clayton, Judith, and Olga Rodriguez. 2014. “Develop- Trends in Child Malnutrition: UNICEF/WHO/World ment, Discouragement, or Diversion? New Evidence on Bank Group Joint Child Malnutrition Estimates, Key the Effects of College Remediation Policy.” Education Findings of the 2016 Edition.” UNICEF, New York; WHO, Finance and Policy 10 (1): 4–45. Geneva; World Bank, Washington, DC. http://www.who Scrivener, Susan, Dan Bloom, Allen LeBlanc, Christina Pax- .int/nutgrowthdb/estimates2015/en/. son, Cecilia Elena Rouse, and Colleen Sommo. 2008. “A Valerio, Alexandria, María Laura Sánchez Puerta, Namrata Good Start: Two-Year Effects of a Freshmen Learning Raman Tognatta, and Sebastián Monroy-Taborda. 2016. Community Program at Kingsborough Community Col- “Are There Skills Payoffs in Low- and Middle-Income lege.” With Jenny Au, Jedediah J. Teres, and Susan Yeh. Countries? Empirical Evidence Using STEP Data.” Policy Opening Doors Project, MDRC, New York, March. Research Working Paper 7879, World Bank, Washington, Scrivener, Susan, Colleen Sommo, and Herbert Collado. DC. 2009. “Getting Back on Track: Effects of a Community Visher, Mary G., Kristin F. Butcher, and Oscar S. Cerna. 2010. College Program for Probationary Students.” Opening “Guiding Developmental Math Students to Campus Ser- Doors Project, MDRC, New York. vices: An Impact Evaluation of the Beacon Program at Serneels, Pieter, and Stefan Dercon. 2014. “Aspirations, Pov- South Texas College.” With Dan Cullinan and Emily erty, and Education: Evidence from India.” Young Lives Schneider. Report, MDRC, New York. Working Paper 125, Young Lives, Oxford Department Visher, Mary G., Emily Schneider, Heather Wathington, and of International Development, University of Oxford, Herbert Collado. 2010. “Scaling Up Learning Communi- Oxford, U.K. ties: The Experience of Six Community Colleges.” Report, Shonkoff, Jack, Pat Levitt, W. T. Boyce, Judy Cameron, Greg National Center for Postsecondary Research, Teachers Duncan, N. A. Fox, Megan Gunnar, et al. 2010. “Persistent College, Columbia University, New York. Fear and Anxiety Can Affect Young Children’s Learning and Development.” Working Paper 9, Center on the Walker, Susan P., Theodore D. Wachs, Julie Meeks Gardner, Developing Child, Harvard University, Cambridge, MA. Betsy Lozoff, Gail A. Wasserman, Ernesto Pollitt, Julie A. Skoufias, Emmanuel. 2016. “Synergies in Child Nutrition: Carter, et al. 2007. “Child Development: Risk Factors for Interactions of Food Security, Health and Environment, Adverse Outcomes in Developing Countries.” Lancet 369 and Child Care.” Policy Research Working Paper 7794, (9556): 145–57. World Bank, Washington, DC. Whitebread, David, Martina Kuvalja, and Aileen O’Connor. Thompson, Ross A., and Charles A. Nelson. 2001. “Develop- 2015. “Quality in Early Childhood Education: An Interna- mental Science and the Media: Early Brain Develop- tional Review and Guide for Policy Makers.” With contri- ment.” American Psychologist 56 (1): 5–15. butions from Qatar Academy. WISE 20, World Innova- Tukundane, Cuthbert, Alexander Minnaert, Jacques Zeelen, tion Summit for Education, Qatar Foundation, Doha. and Peter Kanyandago. 2015. “A Review of Enabling Fac- WIDE (World Inequality Database on Education). 2017. Com- tors in Support Intervention Programmes for Early School pletion indicators. http://www.education-inequalities.org. Leavers: What Are the Implications for Sub-Saharan Wilson, Sandra Jo, and Emily E. Tanner-Smith. 2013. “Drop- Africa?” Children and Youth Services Review 52: 54–62. out Prevention and Intervention Programs for Improv- Tukundane, Cuthbert, Jacques Zeelen, Alexander Minnaert, ing School Completion among School-Aged Children and Peter Kanyandago. 2014. “  ‘I Felt Very Bad, I Had and Youth: A Systematic Review.” Journal of the Society for Self-Rejection’: Narratives of Exclusion and Marginalisa- Social Work and Research 4 (4): 357–72. tion among Early School Leavers in Uganda.” Journal of Windisch, Hendrickje Catriona. 2015. “Adults with Low Youth Studies 17 (4): 475–91. Literacy and Numeracy Skills: A Literature Review on Tyler, John H., and Magnus Lofstrom. 2009. “Finishing High Policy Intervention.” OECD Education Working Paper School: Alternative Pathways and Dropout Recovery.” 123, Organisation for Economic Co-operation and Devel- Future of Children 19 (1): 77–103. opment, Paris. UIS (UNESCO Institute of Statistics). 2017. Education indica- World Bank. 2015. World Development Report 2015: Mind, tors. http://data.uis.unesco.org. Society, and Behavior. Washington, DC: World Bank. There is no learning without prepared, motivated learners | 129 ————. 2017. World Development Indicators (database). Rigorous Research and Promising Trends in Develop- World Bank, Washington, DC. http://data.worldbank mental Education.” NCPR working paper, National Cen- .org/data-catalog/world-development-indicators. ter for Postsecondary Research, Teachers College, World Policy Analysis Center. Various years. Is Education Columbia University, New York. Tuition-Free? (database). World Policy Analysis Center, Zachry Rutschow, Elizabeth M., and Emily Schneider. 2011. Fielding School of Public Health, University of Califor- “Unlocking the Gate: What We Know About Improving nia, Los Angeles. http://www.worldpolicycenter.org Developmental Education.” MDRC, New York, June. /policies/is-education-tuition-free/is-beginning-secondary Zhao, Chun-Mei, and George D. Kuh. 2004. “Adding Value: -education-tuition-free. Learning Communities and Student Engagement.” Zachry, Elizabeth M. 2008. “Promising Instructional Research in Higher Education 45 (2): 115–38. Reforms in Developmental Education: A Case Study Zuilkowski, Stephanie Simmons, Matthew C. H. Jukes, and of Three Achieving the Dream Colleges.” With Emily Margaret M. Dubeck. 2016. “  ‘I Failed, No Matter How Schneider. MDRC, New York. Hard I Tried’: A Mixed-Methods Study of the Role of Zachry, Elizabeth M., and Emily Schneider. 2010. “Building Achievement in Primary School Dropout in Rural Kenya.” Foundations for Student Readiness: A Review of International Journal of Educational Development 50: 100–07. 130 | World Development Report 2018 Teacher skills and motivation both matter (though many education systems act like they don’t) 6 For students to learn, teachers have to teach effectively—but many education systems pay little attention to what teachers know or what they do in the classroom. Focusing on teachers’ skills and motivation can pay off. After prepared and motivated learners, equipped Most teacher training and motivated teachers are the most fundamental ingredient of learning. Teachers are also the largest is ineffective, but some budget item, with their salaries accounting for over approaches work three-quarters of the education budget at the primary level in low- and middle-income countries.1 Yet many In-service professional development requires signifi- education systems put in classrooms teachers who cant time and resources. A survey of 38 developed and have little mastery of the subjects they are to teach— developing countries found that 91 percent of teach- especially in classrooms serving poor children.2 Once ers had participated in professional development in in place, most teachers take part in some professional the previous 12 months.4 Two-thirds of World Bank development, but much of it is inconsistent and overly projects with an education component in the last theoretical. Meanwhile, education systems often lack decade incorporated teacher professional develop- effective mechanisms to mentor and motivate teach- ment. Developing countries spend many millions a ers.3 Such failures can be illuminated through models year to strengthen teachers.5 of human behavior—which also point to solutions But a lot of teacher professional development goes (table 6.1). A synthesis of the evidence in these areas unevaluated—and much of it may be ineffective. One reveals three principles that are key to achieving team of teacher training experts in the United States learning success through teachers: characterized professional development in the coun- try as “episodic, myopic, and often meaningless.”6 • To be effective, teacher training needs to be individ- Teacher training in low- and middle-income coun- ually targeted and repeated, with follow-up coach- tries is often short and of low quality.7 Countries often ing, often around a specific pedagogical technique. have many training programs under way at the same • To avoid learners falling behind to the point where time—in some cases dozens—with little to show for they cannot catch up, teaching needs to be pitched them (box 6.1). to the level of the student. Though preservice teacher training is important • Increasing teacher motivation with incentives can in providing basic skills (box 6.2), evidence on teacher increase learning if the incentivized actions are training credentials is mixed. Much of the limited within teachers’ capacity and if the failure to per- evidence on teacher credentials, generally from high- form those actions has impeded learning. income countries, indicates they have no or extremely Teacher skills and motivation both matter (though many education systems act like they don’t) | 131 Table 6.1 Models of human behavior can guide actions to improve teaching: Some examples Models that identify a Approaches that address the Synthesis principle Where this fails mechanism behind this failure modeled mechanism Provide individually Much teacher training is Simple optimization (by In India, a program with limited targeted and repeated one-off, with little to no government) with information preservice but repeated follow- teacher training, with follow-up coaching in the failure: Follow-up coaching is more up for community teachers led to follow-up coaching. classroom. costly than centrally delivered sizable learning gains. training, and centrally delivered training may give the impression of In the United States, programs effectiveness by changing teacher associated with a specific knowledge but not practice. pedagogical technique were General pedagogical training twice as effective as general may be cheaper than training in pedagogical training. specific techniques, and evidence on relative effectiveness is recent. Pitch teaching to the In many countries, most Information failure: Policy In India and Kenya, reorganizing level of the student. students fall far behind the makers may have an imperfect classes by ability improved curriculum, and, facing understanding of how little many learning. large, heterogeneous students are learning. classes, teachers have Behavioral (mental models): In India, complementing teachers difficulty teaching at a Teachers may believe that lower- with dynamic computer-assisted level that allows students performing learners cannot learning programs that adapt to to learn. succeed; curriculums may be learners’ ability levels improved optimistically pitched higher than math ability. Teachers receive most students can keep up with. explicit guidance to teach students at their level. Strengthen teacher In many systems, teachers Principal-agent: If the education Teacher financial incentives have motivation by have few incentives system signals that learning is not been effective in countries with incentivizing actions (financial or professional) valued, teachers will not have the high absenteeism, such as India that are within teachers’ for good performance same incentives as students and and Kenya. capacity and that are beyond their intrinsic parents have. essential to learning. motivation. Source: WDR 2018 team. small effects on student learning.8 Simple statistical Specificity means teacher training programs are most associations across francophone Africa suggest a pos- effective when they teach pedagogy specific to a sub- itive relationship between teacher preparation and ject area (say, how to effectively teach a mathematics student performance, but that relationship could be class). Continuity means teachers receive significant driven by other factors, such as strategic placement continual support—not one-off workshops.12 of good teachers in desirable areas (where students In teacher training programs, the inclusion of would perform well in any case).9 Preparing teachers follow-up visits in school leads to higher learning gains. better is crucial, but the political economy challenges To bridge the gap between learning new methods in to doing so may be greater than for in-service training, training and implementing them in practice, devel- and the evidence is more limited. The same principles oping countries should make more use of follow-up that lead to effective in-service training serve as use- visits in which trainers observe and support teachers ful starting points for improving preservice training. in the classroom.13 In Africa, a range of programs Is there hope for in-service training or professional with long-term teacher mentoring and coaching has development? Decidedly yes. Experience from high- shown sizable learning effects.14 In India, a program income countries shows that practicality, specificity, that provided little initial training to teachers but and continuity are key to effective teacher profes- then provided support throughout the year signifi- sional development.10 Practicality means teachers are cantly increased both math and language ability, with trained using concrete methods as opposed to theoret- the largest gains for those students who were per- ical constructs, and the training is classroom-based.11 forming poorly at the outset.15 Teachers in Shanghai, 132 | World Development Report 2018 Box 6.1 The landscape of in-service teacher training The quality of in-service teacher training varies dramatically across five regions, the median number of follow-up visits across countries, but much of the training does not align is fewer than one per teacher. Many in-service training pro- with practices that are associated with better student per- grams (50 percent among a sample of programs) evaluate formance.a One good practice of in-service teacher training their success based on teacher knowledge at the end of the involves follow-up visits to teachers’ classrooms to provide training; far fewer (25 percent) seek to assess their impact ongoing support. Among 100 teacher training programs on student learning.b Source: WDR 2018 team. a. Popova, Evans, and Arancibia (2016). b. Popova, Breeding, and Evans (2017). Box 6.2 What works in preservice teacher training? In New York City, teachers who participated in teacher learning for students.b This finding calls into question education programs that focused on practical classroom the value of preservice training. However, the alternative work and on the curriculum of the first year produced routes often replace preservice education with more care- significantly better results among first-year teachers than ful selection of teachers (such as in Teach for America) or programs that did not.a At the same time, systems that with more performance-oriented contracts (such as those have introduced alternative routes to teaching—routes for contracted community teachers). Thus preservice edu- such as Teach for America or community-teacher programs cation remains important for most education systems and that skip regular preservice education—have not reduced will likely yield better results with more practical training. Source: WDR 2018 team. a. Boyd and others (2009). b. Duflo, Dupas, and Kremer (2015); Glazerman, Mayer, and Decker (2006). China—where performance is high by global stan- budget to deliver at scale. But teachers will not learn dards—participate in ongoing Teaching-Research without receiving high-quality teaching themselves. Groups, which provide development, mentoring, and A country facing this conundrum may be better peer evaluation based on classroom observation.16 served by delivering high-quality training in stages Likewise, training associated with a specific ped- rather than ineffective training to all in the short run. agogical technique tends to be more effective. Across educational interventions in the United States, pro- grams teaching a specific pedagogical method have Helping teachers teach to more than twice the impact of programs focused on the level of the student has general pedagogy.17 Globally, specific guidance is cru- cial for low-skilled teachers, who may lack the ability proven effective to be effective even when motivated.18 At times, in set- In many countries facing the learning crisis, it may tings where teachers have limited skills, this involves be that only students who start at the highest levels providing lesson plans that are highly scripted, outlin- of learning are able to keep learning. This is in part ing concrete steps for teachers.19 Many countries will because teachers tend to teach to the most advanced protest that high-quality in-service professional devel- students in a class.20 These students are the easiest opment—repeated, with follow-up visits in school, to teach, and when teachers solicit answers from often around a specific technique—is beyond their students, the high performers are the most likely Teacher skills and motivation both matter (though many education systems act like they don’t) | 133 Figure 6.1 Only a small fraction of teaching the students in the lower-performing group, learners keeps up with the curriculum and higher-quality teachers may be assigned to the higher-performing classes because these students Probability of a correct answer on a math test, by grade, relative to curriculum standards, Andhra Pradesh, India may be easier to teach and so the assignment appears to be a reward. 100 In school systems with very low learning levels, Grade 5 standard ability grouping has had positive impacts on both 90th percentile 80 lower- and higher-performing students. In Kenya, grouping students into classes by ability led to 60 Average improved outcomes across the board, with the high- Percent Grade 1 standard est impacts among learners with more motivated 40 teachers.25 In India, schools reorganized classes by group for just an hour a day and observed major 20 gains in learning.26 Much of the rest of the evidence 10th percentile comes from the United States. Studies that relied on a 0 credible counterfactual found that grouping students 1 2 3 4 5 by ability either helps some students or at least has Grade no adverse impact.27 In low-performing education Source: WDR 2018 team, using data from Muralidharan and Zieleniak (2013). systems, the lowest-performing students learn little Data at http://bit.do/WDR2018-Fig_6-1. to nothing (figure 6.1), so allowing teachers to target pedagogy may have a positive net effect. Another way to help teachers teach to the level of to volunteer them. That leaves behind the students the student is to help them conduct better diagnos- who entered the class with less knowledge. Indeed, tics. In Liberia, an intervention that taught teachers to Kenyan school dropouts identified that problem as a better evaluate their students was effective, especially primary reason for leaving school.21 Another reason when combined with training and additional materi- that many students fall behind is that in many coun- als. So was a similar program in Malawi.28 In Singa- tries the curriculum may simply be too ambitious.22 pore, students take screening tests at the beginning of Teachers feel constrained to teach to the curriculum grade 1, and those who are behind in reading receive even when students have trouble keeping up.23 additional support daily.29 By contrast, an interven- A key principle in leaving no learner behind is to tion in India that merely provided formative evalu- help teachers teach to the level of their students. This ation was not effective; nor was another program in technique has been successful in different formats India that provided diagnostic reports and written across a range of scenarios, whether by using com- suggestions on how to use the reports to strengthen munity teachers to provide remedial lessons to the teaching.30 Clearly, helping teachers to better under- lowest performers, reorganizing classes by ability, or stand their students’ ability levels is worthwhile, but using technology to adapt lessons.24 In many cases, it if teachers lack the tools to respond effectively or the does not require a significantly greater teacher effort, incentives to do so—given that teaching students at but rather relies on restructuring classes or providing multiple levels is challenging—then it may not be remedial lessons for the lowest performers. A related sufficient. The diagnostics can work where a system principle of effective instruction is to reach students is in place to follow up, as well as where teacher moti- by teaching them in their mother tongue (box 6.3). vation is less of a binding constraint (box 6.4). Grouping students by ability may allow teachers New technologies offer promising ways to help with to more effectively target teaching to the levels of stu- teaching to the level of the student. Computer-assisted dents in their classes. The theoretical effects of such learning programs can permit students to go at their grouping are mixed. The positive effects of better- own pace or adjust the level of instruction based on an targeted teaching have a potential downside: the initial screening test.31 More advanced software can adverse effect for lower-performing students of no not only screen students initially but also dynamically longer learning from their higher-performing peers. adjust questions based on ongoing performance. Furthermore, in early grades in particular, student Although the overall evidence on computer-assisted ability is not always easy to measure, so separating instruction is markedly mixed, such a dynamic learn- students by ability can put students on the wrong ing program among secondary school students in track. Teachers may also reduce their efforts when Delhi, India, led to striking gains in both mathematics 134 | World Development Report 2018 Box 6.3 Reaching learners in their own language Children learn to read most effectively in the language also performed better in English reading later on.f On the they speak at home—their mother tongue. In Kenya, stu- other hand, results from a first-language program in Kenya dents in early grades had higher reading comprehension do not show better outcomes in the second language when their teachers had training and materials in mother compared with a second-language literacy program only tongue instruction.a Students participating in a pilot in rural (though the program lasted only one year).g Philippines, where they received instruction in their local lan- But in countries with many languages, mother tongue guage, showed significantly higher reading and math scores instruction can be overwhelming to implement, and a lan- than students in traditional schools, which used English guage “mismatch” can result in learners being left behind. and Filipino.b In Ethiopia, students in schools affected by a Filipinos speak more than 180 different languages, Kenyans reform to implement mother tongue instruction were subse- speak more than 70, and Peruvians speak nearly 100. In 98 quently more likely to be in the appropriate grade for their countries worldwide, the chance that two randomly selected age.c Beyond its direct learning impacts on them, students individuals speak the same mother tongue is under 50 receiving instruction in their mother tongue are more likely percent (map B6.3.1).h In communities with a dominant lan- to attend and persist in school, as demonstrated by data from guage group, the choice of that language for mother tongue 26 countries.d instruction may marginalize minority children. Even in coun- The increased skill from learning to read in mother tries with few languages, teachers generally have little train- tongue can translate into greater skill in a second language. ing in mother tongue instruction, and the materials available Parents and policy makers sometimes object to mother for mother tongue instruction may be limited and of lower tongue instruction on the grounds that the mother tongue quality than materials in the lingua franca.i In communities is not a practical language for the labor market. Yet in South with multiple mother tongues, schools may divide classes Africa, students instructed in their mother tongue in early by mother tongue, but this division can act as segregation.j grades actually performed better in English proficiency in Mother tongue instruction may be an unambiguous benefit later grades.e Likewise in pilot interventions in Malawi and for countries with a limited number of mother tongues, such the Philippines, students instructed in their mother tongue as Burundi or Haiti, but the initiative still involves a major Map B6.3.1 Linguistic diversity around the world Probability that two randomly selected individuals speak a different language Low High Source: WDR 2018 team, using data from Ethnologue (2015). Data at http://bit.do/WDR2018-Map_B6-3-1. IBRD 43166 | SEPTEMBER 2017 (Box continues next page) Teacher skills and motivation both matter (though many education systems act like they don’t) | 135 Box 6.3 Reaching learners in their own language (continued) investment in materials and teacher training. In more diverse overall. In some cases, they may opt for better-selected and locales, governments will need to weigh the gains and the better-trained teachers who receive more support in teach- costs associated with mother tongue instruction against ing students at their level, regardless of the language they those of competing investments in higher-quality education speak. Source: WDR 2018 team. a. Piper, Zuilkowski, and Ong’ele (2016). f. Shin and others (2015); Walter and Dekker (2011). b. Walter and Dekker (2011). g. Piper, Zuilkowski, and Ong’ele (2016). c. Seid (2016). h. Ethnologue (2015). d. Smits, Huisman, and Kruijff (2008). i. Ong’uti, Aloka, and Raburu (2016); RTI International (2016). e. Taylor and von Fintel (2016). j. Metila, Pradilla, and Williams (2016). Box 6.4 Using diagnostic data to deliver better learning in Latin America Mexico’s Colima state implemented a learning improvement implementation of the plan. Student performance improved program in low-performing public schools using student in both language and math, but only several months performance on a national exam. Each school was assigned after the program was launched.a A similar program in a technical adviser who visited schools three times a month Argentina—distributing reports on the learning outcomes to train teachers on analyzing the test information, as well as of students to public primary schools to inform teachers on understanding the reasons for poor performance. Based of the strengths and weaknesses of their students—also on the analysis, the adviser—working with school directors increased learning. Students in those schools reported that and teachers—developed a school-specific plan to address their teachers were more active in interacting with students identified problems and provided follow-up support during in their classrooms and were less likely to leave early.b Source: WDR 2018 team. a. de Hoyos, Garcia-Moreno, and Patrinos (2017). b. de Hoyos, Ganimian, and Holland (2016). and language.32 Teaching at the level of the student is teachers for performing well nor penalize them for not a novel idea, but a range of new evidence is show- performing poorly. Teachers need to be treated as pro- ing how it can be implemented—even at scale—in fessionals—and good professionals receive support developing countries. and respect, but are also held to high expectations. A ­system that does not pay attention to what its teachers are doing does not afford teachers the Teacher motivation and respect they deserve (box 6.5). incentives make a difference, Over the long run, the best way to strengthen even with few inputs teacher ability and motivation may be to attract capa- ble, intrinsically motivated people into the profession. No amount of training or inputs can substitute for In many countries and economies, the youth who plan teacher motivation. Because of high teacher absen- to go into teaching are not among the highest academic teeism in many countries, fostering effort is a serious performers (figure 6.2). In Finland, teaching is a cov- challenge. Moreover, even when they are in school, eted profession, largely because teachers receive great teachers are often not in class teaching. Yet edu- respect, are well trained, are reasonably paid, and have cation systems in many countries neither reward autonomy to implement teaching standards.33 Across 136 | World Development Report 2018 Box 6.5 Would raising teachers’ salaries increase their motivation? In many countries, teachers are paid less than other com- pay increased teacher satisfaction, but it had no effect on parably educated professionals.a Would raising their sal- either measurable effort or student performance for existing aries lead to higher motivation and better performance? teachers.b Though higher salaries could attract more capable Indonesia doubled pay for certified teachers, using a ran- candidates to the profession over time, raising salaries is no domized controlled trial to evaluate the impact. Doubling quick fix for shortcomings in motivation or effort. Source: WDR 2018 team. a. Mizala and Ñopo (2016); OECD (2016a). b. de Ree and others (forthcoming). many countries, average teacher pay has fallen relative Better selection and retention policies will result to that of other professions. At the same time, the wage in better teachers. More meritocratic hiring—say, distribution in teaching has narrowed. High-ability based on a test instead of patronage—could improve candidates may be less attracted by a narrow pay student learning.35 One proposal would be to intro- structure because it gives them little opportunity to duce a teaching apprenticeship of three to five years, reap professional rewards from high performance.34 allowing systems to identify effective teachers.36 Restructuring teacher pay both to remunerate compet- The least effective teachers could then be transi- itively and to provide returns to good performance— tioned out of the teaching force. In the United States, whether directly through pay or indirectly through proposals to phase out the least effective teachers promotion or retention—may improve the quality of suggest that the gains to learners over time would candidates entering the teaching profession. But this is be sub­ stantial: replacing the least effective 7–12 a long-term solution, not a quick fix, and even the best percent of teachers could bridge the gap between candidates need a supportive system to maintain their U.S. student performance and that of Finland.37 Esti- skills and effort over time. mates of teacher value added in other countries are Figure 6.2 Prospective engineers typically score higher than prospective teachers on PISA tests PISA 2015 scores for participating countries and economies, by subject and self-identified prospective occupation a. Mathematics b. Reading 600 600 500 500 Score Score 400 400 300 300 Prospective engineers Prospective teachers Source: WDR 2018 team, using data from OECD (2016b). Data at http://bit.do/WDR2018-Fig_6-2. Note: PISA = Programme for International Student Assessment. Teacher skills and motivation both matter (though many education systems act like they don’t) | 137 comparable, suggesting similarly large gains around of neglectful ones are just two important parts of a the world to improved teacher selection.38 broader spectrum of accountability interventions. At Education systems need to build accountability to the same time, teachers in many environments face align incentives between teachers and others. Teach- multiple demands beyond teaching, as well as risks ers have incentives and information that are distinct such as late payment of salaries and even physical from those of students, parents, and administrators, danger (box 6.6). It can be tempting, in light of data and mental models and social expectations affect the on high teacher absenteeism and low teacher skills, decisions of all actors. In the absence of accountability to blame teachers for many of the faults of education to provide motivation, teachers may minimize their systems. But these systems often ask far more of efforts even as learners and parents wish for them teachers than teaching—and at times offer relatively to exert more. In Argentina and Uganda, more than little in return.42 one-third of teachers surveyed do not see themselves Financial and nonfinancial incentives are one as responsible for their students’ learning; in Senegal, possible mechanism for teacher motivation. In the share is more than half.39 India, students performed better in primary schools Teacher motivation works through various behav- that provided teachers with financial incentives for ioral mechanisms and comes in multiple forms.40 The higher reading and mathematics scores.43 Students fact that another person may observe their perfor- also scored higher in science and social studies, mance offers a form of professional motivation. So do despite no financial incentives being offered in evaluations, where teachers expect their performance those areas. Other financial incentive programs were to be assessed, with the associated consequences.41 successful in two districts of Kenya and elsewhere Financial incentives for successful teachers and firing in India.44 In the United States, by contrast, teacher Box 6.6 One factor undermining teaching: Poor working conditions Analyses of the proximate causes of lack of learning in Teachers in developing countries also face difficult low- and middle-income countries often point to teachers. working and living conditions. A lack of school infrastruc- Evidence suggests that in many countries teachers are absent ture and equipment often handicaps their efforts.f Many for an astonishing number of school days and know too little teachers take on other jobs to support themselves and their about the subjects they are to teach. For this reason, students families.g The situation is even worse for teachers in remote and other stakeholders may want and deserve more from and rural areas, who have to travel long distances to work teachers—but teachers also deserve more from the systems and collect their salary.h that employ them.a Over the last few decades, the status of And then there are the widely implemented curriculum the teaching profession has declined across the world in terms reforms that require teachers to equip students with new of pay, respect, and working conditions.b Because of the rapid skills and employ better pedagogy, but often without expansion in access to education, teachers in developing coun- giving teachers sufficient training and supportive teaching tries often lead oversized, multigrade classes.c The teacher materials.i In such cases, teachers are expected to perform shortage increases workloads and requires long working as professionals, but education systems fail to offer them hours, sometimes including double shifts.d Moreover, teachers professional development opportunities and create a pro- often have duties outside classrooms, such as coordinating fessional culture for them.j the activities of parent-teacher associations, running extra­ curricular activities, and performing administrative tasks.e Source: WDR 2018 team. a. Evans and Yuan (2017). e. Guajardo (2011); Liu and Onwuegbuzie (2012); Luschei and Chudgar (2017). b. Dolton and Marcenaro-Gutierrez (2011); Hammett (2008); Harris-Van f. Alcázar and others (2006); Gamero Burón and Lassibille (2016); Urwick Keuren and Silova (2015). and Kisa (2014). c. Gamero Burón and Lassibille (2016); Guajardo (2011); Ramachandran, g. Urwick and Kisa (2014). Bhattacharjea, and Sheshagiri (2008). h. Gamero Burón and Lassibille (2016). d. Ávalos and Valenzuela (2016); Gamero Burón and Lassibille (2016); Liu i. Peng and others (2014); Urwick and Kisa (2014). and Onwuegbuzie (2012); Luschei and Chudgar (2017); Osei (2006); j. Mooij (2008). Urwick and Kisa (2014). 138 | World Development Report 2018 financial incentives did not improve test scores in With financial incentives, the devil is in the details. several states.45 However, large financial incentives Incentives can be based on teacher inputs such as for teachers did increase student learning in the attendance or on outputs such as student learning. District of Columbia, United States.46 In Mexico and They can be based on reaching an absolute level of Tanzania, teacher financial incentives were effective achievement or on gains. They can be available to all only in conjunction with another intervention.47 who reach a goal, or they can be competitive across One interpretation of this scattered evidence is that schools. They will vary in size relative to teacher sal- financial incentives are most likely to be effective aries. The evidence on these design elements is still when teachers can take straightforward actions to limited, but they merit careful consideration, taking improve learning. In environments with high teacher into account local institutions. absenteeism from school or from the classroom while Likewise, the precise shape of a system’s overall at school, it is likely to be clear to teachers that they incentive structure will vary by context. In some places, can improve learning by simply coming to school financial incentives may be worth piloting. In others, and spending more time teaching. Alternatively, in increased community accountability may be effective. environments like that found in the United States The mixed evidence on these interventions suggests a where teacher absenteeism is minimal, the specific need to examine carefully the context and to test pro- actions that teachers should take to improve learn- grams locally. But while details will vary, no education ing may be less obvious and less easy to implement. system will be successful unless it provides incen- Nonfinancial incentives may include providing suc- tives—whether implicit or explicit—for teacher effort. cessful teachers with special recognition. Evidence of the effectiveness of these incentives in education *** is limited, although there is suggestive evidence in Over time, education systems perform best when other sectors—for example among health workers in their teachers are respected, prepared, selected based Zambia, where public recognition of worker achieve- on merit, and supported in their work. Countries ment markedly improved performance.48 should work toward these objectives. But in the short Financial incentives can also create challenges. In run, countries can take actions to strengthen the Kenya, responding to a student incentive program, performance of teachers. They can improve the qual- teachers taught specifically to the test, potentially ity of professional development, shifting resources neglecting more holistic learning. In a teacher to the kinds of professional development that will incentive program in Mexican secondary schools, a change teacher performance in the classroom. They significant portion of the identified increase in stu- can support teachers in teaching to the level of the dent learning was attributed to student cheating.49 student. They can provide a professional structure so In the United States, teacher cheating rose strongly that teachers feel motivated to apply what they know. when incentives were increased.50 And when teacher Teachers are key to learners’ education. Making them incentive programs are removed, the results can also more effective in both the short and the long run is be adverse.51 an excellent investment. Notes 1. UIS (2017). 10. Popova, Evans, and Arancibia (2016). 2. Bold and others (forthcoming); Tandon and Fukao (2015); 11. Walter and Briggs (2012). World Bank (2016). 12. Darling-Hammond and others (2009); Yoon and others 3. Bruns and Luque (2015); Mulkeen (2010). (2007). 4. Strizek and others (2014). 13. Kraft, Blazar, and Hogan (2016); Popova, Evans, and 5. Calderón (2014); World Bank (2014, 2016). Arancibia (2016). 6. Darling-Hammond and others (2009). 14. Conn (2017). 7. Hammett (2008); Lauwerier and Akkari (2015). 15. Banerjee and others (2007). 8. Aaronson, Barrow, and Sander (2007); Buddin and 16. Liang, Kidwai, and Zhang (2016). Zamarro (2009); Goldhaber (2007); Rivkin, Hanushek, 17. Fryer (2017). and Kain (2005). 18. Ganimian and Murnane (2016). 9. Michaelowa (2001). Teacher skills and motivation both matter (though many education systems act like they don’t) | 139 19. He, Linden, and MacLeod (2008, 2009); Lucas and others Ávalos, Beatrice, and Juan Pablo Valenzuela. 2016. “Education (2014); Spratt, King, and Bulat (2013). for All and Attrition/Retention of New Teachers: A Tra- 20. Abadzi and Llambiri (2011); Ciaccio (2004); Leder (1987). jectory Study in Chile.” International Journal of Educational 21. Zuilkowski, Jukes, and Dubeck (2016). Development 49: 279–90. 22. Pritchett and Beatty (2015). Banerjee, Abhijit Vinayak, Rukmini Banerji, James Berry, 23. Banerjee and others (2016). Esther Duflo, Harini Kannan, Shobhini Mukerji, Marc 24. Banerjee and others (2007, 2016); Duflo, Dupas, and Shotland, et al. 2016. “Mainstreaming an Effective Inter- Kremer (2011); Kiessel and Duflo (2014); Muralidharan, vention: Evidence from Randomized Evaluations of Singh, and Ganimian (2016). ‘Teaching at the Right Level’ in India.” NBER Working 25. Cummins (2016); Duflo, Dupas, and Kremer (2011). Paper 22746, National Bureau of Economic Research, 26. Banerjee and others (2016). Cambridge, MA. 27. Figlio and Page (2002); Lefgren (2004); Zimmer (2003). Banerjee, Abhijit Vinayak, Shawn Cole, Esther Duflo, and 28. Bolyard (2003); Piper and Korda (2010). Leigh Linden. 2007. “Remedying Education: Evidence 29. OECD (2011). from Two Randomized Experiments in India.” Quarterly 30. Aaronson, Barrow, and Sander (2007); Duflo and others Journal of Economics 122 (3): 1235–64. (2014); Muralidharan and Sundararaman (2010). Behrman, Jere R., Susan W. Parker, Petra E. Todd, and 31. Banerjee and others (2007); Carrillo, Onofa, and Ponce Kenneth I. Wolpin. 2015. “Aligning Learning Incentives (2010). of Students and Teachers: Results from a Social Experi- 32. Muralidharan, Singh, and Ganimian (2016). ment in Mexican High Schools.” Journal of Political Econ- 33. Sahlberg (2011). omy 123 (2): 325–64. 34. Jackson (2012). Bold, Tessa, Deon Filmer, Gayle Martin, Ezequiel Molina, 35. Estrada (2016). Brian Stacy, Christophe Rockmore, Jakob Svensson, et al. 36. Muralidharan (2016). Forthcoming. “Enrollment without Learning: Teacher 37. Hanushek (2011). Effort, Knowledge, and Skill in Primary Schools in 38. Buhl-Wiggers and others (2017). Africa.” Journal of Economic Perspectives. 39. Sabarwal, Abu-Jawdeh, and Masood (2017). Bolyard, K. J. 2003. “Linking Continuous Assessment and 40. Gill, Lerner, and Meosky (2016). Teacher Development: Evaluating a Model of Continu- 41. Lerner and Tetlock (1999). ous Assessment for Primary Schools in Malawi.” EQUIP1 42. Mizala and Ñopo (2016); OECD (2016a). Continuous Assessment, Educational Quality Improve- 43. Muralidharan (2012); Muralidharan and Sundararaman ment Program, U.S. Agency for International Develop- (2011). ment, Washington, DC. 44. Duflo, Hanna, and Ryan (2012); Glewwe, Ilias, and Boyd, Donald J., Pamela L. Grossman, Hamilton Lankford, Kremer (2010). Susanna Loeb, and James Wyckoff. 2009. “Teacher Prepa- 45. Fryer (2013); Glazerman, McKie, and Carey (2009); ration and Student Achievement.” Educational Evaluation Springer and others (2010). and Policy Analysis 31 (4): 416–40. 46. Dee and Wyckoff (2015). Bruns, Barbara, and Javier Luque. 2015. Great Teachers: How to 47. Behrman and others (2015); Mbiti, Muralidharan, and Raise Student Learning in Latin America and the Caribbean. Schipper (2016). With Soledad De Gregorio, David K. Evans, Marco 48. Ashraf, Bandiera, and Jack (2014). Fernández, Martin Moreno, Jessica Rodriguez, Guillermo 49. Behrman and others (2015). Toral, and Noah Yarrow. Latin American Development 50. Jacob and Levitt (2003). Forum Series. Washington, DC: World Bank. 51. Jinnai (2016); Visaria and others (2016). Buddin, Richard, and Gema Zamarro. 2009. “Teacher Qualifi- cations and Student Achievement in Urban Elementary Schools.” Journal of Urban Economics 66 (2): 103–15. References Buhl-Wiggers, Julie, Jason T. Kerwin, Jeffrey A. Smith, and Aaronson, Daniel, Lisa Barrow, and William Sander. 2007. Rebecca Thornton. 2017. “The Impact of Teacher Effec- “Teachers and Student Achievement in the Chicago Pub- tiveness on Student Learning in Africa.” Paper presented lic High Schools.” Journal of Labor Economics 25 (1): 95–135. at RISE Annual Conference 2017, Center for Global Devel- Abadzi, Helen, and Stavri Llambiri. 2011. “Selective Teacher opment, Washington, DC, June 15–16. http://www.rise Attention in Lower-Income Countries: A Phenomenon programme.org/sites/www.riseprogramme.org/files Linked to Dropout and Illiteracy?” Prospects 41 (4): 491–506. /Buhl-Wiggers%20The%20Impact%20of%20Teacher%20 Alcázar, Lorena, F. Halsey Rogers, Nazmul Chaudhury, Effectiveness%202017-04-30_0.pdf. Jeffrey Hammer, Michael R. Kremer, and Karthik Calderón, David. 2014. “Invertir en la Formación Docente.” Muralidharan. 2006. “Why Are Teachers Absent? Probing Animal Político (blog), November 13. http://www.animal Service Delivery in Peruvian Primary Schools.” Interna- politico.com/blogueros-aprender-es-mi-derecho/2014/11 tional Journal of Educational Research 45 (3): 117–36. /13/invertir-en-la-formacion-docente/. Ashraf, Nava, Oriana Bandiera, and B. Kelsey Jack. 2014. “No Carrillo, Paul, Mercedes Onofa, and Juan Ponce. 2010. “Infor- Margin, No Mission? A Field Experiment on Incentives mation Technology and Student Achievement: Evidence for Public Service Delivery.” Journal of Public Economics from a Randomized Experiment in Ecuador.” IDB 120: 1–17. 140 | World Development Report 2018 Working Paper IDB-WP-223, Inter-American Develop- Estrada, Ricardo. 2016. “Crony Education: Teacher Hiring ment Bank, Washington, DC. and Rent Extraction.” Working paper, European Univer- Ciaccio, Joseph. 2004. Totally Positive Teaching: A Five-Stage sity Institute, San Domenico di Fiesole, Italy. Approach to Energizing Students and Teachers. Alexandria, Ethnologue. 2015. Ethnologue: Languages of the World VA: Association for Supervision and Curriculum (database). 18th ed., SIL International, Dallas. http:// Development. www.ethnologue.com/18/. Conn, Katharine M. 2017. “Identifying Effective Education Evans, David K., and Fei Yuan. 2017. “The Working Condi- Interventions in Sub-Saharan Africa: A Meta-Analysis tions of Teachers in Low- and Middle-Income Countries.” of Impact Evaluations.” Review of Educational Research WDR 2018 background paper, World Bank, Washing- (May 26). http://journals.sagepub.com/doi/abs/10.3102 ton, DC. /0034654317712025. Figlio, David N., and Marianne E. Page. 2002. “School Choice Cummins, Joseph R. 2016. “Heterogeneous Treatment and the Distributional Effects of Ability Tracking: Does Effects in the Low Track: Revisiting the Kenyan Primary Separation Increase Inequality?” Journal of Urban Econom- School Experiment.” Economics of Education Review ics 51 (3): 497–514. 56 (February): 40–51. Fryer, Roland G., Jr. 2013. “Teacher Incentives and Student Darling-Hammond, Linda, Ruth Chung Wei, Alethea Andree, Achievement: Evidence from New York City Public Nikole Richardson, and Stelios Orphanos. 2009. “Profes- Schools.” Journal of Labor Economics 31 (2): 373–407. sional Learning in the Learning Profession: A Status —— — —. 2017. “The Production of Human Capital in Devel- Report on Teacher Development in the United States oped Countries: Evidence from 196 Randomized Field and Abroad.” National Staff Development Council, Experiments.” In Handbook of Field Experiments, edited Dallas. by Abhijit Vinayak Banerjee and Esther Duflo, Vol. 2, Dee, Thomas S., and James Wyckoff. 2015. “Incentives, Selec- 95–322. Handbooks in Economics Series. Amsterdam: tion, and Teacher Performance: Evidence from Impact.” North-Holland. Journal of Policy Analysis and Management 34 (2): 267–97. Gamero Burón, Carlos, and Gérard Lassibille. 2016. “Job de Hoyos, Rafael E., Alejandro J. Ganimian, and Peter A. Satisfaction among Primary School Personnel in Mada- Holland. 2016. “Teaching with the Test: Experimental gascar.” Journal of Development Studies 52 (11): 1628–46. Evidence on Diagnostic Feedback and Capacity-Building Ganimian, Alejandro J., and Richard J. Murnane. 2016. for Schools in Argentina.” Working paper, World Bank, “Improving Education in Developing Countries: Lessons Washington, DC. from Rigorous Impact Evaluations.” Review of Educational de Hoyos, Rafael E., Vicente A. Garcia-Moreno, and Harry Research 86 (3): 719–55. Anthony Patrinos. 2017. “The Impact of an Accountability Gill, Brian P., Jennifer S. Lerner, and Paul Meosky. 2016. Intervention with Diagnostic Feedback: Evidence from “Reimagining Accountability in K–12 Education.” Behav- Mexico.” Economics of Education Review 58: 123–40. ioral Science and Policy 2 (1): 57–70. de Ree, Joppe, Karthik Muralidharan, Menno Pradhan, Glazerman, Steven, Daniel Mayer, and Paul Decker. 2006. and Halsey Rogers. Forthcoming. “Double for Nothing? “Alternative Routes to Teaching: The Impacts of Teach Experimental Evidence on an Unconditional Teacher for America on Student Achievement and Other Out- Salary Increase in Indonesia.” Quarterly Journal of comes.” Journal of Policy Analysis and Management 25 (1): Economics. 75–96. Dolton, Peter, and Oscar Marcenaro-Gutierrez. 2011. “2013 Glazerman, Steven, Allison McKie, and Nancy Carey. 2009. Global Teacher Status Index.” With Vikas Pota, Marc “An Evaluation of the Teacher Advancement Program Boxser, and Ash Pajpani. Varkey Gems Foundation, (TAP) in Chicago: Year One Impact Report, Final Report.” London. https://www.varkeyfoundation.org/sites/default Mathematica Policy Research, Princeton, NJ. /files/documents/2013GlobalTeacherStatusIndex.pdf. Glewwe, Paul W., Nauman Ilias, and Michael R. Kremer. Duflo, Esther, James Berry, Shobhini Mukerji, and Marc 2010. “Teacher Incentives.” American Economic Journal: Shotland. 2014. “A Wide Angle View of Learning: Evalua- Applied Economics 2 (3): 205–27. tion of the CCE and LEP Programmes in Haryana.” Goldhaber, Dan. 2007. “Everyone’s Doing It, but What Does 3ie Grantee Final Report, International Initiative for Teacher Testing Tell Us about Teacher Effectiveness?” Impact Evaluation, New Delhi. Journal of Human Resources 42 (4): 765–94. Duflo, Esther, Pascaline Dupas, and Michael R. Kremer. 2011. Guajardo, Jarret. 2011. “Teacher Motivation: Theoretical “Peer Effects, Teacher Incentives, and the Impact of Framework, Situation Analysis of Save the Children Tracking: Evidence from a Randomized Evaluation in Country Offices, and Recommended Strategies.” Teacher Kenya.” American Economic Review 101 (5): 1739–74. Motivation Working Group, Save the Children, Fair­ - —— ——. 2015. “School Governance, Teacher Incentives, and field, CT. Pupil-Teacher Ratios: Experimental Evidence from Hammett, Daniel. 2008. “Disrespecting Teacher: The Decline Kenyan Primary Schools.” Journal of Public Economics 123 in Social Standing of Teachers in Cape Town, South (March): 92–110. Africa.” International Journal of Educational Development Duflo, Esther, Rema Hanna, and Stephen P. Ryan. 2012. 28 (3): 340–47. “Incentives Work: Getting Teachers to Come to School.” Hanushek, Eric A. 2011. “Valuing Teachers: How Much Is a American Economic Review 102 (4): 1241–78. Good Teacher Worth?” Education Next 11 (3): 40–45. Teacher skills and motivation both matter (though many education systems act like they don’t) | 141 Harris-Van Keuren, Christine, and Iveta Silova. 2015. “Imple- Teacher Distribution in Developing Countries: Teachers of menting EFA Strategy No. 9: The Evolution of the Status Marginalized Students in India, Mexico, and Tanzania, edited of the Teaching Profession (2000–2015) and the Impact by Thomas F. Luschei and Amita Chudgar, 87–107. New on the Quality of Education in Developing Countries, York: Palgrave Macmillan. Three Case Studies.” With Suzanne McAllister. Report Mbiti, Isaac M., Karthik Muralidharan, and Youdi Schipper. ED/EFA/MRT/2015/PI/08, background paper, Education 2016. “Inputs, Incentives, and Complementarities in Pri- for All Global Monitoring Report, United Nations Educa- mary Education: Experimental Evidence from Tanzania.” tional, Scientific, and Cultural Organization, Paris. Working paper, University of California at San Diego. He, Fang, Leigh L. Linden, and Margaret MacLeod. 2008. Metila, Romylyn A., Lea Angela S. Pradilla, and Alan B. Wil- “How to Teach English in India: Testing the Relative Pro- liams. 2016. “The Challenge of Implementing Mother ductivity of Instruction Methods within the Pratham Tongue Education in Linguistically Diverse Contexts: English Language Education Program.” Working paper, The Case of the Philippines.” Asia-Pacific Education Columbia University, New York. Researcher 25 (5–6): 781–89. ————. 2009. “A Better Way to Teach Children to Read? Evi- Michaelowa, Katharina. 2001. “Primary Education Quality in dence from a Randomized Controlled Trial.” Working Francophone Sub-Saharan Africa: Determinants of paper, Columbia University, New York. Learning Achievement and Efficiency Considerations.” Jackson, C. Kirabo. 2012. “Recruiting, Retaining, and Creat- World Development 29 (10): 1699–1716. ing Quality Teachers.” Nordic Economic Policy Review 3 (1): Mizala, Alejandra, and Hugo Ñopo. 2016. “Measuring the 61–104. Relative Pay of School Teachers in Latin America 1997– Jacob, Brian A., and Steven D. Levitt. 2003. “Rotten Apples: An 2007.” International Journal of Educational Development Investigation of the Prevalence and Predictors of Teacher 47: 20–32. Cheating.” Quarterly Journal of Economics 118 (3): 843–78. Mooij, Jos. 2008. “Primary Education, Teachers’ Profession- Jinnai, Yusuke. 2016. “To Introduce or Not to Introduce Mon- alism and Social Class about Motivation and Demotiva- etary Bonuses: The Cost of Repealing Teacher Incen- tion of Government School Teachers in India.” Interna- tives.” Economics and Management Series EMS-2016-08, tional Journal of Educational Development 28 (5): 508–23. IUJ Research Institute, International University of Japan, Mulkeen, Aidan G. 2010. Teachers in Anglophone Africa: Issues in Minamiuonuma, Niigata Prefecture. http://www.iuj.ac Teacher Supply, Training, and Management. Development .jp/research/workingpapers/EMS_2016_08.pdf. Practice in Education Series. Washington, DC: World Kiessel, Jessica, and Annie Duflo. 2014. “Cost Effectiveness Bank. Report: Teacher Community Assistant Initiative (TCAI).” Muralidharan, Karthik. 2012. “Long-Term Effects of Teacher IPA Brief, Innovation for Poverty Action, New Haven, CT. Performance Pay: Experimental Evidence from India.” Kraft, Matthew A., David Blazar, and Dylan Hogan. 2016. Working paper, University of California at San Diego. “The Effect of Teacher Coaching on Instruction and ——— —. 2016. “A New Approach to Public Sector Hiring in Achievement: A Meta-Analysis of the Causal Evidence.” Working paper, Brown University, Providence, RI. India for Improved Service Delivery.” Working paper, Lauwerier, Thibaut, and Abdeljalil Akkari. 2015. “Teachers University of California at San Diego. and the Quality of Basic Education in Sub-Saharan Muralidharan, Karthik, Abhijeet Singh, and Alejandro J. Africa.” ERF Working Paper 11, Education Research and Ganimian. 2016. “Disrupting Education? Experimental Foresight, Paris. Evidence on Technology-Aided Instruction in India.” Leder, Gilah C. 1987. “Teacher Student Interaction: A Case NBER Working Paper 22923, National Bureau of Eco- Study.” Educational Studies in Mathematics 18 (3): 255–71. nomic Research, Cambridge, MA. Lefgren, Lars. 2004. “Educational Peer Effects and the Chicago Muralidharan, Karthik, and Venkatesh Sundararaman. 2010. Public Schools.” Journal of Urban Economics 56 (2): 169–91. “The Impact of Diagnostic Feedback to Teachers on Lerner, Jennifer S., and Philip E. Tetlock. 1999. “Accounting Student Learning: Experimental Evidence from India.” for the Effects of Accountability.” Psychological Bulletin 125 Economic Journal 120 (546): F187–F203. (2): 255–75. ——— —. 2011. “Teacher Performance Pay: Experimental Evi- Liang, Xiaoyan, Huma Kidwai, and Minxuan Zhang. 2016. dence from India.” Journal of Political Economy 119 (1): How Shanghai Does It: Insights and Lessons from the Highest- 39–77. Ranking Education System in the World. Directions in Devel- Muralidharan, Karthik, and Yendrick Zieleniak. 2013. “Mea- opment: Human Development Series. Washington, DC: suring Learning Trajectories in Developing Countries World Bank. with Longitudinal Data and Item Response Theory.” Liu, Shujie, and Anthony J. Onwuegbuzie. 2012. “Chinese Paper presented at Young Lives Conference, Oxford Teachers’ Work Stress and Their Turnover Intention.” University, Oxford, U.K., July 8–9. International Journal of Educational Research 53: 160–70. OECD (Organisation for Economic Co-operation and Devel- Lucas, Adrienne M., Patrick J. McEwan, Moses Ngware, and opment). 2011. Strong Performers and Successful Reformers Moses Oketch. 2014. “Improving Early-Grade Literacy in in Education: Lessons from PISA for the United States. Paris: East Africa: Experimental Evidence from Kenya and OECD. Uganda.” Journal of Policy Analysis and Management 33 (4): ——— —. 2016a. Education at a Glance 2016: OECD Indicators. 950–76. Paris: OECD. Luschei, Thomas F., and Amita Chudgar. 2017. “Supply-Side ——— —. 2016b. PISA 2015 Results: Excellence and Equity in Educa- Explanations for Inequitable Teacher Distribution.” In tion. Vol. 1. Paris: OECD. 142 | World Development Report 2018 Ong’uti, Charles Onchiri, Peter J. O. Aloka, and Pamela Learning English as a Second Language.” Bilingual Raburu. 2016. “Factors Affecting Teaching and Learning Research Journal 38 (3): 255–74. in Mother Tongue in Public Lower Primary Schools in Smits, Jeroen, Janine Huisman, and Karine Kruijff. 2008. Kenya.” International Journal of Psychology and Behavioral “Home Language and Education in the Developing Sciences 6 (3): 161–66. World.” Report 2009/ED/EFA/MRT/PI/21, background Osei, George M. 2006. “Teachers in Ghana: Issues of Train- paper, Education for All Global Monitoring Report, ing, Remuneration and Effectiveness.” International Jour- United Nations Educational, Scientific, and Cultural nal of Educational Development 26 (1): 38–51. Organization, Paris. Peng, Wen J., Elizabeth McNess, Sally Thomas, Xiang Rong Spratt, Jennifer, Simon King, and Jennae Bulat. 2013. “Inde- Wu, Chong Zhang, Jian Zhong Li, and Hui Sheng Tian. pendent Evaluation of the Effectiveness of Institut pour 2014. “Emerging Perceptions of Teacher Quality and l’Education Populaire’s ‘Read-Learn-Lead’ (RLL) Program Teacher Development in China.” International Journal of in Mali: Endline Report.” Research Triangle Institute, Educational Development 34: 77–89. Research Triangle Park, NC. Piper, Benjamin, and Medina Korda. 2010. “EGRA Plus: Springer, Matthew G., Dale Ballou, Laura Hamilton, Liberia, Program Evaluation Report.” Research Triangle Vi-Nhuan Le, J. R. Lockwood, Daniel F. McCaffrey, Institute, Research Triangle Park, NC. Matthew Pepper, et al. 2010. “Teacher Pay for Perfor- Piper, Benjamin, Stephanie S. Zuilkowski, and Salome mance: Experimental Evidence from the Project on Ong’ele. 2016. “Implementing Mother Tongue Instruc- Incentives in Teaching.” National Center on Performance tion in the Real World: Results from a Medium-Scale Incentives, Vanderbilt University, Nashville, TN. Randomized Controlled Trial in Kenya.” Comparative Strizek, Gregory A., Steve Tourkin, Ebru Erberber, and Pat- Education Review 60 (4): 776–807. rick Gonzales. 2014. “Teaching and Learning Interna- Popova, Anna, Mary E. Breeding, and David K. Evans. 2017. tional Survey (TALIS) 2013: U.S. Technical Report.” NCES “Global Landscape of In-Service Teacher Professional 2015–010, National Center for Education Statistics, Insti- Development Programs: The Gap between Evidence and tute of Education Sciences, U.S. Department of Educa- Practice.” Background paper, World Bank, Washington, tion, Washington, DC. DC. Tandon, Prateek, and Tsuyoshi Fukao. 2015. Educating the Popova, Anna, David K. Evans, and Violeta Arancibia. 2016. Next Generation: Improving Teacher Quality in Cambodia. “Training Teachers on the Job: What Works and How to Directions in Development: Human Development Series. Measure It.” Policy Research Working Paper 7834, World Washington, DC: World Bank. Bank, Washington, DC. Taylor, Stephen, and Marisa von Fintel. 2016. “Estimating the Pritchett, Lant, and Amanda Beatty. 2015. “Slow Down, You’re Impact of Language of Instruction in South African Pri- Going Too Fast: Matching Curricula to Student Skill mary Schools: A Fixed Effects Approach.” Economics of Levels.” International Journal of Educational Development Education Review 50: 75–89. 40: 276–88. UIS (UNESCO Institute for Statistics). 2017. Education (data- Ramachandran, Vimala, Suman Bhattacharjea, and K. M. base). UIS, Montreal. http://data.uis.unesco.org/. Sheshagiri. 2008. “Primary School Teachers: The Twists Urwick, James, and Sarah Kisa. 2014. “Science Teacher Short- and Turns of Everyday Practice.” Working paper, Educa- age and the Moonlighting Culture: The Pathology of the tional Resource Unit, New Delhi. Teacher Labour Market in Uganda.” International Journal Rivkin, Steven G., Eric A. Hanushek, and John F. Kain. 2005. of Educational Development 36: 72–80. “Teachers, Schools, and Academic Achievement.” Econo- Visaria, Sujata, Rajeev Dehejia, Melody M. Chao, and metrica 73 (2): 417–58. Anirban Mukhopadhyay. 2016. “Unintended Conse- RTI International. 2016. “Survey of Children’s Reading Mate- quences of Rewards for Student Attendance: Results rials in African Languages in Eleven Countries: Final from a Field Experiment in Indian Classrooms.” Econom- Report.” EdData II: Data for Education Research and ics of Education Review 54: 173–84. Programming in Africa, U.S. Agency for International Walter, Catherine, and Jessica Briggs. 2012. “What Profes- Development, Washington, DC. sional Development Makes the Most Difference to Sabarwal, Shwetlena, Malek Abu-Jawdeh, and Eema Masood. Teachers?” Department of Education, University of 2017. “Understanding Teacher Effort: Insights from Oxford, Oxford, U.K. Cross-Country Data on Teacher Perceptions.” Back- Walter, Stephen L., and Diane E. Dekker. 2011. “Mother ground paper, World Bank, Washington, DC. Tongue Instruction in Lubuagan: A Case Study from the Sahlberg, Pasi. 2011. Finnish Lessons: What Can the World Learn Philippines.” International Review of Education 57 (5–6): from Educational Change in Finland? New York: Teachers 667–83. College Press. World Bank. 2014. “Teacher Training (% of Total Education Seid, Yared. 2016. “Does Learning in Mother Tongue Matter? Expenditure).” EdStats: Education Statistics (database). Evidence from a Natural Experiment in Ethiopia.” World Bank, Washington, DC. http://datatopics.world Economics of Education Review 55: 21–38. bank.org/education/. Shin, Jaran, Misty Sailors, Nicola McClung, P. David —— — —. 2016. “Assessing Basic Education Service Delivery in Pearson, James V. Hoffman, and Margaret Chilimanjira. the Philippines: The Philippines Public Education 2015. “The Case of Chichewa and English in Malawi: The Expenditure Tracking and Quantitative Service Delivery Impact of First Language Reading and Writing on Study.” Report AUS6799. Washington, DC: World Bank. Teacher skills and motivation both matter (though many education systems act like they don’t) | 143 Yoon, Kwang Suk, Teresa Duncan, Silvia Wen-Yu Lee, Beth Zimmer, Ron. 2003. “A New Twist in the Educational Track- Scarloss, and Kathy L. Shapley. 2007. “Reviewing the ing Debate.” Economics of Education Review 22 (3): 307–15. Evidence on How Teacher Professional Development Zuilkowski, Stephanie Simmons, Matthew C. H. Jukes, and Affects Student Achievement.” Issues and Answers Margaret M. Dubeck. 2016. “ ‘I Failed, No Matter How Report REL 2007–033, Regional Educational Laboratory Hard I Tried’: A Mixed-Methods Study of the Role of Southwest, National Center for Education Statistics, Achievement in Primary School Dropout in Rural Institute of Education Sciences, U.S. Department of Edu- Kenya.” International Journal of Educational Development cation, Washington, DC. 50: 100–107. 144 | World Development Report 2018 Everything else should strengthen the teacher-learner interaction 7 Investments in school inputs, management, and governance often are not guided by how well they improve the teacher-learner relationship. To be effective, they should be. Learners and teachers have a more productive Ensure that information and communication tech- •  learning relationship when supported by learning nology (ICT) can be implemented in current sys- materials and other inputs. Most countries, from the tems. Otherwise, it will be ineffective. lowest- to the highest-income, are seeking to incorpo- Recognize that school management and gover- •  rate technology into their classrooms and education nance reform, along with community monitoring, systems. But technology is merely the most discussed can achieve more learning only if they affect inter- of a range of inputs intended to improve the teacher- action between teachers and learners. learner relationship, from pencils and textbooks in the hands of learners to the walls and roofs of school buildings. Good school management also focuses on Technological interventions supporting students learning from teachers. Yet tech- increase learning—but only nology—along with other physical inputs—often fails to support the work that teachers and students do, and if they enhance the teacher- the potential of school leaders and community mem- learner relationship bers often goes unrealized. Technology can strengthen learning.1 Software can be This chapter lays out evidence for the most effec- highly effective if it allows students to learn at their tive use of these complementary inputs in places own pace and, in the best cases, adapts dynamically where the gap between evidence and practice is larg- to their knowledge.2 A game-based computer-assisted est. In many cases, the failures observed can be illu- learning program in Qinghai, China, intended to minated through models of human behavior, which improve student language scores not only did that, but also point the way to solutions (table 7.1). A synthesis also improved students’ knowledge of mathematics.3 of the evidence in these areas reveals three principles Technology is about much more than giving that are keys to success in achieving learning through computers to students. ICT interventions include a school investments: wide range of technological monitoring and infor- •  Ensure that other inputs—including new tech- mation systems at all levels of education, from indi- nology—complement teachers, thereby making vidual students to education systems. Computers teaching more effective. Taking this approach, and computer-assisted learning software, as well as rather than seeking to circumvent teachers, can online platforms such as Google Classroom, Black- increase learning. board, and Brazil’s Education Connection, enable Everything else should strengthen the teacher-learner interaction | 145 Table 7.1 Models of human behavior can guide actions to improve the effectiveness of school inputs and governance: Some examples Models that identify a Approaches that address Synthesis principle Where this fails mechanism behind this failure the modeled mechanism Additional inputs should Inputs like laptops are sometimes Information failure: Policy New books and materials complement rather than used to circumvent the teacher- makers seek to circumvent have been ineffective in substitute for teachers. learner relationship but fail to poorly functioning teacher- many places, but in Liberia deliver learning benefits. learner relationships without they increased student evidence on an alternative learning when combined model of learning. with teacher training. Technologies must be Education technology investments Behavioral (optimism bias): In India, computer-assisted implementable in the current routinely fail because there is Policy makers project unrealistic learning has dramatically education system to achieve limited capacity to maintain them, technological progress. improved learning outcomes more learning. or the infrastructure needed for in dedicated technology them to work effectively does not centers. exist. School governance reform School governance reforms and Information failure: Community In Mexico, community and community monitoring community monitoring often fail members often do not observe engagement over time, with improve learning only if they to take into account community the most important part of decentralization of real affect the teacher-learner capacity. the learning process—what decision-making power, has interaction. happens in the classroom. been effective. Source: WDR 2018 team. learners and parents to communicate with teachers Figure 7.1 Information and about assignments and materials, and they offer communication technology has free materials that educators and parents can use in had a mixed impact on learning designing age-appropriate development activities.4 Distribution of the effects of education technology on These platforms include interactive whiteboards, text student learning, by type messages to support teachers, and televised programs to improve instructional quality in areas with limited 100 Positive effect access to trained teachers.5 Though ICT offers potentially significant gains 80 for education, the effects of tested interventions have varied greatly. Some programs have been extremely impressive, such as a dynamic computer-assisted 60 No effect learning program for secondary school students in Percent India that increased math and language scores more than most other learning interventions tested there 40 or elsewhere.6 But others, such as the One Laptop Per Child programs in Peru and Uruguay, have shown no impact on student reading or math ability.7 Indeed, 20 the vast majority of ICT interventions have had Negative effect either no impact or—as with certain hardware inter- ventions—a negative impact on student learning 0 (figure 7.1).8 Software Hardware Moreover, current evidence likely overestimates Source: WDR 2018 team, using data from Muralidharan, Singh, and Ganimian the effectiveness of ICT interventions in education (2016, annex 2). Data at http://bit.do/WDR2018-Fig_7-1. because many fail—or stumble badly—before being implemented. In Haiti, a program to use smart- Laptop Per Child initiative faced years of delays in phones to monitor teacher attendance had no effect several states. And a year after the laptops made on teacher attendance or student outcomes because it to classrooms, more than 40 percent of teachers implementation proved untenable.9 Brazil’s One reported never or rarely using them in classroom 146 | World Development Report 2018 activities.10 Rich countries face the same challenges: class time. Students under that model performed sig- beyond education, almost a fifth of public sector ICT nificantly worse than students left with their regular projects in the United Kingdom have had cost over- teachers. In the other approach, where students used runs of more than 25 percent, and the typical project the program after school, there were sizable gains, takes 24 percent longer to implement than initially especially for the poorest performers.16 Another exam- expected.11 It is crucial to focus on technologies that ple of technology that complements teachers is a series are truly feasible in existing systems. In rural areas, of prepared videos of high-quality lessons—such as technology may be more attractive because of weak Brazil’s Telecurso—which can be used in a classroom. education systems, but at the same time those weak Technology holds some promise in fragile set- systems—with their limited access to electricity or tings, such as those afflicted by war or epidemic, to the internet—have the least capacity to support edu- maintain a connection to formal education. During cation technology interventions. the 2014–15 Ebola epidemic in Sierra Leone, schools With such varied returns and so many challenges shut down for eight months, but the government to implementation, why is there so much investment launched an emergency education program with in education technology? Both principal-agent rela- lessons five days a week. A 30-minute lesson over tionships and behavioral biases likely play a role. The the radio is unlikely to have a deep learning impact, principal-agent model is relevant because public offi- but this kind of program may help children stay cials may derive political returns from flashy techno- connected to learning.17 Sudan’s Can’t Wait to Learn logical interventions, independent of their usefulness program, which provides out-of-school children with for better learning. Thus their personal incentives (to computer tablets loaded with learning games, has make highly visible investments) may diverge from shown positive learning impacts in mathematics and the goals of students (to learn). Cognitive bias may is now being tested on a large scale in areas receiv- also be a factor, with individuals being unrealistically ing Syrian refugees.18 In places where teachers are optimistic. In fact, there is a long history of over­ unavailable, such approaches may be the best option. estimating the transformative nature of technology Impacts on literacy and numeracy are not the only in schools, going back to Thomas Edison asserting in measures of success: technology can also promote 1913 that “books will soon be obsolete in schools. . . . digital skills. As more jobs require digital literacy, Our school system will be completely changed in the the opportunity to acquire those skills is an end next ten years.” Edison predicted that books would be in itself. Students with more access to computers entirely replaced by silent films.12 Half a century later, at home have better computer skills.19 And though as computers gained traction, some scholars won- Peru’s One Laptop Per Child program had no effect on dered if they might replace teachers at some point.13 academic achievement or cognitive skills, students Of course, schools in technology-rich environments did significantly improve their knowledge of how to do look different from those elsewhere: students use laptops.20 In such cases, clarity of purpose is key. might do their work on interactive displays rather Obviously, youth need computers to learn how to use than on paper. But technology has for the most part them. But as tools for teaching reading and numeracy, not been particularly disruptive in education. The evidence on their usefulness is mixed. buildings, the processes of the school day, and the interactions between teachers and students are very similar to those of a century ago.14 Other inputs bring learners Technologies that complement teachers work bet- to school—but promote ter than technologies that substitute for them. Many learning only if they target students have poorly prepared teachers with limited training and motivation, and education systems have teaching and learning been tempted to use technology to circumvent these Building schools can increase enrollment in places teachers. Most such attempts have failed. By contrast, with few schools, especially for girls. In Afghanistan, using technology to complement teachers offers the provision of community-based schools in just more promise.15 Consider a computer-assisted learning over a dozen communities increased enrollment program in Gujarat, India, that was implemented in massively, effectively eliminating the gender gap in two ways. One approach pulled students out of reg- enrollment.21 In Burkina Faso, a program to construct ular classes to use computer-based math programs— schools with modern amenities increased enroll- in other words, the program substituted for regular ment by a large margin, with the biggest impacts for Everything else should strengthen the teacher-learner interaction | 147 girls. Even beyond building entire schools, building School management and latrines—particularly gender-specific ones—signifi- cantly increased enrollment of adolescent girls in governance are crucial, and India.22 But in places where learners have relatively involving communities can easy access to schools, additional schools will not help overcome incentive be the most cost-effective way to raise access or to improve learning. problems and information Even in places lacking infrastructure, providing failures—but only if it does not necessarily lead to more learning. The communities have capacity Afghanistan and Burkina Faso programs boosted learning, while India’s did not. Why? Constructing Schools with better management have better test a school where children previously had no access scores.28 Schools vary significantly in management directly alters the learning process by creating a place quality (figure 7.2), and school leadership plays a cru- to learn that did not exist before. Building latrines cial role in school performance. Effective leadership makes school a safe space, and so it makes children means having school principals who are actively (especially girls) better able to be at school—but involved in helping teachers solve problems, includ- because it does not affect what happens in the class- ing by providing instructional advice.29 It also means room, it may not affect learning. having principals who set goals with teachers to School feeding gets children to school, but it does prioritize and achieve high levels of learning. These not always improve learning. The most consistent factors are associated with the highest levels of stu- impact of school-based meal programs has been more dent learning, and they confirm that effective school children in school, such as in Burkina Faso, Kenya, and leadership improves the quality of teacher-learner Peru.23 At school age, providing meals contributes less interactions. A major school district in the United to brain development than earlier in the child’s life, States improved student learning by training school but it could still increase learning through improved attention and energy. However, if meals are offered Figure 7.2 Schools vary significantly in during normal school hours, they reduce time on task. management quality In Kenya and Peru, meals took significant time away from the classroom, and so they had an ambiguous Average school management score by country, relative to top-performing country, participating countries net effect. Impacts on measured learning are mixed, with positive effects in Burkina Faso and Peru. United Kingdom Similarly, simply increasing the materials avail- able at schools does not improve learning if the Sweden materials do not improve teacher-learner interaction. Canada Providing more textbooks in Sierra Leone in 2008 did United States not result in those books being used in the classroom because administrators put most of the books in Germany storage—potentially to hedge against future textbook Mexico shortfalls.24 Another textbook program, in Kenya, had Brazil no impact on learning, most likely because most stu- dents did not fully understand the language in which Italy the books were written.25 Simply providing desktop Haiti computers to classrooms in Colombia—where they were not well integrated in the curriculum—likewise India had no impact on learning.26 It seems obvious that Tanzania resources have to be used to have an impact, but many 0 20 40 60 80 100 interventions that provide inputs fail exactly because Percent insufficient thought is given to how resources will be Source: Bloom and others (2015). Data at http://bit.do/WDR2018-Fig_7-2. used. Infrastructure and other inputs are essential, Note: The school management score is a combination of 14 basic management but they work only when they serve the relationship practices, each rated from 1 to 5. Schools with higher scores have more between teaching and learning.27 structured management practices. 148 | World Development Report 2018 Box 7.1 Training better school principals in Jamaica Training can improve the quality of school management. program, with mentoring and coaching from experienced In Jamaica, the government invested in a school principal school leaders. The training modules subsequently received training program with key characteristics that likely led to high ratings for relevance from participants. Although better management. The program was based on analysis of the program has not been evaluated with a comparison principals’ weaknesses. Principals were trained to provide group, both the principals themselves and the teachers in feedback to teachers on their performance, as well as to their schools report major gains in management quality. use data to evaluate the learning needs of students. The Teachers say they are twice as likely to be observed in their program also provided practical experience: after initial classrooms and to have the principal work with them to training, principals spent three months implementing the develop short-term goals.a Source: WDR 2018 team. a. Nannyonjo (2017). principals in three sets of skills: how to give feedback school autonomy is beneficial to student learning in to teachers on lesson plans; how to support teachers high-income countries but detrimental in developing in regular learner assessments, as well as to provide countries.33 At the micro level, a school-based man- feedback on action plans to improve student perfor- agement intervention in The Gambia improved test mance; and how to, through classroom observation, scores only in communities with high literacy rates give feedback on teacher performance.30 In Mada- among parents.34 A similar result was observed in the gascar, clarifying the management roles of district impacts of a school grant program in Niger.35 Several officers, school principals, and teachers and provid- of these programs did not last more than a year or ing them with coaching and supervision improved two, in some cases because the programs were pilots student outcomes, at least in schools where the heads and in others because of unstable education policies. had good performance incentives.31 Likewise, in Without time for communities to learn how to effec- Jamaica training and mentoring principals improved tively engage in school management, impacts on school management (box 7.1). learning are unlikely. Because communities are more Many countries have decentralized some elements readily able to monitor school enrollment than learn- of their education systems, often called school-based ing, school-based management may increase access management. Providing schools and communities even in low-capacity communities, as happened in with decision-making power and resources can solve Burkina Faso.36 two problems. First, by giving local school leaders and Community monitoring will not help learning if parents more direct influence over teachers and other it does not affect what happens in the classroom. A school representatives, it may make teachers more range of interventions seek to increase community immediately responsive to student needs. Contrast monitoring of schools by sharing school information this with supervision by a ministry of education with parents. The structure of these programs varies, representative based far away, who has little ability from the parents themselves collecting data on teacher to bring shirking teachers to account. Second, schools attendance or school performance, to the education and communities may have better information about systems disseminating prepared data to parents, to the needs of local schools, which, along with access to supplementing information with facilitated meetings discretionary resources, means they can more nimbly in which parents and teachers can discuss grievances meet those needs. and lay out courses of action.37 But parents are rarely School-based management programs improve in the classroom, and even when they are, they can- learning when the community has the capacity to not necessarily identify good classroom practice. This make and implement smarter decisions.32 Data on may explain why the growing evidence on these pro- 1 million students from 42 countries suggest that grams reveals mixed results.38 For example, in Andhra Everything else should strengthen the teacher-learner interaction | 149 Pradesh, India, providing community members with outcomes only when the grants are unanticipated. report cards on school performance failed to increase When parents have known the grants were coming, either parental engagement or student learning.39 they have reduced their own investments in educa- Successful community monitoring increases tion—not a recipe for long-term learning impacts.44 accountability through feedback loops between multi- By themselves, grants act in much the same way as ple stakeholders. Low-stakes accountability programs other interventions that simply increase resources in have improved student learning in Mexico, Pakistan, schools. There is no guarantee that they will improve and Uganda.40 Though some monitoring programs learning. succeed and others fail for multiple reasons, success- But grants can be leveraged in the context of ful programs, such as those in Mexico and Uganda, broader school-based management programs to do not reach out to only one group, but rather share improve school outcomes. In Tanzania, grants alone information explicitly with school leaders and teach- had no impact on student learning, but grants com- ers, as well as with communities and parents. (The bined with teacher incentives did improve learning. In Pakistan experiment is an exception, focusing only on Niger, grants alone had little impact, but grants with parents.) Parents alone cannot enact accountability, training improved both student learning and parent and better information in the hands of school officials support for schools.45 Likewise, grants alone had no helps. To improve learning, parents and communities impact on learning in Indonesia, but they did improve need to be able to harness increased information to learning when the program also linked school man- hold teachers and schools more accountable. agement committees to village councils, seeking to School grant programs—in which schools receive resolve the principal-agent problem.46 When grants regular influxes of resources and more autonomy are included in larger programs to encourage commu- over budget allocations—are one type of school-based nity school councils to engage in school management, management program. In Haiti, a program that pro- the combination can improve learning.47 vided schools with grants based on the number of stu- dents enrolled significantly increased enrollment.41 *** From this perspective, school grants can effectively Whether an education input is a physical item, such deliver money to schools, and schools need money as a tablet or textbook, or a process, such as school to function. But most programs that simply deliver management and leadership, it will improve learning grants to schools do not increase learning. Just distrib- only if it directly improves the quality of teacher- uting grants to schools—as in The Gambia, Indonesia, learner interactions. Without that, more inputs will and Tanzania—had no effect on student learning.42 In pile onto an ineffective process and fail to have the Senegal, observed impacts on learning appeared only desired impact. But used strategically, inputs can work for a subset of children, then disappeared by the next together with prepared learners and knowledgeable, year.43 Some school grant programs increase learning motivated teachers to produce high levels of learning. Notes 1. McEwan (2015). adaptive educational technology, and new evaluations 2. Banerjee and others (2007); Carrillo, Onofa, and Ponce are expected to be published in late 2017. (2010); Muralidharan, Singh, and Ganimian (2016). 8. Bulman and Fairlie (2016); Muralidharan, Singh, and 3. Lai and others (2012). Ganimian (2016). 4. Esteves Pereira and Cabral (2016); “Planning Educational 9. Adelman and others (2015). Activities for Children (PEACH),” Georgia Department 10. Lavinas and Veiga (2013). of Early Care and Learning, Atlanta, http://www.peach 11. Budzier and Flyvbjerg (2012). .decal.ga.gov/app/. 12. Smith (1913, 24). 5. Jukes and others (2017); Wolff and others (2002). 13. Bellissant (1970); Goodlad (1969). 6. Muralidharan, Singh, and Ganimian (2016). 14. Pritchett (2013). 7. Cristia and others (2017); de Melo, Machado, and 15. Snilstveit and others (2016). Miranda (2014). For Uruguay, the evaluation covers 16. Linden (2008). math and reading impacts in the early years of the 17. Powers (2016). program, when its main objective was to provide 18. War Child Holland (2016). equipment and connectivity for schools; the program 19. Kuhlemeier and Hemker (2007). evolved since then to add ICT training for teachers and 20. Beuermann and others (2015). 150 | World Development Report 2018 21. Burde and Linden (2013). Barr, Abigail, Frederick Mugisha, Pieter Serneels, and 22. Adukia (2017). Andrew Zeitlin. 2012. “Information and Collective Action 23. Cueto and Chinen (2008); Kazianga, de Walque, and in Community-Based Monitoring of Schools: Field and Alderman (2009); Vermeersch and Kremer (2005). Lab Experimental Evidence from Uganda.” Working 24. Sabarwal, Evans, and Marshak (2014). paper, Georgetown University, Washington, DC. 25. Glewwe, Kremer, and Moulin (2009). Barrera-Osorio, Felipe, and Leigh L. Linden. 2009. “The Use 26. Barrera-Osorio and Linden (2009). and Misuse of Computers in Education: Evidence from a 27. Ganimian and Murnane (2016). Randomized Experiment in Colombia.” Policy Research 28. Bloom and others (2015). Working Paper Series 4836, World Bank, Washington, DC. 29. Robinson, Lloyd, and Rowe (2008); Waters, Marzano, Beasley, Elizabeth, and Elise Huillery. 2017. “Willing but and McNulty (2003). Unable? Short-Term Experimental Evidence on Parent 30. Fryer (2017). Empowerment and School Quality.” World Bank Economic 31. Lassibille (2016). Review 31 (2): 531–52. 32. Carr-Hill and others (2015). Bellissant, Camille. 1970. “Teaching and Learning Lan- 33. Hanushek, Link, and Woessmann (2013). guages.” In IFIP World Conference on Computer Education, 34. Blimpo, Evans, and Lahire (2015). Vol. 3, edited by Bob Scheepmaker and Karl L. Zinn, 35. Beasley and Huillery (2017). 145–48. New York: Science Associates International. 36. Sawada and others (2016). https://stacks.stanford.edu/file/druid:jd969fg9400/jd969 37. Read and Atinc (2016). fg9400.pdf. 38. Cheng and Moses (2016); Read and Atinc (2016). Beuermann, Diether W., Julian Cristia, Santiago Cueto, Ofer 39. Banerjee and others (2010). Malamud, and Yyannu Cruz-Aguayo. 2015. “One Laptop 40. Andrabi, Das, and Khwaja (2017); Barr and others (2012); Per Child at Home: Short-Term Impacts from a Random- de Hoyos, Garcia-Moreno, and Patrinos (2015). ized Experiment in Peru.” American Economic Journal: 41. Adelman and Holland (2015). Applied Economics 7 (2): 53–80. 42. Blimpo, Evans, and Lahire (2015); Mbiti and others (2017); Blimpo, Moussa P., David K. Evans, and Nathalie Lahire. Pradhan and others (2014). 2015. “Parental Human Capital and Effective School 43. Carneiro and others (2015). Management.” Policy Research Working Paper 7238, 44. Das and others (2013). World Bank, Washington, DC. 45. Kozuka (2017); Mbiti and others (2017). Bloom, Nicholas, Renata Lemos, Raffaella Sadun, and 46. Pradhan and others (2014). John Van Reenen. 2015. “Does Management Matter in 47. Gertler, Patrinos, and Rubio-Codina (2012); Santibañez, Schools?” Economic Journal 125 (584): 647–74. Abreu-Lastra, and O’Donoghue (2014). Budzier, Alexander, and Bent Flyvbjerg. 2012. “Overspend? Late? Failure? What the Data Says about IT Project Risk in the Public Sector.” In Commonwealth Governance Hand- References book 2012/13: Democracy, Development, and Public Administra- tion, edited by Andrew Robertson and Rupert Jones-Parry, Adelman, Melissa, Moussa P. Blimpo, David K. Evans, 145–47. London: Commonwealth Secretariat. Atabanam Simbou, and Noah Yarrow. 2015. “Can Infor- Bulman, George, and Robert W. Fairlie. 2016. “Technology mation Technology Improve School Effectiveness in and Education: Computers, Software, and the Internet.” Haiti? Evidence from a Field Experiment.” Working In Handbook of the Economics of Education, Vol. 5, edited by paper, World Bank, Washington, DC. Eric A. Hanushek, Stephen J. Machin, and Ludger Adelman, Melissa A., and Peter Holland. 2015. “Increasing Woessmann, 239–80. Handbooks in Economics Series. Access by Waiving Tuition: Evidence from Haiti.” Policy Amsterdam: North-Holland. Research Working Paper 7175, World Bank, Washing- Burde, Dana, and Leigh L. Linden. 2013. “Bringing Education ton, DC. to Afghan Girls: A Randomized Controlled Trial of Adukia, Anjali. 2017. “Sanitation and Education.” American Village-Based Schools.” American Economic Journal: Applied Economic Journal: Applied Economics 9 (2): 23–59. Economics 5 (3): 27–40. Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2017. Carneiro, Pedro, Oswald Koussihouèdé, Nathalie Lahire, “Report Cards: The Impact of Providing School and Child Costas Meghir, and Corina Mommaerts. 2015. “Decen- Test Scores on Educational Markets.” American Economic tralizing Education Resources: School Grants in Sene- Review 107 (6): 1535–63. gal.” NBER Working Paper 21063, National Bureau of Banerjee, Abhijit Vinayak, Rukmini Banerji, Esther Duflo, Economic Research, Cambridge, MA. Rachel Glennerster, and Stuti Khemani. 2010. “Pitfalls of Carr-Hill, Roy, Caine Rolleston, Tejendra Pherali, Rebecca Participatory Programs: Evidence from a Randomized Schendel, Edwina Peart, and Emma Jones. 2015. The Evaluation in Education in India.” American Economic Effects of School-Based Decision Making on Educational Out- Journal: Economic Policy 2 (1): 1–30. comes in Low- and Middle-Income Contexts: A Systematic Banerjee, Abhijit Vinayak, Shawn Cole, Esther Duflo, and Review. 3ie Grantee Final Review. London: International Leigh Linden. 2007. “Remedying Education: Evidence Initiative for Impact Evaluation. from Two Randomized Experiments in India.” Quarterly Carrillo, Paul, Mercedes Onofa, and Juan Ponce. 2010. “Infor- Journal of Economics 122 (3): 1235–64. mation Technology and Student Achievement: Evidence Everything else should strengthen the teacher-learner interaction | 151 from a Randomized Experiment in Ecuador.” IDB Work- Jukes, Matthew C. H., Elizabeth L. Turner, Margaret M. ing Paper IDB-WP-223, Inter-American Development Dubeck, Katherine E. Halliday, Hellen N. Inyega, Bank, Washington, DC. Sharon Wolf, Stephanie Simmons Zuilkowski, et al. 2017. Cheng, Xuejiao Joy, and Kurt Moses. 2016. Promoting “Improving Literacy Instruction in Kenya through Transparency through Information: A Global Review of School Teacher Professional Development and Text Messages Report Cards. Ethics and Corruption in Education Series. Support: A Cluster Randomized Trial.” Journal of Research Paris: International Institute for Educational Planning, on Educational Effectiveness 10 (3): 449–81. United Nations Educational, Scientific, and Cultural Kazianga, Harounan, Damien de Walque, and Harold Organization. Alderman. 2009. “Educational and Health Impacts of Cristia, Julián, Pablo Ibarrarán, Santiago Cueto, Ana San- Two School Feeding Schemes: Evidence from a Random- tiago, and Eugenio Severín. 2017. “Technology and Child ized Trial in Rural Burkina Faso.” Policy Research Work- Development: Evidence from the One Laptop Per Child ing Paper 4976, World Bank, Washington, DC. Program.” American Economic Journal: Applied Economics Kozuka, Eiji. 2017. “Enlightening Communities and Parents 9 (3): 295–320. for Improving Student Learning: Evidence from Ran- Cueto, Santiago, and Marjorie Chinen. 2008. “Educational domized Experiment in Niger.” Working Paper, JICA Impact of a School Breakfast Programme in Rural Peru.” Research Institute, Tokyo. International Journal of Educational Development 28 (2): Kuhlemeier, Hans, and Bas Hemker. 2007. “The Impact of 132–48. Computer Use at Home on Students’ Internet Skills.” Das, Jishnu, Stefan Dercon, James Habyarimana, Pramila Computers and Education 49 (2): 460–80. Krishnan, Karthik Muralidharan, and Venkatesh Sunda- Lai, Fang, Linxiu Zhang, Qinghe Qu, Xiao Hu, Yaojiang Shi, raraman. 2013. “School Inputs, Household Substitution, Matthew Boswell, and Scott Rozelle. 2012. “Does and Test Scores.” American Economic Journal: Applied Eco- Computer-Assisted Learning Improve Learning Out- nomics 5 (2): 29–57. comes? Evidence from a Randomized Experiment in de Hoyos, Rafael E., Vicente A. Garcia-Moreno, and Harry Public Schools in Rural Minority Areas in Qinghai, Anthony Patrinos. 2015. “The Impact of an Accountability China.” REAP Working Paper 237, Rural Education Action Intervention with Diagnostic Feedback: Evidence from Program, Freeman Spogli Institute, Stanford University, Mexico.” Policy Research Working Paper 7393, World Stanford, CA. Bank, Washington, DC. Lassibille, Gérard. 2016. “Improving the Management Style de Melo, Gioia, Alina Machado, and Alfonso Miranda. 2014. of School Principals: Results from a Randomized Trial.” “The Impact of a One Laptop Per Child Program on Education Economics 24 (2): 121–41. Learning: Evidence from Uruguay.” IZA Discussion Paper Lavinas, Lena, and Alinne Veiga. 2013. “Brazil’s One Laptop 8489, Institute for the Study of Labor, Bonn, Germany. Per Child Program: Impact Evaluation and Implementa- Esteves Pereira, Lucia Helena, and Isabel Cabral. 2016. “Ges- tion Assessment.” Cadernos de Pesquisa 43 (149). tão Escolar: A Opinião dos Profissionais de Educação Linden, Leigh L. 2008. “Complement or Substitute? The sobre o Sistema de Tecnologia do Estado do Rio de Effect of Technology on Student Achievement in India.” Janeiro.” Regae, Revista de Gestão e Avaliação Educacional Edited by Michael Trucano. InfoDev Working Paper 17, 4 (7): 47–60. https://periodicos.ufsm.br/regae/article/view World Bank, Washington, DC. /14875. Mbiti, Isaac M., Karthik Muralidharan, Mauricio Romero, Fryer, Roland G., Jr. 2017. “Management and Student Youdi Schipper, Constantine Manda, and Rakesh Rajani. Achievement: Evidence from a Randomized Field Exper- 2017. “Inputs, Incentives, and Complementarities in Edu- iment.” NBER Working Paper 23437, National Bureau of cation: Experimental Evidence from Tanzania.” Working Economic Research, Cambridge, MA. paper, University of California at San Diego. Ganimian, Alejandro J., and Richard J. Murnane. 2016. McEwan, Patrick J. 2015. “Improving Learning in Primary “Improving Education in Developing Countries: Lessons Schools of Developing Countries: A Meta-Analysis of from Rigorous Impact Evaluations.” Review of Educational Randomized Experiments.” Review of Educational Research Research 86 (3): 719–55. 85 (3): 353–94. Gertler, Paul J., Harry Anthony Patrinos, and Marta Muralidharan, Karthik, Abhijeet Singh, and Alejandro J. Rubio-Codina. 2012. “Empowering Parents to Improve Ganimian. 2016. “Disrupting Education? Experimental Education: Evidence from Rural Mexico.” Journal of Devel- Evidence on Technology-Aided Instruction in India.” opment Economics 99 (1): 68–79. NBER Working Paper 22923, National Bureau of Eco- Glewwe, Paul W., Michael R. Kremer, and Sylvie Moulin. nomic Research, Cambridge, MA. 2009. “Many Children Left Behind? Textbooks and Test Nannyonjo, Harriet. 2017. “Building Capacity of School Lead- Scores in Kenya.” American Economic Journal: Applied Eco- ers: Strategies That Work, Jamaica’s Experience.” Work- nomics 1 (1): 112–35. ing paper, World Bank, Washington, DC. Goodlad, John I. 1969. “Computers and the Schools in Mod- Powers, Shawn. 2016. “The Impact of Ebola on Education in ern Society.” Proceedings of the National Academy of Sciences Sierra Leone.” Education for Global Development (blog), of the United States of America 63 (3): 595–603. May 4. http://blogs.worldbank.org/education/impact Hanushek, Eric A., Susanne Link, and Ludger Woessmann. -ebola-education-sierra-leone. 2013. “Does School Autonomy Make Sense Everywhere? Pradhan, Menno, Daniel Suryadarma, Amanda Beatty, Panel Estimates from PISA.” Journal of Development Eco- Maisy Wong, Arya Gaduh, Armida Alisjahbana, and nomics 104: 212–32. Rima Prama Artha. 2014. “Improving Educational Quality 152 | World Development Report 2018 through Enhancing Community Participation: Results Smith, Frederick James. 1913. “The Evolution of the Motion from a Randomized Field Experiment in Indonesia.” Picture, VI: Looking into the Future with Thomas A. American Economic Journal: Applied Economics 6 (2): 105–26. Edison.” New York Dramatic Mirror (July 3). Pritchett, Lant. 2013. The Rebirth of Education: Schooling Ain’t Snilstveit, Birte, Jennifer Stevenson, Radhika Menon, Learning. Washington, DC: Center for Global Develop- Daniel Phillips, Emma Gallagher, Maisie Geleen, Han- ment; Baltimore: Brookings Institution Press. nah Jobse, et al. 2016. “The Impact of Education Pro- Read, Lindsay, and Tamar Manuelyan Atinc. 2016. “Infor­ grammes on Learning and School Participation in Low- mation for Accountability: Transparency and Citizen and Middle-Income Countries: A Systematic Review Engagement for Improved Service Delivery in Education Summary Report.” 3ie Systematic Review Summary 7, Systems.” Global Economy and Development Working International Initiative for Impact Evaluation, London. Paper 99, Brookings Institution, Washington, DC. http://www.3ie impact.org/media/filer_public/2016/09/20 Robinson, Viviane M. J., Claire A. Lloyd, and Kenneth J. /srs7-education-report.pdf. Rowe. 2008. “The Impact of Leadership on Student Out- Vermeersch, Christel M. J., and Michael R. Kremer. 2005. comes: An Analysis of the Differential Effects of Leader- “School Meals, Educational Achievement, and School ship Types.” Educational Administration Quarterly 44 (5): Competition: Evidence from a Randomized Evaluation.” 635–74. Policy Research Working Paper 3523, World Bank, Wash- Sabarwal, Shwetlena, David K. Evans, and Anastasia Mar- ington, DC. shak. 2014. “The Permanent Input Hypothesis: The Case War Child Holland. 2016. “Can’t Wait to Learn.” War Child of Textbooks and (No) Student Learning in Sierra Leone.” Holland, Amsterdam. https://www.warchild.nl/sites Policy Research Working Paper 7021, World Bank, Wash- /default/files/bijlagen/node_13537/27-2016/2016_6_pager ington, DC. _cant_wait_to_learn_english.pdf. Santibañez, Lucrecia, Raúl Abreu-Lastra, and Jennifer L. Waters, Tim, Robert J. Marzano, and Brian McNulty. 2003. O’Donoghue. 2014. “School Based Management Effects: “Balanced Leadership: What 30 Years of Research Tells Resources or Governance Change? Evidence from Mex- Us about the Effect of Leadership on Student Achieve- ico.” Economics of Education Review 39 (April): 97–109. ment.” McRel Working Paper, McRel International, Sawada, Yasuyuki, Takeshi Aida, Andrew Griffen, Harounan Denver. Kazianga, Eiji Kozuka, Haruko Nogushi, and Yasuyuki Wolff, Laurence, Claudio de Moura Castro, Juan Carlos Todo. 2016. “On the Role of Community Management in Navarro, and Norma García. 2002. “Television for Sec- Correcting Market Failures of Rural Developing Areas: ondary Education: Experience of Mexico and Brazil.” Evidence from a Randomized Field Experiment of In Technologies for Education: Potentials, Parameters, and COGES Project in Burkina Faso.” Paper presented at Prospects, edited by Waddi D. Haddad and Alexandria annual meeting of Agricultural and Applied Economics Draxler, 144–52. Washington, DC: Academy for Educa- Association, Boston, July 31–August 2. http://ageconsearch tional Development; Paris: United Nations Educational, .umn.edu/bitstream/236323/2/SelectedPaper_9662.pdf. Scientific, and Cultural Organization. Everything else should strengthen the teacher-learner interaction | 153 8 Build on foundations by linking skills training to jobs After leaving school—whether as dropouts or graduates—many young people land jobs with limited prospects. But training offers a way out. How can successful job skills training programs be replicated? How can they be made available, affordable, and effective for the many young job seekers moving from school to work? Young people around the world face substantial chal- Technical and vocational education and training •  lenges in their transition from school to work. Many (TVET) offers a viable path, but only when programs of them, especially youth from disadvantaged back- are designed and implemented in partnership with grounds, leave formal education prematurely, lacking employers. the foundational skills needed to succeed on the job. In other words, the learning crisis manifests itself in the labor market. As a result, many become unemployed or Workplace training can help stuck in low-wage, unstable, informal-sector jobs that young people develop skills, offer them few opportunities to strengthen their skills. But the same can happen even to secondary school yet few benefit from it graduates, if they cannot fulfill labor market needs. Workplace training deepens workers’ skills and raises When young people leave formal education, they firms’ productivity.2 It can increase workers’ output usually take one of three paths to employment. Some by 10 percent or more, which is similar in magnitude join the labor market without any further education or to the payoff from investments in physical capital.3 In training. For them, workplace training is an important Latin America and the Caribbean, a 1 percent increase in way to build skills. Others enroll in formal technical the proportion of trained workers in large firms raised or vocational training programs that build the skills productivity by 0.7 percent.4 In Mexico, investments in required for specific fields or occupations of interest.1 training increased productivity and firm-level wages These programs usually result in a formal technical by 4–7 percent for manufacturing workers.5 Similarly, qualification or an industry-recognized certification. returns were 7.7 percent in Malaysia and 4.5 percent Finally, a smaller group postpones looking for work or in Thailand for workers holding a secondary educa- enrolling in further education and training. Three types tion qualification or more.6 In Kenya and Zambia, of job training programs can help youth improve along workplace training was associated with a 20 percent these paths: increase in the wages of manufacturing workers.7 Despite its potential benefits, young workers Workplace training can benefit both workers and •  rarely receive workplace training. In developing firms, yet it is not widely available to young adults. countries, the percentage of working-age adults par- Short-term job training programs often have lim- •  ticipating in work-related training ranges from 20 ited impacts, but careful program design could help percent in urban Bolivia and Colombia to less than improve outcomes. 10 percent in the Lao People’s Democratic Republic 154 | World Development Report 2018 Figure 8.1 Few benefit from workplace training, and those who do tend to already have better literacy or education Workplace training participation in last 12 months, participating countries (2011–14) a. By age b. By literacy proficiency a c. By education Armenia Armenia Armenia Bolivia Bolivia Bolivia Colombia Colombia Colombia Georgia Georgia Georgia Ghana Ghana Ghana Kenya Kenya Kenya Lao PDR Lao PDR Lao PDR Macedonia, FYR Macedonia, FYR Macedonia, FYR Sri Lanka Sri Lanka Sri Lanka Ukraine Ukraine Ukraine Vietnam Vietnam Vietnam Yunnan (China) Yunnan (China) Yunnan (China) 0 5 10 15 20 25 0 5 10 15 20 25 0 5 10 15 20 25 Percent Percent Percent Ages 25–44 Level 2 and above Upper secondary and above Ages 15–24 Level 1 and below Lower secondary and below Source: WDR 2018 team, using data from World Bank’s STEP Skills Measurement Program (http://microdata.worldbank.org/index.php/catalog/step/about). Data at http://bit.do/ WDR2018-Fig_8-1. Note: Respondents were asked, “In the past 12 months, have you participated in any training courses, such as work-related training or private skills training, that lasted at least 5 days/ 30 hours (not part of the formal educational system)?” Low proficiency is defined as level 1 and below on the literacy assessments and indicates limited understanding of texts. Medium to high proficiency is defined as level 2 and above and indicates the ability to integrate, evaluate, and interpret information from a variety and complexity of text materials. No literacy proficiency data available for Lao PDR; Macedonia, FYR; Sri Lanka; or Yunnan (China). a.  and Vietnam.8 Training participation is even lower a way to upgrade their skills in a workplace setting. for young people with incomplete education, limited In these noncertification-granting arrangements, skills, or short employment tenure.9 In Peru, fewer learning takes place while a young person works than one in five young workers receive training in alongside an experienced craftsperson over a period their first year on the first job.10 Employers’ decisions of time.13 Though available in many parts of the to invest in training are affected by potential produc- world, informal apprenticeships are most common in tion improvements, worker turnover, and a firm’s Sub-Saharan Africa. For example, in Benin, Cameroon, overall management practices.11 Training participa- Côte d’Ivoire, and Senegal, informal apprenticeships tion is lower not just for young workers generally account for almost 90 percent of the training that pre- (figure 8.1, panel a), but especially for young workers pares workers for crafts jobs, as well as employment in with limited literacy proficiency or education quali- some trades (such as carpentry, welding, hairdressing, fications (figure 8.1, panels b and c). Yet workplace plumbing, tailoring, masonry, and weaving).14 Infor- training can be especially beneficial for young adults. mal apprentices are more likely to be young people A cross-country analysis of 38 workplace training with limited formal education from disadvantaged studies finds an average wage increase of 7.2 percent socioeconomic backgrounds.15 These apprenticeships for workers under 35, compared with 4.9 percent for vary widely in their institutional setup, training con- workers over 35.12 tent, working conditions, and financial arrangements. Informal apprenticeships, which can be thought of However, most are nested within community cus- as informal workplace traineeships, offer young people toms, norms, and traditions. Experimental evidence Build on foundations by linking skills training to jobs | 155 on their effectiveness is scant. Evaluations from Sen- Empowerment and Livelihoods for Adolescents pro- egal have shown positive effects on labor market out- gram targeting young women shows encouraging comes, but limited effects on general cognitive skills.16 impacts on graduates’ employment prospects.25 Simi- But early evidence from an apprenticeship program in larly, Nepal’s Adolescent Girls Employment Initiative Côte d’Ivoire that formalized part of the process shows increases nonfarm employment by 13–19 percentage improvements in the labor market outcomes and psy- points for participants.26 In the Dominican Republic, chological well-being of disadvantaged youth.17 evaluations of Programa Juventud y Empleo, a skills Unlocking the potential of informal appren- training program that targets low-income, low-skilled, ticeships requires up-to-date master trainers and out-of-school youth with less than a secondary edu- recognition of apprentices’ training tenure and cation, increased both employment and earnings.27 performance. Too often, master trainers lack the Promising results from interventions in Colombia, information, capacity, and incentives to adapt to new the Dominican Republic, Liberia, Nepal, and Peru are workplace practices. This can lead to apprentices identifying effective approaches to improving young learning obsolete workplace practices.18 Also, because women’s aspirations, socioemotional skills, and labor informal apprenticeships are rarely recognized by market outcomes.28 the formal training system, they offer limited labor Successful short-term job training programs market mobility.19 One way to mitigate this issue is offer more than skills training. Programs that focus to integrate informal apprenticeships into the formal on developing multiple skills and that complement training system, allowing for skills reengagement training with wraparound services such as career with further education and training. In Malawi and guidance, mentoring, and job search assistance have Tanzania, for example, competency-based skills certi- better odds of success.29 For example, comprehensive fication offers a pathway for young workers who have training schemes that emphasize technical skills, been apprentices to be acknowledged for their skills.20 life skills, and internships show positive effects in Kenya, Brazil, and Nepal.30 In Kenya, the Ninaweza Youth Empowerment program, which integrates Short-term job training information and communication technology (ICT), offers opportunities, but life skills, internship training, and job placement support for youth, shows positive impacts on labor most programs fail to deliver market outcomes.31 Similarly, in Brazil the Galpão Many short-term job training programs—which usu- Aplauso program has improved outcomes through a ally last between two weeks and six months—do not combination of  vocational, academic, and life skills meet labor market needs. Meta-analyses of programs training.32 In Nepal, the Employment Fund prioritizes from around the world find that less than a third have comprehensive training programs for youth who are positive, significant impacts on employment and underemployed or unemployed.33 earnings.21 Though the estimated effects of short-term programs are somewhat larger in developing coun- tries, they remain small. Skills training that focuses TVET can prepare young on helping participants accumulate the human capi- people for work, but early tal needed to transition to labor markets can generate sorting into TVET can limit positive returns, but, given their short duration and heterogeneous quality, these short-term programs career growth rarely have impacts as large as the returns from TVET can yield wages on par with equivalent levels completing a formal education.22 Many programs are of general education. Usually lasting from six months poorly designed and implemented, or don’t interest to three years, TVET can be delivered in the dedi- the hard-to-reach young people who might need cated streams of lower secondary, upper secondary, skills upgrading the most.23 The economic rationale or tertiary schools.34 In Brazil, workers with upper for investing in training is often tenuous: in Liberia, secondary TVET earn wages about 10 percent higher for example, it can cost up to 50 times the resulting than those of workers with a general secondary monthly income gain, meaning that recovering the education.35 In Indonesia, returns to public TVET are investment would take 12 years.24 positive for all, and greater for women.36 But despite But short-term training interventions do show encouraging results, TVET programs in many devel- some positive results when targeting disadvantaged oping countries remain an unattractive alternative groups, such as low-skilled women. In Uganda, the 156 | World Development Report 2018 Figure 8.2 Most vocational training students enroll during upper secondary school Gross enrollment rates in general education and technical and vocational education and training (TVET), lower and upper secondary students (circa 2010) 100 80 60 Percent 40 20 0 Lower Upper Lower Upper Lower Upper Lower Upper Lower Upper Lower Upper Lower Upper East Asia Europe and Latin America Middle East South Asia Sub-Saharan World and Pacific Central Asia and the and North Africa Caribbean Africa Left the school system Enrolled in general education Enrolled in vocational education Source: WDR 2018 team, using data from UIS (2016). Data at http://bit.do/WDR2018-Fig_8-2. for young people, with often poor program quality or conveys early advantages in the labor market, but the labor market relevance. advantages dissipate over time. Some workers end up Putting students on a technical track too early can outdated in their occupation-specific skills, making limit their career opportunities for a lifetime. Young them more vulnerable to job loss.39 Though the appro- people need to master foundational skills—reading, priate balance is bound to be country-specific, TVET writing, numeracy, critical thinking, and problem- should not lock participants into narrow occupations solving—to participate meaningfully in TVET. They that are likely to change in unanticipated ways. also need to be mature enough to express career preferences that might have long-term consequences. Countries that have delayed streaming into TVET have Successful job training shown that such changes can lead to improvements. programs share several In Poland, delaying vocational education by a year improved students’ academic performance.37 Problems features with early tracking are exacerbated in systems that Though the evidence available on workplace training do not allow students to go back and forth between and job training interventions—whether short-term general and technical education, leaving technical or long-term—is limited, some features are consistent graduates with limited opportunities to reengage in across successful programs. To the extent possible, further education or training.38 Despite such concerns, the principles discussed in this section are distilled most enrollment in TVET occurs at the start of upper from experimental evidence. But because of the secondary school (figure 8.2). Of equal concern is that, shortage of rigorous research on interventions in in most regions, more young people are leaving formal developing countries, this section also integrates rel- education than are continuing in either general educa- evant findings from different types of studies (non­ tion or TVET—a fact that underscores the importance experimental, systematic, qualitative). of acquiring robust foundational skills early on as a basis for learning on the job and throughout life. Establishing partnerships before training is Developing narrow vocational skills can expedite designed workers’ transitions into the labor force, but broader Sectoral training programs40 partner learners with general skills can help them adapt more easily to employers early and sustain their commitment.41 technological change. Evidence from advanced These programs set up partnerships between inter- economies indicates that narrow technical education mediary institutions—usually network aggregators Build on foundations by linking skills training to jobs | 157 or nonprofits with industry-specific expertise—and arrangements, and an assessment to verify acquired employers in an industry to anticipate job openings, skills.51 A strong partnership between the education design program content, and maximize potential system and industry is crucial to integrate firm placement. Sectoral programs focus on supporting resources, share risk burdens, develop industrywide individuals to enter careers rather than jobs. To do so, skill standards, and deliver apprenticeship training at programs integrate information on career pathways scale. to help participants identify the credentials that are Studies show positive results for both firms and the needed for an occupation and that can be pursued to individuals who complete formal apprenticeships.52 move from an entry-level job to a longer-term career.42 In the United States, a study looking at gains from Success factors include having high-quality intermedi- secondary TVET, postsecondary TVET, and appren- aries, along with comprehensive recruitment services, ticeship programs in the states of Virginia and Wash- to generate good matches among prospective partici- ington found positive gains from all three—especially pants, programs of study, and targeted occupations.43 apprenticeships.53 Studies in Canada, Germany, Sectoral training programs can improve labor Switzerland, and the United States find that employ- market outcomes, raise productivity, and reduce ers recover initial apprentice costs in the short to employee turnover. Among three U.S. sectoral training medium term.54 In Brazil, graduates of a large formal programs—Wisconsin Regional Training Partnership apprenticeship program (Lei do Aprendiz) are more (Milwaukee), Jewish Vocational Service (Boston), Per likely to find permanent, higher-paying jobs, with Scholas (New York City)—participants saw 18 percent larger gains for less educated workers.55 In Malawi, an higher average earnings over a two-year period.44 Sim- innovative formal apprenticeship program targeting ilarly, the Year Up program, which targets vulnerable young women broadened their opportunities to serve youth in several U.S. states, has produced high levels as assistant schoolteachers; graduates gained higher of completion, participation in internships, employ- skills and community standing.56 ment, and earnings.45 Finally, the Generation pro- gram—focusing on low-skilled youth in India, Kenya, Identifying capable teachers Mexico, Spain, and the United States—has resulted in Successful approaches to training depend on capa- high job placement and employer satisfaction.46 Other ble teachers57 with industry expertise who can tailor potential approaches to engaging employers in train- training to meet job requirements.58 Students’ gaps in ing include entering into public-private partnerships foundational skills and lack of motivation intensify with multinational corporations, establishing effec- the complexity of teachers’ roles and responsibilities.59 tive national workforce development initiatives, and The global shift toward competency-based standards fostering workplace training provision through mech- in training, assessment, and certification amplifies anisms such as training funds and tax incentives.47 the importance of capable, involved teachers.60 A study of 10 polytechnics in Ghana highlights the importance Combining classroom with workplace of having teachers able to offer constructive feedback learning as students work through competency modules.61 Formal apprenticeships are a common way to However, often teachers lack industry qualifications combine classroom with workplace learning; such or up-to-date pedagogical expertise, especially when programs are often referred to as “learning while it comes to teaching using a competency-based skills earning.” Formal apprentice programs can last from approach. A study of teaching practices in technical one to three years and take place at the secondary or vocational colleges in Malaysia highlights the diffi- postsecondary level or as an alternative to upper sec- culties that teachers face in moving from assessing ondary education, giving students the opportunity to a student’s knowledge to assessing occupational and engage in industry-supervised workplace practices.48 task-specific competency.62 For programs targeting secondary students, special Yet in many countries, structured professional attention is required to ensure apprentices hone support is not available to ensure that TVET teach- foundational skills, as well as occupation-specific ers remain current on curriculums and industry skills, to avoid overly narrow specialization.49 Formal changes.63 But they could: a review of vocational edu- apprentices are typically paid less than the market cation systems in 10 countries in the Middle East and wage.50 Good-practice apprenticeships offer structured North Africa (MENA) finds active experimentation training, a professional trainer to oversee appren- with innovative models to build career structures tices, a written contract that stipulates training that reflect common norms, values, and standards to 158 | World Development Report 2018 professionalize TVET teaching. Six of the 10 countries trajectories, and outcomes.68 Career guidance poli- have developed occupational standards for technical cies are a priority across 28 European countries, yet teachers to recognize career progression, though it the scope and depth of programs vary substantially, is too early to tell whether or how the new standards highlighting the need for a well-articulated vision, are influencing student outcomes.64 Other countries, cohesive strategy, and robust quality assurance such as Ethiopia and Lao PDR, are experimenting mechanisms linked to funding.69 Few member coun- with introducing standards and expanding the tries of the Organisation for Economic Co-operation qualifications for technical training instructors. But and Development have program standards to mon- getting robust information on program effectiveness itor the quality of services, especially for programs is difficult because most interventions are not evalu- delivered by private providers. This results in an ated for impact.65 overreliance on staff qualifications as indicators of quality.70 Making student support services and Successful career guidance programs have clear comprehensive information available for objectives and outcome measurement to track pro- decision making gram performance. They also have different pathways Career information is an important part of training for participants from a diversity of backgrounds, so programs, helping students identify opportunities, skilled career guidance staff can tailor skills develop- stay on course, and transition into a career.66 Career ment trajectories according to need.71 information interventions are usually grouped into career education programs, which might include pro- *** viding direction on coursework selection, and career Successful job training programs are typically based planning, which is usually provided on an individual on strong ties with employers, with curriculums basis.67 Career information can be especially useful taught by teachers who have both industry experi- for students who lack family or social networks that ence and up-to-date pedagogical expertise. These can provide meaningful direction. Since the early programs also tend to reinforce foundational skills, 2000s, countries in the European Union have been integrate classroom instruction with workplace experimenting with mechanisms to integrate career learning, and offer certifications that can be further guidance with national lifelong learning strategies built on. These features keep career paths open for in order to align with the Lisbon Strategy and the graduates. Though job training programs can yield strategic framework for European cooperation on positive outcomes, a key lesson is that trainees education and training. still need strong foundational skills—cognitive and Still, evidence is limited on how career infor- socioemotional—before moving into specialized mation initiatives affect students’ choices, training streams. Notes 1. Pre-employment job training programs can be grouped 3. Dearden, Reed, and Van Reenen (2006); De Grip and into (1) short-term programs of less than six months Sauermann (2012); Konings and Vanormelingen (2015); that focus on vocational subjects and (2) longer-term Saraf (2017). technical and vocational education and training (TVET) 4. González-Velosa, Rosas, and Flores (2016). programs of more than six months that are mapped to 5. Tan and López-Acevedo (2003). formal education system levels. 6. Almeida and de Faria (2014). 2. Formal workplace training refers to supervised skills 7. Rosholm, Nielsen, and Dabalen (2007). development activity that links knowledge gained in 8. Roseth, Valerio, and Gutiérrez (2016). the workplace with the needs of business firms (see ILO 9. Almeida and Aterido (2010); Cabrales, Dolado, and Mora 2010). A comparison across workplace models is difficult (2014); Sousounis and Bladen-Hovell (2010). because of the heterogeneity in design, implementation, 10. Cavero and Ruiz (2016). and effectiveness of training schemes. See Acemoglu 11. Saraf (2017). and Pischke (1996); Almeida, Behrman, and Robalino 12. Haelermans and Borghans (2012). (2012); Almeida and Carneiro (2009); Bassanini and oth- 13. ILO (2012). ers (2005); Blundell and others (1999); Dearden, Reed, 14. ILO (2012). and Van Reenen (2006); and Haelermans and Borghans 15. Adams and others (2009); Darvas, Farvara, and Arnold (2012). (2017); ILO (2012). Build on foundations by linking skills training to jobs | 159 16. Aubery, Giles, and Sahn (2017). education), and trainers (workplace training). See 17. World Bank (2016). Axmann, Rhoades, and Nordstrum (2015) and Stanley, 18. ILO (2012). Adubra, and Chakroun (2014). 19. ILO (2012). 58. Axmann, Rhoades, and Nordstrum (2015); Biavaschi and 20. Aggarwal, Hofmann, and Phiri (2010); Nübler, Hofmann, others (2012); Grollmann (2008); Maclean and Lai (2011). and Greiner (2009). 59. Hodge (2016). 21. Kluve and others (2016); McKenzie (2017). 60. Guthrie and others (2009); ILO (2010). Experimental 22. McKenzie (2017). studies evaluating the impact of different approaches to 23. Blattman and Ralston (2015); Kluve and others (2016); training and supporting the professional development LaLonde (2003); McKenzie (2017). of vocational teachers are extremely rare. 24. Adoho and others (2014); Blattman and Ralston (2015); 61. Boahin and Hofman (2014). McKenzie (2017). 62. Azmanirah and others (2014). 25. Bandiera and others (2014). 63. Axmann, Rhoades, and Nordstrum (2015). 26. Chakravarty and others (2016). 64. OECD (2010, 2014); UNESCO (2014). 27. Card and others (2011). 65. Gerds (2009); Kingombe (2012); Soysouvanh (2013). 28. Fox and Kaul (2017). 66. OECD and EC (2004); Watts and Sultana (2004). 29. Eichhorst and others (2012); Fares and Puerto (2009); 67. OECD (2010). Kluve and others (2016). 68. Hooley (2014); Hooley and Dodd (2015); Kluve and others 30. Fox and Kaul (2017). (2016); OECD (2010); Sultana and Watts (2008). 31. IYF (2013). 69. Watts, Sultana, and McCarthy (2010). 32. Calero and others (2014). 70. OECD and EC (2004). 33. Chakravarty and others (2016). 71. OECD and EC (2004). 34. OECD (2014); Tan and Nam (2012). 35. Almeida and others (2015). 36. Newhouse and Suryadarma (2011). References 37. Jakubowski and others (2016). Acemoglu, Daron, and Jorn-Steffen Pischke. 1996. “Why Do 38. Biavaschi and others (2012). Firms Train? Theory and Evidence.” NBER Working 39. Hampf and Woessmann (2016); Hanushek and others Paper 5605, National Bureau of Economic Research, (2017). Cambridge, MA. 40. Sectoral training programs are defined as partnership Adams, Arvil V., Harold Coulombe, Quentin Wodon, and arrangements between the government, employers, Setarah Razmara. 2009. “Education, Skills, and Labor and nonprofit organizations set up to train unemployed Market Outcomes in Ghana.” Working paper, World or underemployed adults. They usually target young Bank, Washington, DC. adults who have incomplete upper secondary or tertiary Adoho, Franck M., Shubha Chakravarty, Dala T. Korkoyah, Jr., education qualifications whose skills are relatively low, Mattias K. A. Lundberg, and Afia Tasneem. 2014. “The as well as disadvantaged young workers seeking skills Impact of an Adolescent Girls Employment Program: upgrading programs to reenter the labor force or move The EPAG Project in Liberia.” Policy Research Working into higher-quality jobs. Paper 6832, World Bank, Washington, DC. 41. CED (2015); Conway and Giloth (2014); King (2014); Aggarwal, Ashwani, Christine Hofmann, and Alexander Martinson (2010); NGA (2013). Phiri. 2010. “A Study on Informal Apprenticeship in 42. Bragg, Dresser, and Smith (2012). Malawi.” Employment Report 9, International Labour 43. King (2014); Maguire and others (2010). Office, Geneva. 44. Maguire and others (2010). Almeida, Rita Kullberg, Leandro Anazawa, Naercio Menezes 45. Roder and Elliott (2011). Filho, and Lígia Maria De Vasconcellos. 2015. “Investing in 46. Mourshed, Farrell, and Barton (2013). Technical and Vocational Education and Training: Does It 47. Dunbar (2013); Tan and others (2016). Yield Large Economic Returns in Brazil?” Policy Research 48. Fazio, Fernández-Coto, and Ripani (2016); Mieschbue- Working Paper 7246, World Bank, Washington, DC. hler and Hooley (2016); Neumark and Rothstein (2006). Almeida, Rita Kullberg, and Reyes Aterido. 2010. “Investment 49. OECD (2010). in Job Training: Why Are SMES Lagging So Much 50. Biavaschi and others (2012); Smith and Kemmis (2013). Behind?” Policy Research Working Paper 5358, World 51. Cumsille (2016); Fazio, Fernández-Coto, and Ripani Bank, Washington, DC. (2016); Smith and Kemmis (2013). Almeida, Rita Kullberg, Jere R. Behrman, and David Robalino, 52. Dietrich, Pfeifer, and Wenzelmann (2016); Hollenbeck eds. 2012. The Right Skills for the Job? Rethinking Training (2008); Lerman (2014); Smith and Kemmis (2013). Policies for Workers. Report 70908, Human Development 53. Hollenbeck (2008). Perspectives Series. Washington, DC: World Bank. 54. Lerman (2013, 2014). Almeida, Rita Kullberg, and Pedro Manuel Carneiro. 2009. 55. Corseuil, Foguel, and Gonzaga (2014). “The Return to Firm Investments in Human Capital.” 56. Safford and others (2013). Labour Economics 16 (1): 97–106. 57. Here teachers is defined broadly to include teachers Almeida, Rita Kullberg, and Marta Lince de Faria. 2014. “The (secondary education), instructors (postsecondary Wage Returns to on-the-Job Training: Evidence from 160 | World Development Report 2018 Matched Employer-Employee Data.” IZA Journal of Labor Janeiro.” IZA Discussion Paper 8210, Institute for the and Development 3 (1): 1–33. Study of Labor, Bonn, Germany. Aubery, Frédéric, John Giles, and David E. Sahn. 2017. “Do Card, David, Pablo Ibarrarán, Ferdinando Regalia, David Apprenticeships Provide Skills Beyond the Master’s Rosas-Shady, and Yuri Soares. 2011. “The Labor Market Trade? Evidence on Apprenticeships, Skills and the Impacts of Youth Training in the Dominican Republic.” ­ Transition to Work in Senegal.” World Bank, Washing- Journal of Labor Economics 29 (2): 267–300. ton, DC. Cavero, Denice, and Claudia Ruiz. 2016. “Do Working Condi- Axmann, Michael, Amy Rhoades, and Lee Nordstrum. 2015. tions in Young People’s First Jobs Affect Their Employ- “Vocational Teachers and Trainers in a Changing World: ment Trajectories? The Case of Peru.” Work4Youth 33, The Imperative of High-Quality Teacher Training Sys- International Labour Office, Geneva. tems.” With contributions from Josée-Anne La Rue and CED (Committee for Economic Development of the Confer- Michelle Byusa. Employment Working Paper 177, Inter- ence Board). 2015. “The Role of Business in Promoting national Labour Organization, Geneva. Educational Attainment: A National Imperative.” CED, Azmanirah Ab Rahman, Nurfirdawati Muhamad Hanafi, Arlington, VA. Marina Ibrahim Mukhtar, and Jamil Ahmad. 2014. Chakravarty, Shubha, Mattias K. A. Lundberg, Plamen “Assessment Practices for Competency Based Education Nikolov Danchev, and Juliane Zenker. 2016. “The Role of and Training in Vocational College, Malaysia.” Procedia: Training Programs for Youth Employment in Nepal: Social and Behavioral Sciences 112: 1070–76. Impact Evaluation Report on the Employment Fund.” Bandiera, Oriana, Niklas Buehren, Robin Burgess, Markus P. Policy Research Working Paper 7656, World Bank, Wash- Goldstein, Selim Gulesci, Imran Rasul, and Munshi ington, DC. Sulaiman. 2014. “Women’s Empowerment in Action: Evi- Conway, Maureen, and Robert P. Giloth. 2014. Connecting dence from a Randomized Control Trial in Africa.” CSAE People to Work: Workforce Intermediaries and Sector Strate- Working Paper WPS/2014–30, Centre for the Study of gies. New York: Aspen Institute. African Economies, Department of Economics, Univer- Corseuil, Carlos Henrique, Miguel Foguel, and Gustavo Gon- sity of Oxford, Oxford, U.K.. zaga. 2014. “Apprenticeship as a Stepping Stone to Better Bassanini, Andrea, Alison L. Booth, Giorgio Brunello, Maria Jobs: Evidence from Brazilian Matched Employer- De Paola, and Edwin Leuven. 2005. “Workplace Training Employee Data.” Texto para Discussão 651, Departamento in Europe.” IZA Discussion Paper 1640, Institute for the de Economia, Centro de Ciências Sociais, Pontifícia Study of Labor, Bonn, Germany. Universidade Católica do Rio de Janeiro, Rio de Janeiro. Biavaschi, Costanza, Werner Eichhorst, Corrado Giulietti, Cumsille, Belén. 2016. “Educación Técnico Vocacional Michael Jan Kendzia, Alexander Muravyev, Janneke Secundaria: Beneficios y Desafíos para los Sistemas Pieters, Núria Rodríguez-Planas, et al. 2012. “Youth Educativos.” Technical Note, Inter-American Dialogue, Unemployment and Vocational Training.” IZA Discus- Washington, DC. sion Paper 6890, Institute for the Study of Labor, Bonn, Darvas, Peter, Marta Farvara, and Tamara Arnold. 2017. Germany. Stepping Up Skills in Urban Ghana: Snapshot of the STEP Blattman, Christopher, and Laura Ralston. 2015. “Generating Skills Measurement Survey. Directions in Development: Employment in Poor and Fragile States: Evidence from Human Development Series. Washington, DC: World Labor Market and Entrepreneurship Programs.” Work- Bank. ing paper, Columbia University, New York. Dearden, Lorraine, Howard Reed, and John Van Reenen. Blundell, Richard, Lorraine Dearden, Costas Meghir, and 2006. “The Impact of Training on Productivity and Barbara Sianesi. 1999. “Human Capital Investment: The Wages: Evidence from British Panel Data.” Oxford Bulletin Returns from Education and Training to the Individual, of Economics and Statistics 68 (4): 397–421. the Firm and the Economy.” Fiscal Studies 20 (1): 1–23. De Grip, Andries, and Jan Sauermann. 2012. “The Effects of Boahin, Peter, and W. H. Adriaan Hofman. 2014. “Perceived Training on Own and Co-worker Productivity: Evidence Effects of Competency-Based Training on the Acquisi- from a Field Experiment.” Economic Journal 122 (560): tion of Professional Skills.” International Journal of Educa- 376–99. tional Development 36 (May): 81–89. Dietrich, Hans, Harald Pfeifer, and Felix Wenzelmann. 2016. Bragg, Debra D., Laura Dresser, and Whitney Smith. 2012. “The More They Spend, the More I Earn? Firms’ Training “Leveraging Workforce Development and Postsecondary Investments and Post-Training Wages of Apprentices.” Education for Low-Skilled, Low-Income Workers: Les- Economics of Education Working Paper 116, Institute for sons from the Shifting Gears Initiative.” New Directions Strategy and Business Economics, University of Zurich, for Community Colleges 157: 53–66. Switzerland. Cabrales, Antonio, Juan José Dolado, and Ricardo Mora. 2014. Dunbar, Muriel. 2013. “Engaging the Private Sector in “Dual Labour Markets and (Lack of) on-the-Job Training: Skills Development: Final.” Health and Education PIAAC Evidence from Spain and Other EU Countries.” Advice and Resource Team, Oxford Policy Management, IZA Discussion Paper 8649, Institute for the Study of Oxford, U.K. Labor, Bonn, Germany. Eichhorst, Werner, Núria Rodríguez-Planas, Ricarda Calero, Carla, Carlos Henrique Corseuil, Veronica Gonzales, Schmidl, and Klaus F. Zimmermann. 2012. “A Roadmap Jochen Kluve, and Yuri Soares. 2014. “Can Arts-Based to Vocational Education and Training Systems around Interventions Enhance Labor Market Outcomes among the World.” IZA Discussion Paper 7110, Institute for the Youth? Evidence from a Randomized Trial in Rio de Study of Labor, Bonn, Germany. Build on foundations by linking skills training to jobs | 161 Fares, Jean, and Olga Susana Puerto. 2009. “Towards Com- Hooley, Tristram, and Vanessa Dodd. 2015. “The Economic prehensive Training.” Social Protection and Labor Dis- Benefits of Career Guidance.” Research paper, Careers cussion Paper 0924, World Bank, Washington, DC. England, Chorley, Lancashire, U.K. Fazio, María Victoria, Raquel Fernández-Coto, and Laura ILO (International Labour Organization). 2010. “Teachers Ripani. 2016. “Apprenticeships for the XXI Century: and Trainers for the Future: Technical and Vocational A Model for Latin America and the Caribbean?” Education and Training in a Changing World.” Report Inter-American Development Bank, Washington, DC. prepared for Global Dialogue Forum on Vocational Edu- Fox, Louise M., and U. Kaul. 2017. “What Works for Youth cation and Training, Geneva, September 29–30. Employment in Low-Income Countries?” USAID, Wash- — —— —. 2012. “Upgrading Informal Apprenticeship: A ington, DC. Resource Guide for Africa.” Geneva, International Gerds, Peter. 2009. “Standards for Occupation-Directed Pro- Labour Office. fessional Development of TVET Personnel in Developing IYF (International Youth Foundation). 2013. “A Summative Countries.” In International Handbook of Education for the Report.” Vol. 1, “Testing What Works in Youth Employ- Changing World of Work: Bridging Academic and Vocational ment: Evaluating Kenya’s Ninaweza Program.” IYF, Learning, edited by Rupert Maclean and David N. Wilson, Baltimore. Vol. 3, 1407–22. Bonn, Germany: UNESCO-UNEVOC Jakubowski, Maciej, Harry Anthony Patrinos, Emilio Ernesto International Center for Education; Dordrecht, the Neth- Porta, and Jerzy Wiśniewski. 2016. “The Effects of Delay- erlands: Springer Science+Business. ing Tracking in Secondary School: Evidence from the González-Velosa, Carolina, David Rosas, and Roberto Flores. 1999 Education Reform in Poland.” Education Economics 2016. “On-the-Job Training in Latin America and the 24 (6): 557–72. Caribbean: Recent Evidence.” In Firm Innovation and Pro- King, Christopher T. 2014. “Sectoral Workforce and Related ductivity in Latin America and the Caribbean: The Engine of Strategies: What We Know and What We Need to Know.” Economic Development, edited by Matteo Grazzi and Carlo In Connecting People to Work: Workforce Intermediaries and Pietrobelli, 137–66. Washington, DC: Inter-American Sector Strategies, edited by Maureen Conway and Robert P. Development Bank; New York: Springer Nature. Giloth, 209–38. New York: Aspen Institute. Grollmann, Philipp. 2008. “The Quality of Vocational Teach- Kingombe, Christian. 2012. “Lessons for Developing Coun- ers: Teacher Education, Institutional Roles, and Profes- tries from Experience with Technical and Vocational sional Reality.” European Educational Research Journal 7 (4): Education and Training.” In Economic Challenges and Policy 535–47. Issues in Early Twenty-First-Century Sierra Leone, edited by Guthrie, Hugh, Roger Harris, Michele Simons, and Tom Omotun E. G. Johnson, 278–365. London: International Karmel. 2009. “Teaching for Technical and Vocational Growth Centre. Education and Training (TVET).” In International Hand- Kluve, Jochen, Olga Susana Puerto, David A. Robalino, Jose book of Research on Teachers and Teaching, Part 1, edited Manuel Romero, Friederike Rother, Jonathan Stöterau, by Lawrence J. Saha and Anthony Gary Dworkin, Felix Weidenkaff, et al. 2016. “Do Youth Employment 851–63. Springer International Handbooks of Education Programs Improve Labor Market Outcomes? A System- Series 21. New York: Springer Science+Business. atic Review.” IZA Discussion Paper 10263, Institute for Haelermans, Carla, and Lex Borghans. 2012. “Wage Effects of the Study of Labor, Bonn, Germany. On-the-Job Training: A Meta-Analysis.” British Journal of Konings, Jozef, and Stijn Vanormelingen. 2015. “The Impact Industrial Relations 50 (3): 502–28. of Training on Productivity and Wages: Firm-Level Evi- Hampf, Franziska, and Ludger Woessmann. 2016. “Vocational dence.” Review of Economics and Statistics 97 (2): 485–97. vs. General Education and Employment over the Life- LaLonde, Robert J. 2003. “Employment and Training Pro- Cycle: New Evidence from PIAAC.” IZA Discussion Paper grams.” In Means-Tested Transfer Programs in the United 10298, Institute for the Study of Labor, Bonn, Germany. States, edited by Robert A. Moffitt, 517–86. Chicago: Hanushek, Eric A., Guido Schwerdt, Ludger Woessmann, University of Chicago Press. and Lei Zhang. 2017. “General Education, Vocational Edu- Lerman, Robert I. 2013. “Skill Development in Middle Level cation, and Labor-Market Outcomes over the Life-Cycle.” Occupations: The Role of Apprenticeship Training.” IZA Journal of Human Resources 52 (1): 48–87. Policy Paper 61, Institute for the Study of Labor, Bonn, Hodge, Steven. 2016. “After Competency-Based Training: Germany. Deepening Critique, Imagining Alternatives.” Interna- — —— —. 2014. “Do Firms Benefit from Apprenticeship Invest- tional Journal of Training Research 14 (3): 171–79. ments? Why Spending on Occupational Skills Can Yield Hollenbeck, Kevin. 2008. “State Use of Workforce System Economic Returns to Employers.” IZA World of Labor, Net Impact Estimates and Rates of Return.” Paper pre- Institute for the Study of Labor, Bonn, Germany. sented at Association for Public Policy Analysis and Maclean, Rupert, and Ada Lai. 2011. “Editorial: The Future Management Conference, “The Next Decade: What Are of Technical and Vocational Education and Training: the Big Policy Challenges?” Los Angeles, November 6–8. Global Challenges and Possibilities.” International Journal Hooley, Tristram. 2014. “The Evidence Base on Lifelong of Training Research 9 (1–2): 2–15. Guidance: A Guide to Key Findings for Effective Policy Maguire, Sheila, Joshua Freely, Carol Clymer, Maureen Con- and Practice.” ELGPN Tools 3, European Lifelong Guid- way, and Deena Schwartz. 2010. “Tuning In to Local ance Policy Network, Finnish Institute for Educational Labor Markets: Findings from the Sectoral Employment Research, University of Jyväskylä, Jyväskylä, Finland. Impact Study.” Public/Private Ventures, Philadelphia. 162 | World Development Report 2018 Martinson, Karin. 2010. “Partnering with Employers to Pro- Smith, Erica, and Ros Brennan Kemmis. 2013. “Towards a mote Job Advancement for Low-Skill Individuals.” Model Apprenticeship Framework: A Comparative National Institute for Literacy, Washington, DC. Analysis of National Apprenticeship Systems.” World McKenzie, David J. 2017. “How Effective Are Active Labor Bank and International Labour Office, New Delhi. Market Policies in Developing Countries? A Critical Sousounis, Panos, and Robin Bladen-Hovell. 2010. “Per- Review of Recent Evidence.” Policy Research Working sistence in the Determination of Work-Related Training Paper 8011, World Bank, Washington, DC. Participation: Evidence from the BHPS, 1991–1997.” Mieschbuehler, Ruth, and Tristram Hooley. 2016. “World- Economics of Education Review 29 (6): 1005–15. Class Apprenticeship Standards: Report and Recommen- Soysouvanh, Boualinh. 2013. “Development of Standards for dations.” International Centre for Guidance Studies, Vocational Teachers at Bachelor Level in Lao PDR.” College of Education, University of Derby, Derby, U.K. Research and Development Series 2, Regional Coopera- Mourshed, Mona, Diana Farrell, and Dominic Barton. 2013. tion Platform for Vocational Teacher Education in Asia, “Education to Employment: Designing a System That Shanghai. Works.” McKinsey Center for Government, Copenhagen. Stanley, Julian, Edem Adubra, and Borhene Chakroun. 2014. Neumark, David, and Donna Rothstein. 2006. “School-to- Technical and Vocational Teachers and Trainers in the Arab Career Programs and Transitions to Employment and Region: A Review of Policies and Practices on Continuous Pro- Higher Education.” Economics of Education Review 25 (4): fessional Development. Paris: United Nations Educational, 374–93. Scientific, and Cultural Organization. Newhouse, David, and Daniel Suryadarma. 2011. “The Value Sultana, Ronald G., and Anthony Gordon Watts. 2008. of Vocational Education: High School Type and Labor “Career Guidance in the Middle East and North Africa.” Market Outcomes in Indonesia.” World Bank Economic International Journal for Educational and Vocational Guidance Review 25 (2): 296–322. 8 (1): 19–34. NGA (National Governors Association). 2013. “State Sector Tan, Hong, and Gladys López-Acevedo. 2003. “Mexico: Strategies Coming of Age: Implications for State Work- In-Firm Training for the Knowledge Economy.” Policy force Policymakers.” NGA Center for Best Practices, Research Working Paper 2957, World Bank, Washing- NGA, Washington, DC. https://www.nga.org/files/live ton, DC. /sites/NGA/files/pdf/2013/1301NGASSSReport.pdf. Tan, Jee-Peng, Kiong Hock Lee, Ryan Flynn, Viviana V. Nübler, Irmgard, Christine Hofmann, and Clemens Greiner. Roseth, and Yoo-Jeung Joy Nam. 2016. Workforce Develop- 2009. “Understanding Informal Apprenticeship: Find- ment in Emerging Economies: Comparative Perspectives on ings from Empirical Research in Tanzania.” Employment Institutions, Praxis, and Policies. Directions in Develop- Working Paper 32, International Labour Organization, ment: Human Development Series. Washington, DC: Geneva. World Bank. OECD (Organisation for Economic Co-operation and Devel- Tan, Jee-Peng, and Yoo-Jeung Joy Nam. 2012. “Pre- opment). 2010. Learning for Jobs: Synthesis Report. OECD Employment Technical and Vocational Education and Reviews of Vocational Education and Training Series. Training: Fostering Relevance, Effectiveness, and Effi- Paris: OECD. ciency.” In The Right Skills for the Job? Rethinking Training —— — —. 2014. Skills beyond School: Synthesis Report. OECD Policies for Workers, edited by Rita Kullberg Almeida, Jere Reviews of Vocational Education and Training Series. R. Behrman, and David Robalino, 67–103. Report 70908, Paris: OECD. Human Development Perspectives Series. Washington, OECD (Organisation for Economic Co-operation and Devel- DC: World Bank. opment) and EC (European Commission). 2004. “Career UIS (UNESCO Institute for Statistics). 2016. “Education.” Guidance: A Handbook for Policy Makers.” OECD, Paris. Montreal. http://uis.unesco.org. Roder, Anne, and Mark Elliott. 2011. “A Promising Start: Year UNESCO (United Nations Educational, Scientific, and Cul- Up’s Initial Impacts on Low-Income Young Adults’ tural Organization). 2014. “Technical and Vocational Careers.” Economic Mobility Corporation, New York. Teachers and Trainers in the Arab Region: A Review of Roseth, Viviana V., Alexandria Valerio, and Marcela Gutiérrez. Policies and Practices on Continuous Professional Devel- 2016. Education, Skills, and Labor Market Outcomes: Results opment.” UNESCO, Paris. from Large-Scale Adult Skills Surveys in Urban Areas in 12 Watts, Anthony Gordon, and Ronald G. Sultana. 2004. Countries. STEP Skills Measurement Series. Washington, “Career Guidance Policies in 37 Countries: Contrasts and DC: World Bank. Common Themes.” International Journal for Educational Rosholm, Michael, Helena Skyt Nielsen, and Andrew and Vocational Guidance 4 (2–3): 105–22. Dabalen. 2007. “Evaluation of Training in African Enter- Watts, Anthony Gordon, Ronald G. Sultana, and John prises.” Journal of Development Economics 84 (1): 310–29. McCarthy. 2010. “The Involvement of the European Safford, Kimberly, Deborah Cooper, Freda Wolfenden, and Union in Career Guidance Policy: A Brief History.” Inter- Joyce Chitsulo. 2013. “  ‘Give Courage to the Ladies’: national Journal for Educational and Vocational Guidance Expansive Apprenticeship for Women in Rural Malawi.” 10 (2): 89–107. Journal of Vocational Education and Training 65 (2): 193–207. World Bank. 2016. “Projet Emploi Jeune et Développement Saraf, Priyam. 2017. “Returns, Barriers, and Policy Outcomes des Compétences (PEJEDEC); Evaluation d’impact du to On-the-Job Training: Creating Gains for Workers, programme de Travaux à Haute Intensité de Main Firms, and Society.” Background paper, World Bank, d’œuvre (THIMO): Résultats intermédiaires à court Washington, DC. terme.” World Bank, Washington, DC. Build on foundations by linking skills training to jobs | 163 SPOTLIGHT 5 Technology is changing the world of work: What does that mean for learning? The nature of work is changing. Within countries, digital imaging, for example. And, of course, technol- jobs have been shifting across sectors—sometimes on ogy creates new jobs as well. So technology eliminates a massive scale. Some shifts have been out of agricul- some jobs, creates others, and increases the returns to ture. In what are now high-income countries, people yet others.5 have shifted out of agriculture dramatically over the The impact of technology on jobs varies dramati- last half-century. In the Republic of Korea, the share cally across countries. For rich countries, predictions of workers in farm jobs fell from 80 percent in 1950 to range from dire (in which “robot overlords” take over less than 7 percent in 2009. In Chile, the share of farm most jobs) to the much more modest estimate that workers fell from 30 percent to under 15 percent in the 9 percent of jobs in rich countries could currently be same period. Other shifts have been out of industrial automated.6 For low-income countries, where tech- production. In the United States, the share of workers nological penetration is much lower, the impact of in manufacturing halved between 1950 and 2009.1 technology on work will likely be more incremental In low- and middle-income countries, the shift is (figure S5.1). Small-scale agriculture and household ongoing. Across Sub-Saharan Africa, employment in enterprises will not be automated in the near future, agriculture is expected to drop nearly 10 percent this especially in countries such as Nicaragua, where less decade, with a large rise in the numbers of people run- than 20 percent of households have access to the ning small household businesses.2 internet, or Liberia, where less than 10 percent of the Technology—including digital technology—is cen- population has access to electricity.7 Those numbers tral to these changes. Eighty-five percent of the will surely grow; greater access to technology will population worldwide now has access to electricity. enable more poor nations and individuals to access Digital technologies penetrate most corners of the those sectors that see high returns to technological world, with one mobile phone subscription per person growth. But in the short run, technology will change globally, and 4 in 10 persons connected to the inter- the demand for skills much more in countries that net.3 As the World Development Report 2016 points out, have the infrastructure to support automation. “With rising computing power, combined with the Individuals who enter the workforce with better connectivity and informational value of the internet, technological skills will see benefits. Because tech- digital technologies are taking on more tasks.”4 This nology affects different workers in different ways, is particularly true for routine tasks that are easy to those who emerge from the education system with automate such as a cashier’s job. But other jobs—such technological skills are more likely to be able to as a teacher’s—are not easy to automate. Technology enter those professions (high-skilled, high-paying) ultimately substitutes for some workers. For workers that are gaining from technology.8 Around the world, whose jobs are not replaced, such as hairdressers or the rise of information technology is increasing the surgeons, technology has varied effects. While it may demand for high-skilled graduates who can use that leave the hairdresser relatively untouched, it can make technology effectively.9 That rising demand translates the surgeon dramatically more productive—with into higher wages.10 Because this dynamic can widen 164 | World Development Report 2018 Figure S5.1 Technology use has increased dramatically over the past decade—but remains low in many countries Percentage of population who have access to electricity (2005–14) and who use the internet (2005–15), by country income group a. Access to electricity b. Internet users 100 100 80 80 60 60 Percent Percent 40 40 20 20 0 0 2005 2008 2011 2014 2005 2008 2011 2014 High-income countries Middle-income countries Low-income countries World Source: WDR 2018 team, using data from World Bank (2017a) for electricity access and ITU (2016a) for internet users. Data at http://bit.do/WDR2018-Fig_S5-1. inequality, ensuring that much of the population has Chile, more than 90 percent used computers at school access to these skills is essential. Globally, 85 percent and two-thirds had access to computers at home. of countries include computer skills in their curricu- Although two-thirds of them were able to search for lums for upper secondary school. But some regions information online, only half could organize informa- lag, with Sub-Saharan Africa at only 50 percent, and tion (such as arranging folders on the computer). Less much lower at lower levels of schooling.11 Beyond than one-third could produce information (such as the benefits to individuals, a population with strong writing an email with adequate content).15 Individuals technical skills is more likely to attract international need structured training in computing skills if they industries that require those skills, such as modern are to reap the returns of the technological revolution. manufacturing. Students entering the workforce need better crit- In environments with extremely limited access to ical thinking and socioemotional skills. The ability to computing technology, simple exposure can make a use technology is one way for them to take advantage difference, but the skills that students gain are not the of technological advancement. But another is to excel skills they need. Replacing traditional textbooks with at those skills that technology carries out less well. laptops equipped with electronic textbooks neither Those include higher-order cognitive skills and inter- helped nor harmed reading ability in Honduras, but in personal, socioemotional skills.16 In the United States, an environment where only 7 percent of students nor- jobs that require high socioemotional skills (such as mally use the internet at school, the laptops allowed nurses or social workers) are growing and jobs that many more of them to develop the ability to search require high socioemotional skills along with high for content online and do basic word processing.12 cognitive ability (such as financial managers) are In Romania, vouchers to purchase standard home growing the fastest.17 Education systems are begin- computers improved very basic general computing ning to learn how to cultivate socioemotional skills skills.13 But such skills that are gained from mere in learners: Recent efforts in Peru and Turkey have exposure may not be the skills needed to succeed resulted not only in better socioemotional ability in the marketplace. Distributing simple laptops for but also in better academic performance.18 It is not home use in Peru made learners more competent on enough to train learners to use computers: to navi- those laptops, but that did not translate to better skills gate a rapidly changing world, they have to interact on other, general-use operating systems (such as Mic- effectively with others, think creatively, and solve rosoft Windows).14 Among high school students in problems. Technology is changing the world of work: What does that mean for learning? | 165 All of those skills that help individuals succeed in digital content only if they have strong literacy skills. rapidly changing economies are built on the same They can program new online applications only if foundations of literacy and numeracy. It may be they have confident numeracy skills. Socioemotional tempting to divert resources from the development skills like grit, which are most malleable in childhood, of foundational skills into the technological skills, can be practiced and strengthened in the service of higher-order cognitive skills, and socioemotional gaining strong foundational skills. Higher-order cog- skills needed in the 21st century, which seem more nitive skills involve consuming information using novel and exciting. But these are complements to literacy and numeracy skills and combining it in new foundational skills, not substitutes for them—they ways. Innovations in developing 21st-century skills can only be built on a solid foundation. Workers can are much needed, but these skills work best in con- search effectively for digital information or create junction with strong foundational abilities. Notes 1. Handel (2012). Beuermann, Diether W., Julian Cristia, Santiago Cueto, Ofer 2. Fox and others (2013). Malamud, and Yyannu Cruz-Aguayo. 2015. “One Laptop 3. ITU (2016b); World Bank (2017a). Per Child at Home: Short-Term Impacts from a Random- 4. World Bank (2016, 120). ized Experiment in Peru.” American Economic Journal: 5. World Bank (2016). Applied Economics 7 (2): 53–80. 6. Arntz, Gregory, and Zierahn (2016); Drum (2013). Claro, Magdalena, David D. Preiss, Ernesto San Martín, 7. ITU (2016b); World Bank (2017a). Ignacio Jara, J. Enrique Hinostroza, Susana Valenzuela, 8. World Bank (2016). Flavio Cortes, et al. 2012. “Assessment of 21st Century 9. Autor, Katz, and Krueger (1998); Michaels, Natraj, and ICT Skills in Chile: Test Design and Results from High Van Reenen (2014). School Level Students.” Computers and Education 59 (3): 10. Falck, Heimisch, and Wiederhold (2016). 1042–53. 11. UIS (2017). Deming, David J. Forthcoming. “The Growing Importance of 12. Bando and others (2017). Social Skills in the Labor Market.” Quarterly Journal of 13. Malamud and Pop-Eleches (2011). Economics. 14. Beuermann and others (2015). Drum, Kevin. 2013. “Welcome, Robot Overloads; Please Don’t 15. Claro and others (2012). Fire Us? Smart Machines Probably Won’t Kill Us All, but 16. World Bank (2016). They’ll Definitely Take Our Jobs and Sooner Than You 17. Deming (forthcoming). Think.” Mother Jones (May/June). http://www.motherjones 18. Alan, Boneva, and Ertac (2015); World Bank (2017b). .com/media/2013/05/robots-artificial-intelligence-jobs -automation. Falck, Oliver, Alexandra Heimisch, and Simon Wiederhold. References 2016. “Returns to ICT Skills.” OECD Education Working Paper 134, Organisation for Economic Co-operation and Alan, Sule, Teodora Boneva, and Seda Ertac. 2015. “Ever Development, Paris. Failed, Try Again, Succeed Better: Results from a Fox, Louise M., Cleary Haines, Jorge Huerta Muñoz, and Randomized Educational Intervention on Grit.” HCEO Alun H. Thomas. 2013. “Africa’s Got Work to Do: Employ- Working Paper 2015–009, Human Capital and Economic ment Prospects in the New Century.” IMF Working Opportunity Global Working Group, Economics Paper WP/13/201, International Monetary Fund, Wash- Research Center, University of Chicago. ington, DC. Arntz, Melanie, Terry Gregory, and Ulrich Zierahn. 2016. Handel, Michael J. 2012. “Trends in Job Skill Demands in “The Risk of Automation for Jobs in OECD Countries: A OECD Countries.” OECD Social, Employment, and Comparative Analysis.” OECD Social, Employment, and Migration Working Paper 143, Organisation for Eco- Migration Working Papers 189, Organisation for Eco- nomic Co-operation and Development, Paris. nomic Co-operation and Development, Paris. ITU (International Telecommunication Union). 2016a. Autor, David H., Lawrence F. Katz, and Alan B. Krueger. 1998. World Telecommunication/ICT Indicators Database. “Computing Inequality: Have Computers Changed the ITU, Geneva. http://www.itu.int/en/ITU-D/Statistics Labor Market?” Quarterly Journal of Economics 113 (4): /Pages/publications/wtid.aspx. 1169–1213. —— — —. 2016b. Measuring the Information Society Report 2016. Bando, Rosangela, Francisco Gallego, Paul Gertler, and Dario Geneva: ITU. http://www.itu.int/en/ITU-D/Statistics Romero Fonseca. 2017. “Books or Laptops? The Effect of /Documents/publications/misr2016/MISR2016-w4.pdf. Shifting from Printed to Digital Delivery of Educational Malamud, Ofer, and Cristian Pop-Eleches. 2011. “Home Content on Learning.” Economics of Education Review. In Computer Use and the Development of Human Capital.” press. Quarterly Journal of Economics 126 (2): 987–1027. 166 | World Development Report 2018 Michaels, Guy, Ashwini Natraj, and John Van Reenen. 2014. ——— —. 2017a. World Development Indicators (database). “Has ICT Polarized Skill Demand? Evidence from Eleven World Bank, Washington, DC. http://data.worldbank Countries over Twenty-Five Years.” Review of Economics .org/data-catalog/world-development-indicators. and Statistics 96 (1): 60–77. ——— —. 2017b. “Peru: If You Think You Can Get Smarter, You UIS (UNESCO Institute for Statistics). 2017. “Curriculum and Will.” What We Do: Projects and Operations (April 25), World ICT in Education.” UIS, Montreal. http://data.uis.unesco Bank, Washington, DC. http://www.worldbank.org/en .org. /results/2017/04/25/peru-if-you-think-you-can-get World Bank. 2016. World Development Report 2016: Digital -smarter-you-will?CID=POV_TT_Poverty_EN_EXT. Dividends. Washington, DC: World Bank. Technology is changing the world of work: What does that mean for learning? | 167 PART IV Making the system work for learning at scale 9 Education systems are misaligned with learning 10 Unhealthy politics drives misalignments 11 How to escape low-learning traps 9 Education systems are misaligned with learning Education systems are often poorly aligned with learning goals. These misalignments are driven in part by technical complexities: education systems simultaneously pursue many (often conflicting) goals, with the many system actors continually interacting in complex ways. Compounding these technical challenges is the limited policy implementation capacity of the many government agencies responsible for learning. Kenya’s government discovered just how difficult it teachers on a contract basis violated constitutional is to turn successful small-scale interventions into rights to equal pay for equal work. The ruling led to systemwide improvements in learning. In the late guarantees from the government to gradually absorb 2000s, even though access to primary schooling was all contract teachers into the civil service and provide high, many children failed to acquire even basic skills. them with the same employment protections. These The government argued that large classes, with their developments significantly changed the employ- overburdened teachers, lay behind these disappoint- ment prospects of contract teachers. In particular, ing results. But a constrained education budget meant they weakened the link between performance and that hiring more civil service teachers to address these the chances of contract renewal—the main channel problems was not an option. Instead, in 2009 the gov- through which the original trial had improved stu- ernment hired 18,000 temporary contract teachers. dent learning. At the same time, the ministry also The new program shared many of the same features struggled to implement the program. Government- of an earlier pilot experiment by a nongovernmen- employed contract teachers were paid on average tal organization that provided government schools three months late, hurting student learning. with contract teachers.1 The pilot reduced class sizes, This example illustrates a more common finding leading to improved learning outcomes for students that working at scale is not the same as “scaling up.”3 taught by the new contract teachers. Moreover, these Similar difficulties in changing teacher employment gains were achieved at a cost well below the cost of conditions in government schools have occurred the alternative of hiring more civil service teachers. in many other countries, despite evidence from But unlike the pilot intervention, the government pilot programs showing their potential to improve program failed to deliver any improvements in learn- learning.4 These examples show that implementing ing.2 A combination of union resistance and lack of interventions at scale can also induce responses from Ministry of Education capacity to manage contract other actors or parts of an education system that can teachers underpinned the program’s lack of impact. alter the potential impacts on learning. The Kenyan teachers’ union successfully challenged In many countries, education systems suffer from the program in the courts, arguing that hiring two related weaknesses. First, systems are not well 170 | World Development Report 2018 aligned with the overall goal of learning; other goals Figure 9.1 Technical and political barriers pull can detract from, and in some cases compete with, education systems away from the goal of learning efforts to improve learning outcomes. Second, the elements of an education system are often incompat- ible or incoherent. For example, government funding allocations sometimes fail to provide the resources schools need to improve learning. Even when school r s Le he ar funding is available, the rules governing its use often c ne leave little flexibility for schools to use it in ways tai- a Te rs lored to the specific needs of students. Technical and political factors underlie these system weaknesses. Getting all parts of an education system to work together is difficult, and the agencies responsible for designing, implementing, and eval- uating education policies often lack the capacity to LEARNING take on this role. For example, timely information on student learning outcomes is not available in many low-income countries, making it harder to design Sch appropriate interventions and to monitor their oo ts pu effectiveness. The interests of system actors can also lm na in contribute to misalignments. For example, calls to ol a ge o devolve control over resources to schools are some- me S ch times resisted because private textbook providers fear nt losing out on lucrative centralized contracts.5 Failure to tackle these technical and political constraints can trap countries in a low-learning, low- Source: WDR 2018 team. accountability, high-inequality equilibrium. When different parts of a system fail to work together, edu- cation outcomes will fall far short of what is possible. • Information and metrics. Accurate, credible infor- When actors in the system interact to pursue many mation on learning is often unavailable. This goals, the mechanisms that hold them accountable for can divert attention from learning and hinder learning are weakened. And where powerful groups monitoring and evaluation of interventions can divert resources to align with their own inter- aimed at improving outcomes. ests, education systems can exacerbate inequalities. • Finance. Education funding is sometimes inad- Together, these factors can pull an education system equate and often allocated in ways inconsistent out of alignment with the overall goal of learning with a goal of providing equitable opportunities (figure 9.1). for effective learning. • Incentives. The motivation and incentives of Misalignments and system actors are often only weakly linked to student learning. incoherence impede learning Taking a systems approach can help to identify the Learning objectives and responsibilities elements that are incoherent with each other or mis- Though most education systems recognize learning aligned with learning (box 9.1). Though every educa- as a central goal, it often receives less prominence tion system faces its own challenges, incoherence and than other objectives. Looking beyond high-level misalignments tend to occur across four elements: policy documents often reveals the objectives that matter most in the day-to-day affairs of education Learning objectives and responsibilities. Clearly •  agencies. Bangladesh has made progress in linking articulated learning goals are often missing. But education sector objectives explicitly to government even when they exist, the roles and responsibili- budgets—for example, budget documents link allo- ties of different system actors in achieving them cations to specific activities aimed at improving are unclear, resulting in limited accountability. education outcomes. However, the government’s key Education systems are misaligned with learning | 171 Box 9.1 It’s all about (education) systems What’s an education system? teachers despite lower levels of education, training, and An education system is a collection of “institutions, actions pay.d This finding suggests that some poor performance is and processes that affect the ‘educational status’ of citizens driven not so much by a teacher’s individual capacity but in the short and long run.”a Education systems are made by the organizational setting—incentives, accountability up of a large number of actors (teachers, parents, politi- mechanisms, power relations—in which government teach- cians, bureaucrats, civil society organizations) interacting ers operate. A systems approach aims to identify these with each other in different institutions (schools, ministry underlying factors so that policy design can tackle the departments) for different reasons (developing curricu- deeper causes of poor performance. lums, monitoring school performance, managing teachers). A systems approach can also highlight where system All these interactions are governed by rules, beliefs, and elements are incoherent. For example, curriculum improve- behavioral norms that affect how actors react and adapt to ments may lead to few improvements in student learning if changes in the system.b other parts of the system (such as assessment or teacher development) fail to adapt. A systems view can reveal how Why is it useful to take a systems approach? changes in one part of the system affect other subsys- A systems approach takes into account the interactions tems and support better alignment and ultimately better between the parts of an education system. In doing so, it outcomes.e seeks to understand how they work together to drive sys- A systems approach is also better suited to working with tem outcomes, instead of focusing on specific elements in the complexity of education systems. The many objectives isolation.c It can help assess whether different actors and that education systems tend to pursue at the same time, subsystems align with education goals and shed light on coupled with the many different actors involved in pursuing the underlying drivers of system performance. For exam- these objectives, make it difficult to predict how different ple, limited teacher capacity is often highlighted as a major interventions will affect learning. A systems approach shifts cause of poor performance. But trials introducing contract the focus away from interventions designed to address teachers into schools have shown that they can deliver specific problems, toward the broader changes required to the same or better learning outcomes than government improve learning sustainably. Source: WDR 2018 team. a. Moore (2015, 1). b. World Bank (2003). c. Bowman and others (2015). d. Bruns, Filmer, and Patrinos (2011). e. Newman, King, and Abdul-Hamid (2016). performance indicators mostly deal with access and that schools received the right books.7 Delivering early completion; only 1 of the 12 indicators targets learn- childhood development services typically requires ing. Moreover, that indicator tracks literacy rates coordination among several government agencies, in the population over age 15, which is insensitive including health and education ministries. Managing to changes in school performance over the medium these many agencies is challenging. In the early 1990s, term.6 constant shifts in responsibility for early childhood Even where learning is a clear goal, the way edu- development in Ghana resulted in inadequate stew- cation systems are organized sometimes hampers ardship of these services.8 performance. Because tasks are often fragmented across education departments and government agen- Information and metrics cies, it can be hard to identify who is accountable for Systems often lack the information needed to sup- outcomes. In Romania, responsibilities for textbook port the design and implementation of reform. provision were split among four different agencies, Education management information systems cover yet none of them was solely responsible for ensuring a wide range of indicators on service delivery, but in 172 | World Development Report 2018 many countries they do not routinely include data (figure 9.2). Public expenditure reviews and other on learning. India’s District Information System for studies reveal similar patterns across subnational Education (DISE) is designed to provide report cards administrations and even across schools (spotlight 6). for districts, but of the 980 data points reported, none The weak link between spending and learning is a covers student learning.9 That omission can make it feature of the different environments in which educa- difficult for systems to track interventions to improve tion systems operate. Systems with higher corruption learning, for parents to demand better services from or lower bureaucratic quality are less likely to use politicians or directly from schools, and for agencies resources effectively to raise learning.11 to design effective policies to improve learning. These simple correlations also suggest that many education systems are delivering learning outcomes Finance well below what is possible given current levels of Public spending does not correlate strongly with funding. In India, excess teacher absenteeism in the learning. The link between spending and learning public sector is estimated to cost US$1.5 billion a year. differs enormously, even among countries at similar If teacher accountability systems were more strongly levels of economic development. In 2015 Peru spent aligned with learning, teacher attendance would 28 percent less per student than the Dominican improve, allowing the system to achieve higher levels Republic, but it had Programme for International of learning at the same cost.12 Student Assessment (PISA) mathematics scores that Improvements in learning are unlikely when addi- were more than half a standard deviation higher.10 tional resources are allocated like past funding. The More generally, cross-country correlations between composition of education spending in many coun- public spending and learning levels are weak and tries is suboptimal. Funding for teacher salaries often statistically insignificant after controlling for income absorbs more than 80 percent of education budgets in per capita. Moreover, for any given level of spending low-income countries, leaving little room for spend- there is a wide range of outcomes. Even changes in ing in other areas. Using additional funding to shift public education spending over time sometimes spending patterns to ensure that teachers have the result in unexpected outcomes. For example, Bulgar- complementary inputs needed—such as textbooks ia’s PISA mathematics scores increased between 2009 and in-service training—would improve alignment and 2015, despite reductions in spending per student and significantly aid learning.13 Figure 9.2 Simple associations between education spending and learning are weak a. Spending and learning outcomes b. Changes in spending and learning 100 4 conditional on GNI per capita (log, PPP) PISA mathematics score, 2009–15 PER PISA 2015 mathematics score, Average annual % change in 50 2 IDN BGR POL COL LTU 0 0 GBR CZE BRA HUN −50 −2 KOR AUS SVK y = 2.37 + 0.001x y = –0.02 + 0.05x t = 0.44 t = 1.35 −100 R 2 = 0.004 −4 R 2 = 0.05 −4,000 −2,000 0 2,000 4,000 6,000 −10 −5 0 5 10 Annual government spending on primary and Average annual % change in secondary public education, per student (PPP), spending per secondary student, 2009–15 conditional on GNI per capita (log, PPP) Sources: WDR 2018 team, using data from OECD (2016); UIS (2017); World Bank (2017a). Data at http://bit.do/WDR2018-Fig_9-2. Note: AUS = Australia; BGR = Bulgaria; BRA = Brazil; COL = Colombia; CZE = Czech Republic; GBR = United Kingdom; HUN = Hungary; IDN = Indonesia; KOR = Republic of Korea; LTU = Lithuania; PER = Peru; POL = Poland; SVK = Slovak Republic. GNI = gross national income; PISA = Programme for International Student Assessment; PPP = purchasing power parity U.S. dollars. Education systems are misaligned with learning | 173 Incentives evaluate teacher performance, those mechanisms are Education system actors face many incentives, but often disconnected from decisions on professional only some of these incentives are aligned with learn- development. Edo State in Nigeria conducts annual ing. System actors are motivated by a range of factors performance evaluations, but these evaluations do that affect how they carry out their duties.14 Profes- not affect decisions on teacher promotions, nor do sional rewards—the social status afforded to their they lead to sanctions or rewards for teachers based occupation, the ability to develop new competencies, on their performance.16 intrinsic motivation—are all important factors driv- ing behavior. Financial rewards and accountability Coherence matters: Getting all parts of the mechanisms, such as feedback from parents or from system working together managers, can also affect how system actors perform. Ensuring that the parts of an education system work Though some of these factors that motivate system together is as important as ensuring alignment actors are aligned with learning, some are not. For toward learning. Even if a country has prioritized stu- example, salaries and career progression are often dent learning, established reasonable learning met- determined largely by a combination of qualifica- rics, and aligned funding with incentives, it still needs tions and experience, despite these characteristics to ensure that system elements are coherent (box 9.2). having only a weak relationship with learning.15 Even If a country adopts a new curriculum that places where countries have invested in mechanisms to greater emphasis on active learning and creative Box 9.2 Aligning all the ingredients for effective teaching in Shanghai When 15-year-old students in Shanghai, China, outscored scale allows high-performing and long-serving teachers their peers in every other education system in the 2012 to earn significantly more than new teachers. Adequate Programme for International Student Assessment (PISA), financing keeps teaching workloads relatively low, they sparked global interest in figuring out how Shanghai giving teachers the time to develop and prepare lesson did it. One lesson is that coherence among key system plans. elements, all aligned toward learning, has made Shanghai’s 4. Incentives. Because of this attractive compensation teacher workforce particularly effective: package and the high societal respect for teachers, Shanghai can attract skilled, able candidates to teaching. 1.  Learning objectives and responsibilities. Learning stan- Incentives—both monetary and nonmonetary—encou- dards lay out clearly the competencies that students are rage teachers to maintain high standards and continue expected to master in each grade. Teachers are expected improving their teaching skills. For example, high- to translate these standards into detailed lesson plans, performing teachers are recognized through the title so that students can learn the curriculum effectively. of “model teacher,” and a (small) share of a teacher’s 2. Information and metrics. Based on the learning stan- overall pay is based on performance. Teachers also dards, schools routinely assess student progress. The have opportunities to act on these incentives, thanks results of these assessments are fed directly into the to a well-established professional development system classroom, where teachers use them to adjust lesson aligned with their needs. For example, school leaders plans and schedule additional time for areas in which draw on their close monitoring of teachers to develop students are weak. Student assessments are also an targeted training plans for individual teachers. important input to a comprehensive system to monitor, evaluate, and support teachers. No two education systems are alike, and attempting to 3. Finance. The salary and benefits package for teachers exactly replicate Shanghai’s system of teacher manage- in Shanghai is generous compared with those in other ment in other countries is unlikely to work. Still, the core parts of China. In fact, it is comparable with those of principle likely applies anywhere: aligning the various parts other professional occupations. Moreover, the salary of the system coherently toward learning pays off. Source: WDR 2018 team, based on Liang, Kidwai, and Zhang (2016). 174 | World Development Report 2018 thinking, the curriculum alone will not change much. The national government in Manila manages the sys- Teachers need training so that they use more active tem through a network of more than 200 division and learning methods, and they need to care enough to 2,500 district education offices. These offices oversee make the change—given that teaching to the new over 600,000 public school teachers, or more than 40 curriculum could be much more demanding than old percent of the public sector workforce. Even routine methods that favor rote learning. Even if teachers are tasks involve coordination between many parts of the on board with curriculum reform, students and their system. For example, management of public school families could weaken its effects if an unreformed operational funds relies on student data from the cen- examinations system creates misaligned incentives. tral office. Once schools have their allocations, they In the Republic of Korea, efforts to introduce a more issue about 500,000 checks and generate as many student-centered curriculum—one that encourages spending reports, each detailing individual spending greater creativity—have sometimes conflicted with items. The monitoring of these financial flows alone pressure on students to succeed on the all-important puts a significant strain on the system, even though university entrance examinations.17 they account for less than 5 percent of government The need for coherence between different parts of education spending.20 an education system makes it risky to borrow from Three characteristics of complex education sys- other countries. Education policy makers often scruti- tems magnify the technical challenges of managing nize higher-performing systems to identify what they them. First, systems are opaque. Many of the goals could borrow to improve learning outcomes in their pursued by these actors are hard to observe, as are own systems. Indeed, the search for the secret ingre- many of the interactions among the actors, whether dient behind Finland’s record of learning led in the they take place in the classroom or in the bureaucracy. 2000s to a swarm of visiting delegations in what has Second, systems are “sticky”: reforms to improve been dubbed “PISA tourism.” Finland’s system gives learning are hard to launch, and they take time to its well-educated teachers considerable autonomy, bear fruit. Third, implementing reforms successfully so they are able to tailor their teaching to the needs requires capacity that many bureaucracies lack. of their students. But lower-performing systems that simply import Finland’s teacher autonomy into Many goals and actors make education their own contexts—contrary to the advice of Finn- systems opaque ish educators who emphasize coherence—are likely Education systems typically have a range of goals, to be disappointed. If teachers are poorly prepared, including equipping students with the skills needed unmotivated, and loosely managed, then giving them for the labor market, advancing social equity, and greater autonomy will likely compound the problem. teaching children the norms, beliefs, and histories South Africa discovered this in the 1990s and 2000s, of their community. But education systems can have when it adopted a curriculum approach that set objec- other goals that can hamper efforts to improve learn- tives centrally but left implementation up to teachers. ing. For example, politicians sometimes view educa- The approach failed in many schools, in part because tion systems as a tool for rewarding their supporters it proved to be a poor fit for the capacity of teachers with civil service jobs, or for impressing voters with and the resources they had at their disposal.18 This school construction programs that are visible but not example illustrates why coherence between different strategically planned. These goals can be misaligned system elements and the development of home- with learning, leaving schools with buildings they grown solutions are so important. cannot use and teachers who are not proficient.21 Where these goals compete with other goals, the result is that the overall education system and its Technical complexities make actors are not aligned toward learning. it hard to align education Managing the system to improve learning is dif- ficult. Promoting learning in the classroom involves systems with learning significant discretion for teachers, who must use their Every day 23 million children—a fifth of the popu- professional assessment to tailor their teaching to the lation—attend one of the 47,000 public elementary needs of their students. Teaching also involves regular, and high schools in the Philippines.19 When their repeated interactions between students and teachers parents are included, about two-thirds of Filipinos over a relatively long period. These characteristics— interact with the school system on a regular basis. coupled with a dearth of information and metrics on Education systems are misaligned with learning | 175 student outcomes at the school level—make it hard to easily monitored investments aimed at expanding manage and monitor learning. These challenges may access. By contrast, investments to raise teacher com- be exacerbated if private schools are a major player, petence or improve the curriculum are less visible, and because those schools typically operate outside the monitoring their impact on student learning is more direct control of the public system (box 9.3). difficult. Such challenges can sometimes prompt edu- Some things are easier to monitor.22 School build- cation systems to emphasize improvements in access ing and cash transfer programs are highly visible and over improvements in quality.23 Even when systems Box 9.3 Can private schooling be aligned to learning for all? Private schools play a major role in education, even for the safer than public schools.g Furthermore, although public poor. Globally, roughly one in eight primary school students primary education is formally free in the vast majority of attends a private school. At the secondary level, the num- countries, many informal fees remain, reducing the cost ber rises to one in four among middle-income countries difference between public and private schools. (table B9.3.1).a The numbers are similar for low-income But there is no consistent evidence that private schools countries, where they may be underestimates if informal deliver better learning outcomes than public schools, or the schools are undercounted.b In some places, the share of opposite. In Colombia, India, and the United States, exper- students attending private schools is much higher than imental evaluations of the consequences of enrolling in a these global figures. In one Nigerian state, 57 percent of private versus a public school show mixed results.h In some all basic education learners attend private schools.c These contexts, private schools may deliver comparable learning enrollments are not limited to high-income households. levels at lower cost than public systems, often by paying In slum communities in Nairobi, Kenya, 43 percent of the lower teacher salaries.i Even so, lower teacher salaries may poorest quintile of families send their children to private reduce the supply of qualified teachers over time. schools. This is higher than the proportion among the rich- Much of the evidence cited in this debate is nonexper- est quintile of families in nonslum communities who send imental, so it may conflate the effects of private schools their children to private schools (35 percent).d In Jamaica, themselves with the effects of the type of students who 10 percent of learners from the poorest economic groups enroll in private schools. Comparisons across 40 countries enroll in private schools.e that seek to adjust for these differences in student char- Low-income households are willing to make this sac- acteristics find no private school advantage in the vast rifice because they perceive that private schools deliver majority of countries.j Moreover, little rigorous research has better education at comparable cost. In many countries, assessed the effects of private schooling on students’ values parents say that teacher absenteeism is lower in private or on the long-term health of the public school system. schools and that learning outcomes are better.f In Jamaica From a public policy perspective, how should gov- and South Africa, parents suggest that private schools are ernments view the growth in private schooling? Should Table B9.3.1 Private providers account for a significant share of school enrollment Percentage of learners enrolled in private education, by country income group (2014) Country income group Preprimary Primary Secondary Low-income 57 14 20 Middle-income 42 13 25 High-income 42 12 20 Source: World Bank (2017a). (Box continues next page) 176 | World Development Report 2018 Box 9.3 Can private schooling be aligned to learning for all? (continued) governments encourage its growth, whether by removing with the learning crisis, some have turned to public-private restrictions on new schools or even by providing public partnerships in which they provide private schools with subsidies that allow more students to enroll in private resources. In Pernambuco, Brazil, the state government schools? Is there a trade-off between the short-term is seeking to place half of the state’s students in govern- growth of private schools and the long-term health of the ment-funded private schools.p In Uganda, the government education system? provided hundreds of private schools with the resources Private schools offer a variety of potential benefits. A needed to meet the growing demand for secondary edu- straightforward one is proximity: new private schools can cation.q In some cases, this means private providers essen- fill a gap when the nearest public schools are far away, tially mirror public schools in terms of education policy, or when there is demand to expand faster than public such as in the government-“aided” schools in India.r But in infrastructure can be built.k As for cost, in China, Ghana, other cases, such as in voucher schools in the United States and Kenya some private schools are comparable in cost to or Liberia’s Partnership Schools pilot, publicly funded pri- the public alternative.l Private schools can also innovate vate schools have significant leeway in how they run their in ways that public schools cannot because they operate schools, letting student learning results be the measure of under fewer constraints. Moreover, private schools can ful- quality.s In Uganda, public resources increased the quality fill niches for families with preferences different from the of private schools, and public-private partnerships are government’s—for example, if parents value single-sex or likely a useful strategy if countries seek to expand enroll- religious education. Private schools may also have lower ments dramatically in a short time.t rates of teacher absence, such as in four countries in Sub- But overseeing private schools may be no easier than Saharan Africa.m In those schools, nonperforming teachers providing quality schooling. The key challenge for policy can be let go more easily than in public schools, increas- makers is to develop a policy and regulatory framework ing their accountability. Finally, competition from private that ensures access for all children, protects families from schools could improve the performance of nearby public exploitation, and establishes an environment that encour- schools.n ages education innovation. Managing a regulatory frame- But these benefits come with many risks. Private work to achieve this is difficult: the same technical and schools may skim off the higher-income students who political barriers that education systems face more gen- are easiest and most profitable to teach, leaving only the erally come into play. From a technical perspective, devel- more disadvantaged students in the public system.o Private oping a framework to accommodate the diverse nature schooling may also deepen social cleavages along dimen- of nonstate provision is complicated. In Bangladesh, for sions other than income if it causes students to be sorted example, there are 11 separate categories for the nonstate by language, ethnicity, or religion. Because families are provision of presecondary education (figure B9.3.1). Unlike not necessarily knowledgeable about pedagogy, private government schools that are relatively homogeneous, schools can induce them to make choices that slow student nonstate provision reflects many different philosophies or learning—for example by discouraging mother tongue approaches to education. The capacity of education agen- instruction. And because families cannot evaluate quality cies to effectively align incentives and monitor services or learning perfectly, private providers may try to take is often limited, and assessing quality in contexts where advantage of them to increase profits or achieve their other education is provided in very different ways requires added goals. Finally, even if the expansion of private schooling skills. Though neither is easy, governments may deem it brings short-term benefits, it can undermine the political more straightforward to provide quality education than to constituency for effective public schooling in the longer regulate a disparate collection of providers that may not term. It is impossible to make any global statement about have the same objectives. whether the benefits or risks dominate. The bottom line is that countries need to ensure that Experience with public-private partnerships is growing. private schooling does not undermine learning for all. As governments face their own limited capacity to cope Different countries make different choices on private (Box continues next page) Education systems are misaligned with learning | 177 Box 9.3 Can private schooling be aligned to learning for all? (continued) delivery, acting on a variety of motivations. But if they do change in a delivery mechanism. Governments may choose allow or even encourage private schooling, they need to to contract out some service delivery, but they should never remain alert to all the risks just outlined. The problems out- contract out the responsibility for ensuring that all children lined in this Report do not disappear simply because of a and youth have the opportunity to learn. Figure B9.3.1 In Bangladesh, there are 11 different kinds of nonstate providers of presecondary education Number of institutions under each nongovernment provider of presecondary education (2016) Kindergarten BRAC learning centers High madrasha attached NGO schools Ebtedayee madrasha Nongovernment Other NGO learning centers Mosque-based learning centers Temple-based learning centers Quami Tea garden 0 5,000 10,000 15,000 20,000 25,000 Number of institutions Source: WDR 2018 team, using data from Directorate of Primary Education, Bangladesh (2016). Data at http:// bit.do/WDR2018-Fig_B9-3-1. Note: NGO = nongovernmental organization. Source: WDR 2018 team. a. World Bank (2017a). l. Heyneman and Stern (2014). b. D. Capital Partners (2016). m. Bold and others (2017). c. Härmä (2013). n. de la Croix and Doepke (2009); Kosec (2014); Sandström and d. Oketch and others (2010). Bergström (2005). e. Heyneman and Stern (2014). o. Akaguri (2014); Härmä (2011). f. Day Ashley and others (2014); Heyneman and Stern (2014). p. “Educação Integral,” Secretaria de Educação, State of Pernambuco, g. Heyneman and Stern (2014). Várzea, Recife, Brazil, http://www.educacao.pe.gov.br/portal h. Kingdon (2017); Urquiola (2016). /?pag=1&men=70. i. Andrabi, Das, and Khwaja (2008); Day Ashley and others (2014); q. Barrera-Osorio and others (2016). Muralidharan and Sundararaman (2015). r. Kingdon (2017). j. Sakellariou (2017). s. Economist (2017). k. Oketch and others (2010); Tooley (2005). t. Barrera-Osorio and others (2016). to monitor student learning are effective, they can linear relationships from cause to effect. The multiple sometimes lead to biases toward better-performing interactions that characterize teaching and learning students, short-term test preparation, or a narrow and the almost continuous feedback that they provide focus on subjects that are explicitly tested. can result in teachers, parents, and students adapting The multiplicity of actors and institutions in an their behavior in unpredictable ways. For example, education system makes the outcomes of efforts to the introduction of school grants in Andhra Pradesh, improve learning unpredictable.24 Learning is a com- India, and in Zambia failed to improve student learn- plex process that is difficult to break down into simple ing in the long term because parents reduced their 178 | World Development Report 2018 financial support in anticipation of the increase in efficiently, and can troubleshoot in real time—all of government funding.25 Reducing the financial burden which are in short supply in many systems. Moreover, on parents may be a desirable effect of these grants, behavioral economics highlights many cognitive pit- but it was not their primary intent. More generally, falls that policy makers commonly face in complex many factors outside the classroom and the school operating environments. These include difficulty in system, including health and economic shocks, can evaluating policy effectiveness when faced with too alter the impact of interventions aimed at improv- many options; loss aversion, or the tendency to feel ing learning. Failure to learn and adjust policies in failures more intensely than successes, which makes response to such changes often means that interven- policy makers wary of experimentation; biases that tions do not work as planned. lead to selective use of information to reinforce exist- ing views; and relational bias, which makes it harder Education systems are “sticky” for officials with elite educational backgrounds to Education systems are slow to change. Some of the grasp the challenges of mass education.28 best-known successes in reforming systems, such as Education agencies often lack the capabilities in Chile or Finland, took decades from initiation to needed to deal with these complexities.29 A recent fruition. Even at the micro level, such as in schools assessment shows how multitasking and fragmen- in the United States that enacted comprehensive tation within education agencies can blur lines of school reform, it took 8–14 years for the full effects to accountability for learning. In Cyprus, because of the be felt.26 These long time frames present two further absence of a department for human resources and challenges to better aligning education systems with general administration, pedagogical departments had learning. First, to improve learning, policies usually to manage these responsibilities, diverting time from have to remain relatively consistent. This is difficult developing programs and policies.30 Public expendi- under normal circumstances: changes in government, ture and financial accountability assessments also volatile funding, and shifts in the overall economic highlight the low capacity in many developing coun- context all threaten the sustainability of policies.27 But tries in key areas. For example, only about half of the staying the course is even more challenging when 72 low- and middle-income countries assessed since the reforms fail to show any benefits in the short 2010 had any system in place to ensure that resources run. Second, the long lags make program evaluation intended for schools, health clinics, and other service more difficult, because attributing improvements to delivery units reached the front lines.31 specific interventions is especially challenging when their impacts emerge only in the long run. *** Technical challenges and lack of implementation Implementation capacity to improve capacity result in misaligned education systems. learning at scale is often lacking When countries are unable to overcome these chal- Opacity and stickiness make technical alignment hard lenges, their education systems deliver levels of learn- enough to achieve; weaknesses in implementation ing far below what is possible. But tackling the techni- capacity make the task even more daunting. Success- cal barriers to better learning is only part of the battle. ful implementation depends on effective leadership, To break out of low-learning equilibriums, countries coordination between education agencies, and imple- must also address the political constraints that are mentation teams that are motivated, use resources often at the heart of these technical misalignments. Notes 1. Duflo, Dupas, and Kremer (2015). 8. World Bank (2015b). 2. Bold and others (2013). 9. See DISE website, http://udise.in/. 3. Acemoglu (2010). 10. OECD (2016); UIS (2017). 4. Béteille and Ramachandran (2016); Bruns, Filmer, and 11. Rajkumar and Swaroop (2008); Suryadarma (2012). Patrinos (2011); Duthilleul (2005). 12. Muralidharan and others (2017). 5. Hallak and Poisson (2007); Transparency International 13. Pritchett and Filmer (1999). (2009). 14. Bruns and Luque (2015); Finan, Olken, and Pande (2015). 6. Ministry of Finance, Bangladesh (2017). 15. Glewwe and others (2011); OECD (2009). 7. World Bank (2010). 16. World Bank (2015a); Reboot (2013). Education systems are misaligned with learning | 179 17. King and Rogers (2014); Park (2016). and Achievement: A Meta-Analysis.” Review of Educational 18. Chisholm and Leyendecker (2008); Todd and Mason Research 73 (2): 125–230. (2005). Bowman, Kimberly, John Chettleborough, Helen Jeans, 19. See Uy (2017) for number of children; Department of Jo Rowlands, and James Whitehead. 2015. “Systems Education, Philippines (2015) for number of schools. Thinking: An Introduction for Oxfam Programme Staff.” 20. World Bank (2016). Oxfam, Oxford, U.K. 21. Harding and Stasavage (2014); Pierskalla and Sacks Bruns, Barbara, Deon Filmer, and Harry Anthony Patrinos. (2015); Wales, Magee, and Nicolai (2016). 2011. Making Schools Work: New Evidence on Accountability 22. Holmstrom and Milgrom (1991). Reforms. Human Development Perspectives Series. 23. Harding and Stasavage (2014). Washington, DC: World Bank. 24. Burns and Köster (2016); Snyder (2013). Bruns, Barbara, and Javier Luque. 2015. Great Teachers: How to 25. Das and others (2013). Raise Student Learning in Latin America and the Caribbean. 26. Although these comprehensive school reforms differed With Soledad De Gregorio, David K. Evans, Marco across schools, they all targeted reorganizing entire Fernández, Martin Moreno, Jessica Rodriguez, Guill- schools in a coordinated way rather than implementing ermo Toral, and Noah Yarrow. Latin American Develop- single or specialized interventions. See Borman and ment Forum Series. Washington, DC: World Bank. others (2003). Burns, Tracey, and Florian Köster, eds. 2016. Educational 27. World Bank (2017b). Research and Innovation: Governing Education in a Complex 28. Thaler and Sunstein (2008); World Bank (2015c). World. Paris: Organisation for Economic Co-operation 29. Pritchett and Woolcock (2004). and Development. 30. World Bank (2014). Chisholm, Linda, and Ramon Leyendecker. 2008. “Curricu- 31. PEFA (2017). lum Reform in Post-1990s Sub-Saharan Africa.” Interna- tional Journal of Educational Development 28 (2): 195–205. Referencesa D. Capital Partners. 2016. “The Impact of Private Investment Acemoglu, Daron. 2010. “Theory, General Equilibrium, and in Education in Sub-Saharan Africa: How Can Private Political Economy in Development Economics.” Journal of Capital Shape Education across the Continent?” Economic Perspectives 24 (3): 17–32. D. Capital Partners, Dalberg Global Development Advi- Akaguri, Luke. 2014. “Fee-Free Public or Low-Fee Private sors, London. http://www.dalberg.com/wp-content Basic Education in Rural Ghana: How Does the Cost /uploads/2016/05/Private-investments-in-SSA-Education Influence the Choice of the Poor?” Compare: A Journal of .pdf. Comparative and International Education 44 (2): 140–61. Das, Jishnu, Stefan Dercon, James Habyarimana, Pramila Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2008. Krishnan, Karthik Muralidharan, and Venkatesh Sunda- “A Dime a Day: The Possibilities and Limits of Private raraman. 2013. “School Inputs, Household Substitution, Schooling in Pakistan.” Comparative Education Review and Test Scores.” American Economic Journal: Applied Eco- 52 (3): 329–55. nomics 5 (2): 29–57. Barrera-Osorio, Felipe, Pierre Gaspard de Galbert, James P. Day Ashley, Laura, Claire Mcloughlin, Monazza Aslam, Habyarimana, and Shwetlena Sabarwal. 2016. “Impact of Jakob Engel, Joseph Wales, Shenila Rawal, Richard Public-Private Partnerships on Private School Perfor- Batley, et al. 2014. “The Role and Impact of Private mance: Evidence from a Randomized Controlled Trial in Schools in Developing Countries.” Education Rigorous Uganda.” Policy Research Working Paper 7905, World Literature Review, U.K. Department for International Bank, Washington, DC. Development, London. https://www.gov.uk/government Béteille, Tara, and Vimala Ramachandran. 2016. “Contract /uploads/system/uploads/attachment_data/file/439702 Teachers in India.” Economic and Political Weekly 51 (25): /private-schools-full-report.pdf. 40–47. de la Croix, David, and Matthias Doepke. 2009. “To Segre- Bold, Tessa, Deon Filmer, Gayle Martin, Ezequiel Molina, gate or to Integrate: Education Politics and Democracy.” Brian Stacy, Christophe Rockmore, Jakob Svensson, et al. Review of Economic Studies 76 (2): 597–628. 2017. “What Do Teachers Know and Do? Does It Matter? Department of Education, Philippines. 2015. “Fact Sheet— Evidence from Primary Schools in Africa.” Policy Basic Education Statistics.” Manila. Research Working Paper 7956, World Bank, Washington, Directorate of Primary Education, Bangladesh. 2016. Annual DC. Primary School Census. Dhaka, Bangladesh: Information Bold, Tessa, Mwangi S. Kimenyi, Germano Mwabu, Alice Management Division and Monitoring and Evalua- Ng’ang’a, and Justin Sandefur. 2013. “Scaling Up What tion Division, DPE, Ministry of Primary and Mass Works: Experimental Evidence on External Validity in Education. Kenyan Education.” CSAE Working Paper WPS/2013-04, Duflo, Esther, Pascaline Dupas, and Michael R. Kremer. 2015. Centre for the Study of African Economies, University “School Governance, Teacher Incentives, and Pupil- of Oxford, Oxford, U.K. Teacher Ratios: Experimental Evidence from Kenyan Borman, Geoffrey D., Gina M. Hewes, Laura T. Overman, Primary Schools.” Journal of Public Economics 123 (March): and Shelly Brown. 2003. “Comprehensive School Reform 92–110. a.  References to titles of publications that include South Korea Duthilleul, Yael. 2005. “Lessons Learnt in the Use of refer to the Republic of Korea. ‘Contract’ Teachers: Synthesis Report.” International 180 | World Development Report 2018 Institute for Educational Planning, United Nations Ministry of Finance, Bangladesh. 2017. “Medium-Term Bud- Educational, Scientific, and Cultural Organization, Paris. getary Framework (MTBF) 2017–18 to 2019–20.” Finan- Economist. 2017. “Ashes to Classes: Liberia’s Bold Experiment cial System Management Unit, Finance Division, in School Reform.” February 23. http://www.economist Ministry of Finance, Dhaka, Bangladesh. https://mof .com/news/middle-east-and-africa/21717379-war-scorched .gov.bd/en/index.php?option=com_content&view= -state-where-almost-nothing-works-tries-charter-schools article&id=397&Itemid=1. -liberias. Moore, Mark. 2015. “Creating Efficient, Effective, and Just Finan, Frederico S., Benjamin A. Olken, and Rohini Pande. Educational Systems through Multi-Sector Strategies of 2015. “The Personnel Economics of the State.” NBER Reform.” RISE Working Paper 15/004, Research on Working Paper 21825, National Bureau of Economic Improving Systems of Education, Blavatnik School of Research, Cambridge, MA. Government, Oxford University, Oxford, U.K. Glewwe, Paul W., Eric A. Hanushek, Sarah D. Humpage, and Muralidharan, Karthik, Jishnu Das, Alaka Holla, and Aakash Renato Ravina. 2011. “School Resources and Educational Mohpal. 2017. “The Fiscal Cost of Weak Governance: Evi- Outcomes in Developing Countries: A Review of the Lit- dence from Teacher Absence in India.” Journal of Public erature from 1990 to 2010.” NBER Working Paper 17554, Economics 145: 116–35. National Bureau of Economic Research, Cambridge, MA. Muralidharan, Karthik, and Venkatesh Sundararaman. 2015. Hallak, Jacques, and Muriel Poisson. 2007. Corrupt Schools, “The Aggregate Effect of School Choice: Evidence from Corrupt Universities: What Can Be Done? Paris: Interna- a Two-Stage Experiment in India.” Quarterly Journal of tional Institute for Educational Planning, United Nations Economics 130 (3): 1011–66. Educational, Scientific, and Cultural Organization. Newman, John L., Elizabeth M. King, and Husein Abdul- Harding, Robin, and David Stasavage. 2014. “What Democ- Hamid. 2016. “The Quality of Education Systems and racy Does (and Doesn’t Do) for Basic Services: School Education Outcomes.” Background Paper: The Learning Fees, School Inputs, and African Elections.” Journal of Pol- Generation, International Commission on Financing itics 76 (1): 229–45. Global Education Opportunity, New York. Härmä, Joanna. 2011. “Low Cost Private Schooling in India: OECD (Organisation for Economic Co-operation and Devel- Is It Pro Poor and Equitable?” International Journal of opment). 2009. Evaluating and Rewarding the Quality of Educational Development 31 (4): 350–56. Teachers: International Practices. Paris: OECD. —— ——. 2013. “Access or Quality? Why Do Families Living in —— — —. 2016. PISA 2015 Results: Excellence and Equity in Educa- Slums Choose Low-Cost Private Schools in Lagos, tion. Vol. 1. Paris: OECD. Nigeria?” Oxford Review of Education 39 (4): 548–66. http:// Oketch, Moses, Maurice Mutisya, Moses Ngware, and Alex www.tandfonline.com/doi/full/10.1080/03054985.2013 C. Ezeh. 2010. “Why Are There Proportionately More .825984. Poor Pupils Enrolled in Non-state Schools in Urban Heyneman, Stephen P., and Jonathan M. B. Stern. 2014. “Low Kenya in Spite of FPE Policy?” International Journal of Cost Private Schools for the Poor: What Public Policy Is Educational Development 30 (1): 23–32. Appropriate?” International Journal of Educational Develop- Park, Rufina Kyung Eun. 2016. “Preparing Students for South ment 35 (March): 3–15. Korea’s Creative Economy: The Successes and Challenges Holmstrom, Bengt, and Paul Milgrom. 1991. “Multitask of Educational Reform” [refers to Republic of Korea]. Principal-Agent Analyses: Incentive Contracts, Asset Research report, Asia Pacific Foundation of Canada, Ownership, and Job Design.” Journal of Law, Economics, Vancouver. and Organization 7 (January): 24–52. PEFA. 2017. “Assessment Pipeline.” PEFA, World Bank, Wash- King, Elizabeth, and F. Halsey Rogers. 2014. “Intelligence, ington, DC. https://pefa.org/assessments/listing. Personality, and Creativity: Unleashing the Power of Pierskalla, Jan, and Audrey Sacks. 2015. “Personnel Politics: Intelligence and Personality Traits to Build a Creative Elections, Clientelistic Competition, and Teacher Hiring and Innovative Economy.” Paper presented at Republic in Indonesia.” Working paper, World Bank, Jakarta, of Korea–World Bank symposium, “Achieving HOPE Indonesia. (Happiness of People through Education): Innovation Pritchett, Lant, and Deon Filmer. 1999. “What Education in Korean Education for a Creative Economy,” Seoul, Production Functions Really Show: A Positive Theory of November 4. Education Expenditures.” Economics of Education Review Kingdon, Geeta Gandhi. 2017. “The Private Schooling Phe- 18 (2): 223–39. nomenon in India: A Review.” IZA Discussion Paper Pritchett, Lant, and Michael Woolcock. 2004. “Solutions 10612, Institute for the Study of Labor, Bonn, Germany. When the Solution Is the Problem: Arraying the Disarray Kosec, Katrina. 2014. “Relying on the Private Sector: The in Development.” World Development 32 (2): 191–212. Income Distribution and Public Investments in the Rajkumar, Andrew Sunil, and Vinaya Swaroop. 2008. “Public Poor.” Journal of Development Economics 107 (March): Spending and Outcomes: Does Governance Matter?” 320–42. Journal of Development Economics 86 (1): 96–111. Liang, Xiaoyan, Huma Kidwai, and Minxuan Zhang. 2016. Reboot. 2013. “Addressing Teacher Absenteeism in Edo State: How Shanghai Does It: Insights and Lessons from the Highest- Summary of Findings.” Reboot, Abuja, Nigeria. Ranking Education System in the World. Directions in Devel- Sakellariou, Chris. 2017. “Private or Public School Advan- opment: Human Development Series. Washington, DC: tage? Evidence from 40 Countries Using PISA 2012– World Bank. Mathematics.” Applied Economics 49 (29): 2875–92. Education systems are misaligned with learning | 181 Sandström, F. Mikael, and Fredrik Bergström. 2005. “School Wales, Joseph, Arran Magee, and Susan Nicolai. 2016. “How Vouchers in Practice: Competition Will Not Hurt You.” Does Political Context Shape Education Reforms and Journal of Public Economics 89 (2–3): 351–80. Their Success? Lessons from the Development Progress Snyder, Sean. 2013. “The Simple, the Complicated, and the Project.” ODI Dimension Paper 06, Overseas Develop- Complex: Educational Reform through the Lens of Com- ment Institute, London. plexity Theory.” OECD Education Working Paper 96, World Bank. 2003. World Development Report 2004: Making Organisation for Economic Co-operation and Develop- Services Work for Poor People. Washington, DC: World ment, Paris. Bank; New York: Oxford University Press. Suryadarma, Daniel. 2012. “How Corruption Diminishes the ——— —. 2010. “Romania Functional Review: Pre-University Effectiveness of Public Spending on Education in Indo- Education Sector, Final Report.” Report 74287, World nesia.” Bulletin of Indonesian Economic Studies 48 (1): 85–100. Bank, Washington, DC. Thaler, Richard H., and Cass R. Sunstein. 2008. Nudge: ——— —. 2014. “Analysis of the Function and Structure of the Improving Decisions about Health, Wealth, and Happiness. Ministry of Education and Culture of the Republic of New Haven, CT: Yale University Press. Cyprus.” World Bank, Washington, DC. Todd, Alexa, and Mark Mason. 2005. “Enhancing Learning ——— —. 2015a. “Governance and Finance Analysis of the Basic in South African Schools: Strategies beyond Outcomes- Education Sector in Nigeria.” Report ACS14245, World Based Education.” International Journal of Educational Bank, Washington, DC. Development 25 (3): 221–35. ——— —. 2015b. World Bank Support to Early Childhood Develop- Tooley, James. 2005. “Private Schools for the Poor: Education ment: An Independent Evaluation. Washington, DC: World Where No One Expects It.” Education Next 5 (4): 11. Bank. Transparency International. 2009. “Corruption in the ——— —. 2015c. World Development Report 2015: Mind, Society, and Education Sector.” Working Paper 04, Transparency Behavior. Washington, DC: World Bank. International, Berlin. ——— —. 2016. Assessing Basic Education Service Delivery in the UIS (UNESCO Institute for Statistics). 2017. Education Philippines: The Philippines Public Education Expenditure (database). UIS, Montreal. http://data.uis.unesco.org/. Tracking and Quantitative Service Delivery Study. Report Urquiola, Miguel. 2016. “Competition among Schools: Tra­ AUS6799. Washington, DC: World Bank. ditional Public and Private Schools.” In Handbook of ——— —. 2017a. World Development Indicators (database). the Economics of Education, Vol. 5, edited by Eric A. World Bank, Washington, DC. http://data.worldbank Hanushek, Stephen J. Machin, and Ludger Woessmann, .org/products/wdi. 209–37. Handbooks in Economics Series. Amsterdam: ——— —. 2017b. World Development Report 2017: Governance North-Holland. and the Law. Washington, DC: World Bank. Uy, Jocelyn R. 2017. “DepEd Braces for 22m Students.” News- info. June 2. http://newsinfo.inquirer.net/901949/deped -braces-for-22m-students. 182 | World Development Report 2018 SPOTLIGHT 6 Spending more or spending better—or both? Good teachers, conducive learning environments, reliable assessment systems, and innovative learning technologies all cost money. And as more students progress further in school, financing needs will rise. Yet more funding leads to better learning only if it is used well, with an intentional focus on learning outcomes. Patterns of public education to use education to create shared values; delivering education directly, or at least financing it, gives them spending leverage to ensure this happens. But in all these cases, Decisions over how to allocate public spending inevi- governments must weigh benefits against the costs tably require difficult trade-offs. When deciding how of investing in education rather than in some other to spend scarce resources, governments have to weigh area—and they must decide how to spend within the the costs and benefits of different spending decisions, education sector. both of which are typically estimated with large mar- As their spending patterns show, countries are gins of error. Governments also have to weigh the increasingly willing to invest in education. Whether short- and long-term benefits of different spending because of the public economics calculations just dis- choices. Should they spend more on urban infra- cussed or for other political reasons, countries have structure improvements to reduce air pollution in the devoted a rising share of their national income to edu- future, or should they invest today in better primary cation (figure S6.1). In 2012 about two-thirds of coun- health care services to treat respiratory infections? tries that reported information on spending devoted Spending on education is subject to this same over 4 percent of national income to public spending calculus. Education’s many potential benefits for both on education. Education also typically absorbs the larg- individuals and societies (see chapter 1) make it a est single share of a government’s budget, averaging strong candidate for public support. In fact, the obliga- about 15 percent of the budget across low- and middle- tion to provide equitable education is often enshrined income countries. In some countries, the investment in law. Although the high returns to education mean in education is still low, indicating scope for further that many students are willing to bear the costs them- prioritization, but the aggregate trends suggest that selves, there are strong rationales for public financing governments recognize the importance of education. of at least some parts of the education system. First, a concern for fairness induces countries to subsidize Does more spending education for children and youth from the poorest households, because their families may be unable or improve learning outcomes? unwilling to finance their education. Second, because While there is a strong rationale for public invest- education has positive spillovers for others—such as ment in education, the relationship between spend- when it reduces the propensity to commit crime— ing and learning outcomes is often weak. In global individuals may underinvest in their own education, learning assessments, for example, although higher from society’s perspective. Third, governments want per-student spending initially appears to lead to more Spending more or spending better—or both? | 183 Figure S6.1 Governments devote a large share of their budgets to education a. Government education spending as b. Government education spending as percentage percentage of GNP (1999 and 2012) of total government spending (1999 and 2012) 15 15 10 10 Percent Percent 5 5 0 0 Low- Lower- Upper- High- Low- Lower- Upper- High- income middle- middle- income income middle- middle- income income income income income 1999 2012 Source: UNESCO (2015). Data at http://bit.do/WDR2018-Fig_S6-1. Note: Median values are shown. GNP = gross national product. learning at the poorer end of the global income scale, a commonly used mechanism in many countries to the correlation largely disappears once controlling for provide schools with the resources needed to support countries’ per capita income. This finding suggests school improvement. Although grants have often that the correlation is driven more by economic devel- increased student enrollment and retention, they opment than by the level of public spending.1 have had relatively limited effects on learning. For Regional learning assessments—which include example, recent evaluations in Indonesia and Tanza- many more low- and lower-middle-income countries— nia found that school grants alone did not increase also show how inconsistent the association between student learning.4 spending and learning can be. For example, public spending per primary school student increased over the 2000s in both Kenya and Lesotho; yet student Weak links in the learning outcomes improved in Lesotho but declined spending-learning chain in Kenya (figure S6.2, panels a and b). Guate­ mala There are five main reasons why spending does not improved student learning significantly between always lead to better and more equal student learning 2006 and 2013, even though per-student spending outcomes:5 declined over the same period (figure S6.2, panels c and d). Comparing across regions within a country • Spending is not allocated equitably. often reveals similar patterns. In Indonesia during • Funds do not reach schools or are not used for their the 2000s, the link between changes in district edu- intended purposes. cation spending and secondary school examination • Public spending can substitute for private spending. results was very weak.2 These findings indicate that • Decisions on the use of public funding are not education systems, and even schools within the same coherently aligned with learning. system, vary in their ability to translate increased • Government agencies lack the capacity to use fund- spending into better learning outcomes. ing effectively. Providing more resources directly to schools has also had mixed effects on learning in different Public spending is often allocated in ways that exclude environments. A review of two decades of research poor and marginalized children, reducing its over- reveals that the association between many school- all impact on learning. Overall, public education level resources (such as textbooks) and student expenditure tends to favor wealthier, more powerful outcomes is variable.3 School grants have become groups (table S6.1). Poorer households do tend to 184 | World Development Report 2018 Figure S6.2 The relationship between changes in public education spending and student learning is often weak Changes in public education spending and in sixth-grade mathematics learning outcomes, selected countries a. Changes in public b. Changes in sixth-grade education spending mathematics learning outcomes Spending per student, PPP 500 Average SACMEQ score 550 400 300 500 200 450 100 0 400 2000 2005 2010 2000 2005 2010 Kenya Lesotho Malawi c. Changes in public d. Changes in sixth-grade education spending mathematics learning outcomes 1,400 540 Spending per student, PPP Average LLECE score 1,200 520 1,000 500 800 480 600 460 2006 2010 2014 2006 2010 2014 Guatemala Paraguay Peru Sources: WDR 2018 team, using data from UIS (2016) for spending and from World Bank (2017) for student learning. Data at http://bit.do/WDR2018-Fig_S6-2. Note: Per student spending is reported in purchasing power parity (PPP) U.S. dollars. Student learning data are derived from data collected by the Southern and Eastern Africa Consortium for Monitoring Educational Quality (SACMEQ) and the Latin American Laboratory for Assessment of the Quality of Education (LLECE). For each country, the two plotted data points reflect the years for which data are available. receive a greater share of public spending on primary differences across socioeconomic groups, since they education because they tend to have more children typically do not account for the fact that students than wealthier households. But public spending on from poorer families tend to receive lower-quality secondary and tertiary education overwhelmingly schooling than those from wealthier families. Allo- favors wealthier groups, because by the time students cating resources more equitably could therefore raise reach those levels, many of the poor have already left average learning levels. school. In Zambia, 39 percent of secondary education Public funds sometimes fail to reach schools or are not spending was allocated to the richest fifth of house- used as intended. In 2013–14, almost a third of school holds, compared with only 8 percent for the poorest. capitation grants failed to reach Zambian primary The gap is even wider at the tertiary level, where 86 schools.6 In the Philippines, in 2013 about a quarter percent of all public spending is captured by the rich- of similar funds did not reach primary and lower est households. These estimates likely understate the secondary schools.7 In Zambia, funds were diverted Spending more or spending better—or both? | 185 Table S6.1 Inequalities in public education spending are common Incidence of public education spending by household income quintile, selected countries and years Percent Primary Secondary Tertiary Total Country Year(s) Poorest Richest Poorest Richest Poorest Richest Poorest Richest Bangladesh 2010 27 13 13 23 2 55 20 20 Burundi 2006 23 13 12 27 4 59 15 29 Congo, Rep. 2011 21 16 18 18 1 62 — — Ghana 2007 19 13 13 20 4 65 12 34 Honduras 2004 31  6  5 20 1 67 — — Indonesia 2007 26 11 15 19 4 57 20 23 Pakistan 2007–08 25 11 16 23 9 55 17 28 Thailand 2011 25 14 — — 1 73 20 26 Uganda 2009–10 19 15  6 38 1 68 — — Zambia 2010 22 14  8 39 0 86 15 31 Sources: Bangladesh: World Bank (2013a); Burundi: Tsimpo and Wodon (2014); the Republic of Congo: World Bank (2014); Ghana: Wodon (2012); Honduras: Gillingham, Newhouse, and Yackovlev (2008); Indonesia: Wika and Widodo (2012); Pakistan: Asghar and Zahra (2012); Thailand: Buracom (2016); Uganda: Guloba (2011); Zambia: World Bank (2016b). Note: Poorest (richest) refers to the poorest (richest) 20 percent of households. Estimates for secondary in Ghana and the Republic of Congo are for lower secondary. Primary estimates for Thailand also include secondary. — = data not available. for other uses, including to fund district-level operat- because parents reduced their own financial support ing costs. In the Philippines, while district education in anticipation of increased government funding.11 offices reported using some of the funds to pay school Decisions on how to use public resources often lack expenses, this use was not recorded and schools coherent alignment with learning. The evidence on ways had no way of monitoring the spending. Schools to improve learning is growing, suggesting ways to in the Philippines that served poorer students also use funding more effectively. Also important is ensur- received a smaller share of their intended allocation ing that the mix of inputs and interventions that are than schools serving wealthier students.8 Even when funded work together well. Many education systems resources are delivered to schools, they are sometimes find this difficult. For example, more classrooms may not used. In Sierra Leone, a 2008 program success- be built, but there are insufficient funds to hire the fully increased the delivery of textbooks to schools, teachers needed to use them. Teachers are present but the textbooks had no impact on learning because in classrooms, but they lack the learning materials they were stored as a hedge against future shortfalls needed to teach effectively. Improving coherence rather than distributed to students.9 is not just about the mix of inputs, but also about Household spending can also affect the link between the systems that manage these inputs. In Tanzania, public spending and outcomes. Taking account of house- grants given to schools were ineffective on their hold spending on education can alter the picture own, but combining grants with teacher incentives of overall spending across countries. Government ensured the grants were used effectively to improve expenditure as a proportion of gross domestic prod- student learning.12 In Indonesia, school grants uct (GDP) in Nepal is much lower than in Vietnam. improved learning only when they were combined However, when all public and private spending on with measures to link school committees with village education is taken into account, spending in Nepal is authorities.13 much higher.10 Households can also react to increases The government agencies responsible for managing edu- in public education spending by lowering their own cation often lack the capacity to use resources effectively. The contributions. For example, the introduction of school Philippines recently embarked on an ambitious edu- grants in India and Zambia had no effect on learning cation reform backed up with significant increases in 186 | World Development Report 2018 public investment. A central element of the program is The key will be to use those additional resources the introduction of two additional years of secondary in ways that improve learning, especially for disad- education, which in turn requires the rehabilitation vantaged children. Costing exercises are sometimes and expansion of school infrastructure to provide the misinterpreted as implying that more spending is all places needed for senior high school. Despite a 19-fold that is needed. But because there is no certainty that increase in the infrastructure budget between 2005 spending will lead to better outcomes, spending bet- and 2015, lack of government capacity to manage ter will also be essential—as the Education Commis- such a massive school building program has meant sion emphasizes. When education is funded using that a large share of the resources remained unspent. resources diverted from other pressing public needs In 2014 only 64 percent of the infrastructure budget such as health or infrastructure, or funded through was committed. And even where classrooms were debt to be repaid by the next generation, it is crucial built, school principals have been largely unsatisfied that spending be oriented toward what will improve with their quality.14 learning for all. How to achieve this is the focus of this World Development Report. Spending to improve Spending more can be an important first step to spending better, but, again, increasing spending learning alone is not sufficient to improve learning. The Achieving education goals, whether national or politics of education reform sometimes requires global, will certainly require more spending in the compensating stakeholders who might lose out, coming decades. The Education Commission esti- or spending more to lay the foundations for future mates that low- and middle-income countries will reform. For example, addressing the low pay of teach- have to increase spending by 117 percent between ers in Peru was an important prerequisite for intro- 2015 and 2030 to enable most children to complete ducing the reforms (such as linking teacher career primary and secondary education with minimum paths to performance) that underpinned improve- levels of learning, as the Sustainable Development ments in learning outcomes.16 However, in other Goals call for.15 Reliably estimating such global costs cases strategies of this kind have worked less well. is difficult because doing so requires accurate infor- A 2006 education finance law in Argentina aimed at mation on many aspects of country systems that is reversing declines in quality led to a near-doubling of often unavailable. It also requires making assump- education spending as a share of GDP (from 3.5 to 6 tions—for example, about optimal class size—that, percent) between 2005 and 2013. The new resources while valid for some countries, may not apply to were used to increase teacher hiring, raise teacher others. Notwithstanding these difficulties, exercises pay, and improve school infrastructure. Yet despite of this kind offer useful information on what school these improvements in inputs, learning outcomes expansion of reasonable quality might cost. That have improved only marginally in recent years and information indicates that, even with greater effi- are still below 2003 levels.17 These experiences high- ciencies, it will be impossible to extend schooling for light the need to strengthen the links in the spending- hundreds of millions of students without investing learning chain, if more spending is to lead to better more in education. learning outcomes. Notes 1. See chapter 9 and Altinok (2010). 10. UIS (2016). 2. World Bank (2013b). 11. Das and others (2013). This may be beneficial if it reduces 3. Glewwe and others (2011). financial burdens on parents, but that was not the pri- 4. Mbiti, Muralidharan, and Schipper (2016); Pradhan and mary purpose of these grants. others (2014). 12. Mbiti, Muralidharan, and Schipper (2016). 5. See Filmer, Hammer, and Pritchett (2000) for a similar 13. Pradhan and others (2014). analysis of health spending. 14. World Bank (2016a). 6. World Bank (2016c). 15. This includes only projected costs of primary and second- 7. World Bank (2016a). ary education. See Education Commission (2016, table 3). 8. See Policy Note 5, figure 8 in World Bank (2016a). 16. Bruns and Schneider (2016). 9. Sabarwal, Evans, and Marshak (2014). 17. de Hoyos, Holland, and Troiano (2015). Spending more or spending better—or both? | 187 References Altinok, Nadir. 2010. “Do School Resources Increase School Artha. 2014. “Improving Educational Quality through Quality?” Brussels Economic Review 51 (4): 435–58. Enhancing Community Participation: Results from a Asghar, Zahid, and Mudassar Zahra. 2012. “A Benefit Inci- Randomized Field Experiment in Indonesia.” American dence Analysis of Public Spending on Education in Paki- Economic Journal: Applied Economics 6 (2): 105–26. stan Using PSLM Data.” Lahore Journal of Economics 17 (2): Sabarwal, Shwetlena, David K. Evans, and Anastasia Mar- 111–36. shak. 2014. “The Permanent Input Hypothesis: The Case Bruns, Barbara, and Ben Ross Schneider. 2016. “Managing of Textbooks and (No) Student Learning in Sierra Leone.” the Politics of Quality Reforms in Education: Policy Les- Policy Research Working Paper 7021, World Bank, Wash- sons from Global Experience.” Background Paper: The ington, DC. Learning Generation, International Commission on Tsimpo, Clarence, and Quentin Wodon. 2014. “Measuring Financing Global Education Opportunity, New York. the Benefit Incidence of Public Spending for Education Buracom, Ponlapat. 2016. “The Distributional Effects of in Burundi.” Global Partnership for Education, World Social Spending in Thailand: Evidence from a New Data- Bank, Washington, DC. base.” Asian Politics and Policy 8 (2): 263–79. UIS (UNESCO Institute for Statistics). 2016. “Who Pays for Das, Jishnu, Stefan Dercon, James Habyarimana, Pramila What in Education? The Real Costs Revealed through Krishnan, Karthik Muralidharan, and Venkatesh Sunda- National Education Accounts.” UIS, Montreal. raraman. 2013. “School Inputs, Household Substitution, UNESCO (United Nations Educational, Scientific, and Cul- and Test Scores.” American Economic Journal: Applied tural Organization). 2015. “Global Monitoring Report Economics 5 (2): 29–57. 2015: Education for All 2000–2015: Achievements and de Hoyos, Rafael E., Peter A. Holland, and Sara Troiano. 2015. Challenges.” UNESCO, Paris. “Understanding the Trends in Learning Outcomes in Wika, Gek Sintha Mas Jasmin, and Tri Widodo. 2012. “Distri- Argentina, 2000 to 2012.” Policy Research Working Paper bution of Government Spending on Education in Indo- 7518, World Bank, Washington, DC. nesia.” MPRA Paper 79501, Munich Personal RePEc Education Commission. 2016. The Learning Generation: Invest- Archive, University Library of Munich. ing in Education for a Changing World. New York: Inter­ Wodon, Quentin, ed. 2012. Improving the Targeting of Social national Commission on Financing Global Education Programs in Ghana. World Bank Study Series. Washing- Opportunity. ton, DC: World Bank. Filmer, Deon, Jeffrey S. Hammer, and Lant H. Pritchett. 2000. World Bank. 2013a. “Bangladesh Education Sector “Weak Links in the Chain: A Diagnosis of Health Policy Review—Seeding Fertile Ground: Education That Works in Poor Countries.”  World Bank Research Observer 15 (2): for Bangladesh.” Report 86237, World Bank, Dhaka, 199–224. Bangladesh. Gillingham, Robert, David Newhouse, and Irene Yackovlev. —— ——. 2013b. “Spending More or Spending Better: Improv- 2008. “The Distributional Impact of Fiscal Policy in Hon- ing Education Financing in Indonesia.” Report 73050-ID, duras.” IMF Working Paper WP/08/168, International World Bank, Jakarta, Indonesia. Monetary Fund, Washington, DC. —— ——. 2014. Republic of Congo: Enhancing Efficiency in Educa- Glewwe, Paul W., Eric A. Hanushek, Sarah D. Humpage, and tion and Health Public Spending for Improved Quality Service Renato Ravina. 2011. “School Resources and Educational Delivery for All. Report AUS5649. Washington, DC: World Outcomes in Developing Countries: A Review of the Lit- Bank. erature from 1990 to 2010.” NBER Working Paper 17554, —— ——. 2016a. Assessing Basic Education Service Delivery in the National Bureau of Economic Research, Cambridge, MA. Philippines: The Philippines Public Education Expenditure Guloba, Madina. 2011. “Public Expenditure in the Educa- Tracking and Quantitative Service Delivery Study. Report tion Sector in Uganda: A Benefit Incidence Analysis AUS6799. Washington, DC: World Bank. (Phase 2).” Research Series 110, Economic Policy Research —— ——. 2016b. “Education Public Expenditure Review in Centre, Kampala, Uganda. Zambia.” Education Global Practice Series, World Bank, Mbiti, Isaac M., Karthik Muralidharan, and Youdi Schipper. Washington, DC. 2016. “Inputs, Incentives, and Complementarities in Pri- —— ——. 2016c. “Education Sector Public Expenditure Track- mary Education: Experimental Evidence from Tanzania.” ing and Service Delivery Survey in Zambia.” Education Working paper, University of California at San Diego. Global Practice Series, World Bank, Washington, DC. Pradhan, Menno, Daniel Suryadarma, Amanda Beatty, Maisy —— ——. 2017. EdStats: Education Statistics (database). Wash- Wong, Arya Gaduh, Armida Alisjahbana, and Rima Prama ington, DC. datatopics.worldbank.org/education/. 188 | World Development Report 2018 Unhealthy politics drives misalignments 10 Politics can intensify misalignments in education systems, when the vested interests of stakeholders divert systems away from learning. This can happen at various stages, from setting policy goals to designing, implementing, evaluating, and sustaining reforms. Even when many individual actors are committed to learning, a system can remain stuck in a low-learning trap. Education systems are complex. Aligning an educa- even when there is consensus on technical design, tion system’s goals, financing, and incentives with students may not benefit. For example, in 1996–97 student learning is difficult for technical reasons. But the superintendent of New York City’s District there are also political reasons systems do not prior- 29 rigged a $6 million contract, awarding it to a itize student learning. Political impetus to fix mis- computer company affiliated with a politically con- alignments can help achieve important educational nected property developer. In return, the company objectives—as it has in Chile, England, and India (see gave the superintendent expensive gifts, while chapter 11)—but unhealthy politics can make things delivering archaic or nonfunctioning computers worse. Too often, education interventions, whether to students. Teachers had been counting on decent big reforms or day-to-day implementation steps, are computers to help their students in math; without compromised because powerful individuals or groups the computers, the students lost out.2 can make others act in ways that serve private inter- In 2009 Mexico’s federal government introduced •  ests rather than the collective good.1 Powerful actors a plan for competitive recruitment of teachers, frequently benefit from the status quo and devise whereby all candidates were required to take a test mechanisms to preserve it, regardless of the impact covering content knowledge, pedagogical mastery, on system performance. These mechanisms result in and ethics. Designing the tests was technically dif- actors being trapped in low-learning equilibriums. ficult. But the technical challenges paled next to the political impediments created by local affiliates of Unhealthy politics can Mexico’s teachers’ union, the Sindicato Nacional de Trabajadores de la Educación (SNTE), which has 1.4 intensify misalignments in million members. The policy change meant that the education systems opportunity for patronage-driven hiring would van- Many education systems encounter political imped- ish. Because of strong opposition from the SNTE, iments and rent-seeking, making alignment much the reform was diluted, making it applicable only harder to achieve. Consider these examples: to a small pool of vacancies. Estimates suggest that up to 85 percent of hiring in 2010 was discretionary •  Using computers to educate students requires dif- rather than competitive. Recent evidence indicates ficult technical decisions on program design. But that the teachers hired through discretionary Unhealthy politics drives misalignments | 189 methods were much less effective at improving interventions threaten interests, whether they be a student learning than those hired competitively.3 person’s financial, ideological, or status-related inter- Vyapam (http://www.vyapam.nic.in) is the •  ests, resistance from different parts of the system can government-run professional examination board be expected. The net effect: the system is pulled away in Madhya Pradesh, India. It conducts large-scale from a focus on learning (figure 10.1). entrance tests for admission into courses such as Education systems are vulnerable to political inter- medicine and for recruitment into state government ference because they are opaque and because teachers jobs such as the police. Designing entrance tests constitute a large base of government employees. The and ranking candidates are technically challenging opacity of education systems, coupled with uncer- when there can be more than 100,000 candidates. But tainty about how a specific education policy will affect political economy factors intrude as well: recently, learning, is fertile ground for contestation of reforms. rent-seeking is alleged to have undermined the goal Teachers—the most important factor in learning— of fair, transparent admissions. In 2013 an indepen- have traditionally been important grassroots political dent probe exposed a potential multibillion-dollar actors, because of their geographic spread and regular scheme in which senior politicians and government interaction with parents. Two characteristics make officials had allegedly set up a system allowing teachers especially attractive as patronage appointees. unqualified candidates to pay bribes, often to middle- First, entry costs to the profession are often low. Sec- men, to receive high rankings in entrance tests.4 In ond, the impact of incompetent patronage appointees 2015, the Supreme Court of India transferred the case on learning is not immediately visible, so it has few from the state government to the country’s premier reputational consequences for politicians, especially investigative agency, the Central Bureau of Investiga- if they are already operating on a short time horizon.6 tion, which is currently pursuing the investigation. Because of the size of the teaching force, teach- ers’ unions can be politically important. The political Education systems involve many stakeholders with power of a union depends on how effectively its leader- multiple, often contradictory, interests.5 These systems ship can mobilize teachers, which varies widely within are not just about students, teachers, or principals. and across countries. In many countries, not all teach- They also involve politicians, bureaucrats, the judi- ers are union members or engaged in union activity.7 ciary, private players, and more. Participants linked to Whether union activity helps or hinders education these institutions have a vested interest in how the sys- reform ultimately depends on several factors (box 10.1). tem works, including its structure and funding. A text- book supplier may want to provide a quality product, but it also cares about profits. A politician may want Multiple actors and interests: to make teachers accountable for student learning, but Pulling the system out of also realizes the electoral risks of teacher opposition. alignment at each A bureaucrat may support meritocratic admissions, but also accepts a “token of appreciation” for ensuring step of the policy cycle the admission of an acquaintance’s child to a desirable Personal interests influence reform at every step. school. A parent may want to complain about a teacher, Vested interests—of teachers, principals, bureaucrats, but worries that her child could suffer retaliation. politicians, parents, students, the judiciary, civil soci- Vested interests are not confined to private or ety organizations, the private sector—are influential rent-seeking interests. Actors in education systems at every step of the education policy cycle. Broadly, are often driven by their values or ideology, especially these steps are setting policy goals, designing poli- when the consequences of education policies are not cies, implementing policies, evaluating policies, and readily apparent. Examples include a commitment to sustaining policy reforms. The forces that detract public schools versus public-private choice, secular from alignment tend to be magnified in conflict set- education versus religious, and accountability for tings (box 10.2). test scores versus a focus on teacher qualifications. In addition, education systems can be used by dominant Setting policy goals ethnic groups—especially in multilingual or multi- In many cases, policies are not chosen for their religious societies—to promote their positions while effectiveness in improving learning. Often, they are suppressing minorities. guided instead by the vested interests of powerful Multiple interests jeopardize learning goals. Bal- actors. Policies to hire teachers tend to be popular ancing multiple interests is difficult. When education with politicians, teachers, and parents because they 190 | World Development Report 2018 Figure 10.1 Contradictory interests detract from learning objectives CiviC l iv il so ians orgS o anic ietciet tic P oli z at y Oy ion r s gs. Pe Le rs er ar r to e s/c ch ec ne om te s a rs Te mun Priva ities LEARNING Sch B u re s ia ry ut oo np auc lm an dic ag li o rat ho Ju em Sc s ent In te rn s at i or on ct al a ra c to r s Othe Source: WDR 2018 team. Box 10.1 How do teachers’ unions affect learning? Teachers’ unions are important institutions for protecting the It is impossible to say that unions always help or harm rights of teachers, but do they matter for student learning? student learning; it depends on their characteristics and The quantitative literature identifies situations in which behaviors, as well as the context in which they operate. All unions may have undermined high-quality teaching and countries have unions, but they vary in membership and learning. By fighting for higher salaries while protecting number. Figure B10.1.1 shows the wide variation in teacher incumbent teachers from outside competition, unions some- unionization across countries. Some countries, such as times stifle the formation of an effective teaching cadre.a Finland and Mexico, have one dominant teachers’ union, A study in India finds that union membership is negatively whereas others, such as India and South Africa, have several. correlated with student achievement.b However, hidden There are also institutional variations in teachers’ unions, behind large-scale correlations is evidence of union behavior such as differences in internal organization, stability, and that has been beneficial for education reform efforts, includ- party affiliation. In the United States, some have argued ing efforts by the Zambia National Education Coalition, the that teachers’ unions resist education reforms because Uganda National Teachers’ Union, and the Confederación union leaders represent the median teacher, and if leaders Nacional de Maestros de Educación Rural de Bolivia. supported these reforms, they would be voted out.c On the (Box continues next page) Unhealthy politics drives misalignments | 191 Box 10.1 How do teachers’ unions affect learning? (continued) Figure B10.1.1 Teacher unionization varies across countries Union membership as a percentage of total teachers, selected countries (2012–15) 100 80 60 Percent 40 20 0 Mexico Finland Pakistan South Africa Uganda Kenya Korea, Rep. Source: Shrestha (2017). Data at http://bit.do/WDR2018-Fig_B10-1-1. Note: Bars represent the ratio of union members to teachers. In Mexico, because the union includes a sufficient number of retirees and nonteaching staff, the ratio exceeds 100 percent. other hand, evidence from Argentina and Mexico suggests In summary, the outcome of union behavior will depend that union behavior (and ability to resist reform) depends on how the proposed reform aligns with the interplay on the influence of partisan identities, organizational of a union’s goals, quantitative strength, stability, and stra- fragmentation, and the competition for union leadership.d tegic alliances. Sources: WDR 2018 team, based on: Carnoy (2007); Eberts and Stone (1987); Hoxby (1996); Kingdon and Teal (2010); Moe (2001, 2011); Murillo (1999, 2012); Shrestha (2017). a. Hoxby (1996). b. Kingdon and Teal (2010). c. Moe (2011). d. Murillo (1999). bring visible, immediate benefits. Likewise, large- performance appraisal, accompanied by tools to help scale school construction programs tend to attract teachers improve. Yet, because such reforms could considerable support. In a diverse range of countries expose poorly performing teachers, the reforms rarely (Cambodia, Colombia, Mozambique), policy makers reach the policy arena. Other policies that threaten have invested in building preschools instead of in teaching jobs include school consolidation or closing. less visible but more effective process-oriented early Such policies have been difficult to implement, because childhood initiatives, such as programs to improve parental support for local schools makes it politically parent-child interactions. In Bangladesh, until recently infeasible to close small, high-cost rural schools.9 In it was much easier to unite elites around the need for Bulgaria, school principals have been reluctant to let mass education than around raising educational stan- teachers go, despite declining school-age populations. dards in schools.8 In several countries, strong teachers’ unions have pre- It is also difficult to adopt a policy goal that threat- vented large-scale teacher redundancies.10 ens or reconfigures jobs, as is true for most quality- enhancing education policies. For example, an alter- Designing policies native to the politically popular policy of reducing Even when the goal of a policy is to improve stu- class size would be to introduce serious teacher dent learning, its final design often reflects what 192 | World Development Report 2018 Box 10.2 How politics can derail learning in conflict-affected states Conflict-affected regions face important political economy of the security challenge of examining and addressing local constraints in developing their education systems. Violent differences during a violent conflict. Domestic considerations conflict hampers learning in an immediate sense when create challenges as well, as in decisions about the medium schools, students, and teachers are targeted, and also of instruction in schools. For example, approaches that over the long term when security issues divert attention guarantee the right of all children to be educated in their and resources from schools. The “security first” approach “own” language can be used by vested groups to segregate often hides the vested interests in the security sector— communities, as happened in Bosnia and Herzegovina in powerful military and political actors, as well as external the 1990s. Political economy challenges can also occur within political interests—which have agendas that overshadow healthy democracies that have conflict regions. Insurgency- development. affected parts of Chhattisgarh, India, have found it difficult Policy in politically weak or fragile conflict-affected states to implement education reforms aimed at improving teacher can be influenced by both external and internal power rela- accountability and student learning. A key concern has been tions. External aid agencies are often handicapped by the an overall lack of funding, payment delays, and interruption difficulty of delivering aid in violent or insecure contexts. This of teachers’ pay. Reduced funding may steer systems toward difficulty usually leads to an emphasis on generalized educa- employing patronage hires, allowing less qualified and often tional frameworks rather than context-specific ones because uncertified teachers to replace trained teachers. Sources: WDR 2018 team, based on: Bensalah (2002); De Herdt, Titeca, and Wagemakers (2010); Magill (2010); Mosselson, Wheaton, and Frisoli (2009); Novelli and others (2014); Rose and Greeley (2006); Shields and Rappleye (2008). powerful interests want, which can undermine the Evaluation to establish standardized procedures for goal. Decentralization policies aim to increase policy monitoring school performance and establishing responsiveness and accountability, but many times the support needs of schools.12 Though the policy they delegate accountability for results without the was meant to be supportive, premised on building authority or resources to achieve them. In Indone- collaborations and mentorship, the South African sia, Pakistan, and some Latin American countries, Democratic Teachers’ Union—the country’s largest— major decentralization efforts have struggled (at remained opposed to it. Many of the union’s chap- least initially) to find the right balance between cen- ters blocked the adoption of the policy in schools in tral and local funding, or between central and local their areas. A similar situation occurred in Mexico authority.11 Central authorities often attempt to limit in 2012.13 the power of lower units of government because Well-intentioned reforms may threaten the legal local governments—being closer to the people—can entitlements of individuals—and when, understand- threaten the political power of more distant govern- ably, they turn to the courts for redress, reforms risk ments. At the same time, local governments may be being stalled. In Peru, unions resisted a new law on unwilling to assume greater responsibility or adopt teacher evaluations by challenging its constitutional- national norms—for example, on the inclusion of ity.14 The ensuing court process then delayed imple- marginalized groups. mentation of the first round of evaluations. Though the court eventually upheld the law, for political Implementing policies reasons the union was given a major concession: the Policy makers may face little resistance when sign- law applied only to newly hired teachers. Similarly, in ing off on a policy, but implementation can be com- 2002 teachers in Andhra Pradesh, India, stalled imple- promised if the policy threatens powerful interests. mentation of a policy on teacher transfers by filing a Policies designed to measure teacher performance court case.15 have been particularly difficult to implement. In Parents can also make it difficult to implement 2000 South Africa’s (then) Department of Education learning-focused policies. A common example is introduced the National Policy on Whole-School parents helping children to cheat on examinations, Unhealthy politics drives misalignments | 193 which makes it hard to measure student learning. Sustaining policy reforms In 2015 the global media broadcast images of family Even when difficult reforms are implemented, they members in Bihar, India, handing cheat sheets to can be undone. Reversal can be incremental, with children inside a building taking exams.16 Perhaps policy makers softening elements to appease specific parents are aware that their children have not learned groups. In the late 1990s, the government of Madhya much in school, leaving them uncompetitive against Pradesh, India, began hiring teachers from the newly better-prepared or more affluent children. created shiksha karmi cadre, under which all new teach- ers were to be locally recruited and put on 10-month Evaluating policies contracts. In response, teacher applicants filed court Indicators of the effectiveness of policies are often cases arguing that the policy violated their constitu- chosen in a way that lets powerful groups off the tional rights, which emphasize that no citizen can be hook. When a policy fails, frontline bureaucrats ineligible for office based on criteria such as place of or principals may face repercussions regardless of birth. Burdened with litigation and pressure, the gov- whether failure was in their control. As a result, deci- ernment redesigned the policy, making concessions sions on what to measure and track are less a reflec- on local recruitment and qualifications.21 Similarly, tion of what the education system values than of who in São Paulo, Brazil, reforms of teacher career tracks is willing to be held accountable for what. For exam- introduced in 2009 were gradually undone by 2011 ple, India’s landmark Right to Education Act (Act No. under a new education minister.22 35, 2009) did not originally contain any measure of Reversal can be sudden. In Ghana, an early child- teacher effectiveness or of student learning (although hood care and development body was set up under subsequent rules and amendments have sought to the office of the president, with high-level support. introduce the quality dimension). Similarly, accred- But a change in administration put the office under itation systems in higher education tend to focus the Ministry of Gender, Children, and Social Protec- on inputs—such as number of classrooms, amount tion, lowering the priority and visibility given to early of equipment, or faculty-student ratios—instead of childhood issues.23 In República Bolivariana de Ven- what students have learned or whether they become ezuela, decades of reforms that had created a strong employed.17 Such an approach limits liability, but jeop- higher education system were reversed when a new ardizes learning goals. populist government set the goal of universalizing Data can be manipulated. Even when indicators higher education. That effort, without prepared stu- track meaningful variables, data quality may be com- dents, adequate faculty, or the appropriate infrastruc- promised. Data on outcomes can be gamed; decisions ture, has weakened the country’s education system.24 on who collects data and how often are made using These cases raise two important issues. First, why subjective criteria. Gaming might take the form of do parents and students have such a limited voice candidates hiring test takers, parents facilitating in influencing the vested interests that jeopardize cheating, teachers misreporting student test scores, quality-enhancing reforms? It could be that those most or government officials encouraging teachers to mod- likely to benefit from reforms—especially parents and ify test scores.18 In several countries, comparisons of students—are often poorly organized. Moreover, the national enrollment data with household survey data immediate gains of any proposed policy tend to be find systematic discrepancies, with official statistics uncertain, making it harder to mobilize support for the sometimes exaggerating progress.19 reform. Parents may also find that the potential ram- A subtler barrier to effective monitoring and ifications of opposing a teacher or politician could be evaluation is when governments collect mountains formidable for their children. By contrast, those who of data but not in a format that facilitates decision stand to lose from reforms tend to be better aware of making. In some countries, the many efforts to col- their losses and, in many cases, are better organized for lect data on indicators create the illusion that policy collective action.25 makers are actively engaged in data-driven decision Second, more generally, why do these low-learning making to improve school quality. But by the time equlibriums persist? For every teacher, bureaucrat, data entry is completed, it is time for the next round politician, judge, or businessperson who jeopardizes of data collection. No serious analysis is conducted, learning, there are several who feel deeply account- feedback is not provided to schools, or the data are too able for student learning and act to strengthen edu- broad to be useful.20 Such instances devalue data in cation systems. Yet individual actors find it hard to decision making. escape these traps. Why? 194 | World Development Report 2018 Trapped in low-accountability, Figure 10.2 Interdependencies characterize the relationship between teachers and politicians low-learning equilibriums The formal rules of the game—that is, the laws Promise of electoral support and policies governing education systems—already reflect power asymmetries.26 When specific policy goals are chosen, when finance is allocated to certain tasks, when teachers’ unions bargain for concessions, Teachers Job or desirable transfer Politicians preexisting power asymmetries and struggles are expressed through policy. Source: WDR 2018 team, based on Béteille (2009). But such decisions also reveal the informal con- tracts that determine which formal rules are chosen or followed. Unwritten codes of conduct derive from But in another context, dependence patterns could the values, expectations, and cultural norms in a reverse. In 2007, when teachers in Rajasthan threat- social setting, and they are important in determining ened the ruling party with electoral sabotage, they the extent, nature, and strength of politics in that set- were the ones who wielded power, with the ruling ting.27 In Indonesia, where older colleagues are treated party dependent on them for victory. But in another with considerable courtesy, school mergers have often context at the same time, the ruling party controlled been delayed informally until principals who stood to individual teachers through patronage-based appoint- lose their jobs retired.28 In rural Rajasthan, India, field ments and transfers (figure 10.2). Because these research finds that teachers often have to pay bribes to opposing relations occurred simultaneously, the get needed services, such as a transfer. Interestingly, distinctions between who was more dependent and the norm differs by gender: male teachers make the who was accountable to whom became blurred.33 Such payments directly, while female teachers typically go interdependencies govern relationships between through a male relative.29 various participants in education systems, such as The widespread operation of informal networks parent-teacher or bureaucrat-middleman interactions. reveals a lack of generalized trust within systems. Interdependencies can become coercive and Unwritten codes of conduct between individuals entrenched. This happens when actors are unable can thrive only if there is sufficient trust between to break out of informal contracts. In the Vyapam them. Each must trust that the other will behave as case in Madhya Pradesh, India, several bureaucrats, expected. Yet as individuals cultivate personalized fearing adverse career repercussions, allegedly joined trust-based relationships—often undermining learn- the scam, making it much worse than otherwise ing or equity goals in the process—overall trust in the possible. Then others joined—with middlemen pur- system suffers.30 portedly profiting off the connections made between As systems grow more complex and the number the various players. What started out as a small-time of actors and interactions increases, uncertainty operation allegedly became institutionalized (albeit multiplies. Trusting others becomes increasingly dif- informally) as people began to believe they would lose ficult. Creating reciprocal obligations helps manage out if they questioned the status quo.34 Likewise, in the uncertainty.31 These obligations do not need to be New York City the unwritten power of school board spelled out; the social setting ensures they are under- members forced superintendents and principals to stood. During the Suharto era in Indonesia, teachers routinely allow wasteful practices.35 This pattern were required to display “mono-loyalty” to the state repeats itself across cases, countries, and time. and teach compulsory courses in the state ideology, As participants get trapped in unhealthy inter- Pancasila.32 If they did not, they knew they risked dependencies, they devise mechanisms to protect demotion or transfer to schools in undesirable areas. themselves from undue blame and punishmen—and In SNTE-dominant parts of Mexico, teachers knew avoid taking risks. Fearing repercussions for uncoop- that if they did not support the SNTE, they risked erative behavior, actors make choices that provide the unfavorable transfers or being sidelined. appearance of change—for example, when a politician Reciprocal obligations complicate accountability. presides over school openings but does not address Power relations between entities and groups depend teacher absenteeism, a judge delays case hearings on context. One group may be more dependent on endlessly, or a parent sits on a dormant school com- another—and therefore less powerful—in one context. mittee. These actors become averse to taking risks or Unhealthy politics drives misalignments | 195 innovating. Such behavior coexists with a perverse quo are likely to put themselves at considerable pro- form of information management. For fear of being fessional risk. The system leaves them little choice wrongly implicated in illegal behavior, officials some- but to conform. The problem is not limited to specific times generate mountains of paper, files, and data, individuals, but arises from the multiple interests of paralyzing the system instead of providing relevant actors and the underlying incentives in education information.36 The opacity, stickiness, and low capac- systems. The accountability needed to ensure student ity of education systems make it easier to exaggerate learning becomes secondary. accomplishments and cover up performance problems. Abdicating responsibility and avoiding blame *** erode an education system’s ability to function, This is the story of unhealthy politics.37 Healthy pol- thereby perpetuating a low-accountability, low- itics can generate the momentum for reform and learning equilibrium. Teachers, bureaucrats, judges, deliver results for education outcomes, as chapter 11 or politicians who fail to cooperate with the status shows. Notes References 1. World Bank (2017). Albornoz, Orlando. 2007. “Recent Changes in Venezuelan 2. Segal (2005). Higher Education.” International Higher Education 48 3. Bruns and Luque (2015); Estrada (2016). (Summer): 18–19. 4. Hindustan Times (2015); Sethi (2015). Bailey, Frederick George. 1969. Strategems and Spoils: A Social 5. Grindle (2004); Moe and Wiborg (2017). Anthropology of Politics. Pavilion Series: Social Anthropol- 6. Bruns and Schneider (2017). ogy. New York: Schocken Books. 7. Moe and Wiborg (2017); Murillo (1999). Bensalah, Kacem, ed. 2002. “Guidelines for Education in Sit- 8. Hossain and others (2017). uations of Emergency and Crisis: EFA Strategic Plan- 9. Forgy (2009). ning.” Division of Policies and Strategies of Education, 10. Forgy (2009); Pepinsky, Pierskalla, and Sacks (2017). United Nations Educational, Scientific, and Cultural 11. Indonesia: Rosser and Fahmi (2016); World Bank (2007); Organization, Paris. Pakistan: Dundar and others (2014); Ghaus-Pasha (2011); Béteille, Tara. 2009. “Absenteeism, Transfers and Patronage: Latin America: Willis, da CB Garman, and Haggard The Political Economy of Teacher Labor Markets in (1999). India.” PhD dissertation, Stanford Graduate School of 12. Taylor, Muller, and Vinjevold (2003). Education, Stanford University, Stanford, CA. 13. Bruns and Luque (2015). —— — —. 2015. “Fixers in India’s Teacher Labor Markets.” Asian 14. Bruns and Luque (2015). Survey 55 (5): 942–68. 15. Sharma and Ramachandran (2009). Bill and Melinda Gates Foundation. 2015. “Teachers Know 16. CNN.com (2015). Best: Making Data Work for Teachers and Students.” 17. Stevens and Kirst (2015). https://s3.amazonaws.com/edtech-production/reports 18. Jhingran (2016); Levitt and Dubner (2010). /Gates-TeachersKnowBest-MakingDataWork.pdf. 19. Sandefur and Glassman (2015). Bjork, Christopher. 2006. “Decentralisation in Education, 20. Bill and Melinda Gates Foundation (2015); Jhingran Institutional Culture, and Teacher Autonomy in Indone- (2016). sia.” In Decentralisation and Privatisation in Education: The 21. Sharma (1999). Role of the State, edited by Joseph Zajda, 133–50. Dordrecht, 22. Bruns and Luque (2015). the Netherlands: Springer. 23. IEG (2015). Bruns, Barbara, and Javier Luque. 2015. Great Teachers: 24. Albornoz (2007). How to Raise Student Learning in Latin America and 25. Grindle (2004). the Caribbean. With Soledad De Gregorio, David K. Evans, 26. World Bank (2017). Marco Fernández, Martin Moreno, Jessica Rodriguez, 27. Bailey (1969); Easton (1979); Greif (2006); North (1991). Guillermo Toral, and Noah Yarrow. Latin American 28. Rosser and Fahmi (2016). Development Forum Series. Washington, DC: World 29. Béteille (2015). Bank. 30. Burns, Köster, and Fuster (2016); Gambetta (1988). Bruns, Barbara, and Ben Ross Schneider. 2017. “Reforming 31. Gouldner (1960). Education Quality: Difficult Reforms Facing Disruptive 32. Bjork (2006). Opponents.” Paper presented at the Workshop on the 33. Béteille (2009). Political Economy of Education Reform, Mexico City, 34. Sethi (2015). February 3–4. 35. Segal (2005). Burns, Tracey, Florian Köster, and Marc Fuster. 2016. Educa- 36. Mathur (2010). tion Governance in Action: Lessons from Case Studies. Educa- 37. World Bank (2016). tional Research and Innovation Series. Paris: Centre for 196 | World Development Report 2018 Educational Research and Innovation, Organisation for Hoxby, Caroline Minter. 1996. “How Teachers’ Unions Affect Economic Co-operation and Development. Education Production.” Quarterly Journal of Economics Carnoy, Martin. 2007. Cuba’s Academic Advantage: Why Stu- 111 (3): 671–718. dents in Cuba Do Better in School. With Amber K. Gove and IEG (Independent Evaluation Group). 2015. World Bank Jeffery H. Marshall. Stanford, CA: Stanford University Support to Early Childhood Development: An Independent Press. Evaluation. What Works Series. Washington, DC: World CNN.com. 2015. “Bihar Cheating Scandal: What Parents in Bank. India Will Do for Good Grades.” March 20. http://www Jhingran, Dhir. 2016. “Data Collection Alone Cannot Improve .cnn.com/2015/03/20/asia/india-cheating-parents-school Learning Outcomes in State-Run Schools.” Hindustan -tests/. Times, November 25. http://www.hindustantimes.com De Herdt, Tom, Kristof Titeca, and Inge Wagemakers. 2010. /analysis/data-collection-alone-cannot-improve-learning “Making Investment in Education Part of the Peace Div- -outcomes-in-state-run-schools/story-xGSCZ8yXxMtEl idend in the DRC.” Paper presented at Chronic Poverty Q3qa8Cn4L.html. Research Centre conference, “Ten Years of War against Kingdon, Geeta, and Francis Teal. 2010. “Teacher Unions, Poverty,” Manchester, U.K., September 8–10. Teacher Pay, and Student Performance in India: A Pupil Dundar, Halil, Benoît Millot, Yevgeniya Savchenko, Harsha Fixed Effects Approach.” Journal of Development Economics Aturupane, and Tilkaratne A. Piyasiri. 2014. Building the 91 (2): 278–88. Skills for Economic Growth and Competitiveness in Sri Lanka. Levitt, Steven D., and Stephen J. Dubner. 2010. Freakonomics: Directions in Development: Human Development Series. Il Calcolo dell’Incalcolabile. Milan: Sperling and Kupfer. Washington, DC: World Bank. Magill, Clare. 2010. “Education and Fragility in Bosnia and Easton, David. 1979. A Framework for Political Analysis. Herzegovina.” Research Paper IIEP, Education in Emer- Chicago: University of Chicago Press. gencies and Reconstruction Series, International Insti- Eberts, Randall W., and Joe A. Stone. 1987. “Teacher Unions tute for Educational Planning, United Nations Educa- and the Productivity of Public Schools.” Industrial and tional, Scientific, and Cultural Organization, Paris. Labor Relations Review 40 (3): 354–63. Mathur, Nayanika. 2010. “Paper Tiger? The Everyday Life Estrada, Ricardo. 2016. “Crony Education: Teacher Hiring of the State in the Indian Himalaya.” PhD dissertation, and Rent Extraction.” Working paper, European Univer- University of Cambridge, Cambridge, U.K. sity Institute, San Domenico di Fiesole, Italy. Moe, Terry M. 2001. “Teachers Unions and the Public Forgy, Larry. 2009. “Per Student Financing in ECA School Schools.” In A Primer on America’s Schools, edited by Terry Systems.” Europe and Central Asia Knowledge Brief 6, M. Moe, 151–83. Stanford, CA: Hoover Institution Press. World Bank, Washington, DC. —— — —. 2011. Special Interest: Teachers Unions and America’s Gambetta, Diego. 1988. Trust: Making and Breaking Cooperative Public Schools. Washington, DC: Brookings Institution Relations. Oxford, U.K.: Blackwell Publishers. Press. Ghaus-Pasha, Aisha. 2011. “Fiscal Implications of the 18th Moe, Terry M., and Susanne Wiborg. 2017. The Comparative Amendment: The Outlook for Provincial Finances.” Politics of Education: Teachers Unions and Education Systems Policy Notes Series on Pakistan No. 1, World Bank, around the World. Cambridge, U.K.: Cambridge University Washington, DC. Press. Gouldner, Alvin Ward. 1960. “The Norm of Reciprocity: Mosselson, Jacqueline, Wendy Wheaton, and Paul St. John A Preliminary Statement.” American Sociological Review Frisoli. 2009. “Education and Fragility: A Synthesis of 25 (2): 161–78. the Literature.” Journal of Education for International Devel- Greif, Avner. 2006. Institutions and the Path to the Modern Econ- opment 4 (1): 1–17. omy: Lessons from Medieval Trade. Political Economy of Murillo, Maria Victoria. 1999. “Recovering Political Dynam- Institutions and Decisions Series. Cambridge, U.K.: ics: Teachers’ Unions and the Decentralization of Educa- Cambridge University Press. tion in Argentina and Mexico.” Latin American Politics and Grindle, Merilee Serrill. 2004. Despite the Odds: The Conten- Society 41 (1): 31–57. tious Politics of Education Reform. Princeton, NJ: Princeton —— — —. 2012. “Teachers Unions and Public Education.” University Press. Perspectives on Politics 10 (1): 134–36. Hindustan Times. 2015. “25 Mystery Deaths and 2,000 Arrests: North, Douglass C. 1991. “Institutions.” Journal of Economic All about MP’s Vyapam Scam.” June 29. http://www Perspectives 5 (1): 97–112. .hindustantimes.com/bhopal/25-mystery-deaths-and-2 Novelli, Mario, Sean Higgins, Mehmet Ugur, and Oscar Vali- -000-arrests-all-about-mp-s-vyapam-scam/story-Y3d ente. 2014. “The Political Economy of Education Systems LEQdkEsVyCwuPQXxaXI.html. in Conflict-Affected Contexts: A Rigorous Literature Hossain, Naomi, Mirza Hassan, Md Ashikur Rahman, Review.” U.K. Department for International Develop- Khondokar Shakhawat Ali, and M. Sajidul Islam. 2017. ment, London. “The Problem with Teachers: The Political Settlement Pepinsky, Thomas B., Jan H. Pierskalla, and Audrey Sacks. and Education Quality Reforms in Bangladesh.” ESID 2017. “Bureaucracy and Service Delivery.” Annual Review of Working Paper 86, Effective States and Inclusive Devel- Political Science 20 (1): 249–68. opment Research Centre, Global Development Institute, Rose, Pauline, and Martin Greeley. 2006. “Education in Frag- School of Environment, Education, and Development, ile States: Capturing Lessons and Identifying Good Prac- University of Manchester, Manchester, U.K. tice.” DAC Fragile States Group, Development Assistance Unhealthy politics drives misalignments | 197 Committee, Organisation for Economic Co-operation Shrestha, Unika. 2017. “Beyond Politics: The Role of Teachers’ and Development, Paris. Unions in Promoting Quality Education.” Background Rosser, Andrew J., and Mohamad Fahmi. 2016. “The Political note, World Bank, Washington, DC. Economy of Teacher Management in Decentralized Stevens, Mitchell, and Michael Kirst. 2015. Remaking College: Indonesia.” Policy Research Working Paper 7913, World The Changing Ecology of Higher Education. Stanford, CA: Bank, Washington, DC. Stanford University Press. Sandefur, Justin, and Amanda Glassman. 2015. “The Political Taylor, Nick, Johan Muller, and Penny Vinjevold. 2003. Get- Economy of Bad Data: Evidence from African Survey ting Schools Working: Research and Systemic School Reform in and Administrative Statistics.” Journal of Development South Africa. Cape Town: Pearson Education South Africa. Studies 51 (2): 116–32. Wiborg, Susanne. 2017. “Teachers Unions in the Nordic Segal, Lydia G. 2005. Battling Corruption in America’s Public Countries: Solidarity and the Politics of Self-Interest.” Schools. Cambridge, MA: Harvard University Press. The Comparative Politics of Education: Teachers Unions and Sethi, Aman. 2015. “The Mystery of India’s Deadly Exam Education Systems around the World, edited by Terry M. Scam.” Guardian, December 17. https://www.theguardian Moe and Susanne Wiborg, 144–91. Cambridge, U.K.: .com/world/2015/dec/17/the-mystery-of-indias-deadly Cambridge University Press. -exam-scam. Willis, Eliza, Christopher da CB Garman, and Stephan Sharma, Rashmi. 1999. “What Manner of Teacher: Some Haggard. 1999. “The Politics of Decentralization in Latin Lessons from Madhya Pradesh.” Economic and Political America.” Latin American Research Review 34 (1): 7–56. Weekly 34 (25): 1597–1607. World Bank. 2007. “Spending for Development: Making Sharma, Rashmi, and Vimala Ramachandran, eds. 2009. the Most of Indonesia’s New Opportunities; Indonesia The Elementary Education System in India: Exploring Institu- Public Expenditure Review 2007.” World Bank, Washing- tional Structures, Processes, and Dynamics. New Delhi: ton, DC. Routledge. —— ——. 2016. Making Politics Work for Development: Harnessing Shields, Robin, and Jeremy Rappleye. 2008. “Differentiation, Transparency and Citizen Engagement. Policy Research Development, (Dis)Integration: Education in Nepal’s Report Series. Washington, DC: World Bank. ‘People’s War.’” Research in Comparative and International —— ——. 2017. World Development Report 2017: Governance and the Education 3 (1): 91–102. Law. Washington, DC: World Bank. 198 | World Development Report 2018 How to escape low-learning traps 11 Tackling the technical and political constraints that misalign education systems requires action on three fronts: investing in better information on learning; mobilizing coalitions for learning; and adopting a more iterative, adaptive approach to change. Since 1995, England has substantially improved the Reforms that improve learning rely on good literacy and numeracy skills of primary schoolchildren strategies—both political and technical. This chapter using good political strategy and sound technical solu- draws lessons from various experiences to identify tions.1 As a result, the proportion of students in grade 4 how opportunities for reform emerge and how poli- reaching the intermediate benchmark in the Trends in ticians, bureaucrats, parents, and students can seize International Mathematics and Science Study (TIMSS) them. It focuses on three entry points for addressing assessment of mathematics shot up from 54 percent systemic political and technical challenges: improving in 1995 to 80 percent in 2015 (figure 11.1)—an achieve- information, building coalitions and strengthening ment matched by few other countries.2 Poor education incentives, and encouraging innovation and agility. outcomes had become an important issue in the 1997 Most countries need all three. national elections, and the new government responded with a national strategy at the start of its term in 1998.3 At the heart of the reforms was a redesign of how Improving information teachers taught. The new strategy set clear targets for Addressing weaknesses in education systems is dif- the country, as well as for individual schools, based ficult when accurate, usable information on learning on regular, publicly available data on student achieve- is lacking. Without it, stakeholders cannot hold pol- ment. The targets provided incentives for local educa- iticians and bureaucrats accountable, assess system tion authorities, teachers, and principals. The govern- performance, or design effective policies to improve ment adjusted school inspections to reflect the new learning. Though it might not be enough on its own, curriculum; it also strengthened the links between better information on learning can provide the sub- teacher performance and pay. A revamped profes- stance needed for better political strategies and the sional development program, supported by local liter- evidence base needed for effective policies. acy consultants, helped teachers implement the new strategy. Local governments received substantial new Information can increase political funding for implementation. Literacy and numeracy incentives to improve learning “hours,” introduced as part of the new strategy, sig- The absence of information on learning can weaken nificantly improved early learning outcomes.4 The the political incentives to provide good public ser- program has continued to evolve, with more support vices. Targeted programs or even direct vote buying focused on disadvantaged learners. are sometimes exchanged for political support, How to escape low-learning traps | 199 Figure 11.1 Primary school numeracy has increased dramatically in England TIMSS mathematics scores for grade 4 students, and share of students reaching the intermediate benchmark in TIMSS mathematics assessment a. Average scores for grade 4 students b. Students above and below the intermediate benchmark, England 540 50 40 520 30 Percent Score 500 20 480 10 460 0 1995 2000 2005 2010 2015 1995 2000 2005 2010 2015 England Above intermediate benchmark Participating OECD countries Below intermediate benchmark All participating countries Source: WDR 2018 team, using data from Trends in International Mathematics and Science Study (TIMSS), 1995–2015 (https://timssandpirls.bc.edu/). Data at http://bit.do/WDR2018-Fig_11-1. Note: Students at the intermediate level are able to apply basic mathematical knowledge in straightforward situations; demonstrate an understanding of whole numbers and some understanding of fractions; visualize three-dimensional shapes from two-dimensional representations; and interpret bar graphs, pictographs, and tables to solve simple problems. resulting in poor service delivery.5 Better information increased competition between schools. As a result, can encourage voters to elect politicians who deliver learning outcomes improved in both public and pri- results.6 For example, using a metric that combines vate schools, and private school fees were cut.8 Par- student passing rates with test scores, the federal ents can also use information to pressure schools to government in Brazil sets credible education targets raise standards.9 For example, the provision of report that are widely scrutinized (box 11.1). Meeting these cards has strengthened accountability in some coun- targets increases the chances of an incumbent poli- tries.10 Interventions of this kind work best where tician being reelected and of bureaucrats keeping power relations between actors in an education sys- their jobs.7 This example also highlights the value tem are not highly unequal or organized to support of providing information on learning for areas that patronage networks, and where frontline service correspond with political jurisdictions; because of providers have autonomy to respond to community the overlap, citizens can hold politicians accountable demands.11 When these factors prevent parents’ voices for progress on education targets. But whether infor- from being heard, it can encourage some, especially mation can shift incentives toward a greater focus on middle-class parents, to opt out of the public educa- learning depends on the broader context. For exam- tion system, weakening pressure on governments to ple, better information in just one sector is unlikely improve learning across the system.12 to disrupt patronage networks in countries where Information can also help ensure that resources clientelism is entrenched across the political system. go where they are intended. In the mid-1990s, schools in Uganda received only around a quarter of their Information can also improve incentives in intended per student grant allocations. The govern- schools ment began to publish information on the timing Information on school performance can make local and amount of transfers made to districts for school education systems work better. In many developing capitation grants so that schools could monitor local countries, parents have limited information on the administrators. This move increased the share of grant quality of their local schools. In Pakistan, providing funding reaching schools by reducing capture of fund- parents with information on learning outcomes ing by district offices. Consistent with the feedback 200 | World Development Report 2018 Box 11.1 Using information to align incentives with learning in Brazil From 2000 to 2012, Brazil’s learning outcomes on the region in Brazil. Targets based on this index are used by Programme for International Student Assessment (PISA) system administrators at every level, as well as by parents, showed steady improvement, with gains in some sub- to hold schools and local administrations accountable for jects concentrated among poorer-performing students. learning. Underlying this progress were reforms that strengthened Better information also raised the incentives for pol- accountability for system performance, reduced funding iticians to improve performance. Public awareness of the inequalities across Brazil’s diverse regions, and provided index is high, with the biannual release of IDEB scores cash transfers to the neediest families. Improvements in generating extensive media coverage and debate. This not information underpinned these reforms. only places education quality high on the political agenda, Better information made it much easier to hold edu- but also makes it an important factor when citizens choose cation agencies accountable for learning. A state-level their local representatives. learning assessment introduced in 1995 was extended 10 Crucially, the government also uses the index to target years later to cover all fourth- and eighth-grade students. low-performing schools for additional support and intro- The central government combined assessment results with duce programs to motivate system actors. For example, student promotion rates to create an index of basic edu- schools receive bonuses based on annual improvements in cation quality (Índice de Desenvolvimento da Educação IDEB scores, and evidence suggests this move has contrib- Básica, IDEB) for every school, municipality, state, and uted to better learning. Sources: WDR 2018 team, based on Bruns, Evans, and Luque (2011); Ferraz and Bruns (2012); OECD (2016); Toral (2016). loop described in the next section, schools in areas Oportunidades, Mexico’s conditional cash transfer with better access to newspapers benefited the most.13 program, has endured since 1997 despite political and economic changes. Because they provided solid Good information is also vital for evidence of how the program improved the lives of monitoring, evaluating, and guiding children, impact evaluations were key to the decision systems to continue the program after a new government was System managers need information to monitor and elected in 2000.16 analyze system performance. School supervisors need information on student learning outcomes But many information and knowledge to identify and address poorly performing schools. systems are not serving these purposes Good research and evaluation on programs and poli- Information needed to improve learning is lacking in cies aimed at improving learning can support better many countries. An assessment of capacity to mon- implementation by enabling feedback loops. In the itor progress toward the Sustainable Development early 2000s, Cambodia’s scholarship program sought Goals found that, of 121 countries, a third lacked data to improve learning outcomes for disadvantaged on learning outcomes at the end of primary school, students. An early evaluation of the program found and half had insufficient information on learning at that it improved attainment and narrowed gender the end of lower secondary school.17 Even fewer have gaps in enrollment, but it failed to reach the poorest the data to track these learning outcomes over time. children or improve learning.14 In 2006, as a result of Information systems in the education sector, which these findings, the government improved the targeting are often weak, are rarely used for decision making, of poorer children. It then experimented with using planning, or implementation. the scholarships to encourage learning. Introducing There are many barriers to using information merit-based criteria into student selection increased to improve learning outcomes. In Tanzania, widely enrollment and improved learning, raising mathemat- publicized results from citizen-led learning assess- ics test scores by about 0.17 standard deviations.15 ments influenced public perceptions of education Research and evaluation can also build sup- and shifted the government’s focus toward learning port for effective programs across political cycles. (box 11.2). Yet such direct links between evidence and How to escape low-learning traps | 201 Box 11.2 Citizen-led assessments have raised awareness of the learning crisis in South Asia and Sub-Saharan Africa Citizen-led learning assessments are locally designed mea- The initiatives successfully disseminated their results •  surements of basic reading and mathematics competen- and raised awareness about the learning crisis. They also cies. Typically conducted by networks of civil society orga- increased the focus on learning in government planning nizations, these assessments test children whether they are documents. in or out of school—something that conventional testing In India, partnerships between some state governments •  cannot do. Their goal is to increase awareness of learning and Pratham, an NGO that seeks to improve education outcomes and to encourage stakeholders to take action to quality, have designed interventions to address the prob- improve learning. Citizen-led assessments have been con- lems identified by the Annual Status of Education Report ducted mainly in South Asia and Sub-Saharan Africa. For (ASER) assessment. Moreover, the government of India example, the Campaign for Popular Education (CAMPE)— now holds its National Achievement Survey annually a network of over 1,000 nongovernmental organizations (rather than once every three years) to track learning (NGOs), researchers, and educators in Bangladesh—began more frequently. carrying out assessments of this kind in 1999. Evaluations of these initiatives concluded that: While the assessment results have led to action in some cases, the link to improved learning is not automatic. •  The public finds these assessments more salient than Over the short period that the ASER in India and Uwezoa larger-scale, more complex national assessments, because in Tanzania have been operating, their assessment results the citizen-led assessments focus on a narrower set of do not show any clear overall pattern of increases in basic competencies, starting with recognizing letters and learning—although some Indian states showed significant numbers. improvements between 2010 and 2016. Sources: WDR 2018 team, based on Chowdhury, Choudhury, and Nath (1999); Rath and others (2015); R4D (2015). a. Uwezo means “capability” in Kiswahili. policy making are often missing.18 Some evaluations of particular groups at the expense of better public take too long to inform decision making; others services. In Argentina between 1998 and 2007, news- fail to track key drivers of low system performance. papers that received government funding published Even where usable information exists, government fewer reports on corruption than did others.20 agencies may lack the incentives or capacity to use What are the characteristics of an information it well.19 Independence also matters: reliable, salient system that promotes learning (table 11.1)? First, infor- information can provide incentives for better perfor- mation needs to be credible, politically salient, and mance, but biased media may protect the interests publicly available. Second, clear targets for progress Table 11.1  Principles for making the most of information and the roles that actors can play Principles for making the best use of information Roles that different actors can play Provide regular, credible, politically salient, and •  •  Government institutions: Produce and disseminate publicly available information on learning. national assessment results; conduct in-house •  Set clear targets or expectations for learning, so there evaluations; support education research and is a benchmark for judging performance. evaluation in external research institutes. •  Align information with the political and administrative Civil society and private sector: Produce and •  jurisdictions that have authority to act. disseminate citizen-led learning assessments; use assessments and research to support interventions Build information systems that are responsive to the •  that improve learning. policy cycle and facilitate decision making. Source: WDR 2018 team. 202 | World Development Report 2018 on learning can strengthen incentives by providing policy changes through wide-ranging consultations measures of system performance. Third, meaningful that try to bring together key interest groups.22 Peru’s information on learning needs to be aligned with Business Association for Education organized an political or decision-making power, so that the public information campaign that helped shift public can hold education decision makers more account- opinion to support reforms that began in 2006. Gov- able. Finally, information needs to be usable by policy ernment reformers used information on the poor makers, administrators, and other system actors— learning outcomes of the education system to mobi- that is, it must be timely, accurate, policy relevant, and lize public support for efforts to strengthen teacher sensitive to the policy cycle. accountability, which led to sustained improvements in learning.23 Alliances between education stakehold- Building coalitions and ers have also formed in some countries to realize the right to education through the legal system (box 11.3). strengthening incentives Though mobilization efforts can be successful at Education systems are made up of many actors who rebalancing interests, they may be less successful at pursue interests that do not always align with learn- shifting the interests of those opposed to reforms. ing. Addressing this requires action on two fronts. Education reform is a long process, and well-organized First, coalitions of interest groups are needed to build opposition can derail it, particularly during implemen- a consensus around the actions that will strengthen tation. In Peru, the government successfully mobilized accountability for better learning. This often requires public support to get reforms approved, but it was less mobilizing support from groups that are not actively successful at getting buy-in from teachers, which led involved in agenda-setting or that do not engage with to continued resistance from teachers’ unions during others. Second, the incentives of bureaucrats and implementation. While the broad reform direction other system actors need to align more closely with remained intact and learning improved, this experi- learning (table 11.2). ence highlights a potential trade-off between man- aging the politics of reform and getting implemen- Mobilizing support and building coalitions tation right. When reformers have to devote effort to to improve learning managing opposition, that effort can divert attention System actors have a better chance of enacting from implementing reforms well. Lack of buy-in from reforms when they act collectively. Some actors have important groups deters them from contributing to more power to shift policy toward learning, in part policy design or implementation, thereby undermin- because they are better organized.21 For example, in ing the sustainability of the reform.24 many countries teachers’ unions have a powerful Building broad-based coalitions of stakeholders is voice in debates on reform, whereas the collective important at all stages of the policy cycle. Malaysia voice of parents and students is often muted. created a performance delivery unit to spearhead Mobilizing support and building coalitions of a comprehensive reforms in many sectors, including range of system actors have helped to improve learn- education. The unit uses “labs” that build coalitions of ing. Many countries have built support for proposed stakeholders and involve them in all stages of reform, Table 11.2 Principles for building effective coalitions and the roles that actors can play Principles for building effective coalitions Roles that different actors can play Mobilize support for reforms through clear •  Government institutions: Develop open, inclusive •  articulation of the problems of low learning. spaces to discuss reform and identify technically and •  Develop a political strategy to mobilize support and politically feasible solutions; build the appropriate build long-term coalitions for learning. institutional capacity. •  Avoid direct confrontation in favor of negotiation and Civil society and business organizations: Advocate •  compensation where possible. for better education systems; support community and parent action at all levels to improve outcomes. Encourage strong partnerships between schools and •  communities. Teachers and unions: Advocate for system •  improvements; use system knowledge to engage in Strengthen the capabilities of organizations •  debates on reform. responsible for education services. Source: WDR 2018 team. How to escape low-learning traps | 203 Box 11.3 Using the legal system to press for change With more than 80 percent of national constitutions rec- • The extent to which the legal system has been used to ognizing the right to education, courts have become an press for policy changes depends significantly on the increasingly important arena for holding governments nature of the court system, the presence of support accountable for education policies and practices. structures for legal mobilization, and the ideology of the In recent years, India and Indonesia have seen a signif- courts. icant increase in education rights litigation. In India, this • Using education rights litigation effectively is condi- trend has been driven by the adoption in 2009 of the land- tional on judges who are open to such cases; civil society mark Right of Children to Free and Compulsory Education groups that can help citizens press their claims; and Act. Cases have included demands to ensure equal access broader political mobilization. to education, the fulfillment of minimum service standards, • Policy-oriented litigation has mainly served the interests and assurance that governments will fulfill their spending of poor or marginalized groups, even though sections obligations. Many of these cases have been successful. The of the middle class have been centrally involved in Indian Supreme Court has consistently ruled in favor of much of the litigation. Gains have largely come through upholding quotas for poor children in private schools. The better access to education, although successes have High Court in Uttarakhand required the state government often been at the expense of quality education for the to adopt minimum qualification standards for teachers. middle class. And in Indonesia, parents succeeded in enforcing consti- • Litigation as a strategy for improving learning outcomes tutional provisions that obligated the government to spend has its limitations. Often, judgments need to be enforced 20 percent of its budget on education. by the same public officials who were the target of the These cases have often been brought by individuals or initial lawsuit. Even when judgments are implemented, small groups, with nongovernmental organization (NGO) they are more often about ensuring access than improv- activists and teachers’ unions providing technical and ing learning. Courts typically lack the necessary expertise financial support. An assessment of the impact of litigation on learning, especially where information on learning of this kind in India and Indonesia found the following: outcomes is scarce. Source: WDR 2018 team, based on Rosser and Joshi (2017). from design to implementation.25 Stakeholders typi- At the outset, efforts to build a supporting coalition cally come together in the labs for six to nine weeks were only half-hearted, and despite the learning gains, at the start of reforms to discuss priorities, agree on the reforms have remained unpopular. The election performance indicators, and produce implementa- of a new government in 2015 led to heated debate tion plans. During implementation, minilabs bring on whether to scrap key elements of the original stakeholders together to adjust plans. Programs intro- reforms.27 Building a coalition may require better com- duced under the process are credited with increasing munication strategies—or it may require changing grade 3 literacy rates in Malaysia from 89 percent in the reform design, to one that is second-best techni- 2009 to close to 100 percent in 2012. The approach has cally but easier to implement and sell to stakeholders. been exported to other countries, including India, A gradual, negotiated approach to reform may South Africa, and Tanzania (box 11.4). work better than confrontation. Where coalitions Without efforts to build coalitions for learning, of system actors foster collaboration around shared reforms are less likely to endure. Even if evidence goals, reforms are more likely to succeed. The history shows that the reforms improve learning, their sus- of reforms to improve teaching in Chile demonstrates tainability is at risk when they are misunderstood or how gradual, negotiated reforms can build strong unpopular among system actors. In Poland, large-scale coalitions for change (box 11.5). Since Chile’s return changes in the structure of the education system were to democracy, successive governments have adjusted introduced in 1999 as part of broader decentraliza- the working conditions of teachers to improve their tion reforms. These reforms have been credited with welfare, while also linking pay and career develop- improving student learning outcomes significantly.26 ment more closely to performance. These changes 204 | World Development Report 2018 Box 11.4 Using “labs” to build coalitions for learning Rapidly deteriorating results on school-leaving examina- The lab process made it possible to introduce a com- tions, together with other newly available information plex package of politically sensitive reforms. For example, on poor system performance, motivated policy makers the government introduced monetary and nonmonetary in Tanzania to launch the ambitious Big Results Now in incentives to reward the most improved schools, along Education (BRN) program in 2013. The BRN adopted a with accountability measures that used public examination “service delivery” approach that was first introduced in results to rank schools. The BRN also introduced, for the the United Kingdom in the early 1990s and then adapted first time, a national sample-based assessment to measure successfully in Malaysia in 2009. early grade literacy and numeracy. Communication cam- At the heart of the approach was a six-week-long “lab” paigns succeeded in generating very high levels of public to identify priority reform areas and develop mutually awareness of the BRN’s objectives nationwide. agreed-on delivery plans. The lab brought together all Although the program has been running for only four the key system actors—government officials, academ- years, there are signs that it has begun to improve learning ics, teachers’ unions, development partners, civil society outcomes. However, the program has not been without its organizations—at a level senior enough to ensure follow- difficulties; for example, a recent review highlighted the through. Together, the lab participants drafted nine key difficulties in coordination between the government agen- initiatives, developed step-by-step implementation plans, cies responsible for education. But over the past few years, and assigned responsibilities for those steps. examination results have slowly improved, and primary school students have made gains in early grade reading. Sources: WDR 2018 team, based on Sabarwal, Joshi, and Blackmon (2017); Todd and Attfield (2017); World Bank (2017b). Box 11.5 Reformers in Chile negotiated changes gradually In the early 2000s Chile’s education system registered sig- Figure B11.5.1 Reading scores have nificant, sustained improvements in learning levels. The pro- improved in Chile portion of 15-year-olds who achieved reading scores at or PISA reading scores above a Programme for International Student Assessment (PISA) level of proficiency increased from 52 percent to 500 69 percent between 2000 and 2015 (figure B11.5.1). Much of the improvement was attributable to the 480 Sistema Nacional de Evaluación de Desempeño (National 460 Performance Evaluation System; SNED) program imple- Score mented in 1996. This program began by awarding teacher 440 bonuses based on school-level indicators of performance. In 2004 individual teacher incentives were introduced, 420 based on mandatory performance evaluations of public school teachers. By the end of the 2000s, these incentives 400 accounted for 15–25 percent of the average teacher salary. 2000 2005 2010 2015 Rigorous evaluations of the group-based program revealed OECD average Chile that the incentives significantly improved student learning. Source: WDR 2018 team, using data from the Programme for International The gradual shift from school to individual incentive Student Assessment (PISA) (www.oecd.org/pisa). Data at http://bit.do payments was a pragmatic attempt to address the potential /WDR2018-Fig_B11-5-1. (Box continues next page) How to escape low-learning traps | 205 Box 11.5 Reformers in Chile negotiated changes gradually (continued) opposition of teachers’ unions to performance-related pay. A final factor in the successful adoption of these reforms Before implementing a mandatory program for all teachers, was their inclusion in a broader set of reforms that increased the administration introduced a voluntary individual assess- resources for education and raised teachers’ salaries. SNED ment and incentive system that set a precedent for teacher became part of the teacher professionalism pillar of the Full evaluation. Because these steps allowed time to adjust and School Day reform package. More teachers were covered gain support for the new system, they were key to its success. by the reforms, and the incentive amount was increased. Establishing credibility with the teachers’ union early Salary increases before the start of the program may have on was another key strategy. The Teacher Statute passed in helped to lessen opposition to the mandatory individual 1991 conferred civil service status on teachers, guarantee- pay incentive. ing associated job benefits, protection, and an opportunity As a consequence, the Chilean programs remain one of for centralized wage negotiations. This move sent a positive the few long-running “pay for performance”–type reforms signal to teachers. Trust between the union and the govern- that have been successfully scaled to the national level. In ment increased further through regular discussions on the other contexts, such reforms have often been unpopular, implementation of reforms. As part of these efforts, union but in Chile the reforms continue: in 2016 new legislation members codesigned the performance evaluations used for passed to widen the coverage of the incentive program, the incentive program. while strengthening teacher professional development. Sources: WDR 2018 team, based on Avalos and Assael (2006); Contreras and Rau (2012); Delannoy (2000); Mizala and Schneider (2014); OECD (2016); World Bank (2017a). have contributed to Chile’s steady improvement in but they had a major impact on public spending.30 international learning assessments. By 2011, with less than a third of teachers certified, Negotiations can also include strategies to com- 9 percent of the education budget already went to pensate actors disadvantaged by reform. One such certification allowances.31 strategy is to provide targeted assistance to students harmed by reforms to improve system efficiency. Building partnerships between schools and Additional services for children affected by school clo- communities sures, for example, can ease school consolidations.28 Sustained reform requires strong partnerships Another strategy is to use “dual-track” reforms to pro- between schools and communities. Where incentives tect some incumbents from the negative impacts of for systemwide reform are weak, local action can reforms. For example, pay-for-performance programs substitute. In South Africa, the political and economic in Peru and in the District of Columbia in the United context has constrained efforts to improve education States were initially introduced voluntarily. performance in some provinces, but local progress has Compensating perceived losers can help get been made possible in some schools through strong reforms approved, but that approach comes with partnerships between parents and schools.32 Local risks. In 2005 the Indonesian government introduced partnerships are particularly important in fragile a comprehensive reform program aimed at raising and conflict-affected areas.33 For example, a program the competencies of teachers. Teacher certification that built community-based schools in Afghanistan was the centerpiece of the reforms, with teachers reduced the distance to school, increased enrollment, required to pass a competency test to continue teach- and improved learning outcomes, particularly for ing.29 In exchange for these new obligations, the nego- girls.34 Yet these local partnerships tend to work best tiated agreement provided certified teachers with an when supported by responsive higher-level institu- additional monthly allowance as large as their base tions, which are sorely lacking in fragile environments. salary. But early in implementation, the requirements for certification were diluted because of political pres- Aligning the incentives and capacity of sures, so that teachers were no longer required to pass system actors with learning a competency test. In the end, the reforms had little The success of reforms depends on the ability, incen- impact on teacher competencies or student learning, tives, and motivations of public officials. Managing 206 | World Development Report 2018 education systems effectively requires competent learning. Some parts of the solution to low learning public service–oriented personnel, which in turn are relatively straightforward. Inadequate infrastruc- means commensurate pay and working conditions.35 ture and learning materials, while logistically chal- But if the political economy of education is misaligned lenging, can be addressed directly: the technologies with public goals, candidates with less desirable attri-needed are well known, and most education systems butes may be attracted to public service. In Mexico, have enough experience solving these issues. But teachers were often hired based on political patronage improving what happens in the classroom is much rather than merit, which resulted in lower-quality harder. It involves changing student and teacher hires compared with those in test-based systems.36 behavior, as well as supporting teachers in efforts to Efforts to build the capacity of bureaucracies have tailor their teaching to the needs of their students. been disappointing.37 Even where individual capac- The traditional approaches to reform—in which pre- ity is built successfully, the incentives to use this defined interventions are introduced with little room capacity to develop and implement effective policies to adapt during implementation—are rarely effective. are often absent.38 Put another way, building organi- Learning reforms need a more agile approach, zational capability to improve education outcomes with room for adaptation.40 This is not the same as tends to work best when incentives in education sys- experimenting with different interventions in pilot tems are aligned with the same goals. For example, projects. Rather, it means testing approaches at scale where politicians face stronger incentives to provide in their political and economic contexts and using public goods, this has inspired efforts to build pro- the existing capabilities of implementing agencies. fessional bureaucracies that can deliver better public A recent review of complex public management services.39 reforms, including in education, highlighted the key elements of successful reforms.41 Those reforms started out with a clear articulation of the problem, Encouraging innovation together with an initial set of potential solutions, and and agility then adopted solutions that emerged from experi- FIGURE 11.2 mentation during implementation (figure 11.2). Final Political and technical complexities make it chal- interventions tended to be hybrids, drawing on local lenging to design and implement policies to improve and global evidence. Figure 11.2 Problem-driven iterative adaptation drives successful reforms 1 Define and diagnose problems 1 Redefine and rediagnose problems 1 Redefine and rediagnose problems Repeat IMPROVED 2 2 2 LEARNING 5 Design Redesign Redesign options options options Adapt 3 Implement 4 Evaluate 3 Implement 4 Evaluate 3 Implement 4 Evaluate Source: Adapted from Andrews, Pritchett, and Woolcock (2017). How to escape low-learning traps | 207 Searching for solutions to local problems improve learning. In India, an experiment showed All systems have some parts that work well; these that grouping children by ability and using level- parts can be used to identify technically and polit- appropriate teaching along with continual assess- ically feasible approaches to improving learning. ment improved students’ reading abilities. Recogniz- In Misiones Province, Argentina, student dropout ing that a small-scale experiment was no guarantee of rates were high. But some schools bucked the trend: success in the government system, Pratham—the teachers agreed on informal learning contracts with NGO responsible for the original evaluation—experi- parents instead of blaming them for poor student mented with different approaches to level-appropriate performance. Schools that adopted more constructive teaching in government schools. This experimenta- approaches to parent-teacher relations saw dropout tion tested the assumptions of the original model and fall significantly.42 Schools approach challenges in identified factors behind the earlier success. It then different ways, so analysis of positive outliers could identified two approaches to implementation that be useful for policy making (box 11.6). could work at scale.43 Even in fragile states, where Local innovations, however, may not be enough to system capabilities are limited, iterative approaches close the learning gap between countries. Employing like this have been successful at restoring essential principles from the growing global knowledge can education services (box 11.7). provide useful ideas for improving learning in spe- Policy makers can test policies before introducing cific contexts. A more iterative approach to system them more widely. Whole-system reforms are difficult change can be a way to adapt interventions inspired to evaluate because they lack an appropriate counter- by global experiences to local contexts. factual, making it difficult to trace the impacts of policy change and adapt strategies to improve learning. Small Integrating an iterative and adaptive pilots can overcome these difficulties, but it is hard to approach to policy making and assess whether they will be effective without the atten- implementation tion and nurturing that can occur in a pilot. As a middle Recent examples show how an iterative, adaptive way, China and other countries have tested new poli- approach can strengthen education systems and cies in specific regions.44 Policy makers first identify Box 11.6 High-performing schools in the West Bank and Gaza offer some learning lessons The United Nations Relief and Works Agency (UNRWA) Although preservice training is similar in UNRWA and provides over 300,000 refugees in the West Bank and public schools, UNRWA teachers complete a two-year Gaza with basic education services. In multiple rounds of training program in classroom instruction, resulting international assessments, UNRWA schools outperformed in teaching approaches that are better aligned with public schools, delivering the equivalent of one year’s addi- learning. tional learning despite the lower socioeconomic status of Assessment and evaluation. UNRWA schools have •  UNRWA students and lower per student spending.a Drivers more rigorous, more frequent student assessments and of their better performance include: teacher evaluations than public schools. Effective school leadership. UNRWA invests in develop- •  • G  reater parental involvement in school activities and a ing qualified principals who can support their teachers close partnership between schools, households, and ref- effectively. ugee communities, which contributes to a shared sense of purpose and collaborative mechanisms for monitoring Identifying lessons from high-performing schools is not and support. always easy. Some factors such as school leadership that • M  ore effective teacher support systems. Teachers are drive high performance may be idiosyncratic, making them trained using standards that clearly articulate what hard to replicate. Drawing on large samples of schools can students should know and be able to do in each grade. help identify more generalizable lessons. Source: WDR 2018 team, based on Abdul-Hamid and others (2016). a. This comparison is for UNRWA schools and public schools in Jordan. 208 | World Development Report 2018 Box 11.7 Burundi improved education services by iterating and adapting After a protracted civil war and long peace process in the education system, including provincial education Burundi, a new government and new constitution in directors and parent-teacher associations. 2005 led to a renewed emphasis on public services. Many • Implementation. Senior government officials gave the schools had been destroyed, and management systems team authority to implement its new approach in a sin- had collapsed. As the new government took office, primary gle province. As implementation progressed, the team net enrollment rates stood at just 56 percent, student- regularly adjusted its action plan. classroom ratios were 87:1, and 20 students shared a single • Planning for sustainability. After reviewing the interven- mathematics textbook on average. tion’s performance, senior government officials decided The government prioritized reducing the high student- how to scale up the program to other provinces. textbook ratios and delays in delivery as part of a broader rapid-results initiative that had three stages: The initiative far exceeded its targets. Textbook avail- ability increased, and average delivery times fell from over • Shaping. In this stage, a reform team identified why a year to 60 days. This success led to similar initiatives to there were not enough textbooks. To ensure practical tackle teacher payroll problems, as well as many other ser- solutions, the team comprised stakeholders from across vice delivery problems beyond education. Source: WDR 2018 team, based on Campos, Randrianarivelo, and Winning (2015). the main problems; then they agree on which solutions outcomes are more likely to see greater innovation to subject to experimentation. They develop proposals across the education system (table 11.3).46 for experiments, in part by analyzing solutions adopted in other countries to tackle similar issues, with differ- Good information systems and ent regions trying alternatives. Successful policies broad-based coalitions are also needed are then rolled out to other regions. Belgium and the A capacity to learn from the implementation of new Netherlands have adopted similar approaches.45 innovations is vital. Information systems that pro- Giving stakeholders the authority and autonomy vide rapid, regular, accurate feedback are crucial for to adopt such approaches runs counter to how many more adaptive approaches to improving learning. education agencies operate. Closed systems limit the Some countries are beginning to build these kinds autonomy of system actors and judge performance of capabilities into their education agencies. Peru’s based on compliance with formal rules over resource MineduLAB in the Ministry of Education is a collabo- use, leaving little room for innovation. By contrast, ration between government agencies and experienced more open systems that have a sharper focus on researchers. 47 The lab introduces innovations directly Table 11.3 Principles for encouraging innovation at scale and the roles that actors can play Principles for encouraging innovation and agility in approaches to improving learning Roles that different system actors can play Adopt a more iterative and adaptive approach to the •  •  Government institutions: Develop an enabling design and implementation of policies. environment and incentives for innovation and a •  Identify promising solutions from within the education more iterative approach. system, as well as the global knowledge base. Civil society and private sector providers: Experiment •  •  Establish information systems that provide rapid with different approaches to improving learning. feedback to support implementation. Develop the capability of education agencies, an •  enabling environment, and autonomy to encourage innovation. Source: WDR 2018 team. How to escape low-learning traps | 209 into government schools, and information from min- opportunities to improve broad-based learning. istry systems (rather than individual data collection This context changes infrequently, but when it does exercises) must be used by researchers to evaluate the change it provides opportunities for significant new programs. Results must also be available within changes in education policy. During the martial the same academic year. In MineduLAB’s first year, law period of the 1970s in the Philippines, govern- innovations included providing more comparative ment spending on education fell below 2 percent information on school performance and introducing of the gross domestic product (GDP). In the 1980s, modules to encourage primary school students to the People’s Power Revolution restored democratic adopt a growth mindset. The program is still new, but rule, ushering in a new government that was more its approach is promising. responsive to demands for broader access to edu­ To be sustainable, these approaches need broad sup- cation. Trade liberalization increased the demand port. Though this iterative approach can help in devel- for skilled workers, further raising the incentives oping more effective strategies, it comes with risks for better education. With these societal shifts, for actors in education systems. Politicians can incur public investment in education increased by 2 significant costs if experiments fail or divert resources percentage points of GDP between 1980 and 2000 away from more traditional activities. Students can (figure 11.3). also suffer if new approaches disrupt their schooling Critical junctures often arise from broader decen- without improving it. Yet some risk-taking is vital if tralization and reform efforts, as in the education education systems are to improve learning. Mobilizing reforms in Latin America during the 1990s.48 Beyond stakeholder support and providing space for consulta- shifting responsibility for education services to local tions from the outset can reduce the risks. governments and schools, decentralization can pro- vide opportunities to better align important elements Education systems need to be agile to of education systems. After early decentralization exploit critical moments reforms in Poland, the government introduced Politicians and education system managers also formula-based funding mechanisms to link school need to respond quickly when changes create funding levels more closely to school needs. This Figure 11.3 Trends in public education spending in the Philippines track changes in the broader political and economic context Public education spending as percentage of GDP, and measures of democracy and trade openness, the Philippines (1960–2000) a. Spending and governance b. Spending and trade openness Public education spending (% of GDP) Public education spending (% of GDP) 10 4 4 70 5 Level of democracy 3 3 60 Trade openness 0 2 2 50 −5 1 −10 1 40 1960 1970 1980 1990 2000 1960 1970 1980 1990 2000 Public education spending (% of GDP) Public education spending (% of GDP) Level of democracy Trade openness Source: Ansell (2006). Adapted with permission from Ben W. Ansell; further permission required for reuse. Data at http://bit.do/WDR2018-Fig_11-3. Note: Level of democracy is measured by the polity score, which consists of an evaluation of the competitiveness and openness of elections, the nature of political participation in general, and the extent of checks on executive authority. A high positive score corresponds to strong democratic institutions; negative scores indicate more autocratic systems. Trade openness is measured by the inverted Hiscox Kastner score, which gauges the degree to which a country deviates from an optimal level of imports from a hypothetical protection-free environment. Higher scores indicate greater openness. 210 | World Development Report 2018 shift aligned funding with new realities, helping the also supporting their participation in international system reduce inefficiencies.49 assessments.52 To innovate effectively—as indeed to build coa- Beyond support to measure learning, external litions and use information for reform—education actors can also help build global knowledge on ways systems need strong, competent leadership. Research to diagnose system weaknesses and improve learn- highlights three key attributes of effective leaders. ing. This knowledge base has expanded rapidly, but First, they can clearly articulate problems and pres- more research is needed on how to adapt promising ent clear visions for how to tackle them. Second, interventions to specific contexts. External actors can they mobilize human and financial resources around fund research and encourage collaboration among agreed-on goals and build coalitions to advocate for practitioners, researchers, and government institu- change and support implementation. Finally, effec- tions to build capacity and locally relevant knowledge tive leaders focus on identifying solutions that fit the on effective ways to improve learning. institutional context.50 Encourage flexibility and support reform coalitions How can external actors External actors can also encourage inclusive reforms support initiatives to through project development activities, policy dis- improve learning? cussions, and support to other system actors. Though there has been much progress on the aid effective- Support the creation of objective, ness agenda first agreed on in the Paris Declaration politically salient information in 2005, there is still room for improvement. A key Global education initiatives can improve political aspect of this agenda is building inclusive reforms. incentives for action. The Millennium Develop- But progress in this area has been slow. Across all sec- ment Goals (MDGs) were ssuccessful at mobilizing tors, only about half of countries were judged to have international and domestic actors on development systems for meaningful dialogue with civil society challenges. Though the global impact of the MDGs— organizations. Moreover, dialogue between the pub- including the education goal—is still being debated, lic and private sectors was judged to be difficult and the legitimacy that progress could confer on weak or rarely led to action.53 Tackling these issues is vital for unstable governments was often a powerful incentive the emergence of the coalitions needed to design and for change. Many countries introduced reforms to implement effective policies. expand access to schooling in successful efforts to In education, consultative groups and civil society meet the MDGs. The Sustainable Development Goal organizations could promote more inclusive reforms. (SDG) indicators, which will include a set of compa- The Civil Society Education Fund (CSEF), launched in rable learning measures, could play a similar role by 2009, has supported national education coalitions in motivating countries to shift their focus from school- more than 40 developing countries, and the number ing to learning. of civil society organizations involved in education By supporting improvements in learning assess- planning and policy has expanded rapidly.54 For ment, external actors can help shine a light on low example, the fund has supported the Ghana National learning levels and their causes. For one thing, they Education Campaign Coalition (GNECC) in lobbying can help developing countries participate in regional for more participatory education planning, policy and global assessments, which are an important tool formulation, and monitoring. GNECC members have for opening up spaces for change and influencing worked together to present new findings on educa- policy debates.51 They could also help ensure that test tion issues during annual education review meetings items are linked across countries and across time, and to advocate for change.55 which would allow results of different assessments to be more comparable. External actors can also help Link financing more closely to results that by supporting national assessment efforts, so that lead to learning they can provide more politically salient information While the overall contribution of development on learning. The READ program, a partnership among assistance to country investments in education is development partners, education practitioners, and relatively small, it is important in some low-income low-income country governments, has helped coun- countries (figure 11.4). In 2015 international finance tries strengthen their national assessments, while accounted for 14 percent of education spending in How to escape low-learning traps | 211 Figure 11.4 Most funding for education activities but are not aligned with career development comes from domestic sources, but incentives are likely to be less sustainable. External international finance is important for actors can support alignment by shifting the focus of low-income countries systems toward learning, linking their financing to results rather than the provision of specific inputs or Estimated sources of education spending, by income group (2015) activities. More development partners are using results- 100 based financing in education. These approaches seek to align system components by linking financing to results. They shift the emphasis from inputs toward 80 performance. Some financing is linked directly to student achievement. For example, a U.K. program 60 that supports the education system in Ethiopia pro- Percent vides an agreed-on amount for net increases in the number of students who pass the examination at the 40 end of lower secondary education. The multidonor- financed Big Results Now in Education program in Tanzania links financing to student learning and to 20 intermediate outputs that support improvements in education quality. The ultimate impact of these 0 approaches on system performance is still being Low-income Lower-middle- Upper-middle- evaluated, since they are new. But initial findings income income suggest they have the potential to tackle system-level International finance constraints and improve system performance.57 Households Domestic public *** Source: Education Commission (2016). Data at http://bit.do/WDR2018 There is nothing inevitable about poor learning out- -Fig_11-4. comes, whatever a country’s level of development. Some countries have used well-documented reforms low-income countries. But support is much higher to escape low-learning traps, successfully reorient­ing in some countries. In Mali, development assistance their systems toward learning. Others have achieved accounted for approximately 25 percent of public learning outcomes that far exceed what their develop- education spending between 2004 and 2010. More- ment level would predict, indicating that they escaped over, global estimates of the investments required the trap in the past. Though there is no single recipe to raise learning as part of the SDGs imply a need for achieving broad-based learning, these cases iden- to increase development assistance, particularly to tify three entry points for getting under way. First, low-income countries.56 deploy information and metrics to shine a light on the But external actors must provide financing in hidden exclusion of low learning. Second, build coali- a way that aligns systems with learning. Projects tions that can better align incentives toward learning, aimed at narrow aspects of reform or on specific especially the learning of the most disadvantaged. interventions, run the risk of exacerbating existing Third, commit to innovation and agility, using feed- misalignments, if weaknesses in other parts of the back loops for continuous improvement. None of this system are not tackled at the same time. For exam- is easy, but history shows that achieving education’s ple, projects that support professional development promise will depend on taking up the challenge. 212 | World Development Report 2018 Notes 1. Cassen, McNally, and Vignoles (2015); Stannard and 53. OECD and UNDP (2016). Huxford (2007); Tanner and others (2010). 54. UNESCO (2015). 2. Mullis and others (2016). 55. CSEF (2014). The CSEF is coordinated by the Global 3. The numeracy strategy was introduced in 1999. Campaign for Education (GCE), with funding from the 4. Evaluations of different aspects of the literacy and Global Partnership for Education. numeracy program are summarized in, for example, 56. Education Commission (2016). Machin and McNally (2008); McNally (2015); and Stan- 57. Sabarwal, Joshi, and Blackmon (2017). nard and Huxford (2007). 5. Khemani (2015). 6. Banerjee and others (2011); Brender (2003). References 7. Dias and Ferraz (2017); Toral (2016). Abdul-Hamid, Husein, Harry Anthony Patrinos, Joel Reyes, 8. Andrabi, Das, and Khwaja (2015). Jo Kelcey, and Andrea Diaz Varela. 2016. “Learning in the 9. Barr, Packard, and Serra (2014). Face of Adversity: The UNRWA Education Program for 10. Snilstveit and others (2015). Palestine Refugees.” World Bank Study Series, World 11. Carr-Hill and others (2015); Grandvoinnet, Aslam, and Bank, Washington, DC. Raha (2015). Alonso, Juan Diego, and Alonso Sánchez, eds. 2011. Reforming 12. Banerjee and others (2010); World Bank (2017c). Education Finance in Transition Countries: Six Case Studies in 13. Reinikka and Svensson (2011). Per Capita Financing Systems. World Bank Study Series, 14. Filmer and Schady (2009). World Bank, Washington, DC. 15. Barrera-Osorio and Filmer (2016). Andrabi, Tahir, Jishnu Das, and Asim Ijaz Khwaja. 2015. 16. UNDP (2011). “Report Cards: The Impact of Providing School and Child 17. UIS (2016). Test Scores on Educational Markets.” Policy Research 18. Rath and others (2015). Working Paper 7226, World Bank, Washington, DC. 19. Sutcliffe and Court (2005). Andrews, Matt J. 2015. “Explaining Positive Deviance in Pub- 20. Di Tella and Franceschelli (2011). lic Sector Reforms in Development.” World Development 21. Corrales (1999). 74: 197–208. 22. Bruns and Schneider (2016); Corrales (1999). Andrews, Matt J., Lant Pritchett, and Michael Woolcock. 23. Bruns and Luque (2015). 2013. “Escaping Capability Traps through Problem 24. Bruns and Luque (2015); World Bank (2017c). Driven Iterative Adaptation (PDIA).” World Development 25. Sabel and Jordan (2015); World Bank (2017b). 51: 234–44. 26. Jakubowski (2015); Jakubowski and others (2010). —— ——. 2017. Building State Capability: Evidence, Analysis, Action. 27. Wojciuk (2017). New York: Oxford University Press. 28. Beuchert and others (2016). Ansell, Ben W. 2006. “From the Ballot to the Blackboard: The 29. Chang and others (2013). Redistributive Political Economy of Education.” PhD dis- 30. de Ree and others (2015). sertation, Harvard University. http://users.polisci.umn 31. Chang and others (2013). .edu/~ansell/papers/Ben%20Ansell%20Dissertation.pdf. 32. Levy and others (2016). Avalos, Beatrice, and Jenny Assael. 2006. “Moving from 33. Mansuri and Rao (2013). Resistance to Agreement: The Case of the Chilean 34. Burde and Linden (2012). Teacher Performance Evaluation.” International Journal of 35. Besley and Ghatak (2005); Finan, Olken, and Pande Educational Research 45 (4): 254–66. (2015). Banerjee, Abhijit Vinayak, Rukmini Banerji, James Berry, 36. Estrada (2015). Esther Duflo, Harini Kannan, Shobhini Mukerji, Marc 37. World Bank (2017c). Shotland, and Michael Walton. 2016. “Mainstreaming an 38. Andrews, Pritchett, and Woolcock (2017). Effective Intervention: Evidence from Randomized Eval- 39. Besley and Persson (2009). uations of ‘Teaching at the Right Level’ in India.” CEPR 40. Andrews, Pritchett, and Woolcock (2017). Discussion Paper 11530, Centre for Economic Policy 41. Andrews (2015). Research, London. 42. Green (2016); Pascale, Sternin, and Sternin (2010). Banerjee, Abhijit Vinayak, Rukmini Banerji, Esther Duflo, 43. Banerjee and others (2016). Rachel Glennerster, and Stuti Khemani. 2010. “Pitfalls of 44. Heilmann (2008). Participatory Programs: Evidence from a Randomized 45. Blanchenay (2016). Evaluation in Education in India.” American Economic 46. Andrews, Pritchett, and Woolcock (2013). Journal: Economic Policy 2 (1): 1–30. 47. J-PAL and IPA Perú (2013). Banerjee, Abhijit Vinayak, Selvan Kumar, Rohini Pande, and 48. Grindle (2004). Felix Su. 2011. “Do Informed Voters Make Better Choices? 49. Alonso and Sánchez (2011). Experimental Evidence from Urban India.” Working 50. Leftwich (2009). paper, Harvard University, Cambridge, MA. 51. Devarajan and Khemani (2016). Barr, Abigail, Truman Packard, and Danila Serra. 2014. 52. World Bank (2015). “Participatory Accountability and Collective Action: How to escape low-learning traps | 213 Experimental Evidence from Albania.” European Economic Chang, Mae Chu, Sheldon Shaeffer, Samer Al-Samarrai, Review 68: 250–69. Andrew B. Ragatz, Joppe de Ree, and Ritchie Stevenson. Barrera-Osorio, Felipe, and Deon Filmer. 2016. “Incentivizing 2013. Teacher Reform in Indonesia: The Role of Politics and Schooling for Learning: Evidence on the Impact of Alter- Evidence in Policy Making. Directions in Development: native Targeting Approaches.” Journal of Human Resources Human Development Series. Washington, DC: World 51 (2): 461–99. Bank. Besley, Timothy J., and Maitreesh Ghatak. 2005. “Competi- Chowdhury, A. Mushtaque Raza, Rasheda K. Choudhury, tion and Incentives with Motivated Agents.” American and Samir R. Nath. 1999. Hope Not Complacency: State of Economic Review 95 (3): 616–36. Primary Education in Bangladesh 1999. Dhaka, Bangladesh: Besley, Timothy J., and Torsten Persson. 2009. “The Origins Education Watch, Campaign for Popular Education; of State Capacity: Property Rights, Taxation, and Poli- Dhaka, Bangladesh: University Press. tics.” American Economic Review 99 (4): 1218–44. Contreras, Dante, and Tomás Rau. 2012. “Tournament Incen- Beuchert, Louise Voldby, Maria Knoth Humlum, Helena tives for Teachers: Evidence from a Scaled-Up Interven- Skyt Nielsen, and Nina Smith. 2016. “The Short-Term tion in Chile.” Economic Development and Cultural Change Effects of School Consolidation on Student Achieve- 61 (1): 219–46. ment: Evidence of Disruption?” IZA Discussion Paper Corrales, Javier. 1999. “The Politics of Education Reform: 10195, Institute for the Study of Labor, Bonn, Germany. Bolstering the Supply and Demand, Overcoming Institu- Blanchenay, Patrick. 2016. “Policy Experimentation in Com- tional Blocks.” Report 22549, Education Reform and plex Education Systems.” In Governing Education in a Management Series, World Bank, Washington, DC. Complex World, edited by Tracey Burns and Florian CSEF (Civil Society Education Fund). 2014. “Civil Society Köster, 161–86. Educational Research and Innovation Advocacy: Good Practice Case Studies from Africa.” Series. Paris: Centre for Educational Research and Inno- Global Campaign for Education, Johannesburg. vation, Organisation for Economic Co-operation and Delannoy, Françoise. 2000. “Educational Reforms in Chile, Development. 1980–1998: A Lesson in Pragmatism.” Report 20806, Brender, Adi. 2003. “The Effect of Fiscal Performance on Country Studies, Education Reform and Management Local Government Election Results in Israel: 1989–1998.” Series, World Bank, Washington, DC. Journal of Public Economics 87 (9): 2187–2205. de Ree, Joppe, Karthik Muralidharan, Menno Pradhan, and Bruns, Barbara, David K. Evans, and Javier Luque. 2011. F. Halsey Rogers. 2015. “Double for Nothing? Experimen- Achieving World-Class Education in Brazil: The Next Agenda. tal Evidence on the Impact of an Unconditional Teacher Report 65659. Directions in Development: Human Salary Increase on Student Performance in Indonesia.” Development Series. Washington, DC: World Bank. NBER Working Paper 21806, National Bureau of Eco- Bruns, Barbara, and Javier Luque. 2015. Great Teachers: How to nomic Research, Cambridge, MA. Raise Student Learning in Latin America and the Caribbean. Devarajan, Shantayanan, and Stuti Khemani. 2016. “If Poli- With Soledad De Gregorio, David K. Evans, Marco tics Is the Problem, How Can External Actors Be Part of Fernández, Martin Moreno, Jessica Rodriguez, Guill- the Solution?” Policy Research Working Paper 7761, ermo Toral, and Noah Yarrow. Latin American Develop- World Bank, Washington, DC. ment Forum Series. Washington, DC: World Bank. Dias, Marina, and Claudio Ferraz. 2017. “Voting for Quality? Bruns, Barbara, and Ben Ross Schneider. 2016. “Managing The Impact of School Quality Information on Electoral the Politics of Quality Reforms in Education: Policy Outcomes.” Departamento de Economia, Pontifícia Lessons from Global Experience.” Background Paper: Universidade Católica do Rio de Janeiro, Rio de Janeiro. The Learning Generation, International Commission on Di Tella, Rafael, and Ignacio Franceschelli. 2011. “Govern- Financing Global Education Opportunity, New York. ment Advertising and Media Coverage of Corruption Burde, Dana, and Leigh L. Linden. 2012. “The Effect of Scandals.” American Economic Journal: Applied Economics Village-Based Schools: Evidence from a Randomized 3 (4): 119–51. Controlled Trial in Afghanistan.” NBER Working Paper Education Commission. 2016. The Learning Generation: 18039, National Bureau of Economic Research, Cam- Investing in Education for a Changing World. New York: bridge, MA. International Commission on Financing Global Educa- Campos, Jose Edgardo, Benjamina Randrianarivelo, and Kay tion Opportunity. Winning. 2015. “Escaping the ‘Capability Trap’: Turning Estrada, Ricardo. 2015. “Rules Rather Than Discretion: ‘Small’ Development into ‘Big’ Development.” Interna- Teacher Hiring and Rent Extraction.” EUI Working Paper tional Public Management Review 16 (1): 99–131. MWP 2015/14, Max Weber Program, European Univer- Carr-Hill, Roy, Caine Rolleston, Tejendra Pherali, Rebecca sity Institute, San Domenico di Fiesole, Italy. Schendel, Edwina Peart, and Emma Jones. 2015. The Ferraz, Claudio, and Barbara Bruns. 2012. “Paying Teachers Effects of School-Based Decision Making on Educational to Perform: The Impact of Bonus Pay in Pernambuco, Outcomes in Low- and Middle-Income Contexts: A Systematic Brazil.” Working paper, World Bank, Washington, DC. Review. 3ie Grantee Final Review. London: International Filmer, Deon, and Norbert R. Schady. 2009. “School Enroll- Initiative for Impact Evaluation. ment, Selection, and Test Scores.” Policy Research Work- Cassen, Robert, Sandra McNally, and Anna Vignoles. 2015. ing Paper 4998, World Bank, Washington, DC. Making a Difference in Education: What the Evidence Says. Finan, Frederico S., Benjamin A. Olken, and Rohini Pande. Abingdon, U.K.: Routledge. 2015. “The Personnel Economics of the State.” NBER 214 | World Development Report 2018 Working Paper 21825, National Bureau of Economic Programme). 2016. “Making Development Co-operation Research, Cambridge, MA. More Effective: 2016 Progress Report.” Paris: OECD. Grandvoinnet, Helene, Ghazia Aslam, and Shomikho Raha. Pascale, Richard T., Jerry Sternin, and Monique Sternin. 2015. Opening the Black Box: The Contextual Drivers of Social 2010. The Power of Positive Deviance: How Unlikely Innovators Accountability. New Frontiers of Social Policy Series. Solve the World’s Toughest Problems. Boston: Harvard Busi- Washington, DC: World Bank. ness Press. Green, Duncan. 2016. How Change Happens. Oxford, U.K.: R4D (Results for Development Institute). 2015. “Bringing Oxford University Press. Learning to Light: The Role of Citizen-Led Assessments Grindle, Merilee Serrill. 2004. Despite the Odds: The Conten- in Shifting the Education Agenda.” R4D, Washington, tious Politics of Education Reform. Princeton, NJ: Princeton DC. University Press. Rath, Amitav, Pamela Branch, Dunstan Kishekya, Clement Heilmann, Sebastian. 2008. “Policy Experimentation in Kihinga, Terry Smutylo, and Kornelia Rassmann. 2015. China’s Economic Rise.” Studies in Comparative Interna- Evaluation Twaweza: Tanzania 2009–2014, Final Report. tional Development 43 (1): 1–26. With the assistance of Constance Lim, Yusra Uzair, and Jakubowski, Maciej. 2015. “Opening Up Opportunities: Edu- Maya Kovacevic. SIDA Decentralized Evaluation 2015. cation Reforms in Poland.” IBS Policy Paper 01/2015, Stockholm: Department for Africa, Swedish Interna- Intelligent Business Solutions, Gliwice, Poland. tional Development Cooperation Agency. Jakubowski, Maciej, Harry Anthony Patrinos, Emilio Ernesto Reinikka, Ritva, and Jakob Svensson. 2011. “The Power of Porta, and Jerzy Wiśniewski. 2010. “The Impact of the Information in Public Services: Evidence from Educa- 1999 Education Reform in Poland.” Policy Research tion in Uganda.” Journal of Public Economics 95 (7): 956–66. Working Paper 5263, World Bank, Washington, DC. Rosser, Andrew J., and Anuradha Joshi. 2017. “Using Courts to J-PAL (Abdul Latif Jameel Poverty Action Lab) and IPA Perú Realize Education Rights and Create Opportunities to (Peru Country Office, Innovations for Poverty Action). Improve Learning.” Background paper, World Bank, 2013. “Implementación del Laboratorio de Innovación Washington, DC. Costo-Efectiva de la Política Educativa: MineduLAB.” Sabarwal, Shwetlena, Anuradha Joshi, and William Black- J-PAL, Santo Domingo, Dominican Republic. mon. 2017. “A Review of the World Bank’s Results-Based Khemani, Stuti. 2015. “Buying Votes versus Supplying Public Financing Mechanism Used for Tanzania’s Big Results Services: Political Incentives to Under-Invest in Pro-poor Now in Education Program: A Process Evaluation.” Policies.” Journal of Development Economics 177: 84–93. World Bank, Washington, DC. Leftwich, Adrian. 2009. “Bringing Agency Back In: Politics Sabel, Charles, and Luke Jordan. 2015. “Doing, Learning, and Human Agency in Building Institutions and States, Being: Some Lessons Learned from Malaysia’s National Synthesis and Overview Report.” DLP Research Paper 6, Transformation Program.” Competitive Industries and Developmental Leadership Program, Birmingham, U.K. Innovation Program, World Bank, Washington, DC. Levy, Brian, Robert Cameron, Ursula Hoadley, and Vinothan Snilstveit, Birte, Jennifer Stevenson, Daniel Phillips, Martina Naidoo. 2016. “The Politics of Governance and Basic Vojtkova, Emma Gallagher, Tanja Schmidt, Hannah Education: A Tale of Two South African Provinces.” Occa- Jobse, et al. 2015. Interventions for Improving Learning Out- sional Working Paper 2, Graduate School of Develop- comes and Access to Education in Low- and Middle-Income ment Policy and Practice, University of Cape Town. Countries. 3ie Systematic Review 24, London: Interna- Machin, Stephen, and Sandra McNally. 2008. “The Literacy tional Initiative for Impact Evaluation. http://www.3ie Hour.” Journal of Public Economics 92 (5): 1441–62. impact.org/media/filer_public/2016/07/12/sr24-education Mansuri, Ghazala, and Vijayendra Rao. 2013. Localizing Devel- -review.pdf. opment: Does Participation Work? Policy Research Report Stannard, John, and Laura Huxford. 2007. The Literacy Game: Series. Washington, DC: World Bank. The Story of the National Literacy Strategy. New York: McNally, Sandra. 2015. “Numeracy and Mathematics.” In Routledge. Making a Difference in Education: What the Evidence Says, Sutcliffe, Sophie, and Julius Court. 2005. “Evidence-Based edited by Robert Cassen, Sandra McNally, and A. Policymaking: What Is It? How Does It Work? What Vignoles, 123–34. Abingdon, U.K.: Routledge. Relevance for Developing Countries?” Research Reports Mizala, Alejandra, and Ben Ross Schneider. 2014. “Negotiat- and Studies Series, Overseas Development Institute, ing Education Reform: Teacher Evaluations and Incen- London. tives in Chile (1990–2010).” Governance 27 (1): 87–109. Tanner, Emily, Ashley Brown, Naomi Day, Mehul Kotecha, Mullis, I. V. S., M. O. Martin, P. Foy, and M. Hooper. Natalie Low, Gareth Morrell, Ola Turczuk, et al. 2010. 2016.  “TIMSS 2015 International Results in Mathemat- Evaluation of Every Child a Reader (ECaR). Research Report ics.” TIMSS and PIRLS International Study Center, DFE-RR114. London: U.K. Department for Education. Boston College, Chestnut Hill, MA. http://timssandpirls Todd, Robin, and Ian Attfield. 2017. “Big Results Now! In .bc.edu/timss2015/international-results/. Tanzanian Education: Has the Delivery Approach Deliv- OECD (Organisation for Economic Co-operation and Devel- ered?” U.K. Department for International Development, opment). 2016. PISA 2015 Results: Excellence and Equity in London. Education. Vol. 1. Paris: OECD. Toral, Guillermo. 2016. “When Are Local Governments and OECD (Organisation for Economic Co-operation and Devel- Bureaucrats Held Accountable for the Quality of Public opment) and UNDP (United Nations Development Services? Evidence from Brazil’s Education Sector.” How to escape low-learning traps | 215 MIT Political Science Research Paper 2016-11, Political Docencia Económicas and Massachusetts Institute of Science Department, Massachusetts Institute of Tech- Technology’s conference, “Comparative Political Econ- nology, Cambridge, MA. omy of Education Reforms,” Mexico City, February 2–3. UIS (UNESCO Institute for Statistics). 2016. “Laying the World Bank. 2015. Final READ Trust Fund Report 2008–2015. Foundation to Measure Sustainable Development Goal 4.” Report 101527. Washington, DC: World Bank. Sustainable Development Data Digest, UIS, Montreal. ——— —. 2017a. “Case Study of Chile’s Pay for Performance UNDP (United Nations Development Programme). 2011. Reforms, 1995–2005.” Background note, World Bank, “Mexico: Scaling Up Progresa/Oportunidades, Condi- Washington, DC. tional Cash Transfer Programme.” UNDP, New York. ——— —. 2017b. “Driving Performance from the Center: UNESCO (United Nations Educational, Scientific, and Cul- Malaysia’s Experience with PEMANDU.” Knowledge and tural Organization). 2015. “Civil Society Education Fund Research, Malaysia Development Experience Series, 2013–2014/5.” Biannual Progress and Supervision Report, Global Knowledge and Research Hub, World Bank, January–June 2015, UNESCO, Paris. Washington, DC. Wojciuk, Anna. 2017. “Poland: A Notorious Case of Shock ——— —. 2017c. World Development Report 2017: Governance and Therapy.” Paper presented at Centro de Investigación y the Law. Washington, DC: World Bank. 216 | World Development Report 2018 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 emissions, 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 content. The recycled fiber in our book paper is either unbleached or bleached using totally chlorine-free (TCF), processed chlorine–free (PCF), or enhanced elemental chlorine– free (EECF) processes. More information about the Bank’s environmental philosophy can be found at http://www.worldbank.org/corporateresponsibility. Every year, the World Bank’s World Development Report (WDR) features a topic of central importance to global development. The 2018 WDR— LEARNING to Realize Education’s Promise—is the first ever devoted entirely to education. And the time is right: education has long been critical to human welfare, but it is even more so in a time of rapid economic and social change. The best way to equip children and youth for the future is to make their learning the center of all efforts to promote education. The 2018 WDR explores four main themes: First, education’s promise: education is a powerful instrument for eradicating poverty and promoting shared prosperity, but fulfilling its potential requires better policies—both within and outside the education system. Second, the need to shine a light on learning: despite gains in access to education, recent learning assessments reveal that many young people around the world, especially those who are poor or marginalized, are leaving school unequipped with even the foundational skills they need for life. At the same time, internationally comparable learning assessments show that skills in many middle-income countries lag far behind what those countries aspire to. And too often these shortcomings are hidden—so as a first step to tackling this learning crisis, it is essential to shine a light on it by assessing student learning better. Third, how to make schools work for all learners: research on areas such as brain science, pedagogical innovations, and school management has identified interventions that promote learning by ensuring that learners are prepared, teachers are both skilled and motivated, and other inputs support the teacher-learner relationship. Fourth, how to make systems work for learning: achieving learning throughout an education system requires more than just scaling up effective interventions. Countries must also overcome technical and political barriers by deploying salient metrics for mobilizing actors and tracking progress, building coalitions for learning, and taking an adaptive approach to reform.