92863 world development report MIND, SOCIETY, AND BEHAVIOR A World Bank Group Flagship Report world development report MIND, SOCIETY, AND BEHAVIOR © 2015 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 17 16 15 14 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. 2015. World Development Report 2015: Mind, Society, and Behavior. Washington, DC: World Bank. doi: 10.1596/978-1-4648-0342-0. License: Creative Commons Attribution CC BY 3.0 IGO Translations—If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. Adaptations—If you create an adaptation of this work, please add the following disclaimer along with the attribution: This is an adaptation of an original work by The World Bank. Views and opinions expressed in the adaptation are the sole responsibility of the author or authors of the adaptation and are not endorsed by The World Bank. Third-party content—The World Bank does not necessarily own each component of the content contained within the work. The World Bank therefore does not warrant that the use of any third-party-owned individual component or part contained in the work will not infringe on the rights of those third parties. The risk of claims resulting from such infringement rests solely with you. If you wish to reuse a component of the work, it is your responsibility to determine whether permission is needed for that reuse and to obtain permission from the copyright owner. Examples of components can include, but are not limited to, tables, figures, or images. All queries on rights and licenses should be addressed to the Publishing and Knowledge Division, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ISSN, ISBN, e-ISBN, and DOI: Softcover ISSN: 0163-5085 ISBN: 978-1-4648-0342-0 e-ISBN: 978-1-4648-0343-7 DOI: 10.1596/978-1-4648-0342-0 Hardcover ISSN: 0163-5085 ISBN: 978-1-4648-0344-4 DOI: 10.1596/978-1-4648-0344-4 Cover and interior design: George Kokkinidis, Design Language, Brooklyn, New York Photos and illustrations: Page 53: Stickers used in Kenyan minibuses © Saracen OMD (Nairobi). In James Habyarimana and William Jack. 2011. “Heckle and Chide: Results of a Randomized Road Safety Intervention in Kenya.” Journal of Public Economics 95 (11): 1438–46. Used with permission. Further permission required for reuse. Page 148: Photo of former Brazilian president Lula da Silva © Ricardo Stuckert/Instituto Lula. Used with permission. Further permission required for reuse. Page 177: Reproduced newspaper excerpt © El Espectador (Bogotá), February 28, 1997, A6. Used with permission. Further permission required for reuse. Contents xi Foreword xiii Acknowledgments xvii Abbreviations 1 Overview: Human decision making and development policy 5 Three principles of human decision making 13 Psychological and social perspectives on policy 18 The work of development professionals 21 References 24 Part 1: An expanded understanding of human behavior for economic development: A conceptual framework 25 Introduction 26 Chapter 1: Thinking automatically 26 Two systems of thinking 29 Biases in assessing information 34 Biases in assessing value 36 Choice architecture 37 Overcoming intention-action divides 38 Conclusion 38 Notes 39 References 42 Chapter 2: Thinking socially 43 Social preferences and their implications 49 The influence of social networks on individual decision making 51 The role of social norms in individual decision making v vi CONTENTS 54 Conclusion 55 Notes 55 References 60 Spotlight 1: When corruption is the norm 62 Chapter 3: Thinking with mental models 63 Where mental models come from and why they matter 63 How mental models work and how we use them 65 The roots of mental models 67 The effects of making an identity salient 68 The staying power of mental models 70 Policies to improve the match of mental models with a decision context 72 Conclusion 72 Notes 73 References 76 Spotlight 2: Entertainment education 79 Part 2: Psychological and social perspectives on policy 80 Chapter 4: Poverty 81 Poverty consumes cognitive resources 84 Poverty creates poor frames 85 Social contexts of poverty can generate their own taxes 86 Implications for the design of antipoverty policies and programs 90 Looking ahead 91 References 94 Spotlight 3: How well do we understand the contexts of poverty? 98 Chapter 5: Early childhood development 99 Richer and poorer children differ greatly in school readiness 100 Children need multiple cognitive and noncognitive skills to succeed in school 101 Poverty in infancy and early childhood can impede early brain development 101 Parents are crucial in supporting the development of children’s capacities for learning 103 Parents’ beliefs and caregiving practices differ across groups, with consequences for children’s developmental outcomes 104 Designing interventions that focus on and improve parental competence 108 Conclusion 108 Notes 109 References 112 Chapter 6: Household finance 113 The human decision maker in finance 117 Policies to improve the quality of household financial decisions 123 Conclusion 123 Notes 123 References CONTENTS vii 128 Chapter 7: Productivity 129 Improving effort among employees 134 Recruiting high-performance employees 135 Improving the performance of small businesses 136 Increasing technology adoption in agriculture 139 Using these insights in policy design 140 Notes 140 References 144 Spotlight 4: Using ethnography to understand the workplace 146 Chapter 8: Health 146 Changing health behaviors in the face of psychological biases and social influences 149 Psychological and social approaches to changing health behavior 151 Improving follow-through and habit formation 153 Encouraging health care providers to do the right things for others 155 Conclusion 155 Notes 156 References 160 Chapter 9: Climate change 161 Cognitive obstacles inhibit action on climate change 167 Psychological and social insights for motivating conservation 171 Conclusion 171 Notes 171 References 176 Spotlight 5: Promoting water conservation in Colombia 179 Part 3: Improving the work of development professionals 180 Chapter 10: The biases of development professionals 181 Complexity 182 Confirmation bias 185 Sunk cost bias 186 The effects of context on judgment and decision making 189 Conclusion 190 Notes 190 References 192 Chapter 11: Adaptive design, adaptive interventions 194 Diagnosing psychological and social obstacles 195 Designing an intervention 198 Experimenting during implementation 199 Conclusion: Learning and adapting 199 References 202 Spotlight 6: Why should governments shape individual choices? 205 Index viii CONTENTS Boxes O.1 5 The evolution of thinking in economics about human decision making 10.1 184 The home team advantage: Why experts are consistently biased 10.2 187 A clash of values between development professionals and the local populace: Agricultural reform in Lesotho 10.3 188 It may be difficult for development professionals to accurately predict the views of poor people 11.1 195 Taking the perspective of program beneficiaries through the Reality Check approach 11.2 195 Measurement techniques that can help uncover psychological and social obstacles 11.3 198 Using psychological and social insights and active experimentation in the United Kingdom Figures O.1 7 Automatic thinking gives us a partial view of the world O.2 8 Reframing decisions can improve welfare: The case of payday borrowing O.3 9 What others think, expect, and do influences our preferences and decisions O.4 10 In experimental situations, most people behave as conditional cooperators rather than free riders O.5 11 Thinking draws on mental models O.6 12 Cuing a stigmatized or entitled identity can affect students’ performance O.7 15 There is greater variation across countries in cognitive caregiving than in socioemotional caregiving O.8 16 Clarifying a form can help borrowers find a better loan product O.9 21 Understanding behavior and identifying effective interventions are complex and iterative processes 1.1 28 Framing affects what we pay attention to and how we interpret it 1.2 29 A more behavioral model of decision making expands the standard economic model 1.3 33 Reframing decisions can improve welfare: The case of payday borrowing 1.4 34 Clarifying a form can help borrowers find a better loan product 1.5 35 A small change in the college application process had a huge impact on college attendance 1.6 37 Simplifying voting procedures in Brazil is having positive welfare effects on the poor across generations 2.1 43 What others think, expect, and do influences our own preferences and decisions 2.2 45 Children and young adults most affected by war are more likely to favor members of their own group 2.3 47 Opportunities to punish free riding increase cooperation 2.4 48 In experimental situations, most people behave as conditional cooperators rather than free riders 2.5 49 The power of social monitoring: Pictures of eyes increased contributions to a beverage honor bar 2.6 53 Stickers placed in Kenyan minibuses reduced traffic accidents 3.1 64 What we perceive and how we interpret it depend on the frame through which we view the world around us 3.2 67 Making criminal identity more salient increases dishonesty in prison inmates CONTENTS ix 3.3 68 Cuing a stigmatized or entitled identity can affect students’ performance 3.4 71 Changing disruptive children’s mental models related to trust improved adult outcomes 4.1 82 Poverty is a fluid state, not a stable condition 4.2 83 Financial scarcity can consume cognitive resources 4.3 84 Measuring executive function and fluid intelligence 4.4 88 Targeting on the basis of bandwidth may help people make better decisions S3.1 94 How poor and affluent people in New Jersey view traveling for a discount on an appliance S3.2 95 How World Bank staff view traveling for a discount on a watch S3.3 96 How people in Jakarta, Indonesia, view traveling for a discount on a cell phone S3.4 96 How people in Nairobi, Kenya, view traveling for a discount on a cell phone S3.5 97 How people in Lima, Peru, view traveling for a discount on a cell phone 5.1 99 Variations by wealth in basic learning skills are evident by age three in Madagascar 5.2 100 Abilities in receptive language for three- to six-year-olds vary widely by wealth in five Latin American countries 5.3 101 Unrelenting stress in early childhood can be toxic to the developing brain 5.4 104 There is greater variation across countries in cognitive caregiving than in socioemotional caregiving 5.5 106 A program in rural Senegal encourages parents to engage verbally with their children 5.6 107 Early childhood stimulation in Jamaica resulted in long-term improvements in earnings 6.1 117 Simplifying information can help reduce take-up of payday loans 6.2 119 Changing default choices can improve savings rates 6.3 121 Commitment savings accounts can improve agricultural investment and profit 6.4 122 Popular media can improve financial decisions 7.1 131 Unexpected wage increases can trigger a productivity dividend 7.2 133 Public recognition can improve performance more than financial incentives can 7.3 137 Altering the timing of purchases can be as effective as a subsidy for improving investment 7.4 138 Not noticing a decision can hurt productivity 8.1 148 If a well-known person has a disease, the public might think more seriously about ways to prevent it 8.2 150 Take-up of health products drops precipitously in response to very small fees 8.3 152 Text message reminders can improve adherence to lifesaving drugs 8.4 153 Changing social norms is important but not sufficient for ending open defecation 9.1 163 Worldviews can affect perceptions of the risk posed by climate change 9.2 164 Predicting the effect of rainfall forecasts on the success of growing familiar crops was difficult for farmers in Zimbabwe 9.3 167 Democratic rules can achieve high levels of resource sustainability S5.1 177 The story of Bogotá’s 1997 water supply crisis 10.1 183 How development professionals interpreted data subjectively 10.2 186 How World Bank staff viewed sunk costs x CONTENTS B10.3.1 188 How World Bank staff predicted the views of poor people 11.1 193 Understanding behavior and identifying effective interventions are complex and iterative processes Photo 8.1 148 Former Brazilian president Lula da Silva’s battle with throat cancer was widely covered in the media Tables O.1 6 People have two systems of thinking O.2 13 Examples of highly cost-effective behavioral interventions 1.1 27 People have two systems of thinking 11.1 196 Different obstacles may require different intervention design (Case study: increasing home water chlorination) 11.2 197 Experimental evidence is accumulating on the effectiveness of many psychologically and socially informed designs Foreword As I write, the world is fighting to control the Ebola outbreak in West Africa, a human tragedy that has cost thousands of lives and brought suffering to families and across entire communities. The outbreak is a tragedy not only for those directly affected by the disease but also for their neighbors and fellow citizens. And the indirect, behavioral effects of the Ebola crisis—slowing business activity, falling wages, and rising food prices—will make life even more difficult for millions of people who already live in extreme poverty in that region of the world. Some of these behavioral effects are unavoidable. Ebola is a terrible disease, and quaran- tines and other public health measures are necessary parts of the response. At the same time, it is clear that the behavioral responses we are seeing, not just in West Africa but all over the world, are partly driven by stigma, inaccurate understanding of disease transmission, exaggerated panic, and other biases and cognitive illusions. Sadly, we have seen this happen before, with HIV/AIDS and the SARS and H1N1 influenza outbreaks, and we will likely see it again when we begin to prepare for the next outbreak. Societies are prone to forget what happened, and policy makers tend to focus on the most socially prominent risks, which are not always those that drive disease outbreaks. In light of these risks, this year’s World Development Report—Mind, Society, and Behavior— could not be more timely. Its main message is that, when it comes to understanding and changing human behavior, we can do better. Many development economists and practition- ers believe that the “irrational” elements of human decision making are inscrutable or that they cancel each other out when large numbers of people interact, as in markets. Yet, we now know this is not the case. Recent research has advanced our understanding of the psy- chological, social, and cultural influences on decision making and human behavior and has demonstrated that they have a significant impact on development outcomes. Research also shows that it is possible to harness these influences to achieve development goals. The Report describes an impressive set of results. It shows that insights into how people make decisions can lead to new interventions that help households to save more, firms to increase productivity, communities to reduce the prevalence of diseases, parents to improve cognitive development in children, and consumers to save energy. The promise of this approach to decision making and behavior is enormous, and its scope of application is extremely wide. Let me focus on a few themes. First, it has implications for service delivery. Research shows that small differences in context, convenience, and salience have large effects on crucial choices, such as whether to send a child to school, prevent illness, or save to start a business. That means development practitioners need to focus not only on what interventions are needed but also on how they are implemented. That, in turn, requires implementing agencies to spend more time and resources experimenting, learning, and adapting during the intervention cycle. Second, as the risks and impact of climate change become clearer, we must use every tool at our disposal to confront the challenge. The Report describes how, in addition to taxes and subsidies, behavioral and social insights can help. These include reframing messages to emphasize the visible and palpable benefits of reducing emissions, and the use of social xi xii FOREWORD norms to reduce energy consumption, social networks to motivate national commitments, and analogies to help people grasp climate forecasts. Third, development professionals and policy makers are, like all human beings, subject to psychological biases. Governments and international institutions, including the World Bank Group, can implement measures to mitigate these biases, such as more rigorously diagnos- ing the mindsets of the people we are trying to help and introducing processes to reduce the effect of biases on internal deliberations. The Ebola outbreak makes clear that misunderstanding and miscommunicating risks can have serious repercussions. Quarantining infected individuals can prove sensible, but trying to quarantine nations or entire ethnic groups violates human rights and may actually hin- der efforts to control the outbreak of a disease. This year’s World Development Report provides insight into how to address these and other current challenges and introduces an important new agenda for the development community going forward. Jim Yong Kim President The World Bank Group Acknowledgments This Report was prepared by a team led by Karla Hoff and Varun Gauri and comprising Sheheryar Banuri, Stephen Commins, Allison Demeritt, Anna Fruttero, Alaka Holla, and Ryan Muldoon, with additional contributions from Elisabeth Beasley, Saugato Datta, Anne Fernald, Emanuela Galasso, Kenneth Leonard, Dhushyanth Raju, Stefan Trautmann, Michael Woolcock, and Bilal Zia. Research analysts Scott Abrahams, Hannah Behrendt, Amy Packard Corenswet, Adam Khorakiwala, Nandita Krishnaswamy, Sana Rafiq, Pauline Rouyer, James Walsh, and Nan Zhou completed the team. The work was carried out under the general direc- tion of Kaushik Basu and Indermit Gill. The team received guidance from an Advisory Panel consisting of Daron Acemoglu, Paul DiMaggio, Herbert Gintis, and Cass Sunstein. Stefan Dercon gave insightful comments throughout. Sendhil Mullainathan provided invaluable guidance at the conceptual stages of the Report. Valuable inputs were received from all World Bank Group regions, the anchor networks, the research group, the global practices, the Independent Evaluation Group, and other units. The World Bank Chief Economist Council and the Chief Economist’s Council of Eminent Persons provided many helpful comments. The team would like to acknowledge the generous support for the preparation of the Report by the U.K. Department for International Development; Canada’s Department of Foreign Affairs, Trade, and Development; the Knowledge for Change Program; the Nordic Trust Fund; and the World Bank research support budget. The team also thanks the German Federal Ministry for Economic Cooperation and Development and the Deutsche Gesellschaft für Internationale Zusammenarbeit, which co-organized and hosted the WDR International Policy Workshop in Berlin in December 2013. Consultations were held with the International Monetary Fund; the Organisation for Economic Co-operation and Development; UNICEF and several other United Nations organizations; the Netherlands Ministry of Foreign Affairs; the European Commission; and agencies for development cooperation in Japan (Japan International Cooperation Agency), France (Agence Française de Développement), the United Kingdom (Department for International Development), and the United States (U.S. Agency for International Development). Several other organizations sponsored events to provide feedback on the Report, including Columbia University, Cornell University, the Danish Nudging Network, Experiments in Governance and Politics, Harvard Univer- sity, the International Rescue Committee, the London School of Economics and Political Science, the Overseas Development Institute, Save the Children International, and the U.K. Behavioural Insights Team. Nancy Morrison was the principal editor of the Report. George Kokkinidis was the prin- cipal graphic designer. Timothy Taylor provided valuable editorial guidance. Dana Lane copyedited the Report. The World Bank’s Publishing and Knowledge Division coordinated the design, typesetting, printing, and dissemination of the Report. Special thanks to Denise Bergeron, Mary Fisk, Patricia Katayama, Stephen McGroarty, Stephen Pazdan, and Paschal Ssemaganda, as well as the Translation and Interpretation Unit’s Bouchra Belfqih and her team. The team also thanks Vivian Hon, Jimmy Olazo, and Claudia Sepúlveda for their coordi- nating roles and Vamsee Krishna Kanchi, Swati Mishra, and Merrell Tuck-Primdahl for their xiii xiv ACKNOWLEDGMENTS guidance on communications strategy. Renata Gukovas, Ana Maria Muñoz Boudet, Elizaveta Perova, Rafael Proenca, and Abla Safir reviewed some of the foreign language translations of the overview. The production and logistics team for the Report comprised Brónagh Murphy, Mihaela Stangu, and Jason Victor, with contributions from Laverne Cook and Gracia Sorensen. Sonia Joseph, Liliana Longo, and Joseph Welch were in charge of resource management, and Elena Chi-Lin Lee helped coordinate resource mobilization. Jean-Pierre Djomalieu, Gytis Kanchas, and Nacer Megherbi provided IT support. The Report draws on background papers and notes prepared by Abigail Barr, Nicolas Baumard, Timothy Besley, Thomas Bossuroy, Robert Chambers, Molly Crockett, Jonathan de Quidt, Philippe d’Iribarne, Lina Eriksson, Maitreesh Ghatak, Javier Guillot, Crystal Hall, Johannes Haushofer, Alain Henry, Pamela Jakiela, Nadav Klein, Margaret Levi, Margaret Miller, Juan Jose Miranda Montero, Ezequiel Molina, Owen Ozier, Gael Raballand, Anand Rajaram, Barry Schwartz, Pieter Serneels, Jennifer Stellar, Michael Toman, Magdalena Tsaneva, and Daniel Yoo. The team received expert advice from Yann Algan, Jeannie Annan, Nava Ashraf, Mahzarin Banaji, Abhijit Banerjee, Max Bazerman, Gary Becker, Daniel Benjamin, Cristina Bicchieri, Vicki Bogan, Iris Bohnet, Donald Braman, Colin Camerer, Jeffrey Carpenter, Shantayanan Devarajan, Timothy Evans, Marianne Fay, James Greiner, Luigi Guiso, Jonathan Haidt, David Halpern, Joseph Henrich, Ting Jiang, David Just, Dan Kahan, Ravi Kanbur, Jeffrey Kling, John List, Edouard Machery, Mario Macis, Anandi Mani, Suresh Naidu, Michael Norton, Nathan Nunn, Jacques Rajotte, Todd Rogers, Amartya Sen, Owain Service, Joseph Stiglitz, Jan Svejnar, Ann Swidler, and Danielle Valiquette. Many others inside and outside the World Bank provided helpful comments, made other contributions, and participated in consultative meetings. Despite efforts to be comprehen- sive, the team apologizes for any oversights and expresses its gratitude to all who contributed to this Report. The team would like to thank Dina Abu-Ghaida, Ana Milena Aguilar Rivera, Farzana Ahmed, Ahmad Ahsan, Edouard Al-Dahdah, Inger Andersen, Kevin Arceneaux, Omar Arias, Nina Arnhold, Delia Baldassarri, Luca Bandiera, Arup Banerji, Elena Bardasi, Enis Baris, Antonella Bassani, Andrew Beath, Simon Bell, Robert Beschel, João Biehl, Chris Blattman, Erik Bloom, Zeljko Bogetic, Genevieve Boyreau, Hana Brixi, Stefanie Brodmann, Annette Brown, Busara Center for Behavioral Economics, Alison Buttenheim, Susan Caceres, Oscar Calvo-Gonzalez, Robert Chase, Nazmul Chaudhury, Dandan Chen, Laura Chioda, Ken Chomitz, Luc Christiaensen, Rafael Cortez, Aidan Coville, Debra R. Cubitt, Stefano Curto, Amit Dar, Jishnu Das, Maitreyi Das, Augusto de la Torre, Chris Delgado, Asli Demirgüç-Kunt, Clara de Sousa, Jacqueline Devine, Eric Dickson, Quy-Toan Do, Christopher Eldridge, Yasser El-Gammal, Alison Evans, David Evans, Jorge Familiar, Sharon Felzer, Francisco Ferreira, Deon Filmer, Ariel Fiszbein, Luca Flabbi, Elizabeth Fox, Caroline Freund, Marie Gaarder, Virgilio Galdo, Roberta Gatti, Patricia Geli, Swati Ghosh, Xavier Giné, Hemam Girma, Jack Glen, Markus Goldstein, Alvaro Gonzalez, Karla Gonzalez, Maria Gonzalez de Asis, Pablo Gottret, David Gould, Margaret Grosh, Pelle Guldborg Hansen, Nina Guyon, Oliver Haas, Samira Halabi, Stephane Hallegatte, Mary Hallward-Driemeier, John Heath, Rasmus Heltberg, Jesko Hentschel, Marco Hernandez, Arturo Herrera, Barbara Hewitt, Jane Hobson, Bert Hofman, Stephen Hutton, Leonardo Iacovone, Elena Ianchovichina, Alain Ize, Emmanuel Jimenez, Peter John, Melissa Johns, Sandor Karacsony, Sachiko Kataoka, Lauren Kelly, Stuti Khemani, Igor Kheyfets, Arthur Kleinman, Jeni Klugman, Christos Kostopoulos, Sumir Lal, Somik Lall, Daniel Lederman, Alan David Lee, Arianna Legovini, Philippe Le Houérou, Victoria Levin, Jeffrey Lewis, Evan Lieberman, Kathy Lindert, Audrey Liounis, Gladys Lopez-Acevedo, Luis-Felipe Lopez-Calva, Augusto Lopez-Claros, Xubei Luo, Ashish Makkar, Ghazala Mansuri, Brendan Martin, Maria Soledad Martinez Peria, Gwyneth McClendon, Mike McGovern, Miles McKenna, David McKenzie, Julian Messina, Francesca Moneti, Jonathan Morduch, Juan Manuel Moreno ACKNOWLEDGMENTS xv Olmedilla, Ed Mountfield, Masud Mozammel, Margaret Anne Muir, Florentina Mulaj, Cyril Muller, Carina Nachnani, Evgenij Najdov, Ambar Narayan, Christopher David Nelson, Quynh Nguyen, Son Nam Nguyen, Dan Nielson, Adesinaola Michael Odugbemi, Pedro Olinto, Daniel Ortega, Betsy Paluck, Aaka Pande, Valeria Perotti, Kyle Peters, Josefina Posadas, Gael Raballand, Martín Rama, Biju Rao, Francesca Recanatini, Thomas Rehermann, Melissa Rekas, Dena Ringold, Halsey Rogers, Mattia Romani, Onno Ruhl, James Rydge, Seemeen Saadat, Gady Saiovici, Claudio Santibañez, Indhira Santos, Robert Saum, Eva Schiffer, Sergio Schmukler, Pia Schneider, Andrew Schrank, Ethel Sennhauser, Katyayni Seth, Moses Shayo, Sudhir Shetty, Sandor Sipos, Owen Smith, Carlos Sobrado, Nikola Spatafora, Andrew Stone, Mark Sundberg, Bill Sutton, Jeff Tanner, Marvin Taylor-Dormond, Stoyan Tenev, Hans Timmer, Dustin Tingley, Laura Tuck, Tony Tyrrell, Hulya Ulku, Renos Vakis, Tara Vishwanath, Joachim von Amsberg, Adam Wagstaff, Lianqin Wang, Clay Wescott, Josh Wimpey, Noah Yarrow, and Renee Yuet-Yee Ho. Abbreviations ACC anthropogenic climate change APR annual percentage rate ART antiretroviral therapy BIT Behavioural Insights Team CCT conditional cash transfer CDD community-driven development CDP Carbon Disclosure Project CLTS Community-Led Total Sanitation EE entertainment education FAFSA Free Application for Federal Student Aid HDI Human Development Index HPA hypothalamic-pituitary-adrenal IPCC Intergovernmental Panel on Climate Change KAP knowledge, attitudes, and practices MFI microfinance institution NGO nongovernmental organization OECD Organisation for Economic Co-operation and Development ORS oral rehydration salts ORT oral rehydration therapy PATHS Promoting Alternative Thinking Strategies R&D research and development RCT randomized controlled trial ROSCA rotating savings and credit association RSV relative search volume SES socioeconomic status SMarT Save More Tomorrow SMS short messaging service UNICEF United Nations Children’s Fund WDI World Development Indicators (database) WDR 2015 team team for the 2015 World Development Report WHO World Health Organization xvii OVERVIEW Human decision making and development policy Overview Human decision making and development policy Every person seeks to steer his or her own course, idea that decision making is the product of an interac- and a great deal of development policy aims to supply tion between mind and context—were contacted. The the resources and information people in low- and airplane cockpit was redesigned with close attention middle-income economies require in their voyage to how information is packaged and presented, so through life. But while such an approach is often that it fit the human body and its cognitive abilities. appropriate, it can be incomplete. To understand why, These days, airplane cockpits contain fewer instru- consider a comparison with airplane pilots. During ments than several decades ago because the design the middle decades of the 20th century, a number of of cockpit instrument displays is based on a deeper flight and engine instruments were developed with understanding of human cognitive processes (Wiener the intention of improving how pilots steer their air- and Nagel 1988). craft. But by the 1980s, the multiplying technological The title of this Report, Mind, Society, and Behavior, improvements and additional information had the captures the idea that paying attention to how humans think (the processes of mind) and how history and context shape thinking (the influence of society) can This Report aims to inspire and guide improve the design and implementation of develop- ment policies and interventions that target human the researchers and practitioners choice and action (behavior). To put it differently, development policy is due for its own redesign based who can help advance a new set of on careful consideration of human factors. This Report aims to integrate recent findings on the development approaches based on a psychological and social underpinnings of behavior to make them available for more systematic use by both fuller consideration of psychological researchers and practitioners in development com- munities. The Report draws on findings from many and social influences. disciplines, including neuroscience, cognitive science, psychology, behavioral economics, sociology, political science, and anthropology. In ongoing research, these opposite effect of what the designers had intended: findings help explain decisions that individuals make instead of assisting pilots in steering their courses, in many aspects of development, including savings, airplane cockpits had become increasingly complex investment, energy consumption, health, and child environments in which the technical improvements rearing. The findings also enhance the understand- stressed and even overwhelmed the pilots. Rates of ing of how collective behaviors—such as widespread pilot error rose. Experts in the field of human factors trust or widespread corruption—develop and become design—a multidisciplinary field based on the core entrenched in a society. The findings apply not only OVERVIEW 3 to individuals in developing countries but also to can it occur individually? These examples may seem development professionals, who are themselves prone trivial in comparison to the challenges that govern- to error when decision-making contexts are complex. ments and international organizations face in develop- This approach expands the set of tools and strategies ing countries. Yet they hold an important lesson: when for promoting development and combating poverty. failure affects the profit-making bottom line, product The strength of standard economics is that it places designers begin to pay close attention to how humans human cognition and motivation in a “black box,” actually think and decide. Engineers, private firms, and intentionally simplifying the “messy and mysterious marketers of all stripes have long paid attention to the internal workings of actors” (Freese 2009, 98) by using inherent limits of human cognitive capacity, the role models that often assume that people consider all pos- that social preferences and the context play in our deci- sible costs and benefits from a self-interested perspec- sion making, and the use of mental shortcuts and men- tive and then make a thoughtful and rational decision. tal models for filtering and interpreting information. This approach can be powerful and useful, but in a The development community needs to do the same. number of contexts, it also has a liability: it ignores The body of evidence on decision making in devel- the psychological and social influences on behavior. oping country contexts is still coming into view, and Individuals are not calculating automatons. Rather, many of the emerging policy implications require people are malleable and emotional actors whose deci- further study. Nevertheless, this Report aims to inspire sion making is influenced by contextual cues, local and guide the researchers and practitioners who can social networks and social norms, and shared mental help discover the possibilities and limits of a new set models. All of these play a role in determining what of approaches. For example, can simplifying the enroll- individuals perceive as desirable, possible, or even ment process for financial aid increase participation? “thinkable” for their lives. The new tools based on this Can changing the timing of fertilizer purchases to full consideration of human factors do not displace coincide with harvest earnings increase the rate of existing policy approaches based on affecting self- use? Can providing a role model change a person’s interested personal incentives; rather, they comple- opinion of what is possible in life and what is “right” ment and enhance them. Some of the new approaches for a society? Can marketing a social norm of safe driv- cost very little to implement because they depend on ing reduce accident rates? Can providing information nuances in design or implementation, such as chang- about the energy consumption of neighbors induce ing the timing of cash transfers, labeling something individuals to conserve? As this Report will argue, the differently, simplifying the steps for service take-up, answers provided by new insights into human factors offering reminders, activating a latent social norm, or in cognition and decision making are a resounding reducing the salience of a stigmatized identity. Others yes (see, respectively, Bettinger and others 2012; Duflo, offer entirely new approaches to understanding and Kremer, and Robinson 2011; Beaman and others 2009, fighting poverty. 2012; Habyarimana and Jack 2011; Allcott 2011; Allcott These approaches are already widespread among and Rogers 2014). firms in the private sector, which are often preoccu- From the hundreds of empirical papers on human pied with understanding customer behavior in its nat- decision making that form the basis of this Report, ural contexts. When a company introduces a product, three principles stand out as providing the direction whether a new brand of breakfast cereal, toothpaste, for new approaches to understanding behavior and or cell phone, it is entering a competitive market, designing and implementing development policy. where small differences in usability and user satis- First, people make most judgments and most choices faction mean the difference between product take-up automatically, not deliberatively: we call this “thinking and rejection. In the intensive and interactive design automatically.” Second, how people act and think often phase, the company conducts significant qualitative depends on what others around them do and think: and quantitative research on its customers to under- we call this “thinking socially.” Third, individuals in a stand seemingly peripheral but nonetheless critical given society share a common perspective on making drivers of behavior: When and where do customers sense of the world around them and understanding typically eat breakfast? Are they at home, work, school, themselves: we call this “thinking with mental models.” on a bus, in a train, or in a car? What is the social To illustrate how all three types of thinking mat- meaning of the meal? Does it involve valued rituals? Is ter for development, consider the problems of low it a communal or more private event? Does behavior personal savings and high household debt, which are change need to be coordinated across many people or common across the developing world (and in many 4 WORLD DEVELOPMENT REPORT 2015 high-income countries, as well). Much of economic making comments like “we have neither a dream policy operates on the assumption that increasing sav- nor an imagination” or “we live only for today” (Ber- ings rates requires an increase in the rate of return for nard, Dercon, and Taffesse 2011, 1). In 2010, randomly savers. But other factors beyond the standard variables selected households were invited to watch an hour of of prices, incomes, and regulations also affect saving inspirational videos comprising four documentaries behavior, including automatic thinking that reacts of individuals from the region telling their personal to the framing and perception of choices, the wide- stories about how they had improved their socio- spread tendency to adhere to social norms, and the economic position by setting goals and working hard. mental models of one’s place in life. Field experiments Six months later, the households that had watched the in Kenya, South Africa, and Ethiopia demonstrate the inspirational videos had higher total savings and had relevance of these three principles of human decision invested more in their children’s education, on aver- making to a key development problem. age. Surveys revealed that the videos had increased In Kenya, many households report a lack of cash people’s aspirations and hopes, especially for their chil- as an impediment to investing in preventive health dren’s educational future (Bernard and others 2014). products, such as insecticide-treated mosquito nets. The study illustrates the ability of an intervention to However, by providing people with a lockable metal change a mental model—one’s belief in what is possi- box, a padlock, and a passbook that a household simply ble in the future (Bernard and Taffesse 2014). labels with the name of a preventive health product, The view that labeling, role models, and aspirations researchers increased savings, and investment in these can affect savings is not inconsistent with the view products rose by 66–75 percent (Dupas and Robinson that people respond in predictable ways to changes 2013). The idea behind the program is that although in interest rates or prices and other incentives. The money is fungible—and cash on hand can be spent new approaches do not replace standard economics. at any time—people tend to allocate funds through a But the new approaches enhance our understanding process of “mental accounting” in which they define of the development process and the way develop- categories of spending and structure their spending ment policies and interventions can be designed and behaviors accordingly. What was important about implemented. the metal box, the lock, and the labeled passbook was The mind, society, and behavior framework points to that it allowed people to put the money in a mental new tools for achieving development objectives, as account for preventive health products. The interven- well as new means of increasing the effectiveness of tion worked because mental accounting is one way in existing interventions. It provides more entry points which people are often “thinking automatically” and for policy and new tools that practitioners can draw is an example of a more general framing or labeling on in their efforts to reduce poverty and increase effect in which assigning something to a category shared prosperity. This Report discusses how taking influences how it is perceived. the human factors more completely into account in Conventional financial literacy programs in low- decision making sheds light on a number of areas: the income countries have had limited effects (Xu and persistence of poverty, early childhood development, Zia 2012). In contrast, a recent effort in South Africa household finance, productivity, health, and climate to teach financial literacy through an engaging tele- change. The framework and many examples in the vision soap opera improved the financial choices that Report show how impediments to people’s ability to individuals made. Financial messages were embedded process information and the ways societies shape in a soap opera about a financially reckless charac- mindsets can be sources of development disadvantage ter. Households that watched the soap opera for two but also can be changed. months were less likely to gamble and less likely to The three ways of thinking emphasized here apply purchase goods through an expensive installment plan equally to all human beings. They are not limited to (Berg and Zia 2013). The households felt emotionally those at higher or lower income levels, or to those at engaged with the show’s characters, which made them higher or lower educational levels, or to those in high- more receptive to the financial messages than would income or low-income countries. Numerous examples be the case in standard financial literacy programs. from high-income countries throughout this Report The success of the intervention depended on “think- demonstrate the universality of psychological and ing socially”—our tendency to identify with and learn social influences on decision making. The Report doc- from others. uments the cognitive limitations of people in all walks In Ethiopia, disadvantaged individuals commonly of life, including World Bank staff (see spotlight 3 and report feelings of low psychological agency, often chapter 10). Development professionals themselves OVERVIEW 5 think automatically, think socially, and think with modes, metaphorically, as two distinct systems in mental models and, as a result, may misidentify the the mind: System 1, the automatic system, and System 2, causes of behavior and overlook potential solutions to the deliberative system (Kahneman 2003). Chapter 1 development problems. Development organizations will discuss this division in more detail, but table O.1 could be more effective if practitioners became aware of their own biases and if organizations implemented Box O.1 The evolution of thinking in economics about procedures that mitigate their effects. human decision making For development practitioners, identifying psy- chological and social influences on behavior and Since the foundational work of Adam Smith ([1759, 1776] 1976), economists have constructing policies that work with them—rather explored psychological and social influences on human decision making. John than against them—require a more empirical and Maynard Keynes recognized “money illusion”—the tendency to think of money in experimental approach to policy design. Because nominal rather than in real terms—and used it in his proposed solution to unem- human decision making is so complicated, predicting ployment. He also recognized that many of our long-term investments reflect how beneficiaries will respond to particular interven- “animal spirits”—intuitions and emotions—not cool-headed calculation. Gunnar tions is a challenge. The processes of devising and Myrdal was a student of cultural stagnation. Herbert Simon and F. A. Hayek based implementing development policy would benefit from much of their work on the recognition that human beings can process only so richer diagnoses of behavioral drivers (see spotlight much information at once and are not capable of carefully weighing the costs and 4) and early experimentation in program design that benefits of every possible outcome of their decisions. Albert Hirschman argued anticipates failures and creates feedback loops that that it is useful to remember that people have complex motives; they value coop- allow practitioners to incrementally and continuously eration and loyalty. improve the design of interventions. However, in much of the 20th century, through the work of Paul Samuelson and many others, there was “a steady tendency toward the rejection of hedonistic, Three principles of human introspective, psychological elements” (Samuelson 1938, 344). Milton Friedman, in decision making his famous essay, “On the Methodology of Positive Economics” (1953), and others The organizing framework of part 1 of the Report rests in the 1950s argued persuasively, based on the evidence available at the time, that on three principles of human decision making: think- economists could safely ignore psychological factors in making predictions about ing automatically, thinking socially, and thinking with market outcomes. The individual economic actor could be understood as if he mental models. Although these principles are based behaved like a dispassionate, rational, and purely self-interested agent since indi- on recent groundbreaking research from across the viduals who did not behave that way would be driven out of the market by those social sciences, it is worth emphasizing that the new who did. The assumptions of perfect calculation and fixed and wholly self-regarding research, in some ways, brings the discipline of eco- preferences embedded in standard economic models became taken-for-granted nomics full circle to where it began, with Adam Smith beliefs in many circles. in the late 18th century, and to perspectives that were The past 30 years of research in decision making across many behavioral prominent in the early and middle parts of the 20th and social sciences have led economists to a stage where they measure and century (box O.1). formalize the psychological and social aspects of decision making that many of the foundational contributors to economics believed were important. Empirical First principle: Thinking automatically work demonstrates that people do not make decisions by taking into account In the simplifying assumptions employed in a number all costs and benefits. People want to conform to social expectations. People do of economic models, economic actors consider the full not have unchanging or arbitrarily changing tastes. Preferences depend on the universe of information and environmental cues and context in which they are elicited and on the social institutions that have formed look far into the future to make thoughtful decisions in the interpretive frameworks through which individuals see the world (Basu 2010; the present that will advance their fixed, long-term goals. Fehr and Hoff 2011). Of course, actual human decision making is almost Economics has thus come full circle. After a respite of about 40 years, an eco- never like this (see, for example, Gilovich, Griffin, and nomics based on a more realistic understanding of human beings is being rein- Kahneman 2002; Goldstein 2009). People typically have vented. But this time, it builds on a large body of empirical evidence—microlevel more information than they can process. There are an evidence from across the behavioral and social sciences. The mind, unlike a unmanageably large number of ways to organize the computer, is psychological, not logical; malleable, not fixed. It is surely rational to information that bears on almost any decision. treat identical problems identically, but often people do not; their choices change Thus psychologists have long distinguished two when the default option or the order of choices changes. People draw on mental kinds of processes that people use when thinking: models that depend on the situation and the culture to interpret experiences those that are fast, automatic, effortless, and associa- and make decisions. This Report shows that a more interdisciplinary perspective tive; and those that are slow, deliberative, effortful, on human behavior can improve the predictive power of economics and provide serial, and reflective. Psychologists describe the two new tools for development policy. 6 WORLD DEVELOPMENT REPORT 2015 high-interest debt people will choose to incur. For Table O.1 People have two systems of thinking some of the poorest individuals in many countries, the repeated use of small, short-term, unsecured loans is a Individuals have two systems of thinking—the automatic fact of life; these loans carry interest rates that would be system and the deliberative system. The automatic system influences nearly all our judgments and decisions. over 400 percent if multiplied over a year. Yet the high cost of these loans is often not obvious to borrowers. In Automatic system Deliberative system the United States, creditors called payday lenders offer Considers what automatically Considers a broad set of relevant a short-term loan until the next payday arrives. The comes to mind (narrow frame) factors (wide frame) cost of the loan is typically portrayed as a fixed fee per Effortless Effortful loan—say, $15 for every $100 borrowed for two weeks— Associative Based on reasoning rather than as an effective annual interest rate, or what Intuitive Reflective the cost would be if the loan were repeated over time. Sources: Kahneman 2003; Evans 2008. A field trial in the United States demonstrated the power of framing by testing an intervention that presented the cost of borrowing more transparently provides an overview. Most people think of themselves (Bertrand and Morse 2011). One group received the as primarily deliberative thinkers—but of course they standard envelope from the payday lender, which tend to think about their own thinking processes auto- includes the cash and the paperwork for the loan. The matically and under the influence of received mental envelope stated the amount due and the due date, as models about who they are and how the mind works. shown in figure O.2, panel a. Another group received In reality, the automatic system influences most of our a cash envelope that also showed how the dollar fees judgments and decisions, often in powerful and even accumulate when a loan is outstanding for three decisive ways. Most people, most of the time, are not months, compared to the equivalent fees for borrow- aware of many of the influences on their decisions. ing the same amount with a credit card (figure O.2, People who engage in automatic thinking can make panel b). Those who received the envelope on which what they themselves believe to be large and sys- the costs of the loan were reframed in accumulated tematic mistakes; that is, people can look back on the dollar amounts were 11 percent less likely to borrow choices they made while engaging in automatic think- from the payday lenders in the four months following ing and wish that they had decided otherwise. the intervention. The study captures a key implication Automatic thinking causes us to simplify problems of chapter 1, which is that adjusting what information and see them through narrow frames. We fill in miss- is provided, and the format in which it is provided, can ing information based on our assumptions about the help people make better decisions. world and evaluate situations based on associations that automatically come to mind and belief systems Second principle: Thinking socially that we take for granted. In so doing, we may form a Individuals are social animals who are influenced by mistaken picture of a situation, just as looking through social preferences, social networks, social identities, a small window overlooking an urban park could mis- and social norms: most people care about what those lead someone into thinking he or she was in a more around them are doing and how they fit into their bucolic place (figure O.1). groups, and they imitate the behavior of others almost The fact that individuals rely on automatic think- automatically, as shown in figure O.3. Many people ing has significant implications for understanding have social preferences for fairness and reciprocity and development challenges and for designing the best possess a cooperative spirit. These traits can play into policies to overcome them. If policy makers revise both good and bad collective outcomes; societies that their assumptions about the degree to which people are high in trust, as well as those that are high in cor- deliberate when making decisions, they may be able ruption, require extensive amounts of cooperation (see to design policies that make it simpler and easier for spotlight 1). Chapter 2 focuses on “thinking socially.” individuals to choose behaviors consistent with their Human sociality (the tendency of people to be desired outcomes and best interests. concerned with and associate with each other) adds For example, policy makers can help by paying close a layer of complexity and realism to the analysis attention to such factors as the framing of choices of human decision making and behavior. Because and the default options—an idea referred to as choice many economic policies assume individuals are self- architecture (Thaler and Sunstein 2008). The way that regarding, autonomous decision makers, these policies the cost of borrowing is framed can affect how much often focus on external material incentives, like prices. OVERVIEW 7 However, human sociality implies that behavior is also are cooperating. Figure O.4 shows the results of a influenced by social expectations, social recognition, “public goods game” that was played in eight countries. patterns of cooperation, care of in-group members, It demonstrates that although the proportion of condi- and social norms. Indeed, the design of institutions, tional cooperators versus free riders varies across coun- and the ways in which they organize groups and use tries, conditional cooperators are the dominant type in material incentives, can suppress or evoke motivation every one. In other words, in no society where this for cooperative tasks, such as community development behavior has been studied does the canonical theory of and school monitoring. economic behavior hold (Henrich and others 2001). People often behave as conditional cooperators—that Social preferences and social influences can lead is, individuals who prefer to cooperate as long as others societies into self-reinforcing collective patterns of Figure O.1 Automatic thinking gives us a partial view of the world To make most decisions and judgments, we think automatically. We use narrow framing and draw on default assumptions and associations, which can give us a misleading picture of a situation. Even seemingly irrelevant details about how a situation is presented can affect how we perceive it, since we tend to jump to conclusions based on limited information. 8 WORLD DEVELOPMENT REPORT 2015 Figure O.2 Reframing decisions can improve welfare: The case of payday borrowing a. The standard envelope A payday borrower receives his cash in an envelope. The standard envelope shows only a calendar and the due date of the loan. b. The envelope comparing the costs of the payday loan and credit card borrowing In a field experiment, randomly chosen borrowers received envelopes that showed how the dollar fees accumulate when a payday loan is outstanding for three months, compared to the fees to borrow the same amount with a credit card. How much it will cost in fees or interest if you borrow $300 PAYDAY LENDER CREDIT CARD (assuming two-week fee is $15 per $100 loan) (assuming a 20% APR) If you repay in: If you repay in: 2 weeks $45 2 weeks $2.50 1 month $90 1 month $5 2 months $180 2 months $10 3 months $270 3 months $15 Borrowers who received the envelope with the costs of the loans expressed in dollar amounts were 11 percent less likely to borrow in the next four months compared to the group that received the standard envelope. Payday borrowing decreased when consumers could think more broadly about the true costs of the loan. Source: Bertrand and Morse 2011. Note: APR = annual percentage rate. OVERVIEW 9 behavior. In many cases, these patterns are highly they can be very resistant to change. Social meanings desirable, representing patterns of trust and shared and norms, and the social networks that we are a part values. But when group behaviors influence indi- of, pull us toward certain frames and patterns of collec- vidual preferences and individual preferences com- tive behavior. bine into group behaviors, societies can also end Conversely, taking the human factor of sociality up coordinating activity around a common focal into account can help in devising innovative policy point that is ill-advised or even destructive for the interventions and making existing interventions more community. Racial or ethnic segregation and corrup- effective. In India, microfinance clients who were ran- tion are just two examples (spotlight 1). When self- domly assigned to meet weekly, rather than monthly, reinforcing “coordinated points” emerge in a society, had more informal social contact with one another Figure O.3 What others think, expect, and do influences our preferences and decisions Humans are inherently social. In making decisions, we are often affected by what others are thinking and doing and what they expect from us. Others can pull us toward certain frames and patterns of collective behavior. 10 WORLD DEVELOPMENT REPORT 2015 two years after the loan cycle ended, were more willing step was intended to promote water conservation, it to pool risks, and were three times less likely to default instead increased both water consumption and hoard- on their second loan (Feigenberg, Field, and Pande ing. Recognizing the problem, the city government 2013). In Uganda and Malawi, agricultural extension changed its communication strategy, sent around activities were much more successful when peer farm- volunteers to educate people about the most effective ers were used in training activities (Vasilaky and Leo- conservation measures, and began publicizing daily nard 2013; BenYishay and Mobarak 2014). Individuals water consumption figures and naming individuals generally want to repay their loans and to adopt better who were cooperating with the effort, as well as those technology, but they may have trouble motivating who were not. The mayor appeared in a television ad themselves to do it. By drawing on social motivations, taking a shower with his wife, explaining how the policy can help them reach their goals and protect tap could be turned off while soaping and suggesting their interests. taking showers in pairs. These strategies strengthened The case of a public emergency in Bogotá, Colom- cooperation, and reductions in water use persisted bia, illustrates how policy approaches can both under- long after the tunnel was repaired. mine and nurture cooperative behaviors (spotlight 5). The principle of thinking socially has several impli- In 1997, part of a tunnel providing water to the city cations for policy. Chapter 2 examines the scope for eco- collapsed, triggering a water shortage emergency. The nomic and social incentives in a world where human city government’s first action was to declare a public sociality is a major factor influencing behavior, shows emergency and initiate a communication program how institutions and interventions can be designed to warning inhabitants of the coming crisis. While this support cooperative behavior, and demonstrates how Figure O.4 In experimental situations, most people behave as conditional cooperators rather than free riders The standard economic model (panel a) assumes that people free ride. Actual experimental data (panel b) show that across eight societies, the majority of individuals behave as conditional cooperators rather than free riders when playing a public goods game. The model of free riding was not supported in any society studied. a. Behavior predicted b. Actual behavior revealed in experiments in standard economic model 100 Percentage of the population demonstrating 90 80 contributor behavior 70 60 50 40 30 20 10 0 Colombia Vietnam Switzerland Denmark Russian United Austria Japan Federation States Free riders Conditional cooperators Source: Martinsson, Pham-Khanh, and Villegas-Palacio 2013. Note: Other players did not fit into either of the two categories, which is why the bars do not sum to 100 percent. OVERVIEW 11 social networks and social norms shape behavior and mental models. Mental models affect what individuals can serve as the basis of new kinds of interventions. perceive and how they interpret what they perceive, as shown in figure O.5. There are mental models for how Third principle: Thinking with mental models much to talk to children, what risks to insure, what When people think, they generally do not draw on to save for, what the climate is like, and what causes concepts that they have invented themselves. Instead, disease. Many mental models are useful; others are not they use concepts, categories, identities, prototypes, and contribute to the intergenerational transmission stereotypes, causal narratives, and worldviews drawn of poverty. Mental models come from the cognitive from their communities. These are all examples of side of social interactions, which people often refer to Figure O.5 Thinking draws on mental models Individuals do not respond to objective experience but to mental representations of experience. In constructing their mental representations, people use interpretive frames provided by mental models. People have access to multiple and often conflicting mental models. Using a different mental model can change what an individual perceives and how he or she interprets it. 12 WORLD DEVELOPMENT REPORT 2015 as culture. Culture influences individual decision mak- the revelation evoked a sense of entitlement and “Why ing because it serves as a set of interrelated schemes try?” The simple presence of a stereotype can contrib- of meaning that people use when they act and make ute to measured ability differences, which in turn can choices. These schemes of meaning function like tools reinforce the stereotype and serve as a basis for distinc- for enabling and guiding action (DiMaggio 1997). tion and exclusion, in a vicious cycle. Mental models and social beliefs and practices Finding ways to break this cycle could increase the often become deeply rooted in individuals. We tend to well-being of marginalized individuals enormously. internalize aspects of society, taking them for granted Evidence from a number of contexts suggests that as inevitable “social facts.” People’s mental models invoking positive identities can counteract stereotypes shape their understanding of what is right, what is and raise aspirations. Having individuals contemplate natural, and what is possible in life. Social relations their own strengths has led to higher academic achieve- and structures, in turn, are the basis of socially con- ment among at-risk minorities in the United States, to structed “common sense,” which represents the evi- greater interest in antipoverty programs among poor dence, ideologies, and aspirations that individuals take people, and to an increase in the probability of finding for granted and use to make decisions—and which a job among the unemployed in the United Kingdom in some cases increase social differences. A body of (Cohen and others 2009; Hall, Zhao, and Shafir 2014; writing by anthropologists and other social scientists Bennhold 2013). points out that what people take to be hard evidence and common sense (their basic mental models of their Figure O.6 Cuing a stigmatized or entitled world and how it works) is often shaped by economic identity can affect students’ performance relationships, religious affiliations, and social group High-caste and low-caste boys from villages in India were identities (Bourdieu 1977; Kleinman 2006). Much of randomly assigned to groups that varied the salience of caste that work argues that achieving social change in a identity. When their caste was not revealed, high-caste and situation where mental models have been internalized low-caste boys were statistically indistinguishable in solving may require influencing not only the cognitive deci- mazes. Revealing caste in mixed classrooms decreased the performance of low-caste boys. But publicly revealing caste sion making of particular individuals but also social in caste-segregated classrooms—a marker of high-caste practices and institutions. entitlement—depressed the performance of both high-caste A canonical example of a mental model is a stereotype, and low-caste boys, and again their performance was statistically indistinguishable. which is a mental model of a social group. Stereotypes affect the opportunities available to people and shape 7 processes of social inclusion and exclusion. As a result of stereotypes, people from disadvantaged groups tend to underestimate their abilities (Guyon and Huillery 2014) 6 and may even perform worse in social situations when Number of mazes solved they are reminded of their group membership. In these 5 and other ways, the stereotypes can be self-fulfilling and can reinforce economic differences among groups (for example, see Ridgeway 2011 on gender stereotypes). 4 In India, low-caste boys were essentially just as good at solving puzzles as high-caste boys when caste 3 identity was not revealed, as shown in figure O.6. However, in mixed-caste groups, revealing the boys’ castes before puzzle-solving sessions created a signifi- 2 cant “caste gap” in achievement in which low-caste boys underperformed the high-caste boys by 23 per- 1 cent, controlling for other individual variables (Hoff and Pandey 2006, 2014). Making caste salient to the test takers invoked identities, which in turn affected perfor- 0 Caste not revealed Caste revealed, Caste revealed, mance. The performance of the stigmatized low-caste mixed-caste segregated boys declined relative to the performance of the high- groups groups caste boys. When caste was revealed to the high-caste High caste Low caste boys when they were not mixed with low-caste boys, the high-caste boys underperformed, perhaps because Source: Hoff and Pandey 2014. OVERVIEW 13 These considerations expand the toolkits of policy way improve individual well-being. Table O.2 presents makers in other ways, as well. An increasingly impor- examples of interventions based on a more realistic tant set of development interventions involves the understanding of human behavior that takes human media. Exposure to fiction, such as a serial drama, can factors into account. Drawing on insights from mod- change mental models (see spotlight 2 on entertain- ern behavioral and social sciences can generate new ment education). For example, when people living in kinds of interventions that can be highly cost effective. societies with high fertility were exposed to engaging An expanded understanding of human behavior can soap operas about families with few children, fertil- improve development policy. Whereas part 1 of this ity rates declined (Jensen and Oster 2009; La Ferrara, Report is organized by principles of human behavior, Chong, and Duryea 2012). part 2 is organized by development problems and illus- Shared mental models are persistent and can exert trates how these principles can be applied in a number a major influence on individual choices and aggregate of policy domains. social outcomes. Because mental models are somewhat malleable, interventions can target them to promote Poverty development objectives. Individuals have many differ- Poverty is not only a deficit in material resources but ent and competing mental models that they can bring also a context in which decisions are made. It can to bear on any situation; which one they use depends impose a cognitive burden on individuals that makes on which one the context activates. Policies that expose it especially difficult for them to think deliberatively individuals to new ways of thinking and alternative (Mullainathan and Shafir 2013). Individuals who must understandings of the world can expand the available exert a great deal of mental energy every day just to set of mental models and thus play an important role ensure access to necessities such as food and clean in development. water are left with less energy for careful deliberation than those who, simply by virtue of living in an area Psychological and social with good infrastructure and good institutions, can perspectives on policy instead focus on investing in a business or going to In many cases, a fuller understanding of human deci- school committee meetings. Poor people may thus be sion making can help societies achieve broadly shared forced to rely even more heavily on automatic decision goals like higher savings or better health and in this making than those who are not poor (chapter 4). Table O.2 Examples of highly cost-effective behavioral interventions Intervention Description Outcome Reminders Weekly text messages to remind patients to take their Adherence to a medical regimen HIV drugs in Kenya. Weekly reminders improved the rate of drug adherence to 53% from a baseline of 40%. Nonmonetary gifts Small nonfinancial incentives and prizes—like lentils Immunization rate and metal dinner plates—were combined with a reliable Among children aged 1–3, rates of full immunization were 39% with the immunization provider within the community in India. lentils incentives compared to 18% in the group with only the reliable immunization provision. In areas with no intervention, the rate of full immunization was 6%. Public notices Small stickers were placed in randomly selected buses Traffic accidents encouraging passengers to “heckle and chide” reckless Annual insurance claims rates for accidents declined from 10% to 5%. drivers in Kenya. Making products convenient Chlorine dispensers were provided free of charge at local Take-up of chlorination water sources, and promoters of chlorination to treat The take-up rate was 60% in households with dispensers, compared to water were hired to visit houses in Kenya. 7% for the comparison group. Inspirational messages Poor households were shown videos about how people Aspirations and investment like them had escaped from poverty or improved their Aspirations for children increased. Total savings and investments in socioeconomic status in Ethiopia. schooling were higher after six months. Timing of cash transfers Part of a conditional cash transfer was automatically saved Enrollment in higher education and given as a lump sum at the time when decisions about Enrollment increased in the next school year, without a decline in current school enrollment were made in Colombia. attendance. Sources: Pop-Eleches and others 2011; Banerjee and others 2010; Habyarimana and Jack 2011; Kremer and others 2009; Bernard and others 2014; Barrera-Osorio and others 2011. 14 WORLD DEVELOPMENT REPORT 2015 Sugar cane farmers in India, for example, typically and the deliberative system (for example, the ability to receive their income once a year, at the time of harvest. pay attention). Chapter 5 discusses these issues. The large income difference between just before the In all countries studied to date, whether low, mid- harvest and just after affects financial decision mak- dle, or high income, there is a divergence as early as ing. Right before the harvest, these farmers are much age three in the cognitive and noncognitive skills of more likely to have taken on loans and to have pawned children in households at the bottom of the national some of their belongings. This financial distress takes wealth distribution and those in households at the top. a toll on the cognitive resources that the farmers have The disparity stems in part from problems that policy available before harvest time (Mani and others 2013). can address. Farmers perform worse on the same series of cognitive The problem of insufficient stimulation to children tests before receiving their harvest income than after is of particular concern for low-income countries. A receiving their earnings—a difference in scores that is study of caregiving practices by mothers in 28 develop- equivalent to roughly 10 IQ points. In this sense, pov- ing countries found that socioemotional caregiving did erty imposes a cognitive tax. not vary widely by level of development. In contrast, the amount of cognitive stimulation that mothers provide is systematically lower in countries with lower Drawing on insights from modern measures of economic, health, and education variables, according to the United Nations Human Development behavioral and social sciences can Index (figure O.7). In this study, the level of cognitive stimulation was measured by the number of times generate new kinds of interventions that a caregiver read books, told stories, and engaged in naming, counting, or drawing with the child. When that can be highly cost effective. cognitive stimulation among infants is low, they have fewer and less sophisticated linguistic interactions, which can result in less facility with language and Development policy aimed at reducing or removing impede future scholastic achievement. the cognitive tax on poverty might seek to shift the Very early childhood stimulation has a large impact timing of critical decisions away from periods when on adult success in the labor market, a 20-year study cognitive capacity and energy (bandwidth) are predict- in Jamaica found (Gertler and others 2014). Commu- ably low (such as moving school enrollment decisions nity health workers made weekly home visits to teach closer to periods when income is higher) or targeting mothers how to play and interact with their children assistance to decisions that may require a lot of band- in ways that promote cognitive and emotional develop- width (such as choosing a health insurance plan or ment. Children who were randomly selected to partic- applying to a higher education program). ipate in the program earned 25 percent more as adults Psychological and anthropological research also than those in the control group who did not participate suggests that poverty generates a mental model in the program—enough to close the earnings gap with through which the poor see themselves and their a population that was not disadvantaged. opportunities. In particular, it can dull the capacity to imagine a better life (Appadurai 2004). Evidence Household finance also shows that interventions and policy designs that Making a good financial decision is difficult. It requires alter this mental model so that people can recognize individuals to understand the future cost of money, their own potential more easily—or that at least spare focus on gains and losses evenhandedly, resist the temp- poor people from reminders of their deprivation—can tation to consume too much, and avoid procrastinating. increase important development outcomes such as Recent behavioral and social insights demonstrate the school achievement, labor market participation, and difficulties involved, while also opening avenues for the take-up of antipoverty programs. policy makers to help individuals make decisions that serve their interests and achieve their goals (chapter 6). Child development High consumer debt often results from a form of High stress and insufficient socioemotional and cog- thinking automatically, in which individuals attach nitive stimulation in the earliest years, which tend to much more weight to current consumption through be associated with growing up poor, can impair the borrowing than to the loss of consumption that development of both the automatic decision-making will occur when they have to pay back a loan in the system (for instance, the ability to cope with stress) future. Certain types of financial regulation can help OVERVIEW 15 consumers frame their decisions about borrowing could identify the lowest-cost credit product using in a broader context that encompasses more than the a user-friendly summary sheet designed by the Con- prospect of immediate consumption. This kind of reg- sumer Financial Credit Bureau of Mexico (figure O.8, ulation helps individuals make financial decisions that panel b). they would likely prefer if they had thought delibera- Another set of interventions has focused on savings. tively about them rather than automatically. Some programs have helped individuals attain their An experiment with a low-income population in savings goals through the use of reminders that make Mexico shows how bandwidth constraints may limit the goals more salient. A series of studies in Bolivia, how people process financial information (Giné, Peru, and the Philippines show that simple, timely Martinez Cuellar, and Mazer 2014). Low-income indi- text messages reminding people to save improve viduals from Mexico City were invited to choose the savings rates in line with their goals (Karlan, Morten, best one-year, 10,000 peso loan product (that is, roughly and Zinman 2012). Other programs have helped indi- $800) from a randomized list of loan products resem- viduals increase their savings by offering commitment bling ones locally available. Individuals could earn devices in which consumers voluntarily give up access rewards if they identified the lowest-cost product. As to their savings until they meet a specified target level shown in figure O.8, panel a, only 39 percent of people of savings. When savings accounts were offered in could identify the lowest-cost product when presented the Philippines without the option of withdrawal for with brochures designed by banks for their customers. six months, nearly 30 percent of those offered the But a much larger fraction (68 percent) of individuals accounts accepted them (Ashraf, Karlan, and Yin 2006). Figure O.7 There is greater variation across countries in cognitive caregiving than in socioemotional caregiving Cognitive caregiving activities, shown by the dark bars, tend to be much greater in countries with high Human Development Indexes (HDI) than in countries with low HDI, although there are only slight differences in socioemotional activities (light bars) across countries. The height of the bars with babies on them indicates the average number of cognitive caregiving activities reported by parents in low- and high-HDI countries. Low HDI Medium HDI High HDI Average number of caregiving activities 3 2 1 0 mb go nt Gu e d he Af ea ire R au rra blic e Be d Ar Jam ize Re ica Mo ublic Vi lia Ky zb am Re an Ta ublic me an p. ng a h Se o Ma B rbia nia s ia K Alba R d H kh a eg n ina do ru Ba Ghan an aza ni on r es erz sta an Y Re eg o an iss yz ist Ye ist T o Ga To l ab a ,F ov ce ela etn lad Sie epu ng ral in ’lv Le ail Cô ia, en p p rg ek jik ric -B n, Th nt Mo t U ian r Sy sn Bo Ce Cognitive caregiving activities Socioemotional caregiving activities Source: Bornstein and Putnick 2012. Note: The bar graphs show the number of caregiving activities reported by mothers in the past three days, based on comparable data from 28 developing countries ranked by the United Nations Human Development Index (HDI). The three categories of cognitive caregiving activities measured were reading books; telling stories; and naming, counting, or drawing with the child.  16 WORLD DEVELOPMENT REPORT 2015 and thus miss achieving their own goals (the so-called Figure O.8 Clarifying a form can help borrowers find a intention-action divide). better loan product The gap between intentions and actions inspired an Low-income subjects from Mexico City were invited to classrooms to choose intervention that offered data entry workers in India the the cheapest one-year, $800 (10,000 peso) loan product from a set of five opportunity to select a contract in which each worker products representative of actual credit products offered by banks in Mexico City. They could earn rewards by getting the right answer. When using the banks’ could choose a target for the number of accurately descriptions of their products, only 39 percent of the people could identify the typed fields he or she entered. If a data entry worker cheapest credit product. When using the more straightforward summary sheet, achieved her target, she would be paid at the normal 68 percent could identify the cheapest credit. piece rate. If she missed her own target, however, she a. Bank leaflets b. Summary sheet would be paid at a lower rate. If people can simply do what they intend to do, there is no benefit to choosing this kind of contract because workers do not increase their pay if they meet the target, but lower their pay if they do not. But if workers recognize that there is a gap between intentions and actions, the commitment con- tract can serve a useful purpose. Because effort has a cost in the present and a reward in the future, individu- als may spend less time on effort than their deliberative minds would prefer. The commitment contract gives the individual an incentive to work harder than she might in the current moment when the work needs to be done. In the case of the data entry workers in India, about one-third chose the commitment contract— indicating that some of the workers themselves had a demand for commitment devices. The self-chosen commitment contracts did increase effort. Workers who opted for them increased their productivity by an amount equivalent to what would have been expected 39% of people could identify the cheapest 68% of people could identify the cheapest from an 18 percent increase in piece-rate wages (Kaur, loan product on the information leaflets loan product on a more straightforward from banks. summary sheet. Kremer, and Mullainathan 2014). The way that an identical level of pay is described = 10 people can also affect productivity. Take performance pay for teachers, in which teachers are paid a bonus at the end of the year that depends on the academic performance Source: Giné, Martinez Cuellar, and Mazer 2014. or improvement of their students. This kind of inter- vention failed to improve test scores in low-income neighborhoods in the U.S. city of Chicago (Fryer and After one year, individuals who had been offered and others 2012). Another variant of the program, how- had used the accounts increased savings by 82 percent ever, altered the timing of the bonuses and cast them more than a control group that was not offered such as losses rather than as gains. At the beginning of accounts. These and other studies show that psycho- the school year, teachers were given the amount that logical and social factors impede financial decision administrators expected the average bonus to be. If making and that interventions that target these factors their students’ performance turned out to be above can help individuals achieve financial goals. average at the end of the year, they would receive an additional payment. If student performance was below Productivity average, however, they would have to return the dif- Automatic thinking, social thinking, and thinking ference between what they received at the beginning with mental models also play a large role in worker and the final bonus they would have received had motivation and the investment decisions of farmers their students performed above average. This loss- and entrepreneurs (chapter 7). Even when monetary framed bonus improved test scores substantially. As incentives are strong, individuals may not exert the these examples suggest, well-designed interventions amount of effort that they intend, unless or until a that take into account people’s tendencies to think deadline or payday looms. For example, workers may automatically, socially, and with mental models can frame the decision to work at each moment narrowly improve productivity. OVERVIEW 17 Health rehydration therapy (ORT). While ORT saves lives by The decisions people make about their health and their preventing dehydration, it does not stop the symp- bodies emerge out of a tangle of information, the avail- toms of diarrhea, making the benefits less easy to per- ability and prices of health goods and services, social ceive. The Bangladesh Rural Advancement Committee norms and pressures, mental models of the causes of tackled the barriers to take-up of ORT by designing a disease, and willingness to try certain interventions. home-based approach, in which community health By recognizing this broad array of human factors, workers were employed to teach mothers how to make development policy involving health can in some cases ORT solutions at home in face-to-face social inter- dramatically improve its results (chapter 8). actions that explained the value of the therapy. This Consider the problem of open defecation. About and similar campaigns boosted the adoption of ORT in 1 billion people defecate in the open, and defecation Bangladesh and across South Asia. has been linked to infections in children that lead to Initiatives to increase the use of health products and stunted growth and in some cases death. A standard services often rely on subsidies, another area where approach is to provide information, along with goods psychological and social insights matter. Individuals at a subsidized cost—in this case, to construct toilets. may be willing to adopt and use health products if they But even with these changes in place, new sanitation are free but almost completely unwilling to use them norms are also needed to end this unhealthy practice. when prices are just above zero (Kremer and Glenner- Government officials in Zimbabwe developed “com- ster 2011). The reason is that prices for health products munity health clubs” to create community structures have many meanings in addition to the quantity of that served as a source of group endorsement for new payment required in an exchange. A product that is sanitation norms (Waterkeyn and Cairncross 2005). free triggers an emotional response, and it may convey A related approach to creating new norms with a social norm that everyone should be and will be using some promising anecdotal evidence is Community- it. Setting prices at zero, however, can promote waste Led Total Sanitation (CLTS). One core element of this if people take the product but do not use it. Research approach is that CLTS leaders work with community on this topic in developing countries is recent, but members to make maps of dwellings and the locations the emerging message is that if products are valuable where individuals defecate in the open. The facilitator enough to subsidize, there may be significant payoffs uses a repertoire of exercises to help people recognize to setting prices at zero and not just close to zero. the implications of what they have seen for the spread The choices of health care providers also arise of infections and to develop new norms accordingly. A from a complicated tangle of factors, including the recent and systematic study of CLTS in villages in India scientific information at their disposal, how much and and Indonesia provides evidence of the initiative’s how they are paid, and professional and social norms. value as well as its limitations. The CLTS programs Simply reminding providers of the social expectations were found to decrease open defecation by 7 and 11 surrounding their performance can improve it. For percent from very high levels in Indonesia and India, example, clinicians in urban Tanzania significantly respectively, compared to control villages. But where increased their effort when a visiting peer simply CLTS was combined with subsidies for toilet construc- asked them to improve their care (Brock, Lange, and tion, its impact on toilet availability within households Leonard, forthcoming), even though the visit conveyed was much higher. These findings suggest that CLTS no new information, did not change incentives, and can complement, but perhaps not substitute for, pro- imposed no material consequences. While developing grams that provide resources for building toilets (Patil and enhancing professional and social norms in health and others 2014; Cameron, Shah, and Olivia 2013). care is not simple and the same solution will not Mental models of the body also affect health work everywhere, there are many examples in which choices and behaviors. Beliefs about the causes of leadership has transformed social expectations and sterility, autism, and other conditions influence par- improved performance. ents’ decisions to vaccinate their children, as well as to adopt appropriate therapies. In India, 35–50 Climate change percent of poor women report that the appropri- Responding to climate change is one of the defining ate treatment for a child with diarrhea is to reduce challenges of our time. Poor countries and communi- fluid intake, which makes sense if the prevailing ties are generally more vulnerable to the effects of cli- mental model attributes the cause of diarrhea to mate change and will also bear significant costs during too much fluid (so the child is “leaking”) (Datta and transitions to low-carbon economies. Addressing cli- Mullainathan 2014). However, there is a low-cost and mate change requires individuals and societies not only extraordinarily successful therapy for diarrhea: oral to overcome complex economic, political, technological, 18 WORLD DEVELOPMENT REPORT 2015 and social challenges but also to get around a number procedures to mitigate them. Second, seemingly small of cognitive illusions and biases (chapter 9). Individu- details of design can sometimes have big effects on als ground their views of climate on their experience individuals’ choices and actions. Moreover, similar of recent weather. Ideological and social allegiances challenges can have different underlying causes; solu- can result in confirmation bias, which is the tendency of tions to a challenge in one context may not work in individuals to interpret and filter information in a man- another. As a result, development practice requires an ner that supports their preconceptions or hypotheses. iterative process of discovery and learning. Multiple Individuals tend to ignore or underappreciate informa- psychological and social factors can affect whether a tion presented in probabilities, including forecasts for policy succeeds; while some of these may be known seasonal rainfall and other climate-related variables. before implementation, some will not be. This means Human beings are far more concerned with the present that an iterative process of learning is needed, which than with the future, and many of the worst impacts of in turn implies spreading resources (time, money, and climate change could take place many years from now. expertise) over several cycles of design, implementa- People tend to avoid action in the face of the unknown. tion, and evaluation. Self-serving bias—the tendency of individuals to prefer principles, particularly principles regarding fairness, Development professionals that serve their interests—makes it hard to reach inter- While the goal of development is to end poverty, national agreements on how to share the burdens of development professionals are not always good at mitigating and adapting to climate change. predicting how poverty shapes mindsets. The WDR Psychological and social perspectives also expand 2015 team administered a randomized survey to exam- the menu of options for addressing climate change. ine judgment and decision making among World One option is to use policy to foster new habits of Bank staff. Although 42 percent of Bank staff predicted energy use. In a study of the effect of an eight-month that most poor people in Nairobi, Kenya, would agree period of compulsory electricity rationing in Brazil, with the statement that “vaccines are risky because evidence shows that the policy led to a persistent they can cause sterilization,” only 11 percent of the poor reduction in electricity use, with consumption 14 people sampled (defined in this case as the bottom percent lower even 10 years after rationing ended. third of the wealth distribution in that city) actually Household data on the ownership of appliances and agreed with the statement. Similarly, staff predicted on consumption habits indicate that a change in habits that many more poor residents of Jakarta, Indonesia, was the main reason for the decrease in consumption and Lima, Peru, would express feelings of helpless- (Costa 2012). ness and lack of control over their future than actu- An energy conservation program in the United ally did, according to the WDR 2015 team survey. States illustrates how social comparisons can also This finding suggests that development profession- influence energy consumption. The company running als may assume that poor individuals may be less the program, Opower, mailed “home energy reports” autonomous, less responsible, less hopeful, and less to hundreds of thousands of households; these reports knowledgeable than they in fact are. Beliefs like these compared a household’s electricity use to the amount about the context of poverty shape policy choices. It is used by others in the neighborhood in the same time important to check mental models of poverty against period. This simple information led to a 2 percent reduc- reality (chapter 10). tion in energy consumption, which was equivalent The WDR 2015 team survey also studied the ways to reductions resulting from short-term increases in in which ideological and political outlooks affect how energy prices of 11–20 percent and a long-term increase World Bank staff members interpret data. Survey of 5 percent (Allcott 2011; Allcott and Rogers 2014). respondents were presented with identical data in two different contexts and then were asked to identify the The work of development conclusion that the data best supported. One context professionals was politically and ideologically neutral: the question Recognizing the human factor in decision making and asked which of two skin creams was more effective. behavior has two interrelated repercussions for the The second context was more politically and ideolog- practice of development. First, experts, policy makers, ically charged: the question asked whether minimum and development professionals, like everyone else, are wage laws reduce poverty. The survey found that World themselves subject to the biases and mistakes that can Bank staff members were more likely to get the right arise from thinking automatically, thinking socially, answer in the skin cream context than in the minimum and using mental models. They need to be more aware wage context, even though the data were the same in of these biases, and organizations should implement both. One might be tempted to add that this occurred OVERVIEW 19 even though many World Bank staff members are To see the usefulness of this approach, consider the highly trained experts on poverty, but in reality this problem of diarrheal disease and some experiments occurred because World Bank staff members are highly implemented in Kenya to learn about cost-effective trained on that topic. Faced with a demanding calcula- methods to tackle it (Ahuja, Kremer, and Zwane 2010). tion, they interpreted new data in a manner consistent Bacteria-laden water is a major contributor to the bur- with their prior views, about which they felt confident. den of disease among children and can lead to lifelong This survey followed the line of inquiry developed by physical and cognitive impairment. Lack of access to Kahan and others (2013). clean water was diagnosed as a problem. Thus an early One way to overcome the natural limitations on intervention aimed at improving the infrastructure at judgment among development professionals may be households’ water sources, which are naturally occur- to borrow and adapt certain methods from industry. ring springs. The springs were susceptible to contam- Dogfooding is a practice in the technology industry in ination, such as fecal matter from the surrounding which company employees themselves use a product to experience it and discover its flaws. They work out its kinks before releasing it to the marketplace. Pol- Development professionals are icy designers could try to go through the process of signing up for their own programs, or trying to access themselves subject to the biases and existing services, as a way of diagnosing problems firsthand. Similarly, the practice of red teaming, used mistakes that can arise from thinking in both the military and the private sector, could help uncover weaknesses in arguments before big decisions automatically, thinking socially, and are made and programs are designed. In red teaming, an outside group has the role of challenging the plans, using mental models. They need to procedures, capabilities, and assumptions of an oper- ational design, with the goal of taking the perspective be more aware of these biases, and of potential partners or adversaries. Red teaming is based on the insight, from social psychology, that organizations should implement group settings motivate individuals to argue vigor- ously. Group deliberation among people who disagree procedures to mitigate them. but who share a common interest in finding the truth can divide cognitive labor efficiently, increase the odds that the best design will come to light, and mitigate the environment. To reduce contamination, the springs effects of “groupthink.” were covered with concrete so that water flowed from an above-ground pipe rather than seeping from the Adaptive design, adaptive interventions ground. While this measure considerably improved Because a number of competing factors may sway water quality at the source, it had only moderate decision making in a particular context and because effects on the quality of the water consumed at home development professionals themselves may be prone because the water was easily recontaminated while it to certain biases when assessing a situation, diagnosis was being carried or stored. and experimentation should be part of a continuous Thus the problem was redefined this way: house- process of learning (chapter 11). Institutional mech- holds did not adequately treat their water at home. anisms of development research and policy should Another iteration of experiments demonstrated that ensure space for sound diagnosis and for effective feed- providing free home delivery of chlorine or discount back loops for adapting programs that align with the coupons that could be redeemed in local shops elicited evidence gathered during implementation. This step high take-up of the water treatment product at first but might require changing institutional mental models failed to generate sustained results. People needed to and increasing an organization’s tolerance for failure. chlorinate their water when they returned home from In many cases, the initial diagnosis may be incorrect or the spring, and they needed to continue to go to the may be only partially successful. Only through imple- store to purchase the chlorine when their initial sup- mentation will this become clear. However, instead of plies ran out. penalizing failure or burying findings of failure, orga- These results suggested yet another diagnosis of nizations need to recognize that the real failures are the problem: households cannot sustain the use of policy interventions in which learning from experience water treatment over time. This led to the design does not happen. of free chlorine dispensers next to the water source, 20 WORLD DEVELOPMENT REPORT 2015 which made water treatment salient (the dispenser in the business of shaping individual choices? There served as a reminder right when people were thinking are three basic reasons, as discussed in spotlight 6. about water) and convenient (there was no need to First, shaping choices can help people obtain their make a trip to the store, and the necessary agitation own goals. Reminders to save or take medicine help and wait time for the chlorine to work automatically people who are otherwise caught up in life achieve occurred during the walk home). It also made water objectives that they themselves have set. Commitment treatment a public act, which could be observed by contracts, which markets underprovide, can reinforce whoever was at the spring at the time of water collec- decisions to adopt healthful behaviors. Matching the tion, allowing for social reinforcement of using water timing of social transfers to the timing of charges for treatment. These dispensers proved to be the most school enrollment, or making it easier to buy fertilizer cost-effective method for increasing water treatment at harvest time when cash is at hand, can help over- and averting diarrheal incidents (Abdul Latif Jameel come the divide between intentions and actions for Poverty Action Lab 2012). people who may be forgetful or possess insufficient willpower (that is to say, all of us). Many development policies that operate at the boundary of economics and Multiple behavioral and social factors can psychology can be understood in those terms. Second, individuals’ preferences and immediate affect whether a policy succeeds. Thus aims do not always advance their own interests. Individuals might choose differently, in ways more development practice requires an iterative consistent with their highest aspirations, if they had more time and scope for reflection. Third, socially process of discovery and learning, which reinforced practices and mental models can block choices that enhance agency and promote well-being implies spreading time, money, and and thus prevent individuals from even conceiving of certain courses of action—as when discrimination expertise over several cycles of design, can sometimes lead people, understandably, to adopt low aspirations. Governments should act when inad- implementation, and evaluation. equate engagement, situational framing, and social practices undermine agency and create or perpetuate poverty. Although development actors have legitimate Results like these, as well as the process of continu- differences on some of these issues and place different ous investigation used to establish them, are encourag- weights on individual freedoms and collective goals, ing. So is the realization that a more complete consid- widely shared and ratified human rights constitute a eration of the psychological and social factors involved guiding principle for addressing these trade-offs. in decision making may offer “low-hanging fruit”— Not every psychological or social insight calls for that is, policies with relatively large gains at relatively more government intervention; some call for less. low cost. But given that small changes in design and Because policy makers are themselves subject to cog- implementation can have large consequences for the nitive biases, they should search for and rely on sound success of an intervention, ongoing experimentation evidence that their interventions have their intended will be crucial. Analysis of existing or newly collected effects, and allow the public to review and scrutinize data and field observations will generate hypotheses their policies and interventions, especially those that that can inform the design of possible interventions. aim to shape individual choice. Still, it is not the case Multi-armed interventions—interventions that vary that when governments refrain from action, individ- a number of parameters, such as the frequency of uals freely and consistently make choices in their reminders or the method of rewarding effort—can shed own best interest, uninfluenced by anyone else. Any light on which ones are more effective in meeting the number of interested parties exploit people’s tendency social objective. The learning that takes place during to think automatically, succumb to social pressure, implementation should then feed back into redefining, and rely on mental models (Akerlof and Shiller, forth- rediagnosing, and redesigning programs in a cycle of coming), including moneylenders, advertisers, and continued improvement (figure O.9). elites of all types. In that context, governmental inac- Before policy makers launch initiatives to help indi- tion does not necessarily leave space for individual viduals with decision making, they should confront freedom; rather, government inaction may amount to a normative question: Why should governments be an indifference to the loss of freedom (Sunstein 2014). OVERVIEW 21 Figure O.9 Understanding behavior and identifying effective interventions are complex and iterative processes In an approach that incorporates the psychological and social aspects of decision making, the intervention cycle looks different. The resources devoted to definition and diagnosis, as well as to design, are greater. The implementation period tests several interventions, each based on different assumptions about choice and behavior. One of the interventions is adapted and fed into a new round of definition, diagnosis, design, implementation, and testing. The process of refinement continues after the intervention is scaled up. & diagnose fine & rediagnose Define Rede t ap Ad Design nt & evaluate leme Imp Source: WDR 2015 team. This Report seeks to accelerate the process of Allcott, Hunt, and Todd Rogers. 2014. “The Short-Run applying the new insights into decision making to and Long-Run Effects of Behavioral Interventions: development policy. The possibilities and limits of this Experimental Evidence from Energy Conservation.” American Economic Review 104 (10): 3003–37. doi: 10.1257 approach—based on viewing people more fully and /aer.104.10.3003. recognizing that a combination of psychological and Appadurai, Arjun. 2004. “The Capacity to Aspire: Culture social forces affects their perception, cognition, deci- and the Terms of Recognition.” In Culture and Public sions, and behaviors—are not yet completely known. Action, edited by Vijayendra Rao and Michael Walton, The research presented in the Report comes from an 59–84. Palo Alto, CA: Stanford University Press. active, exciting, and unsettled field. This Report is Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying only the beginning of an approach that could even- Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of tually alter the field of development economics and Economics 121 (2): 635–72. enhance the effectiveness of development policies and Banerjee, Abhijit Vinayak, Esther Duflo, Rachel Glenner- interventions. ster, and Dhruva Kothari. 2010. “Improving Immuni- sation Coverage in Rural India: Clustered Randomised References Controlled Evaluation of Immunisation Campaigns with and without Incentives.” BMJ 340. doi: http://dx Abdul Latif Jameel Poverty Action Lab. 2012. “Cleaner .doi.org/10.1136/bmj.c2220. Water at the Source.” J-PAL Policy Briefcase (Septem- Barrera-Osorio, Felipe, Marianne Bertrand, Leigh L. Linden, ber). http://www.povertyactionlab.org/publication and Francisco Perez-Calle. 2011. “Improving the Design /cleaner-water-source. of Conditional Transfer Programs: Evidence from a Ahuja, Amrita, Michael Kremer, and Alix Peterson Zwane. Randomized Education Experiment in Colombia.” 2010. “Providing Safe Water: Evidence from Random- American Economic Journal: Applied Economics 3 (2): 167–95. ized Evaluations.” Annual Review of Resource Economics Basu, Kaushik. 2010. Beyond the Invisible Hand: Groundwork 2 (1): 237–56. for a New Economics. Princeton, NJ: Princeton Univer- Akerlof, George A., and Robert Shiller. Forthcoming. sity Press. “Phishing for Phools.” Unpublished manuscript. Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, Allcott, Hunt. 2011. “Social Norms and Energy Conser- Rohini Pande, and Petia Topalova. 2009. “Powerful vation.” Journal of Public Economics 95 (9): 1082–95. doi: Women: Does Exposure Reduce Bias?” Quarterly Journal 10.1016/j.jpubeco.2011.03.003. of Economics 124 (4): 1497–1540. 22 WORLD DEVELOPMENT REPORT 2015 Beaman, Lori, Esther Duflo, Rohini Pande, and Petia Topa- Costa, Francisco. 2012. “Can Rationing Affect Long-Run lova. 2012. “Female Leadership Raises Aspirations and Behavior? Evidence from Brazil.” Unpublished man- Educational Attainment for Girls: A Policy Experiment uscript. http://idei.fr/doc/conf/bee2012/Costa_Ration in India.” Science 335 (6068): 582–86. ing_120831.pdf. Bennhold, Katrin. 2013. “Britain’s Ministry of Nudges.” New Datta, Saugato, and Sendhil Mullainathan. 2014. “Behav- York Times, Dec. 7. http://www.nytimes.com/2013/12/08 ioral Design: A New Approach to Development Policy.” /business/international/britainsministry-of-nudges Review of Income and Wealth 60 (1): 7–35. .html?pagewanted=all&_r. DiMaggio, Paul. 1997. “Culture and Cognition.” Annual BenYishay, Ariel, and A. Mushfiq Mobarak. 2014. “Social Review of Sociology 23 (1): 263–87. doi: 10.1146/annurev Learning and Communication.” Working Paper 20139, .soc.23.1.263. National Bureau of Economic Research, Cambridge, Duflo, Esther, Michael Kremer, and Jonathan Robinson. MA. 2011. “Nudging Farmers to Use Fertilizer: Theory and Berg, Gunhild, and Bilal Zia. 2013. “Harnessing Emotional Experimental Evidence from Kenya.” American Eco- Connections to Improve Financial Decisions: Evaluat- nomic Review 101 (6): 2350–90. ing the Impact of Financial Education in Mainstream Dupas, Pascaline, and Jonathan Robinson. 2013. “Why Media.” Policy Research Working Paper 6407, World Don’t the Poor Save More? Evidence from Health Sav- Bank, Washington, DC. ings Experiments.” American Economic Review 103 (4): Bernard, Tanguy, Stefan Dercon, Kate Orkin, and Alema- 1138–71. yehu Seyoum Taffesse. 2014. “The Future in Mind: Evans, Jonathan St. B. T. 2008. “Dual-Processing Accounts Aspirations and Forward-Looking Behaviour in Rural of Reasoning, Judgment, and Social Cognition.” Annual Ethiopia.” Working Paper, Centre for the Study of Review of Psychology 59 (January): 255–78. African Economies, University of Oxford. Fehr, Ernst, and Karla Hoff. 2011. “Introduction: Tastes, Bernard, Tanguy, Stefan Dercon, and Alemayehu Seyoum Castes and Culture: The Influence of Society on Pref- Taffesse. 2011. “Beyond Fatalism: An Empirical Explo- erences.” Economic Journal 121 (556): F396–412. ration of Self-Efficacy and Aspirations Failure in Ethio- Feigenberg, Benjamin, Erica Field, and Rohini Pande. pia.” Discussion Paper 01101, International Food Policy 2013. “The Economic Returns to Social Interaction: Research Institute, Washington, DC. http://www.ifpri Experimental Evidence from Microfinance.” Review of .org/sites/default/files/publications/ifpridp01101.pdf. Economic Studies 80 (4): 1459–83. Bernard, Tanguy, and Alemayehu Seyoum Taffesse. 2014. Freese, Jeremy. 2009. “Preferences.” In The Oxford Handbook “Aspirations: An Approach to Measurement with Val- of Analytical Sociology, edited by Peter Hedström and idation Using Ethiopian Data.” Journal of African Econo- Peter Bearman, 94–114. Oxford, U.K.: Oxford Univer- mies 23 (2): 189–224. doi: 10.1093/jae/ejt030. sity Press. Bertrand, Marianne, and Adair Morse. 2011. “Information Friedman, Milton. 1953. Essays in Positive Economics, Vol. 231. Disclosure, Cognitive Biases, and Payday Borrowing.” Chicago: University of Chicago Press. Journal of Finance 66 (6): 1865–93. Fryer, Roland G., Jr., Steven D. Levitt, John List, and Bettinger, Eric P., Bridget Terry Long, Philip Oreopoulos, Sally Sadoff. 2012. “Enhancing the Efficacy of Teacher and Lisa Sanbonmatsu. 2012. “The Role of Application Incentives through Loss Aversion: A Field Experiment.” Assistance and Information in College Decisions: Working Paper 18237, National Bureau of Economic Results from the H&R Block FAFSA Experiment.” Quar- Research, Cambridge, MA. terly Journal of Economics 127 (3): 1205–42. Gertler, Paul, James Heckman, Rodrigo Pinto, Arianna Bornstein, Marc H., and Diane L. Putnick. 2012. “Cognitive Zanolini, Christel Vermeersch, Susan Walker, Susan and Socioemotional Caregiving in Developing Coun- M. Chang, and Sally Grantham-McGregor. 2014. “Labor tries.” Child Development 83 (1): 46–61. Market Returns to an Early Childhood Stimulation Bourdieu, Pierre. 1977. Outline of a Theory of Practice. Cam- Intervention in Jamaica.” Science 344 (6187): 998–1001. bridge, U.K.: Cambridge University Press. Gilovich, Thomas, Dale Griffin, and Daniel Kahneman. Brock, J. Michelle, Andreas Lange, and Kenneth L. Leonard. 2002. Heuristics and Biases: The Psychology of Intuitive Judg- Forthcoming. “Generosity and Prosocial Behavior in ment. Cambridge, U.K.: Cambridge University Press. Health Care Provision: Evidence from the Laboratory Giné, Xavier, Cristina Martinez Cuellar, and Rafael and Field.” Journal of Human Resources. Keenan Mazer. 2014. “Financial (Dis-)Information: Cameron, Lisa, Manisha Shah, and Susan Olivia. 2013. Evidence from an Audit Study in Mexico.” Policy “Impact Evaluation of a Large-Scale Rural Sanitation Research Working Paper 6902, World Bank, Washing- Project in Indonesia.” Policy Research Working Paper ton, DC. 6360, World Bank, Washington, DC. Goldstein, Daniel G. 2009. “Heuristics.” In The Oxford Hand- Cohen, Geoffrey L., Julio Garcia, Valerie Purdie-Vaughns, book of Analytical Sociology, edited by Peter Hedström Nancy Apfel, and Patricia Brzustoski. 2009. “Recursive and Peter Bearman, 140–67. Oxford, U.K.: Oxford Uni- Processes in Self-Affirmation: Intervening to Close versity Press. the Minority Achievement Gap.” Science 324 (5925): Guyon, Nina, and Elise Huillery. 2014. “The Aspiration- 400–03. Poverty Trap: Why Do Students from Low Social Back- OVERVIEW 23 ground Limit Their Ambition? Evidence from France.” Martinsson, P., N. Pham-Khanh, and C. Villegas-Palacio. Working Paper, Department of Economics, Sciences 2013. “Conditional Cooperation and Disclosure in Po, Paris. Developing Countries.” Journal of Economic Psychology Habyarimana, James, and William Jack. 2011. “Heckle and 34: 148–55. Chide: Results of a Randomized Road Safety Interven- Mullainathan, Sendhil, and Eldar Shafir. 2013. Scarcity: Why tion in Kenya.” Journal of Public Economics 95 (11): 1438–46. Having Too Little Means So Much. New York: Times Books. Hall, Crystal C., Jiaying Zhao, and Eldar Shafir. 2014. “Self- Patil, S. R., B. F. Arnold, A. L. Salvatore, B. Briceno, S. Gan- Affirmation among the Poor: Cognitive and Behavioral guly, J. Colford Jr., and P. J. Gertler. 2014. “The Effect Implications.” Psychological Science 25 (2): 619–25. of India’s Total Sanitation Campaign on Defecation Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin Cam- Behaviors and Child Health in Rural Madhya Pradesh: erer, Ernst Fehr, Herbert Gintis, and Richard McEl- A Cluster Randomized Controlled Trial.” PLoS Med 11 (8): reath. 2001. “In Search of homo economicus: Behavioral e1001709. doi: 10.1371/journal.pmed.1001709. Experiments in 15 Small-Scale Societies.” American Pop-Eleches, Cristian, Harsha Thirumurthy, James Hab- Economic Review 91 (2): 73–78. yarimana, Joshua Graff Zivin, Markus Goldstein, Hoff, Karla, and Priyanka Pandey. 2006. “Discrimination, Damien de Walque, Leslie Mackeen, Jessica Haberer, Social Identity, and Durable Inequalities.” American John Sidle, and Duncan Ngare. 2011. “Mobile Phone Economic Review 96 (2): 206–11. Technologies Improve Adherence to Antiretroviral ————. 2014. “Making Up People: The Effect of Identity Treatment in Resource-Limited Settings: A Random- on Performance in a Modernizing Society.” Journal of ized Controlled Trial of Text Message Reminders.” Development Economics 106: 118–31. AIDS 25 (6): 825–34. Jensen, Robert, and Emily Oster. 2009. “The Power of TV: Ridgeway, Cecilia L. 2011. Framed by Gender: How Gender Cable Television and Women’s Status in India.” Quar- Inequality Persists in the Modern World. Oxford, U.K.: terly Journal of Economics 124 (3): 1057–94. Oxford University Press. Kahan, Dan M., Ellen Peters, Erica Cantrell Dawson, and Samuelson, Paul A. 1938. “The Empirical Implications of Paul Slovic. 2013. “Motivated Numeracy and Enlight- Utility Analysis.” Econometrica 6 (4): 344–56. ened Self-Government.” Working Paper, Cultural Cog- Smith, Adam. (1759) 1976. The Theory of Moral Sentiments. nition Project, Yale Law School, New Haven, CT. London: Millar. Glasgow Bicentenary Edition, edited Kahneman, Daniel. 2003. “Maps of Bounded Rationality: by D. D. Raphael and A. L. Macfie. Oxford, U.K.: Oxford Psychology for Behavioral Economics.” American Eco- University Press. nomic Review 93 (5): 1449–75. —— — —. (1776) 1976. An Inquiry into the Nature and Causes of Karlan, Dean, Melanie Morten, and Jonathan Zinman. the Wealth of Nations. London: Millar. Glasgow Bicente- 2012. “A Personal Touch: Text Messaging for Loan nary Edition, edited by D. D. Raphael and A. L. Macfie. Repayment.” Working Paper 17952, National Bureau of Oxford, U.K.: Oxford University Press. Economic Research, Cambridge, MA. Sunstein, Cass. 2014. Why Nudge? The Politics of Libertarian Kaur, Supreet, Michael Kremer, and Sendhil Mullaina- Paternalism. New Haven, CT: Yale University Press. than. 2014. “Self-Control at Work.” Working Paper, Thaler, Richard H., and Cass R. Sunstein. 2008. Nudge: Harvard University, Cambridge, MA. Improving Decisions about Health, Wealth, and Happiness. Kleinman, Arthur. 2006. What Really Matters: Living a New Haven, CT: Yale University Press. Moral Life amidst Uncertainty and Danger. Oxford, U.K.: Vasilaky, Kathryn, and Kenneth L. Leonard. 2013. “As Good Oxford University Press. as the Networks They Keep? Improving Farmers’ Social Kremer, Michael, and Rachel Glennerster. 2011. “Improv- Networks via Randomized Information Exchange in ing Health in Developing Countries: Evidence from Rural Uganda.” Working Paper, University of Mary- Randomized Evaluations.” In Handbook of Health Eco- land, College Park. nomics, Vol. 2., edited by Thomas G. McGuire, Mark Waterkeyn, Juliet, and Sandy Cairncross. 2005. “Creating V. Pauly, and Pedro P. Barros, 201–315. Amsterdam: Demand for Sanitation and Hygiene through Com- Elsevier. munity Health Clubs: A Cost-Effective Intervention in Kremer, Michael, Edward Miguel, Sendhil Mullainathan, Two Districts in Zimbabwe.” Social Science and Medicine Clair Null, and Alix Peterson Zwane. 2009. “Making 61 (9): 1958–70. Water Safe: Price, Persuasion, Peers, Promoters, or Wiener, E. L., and D. C. Nagel. 1988.  Human Factors in Product Design.” Unpublished manuscript. Aviation. London: Academic Press. La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. Xu, Lisa, and Bilal Zia. 2012. “Financial Literacy in the 2012. “Soap Operas and Fertility: Evidence from Brazil.” Developing World.” Policy Research Working Paper American Economic Journal: Applied Economics 4 (4): 1–31. 6107, World Bank, Washington, DC. Mani, Anandi, Sendhil Mullainathan, Eldar Shafir, and Jiaying Zhao. 2013. “Poverty Impedes Cognitive Func- tion.” Science 341 (6149): 976–80. PART 1 An expanded understanding of human behavior for economic development: A conceptual framework Introduction This first part of this Report presents a framework representations of experience constructed from for understanding and using recent findings on culturally available mental models. People have human decision making. The three chapters in access to multiple and often conflicting mental part 1 develop the three elements of the framework: models, and which one they invoke to make a choice depends on the context. Human 1. Thinking automatically. Much of our thinking is decision making, therefore, is powerfully shaped automatic, not deliberative. It is based on what by both contextual cues and the past experi- effortlessly comes to mind. In contrast to stan- ences of individuals and societies. Showing dard assumptions that we perform complex people new ways of thinking can expand the calculations and consider all possible routes set of mental models they draw on and their of action, humans reach for simple solutions capacity to aspire and can thus increase social and use mental shortcuts much of the time. welfare. Thus minor situational changes can have a large impact on behavior and, ultimately, on the These three elements of the framework are of achievement of development goals. Simplifying first-order importance for development policy, the choice environment can help people make poverty alleviation, and the policy design pro- choices and enact behaviors that benefit them. cess itself. These elements have two important implications: 2. Thinking socially. Humans are not autonomous thinkers or decision makers but deeply social • “Economic man” is a fiction, not a reality. Policies animals. We have innate preferences for altru- that assume that rational decision making will ism, cooperation, and reciprocity, and we are always prevail can go astray in many contexts strongly affected by the social norms and net- and may miss opportunities for low-cost, high- works in our communities. We often want to efficacy interventions. Updating the standard meet others’ expectations of us, and we act on assumptions about human decision making is the basis of shared identities. Recognizing the essential to pushing forward the frontier of devel- importance of social preferences and norms opment policy making. in decision making can help policy makers improve program efficacy and develop new • The interplay of institutions and individuals tools for achieving development objectives. is more complex than is often recognized; yet the potential for temporary interventions and 3. Thinking with mental models. Individuals do not changes in institutions to alter long-standing respond to objective experience but to mental patterns is greater than has been recognized. 25 CHAPTER 1 Thinking automatically Two systems of thinking of thinking is effortless, fast, and largely outside vol- To make a judgment or decision, individuals simplify untary control. The mental reserves for this kind of the problem. They construct a representation in their cognitive activity are vast. This Report refers to this heads and then reach a judgment or decision based on mode as thinking automatically (the automatic system). that simplification. There is a broad consensus in psy- The two systems are also called System 1 (automatic chology that to do this, people use two systems of think- system) and System 2 (deliberative system) (Stanovich ing. Sometimes, they think in a way that is deliberative, and West 2000; Kahneman 2003) (see table 1.1). reflective, and effortful—as when solving a difficult The psychologists Daniel Kahneman and Amos Tver- math problem or in trying to overcome an impulse in sky established that people tend to rely on the automatic acts of self-control. This type of thinking is hard. It is system to make decisions. People evaluate alternatives cognitively taxing and can be exhausting. Our capacity quickly, based on what automatically comes to mind.1 to engage in it is limited. It is difficult to spend even People rarely, if ever, consider all alternatives. Although a few minutes focusing attention in a concentrated often perfectly capable of more careful analysis, people manner. This Report refers to this way of thinking as are hard wired to use just a small part of the relevant thinking deliberatively (the deliberative system). information to reach conclusions. By observing mental processes under controlled experimental conditions, Kahneman and Tversky developed a new understand- We normally think of ourselves in terms ing of human action that helped lay the foundation for the field of behavioral economics—a subfield of econom- of the deliberative system—the conscious ics that draws on the psychological, social, and cultural foundations of human decision making. reasoning self—yet automatic operations Their work dispelled a central cognitive illusion. We normally think of ourselves in terms of the delib- generate complex patterns of ideas that erative system—the conscious reasoning self—yet, in fact, the automatic operations of thinking generate influence nearly all our judgments and complex patterns of ideas that influence nearly all our judgments and decisions. In a recent book, Kahneman decisions. (2011) compares the deliberative system to a supporting character in a play who believes herself to be the hero. The automatic and deliberative systems interact. Most of the time, we use another mode of thinking, The automatic system effortlessly generates impres- with relatively little interference from the deliberative sions and feelings that are the main sources of the system. When we detect anger in the image of a face explicit beliefs and reflective choices of the deliberative or make sense of speech in a fraction of a second, our system. In routine situations, we use the automatic minds are operating in automatic mode. This mode system without much oversight from the deliberative THINKING AUTOMATICALLY 27 system, unless the deliberative system is provoked to ing information, default assumptions and other check it. “mental models” that individuals bring to a problem To see how lightly the deliberative system regulates influence what they pay attention to and how they the automatic system, consider this problem: a bat and interpret what they perceive. Framing in this sense ball cost $1.10. The bat costs $1.00 more than the ball. is a part of decision making. How much does the ball cost? Most people answer “10 cents,” since $1.10 can be easily broken into a sum The first meaning of framing concerns what is done $1 and 10 cents. The automatic system provides a plau- to the decision maker: for example, putting in bold sible response, based on what comes quickly to mind, letters that a payday loan costs $15 for two weeks and before the deliberative system has time to intervene leaving to the small print the fact that the annual inter- and regulate our judgment. The correct answer is est rate is 400 percent. The second meaning of framing 5 cents (since $0.05 + $1.05 = $1.10).2 concerns what the decision maker does. When individuals are under cognitive strain, it is Figure 1.1 depicts an individual looking through a even more difficult to activate the deliberative system. window frame. The frame provides only a very narrow Poverty, time pressure, and financial stress all can view of an urban scene that leads the viewer to imag- cause cognitive strain (see chapter 4). Sugar cane farm- ine it as a park. The figure captures a central feature ers in India offer an example of how financial distress of automatic thinking: what our attention is drawn to can deplete mental resources. The farmers typically and what we focus on are not always the things most receive their income once a year, at the time of har- needed for good decision making. vest. Just before the harvest, 99 percent of the farmers Development practitioners are increasingly using have incurred loans. Just after the harvest, they have the idea of dual-system thinking to address problems received most of the earnings for the season and only of poverty and development, as this World Development 13 percent of farmers are indebted. Their financial dis- Report will discuss. Since people may be powerfully tress before the harvest takes a measurable toll on their influenced by the way that options are described, cognitive resources. Before receiving their harvest simple changes in descriptions of options can some- income, farmers perform worse on a series of cognitive times change behavior. Policies that make it easier to tests than when they take the same tests after receiv- reach the right decisions can sometimes boost welfare ing their earnings, a gap that cannot be explained by substantially and at low cost. This is especially impor- differences before and after harvest in nutrition, physi- tant for individuals living in poverty, as chapter 4 will cal exhaustion, biological stress, or learning. The differ- show. If policy can change which frame people use for ence in scores is roughly equivalent to three-quarters a decision, it can in some cases change the decisions of the cognitive deficit associated with losing an entire they make. night’s sleep (Mani and others 2013). A second broad policy implication of our reliance The idea that people have two systems of thinking is on automatic thinking is the limited power of merely not new and has been anticipated in the work of many providing information. Confirmation bias is the ten- psychologists and philosophers over the centuries dency to automatically interpret information in ways (Frankish and Evans 2009). However, research over that support prior beliefs (Dawson, Gilovich, and the past four decades has vastly expanded our under- Regan 2002). Confirmation bias gives rise to biased standing of the implications for development and, more broadly, for economic policy. One central impli- cation is the power of framing. The term frame applies Table 1.1 People have two systems of to descriptions of decision problems at two levels thinking (Kahneman and Tversky 2000, xiv): Individuals have two systems of thinking—the automatic system and the deliberative system. The automatic system • Description and presentation. The formulation to influences nearly all our judgments and decisions. which decision makers are exposed is called a frame. Automatic system Deliberative system A frame in this sense is the way choices are described Considers what automatically Considers a broad set of relevant and presented. comes to mind (narrow frame) factors (wide frame) • “Mental editing” and interpretation. A frame is also the Effortless Effortful interpretation that decision makers construct for themselves, based on the way they mentally edit Associative Based on reasoning and interpret the information they receive. When Intuitive Reflective situations are complex or ambiguous or entail miss- Sources: Kahneman 2003; Evans 2008. 28 WORLD DEVELOPMENT REPORT 2015 information search, as well. As novelist Jane Austen Persuasion and education must engage with the auto- once wrote, “We each begin probably with a little bias matic system to overcome resistance to new points of and upon that bias build every circumstance in favor of view (see spotlight 2 on entertainment education). This it.”3 Confirmation bias contributes to overconfidence is old news to political consultants and advertisers, and in personal beliefs. People may fail to recognize that policy makers have also surely discovered it from their they do not know what they claim to know, and they own experience. may fail to learn from new information (see chapter 10 This chapter offers a synthesis of the scientific for a discussion of how these biases affect development evidence on the power of the automatic system to professionals and a survey experiment that explores produce systematic behavioral biases. Thirty years ago, possible confirmation bias among World Bank staff). people might reasonably have viewed the findings of Figure 1.1 Framing affects what we pay attention to and how we interpret it To make most decisions and judgments, we use narrow framing and draw on default assumptions and associations, which can give us a misleading picture of the situation. Even seemingly irrelevant details of how a situation is presented can affect our perceptions, since we tend to jump to conclusions based on limited information. THINKING AUTOMATICALLY 29 behavioral economics as a few anomalies. “Sometimes, sions and the failure to adhere to health regimens, take some people are loss averse,” the narrative might have health precautions, and adopt income-increasing tech- gone, “but I don’t behave like that. And it certainly would niques after receiving new information—economists be naïve to design policy based on this assumption.” But have come to recognize the importance of considering over the past few decades, evidence has mounted that the possible impacts on behavior of our dual system automatic thinking cuts across wide swathes of human of thinking, automatic and deliberative, in the design behavior to the point that it can no longer be ignored. and testing of policy. As shown in figure 1.2, panel b, The anomalies that behavioral economics is trying to a more behavioral model of decision making entails explain are not minor and scattered. They are system- several departures from the standard economic model, atic regularities that can be of first-order importance of which two are among the most relevant for policy for health, child development, productivity, resource making: allocation, and the process of policy design itself. The analytical foundations of public policy have • People may process only the information that is traditionally come from standard economic theory. most salient to them, which may lead them to miss In standard economic theory, an important behav- key information and overlook critical consequences. ioral assumption is that people use information in an • There may be a mismatch between intentions and unbiased way and perform careful calculations. The actions (the intention-action divide). Even if people calculations allow them to make choices based on an understand the full consequences of their actions, unbiased consideration of all possible outcomes of they may make decisions that favor the present at alternative choices that might be made. After people the expense of the future, so that they consistently make a choice and observe the outcome, they use the fail to carry out plans that match their goals and ful- information in an unbiased way to make the next fill their interests. decision, and so on. Figure 1.2, panel a, represents this idealized process. Biases in assessing information But confronted with the mounting empirical The world is awash with information, most of which evidence on large and costly errors that people often is irrelevant to any particular decision. When deciding make in critical choices—such as poor financial deci- what to eat for lunch, we must consider how much Figure 1.2 A more behavioral model of decision making expands the standard economic model In the standard economic model (panel a), decision makers use information in an unbiased way and deliberate carefully about all choices and possible consequences. In a more behavioral model (panel b), decision makers may overlook some relevant information because they think automatically as well as deliberatively. a. Standard economic model b. Model of the psychological and social actor ut do ere filt Deliberation Two systems is ation without limit of thinking Some inform Automatic system n n tio tio ma ma Deliberative system Infor Infor Economic Economic outcomes and outcomes Behavior shared mental models Behavior Source: WDR 2015 team. 30 WORLD DEVELOPMENT REPORT 2015 money we have. There are, however, a myriad of things a party just before the accident, the two groups were we are unlikely to find useful to consider, such as the given the same testimony. One group heard the first color of our shirt. When people think about what to get line, below, and the other heard the second line: for lunch, they do not first consider the color of their shirt and then decide it is irrelevant. Shirt color never Version 1: On his way out the door, Sanders staggered enters their deliberative system because their auto- against a serving table, knocking a bowl to the floor. matic system has already decided that it is not impor- tant. And so the individual uses no cognitive energy to Version 2: On his way out the door, Sanders staggered think about it. against a serving table, knocking a bowl of guaca- The automatic system is relying on a framework of mole dip to the floor and splattering guacamole on understanding—a frame, in short—to organize experi- the white shag carpet. ence and distinguish between the things one needs to consider and the things one can ignore. Most frames Did the two groups of students reach a different are adaptive. People could not accomplish anything, or judgment? Should they have? They did, but they should even survive, if they did not have some type of frame in not have, since the information about what was in the place and use some mental shortcuts. A radically sim- bowl was arguably irrelevant to Mr. Sanders’ possible plified set of frames and mental shortcuts can perform drunkenness.4 But those who heard the additional admirably well in many cases (Todd and Gigerenzer detail about the guacamole were more likely to believe 2000). However, sometimes frames lead people to that he was guilty (Reyes, Thompson, and Bower 1980). ignore what is necessary for optimal decision making. A natural interpretation is that the information about the guacamole made the incident more salient. A piece of information is salient when it stands out against Seemingly minor and low-cost policy other pieces of information. Even though students were actively thinking about whether Mr. Sanders was drunk changes may have a large impact on the and attempting to weigh the evidence objectively, their automatic system may have been “telling” some stu- achievement of development goals and dents that this piece of information was decisive.5 Given the role of salience, it will come as no sur- the reduction of poverty. prise that the way in which facts are presented has a great influence on whether they are absorbed and how judgments are reached. What matters is not only the Even if one has the tools of the deliberative system entire set of available information and how each piece with which to assess evidence carefully and accurately, might be logically weighed, but also the sequencing of the automatic system may bias the information that information and the psychological salience of differ- the deliberative system is using. ent types of information. The term for the ease with Shirt color is not usually a relevant factor. However which mental content comes to mind is accessibility on occasion, it might be: for instance, a white shirt that (Kahneman 2003). Automatic thinking is shaped by we did not want to stain. The next sections examine the accessibility of different features of the situation. the biases in judgment that result when relevant Seemingly unimportant features of the context of deci- factors are overlooked. Chapters 2 and 3 will link this sion making—how many choices one must make sense problem to social change, a fundamental aspect of of, whether it resonates with us emotionally, whether development. it activates events in recent memory—can all affect accessibility and therefore judgment (and behavior). Framing When making decisions, people may give greater weight Anchoring than they should to information that has limited, if any, An anchor is an aspect of the environment that has relevance. Consider the case of Mr. Sanders, who ran no direct relevance to a decision but that nonetheless a stop sign while driving and collided with a garbage affects judgments. Anchoring is an extreme example truck. He was accused of being drunk while driving and of automatic thinking. For example, sometimes the last was being tried. Two groups of students were asked to thing that comes to mind has a disproportionate influ- judge Mr. Sanders’ guilt or innocence in a mock jury. ence on decision making. Sometimes the anchor will be Except for the description of Mr. Sanders’ behavior at obvious and appropriate, as in the case of comparison THINKING AUTOMATICALLY 31 shopping. But sometimes the anchor will be inappro- When the dating question came first, the answers priate; the automatic system is grabbing onto anything to the questions were highly correlated, but when it it can to help it in its interpretation of a choice context. was asked second, the responses were uncorrelated. Even subliminal anchors can affect judgment. Evidently, the first question was an anchor for the Consider an experimental study of experts in the response to the second question. The anchor automat- field of law. Experienced jurists participated in a study ically evoked thoughts that affected individuals’ judg- of sentencing decisions (Englich, Mussweiler, and ment about whether or not dating affected happiness. Strack 2006). All the jurists, who were either judges To a surprising degree, the quality of decisions that or experienced lawyers, read a description of a crim- an individual makes in life (like the quality of answers inal case that could end in a jail sentence of up to one to subjective questions on surveys) depends on the year. They were asked what sentence they would hand anchors that happen to be present. Policy makers down, given the facts of the case. Some were told that increasingly take heed of this fact. A change in context a newspaper article had speculated that the sentence that makes one comparison (such as one number, one would be three months, while others were told that fact, one experience, one competitor, or one role model) an article had speculated that the sentence would be particularly salient can change what people choose and nine months. Those jurists given the larger anchor whether a government program is taken up. The power gave significantly longer sentences than those given of framing and anchoring is illustrated by consumers’ the smaller anchor. In a companion study, the anchor decisions in the credit market, discussed next. came not from a newspaper report but from the roll of a pair of dice, rigged to come out three or nine Application: Consumer decisions in when they were rolled in front of the jurist. Again, credit markets the high anchor produced longer sentences than the People in financial distress may resort to borrowing at low one. This finding has been replicated in dozens of extremely high interest rates. This practice has been a experiments. long-standing concern in fighting poverty. Appropri- You can confirm the importance of anchoring ate policy remedies based on standard models would effects by a simple experiment. Ask people to compute, assume that choices are careful and consistent and within five seconds, the product of the numbers one therefore would focus on reducing the risks that the through eight, either as 1 ‫ ן‬2 ‫ ן‬3 ‫ ן‬4 ‫ ן‬5 ‫ ן‬6 ‫ ן‬7 ‫ ן‬8 poor face (and hence the risk of financial distress) and or reversed as 8 ‫ ן‬7 ‫ ן‬6 ‫ ן‬5 ‫ ן‬4 ‫ ן‬3 ‫ ן‬2 ‫ ן‬1. Because on improving the terms on which the poor can borrow your respondents will not have enough time to calcu- (and hence the opportunities to escape distress). But late the full set of products, they will have to estimate the implication of the findings from psychology and the answer. You will almost certainly find that when behavioral economics is that there are additional tar- the sequence starts with small numbers, individuals gets of policy; that is, policy makers can try to improve will estimate the product to be smaller than when the the quality of the decisions that people make that lead sequence starts with big numbers. This experiment to distress or that perpetuate distress. Recent field has been done rigorously (Montier 2007). When the trials among low-income populations in the United sequence started with the small numbers, the median States and Mexico demonstrate the potential for very estimate was 512. When the sequence started with the simple policies to improve financial decision making. large numbers, the median estimate was 2,250. (The correct answer is 40,320.) People jumped to conclu- A field trial on payday borrowing sions based on a very partial view of the problem. In many countries, some of the poorest individuals The power of anchors has implications for survey resort to payday borrowing, for which they incur design and analysis. A prior question, or the inclusion extremely high interest costs. Payday loans (also called of some specific candidate answers in a multiple-choice payday advances) are small, short-term, unsecured question, can influence what information an individ- loans that anyone with a payroll record can normally ual retrieves: this is the automatic system at work. As obtain. Many payday borrowers have no access to alter- an illustration, consider a survey that included these native sources of funds—this is the last resort. For those two questions about personal happiness asked in two individuals, the choice is thus not from whom to bor- different orders (Schwartz, Strack, and Mai 1991): row, but only whether to borrow and, if so, how much. A field trial of payday borrowing in the United States Α. “How happy are you with life in general?” tried to remedy the factors that could potentially lead Β. “How often do you normally go out on a date?” people to borrow more than they would actually want 32 WORLD DEVELOPMENT REPORT 2015 to if they assessed the full costs (Bertrand and Morse about their decisions can sometimes change their 2011). The field trial randomly divided borrowers into behavior. If the field test on payday lending had been groups. A control group received the standard payday an actual policy change in the way information was loan company envelope with the cash and the paper- presented, then individuals would have been exposed work for their loan (figure 1.3, panel a). Another group to the more informative envelopes every time they received a cash envelope that showed, in addition, how visited a payday store instead of just once, and the the dollar fees accumulate when a loan is outstanding effects probably would have been even stronger. And for three months, compared to the equivalent fees for slight alterations in the envelopes might have had borrowing the same amount on a credit card (figure larger effects. Relative to other policy alternatives— 1.3, panel b). The envelopes provided some anchoring to such as subsidies to loans and measures to reduce help borrowers evaluate the cost of payday loans. risk—the intervention has a low cost. Thus it is reason- The experiment incorporated behavioral principles able to consider such interventions as complements to about possible cognitive biases and ways to debias con- more standard policies in the credit market to help the sumers. Whereas the payday loan shops highlight the poor. small dollar cost of the transaction (for example, $15 for a two-week loan of $100), individuals may be misled by Simplification of loan products the apparently low costs and fail to add up in their own Consider next the plight of consumers who have lim- minds the costs over time and thus recognize the high ited experience in a market in which they must choose implicit interest rate of the loans. a product. An experiment in the credit market in Mex- The results of the field experiment suggest that ico City sheds light on the difficulties consumers have borrowers were indeed biased: they were applying (Giné, Martinez Cuellar, and Mazer 2014). Low-income too narrow a decision frame. Compared to the control individuals from Mexico City were invited to choose group, individuals who received the envelope with the the best one-year, 10,000 peso ($800) loan product from “dollar anchor” were 11 percent less likely to borrow a randomized list of loan products representative of from the payday lenders in the four months that fol- the local credit market. Individuals could earn rewards lowed the intervention. if they identified the lowest-cost product. Only 39 per- The findings illustrate the “peanuts effect”: people cent of people could identify the lowest-cost product do not consider the consequences of a small dollar when presented with the actual brochures designed transaction because they view small amounts of by the banks for their customers (figure 1.4, panel a). money as “peanuts”; as a result, they incur high costs But a much larger fraction (68 percent) could identify or forgo lucrative opportunities (Prelec and Loewen- the lowest-cost credit product from a user-friendly stein 1991). Fruit vendors in Chennai, India, provide summary sheet designed by the Consumer Financial a particularly vivid example (Banerjee and Duflo Credit Bureau of Mexico (figure 1.4, panel b). 2011). Each day, the vendors buy fruit on credit to sell Some participants in the experiment received during the day. They borrow about 1,000 rupees (the personalized text messages conveying financial infor- equivalent of $45 in purchasing parity) each morning mation. No text message intervention significantly at the rate of almost 5 percent per day and pay back affected the ability to identify the lowest-cost loan the funds with interest at the end of the day. By forgo- product. In the experiment, only the way that infor- ing two cups of tea each day, they could save enough mation was disclosed on the loan products affected after 90 days to avoid having to borrow and would thus decision making. increase their incomes by 40 rupees a day, equivalent Experimentation on finding the best ways to make to about half a day’s wages. But they do not do that. the nature of their opportunities salient to individuals “The point is that these vendors are sitting under what is an active area of research. Studies include how best appears to be as close to a money tree as we are likely to disseminate information about national employ- to find anywhere,” as Banerjee and Duflo (2011, 191) put ment programs in India (Dutta and others 2014); how it. “Why don’t they shake it a bit more?” The answer best to inform young people and their parents about is clear, in behavioral terms. Thinking as they always the return to higher education (Jensen 2010; Dinkel- do (automatically) rather than deliberatively, the ven- man and Martínez 2014); how best to make people dors fail to go through the exercise of adding up the aware of the risks of AIDS (Dupas 2011); and how best small fees incurred over time to make the dollar costs to increase awareness and use of contraception (Mun- salient enough to warrant consideration. This example shi and Myaux 2006). Chapters in part 2 discuss many illustrates why getting people to think more broadly applications. THINKING AUTOMATICALLY 33 Figure 1.3 Reframing decisions can improve welfare: The case of payday borrowing a. The standard envelope A payday borrower receives his cash in an envelope. The standard envelope shows only a calendar and the due date of the loan. b. The envelope comparing the costs of the payday loan and credit card borrowing In a field experiment, randomly chosen borrowers received envelopes that showed how the dollar fees accumulate when a payday loan is outstanding for three months, compared to the fees to borrow the same amount with a credit card. How much it will cost in fees or interest if you borrow $300 PAYDAY LENDER CREDIT CARD (assuming two-week fee is $15 per $100 loan) (assuming a 20% APR) If you repay in: If you repay in: 2 weeks $45 2 weeks $2.50 1 month $90 1 month $5 2 months $180 2 months $10 3 months $270 3 months $15 Borrowers who received the envelope with the costs of the loans expressed in dollar amounts were 11 percent less likely to borrow in the next four months compared to the group that received the standard envelope. Payday borrowing decreased when consumers could think more broadly about the true costs of the loan. Source: Bertrand and Morse 2011. Note: APR = annual percentage rate. 34 WORLD DEVELOPMENT REPORT 2015 • The connections, if any, that are drawn between the Figure 1.4 Clarifying a form can help borrowers find a better loan product current decision problem and decisions that the individual made earlier Low-income subjects from Mexico City were invited to classrooms to choose • The gap between the period when a decision maker the cheapest one-year, $800 (10,000 peso) loan product from a set of five products representative of actual credit products offered by banks in Mexico forms an intention and the period when he has City. They could earn rewards by getting the right answer. When using the banks’ funds available to pay for it descriptions of their products, only 39 percent of the people could identify the • The salience of a social identity cheapest credit product. When using the more straightforward summary sheet, • The salience of relevant norms.6 68 percent could identify the cheapest credit. Well-thought-out policy can improve development a. Bank leaflets b. Summary sheet outcomes by changing the context of decision making, especially in situations in which even people trained in deliberative thinking might struggle. Several examples related to default options—which are the choices that are selected automatically unless an alternative is spec- ified—are considered next. Default options and other framing effects Many countries all over the world, both rich and poor, seek to remove the impediments students face in obtaining postsecondary education. Policies based on the standard model would focus on lowering the costs and increasing the information about oppor- tunities. But policies based on the psychological and social actor would widen the focus to include framing, broadly understood to include the small details of the consumer’s choice set. A recent study in the United States uncovered the enormous sensitivity of students’ college application decisions to a small change in the 39% of people could identify the cheapest 68% of people could identify the cheapest loan product on the information leaflets loan product on a more straightforward cost of sending test scores to colleges (Pallais, forth- from banks. summary sheet. coming). In 1998, when a popular university readiness examination (the ACT) increased from three to four the = 10 people number of free score reports that test takers could send to colleges, students sent substantially more reports. Figure 1.5 shows that most high school students grad- Source: Giné, Martinez Cuellar, and Mazer 2014. uating before 1998 sent exactly three reports and that most high school students graduating after 1998 sent Biases in assessing value exactly four reports.7 The change in behavior was the Even when individuals make unbiased assessments same for low- and high-income students, which sug- of information, they may make biased assessments gests that the students’ choices were not based on a of value. When people think automatically, the way in deliberative decision that weighed benefits and costs, which their choices are presented and the context in but instead on unthinking acceptance of a default which they make decisions may systematically influ- option: three reports were free, and each additional ence their preferences. Factors that would be unimpor- report would have cost another $6. tant under the standard assumption that people have This is another money tree. While students did unlimited capacities to process information, but that not need to limit the number of schools to which in fact can be quite important, include the following: they applied to the number of free score reports, most students—both low income and high income—did. As • The default option, to which decisions would revert if a result, the low-income students were saving $6 but no other decision was made or no other action taken forgoing $1,700 in lifetime income for each dollar they • The labels on options saved, on average. • The number of options The vast influence of default options on decisions • The sequence in which the options are presented has been widely replicated in many domains, including THINKING AUTOMATICALLY 35 saving and insurance decisions with massive financial Figure 1.5 A small change in the college application consequences.8 Why are these findings so surprising process had a huge impact on college attendance and important? It is not that money trees are every- where, but the findings give us valuable information When the number of free test score reports that a high school student could send to colleges increased from three to four in the United States in 1998, on decision making and on the potential for designing low-income students applied for and attended more selective colleges, which policies that improve welfare. If individuals carefully increased their projected average lifetime income by roughly $10,000, far compared costs and benefits, as standard policy analy- outweighing the $6 cost of sending an additional score report. sis assumes, a switch from three free options to four free options should not affect decisions as long as the 90 costs of doing so are small (which, at $6, they were in 80 Percentage of ACT takers the United States). Defaults can influence choices in a 70 number of ways. Until the moment an individual makes a decision, preferences often are not clearly specified. 60 Since constructing a preference requires effort but 50 accepting the default choice is effortless, people may choose the default. Decision makers might also con- 40 strue the default as a recommendation. A default option 30 is just one example of a frame, broadly defined as a way 20 of structuring choices, that may affect an individual’s behavior by influencing what is salient to him or her 10 and the cognitive costs that a decision entails. 0 1991 1992 1994 1996 1998 2000 2004 Loss aversion High school graduation year In general, people make decisions based on a con- sideration of changes in values from a reference point, Three score reports Four score reports rather than on the basis of absolute values. The refer- Source: Pallais, forthcoming. ence point is the benchmark. When people evaluate whether or not they like something, they tend to implicitly ask themselves, “Compared to what?” It Loss aversion can also be used to influence the turns out that when thinking about something as a behavior of others. In Chicago, for instance, teachers loss, people generally count the difference more than were paid a bonus at the beginning of the school year, they would count it if they thought about the same in advance, but were told they would lose it if students thing as a gain. They feel the losses more acutely than did not meet a threshold level of achievement by the they would feel the gains of a similar size (loss aversion). end of the school year (Fryer and others 2012). These This psychological phenomenon is widespread and teachers expended a substantially greater effort than helps explain a large set of phenomena in financial did teachers who were in all other respects similar but markets (Kahneman and Tversky 1979; Shiller 2000). who could receive the bonus only at the end of the year. Reference points are behind what economists call The potential loss of the bonus was more salient than “money illusion.” Many people prefer a 6 percent the potential gain of the bonus. The change in frame income raise when there is 4 percent inflation to a 3 may also have had a powerful effect by changing the percent raise with no inflation (Shafir, Diamond, and meaning of achieving high results. A gain may have Tversky 1997). They prefer the former option, which is been perceived as a reward for superior performance, expressed in high numerical terms, even though the whereas a loss may have been perceived as a punish- real dollar value of the latter option is higher. Refer- ment for failing to meet a certain performance norm. ence points can mislead when they are established in Policies that increase aspirations may affect behavior terms of nominal rather than real values. in part by changing the benchmark for what is consid- By setting goals, individuals identify a particular ered a loss. Chapter 3 will discuss early work that sug- value as a reference point against which to measure gests that interventions have raised the aspirations and performance. If individuals do not meet the goal, they accordingly changed behaviors among teenage girls in are likely to experience the disappointment as a loss rural India (Beaman and others 2009) and households (Suvorov and van de Ven 2008). Loss aversion may in rural Ethiopia (Bernard and others 2014). Chapter 2 thus make goals a credible and effective instrument will discuss an intervention regarding aspirations for for self-regulation. sex workers in India (Ghosal and others 2013). 36 WORLD DEVELOPMENT REPORT 2015 Factoring in psychological aversion to losses with diet, for example, according to Thaler and Sunstein, reference to the status quo can be important in under- “Putting the fruit at eye level counts as a nudge. Ban- standing the decisions of policy makers, too. Trade pol- ning junk food does not” (2008, 6). Putting the fruit at icy may offer an example. Industries suffering losses eye level is a change in framing. are more likely than others to receive trade protection A component of choice architecture is simplicity. (Trefler 1993; Baron and Kemp 2004). Intuitively, the Too many options or too much complexity may lead prospect of losing tens of thousands of jobs in old individuals to avoid thinking through a decision, to sectors may loom much larger than the prospect of postpone indefinitely making an active decision, or to creating many more jobs under free-market policies make error-ridden decisions. Consider an example in in new sectors (Freund and Ozden 2008). According voting in which individuals may have to make choices to some economists, the reason that political reform in scenarios for which they have limited experience often occurs during crises is that when large numbers and little or no education or training to prepare them. of people have experienced losses, they are more will- ing to gamble to recover what they have lost; that is, Application: Simplification at the ballot they become risk seeking (Weyland 1996). box in Brazil A common policy recommendation to promote devel- Choice architecture opment is to improve public services by increasing A choice architect is someone who organizes the context the political influence of the neediest citizens. But in which people make decisions. Many people are how can it be done? The World Development Report 2004: choice architects, most without realizing it. Think of Making Services Work for Poor People cites the fact that doctors describing the available treatments to patients, “the poor have little clout with politicians” as a cause of matchmakers describing marriage choices, or mon- underprovision of public services. The report devotes eylenders describing loan products. Choice architecture a whole chapter to increasing citizen influence on pol- influences decision making by simplifying the presen- iticians by strengthening “elections, informed voting, tation of options, by automatically evoking particular and other traditional voice mechanisms” (World Bank associations, or by making one option more salient or 2004, 78). After the report was written, a simple way to easier to choose than the alternatives (Thaler and Sun- achieve this objective occurred in Brazil. stein 2008). Federal law in Brazil makes voting compulsory for all citizens aged 18–70.9 Although turnout was thus very high, over 30 percent of votes were blank The policy mechanisms discussed in or error ridden and were therefore discarded in 1994 (Fujiwara 2010, figure 2). Some 42 percent of adult Bra- this chapter include framing, anchoring, zilians had not completed fourth grade. For them, the demands of voting by writing down the names of the simplification, reminders, and candidates on paper ballots were heavy. Beginning in 1998, Brazil introduced electronic voting technology commitment devices. Policy makers can (see figure 1.6). Using the new technology, a voter saw a photo of the candidate he selected. The technology employ these mechanisms to help people provided step-by-step directions that “walked” voters through the process of voting for candidates in the make better decisions, which in turn can many different races and gave them an error message if they incorrectly marked a ballot. The new technol- reduce poverty. ogy reduced the number of error-ridden and under- counted votes among the less educated. The interven- tion effectively enfranchised 11 percent of citizens, When individuals are thinking automatically, a mainly the less educated. After the change, the share mere “nudge” may change their behavior. A nudge is a of valid votes increased to more than 90 percent of policy that achieves behavior change without actually total votes. With more votes of the poor counted, more changing the set of choices. It does not forbid, penalize, candidates from pro-poor parties have been elected to or reward any particular choices. Instead, it points peo- state legislatures. ple toward a particular choice by changing the default An evaluation of this policy change identifies these option, the description, the anchor, or the reference effects by using the fact that when electronic voting point. To encourage people to choose a more healthy technology was introduced in 1998, only municipalities THINKING AUTOMATICALLY 37 with more than a threshold level of voters used the new Figure 1.6 Simplifying voting procedures technology because of the limited supply of devices, in Brazil is having positive welfare effects on the poor across generations while the rest used paper ballots (Fujiwara 2010). Thus the study can compare outcomes for municipalities When Brazil simplified its voting procedures, more poor, just above and just below the threshold to assess the illiterate, and semiliterate voters could cast proper ballots. The increase in the clout of the poor shifted state spending toward effect of the introduction of electronic voting. public health care. As a result, the number of low-birth-weight One of the things that legislators could quickly babies fell, paving the way for better adult health. affect in Brazil is funding for health care. Because of Until 1998, Brazilian elections used only paper ballots. the shift in the political strength of the parties on the Left, state spending on public health care increased by 34 percent over eight years. Public health care is free in Brazil. The shift in the funding, for example, enabled 20 percent more uneducated pregnant women to make regular prenatal visits and improved newborn health (reducing the prevalence of low-weight births by 6 percent). This is a major development success, since But only about 60% of voters had completed fourth grade. Less newborn health, controlling for other factors, predicts than 70% of the votes were correctly filled out. The rest had to lifetime health, education, and income. be discarded. These findings suggest that too little attention has been paid to the unrealistic demands on voters with little education to read instructions and fill out paper ballots at the voting booth. The head designer of Bra- Beginning in 1998, Brazil began to shift to electronic voting, zil’s electronic voting technology described the new where individuals didn’t need to write anything. system’s reduction in error-ridden votes as “a surprise” (Fujiwara 2010, 6). This design change in balloting was To vote for a candidate, an individual typed the candidate’s ID a simple policy that accomplished something quite number into a simple keypad, which called up the candidate’s difficult—a shift in the political clout of the neediest photo. The voter confirmed his choice by pressing the green citizens and a shift in the allocation of public spending button, or canceled a mistake by pressing the orange button. toward health services for the poor. Overcoming intention-action divides This chapter concludes with one additional way in which human behavior systematically departs from that assumed in the standard economic model: individ- The reduction in error-ridden ballots meant the de facto uals have bounded willpower. The deliberative system enfranchisement of 11% of the electorate. can restrain the impulses of the automatic system, but as the chapter has repeatedly emphasized, the deliberative system has limited capacity. Consider the case of HIV/AIDS. A major cause of treatment failure all over the world is incomplete adherence to treat- ment regimens. In many cases, patients will receive With more votes of the poor counted, more candidates from pills from a clinic each month. If taken daily, the pills pro-poor parties were elected in state legislatures, which: will postpone the worst symptoms of the disease for many years. Individuals who understand this and 34% 20% 6% intend to take the pills may nonetheless find it hard to carry out their intention. The press of demands on Quickly increased Increased the fraction Improved newborn states’ budget shares of uneducated pregnant health (reduced by them—caring for their children and earning a living— on public health women with regular 6% the prevalence of impairs their ability to remember to take the pills two care, raising health prenatal visits by 20% low-weight births) expenditures by 34% times each day. over eight years This is one of many instances of a divide between intentions and actions. Underlying many intention- Source: Fujiwara 2010. action divides is present bias, an overweighting of the 38 WORLD DEVELOPMENT REPORT 2015 present relative to the future that results in inconsis- if they do not reach a particular goal. People who are tencies in choices over time. Achieving goals often aware of their own tendency to procrastinate may find requires incurring a cost in the present for a payoff in commitment devices attractive. Commitment devices the future. Since the present is more salient than the helped people save money in a field experiment in the future, people tend to overweight the costs relative to Philippines (Ashraf, Karlan, and Yin 2006) and helped the benefit. The tendency increases the farther away people quit smoking in another field experiment in the deadline lies (see, for example, Shu and Gneezy that country (Giné, Karlan, and Zinman 2010). 2010). Later, individuals feel regret. Policies that create reminders or remove small Conclusion impediments in such areas as savings, adherence We have two systems of thinking: the automatic system to health regimens, and voting in elections have had and the deliberative system. When making decisions, successes in narrowing intention-action divides. To we cannot manage without the automatic system, and improve adherence to HIV/AIDS drug regimens, a it can produce remarkably well-adapted choices at a small-scale study tested the effect of reminders to take trivial cost of effort in decision making. The automatic the antiretroviral medicine (Pop-Eleches and others system draws heavily on default assumptions and 2011). Patients in Kenya were randomly divided into interpretive frames. It is very sensitive to what is salient three groups. No reminders were given to the first and what associations effortlessly come to mind. group, weekly reminders were given to the second This chapter has demonstrated ways in which group, and daily reminders were given to the third development practitioners might make the world group. The reminders were made through a low-cost easier to navigate for people who rely primarily on the messaging system on cell phones dispensed by the automatic system—that is, for everyone. Since every experimenters. The results were promising. Indi- choice set is presented in one way or another, making viduals who received a weekly reminder (through a the crucial aspects of the choice salient and making it low-cost short messaging service, often called SMS) cognitively less costly to arrive at the right decision increased adherence to the drug regimen by 13 percent- (such as choosing the lowest-cost loan product, follow- age points, although a daily reminder had virtually no ing a medical regimen, or investing for retirement) can effect on adherence.10 (Adherence was counted as pos- help people make better decisions. itive if individuals took their drugs at least 90 percent The behavioral perspective on decision making of the days.) The findings suggest that despite SMS suggests that seemingly minor and low-cost policy outages, accidental phone loss, and a dispersed rural changes may have a large impact on the achievement population, the weekly intervention was effective at a of development goals and the reduction of poverty. The very low marginal cost. policy mechanisms discussed in this chapter include In Colombia, the government uses a conditional framing, anchoring, simplification, reminders, and cash transfer (CCT) program under which families of commitment devices. Policy makers can employ these students are paid every two months for attendance at mechanisms to help people make better decisions, school at least 80 percent of the time. Yet there is still a which in turn can reduce poverty. large drop in school enrollment in the higher grades of secondary school and a low rate of matriculation at ter- tiary institutions. Then a simple variation of the CCT Notes was implemented that distributed two-thirds of the 1. Surveys are Kahneman (2003, 2011). A collection “good attendance” funds on the same bimonthly basis of pathbreaking findings is Slovic (1987). Popular but distributed the remaining funds for all the months accounts are Ariely (2008) and Vedantam (2010). in a lump sum upon high school graduation. Students 2. Daniel Kahneman (2003) describes this example in could receive the payment sooner by matriculating at his Nobel Lecture, citing personal communication an institution of higher education. The policy began with Shane Frederick. to work much better. It increased matriculation by 49 3. Cited by Michael Suk-Young Chwe (2014). percentage points (Barrera-Osorio and others 2011). 4. It may not have been completely irrelevant. Know- Commitment devices are an additional promising ing that guacamole was available at the party area of intervention to address present bias. They com- could affect beliefs about how much he drank and bine an awareness of the intention-action divide with how strongly the liquor affected him. an understanding of loss aversion. Commitment devices 5. The finding from this experiment accords with a are strategies whereby people agree to have a penalty theme in literary criticism, in which “irrelevant” imposed on them (that is, they agree to incur a loss) detail adds to believability. Pierre in War and Peace THINKING AUTOMATICALLY 39 notes how a man just before he dies adjusts his Bernard, Tanguy, Stefan Dercon, Kate Orkin, and Alema- blindfold because it is too tight; Orwell notes how a yehu Seyoum Taffesse. 2014. “The Future in Mind: condemned man swerves to avoid a puddle (Wood Aspirations and Forward-Looking Behaviour in Rural 2008). Ethiopia.” Working Paper, Centre for the Study of 6. A review is Schwartz (2013). African Economies, University of Oxford, Oxford, U.K. Bertrand, Marianne, and Adair Morse. 2011. “Information 7. Those graduating in 1998 could send three score Disclosure, Cognitive Biases, and Payday Borrowing.” reports for free if they took the test as 11th graders Journal of Finance 66 (6): 1865–93. and four reports for free if they took the test as 12th Bertrand, Marianne, and Sendhil Mullainathan. 2001. graders. “Do People Mean What They Say? Implications for 8. See Johnson and others (1993); Madrian and Shea Subjective Survey Data.” American Economic Review 91 (2001); Choi, Hardigree, and Thistle (2002); Johnson (2): 67–72. and Goldstein (2003); Thaler and Sunstein (2008). Choi, Seungmook, Don Hardigree, and Paul D. Thistle. 9. A failure to register or vote makes a citizen ineli- 2002. “The Property/Liability Insurance Cycle: A Com- gible to receive several public services until a fine parison of Alternative Models.” Southern Economic Jour- is paid. nal 68 (3): 530–48. 10. The reminders varied in content. Tailoring the Chwe, Michael Suk-Young. 2014. “Scientific Pride and content and implementation of a policy to over- Prejudice.” New York Times, January 31. http://www come psychological resistance or cognitive biases .nytimes.com/2014/02/02/opinion/sunday/scientific -pride-and-prejudice.html?_r=1. requires experimentation. This is a theme through- Dawson, Erica, Thomas Gilovich, and Dennis T. Regan. out the Report. To take another example, a study 2002. “Motivated Reasoning and Performance on the of safe-sex programs in Uganda and Botswana Wason Selection Task.” Personality and Social Psychology suggests that interventions actually can be more Bulletin 28 (10): 1379–87. effective when they establish a collective narrative Dinkelman, Taryn, and A. Martínez. 2014. “Investing in and a shared fate than when they appeal only to Schooling in Chile: The Role of Information about self-interest (Swidler 2009). The success of differ- Financial Aid for Higher Education.” Review of Econom- ent messages can vary across groups, perhaps due ics and Statistics 96 (2): 244–57. to different interpretive frames, which reinforces Dupas, Pascaline. 2011. “Health Behavior in Developing the need for piloting framing interventions, an Countries.” Annual Review of Economics 3 (1): 425–49. idea developed in chapter 11. Dutta, Puja, Rinku Murgai, Martin Ravallion, and Dom- inique van de Walle. 2014. “Right to Work? Assessing References India’s Employment Guarantee Scheme in Bihar.” Ariely, Dan. 2008. Predictably Irrational: The Hidden Forces Report 85414, Equity and Development Series, World That Shape Our Decisions. New York: HarperCollins. Bank, Washington, DC. Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Englich, Birte, Thomas Mussweiler, and Fritz Strack. 2006. Odysseus to the Mast: Evidence from a Commitment “Playing Dice with Criminal Sentences: The Influence Savings Product in the Philippines.” Quarterly Journal of of Irrelevant Anchors on Experts’ Judicial Decision Economics 121 (2): 635–72. Making.” Personality and Social Psychology Bulletin 32 (2): Banerjee, Abhijit, and Esther Duflo. 2011. Poor Economics: 188–200. A Radical Rethinking of the Way to Fight Global Poverty. Evans, Jonathan St. B. T. 2008. “Dual-Processing Accounts New York: PublicAffairs. of Reasoning, Judgment, and Social Cognition.” Annual Baron, Jonathan, and Simon Kemp. 2004. “Support for Review of Psychology 59 (January): 255–78. Trade Restrictions, Attitudes, and Understanding of Frankish, Keith, and Jonathan Evans. 2009. “The Duality Comparative Advantage.” Journal of Economic Psychology of Mind: An Historical Perspective.” In In Two Minds: 25 (5): 565–80. Dual Processes and Beyond, edited by Jonathan Evans and Barrera-Osorio, Felipe, Marianne Bertrand, Leigh L. Keith Frankish, 1–29. Oxford, U.K.: Oxford University Linden, and Francisco Perez-Calle. 2011. “Improving Press. the Design of Conditional Transfer Programs: Evi- Freund, Caroline, and Caglar Ozden. 2008. “Trade Policy dence from a Randomized Education Experiment in and Loss Aversion.” American Economic Review 98 (4): Colombia.” American Economic Journal: Applied Economics 1675–91. 3 (2): 167–95. Fryer Jr., Roland G., Steven D. Levitt, John List, and Sally Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, Sadoff. 2012. “Enhancing the Efficacy of Teacher Incen- Rohini Pande, and Petia Topalova. 2009. “Powerful tives through Loss Aversion: A Field Experiment.” Women: Does Exposure Reduce Bias?” Quarterly Journal Working Paper 18237, National Bureau of Economic of Economics 124 (4): 1497–1540. Research, Cambridge, MA. 40 WORLD DEVELOPMENT REPORT 2015 Fujiwara, Thomas. 2010. “Voting Technology, Political ized Controlled Trial of Text Message Reminders.” Responsiveness, and Infant Health: Evidence from AIDS 25 (6): 825–34. Brazil.” Department of Economics, University of Brit- Prelec, Drazen, and George Loewenstein. 1991. “Decision ish Columbia, Vancouver. Making over Time and under Uncertainty: A Common Ghosal, Sayantan, Smarajit Jana, Anandi Mani, Sandip Approach.” Management Science 37 (7): 770–86. Mitra, and Sanchari Roy. 2013. “Sex Workers, Stigma, Reyes, Robert M., William C. Thompson, and Gordon H. and Self-Belief: Evidence from a Psychological Train- Bower. 1980. “Judgmental Biases Resulting from Dif- ing Program in India.” Working Paper, University of fering Availabilities of Arguments.” Journal of Personal- Warwick, U.K. ity and Social Psychology 39 (1): 2–12. Giné, Xavier, Christian Martinez Cuellar, and Rafael Schwartz, Barry. 2013. “How We Decide.” Background Keenan Mazer. 2014. “Financial (Dis-)Information: paper for the World Development Report 2015. Evidence from an Audit Study in Mexico.” Policy Schwarz, Norbert, Fritz Strack, and Hans-Peter Mai. 1991. Research Working Paper 6902, World Bank, Wash- “Assimilation and Contrast Effects in Part-Whole ington, DC. Question Sequences: A Conversational Logic Analysis.” Giné, Xavier, Dean Karlan, and Jonathan Zinman. 2010. Public Opinion Quarterly 55 (1): 3–23. “Put Your Money Where Your Butt Is: A Commitment Shafir, Eldar, Peter Diamond, and Amos Tversky. 1997. Contract for Smoking Cessation.” American Economic “Money Illusion.” Quarterly Journal of Economics 112 (2): Journal: Applied Economics 2 (4): 213–35. 341–74. Jensen, Robert. 2010. “The (Perceived) Returns to Educa- Shiller, Robert J. 2000. “Measuring Bubble Expectations tion and the Demand for Schooling.” Quarterly Journal and Investor Confidence.” Journal of Psychology and of Economics 125 (2): 515–48. Financial Markets 1 (1): 49–60. Johnson, Eric J., and Daniel Goldstein. 2003. “Do Defaults Shu, Suzanne B., and Ayelet Gneezy. 2010. “Procrastination Save Lives?” Science 302: 1338–39. of Enjoyable Experiences.” Journal of Marketing Research Johnson, Eric J., John Hershey, Jacqueline Meszaros, and 47 (5): 933–44. Howard Kunreuther. 1993. Framing, Probability Distor- Slovic, Paul. 1987. “Perception of Risk.” Science 236 (4799): tions, and Insurance Decisions. New York: Springer. 280–85. Kahneman, Daniel. 2003. “Maps of Bounded Rationality: Stanovich, Keith E., and Richard F. West. 2000. “Advancing Psychology for Behavioral Economics.” American Eco- the Rationality Debate.” Behavioral and Brain Sciences 23 nomic Review 93 (5): 1449–75. (5): 701–17. — —— —. 2011. Thinking, Fast and Slow. New York: Farrar, Suvorov, Anton, and Jeroen van de Ven. 2008. “Goal Setting Straus and Giroux. as a Self-Regulation Mechanism.” Working Paper 0122, Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Center for Economic and Financial Research, Moscow. Theory: An Analysis of Decision under Risk.” Economet- Swidler, Ann. 2009. “Responding to AIDS in Sub- rica 47 (2): 263–91. Saharan Africa.” In Successful Societies: Institutions, Cul- — —— —. 2000. Choices, Values, and Frames. Cambridge, tural Repertoires and Population Health, edited by Peter U.K.: Cambridge University Press and Russell Sage Hall and Michelè Lamont, 128–50. Cambridge, U.K.: Foundation. Cambridge University Press. Madrian, Brigitte C., and Dennis F. Shea. 2001. “The Power Thaler, Richard H., and Cass R. Sunstein. 2008. Nudge: of Suggestion: Inertia in 401(k) Participation and Sav- Improving Decisions about Health, Wealth, and Happiness. ings Behavior.” Quarterly Journal of Economics 116 (4): New Haven, CT: Yale University Press. 1149–87. Todd, Peter M., and Gerd Gigerenzer. 2000. “Précis of Mani, Anandi, Sendhil Mullainathan, Eldar Shafir, and Simple Heuristics That Make Us Smart.” Behavioral and Jiaying Zhao. 2013. “Poverty Impedes Cognitive Func- Brain Sciences 23 (5): 727–41. tion.” Science 341 (6149): 976–80. Trefler, Daniel. 1993. “Trade Liberalization and the Theory Montier, James. 2007. Applied Behavioural Finance. Chich- of Endogenous Protection: An Econometric Study of ester, England: Wiley. U.S. Import Policy.” Journal of Political Economy 101 (1): Munshi, Kaivan, and Jacques Myaux. 2006. “Social Norms 138–60. and the Fertility Transition.” Journal of Development Eco- Vedantam, Shankar. 2010. The Hidden Brain: How Our Uncon- nomics 80 (1): 1–38. scious Minds Elect Presidents, Control Markets, Wage Wars, Pallais, Amanda. Forthcoming. “Small Differences That and Save Our Lives. New York: Random House. Matter: Mistakes in Applying to College.” Journal of Weyland, Kurt. 1996. “Risk Taking in Latin American Eco- Labor Economics. nomic Restructuring: Lessons from Prospect Theory.” Pop-Eleches, Cristian, Harsha Thirumurthy, James Hab- International Studies Quarterly 40 (2): 185–207. yarimana, Joshua Graff Zivin, Markus Goldstein, Wood, James. 2008. How Fiction Works. New York: Farrar, Damien de Walque, Leslie Mackeen, Jessica Haberer, Straus and Giroux. John Sidle, and Duncan Ngare. 2011. “Mobile Phone World Bank . 2004. World Development Report 2004: Making Technologies Improve Adherence to Antiretroviral Services Work for Poor People. Washington, DC: World Treatment in Resource-Limited Settings: A Random- Bank. 2 CHAPTER Thinking socially Humans are deeply social animals. Our beliefs, desires, interests of no one. Yet by the same token, tempo- and behaviors are affected by social preferences, our rary interventions can have large and lasting positive relationships, and the social contexts in which we live effects on a community by shifting a pattern of social and make decisions. We are “group-minded individu- interactions from one suboptimal self-reinforcing als” who see the world from a social as well as an indi- arrangement (or “equilibrium”) to another arrange- vidual perspective; we understand what is in the minds ment that better promotes well-being and becomes of others and often act as if our brains are networked self-sustaining (see spotlight 1, on fighting a social with the brains of other people (Tomasello 2014). norm tolerating corruption). Sociality is also a lever for Human sociality—the tendency among humans to new types of development interventions that harness associate and behave as members of groups—affects the tendencies of individuals to seek social status, to decision making and behavior and has important build and maintain social identities, and to cooperate consequences for development.1 Our social tenden- with others under certain conditions. cies mean that we are not purely selfish and wealth- Policy makers often underestimate the social component in behavior change. The purpose of this chapter is to summarize recent findings on the social Policies can tap people’s social tendencies microfoundations of action and their implications for development policy. To demonstrate that there is a to associate and behave as members of fundamentally social component to thinking and decision making, the discussion begins by examin- groups to generate social change. ing “other-regarding” preferences—including the innate human desire for social status, tendencies to identify maximizing actors, as many economic models and with groups and help others, and propensities to policies assume; rather, we value reciprocity and fair- cooperate with others who are cooperating—and their ness, we are willing to cooperate in the attainment implications for institutional design and development of shared goals, and we have a tendency to develop interventions. Because social networks are the key and adhere to common understandings and rules of pathway through which social influences are transmit- behavior, whether or not they benefit us individually ted, the chapter then considers how social networks and collectively. Since what we do is often contingent affect the development process and interventions that on what others do, local social networks and the ideas, leverage networks to spur social change. Finally, since norms, and identities that propagate through them sociality leads to the informal rules known as social exert important influences on individual behavior norms that coordinate behavior, the chapter examines (see figure 2.1). some of the social outcomes that such norms create A key consequence of sociality for development is and the policies that take account of norms to better that groups and even entire societies can get stuck in achieve development objectives. collective patterns of behavior—such as corruption, Human sociality is like a river running through segregation, and civil war—that arguably serve the society; it is a current that is constantly, if often THINKING SOCIALLY 43 Figure 2.1  What others think, expect, and do influences our own preferences and decisions Humans are inherently social. In making decisions, we are often affected by what others are thinking and doing and what they expect from us. Others can pull us toward certain frames and patterns of collective behavior. imperceptibly, shaping individuals, just as flowing Social preferences and water shapes individual stones in a riverbed. Policy their implications makers can either work with these social currents when designing interventions or ignore them and find Social recognition and the power of themselves swimming upstream. Just as a dam taps a social incentives river’s kinetic energy to generate electricity, interven- Everyone knows that economic incentives can influ- tions can tap sociality to facilitate cost-effective social ence behavior.2 What is less commonly recognized is change. This chapter offers examples of how sociality that social incentives can also exert a powerful effect can serve as a starting point for new kinds of develop- on behavior. In fact, social rewards, such as status and ment interventions. recognition, can motivate people to exert effort and 44 WORLD DEVELOPMENT REPORT 2015 can even substitute for monetary rewards in some are advantageous and exemplary but often difficult situations. In a field experiment in Switzerland, for to quantify. In fact, humans may have an innate, example, researchers disentangled the economically unconscious tendency to reward strong contributors relevant (“instrumental”) and economically irrelevant to group goals with esteem, which helps groups over- (“noninstrumental”) aspects of social rewards by come barriers to collective action (Willer 2009). showing that individuals’ performance improved on A study of contributions to Wikipedia (an online a one-time data entry task when they were told that encyclopedia produced through voluntary efforts) the two people who put in the most effort would be illustrates how bestowals of status may contribute to rewarded with a congratulatory card and a personal the production of collective goods. Contributors who thank-you from the managing director. These noneco- were randomly awarded peer esteem in the form of a nomic rewards increased performance by 12 percent, “Barnstar” (an editing award that is publicly displayed) the equivalent of a hypothetical wage increase of were 60 percent more productive over the course of 35–72 percent, according to previous studies of out- the 90 days following the receipt of the award than put elasticity in gift-exchange experiments (Kosfeld members of a control group, on average (Restivo and and Neckermann 2011). Similarly, salespeople in a van de Rijt 2012). The informal rewards are free to give U.S. company were willing to trade off approximately and carry no immediate material benefits but have a $30,000 in income to achieve membership in the firm’s substantial effect on productivity and may play a key “club” for top performers—the benefits of which were role in sustaining volunteer effort over time. a gold star on their name card, companywide recog- Prestige can incentivize countries, too. When states’ nition, and an e-mail from the chief executive officer different values and norms inhibit cooperation, status (Larkin 2009). awards in the form of participation in international Development interventions can harness the human summits and strategic partnerships may be more desire for status and recognition. In a field experiment effective than conventional strategies of containment in Zambia, hairstylists and barbers recruited by a pub- and integration for achieving cooperation on global lic health organization to sell female condoms in their governance initiatives (Larson and Shevchenko 2010). shops were randomly assigned to one of four groups Chapter 9 on climate change examines the use of sta- receiving different awards based on condom sales tus awards and indicators to motivate policy makers (Ashraf, Bandiera, and Jack, forthcoming). People in and firms. the control group received no rewards, while people in Ranking schemes, which bestow status on exem- the treatment groups received one of the following: a plary states and shame underperforming ones, may 90 percent margin on condom sales; a 10 percent mar- be a cost-effective means of shifting state actions. gin on condom sales; or a nonfinancial reward in the Numerical indicators, such as the World Bank’s Doing form of stars stamped on a publicly displayed chart to Business rankings and the United Nations’ Gender represent each condom sale. The “star treatment” was Empowerment Measure, do not simply provide perfor- designed to make social impact salient by publicizing mance information, but they also serve as “psycholog- the stylist’s contribution to the health of his or her ical rules of thumb” (Sinclair 2005) that simplify and community. After one year, hairdressers in the star frame information (chapter 1) according to an ideology treatment had sold twice as many condoms as hair- of what a “good society” looks like. Indicators enable dressers in any other group, on average. For this group comparisons that motivate a variety of actors, includ- of individuals, the marginal utility of public recogni- ing citizens, nongovernmental organizations (NGOs), tion was higher than the marginal utility of money. elites, bureaucrats, and governments (Davis and oth- When should social awards be used, and how pow- ers 2012). The U.S. “Trafficking in Persons Report,” for erful and enduring are they really? Status awards may example, played an important role in spurring states to be especially useful when the quality of individual criminalize human trafficking, even though the rank- outputs is difficult to measure precisely (Besley and ing system is “hardly scientific” (Kelley and Simmons, Ghatak 2008) and when financial resources are scarce. forthcoming). In a world in which national control Thus many noneconomic organizations, including over policy is valued and information is becoming ever political parties, religious groups, the military, and cheaper to collect, analyze, and disseminate, indicators educational institutions, use status awards to achieve may become important tools for shifting state action. solidarity and elicit contributions to collective goods (Hechter 1987). Firms use employee-of-the-month Altruism, identity, and group dynamics clubs alongside traditional salaries to recognize and Some humans genuinely care about others’ well-being, incentivize contributions to organizational goals that and few of us are selfish all the time. This aspect of THINKING SOCIALLY 45 sociality has been investigated by economists using than nonmembers. In addition, individuals holding a an experimental tool called the “dictator game.” In the formal leadership position in either a farmer group or game, the dictator gets to decide how much of an ini- a village group were more generous toward members tial endowment (say, $10) he would like to give to the of the group in which they were a leader. Experiments second player (in some versions, the dictator’s choices indicated that group members hold leaders to a higher include taking some of the other player’s endowment). standard of caring for fellow group members. Economic theory predicts that the dictator will always While altruism and group identification can sup- make the most self-interested choice. But in experi- port mutual prosperity, they can also set the stage mental situations, fully selfish behavior is the excep- for the in-group favoritism and out-group hostilities tion, not the rule (Forsythe and others 1994; List 2007). that contribute to social unrest. For instance, a recent What determines whether someone acts gener- study indicated that exposure to war between the ages ously or selfishly? Expressions of altruism and other of 7 and 20 was associated with a lasting increase in socially beneficial behavior often depend on the social people’s egalitarian motivations toward their in-group. setting. In the game, dictators are much more gener- The children and young adults most affected by civil ous when they are giving to a charity. They also give wars in Georgia and Sierra Leone were more willing more to welfare recipients who express strong rather to sacrifice their own self-interest in a social choice than weak desires to work (Eckel and Grossman 1996; task to improve equality within their group (thought Fong, Bowles, and Gintis 2006). In Uganda, members of to enhance group cohesion and cooperation) than were coffee-producer cooperatives played the dictator game people who were least affected by violence (see figure and allocated more resources to anonymous members 2.2; Bauer and others 2014). Exposure to war-related of their farmer co-ops than to anonymous covillagers violence can also heighten preferences for military (Baldassarri and Grossman 2013).3 The study controlled solutions as opposed to negotiations, according to the for the effect of social proximity, demonstrating the findings of researchers studying the attitudes of former independent effect of group attachment in which iden- combatants (Grossman, Manekin, and Miodownik, tification with a group causes individuals to perceive forthcoming). Such “war effects” may help explain why even unknown members of the group more positively conflict becomes a persistent state of affairs, although Figure 2.2 Children and young adults most affected by war are more likely to favor members of their own group Children and young adults in postconflict societies played games in which they chose how to share money. Individuals least affected by war behaved in similar ways toward in-group and out-group members, whereas those most affected by war were much more likely to choose the egalitarian option when playing with an in-group member than an out-group member. In both countries, exposure to war increased in-group favoritism. a. Public goods game: Georgia b. Public goods game: Sierra Leone Percentage of egalitarian choices Percentage of egalitarian choices 100 100 75 75 50 50 25 25 0 0 Nonaffected (n = 118) Affected IDPs (n = 75) Least affected (n = 76) Most affected (n = 38) War-exposure group War-exposure group In-group Out-group In-group Out-group Source: Bauer and others 2014. Note: IDPs = internally displaced persons. 46 WORLD DEVELOPMENT REPORT 2015 scientific understanding of the link between exposure vated willingness to reward or punish the behavior of to conflict and group dynamics is limited. others, even at a cost to oneself (see Sobel 2005). If identities are malleable rather than fixed, inter- Economists use a tool called the “ultimatum game” ventions may be able to target social identities as a to study intrinsic reciprocity. The game begins with means of changing behavior. In Liberia, a local non- two players being shown a sum of money, say, $10. One profit organization randomly selected three groups player, the proposer, is instructed to offer some dollar of poor young men with high rates of crime and anti- figure (ranging from $1 to $10) to the second player, social behavior to receive cognitive behavioral ther- who is the responder. If the responder accepts the apy, a cash transfer, or both (Blattman, Jamison, and proposer’s offer, the money is shared according to the Sheridan 2014). Researchers found that the combined offer. But if the responder refuses the offer, each player treatment was associated with large and sustained gets nothing. The self-interest hypothesis suggests decreases in antisocial behavior such as crime and that the proposer should offer the minimum ($1) and violence, as well as modest long-term improvements the responder should accept it—$1 isn’t much, but it is in savings and reduced homelessness, while the cash still a gain. transfer had a short-run but no persistent effect on However, only a minority of people behave in this poverty. The intervention was designed to promote manner. Average proposer offers are often one-third future orientation, self-discipline, and new norms to one-half of the overall amount, and low offers are of nonviolent, cooperative behavior; it was not about routinely rejected by responders (Gintis and others delivering information but about helping individuals 2005). This finding can hold even as the sums grow adopt a new “socially aligned” identity and a related set quite large: when the game was played in Indonesia, of skills and behaviors. proposers continued to make sizable offers when the Cognitive interventions can also help address the amount was approximately three times the partici- psychologically destructive consequences of negative pants’ average monthly expenditures (Cameron 1999). social identities. In India, female sex workers often An interesting twist on the game reveals why this may face considerable stigma and social exclusion, which occur: low offers randomly generated by a computer can lead to self-defeating behavior and attitudes that rather than a person are rarely rejected (Blount 1995, contribute to the persistence of poverty. Working cited in Gintis and others 2005), indicating that it is with an NGO, researchers designed an eight-week uncooperative intentions rather than particular out- training program in which participants met once comes that trigger a desire to punish. Similarly, brain a week for discussion sessions aimed at building imaging studies show that punishing norm violators women’s self-esteem and increasing their sense of activates neural pathways associated with reward agency (Ghosal and others 2013). The randomized processing (de Quervain and others 2004). Language experiment showed that the intervention improved also captures the idea; many cultures have proverbs self-reported measures of self-esteem, agency, and expressing the feeling that “revenge is sweet.” happiness among members of the treatment group Another experimental tool, the “public goods game,” and improved workers’ future-orientation: women shows how critical punishment opportunities are for who particpated in the program were more likely to achieving broader-scale cooperation. The game begins choose a future-oriented savings product and to have with each player privately choosing how much of his visited a doctor, even though the training program or her individual endowment to contribute to a public included no specific mention of health issues. fund. Contributions are then multiplied such that the public good payoff is maximized when players contrib- Intrinsic reciprocity and the attainment of ute their entire endowments. At first, researchers set collective goods up the game so that there was no way for players to A key assumption in standard economics is that public punish low contributors (Fehr and Gächter 2000). In and collective goods are problematic because everyone the first round of play, approximately half the partic- prefers to take advantage of (free ride on) the efforts ipants contributed to the public fund. But over time, of others. Yet experiments show that many people are cooperation unraveled as people realized that others willing to reward others who cooperate, and punish were not “doing their share” and stopped contributing those who do not. There are two different motives that themselves. The result was a very low level of coopera- can explain this behavior: instrumental reciprocity tion after 10 periods. However, when researchers intro- and intrinsic reciprocity. Responding to kindness with duced opportunities for players to award noncontrib- kindness in order to sustain a profitable long-term utors “punishment points” (the laboratory equivalent relationship is an example of instrumental reciprocity. of being able to scold or ostracize a free rider), things In contrast, intrinsic reciprocity is an intrinsically moti- changed. Although punishing was personally costly THINKING SOCIALLY 47 for the individuals doing it, contributions to the public beverage bar (that is, where the consumer is expected good immediately increased, and behavior converged to be on his or her honor and pay for drinks reliably) to almost its full cooperative potential after another 10 in a university department in England (figure 2.5). periods. Figure 2.3 shows the strikingly different pat- Researchers alternated pictures of watchful eyes with terns of cooperation under the two regimes. pictures of flowers above the price list for tea, coffee, The key implication from this body of work is that and milk pasted on a cupboard each week and observed many people are conditional cooperators who prefer to how monetary contributions to the cash box varied cooperate to the degree that others are cooperating. over 10 weeks (Bateson, Nettle, and Roberts 2006). The Figure 2.4 contrasts the assumption from standard results were striking. The contributions per liter of milk economics—that everyone is a free rider—(panel a) consumed (the best available measure of total beverage with the actual distribution of free riders versus condi- consumption) were much higher in “eye weeks” than in tional cooperators observed in eight countries, includ- “flower weeks.” Every time the picture was changed to a ing Colombia and Vietnam, when subjects played pub- pair of eyes, contributions for the week soared. The pre- lic goods games (panel b) (Martinsson, Pham-Khanh, cise nature of the mechanism responsible for the effect and Villegas-Palacio 2013). Although the proportion is unclear: the eyes might have reminded people about of cooperators varies substantially by country, in no the cooperative nature of the honor bar, or they might country do free riders make up a dominant share of have triggered a concern for individual reputation. the population. In other words, the canonical model of Either way, however, the study points to the influence human behavior is not supported in any society that that perceived observation has on behavior. has been studied (Henrich and others 2004). This dynamic may help explain why in Nepal, To investigate whether conditional cooperation among 200 irrigation systems studied, farmer- could help support management of a commons, managed systems achieved higher agricultural yields Rustagi, Engel, and Kosfeld (2010) studied 49 forest user and more equitable distributions of water and were groups in Ethiopia. Combining experimental measures better maintained than government-managed sys- of conditional cooperation and survey measures of tems. Farmers in farmer-managed systems were about monitoring activity, they showed that the percentage of twice as likely as farmers in goverment-managed conditional cooperators varied per group, that groups systems to report that rules were observed and that with a higher share of conditional cooperators were more successful in managing forest commons, and that costly behavior monitoring was a key means by Figure 2.3 Opportunities to punish free riding increase which conditional cooperators enforced cooperation. cooperation In line with theoretical predictions, the conditional Cooperation quickly unravels in a public goods game when individuals cannot cooperators spent the most time conducting forest punish free riding. The introduction of costly punishment opportunities patrols, spending on average 32 hours per month moni- immediately increased cooperation, which converged to almost its full potential toring—1.5 times more than free riders spent. The study after 10 rounds of play. demonstrates that voluntary cooperation can be an 100 important element of commons management. Average contribution rate (%) 90 Voluntary cooperation is fragile because individual willingness to cooperate depends on expectations 80 about the cooperation of others. However, research 70 indicates that people will select into institutions with 60 like-minded cooperators and use efficient punishment to sustain cooperation when they have a chance to do 50 so (Gürerk, Irlenbusch, and Rockenbach 2006; Fehr 40 and Williams 2013). The implication is that policy mak- 30 ers should take into account not only selfish but also cooperative instincts when considering interventions 20 and societal institutions. Building in opportunities 10 for people to observe others’ behaviors—for instance, 0 by making behaviors more public—may be a useful 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 means of bolstering the expectations and therefore the Time periods practice of cooperation. Without punishment opportunity With punishment opportunity To see why, consider what happened when research- ers created the illusion of “being watched” at an honor Source: Fehr and Gächter 2000. 48 WORLD DEVELOPMENT REPORT 2015 Figure 2.4 In experimental situations, most people behave as conditional cooperators rather than free riders The standard economic model (panel a) assumes that people free ride. Actual experimental data (panel b) show that across eight societies, the majority of individuals behave as conditional cooperators rather than free riders when playing a public goods game. The model of free riding was not supported in any society studied. a. Behavior predicted b. Actual behavior revealed in experiments in standard economic model Percentage of the population demonstrating 100 90 80 contributor behavior 70 60 50 40 30 20 10 0 Colombia Vietnam Switzerland Denmark Russian United Austria Japan Federation States Free riders Conditional cooperators Source: Martinsson, Pham-Khanh, and Villegas-Palacio 2013. Note: Other players did not fit into either of the two categories, which is why the bars do not sum to 100 percent. rule infractions were recorded, and farmer-managed cantly improved student learning, student presence, systems were more likely to impose fines for misbe- and teacher presence compared to the government-led havior (Joshi and others 2000, reported in Ostrom process, which had no impact. An experiment indicated 2005). Thus placing control of a resource system and that the participatory process succeeded by increas- associated sanctioning powers in the hands of ben- ing parents’ willingness to contribute to a shared eficiaries rather than the government may improve good—improved school performance. Here, altering outcomes by harnessing people’s natural instincts to the institutional environment was a low-cost means monitor others and respond to watchful observation of eliciting socially beneficial behavior and improving by those around them. school performance. Programs that harness cooper- Similarly, in Uganda, parental participation in a ative motivations in this manner may represent an social accountability process proved more effective alternative to more traditional incentive programs for than a process that was more top down. Researchers school performance. tested two variants of a community monitoring inter- vention involving a school scorecard (Serneels, Zeitlin, Policies to “crowd in” rather than “crowd and Barr 2014). In one variant, a monitoring committee out” cooperation was given a scorecard designed by researchers in con- The apparent ubiquity of preferences for cooperation junction with the Ministry of Education. In the second raises questions about the appropriate role of incen- variant, committees designed their own scorecards in tives in policy. Many policies rest on the assumption a participatory process. Although the scorecards were that external incentives must be used to induce people substantively similar, the participatory variant signifi- to contribute to collective goods. But what if people THINKING SOCIALLY 49 are motivated to cooperate out of altruism or intrinsic Figure 2.5 The power of social monitoring: Pictures of reciprocity, as the previous section suggested that many eyes increased contributions to a beverage honor bar are? A body of research examining this question indi- cates that incentives and other institutional arrange- Voluntary contributions to a beverage honor bar increased when a picture of eyes placed on a cupboard suggested that individuals’ behavior was ments can both “crowd out” and “crowd in” innate pref- being watched. The figure shows the amount of money contributed per erences for cooperation (see Bowles and Polania-Reyes liter of milk each week for 10 weeks, as pictures of eyes were alternated 2012 for a review of this literature; a meta-analysis of with pictures of flowers. Eye images were always associated with higher relevant studies is Deci, Koestner, and Ryan 1999). contributions. The crowding-out phenomenon is illustrated by a study of day-care centers in Israel that began fining 10 parents who arrived late to collect their children. While the intent was to deter tardy pickups, the program 9 increased the number of late-coming parents. What 8 happened? The program put a price on what had previ- Image shown each week ously been a moral behavior (punctuality); it reframed 7 late pickups from a morally inappropriate action into an economically legitimate one in which parents could 6 simply “buy” extra time in a consensual exchange (Gneezy and Rustichini 2000). Relatedly, paying people 5 to participate in the communal task of cutting grass in a schoolyard in Tanzania diminished their satisfaction 4 compared to those who simply volunteered (Kerr, Vard- han, and Jindal 2012). 3 The crowding-in phenomenon was illustrated 2 in figure 2.3. In that case, giving Swiss students the ability to punish (fine) noncooperation “crowded in” 1 preferences for cooperation by penalizing free riding and reassuring people that cooperation was likely to 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 prevail. £ paid per liter of milk consumed Thus incentives and social preferences are some- times complementary and sometimes substitutes, in Source: Adapted from Bateson, Nettle, and Roberts 2006. part because incentives do not simply change prices; they also carry social meanings that depend on the choices (see, for example Simmel 1955; Granovetter relationships among the actors and their preexisting 1985). Social networks are the sets of actors and rela- cultural frameworks (Bowles and Polania-Reyes 2012). tional ties that form the building blocks of human In addition, because preferences for altruistic contri- social experience. Networks provide scope for indi- butions and altruistic punishment vary across indi- viduals to reinforce existing behaviors among one viduals and societies, the optimal policy response may another, but they can also transmit novel information include targeting different behavioral types with dif- and normative pressures, sometimes sparking social ferent interventions (Herrmann, Thöni, and Gächter change.4 The ability of social networks to both stabilize 2008). The implication for policy makers is twofold. and shift patterns of behavior means that they may be First, predicting the effect of an incentive is challeng- able to play an important role in social settings where ing—meaning that testing incentive programs in the formal institutions are lacking. population where they will be deployed is likely to be Social networks are a foundational—and often over- an important step in getting a program right (chapter looked—basis of social order. They are a distinctive 11). Second, altering the institutional environment to form of economic coordination with their own logic, nurture social preferences can be a means of motivat- in which price and authority—the coordinating mech- ing contributions to a collective good. anisms of markets and hierarchies, respectively—are downplayed, but social obligation and reputation loom The influence of social networks large (Powell 1990). Networks also have implications on individual decision making for political and governance outcomes. In India, cities All of us are embedded in networks of social relations in which voluntary associations contained both Hindu that shape our preferences, beliefs, resources, and and Muslim members experienced much less religious 50 WORLD DEVELOPMENT REPORT 2015 violence in the latter half of the 20th century than cit- ing the number or quality of interactions among people ies where such civic links did not exist. Evidence from can build those qualities to support development. In cities experiencing similar types of “conflict shocks” India, researchers found that microfinance clients ran- but different outcomes in terms of riots suggests that domly assigned to meet weekly rather than monthly civic leaders’ interactions within ethnically diverse with their repayment groups had much more informal networks played a critical role in controlling tensions social contact with others in the group even two years (Varshney 2001). after the loan cycle ended, exhibited a greater willing- Social networks can both aid and undermine the ness to pool risk with group members at that time, and achievement of development goals. The success of were three times less likely to default on their second microfinance in developing countries, for instance, loans (Feigenberg, Field, and Pande 2013). is often attributed in part to the close relationships Increasing social ties is also a focus of health among borrowers that channel social pressure to researchers who have explored how relationships encourage repayment (see chapter 6). Yet development help shift health behaviors (see chapter 8). A new goals are also achieved with technologies such as com- program in Lebanon is striving to increase breast- mitment devices that encourage savings by insulating feeding by training female friends of new mothers people from social demands for financial assistance to serve as “support mothers” and soliciting women (see chapter 4). In other words, interventions achieve in the community who have successfully breastfed to their objectives by harnessing some social pressures also serve in this capacity. These “social supports” are and diminishing others. paired with informational breastfeeding classes before birth and professional lactation support to create a multifaceted approach to changing health behavior Taking the effect of social norms into (Nabulsi and others 2014; for a review of the efficacy of community-based interventions for breastfeeding account can lead to better policy design. in low- and middle-income countries, see Hall 2011). Results are not yet available, but the program is a model of how a social network approach may be com- Complicating things even more is the fact that inter- bined with standard informational interventions in an ventions can initiate broad changes in the social fabric attempt to increase program success. that are difficult to foresee. Commitment devices, for A network perspective suggests that increasing the instance, can increase savings among target individ- density of social relations can also improve the civic uals by reducing the social pressure to share financial culture or “social capital” of a community (Putnam, resources with kin, but what if they do so by weakening Leonardi, and Nanetti 1993; for an analysis of the lim- broader sharing norms, thereby compromising individ- itations of this approach as applied to development, see uals’ access to food and child-rearing assistance from Portes and Landolt 2000). Community-driven devel- others during tough times? The trade-off may not be opment (CDD) programs are a class of intervention worth it (Case 2007). In a similar vein, offering insur- founded on this logic. A recent review of CDD projects ance to only those farmers who possess an insurable found that greater community involvement modestly asset (common in developing countries) may actually improves resource sustainability, infrastructure qual- increase the wage risk for landless laborers in Indian ity, and the delivery of health and education services villages by increasing labor demand volatility (Mobarak (Mansuri and Rao 2013). and Rosenzweig 2013). In light of these studies, when Yet increasing social ties with the aim of transform- policies that target social networks are being con- ing civic culture can be challenging for policy makers. sidered, policy makers must pay close attention to how For instance, an analysis of a field experiment in a interventions may affect social relationships in the World Bank program in Sudan aimed at increasing population. civic participation in the wake of a civil war found There are several promising policy interventions that the intervention did not increase altruistic and that harness the influence of social networks to spur cooperative behaviors in lab-in-the-field experiments social change. nor did it increase social network density (Avdeenko and Gilligan 2014). Researchers suggested that CDD Increasing interactions to support new programs too often involve social mobilizers telling behaviors and build civic capacity citizens about the benefits of participation while fail- Some important traits of communities and groups— ing to actually increase social interactions. Both theory such as trust, social cohesion, and cooperation—reside and empirical research suggest that for CDD programs within relationships more than in individuals. Increas- to be effective, practitioners must find ways to help THINKING SOCIALLY 51 citizens actually engage with one another and ways to to talk, when to listen, when to discuss personal mat- help those interactions continue after project incen- ters, when to use contractions, when (and with respect tives disappear (Mansuri and Rao 2013). to what) to purchase insurance” (Sunstein 1996, 914).5 However, social norms are rarely chosen by those Targeting specific individuals to lead and who are subject to them. Many social norms are the amplify social change result of historical circumstances and accumulation Targeting particular types of individuals within a of precedent and are self-reinforcing, regardless of network can make policies more effective and less whether they promote welfare or not (see also chapter costly because they tap into social learning processes 3). Consider a simple example regarding punctuality. that leverage social influence to shift behavior. People Although we often think of punctuality and tardiness learn new ways of doing things from one another, and as innate traits of particular individuals or cultures, they often are more likely to change their behaviors these behaviors are also just “best responses” to the when new practices are embraced by close associates expectations we have about others (Basu and Weibull or others who are most similar or most salient to 2003). If I expect you to be on time and to shame me for them. A randomized experiment in China, for instance, being late, I will arrive on time. But if I expect you to be showed that farmers were more likely to take up 15 minutes late, then I may well prefer to use those weather insurance when they had a friend who had 15 minutes to finish up some paperwork and arrive late participated in an intensive information session about myself. the nature and benefits of the product first; the “net- Such so-called equilibrium behaviors can have very work effect” was half that of attending an information serious consequences for development. For instance, session directly and was equivalent to decreasing the a society can settle into a discriminatory equilibrium average insurance premium by 13 percent (Cai, de in which immigrants do not assimilate because they Janvry, and Sadoulet, forthcoming). The study sug- expect systematic discrimination by the natives, and gests that social networks can amplify the effects of rooted natives are able to identify immigrants who a standard information program to increase adoption have not assimilated and reveal their distaste for them of new products and services. Combining social net- (Adida, Laitin, and Valfort 2014). Although it might be work strategies with a traditional incentive approach that most individuals would prefer a society that fos- is similarly promising. A recent experiment showed tered assimilation and equal treatment, neither the that offering farmers a small performance incentive immigrants nor the rooted natives may have a reason to communicate to peers the benefits of a new seed to change their behavior, given their expectations about technology was a cost-effective means of inducing what others will do. In such a situation, the economic adoption of new agricultural technologies in Malawi integration and success of immigrants remains limited. villages compared to deploying government-employed In the examples above, inefficient social norms may extension workers or strategically chosen lead farmers be maintained simply because of the coordinating role (BenYishay and Mobarak 2014). they play in a society. Yet social norms are also main- tained due to their “grip on the mind” (Elster 1989). The role of social norms in Social norms can evoke strong emotions in people, individual decision making and they often possess an expressive value in the com- Social norms—broadly shared beliefs about what group munities in which they operate. As a result, breaking members are likely to do and ought to do—are infor- a social norm often creates shame and stigma for the mal governance mechanisms that exert a powerful person doing it (Goffman 1959).6 In these ways, social influence on individual decision making and behavior. norms can have large effects on both collective welfare Norms are the “glue” or “cement” of society (Elster and individual agency (Boudet and others 2013). For 1989). Humans are hard wired to develop and adhere to instance, social and legal norms around gender and norms; imitation is one of the key ways humans learn sexuality strongly influence whether women and sex- strategies for interacting in the world (Henrich and ual minorities can be educated and employed, whether Henrich 2007), and young children quickly learn the they can serve as leaders and participate in civic activ- “social rules of the game,” following norms and pun- ities, and under what conditions they bring honor or ishing violators (Rakoczy, Warneken, and Tomasello shame to their families (Klugman and others 2014). 2008). The human propensity to develop norms is so Altering social norms that contribute to undesirable strong that norms emerge for almost every behavior: social outcomes is an obvious policy goal. However, there are norms for “littering, dating, smoking, sing- predicting how norms may interact with policy is dif- ing, when to stand, when to sit, when to show anger, ficult. In a recent field experiment, a civic education when, how, and with whom to express affection, when course in Mali actually widened the gender gap in 52 WORLD DEVELOPMENT REPORT 2015 civic participation by increasing the salience of civic cooperation, a phenomenon discussed earlier in this activity, which increased the social costs for females chapter (Frey and Torgler 2007). When people feel that who participated in civic life (Gottlieb 2014). The inter- the tax system is fair and that others are obeying the vention reduced knowledge gaps, but it exacerbated law, they are much more likely to comply with their gender inequality. obligations (Rothstein 1998). And since most individuals Knowledge about the intersection of policy, social are reciprocators, their decisions in a collective setting norms, and behavior is only just beginning to accu- feed on one another, setting a society on a trend toward mulate, and a great deal more research is needed. This either higher or lower tax compliance (Kahan 2005). section offers a glimpse of some policies that use an Viewing taxpaying through the lens of norm adher- understanding of norms to generate social change. ence, fairness concerns, and reciprocity provides an explanation for why standard tax policies sometimes Designing policy to “work around” the fail and suggests the utility of new types of policies. behavioral effects of social norms Auditing crackdowns that emphasize penalties may In some cases, policy makers may be able to bypass have exactly the opposite of the intended effect, if the behavioral effects of social norms. Consider the the increased sanctions “cue” the idea that evasion is problem of where to locate public schools. In Pakistan, widespread (Sheffrin and Triest 1992). In contrast, pol- many girls who wish to attend school must cross two icies that emphasize the extent of tax compliance and types of social boundaries: caste boundaries and gen- encourage the perception that tax evaders are deviants der boundaries. Low-caste girls may experience stigma may be successful. Tax payments in the state of Min- and face discrimination if they attend a school domi- nesota increased when people were informed of high nated by high castes, and all girls are subject to purdah, compliance rates, but did not increase when people a form of female seclusion that restricts women’s were informed of higher audit rates (Coleman 1996). mobility and social interactions.7 These social con- In the United Kingdom, compliance increased more straints limit educational opportunities for girls. Con- when citizens received letters noting that most people trasting two hypothetical policies, Jacoby and Mansuri in their postal code had already paid their taxes than (2011) show that a policy of providing schools to ham- when the letter did not contain this information about lets dominated by low-caste individuals would increase social norms (BIT 2012). enrollment by almost twice as much as a policy of plac- Policies that use brief communication interventions ing a school in every unserved hamlet, and would do so to correct misperceptions of other people’s behaviors at one-sixth of the cost. and attitudes may be particularly useful in reducing risky behavior when the difficulty of observing a “Marketing” existing social norms to behavior makes it difficult to correctly estimate how shift behavior common it is. In a township in South Africa, a coun- Some behaviors that are important for development, try with one of the highest HIV infection rates in the such as paying taxes and using toilets, vary within a world, men consistently overestimated the prevalence population. And sometimes, people misperceive how and approval of risky sexual behaviors and under- common or how accepted certain behaviors are within estimated the prevalence and approval of protective their community. Where this is the case, “marketing” behaviors. Since expectations about others’ behavior social norms can be an effective and low-cost means of often play into personal decision making, such beliefs increasing awareness of the number of people engag- may constitute a public health concern that could be ing in a behavior and correcting misperceptions about addressed by marketing the desirable social norms the frequency of a behavior. If people understand what (Carey and others 2011). others think and do, they may shift their understand- ing of existing social norms and in turn change their Activating norms to shift behavior own behavior. A powerful example of the utility of activating norms For instance, many policies aimed at increasing tax comes from an effort to reduce traffic deaths. Every revenues are based on the assumption that people are year, about 1.25 million people die from traffic acci- wealth maximizers who will evade their taxes unless dents—more than twice the number of victims from they face the right incentives, such as financial penal- war and violence combined. Ninety percent of the ties and the possibility of jail time. Yet expected penal- road deaths occur in low- and middle-income coun- ties explain very little of the variation in tax compliance tries (Lopez and others 2006). In Kenya, many of the across countries or over time (Cowell 1990). One reason people killed are passengers in minibuses, and people is that taxpaying is a social norm involving conditional are aware of the danger: one-third of respondents to THINKING SOCIALLY 53 Figure 2.6 Stickers placed in Kenyan minibuses reduced traffic accidents English translation of bottom sticker: Hey, will you complain after he causes an accident? BE AWAKE. BE STEADY. SPEAK UP! Source: Habyarimana and Jack 2011. a passenger survey conducted before an intervention tion to duel, and smoking cigarettes have been altered reported having felt that their life was in danger on a through various legal changes (Lessig 1995). recent trip (Habyarimana and Jack 2011). What is more, since people can come to value things Researchers decided to try an inexpensive behav- they experience, legal changes that shift the short-term ioral intervention to reduce accidents. Buses were costs and benefits of action can actually contribute to randomly divided into two groups. In one group, longer-term and self-sustaining behavior changes. nothing was done. In the other group, passengers Recycling programs in North American communities were reminded of their right to a safe ride on public often triggered a great deal of grumbling when they transportation. Stickers posted in the buses encour- were first instituted, and people complied mostly to aged passengers to “heckle and chide” reckless drivers avoid the increased costs of not doing so. But over (figure 2.6). The intervention was a remarkable suc- time, recycling has become a normative behavior in cess. Insurance claims involving injury or death fell by many places, even in areas with low enforcement. half, from 10 percent to 5 percent of claims. Results of Thus behaviors and values can evolve together; formal a driver survey during the intervention suggested that policy instruments that temporarily change prices passenger heckling played a role in improving safety may have long-term effects on preferences and social (Habyarimana and Jack 2009). The cost per year for a norms (Kinzig and others 2013). life saved was about $5.80, making the program even However, the efficacy of law for changing social more cost-effective than childhood vaccination, one of norms has limits. Laws that are greatly at odds with the most cost-effective health interventions available. existing social norms are unlikely to induce desired social changes. The majority of African countries have Changing social norms to shift behavior laws banning female genital cutting, for example, Engineering shifts in social norms is a far from trivial yet the practice remains widespread in many areas task. Yet norms can and do change. One tool for shift- (UNICEF 2013). ing norms is law (Sunstein 1996). Law can change not Informal strategies can also be effective for changing only incentives for action but also the social meaning norms. The use of mass media is one such strategy. In of actions. The social meanings and therefore the a randomized experiment, Rwandese communities lis- desirability of wearing a helmet, declining an invita- tened to radio soap operas containing messages about 54 WORLD DEVELOPMENT REPORT 2015 social conflict and resolution (the treatment group) or A successful example was an intervention in the reproductive health (the control group). Results from United States to reduce bullying in school. Highly con- interviews, focus groups, role-playing exercises, and nected students and “highly salient” clique leaders par- unobtrusive measures of collective decision making ticipated in a program designed to broadcast students’ indicated that the treatment program changed people’s experiences with and reactions to harassment and to perceptions of social norms regarding the appropri- facilitate public discussion on the issue. The “social ref- ateness of open expression and dissenting behavior erents” wrote and read aloud essays about harassment, (Paluck and Green 2009a). Interestingly, the interven- performed skits demonstrating the emotional effects tion altered both perceptions of norms and individ- of bullying, and sold wristbands signaling the wearers’ ual behavior, even though individual attitudes were commitment to reducing harassment. Changing the unchanged. The implication is that targeting social behavior of social referents changed peers’ perceptions norms may be a more fruitful avenue for changing of schools’ collective norms as well as actual harass- prejudiced behaviors than targeting personal beliefs, ment behavior through the mechanism of “everyday although the staying power of such interventions needs interaction” (Paluck and Shepherd 2012). further investigation. Radio soap operas are especially A key to success in many interventions is to iden- interesting because they changed people’s perceptions tify the group or social network within which a rele- of norms in conflict areas, whereas an extensive review vant norm is enforced. Is it the family, the friendship of the literature indicates that many other policies to group, the peer group, the neighborhood, or the entire reduce prejudices have been ineffective (Paluck and community? Green 2009b). Although many developing countries seek to reduce birthrates, for instance, the success of economic incen- tives designed to achieve this outcome, such as free Human sociality is like a river running contraception, has been mixed. One explanation is that fertility is regulated by social norms, so that women through society; it is a current constantly tend to choose the same, socially approved reproduc- tive practices as their most important social referents. shaping individuals, just as flowing water The result is that either most women within a tightly connected social network choose to use contracep- shapes stones in a riverbed. Policy makers tives or very few do. In Bangladesh, the institution of purdah, which limits women’s social interactions to can either work with these social currents other women within their religious group, meant that fertility shifts occurred at the level of these religious or ignore them and find themselves groups rather than across villages, in spite of common family-planning inputs across villages (Munshi and swimming upstream. Myaux 2006). This evidence suggests that fertility transitions may be better viewed as a norm-driven process than as the aggregate outcome of autonomous Some individuals and organizational actors may decisions. Thus the researchers concluded that a pro- be well-suited to leading the charge to change a social gram that encouraged women to meet at a primary norm. An actor who has a passionate interest in chang- health clinic, where they could discuss their options ing the status quo, who is well connected or highly together, might have been more effective than the central to a social network, or who has high status contraceptive program that delivered information and can play a key role in effecting broader change in a inputs to women individually in their homes. society. Such “norm entrepreneurs” can alert people to the existence of a shared complaint and suggest col- Conclusion lective solutions (Sunstein 1996; see also chapter 8 for A great deal of economic policy relies on a model of an example involving quitting smoking). If the norm human behavior that takes little account of human entrepreneur is able to reduce the perceived cost of vio- sociality. Yet humans are innately social creatures, and lating an existing norm, increase the perceived benefit the fact that we are always “thinking socially” has enor- of a new behavior, or create a persuasive new frame for mous implications for decision making and behavior, action by naming, interpreting, and dramatizing social and thus for development. This chapter demonstrates phenomena in new ways, social change can occur very that recognizing the effect of social influences on quickly (see spotlight 5 for an example from Colombia). action can help development practitioners understand THINKING SOCIALLY 55 why standard policies sometimes fail and to develop of the environment, “optimizing behavior” by individ- new interventions to combat poverty and promote uals can lead to very suboptimal social outcomes. As shared prosperity. a result, norm change may sometimes be a necessary Human sociality has several broad implications for component of social change. development interventions. First, economic incentives are not necessarily the best or the only way to motivate Notes individuals. The drive for status and social recognition 1. Sociality, social networks, and social norms also sup- means that in many situations, social incentives can be port the mental models that are internalized—and often used alongside or even instead of economic incentives shared—representations of the world. Mental models to elicit desired behaviors. This is true for both indi- are the primary subject of the next chapter. Although vidual and organizational actors. Moreover, economic there is considerable overlap among the concepts incentives can both “crowd out” intrinsic motivations explored in the two chapters, this chapter addresses the direct social influences on decision making, while and “crowd in” social preferences. The role for incen- chapter 3 focuses on internalized and enduring under- tives in policy is therefore more complicated than is standings of the world and self that often operate inde- generally recognized. pendently of immediate social dynamics. Second, we act as members of groups, for better and 2. Kamenica (2012) provides a review of how behavioral for worse. Sharing and reciprocity among group mem- economics has shaped thinking about incentives. bers and the other-regarding behavior of those who Madrian (2014) discusses uses of incentives informed take on social roles such as “group leader” can contrib- by a behavioral approach in public policy making. ute to the well-being of a community. Interventions 3. Researchers anticipated that subjects would exhibit that increase interactions or create groups among a stronger attachment to their farmer group than to individuals who have a common interest in goals such their village for several reasons: the cooperatives play as loan repayment and breastfeeding may facilitate the a central role in individuals’ welfare; membership in achievement of these objectives. Yet membership in farmer groups is voluntary rather than ascribed; mem- marginalized groups can also lead to the development ber similarity within farmer groups in landholdings, of negative social identities that affected individuals income, age, and the like may promote bonding and would likely not have chosen and would be better off identification. 4. See, for example, Munshi and Rosenzweig (2006); without. From this perspective, cognitive interven- Conley and Udry (2010); Kandpal and Baylis (2013). tions that change identities and self-perceptions can 5. For book-length treatments of social norms, see Bicch- be powerful sources of positive social change. ieri 2006; Posner 2002; Hechter and Opp 2005; Brennan Third, the widespread willingness of individuals and others 2013. to cooperate in the pursuit of shared goals means 6. Of course, some oppositional cultures define them- that institutions and interventions can be designed to selves by breaking the norms of a “dominant group,” harness social preferences. An important lab finding but in such cases, norm breaking becomes itself a nor- replicated across almost every society that has been matively prescribed activity. studied is that most people prefer to cooperate as long 7. For example, the Pakistan Rural Household Survey as others are cooperating. This finding stands in con- (PRHS-II) from 2004–05 revealed that among married trast to the traditional assumption that people prefer to women ages 15–40, 80 percent felt safe alone within shirk social obligations. One implication is that mak- their own settlement while only 27 percent felt safe ing behavior more visible and “marketing” adherence alone outside it (Jacoby and Mansuri 2011). to norms such as paying taxes or using condoms may be a cost-effective means of increasing contributions References to collective goods. Adida, Claire L., David D. Laitin, and Marie-Anne Valfort. 2014. “Muslims in France: Identifying a Discrimina- Finally, human societies develop social norms as a tory Equilibrium.” Journal of Population Economics 27: means of coordinating and regulating behavior as well 1039–86. as expressing community values, and these informal Ashraf, Nava, Oriana Bandiera, and B. Kelsey Jack. Forth- governance mechanisms have profound consequences coming. “No Margin, No Mission? A Field Experiment for societies. However, in contrast to an economic per- on Incentives for Public Service Delivery.” Journal of spective in which social norms enhance the utility of Public Economics. the individuals who uphold them, this chapter suggests Avdeenko, Alexandra, and Michael J. Gilligan. 2014. “Inter- that societies can get stuck in collective patterns of national Interventions to Build Social Capital: Evidence behavior that arguably do not serve anyone’s interests; from a Field Experiment in Sudan.” Policy Research since social norms are often taken-for-granted aspects Working Paper 6772, World Bank, Washington, DC. 56 WORLD DEVELOPMENT REPORT 2015 Baldassarri, Delia, and Guy Grossman. 2013. “The Effect of C. Kalichman. 2011. “Community Norms for HIV Risk Group Attachment and Social Position on Prosocial Behaviors among Men in a South African Township.” Behavior: Evidence from Lab-in-the-Field Experiments.” Journal of Behavioral Medicine 34 (1): 32–40. PLoS ONE 8 (3): 1–9. http://www.plosone.org/article Case, Anne. 2007. “Comment.” In Behavioral Economics and /info %3Adoi%2F10.1371%2Fjournal.pone.0058750. Its Applications, edited by Peter Diamond and Hannu Basu, K., and J. W. Weibull. 2003. “Punctuality: A Cultural Vartiainen. Princeton, N.J.: Princeton University Press. Trait as Equilibrium.” In Economics for an Imperfect Coleman, Stephen. 1996. “The Minnesota Income Tax World: Essays in Honor of Joseph Stiglitz, edited by Rich- Compliance Experiment: State Tax Results.” Research ard Arnott, Bruce Greenwald, Ravi Kanbur, and Barry Report. Minnesota Department of Revenue. http:// Nalebuff, 163–82. Cambridge, MA: MIT Press. www.revenue.state.mn.us/research_stats/research_ Bateson, Melissa, Daniel Nettle, and Gilbert Roberts. reports/19xx/research_reports_content_complnce.pdf. 2006. “Cues of Being Watched Enhance Cooperation in Conley, Timothy G., and Christopher R. Udry. 2010. “Learn- a Real-World Setting.” Biology Letters 2: 412–14. ing about a New Technology: Pineapple in Ghana.” Bauer, Michael, Alessandra Cassar, Julie Chytilova, and American Economic Review 100 (1): 35–69. Joseph Henrich. 2014. “War’s Enduring Effects on the Cowell, Frank A. 1990. Cheating the Government: The Econom- Development of Egalitarian Motivations and In-Group ics of Evasion. Cambridge, MA: MIT Press. Biases.” Psychological Science 25 (1): 47–57. Davis, Kevin, Angelina Fisher, Benedict Kingsbury, and BenYishay, Ariel, and A. Mushfiq Mobarak. 2014. “Social Sally Engle Merry. 2012. Governance by Indicators: Global Learning and Communication.” Working Paper 20139, Power through Classification and Rankings. Vol. 1. Oxford, National Bureau of Economic Research, Cambridge, U.K.: Oxford University Press. MA. Deci, Edward L., Richard Koestner, and Richard M. Ryan. Besley, Timothy, and Maitreesh Ghatak. 2008. “Status 1999. “A Meta-Analytic Review of Experiments Exam- Incentives.” American Economic Review 98 (2): 206–11. ining the Effects of Extrinsic Rewards on Intrinsic Bicchieri, Cristina. 2006. The Grammar of Society. Cam- Motivation.” Psychological Bulletin 125 (6): 627. bridge, U.K.: Cambridge University Press. De Quervain, Dominique J.-F., Urs Fischbacher, Valerie BIT (Behavioural Insights Team). 2012. Applying Behavioural Treyer, Melanie Schellhammer, Ulrich Schnyder, Insights to Reduce Fraud, Error and Debt. London: Cabinet Alfred Buck, and Ernst Fehr. 2004. “The Neural Basis Office, BIT. of Altruistic Punishment.” Science 305 (5688): 1254–58. Blattman, Christopher, Julian C. Jamison, and Margaret Eckel, Catherine C., and Philip J. Grossman. 1996. “Altru- Sheridan. 2014. “Reducing Youth Poverty, Crime and ism in Anonymous Dictator Games.” Games and Eco- Violence: Experimental Evidence from Unconditional nomic Behavior 16 (2): 181–91. Cash Transfers and Behavior Change with High-Risk Elster, Jon. 1989. The Cement of Society. Cambridge, U.K.: Men.” Unpublished paper. Cambridge University Press. Blount, Sally. 1995. “When Social Outcomes Aren’t Fair: Fehr, Ernst, and Simon Gächter. 2000. “Cooperation and The Effect of Causal Attributions on Preferences.” Punishment in Public Goods Experiments.” American Organizational Behavior and Human Decision Processes 63 Economic Review 90 (4): 980–94. (2): 131–44. Fehr, Ernst, and Tony Williams. 2013. “Endogenous Boudet, Ana Maria Muñoz, Patti Petesch, and Carolyn Emergence of Institutions to Sustain Cooperation.” Turk, with Angélica Thumala. 2013. On Norms and http://www.fdewb.unimass.nl/meteor-seminar-et Agency: Conversations about Gender Equality with Women /spring-2014/papers-and-abstracts/Paper_williams.pdf. and Men in 20 Countries. Directions in Development Feigenberg, Benjamin, Erica Field, and Rohini Pande. Series. Washington, DC: World Bank. 2013. “The Economic Returns to Social Interaction: Bowles, Samuel, and Sandra Polania-Reyes. 2012. “Eco- Experimental Evidence from Microfinance.” Review of nomic Incentives and Social Preferences: Substitutes Economic Studies 80 (4): 1459–83. or Complements?” Journal of Economic Literature 50 (2): Fong, Christina, Samuel Bowles, and Herbert Gintis. 2006. 368–425. “Strong Reciprocity and the Welfare State.” In Hand- Brennan, Geoffrey, Lina Eriksson, Robert Goodin, and book of the Economics of Giving, Altruism and Reciprocity, Nicholas Southwood. 2013. Explaining Norms. Oxford, vol. 2, edited by Serge-Christophe Kolm and Jean Mer- U.K.: Oxford University Press. cier Ythier, 1439–64. Amsterdam: Elsevier. Cai, Jinh, Alain de Janvry, and Elisabeth Sadoulet. Forth- Forsythe, Robert, Joel L. Horowitz, Nathan E. Savin, and coming. “Social Networks and the Decision to Insure.” Martin Sefton. 1994. “Fairness in Simple Bargaining American Economic Journal: Applied Economics. Experiments.” Games and Economic Behavior 6 (3): 347–69. Cameron, Lisa A. 1999. “Raising the Stakes in the Ultima- Frey, Bruno S., and Benno Torgler. 2007. “Tax Morale and tum Game: Experimental Evidence from Indonesia.” Conditional Cooperation.” Journal of Comparative Eco- Economic Inquiry 37 (1): 47–59. nomics 35 (1): 136–59. Carey, Kate B., Lori A. J. Scott-Sheldon, Michael P. Carey, Ghosal, Sayantan, Smarajit Jana, Anandi Mani, Sandip Demetria Cain, Regina Mlobeli, Redwaan Vermaak, Mitra, and Sanchari Roy. 2013. “Sex Workers, Stigma Jacqueline Mthembu, Leickness C. Simbayi, and Seth and Self-Belief: Evidence from a Psychological Train- THINKING SOCIALLY 57 ing Program in India.” Unpublished paper. http:// Pakistan.” Policy Research Working Paper 5710, World www2.warwick.ac.uk/fac/soc/economics/staff/sroy Bank, Washington, DC. /sanchari_roy_j-m-p.pdf. Joshi, Neeraj N., Elinor Ostrom, Ganesh P. Shivakoti, and Gintis, Herbert, Samuel Bowles, Robert Boyd, and Ernst Wai Fung Lam. 2000. “Institutional Opportunities and Fehr, eds. 2005. “Moral Sentiments and Material Constraints in the Performance of Farmer-Managed Interests: Origins, Evidence, and Consequences.” In Irrigation Systems in Nepal.” Asia-Pacific Journal of Rural Moral Sentiments and Material Interests: The Foundations Development 10 (2): 67–92. of Cooperation in Economic Life, edited by Herbert Gintis, Kahan, Dan M. 2005. “The Logic of Reciprocity: Trust, Samuel Bowles, Robert Boyd, and Ernest Fehr, 3–40. Collective Action and Law.” In Moral Sentiments and Cambridge, MA: MIT Press. Material Interests: The Foundations of Cooperation in Eco- Gneezy, Uri, and Aldo Rustichini. 2000. “A Fine Is a Price.” nomic Life, edited by Herbert Gintis, Samuel Bowles, Journal of Legal Studies 29 (1): 1–17. Robert Boyd, and Ernst Fehr, 339–78. Cambridge, MA: Goffman, Erving. 1959. The Presentation of Self in Everyday MIT Press. Life. Garden City, NY: Doubleday Anchor Books. Kamenica, Emir. 2012. “Behavioral Economics and the Gottlieb, Jessica. 2014. “Why Women Participate Less Psychology of Incentives.” Annual Review of Economics in Civic Activity: Evidence from Mali.” Unpublished 4 (13): 1–13, 26. paper. http://mgape.files.wordpress.com/2013/06 Kandpal, Eeshani, and Kathy Baylis. 2013. “Expanding /gottlieb_civic-behavior.pdf. Horizons: Can Women’s Support Groups Diversify Granovetter, Mark. 1985. “Economic Action and Social Peer Networks in Rural India?” American Journal of Agri- Structure: The Problem of Embeddedness.” American cultural Economics 95 (2): 360–67. Journal of Sociology 91 (3): 481–510. Kelley, Judith G., and Beth A. Simmons. Forthcoming. Grossman, Guy, Devorah Manekin, and Dan Miodownik. “Politics by Number: Indicators as Social Pressure in Forthcoming. “The Political Legacies of Combat: International Relations.” American Journal of Political Attitudes towards War and Peace among Israeli Ex- Science. Combatants.” International Organization. Kerr, John, Mamta Vardhan, and Rohit Jindal. 2012. “Pro- Gürerk, Özgür, Bernd Irlenbusch, and Bettina Rocken- social Behavior and Incentives: Evidence from Field bach. 2006. “The Competitive Advantage of Sanction- Experiments in Rural Mexico and Tanzania.” Ecological ing Institutions.” Science 312 (5770): 108–11. Economics 73: 220–27. Habyarimana, James, and William Jack. 2009. “Heckle and Kinzig, Ann P., Paul R. Ehrlich, Lee J. Alston, Kenneth Chide: Results of a Randomized Road Safety Interven- Arrow, Scott Barrett, Timothy G. Buchman, Gretchen tion in Kenya.” Working Paper 169, Center for Global C. Daily, Bruce Levin, Simon Levin, Michael Oppen- Development, Washington, DC. http://www.cgdev.org heimer, Elinor Ostrom, and Donald Saari. 2013. “Social /content/publications/detail/1421541/. Norms and Global Environmental Challenges: The ————. 2011. “Heckle and Chide: Results of a Randomized Complex Interaction of Behaviors, Values, and Policy.” Road Safety Intervention in Kenya.” Journal of Public BioScience 63 (3): 164–75. Economics 95 (11): 1438–46. Hall, Jennifer. 2011. “Effective Community-Based Inter- Klugman, Jeni, Lucia Hanmer, Sarah Twigg, Tazeen Hasan, ventions to Improve Exclusive Breast Feeding at Four Jennifer McCleary-Sills, and Julieth Santamaria. 2014. to Six Months in Low- and Low-Middle-Income Coun- Voice and Agency: Empowering Women and Girls for tries: A Systematic Review of Randomised Controlled Shared Prosperity. Washington, DC: World Bank. Trials.” Midwifery 27 (4): 497–502. http://documents.worldbank.org/curated/en/2014/10 Hechter, Michael. 1987. Principles of Group Solidarity. Berke- /20274040/voice-agency-empowering-women-girls ley, CA: University of California Press. -shared-prosperity. Hechter, Michael, and Karl-Dieter Opp, eds. 2005. Social Kosfeld, Michael, and Susanne Neckermann. 2011. “Get- Norms. New York: Russell Sage Foundation. ting More Work for Nothing? Symbolic Awards and Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin Cam- Worker Performance.” American Economic Journal: erer, Ernst Fehr, and Herbert Gintis. 2004. Foundations Microeconomics 3 (3): 86–99. of Human Sociality: Economic Experiments and Ethno- Larkin, I. 2009. “Paying $30,000 for a Gold Star: An Empir- graphic Evidence from Fifteen Small-Scale Societies. Oxford, ical Investigation into the Value of Peer Recognition to U.K.: Oxford University Press. Software Salespeople.” Working Paper, Harvard Busi- Henrich, Joseph, and Natalie Henrich. 2007. Why Humans ness School, Boston. Cooperate: A Cultural and Evolutionary Explanation. Larson, Deborah Welch, and Alexei Shevchenko. 2010. Oxford, U.K.: Oxford University Press. “Status Seekers: Chinese and Russian Responses to Herrmann, Bendikt, Christian Thöni, and Simon Gächter. U.S. Primacy.” International Security 34 (4): 63–95. 2008. “Antisocial Punishment across Societies.” Science Lessig, Lawrence. 1995. “The Regulation of Social Mean- 319: 1362–67. ing.” University of Chicago Law Review 62 (3): 943–1045. Jacoby, Hanan G., and Ghazala Mansuri. 2011. “Crossing List, John A. 2007. “On the Interpretation of Giving in Dic- Boundaries: Gender, Caste and Schooling in Rural tator Games.” Journal of Political Economy 115 (3): 482–93. 58 WORLD DEVELOPMENT REPORT 2015 Lopez, Alan D., Colin D. Mathers, Majid Ezzati, Dean T. Posner, Eric A. 2002. Law and Social Norms. Cambridge, MA: Jamison, and Christopher J. L. Murray. 2006. “Global Harvard University Press. and Regional Burden of Disease and Risk Factors, 2001: Powell, Walter, ed. 1990. “Neither Market nor Hierarchy: Systematic Analysis of Population Health Data.” Lancet Network Forms of Organization.” Research in Organiza- 367 (9524): 1747–57. tional Behavior 12: 295–336. Madrian, Brigitte C. 2014. “Applying Insights from Behav- Putnam, Robert D., Robert Leonardi, and Raffaella Y. Nan- ioral Economics to Policy Design.” Working Paper etti. 1993. Making Democracy Work: Civic Traditions in 20318, National Bureau of Economic Research, Cam- Modern Italy. Princeton, N.J.: Princeton University Press. bridge, MA. Rakoczy, Hannes, Felix Warneken, and Michael Tom- Mansuri, Ghazala, and Vijayendra Rao. 2013. “Localiz- asello. 2008. “The Sources of Normativity: Young ing Development: Does Participation Work?” Policy Children’s Awareness of the Normative Structure of Research Report, World Bank, Washington, DC. Games.” Developmental Psychology 44 (3): 875. Martinsson, Peter, Nam Pham-Khanh, and Clara Restivo, Michael, and Arnout van de Rijt. 2012. “Exper- Villegas-Palacio. 2013. “Conditional Cooperation and imental Study of Informal Rewards in Peer Pro- Disclosure in Developing Countries.” Journal of Eco- duction.” PLoS ONE 7 (3): 34358. doi: 10.1371/journal nomic Psychology 34: 148–55. .pone.0034358. Mobarak, Ahmed Mushfiq, and Mark R. Rosenzweig. Rothstein, Bo. 1998. Just Institutions Matter: The Moral and 2013. “Informal Risk Sharing, Index Insurance, and Political Logic of the Universal Welfare State. Cambridge, Risk Taking in Developing Countries.” American Eco- U.K.: Cambridge University Press. nomic Review 103 (3): 375–80. Rustagi, Devesh, Stefanie Engel, and Michael Kosfeld. Munshi, Kaivan, and Jacques Myaux. 2006. “Social Norms 2010. “Conditional Cooperation and Costly Monitoring and the Fertility Transition.” Journal of Development Eco- Explain Success in Forest Commons Management.” nomics 80 (1): 1–38. Science 330 (6006): 961–65. Munshi, Kaivan, and Mark Rosenzweig. 2006. “Tradi- Serneels, Pieter, Andrew Zeitlin, and Abigail Barr. 2014. tional Institutions Meet the Modern World: Caste, “Participatory Monitoring of Public Services by Ben- Gender, and Schooling Choice in a Globalizing Econ- eficiaries: Insights from Education Services.” Back- omy.” American Economic Review 96 (4): 1225–52. ground note for World Development Report 2015. Nabulsi, Mona, Haya Hamadeh, Hani Tamim, Tamar Sheffrin, Steven M., and Robert K. Triest. 1992. “Can Brute Kabakian, Lama Charafeddine, Nadine Yehya, Dur- Deterrence Backfire? Perceptions and Attitudes in Tax- riyah Sinno, and Saadieh Sidani. 2014. “A Complex payer Compliance.” In Why People Pay Taxes: Tax Compli- Breastfeeding Promotion and Support Intervention in ance and Enforcement, edited by J. Slemrod, 193–218. Ann a Developing Country: Study Protocol for a Random- Arbor: University of Michigan Press. ized Clinical Trial.” BMC Public Health 14 (1): 1–11. doi: Simmel, Georg. 1955. Conflict and the Web of Group Affilia- 10.1186/1471-2458-14-36. tions. New York: Free Press. (Originally published in Ostrom, Elinor. 2005. “Policies That Crowd Out Reciprocity 1922.) and Collective Action.” In Moral Sentiments and Material Sinclair, Ian. 2005. Fostering Now: Messages from Research. Interests: The Foundations of Cooperation in Economic Life, London: Jessica Kingsley Publishers. edited by Herbert Gintis, Samuel Bowles, Robert Boyd, Sobel, Joel. 2005. “Interdependent Preferences and Reci- and Ernst Fehr, 253–76. Cambridge, MA: MIT Press. procity.” Journal of Economic Literature 43 (June): 392–436. Paluck, Elizabeth Levy, and Donald P. Green. 2009a. “Def- Sunstein, Cass R. 1996. “Social Norms and Social Roles.” erence, Dissent, and Dispute Resolution: An Exper- Columbia Law Review 96 (4): 903–68. imental Intervention Using Mass Media to Change Tomasello, Michael. 2014. A Natural History of Human Norms and Behavior in Rwanda.” American Political Thinking. Cambridge, MA: Harvard University Press. Science Review 103 (4): 622–44. UNICEF (United Nations Children’s Fund). 2013. Female —— ——. 2009b. “Prejudice Reduction: What Do We Know? Genital Mutilation/Cutting: A Statistical Overview and A Critical Look at Evidence from the Field and the Lab- Exploration of the Dynamics of Change. New York: oratory.” Annual Review of Psychology 60: 339-67. UNICEF. Paluck, Elizabeth Levy, and Hana Shepherd. 2012. “The Varshney, Ashutosh. 2001. “Ethnic Conflict and Civil Soci- Salience of Social Referents: A Field Experiment on ety.” World Politics 53 (3): 362–98. Collective Norms and Harassment Behavior in a Willer, Robb. 2009. “Groups Reward Individual Sacrifice: School Social Network.” Journal of Personality and Social The Status Solution to the Collective Action Problem.” Psychology 103 (6): 899. American Sociological Review 74 (1): 23–43. Portes, Alejandro, and Patricia Landolt. 2000. “Social Cap- ital: Promises and Pitfalls of Its Role in Development.” Journal of Latin American Studies 32: 529–47. 60 WORLD DEVELOPMENT REPORT 2015 When corruption is the norm Corruption, broadly defined as the use of public office 1976 and 1982 found that refusing to grant favors could for private gain, exists in many forms. Bribery, fraud, subject a public official to complaints filed by constitu- extortion, influence peddling, kickbacks, cronyism, nepo- ents. The norm of corruption was so entrenched that the tism, patronage, embezzlement, vote buying, and election social meaning of an honest official was someone who rigging are all examples of actions that fit under that demanded no more than the going rate as a bribe for pro- umbrella term (see UNDP 2008 for full typology of types of viding a public service (Wade 1985). corruption). A common response to all forms of corruption Pressure to engage in corruption often comes from is to view them as acts committed by autonomous individ- within the bureaucracy. In the Indian example, a highly uals: a bureaucrat takes a bribe; a traffic cop shakes down institutionalized informal system had developed for the a driver; a judge sells his decision. A focus on deterring purchase of transfers from one position to another, with individual corrupt acts provides a powerful foundation for the price dependent on how much the officeholder could reform, yet misses the social element that makes corrup- expect to extract from his constituents for providing tion a persistent problem. agricultural services: “By long-established convention, 8½ Corruption in the social sense is a shared belief that percent of each contract is kicked back to the officers and using public office to benefit oneself and one’s family and clerical staff of the Division—2½ percent to the [executive friends is widespread, expected, and tolerated. In other engineer] . . . 1 percent to the clerical staff and draughts- words, corruption can be a social norm. Moreover, it has men, and 5 percent to the Supervisor and [assistant been the default social norm throughout much of history. engineer] to be split between them” (Wade 1982, 292–93). Only gradually has the principle of equal treatment for all Officials who did not participate risked punishment: Spotlight 1 before the law emerged, and in most states it is still a work supervisors developed a code language to use in reports to in progress (Mungiu-Pippidi 2013). the authorities in charge of promotions to indicate officers who were not willing to extract side payments, identify- Social expectations and mental models ing them as “tactless,” “having no grip over the people,” or perpetuate corruption “unable to manage” (Wade 1985, 483). Those who resisted It is important to understand how the decision to engage might be coaxed into compliance with stories about how in corruption takes place in the mind of a public official. the bribes received were “gifts” from farmers grateful If people believe that the purpose of obtaining office is for how hard they were working on their behalf (Wade to provide one’s family and friends with money, goods, 1982). Ironically, officials who resisted the system might favors, or appointments, then social networks can perpetu- be threatened with bogus public charges of corruption to ate the norm of corruption. Social networks can even serve encourage them to fall into line (Bayley 1966). as a source of punishment for public servants who violate These types of social expectations can become inter- that norm. In Uganda, for instance, reciprocal obligations nalized, as demonstrated in a study that found that when of kinship and community loyalty may have contributed diplomatic immunity meant they had no legal obligation to a governance outcome in which public officials needed to pay for parking violations in New York City, diplomats to use their position to benefit their network in order to be from countries where corruption is high had significantly regarded as good people (Fjeldstad 2005). Holders of public more unpaid fines than those from countries where cor- positions who did not use their influence to assist friends ruption is low (Fisman and Miguel 2007). The finding that and relatives risked derision and disrespect (Fjeldstad, country of origin can predict corrupt actions has been rep- Kolstad, and Lange 2003). licated (Barr and Serra 2010) and suggests that corruption Even people who privately deplore a norm of corrup- is at least in part associated with social norms. tion might go along with it publicly because of perceived social pressure in support of the system. Since people who express different opinions may find themselves treated as Strategies to address corruption outsiders, they will often choose to express support for Where corruption is common, acting corruptly may the status quo simply to avoid the costs of being different become automatic thinking for officials. If so, an appro- (Kuran 1997). Thus societies can get stuck in an equilib- priate countermeasure might be to create novel situations rium in which corruption is the norm, even though pri- to get them to think deliberatively about their behavior vately much of the population would prefer a clean public and reassess their attitudes and mental models about pub- service. lic service. When a nongovernmental organization (NGO) Social pressures can force even clean officials to capit- called 5th Pillar created a zero-rupee note in India with the ulate. In China, for instance, a local official was hounded inscription “I promise to neither accept nor give a bribe” by villagers who pressured him to accept gifts every time for people to hand out when asked for bribes, one official he went home. Told he would be unable to get anything was supposedly “so stunned to receive the note that he accomplished politically by refusing, the official capit- handed back all the bribes he had solicited for providing ulated. He was later arrested on charges of corruption electricity to a village.” Another “stood up, offered tea to (McGregor 2010). Similarly, a study of India between the woman from whom he was trying to extort money, WHEN CORRUPTION IS THE NORM 61 and approved a loan so her granddaughter could go to As this Report argues, fostering collective action is not college” (Panth 2011, 21). purely a matter of incentivizing self-interested individ- Because people behave differently in private from uals. People can be intrinsically motivated to cooperate how they behave when they are (or think they are) being and to punish norm violators. In fact, as experimental observed, transforming opaque corrupt acts into public findings show, “a social norm, especially where there behavior may exert social pressure on officials to uphold is communication between parties, can work as well or their positions as intended. Low-cost Internet platforms nearly as well at generating cooperative behavior as an such as ipaidabribe.com, an NGO initiative by Janaagraha externally imposed set of rules and system of monitoring in India, have made it easier for citizens to publicize and and sanctioning” (Ostrom 2000). Practitioners wishing stigmatize bribery and shame public servants who solicit to fight corruption might therefore wish to experiment bribes, although the impact of such social media initia- with campaigns that emphasize the social norm of clean tives has yet to be evaluated. Newspapers also can make government. corrupt behavior public and empower citizens with infor- mation to monitor officials. In Uganda, corruption by offi- References cials was so extensive that local schools were receiving Barr, Abigail, and Danila Serra. 2010. “Corruption and Culture: only 24 percent, on average, of the central government An Experimental Analysis.” Journal of Public Economics 94 (11): grants to which they were entitled, until newspapers 862–69. began publishing the actual amounts the schools were Bayley, David H. 1966. “The Effects of Corruption in a Develop- supposed to receive. As a result, average funding received ing Nation.” Political Research Quarterly 19 (4): 719–32. by the schools increased to 80 percent of the entitled Economist. 2010. “Georgia’s Mental Revolution.” August 19. Fisman, Raymond, and Edward Miguel. 2007. “Corruption, amount (Reinikka and Svensson 2005). Norms, and Legal Enforcement: Evidence from Diplomatic The persistent nature of long-held mental models may Parking Tickets.” Journal of Political Economy 115 (6): 1020–48. make it challenging to convince the public that gover- Fjeldstad, Odd-Helge. 2005. “Corruption in Tax Administration: nance reforms are real. Thus anticorruption campaigns Lessons from Institutional Reforms in Uganda.” Working may be more successful when their enforcement is highly Paper 2005: 10, Chr. Michelsen Institute, Bergen, Norway. conspicuous, especially when public enforcement action Fjeldstad, Odd-Helge, Ivar Kolstad, and Siri Lange. 2003. Auton- Spotlight 1 is taken against politically powerful individuals widely omy, Incentives and Patronage: A Study of Corruption in the Tanza- believed to be above the law (Rothstein 2005). As an exam- nia and Uganda Revenue Authorities. CMI Report 2003: 9, Chr. Michelsen Institute, Bergen, Norway. ple, when the government of Georgia decided to crack Johnston, Michael, and Sahr John Kpundeh. 2004. Building down on organized crime, it televised “truckloads of heav- a Clean Machine: Anti-Corruption Coalitions and Sustainable ily armed police in ski masks round[ing] up high-profile Reform, Vol. 3466. Washington, DC: World Bank. crime bosses” (World Bank 2012, 15). Social marketing Kuran, Timur. 1997. Private Truths, Public Lies: The Social Con- campaigns to advertise anticorruption efforts, as well as sequences of Preference Falsification. Cambridge, MA: Harvard targeting areas where the government can achieve “quick University Press. wins” of easily observable reductions in corruption, might McGregor, Richard. 2010. The Party: The Secret World of China’s be another way to build citizen support and start shifting Communist Rulers. London: Penguin UK. Mungiu-Pippidi, Alina. 2013. “Becoming Denmark: Historical public perceptions (Recanatini 2013). In Georgia, a public Designs of Corruption Control.” Social Research 80 (4): 1259–86. relations campaign to advertise the reformed traffic safety Ostrom, Elinor. 2000. “Collective Action and the Evolution of police included brand-new uniforms, remodeled police Social Norms.” Journal of Economic Perspectives 14 (3): 137–58. buildings that were open and full of windows to indicate Panth, Sabina. 2011. “Changing Norms Is Key to Fighting Every- transparency, and television commercials portraying civil day Corruption.” CommGAP Discussion Papers, Communi- servants as good people (World Bank 2012). As the head- cation for Governance and Accountability Program, World line of a magazine article explaining the anticorruption Bank, Washington, DC. efforts in Georgia noted, the process involved a “mental Recanatini, Francesca. 2013. “Tackling Corruption and Promot- ing Better Governance: The Road Ahead.” In Anticorruption revolution” (Economist 2010). Seeing things differently Policy: Can International Actors Play a Role? edited by Susan Rose may be a critical component of doing things differently. Ackerman and Paul Carrington, 55–71. Durham, NC: Carolina Viewed through a social lens, changing a social norm Academic Press. about corruption constitutes a collective action problem Reinikka, Ritva, and Jakob Svensson. 2005. “Fighting Corruption rather than simply the repression of deviant behavior to Improve Schooling: Evidence from a Newspaper Cam- (Mungiu-Pippidi 2013). Establishing social action coali- paign in Uganda.” Journal of the European Economic Association tions to unite and mobilize public and private actors with 3 (23): 259–67. overlapping political interests is one promising strategy Rothstein, Bo, ed. 2005. Social Traps and the Problem of Trust. Cambridge, U.K.: Cambridge University Press. that has been pursued in Ghana and in Bangalore, India. UNDP (United Nations Development Programme). 2008. Cor- Providing nonmaterial incentives for participation, such ruption and Development: A Primer. New York: UNDP. as a shared sense of purpose, feelings of solidarity, and Wade, Robert. 1982. “The System of Administrative and Political public prestige, may be particularly important to sus- Corruption: Canal Irrigation in South India.” Journal of Devel- taining a broad coalition with varied interests (Johnston opment Studies 18 (3): 287–328. and Kpundeh 2004). The Internet may make it easier for ————. 1985. “The Market for Public Office: Why the Indian dispersed interests to organize. In Brazil, the campaign at State Is Not Better at Development.” World Development 13 (4): 467–97. avaaz.org collected signatures from 3 million citizens and World Bank. 2012. Fighting Corruption in Public Services: Chron- may have encouraged the legislature to pass a bill prevent- icling Georgia’s Reforms. Directions in Development: Public ing candidates with criminal records from running for Sector Governance. Washington, DC: World Bank. http:// office (Panth 2011). elibrary.worldbank.org/doi/book/10.1596/978-0-8213-9475-5. 3 CHAPTER Thinking with mental models When we think, we generally use concepts that we it would be impossible for people to make most deci- have not invented ourselves but that reflect the shared sions in daily life. And without shared mental models, understandings of our community. We tend not to it would be impossible in many cases for people to question views when they reflect an outlook on the develop institutions, solve collective action problems, world that is shared by everyone around us. An impor- feel a sense of belonging and solidarity, or even under- tant example for development pertains to how people stand one another. Although mental models are often view the need to provide cognitive stimulation to shared and arise, in part, from human sociality (chapter children. In many societies, parents take for granted 2), they differ from social norms, which were discussed that their role is to love their children and keep them in the preceding chapter. Mental models, which need safe and healthy, but they do not view young children not be enforced by direct social pressure, often capture as needing extensive cognitive and linguistic stimu- broad ideas about how the world works and one’s place lation. This view is an example of a “mental model.” 1 in it. In contrast, social norms tend to focus on particu- In some societies, there are even norms against verbal lar behaviors and to be socially enforced. There is immense variation in mental models across societies, including different perceptions of Mental models help people make sense of the way the world “works.” Individuals can adapt their mental models, updating them when they learn that the world—to interpret their environment outcomes are inconsistent with expectations. But this chapter will explain that individuals may hold on to and understand themselves. Mental mental models that have destructive consequences for their lives and may continue to use them to validate models include categories, concepts, their interpretations, even when those models and interpretations are patently false. Individuals can also identities, prototypes, stereotypes, causal hold onto multiple and sometimes even contradictory mental models—drawing on one or another mental narratives, and worldviews. model when the context triggers a particular way of looking at the world. Mental models matter for development because engagement between parents and young children (see they affect decision making. Since a great deal of policy chapter 5). This particular mental model can have huge is based on changing people’s decisions—to save and consequences, even leading to the intergenerational invest, to enroll children in school, to be active citizens, transmission of poverty. to be honest—understanding the role that mental Mental models include categories, concepts, iden- models play in individual decision making opens up tities, prototypes, stereotypes, causal narratives, and the possibility of new levers for policy, while at the worldviews.2 Without mental models of the world, same time highlighting potential problems in design THINKING WITH MENTAL MODELS 63 and implementation. Development interventions can systematically to a group of people, mental models can go wrong when policy designers have a faulty mental entrench poverty. model of how a population will react to a program. This For example, in Ethiopia disadvantaged individ- chapter highlights recent progress in understanding uals sometimes report feelings of low psychological the role that mental models play in economic develop- agency: “We have neither a dream nor an imagina- ment and the implications for policy. tion”; “We live only for today” (Bernard, Dercon, and The first part of the chapter outlines some of the Taffesse 2011, 1). They may believe that they cannot main ways that mental models affect decisions asso- change their future, and that belief limits their ability ciated with development. Mental models bear on the to see opportunities they might have, for example, to evolution of institutions, and on firm behavior and invest. In 2010, a team of researchers set out to discover cognitive performance. The chapter explains how men- whether they could change that mental model. They tal models, which may once have been well adapted to transported video equipment to remote Ethiopian the situation at hand or may once have reflected the villages with camels and four-wheel-drive vehicles. distribution of political power, can persist even when They invited a randomly selected group of villagers to they are no longer adaptive or when the political forces watch an hour of inspirational videos—four documen- that gave rise to them have changed. The chapter then taries in which individuals from the region described discusses implications for policy. how they had improved their socioeconomic positions Institutions and mental models are closely related; by setting goals, making careful choices, persevering, sometimes a change in a mental model requires a and working hard. A survey conducted six months change in an institution. But in some cases, exposure later found that the treatment had increased aspira- to alternative ways of thinking and to new role mod- tions and brought about small but tangible changes els—in real life, in fiction, and through public delib- in behavior: viewers had higher total savings and had eration—can have a measurable influence on mental invested more resources in their children’s schooling models and on behaviors, such as investment and (Bernard and others 2014). education. How mental models work and Where mental models come how we use them from and why they matter At a given moment, there are potentially thousands of Some mental models seem innate among humans. details that could be observed, but we have limited For example, there is some evidence that humans are powers of observation. Mental models affect where innately attuned to the category of “dangerous ani- we direct our attention.3 Mental models provide us mal.” It is easy to condition individuals to fear rats, with default assumptions about the people we inter- but it may be impossible to condition infants to fear act with and the situations we face. As a result, wooden objects and cloth curtains (Bregman 1934). we may ignore information that violates our assump- Humans may also be innately susceptible to finding tions and automatically fill in missing information certain objects or behaviors disgusting, such as incest based on what our mental models suggest is likely to (Haidt 2012). Other mental models are idiosyncratic— be true.4 acquired through experience particular to the individ- The links between perception and automatic think- ual. Many mental models come from experiences that ing are strong, as emphasized by Kahneman (2003) are particular to an environment and thus are widely and discussed in chapter 1. While both involve the shared within one society but not in others (Berger and construction of meaning, in both cases the perceiver Luckmann 1966; Zerubavel 1999). or thinker is not aware of constructing anything. He Mental models enable thought and action, but also imagines that he is responding objectively to the stim- constrain them. When the mental models people use ulus or the situation. Since we are social animals, our are well adapted to the task at hand, they make indi- mental models often incorporate the taken-for-granted viduals better off: “Time and energy are saved, rumi- beliefs and routines of the culture in which we were nation and doubt are reduced, and nothing important raised. One way of thinking about culture is as a set is lost,” note Ross and Nisbett (1991, 77; see also Todd of widely shared tools for perception and construal.5 and Gigerenzer 2000). But mental models may be out The tools may not be fully consistent with one another. of sync with the real world, may substantially limit Thus, as this chapter will show, a given person might the amount of information decision makers use, and exhibit different behaviors when the mental model may cause them to fill in uncertain details of a situa- that is most accessible to him or her changes (Swidler tion with incorrect assumptions. When this happens 1986; DiMaggio 1997). 64 WORLD DEVELOPMENT REPORT 2015 Context can activate a particular mental model, as a theme introduced in chapter 1—individuals tend to illustrated in figure 3.1. The windows provide partial automatically jump to conclusions based on limited views of an urban street. Depending on which window information—and also the main idea of this chapter: a spectator looks out of—a metaphor for which mental thinking processes (both automatic and deliberative, model he uses to interpret the world around him—his as discussed in chapter 1) draw heavily on learned mental representation of the scene will be very dif- mental models. ferent. He is unlikely to be aware that his view might As studies of immigrants show, mental models can be different if he were standing somewhere else or be passed down from generation to generation: mental using a different mental model. This figure illustrates models of trust, gender, fertility, and government, for Figure 3.1 What we perceive and how we interpret it depend on the frame through which we view the world around us Individuals do not respond to objective experience but to their mental representations of experience. In constructing their mental representations, people use interpretive frames provided by mental models. People may have access to multiple and conflicting mental models. Context can activate a particular mental model. Using a different mental model can change the individual’s mental representation of the world around him. THINKING WITH MENTAL MODELS 65 instance, are typically learned from the culture one Agricultural modes of production influence mental grows up in. Social learning processes allow for the models in other ways, as well. Cultivating rice requires intergenerational transfer of mental models.6 A soci- more specialization by teams than cultivating wheat. ety’s past may affect the perceptions and evaluations People from areas of China suited for growing rice of opportunities by current members of the society. tend to have more collectivist views, while people from areas suited for growing wheat tend to have more indi- The roots of mental models vidualist views, controlling for other factors (Talhelm Evidence suggests that historical experience exerts and others 2014). a powerful influence on mental models and, conse- Societies whose current economic activities yield quently, on how individuals understand and react to greater gains to cooperation (for example, because they the world. An example is the legacy of the Atlantic rely on whale-hunting, which requires large teams, as slave trade. Slavery was ubiquitous in many eras and contrasted with economic activities carried out at the in many societies, but the slavery associated with the individual or family level) exhibit more cooperative Atlantic slave trade had some properties that made it behavior in experimental games, Henrich and others especially destructive. The middlemen for the white (2001) show. The link between economic activities slave traders included local Africans. To protect and modes of societal organization has long been rec- themselves from being captured and sold as slaves, ognized, whereas it is only more recently that social individuals needed guns, but to buy guns they needed scientists have demonstrated that economic activi- cash. The main way of obtaining cash was to kidnap ties also shape individuals’ mental models of social someone and sell him into slavery. Thus the Atlantic interaction. slave trade turned brothers against each other, chiefs against subjects, and judges against defendants. The relationship between institutions and Lower levels of trust in some parts of Africa today mental models are related to the intensity of slave trading centuries Some of the best evidence of the impact of mental ago. Regions that were more susceptible to slave raids models on development is that changes in exposure due to accidental features of geography have lower to alternative historical institutions appear to change levels of trust today—trust toward strangers, friends, trajectories of growth, holding constant all other fac- relatives, and institutions (Nunn 2008; Nunn and tors (Guiso, Sapienza, and Zingales 2013; Nunn and Wantchekon 2011). Wantchekon 2011). Much of this work focuses on the Historical modes of production, determined by effect of historical institutions on interpersonal trust. geography or circumstance, also influence mental The weight of a large body of evidence is that trust models. One way to coordinate production within in people outside one’s own family or social group is the household is to use social norms for the division strongly positively related to economic growth.7 Eco- of labor by gender. A technology that increases the nomic efficiency and growth require the exchange of comparative advantage of one sex in a given produc- goods and services to capture economies of scale. Such tive activity increases the benefits to specialization. exchange requires trust: trust that one will be paid as The plough is such a technology. Because it requires a promised, that disputes will be resolved in good faith, great deal of upper-body strength, it gave men a large or, if not resolved, that a third party can step in and comparative advantage in agriculture. Its adoption apply rules in a predictable and fair way. In the absence may have been the source of social norms that are now of trust, microevidence shows that parties will also be used to enforce gender differences in nonagricultural less willing to delegate responsibilities and less willing domains. Today, ethnic groups that used plough-based to specialize, which can result in inefficiency within agricultural techniques in the distant past have more a firm and reduced growth within a country (Bloom, unequal gender norms and stricter norms about gen- Sadun, and Van Reenen 2012). dered activity, as Alesina, Giuliano, and Nunn (2013) Weak constraints on a ruling group are a potent show. The norms persist even when people leave these cause of low national incomes that remain remarkably societies. The children of immigrants to Europe and persistent over time (Acemoglu and Robinson 2012). the United States have gender norms that depend on The standard argument for explaining that persistence whether or not their culture of origin used plough- is that inequality of wealth shapes rules, which then based agricultural technology. Working women are tend to preserve the initial inequality of wealth. But viewed more favorably in societies that did not have social scientists have long argued that institutions the plough than in societies that did, and they rep- have a “schematizing role”; they are not just rules, but resent a higher share of the labor force. also a way of seeing.8 Institutions shape the categories 66 WORLD DEVELOPMENT REPORT 2015 and concepts that people accept as natural and use to much land the traditional landlord caste owns. In interpret the world. villages in which the traditional caste is dominant Recent work on governance in Sierra Leone and (dominant caste members own at least half the land), India can be interpreted in this light. In Sierra Leone, a system of clientelism appears to prevail that does “paramount chiefs,” who are elected for life, rule the not occur in other villages. Workers “sell” their votes villages. The only people who can compete to be a par- to landlords in the dominant caste in exchange for amount chief are those from one of the ruling families insurance and access to the trading network. The vote originally recognized by British colonial authorities. For selling eliminates political competition. Just as in the accidental historical reasons, villages differ in the num- villages in Sierra Leone that have limited competition, ber of such families, ranging from 1 to 12. Villages with in Maharashtran villages with a dominant landlord fewer ruling families and thus less political competition group, governance is worse: in particular, there is a have systematically worse government (in particular, 75–100 percent decline in the availability of pro-poor less secure property rights in land) and systematically national programs. However, surveys indicate that worse development outcomes (such as lower rates of the low-caste individuals tend to view the situation child health, of nonagricultural employment, and of as satisfactory. Low-caste individuals are 14 percent educational attainment), Acemoglu, Reed, and Robin- more likely to say that they trust the large landholders son (2013) find. Standard economic models would pre- in villages in which the government is dominated by dict that individuals in the villages with fewer ruling the landlord class than in villages in which the land- families (and thus poorer development) would have lord class is not dominant. It seems plausible that in less respect for the chiefs’ authority and be less satisfied the traditional oligarchic villages, individuals expect with their government than individuals in the villages little, get what they expect, and so count themselves with many families (and thus generally better gover- not unfairly treated. “Legitimacy follows power,” as nance). But this is not the case. Ribot (2009, 125) notes, but for a reason that may have Acemoglu, Reed, and Robinson (2013) asked respon- more to do with ways of seeing the world than with the dents to agree with one, both, or neither of the fol- quality of services provided. lowing statements: (1) As citizens, we should be more A recent experiment in India sheds more light on active in questioning the actions of leaders; and (2) In the power of institutions to shape perceptions. The our country these days, we should have more respect experiment investigated whether the social interac- for authority. tions in a community can lead community members Chieftaincies with fewer ruling families, which to overlook or fail to seize opportunities for mutually are correlated with worse governance, reported higher beneficial cooperation. To investigate the question, the respect for authority, as measured by higher agree- authors invented a simple variant of the public goods ment with the second statement. One explanation, game and gave anonymous players the opportunity which is emphasized by the authors, is that individuals to vote on a rule to require a minimum contribution have entered into a patron-client relationship, which to the public good (Hoff, Somanathan, and Strimling gives them an interest in perpetuating the traditional 2014). The higher the required contribution (up to the maximum feasible level), the higher each individual’s payoff would be. Villages varied in how high they Most mental models emerge in a society raised the required contribution. The less socially inclusive villages and those with lower trust adopted a through shared experiences, and they can lower rule for contributions. To get independent information on village charac- be passed down across generations. They teristics that might affect cooperation, the authors con- ducted a survey in the villages where the experimental can persist even if they are dysfunctional. subjects lived. The respondents were not related to the subjects of the experiment. Nonetheless, village coop- eration and trust, as measured by the surveys, predict authority. But a study of clientelism in the Indian state cooperation in the experiment. The villages that were of Maharashtra suggests that a mental model may also more inclusive (for example, those in which different play a role in enhancing the legitimacy of the oligarchic social groups collectively celebrated festivals together) institutions (Anderson, Francois, and Kotwal 2014). and in which trust and perceptions of security were In Maharashtra, all villages have democratic rules, higher (for instance, individuals reported that they but the villages differ, by historical accident, in how could leave their bikes unlocked), as well as the villages THINKING WITH MENTAL MODELS 67 in which respondents felt that they had benefited from Figure 3.2 Making criminal identity more government programs, were the villages that raised the salient increases dishonesty in prison public good rule the most. These findings are consis- inmates tent with those of Guiso, Sapienza, and Zingales (2013): history shapes the ability to recognize the possibility Prisoners were asked, in private, to report the outcome of flipping 10 coins and were promised that they could keep for improving outcomes through collective action. those coins for which they reported “heads.” Panel a shows Mental models may create beliefs that limit indi- the distribution of heads in 10 coin tosses that prisoners viduals’ ability to adapt to new opportunities. The reported when their criminal identity was not made salient. Panel b shows the distribution of heads that prisoners culture of honor in the U.S. South is associated with reported when their prisoner identity was made salient. In the sentiment, “If you wrong me, I’ll punish you.” This both figures, the blue curve depicts the distribution of heads cultural response to conflict may have been adaptive in that would be expected if everyone honestly reported the outcomes of his coin tosses. In both figures, the reported an environment without centralized means of protect- distribution is skewed toward a greater number of heads than ing property, so that a man’s willingness to punish the honest behavior predicts. But the distribution of the criminal merest slight was important in deterring theft. Experi- identity treatment is shifted toward more heads (resulting ments show that insulting men in these cultures causes in higher payoffs) than the distribution for the group whose criminal identity was not made salient. a jump in the levels of cortisol (a stress hormone) and testosterone (an aggression hormone) and triggers an a. Criminal identity not made salient urge to punish, whereas it provokes no such responses 25 among men from communities without a culture of honor (Cohen and others 1996). Perhaps not surpris- ingly, a culture of honor impairs the ability of individ- 20 uals to build cooperative conventions: when mistakes Percentage in coordination occur, the hurt party tends to withdraw 15 cooperation, which can lead to less coordination, more misunderstanding, and an unraveling of cooperative 10 behavior (Brooks, Hoff, and Pandey 2014). The effects of making an 5 identity salient An individual’s self-concept consists of multiple iden- 0 0 1 2 3 4 5 6 7 8 9 10 tities (that is, multiple mental models), each associated with different norms that guide behavior (Turner 1985). Number of coin tosses with “heads” One way to test the influence of mental models on behavior is to experimentally manipulate the salience b. Criminal identity made salient of a mental model. A study in a maximum security prison in Zurich, 25 Switzerland, found that increasing the salience of an individual’s criminal identity increased his dishonesty 20 (Cohn, Maréchal, and Noll 2013). The prisoners were asked, in private, to report the outcomes of flipping 10 Percentage 15 coins and were promised that they could keep those coins for which they reported “heads,” while they had 10 to return the coins for which they reported “tails.” Before flipping the coins, the prisoners filled out a survey. Randomly, prisoners received a survey that 5 either included questions about their prisoner identity or that did not. In the group that had been primed to 0 think about themselves as prisoners, more subjects 0 1 2 3 4 5 6 7 8 9 10 cheated (figure 3.2, panel b). Since the decision- Number of coin tosses with “heads” making situation was exactly the same and only the Expected distribution under honest reporting prior set of thoughts triggered by the questionnaire Control Criminal identity had changed, it is clear that context changed the way the subjects evaluated their choices. Cheating was 60 Source: Cohn, Maréchal, and Noll 2013. 68 WORLD DEVELOPMENT REPORT 2015 percent higher in the criminal identity group than in specific to individuals who have a criminal identity. No the control group. similar effect occurred when participants were drawn A follow-up experiment sheds light on why. The from the general population. experiment again randomly assigned the “criminal Research on performance and identity shows that identity” survey to one group of prisoners and the context also influences the ability and motivation control survey to the other group. After completing to learn and to expend effort, with potentially large the survey, prisoners were asked to complete a word effects on human capital formation. Context seems to whose initial letters were, for example, “pol.” Those in trigger beliefs about what one is capable of and what the criminal identity group were almost twice as likely one should achieve. An experiment on caste in India to complete “pol” with a crime-related word, such as finds that context can have an important impact on “police,” than to choose words not related to crime, performance. such as “politics.” The results of the word-completion The experiment assessed the effect of manipu- task show that the criminal identity survey increased lating the salience of caste on children’s intellectual the mental accessibility of crime–related thoughts. performance in classrooms (Hoff and Pandey 2006, The results support the interpretation that individuals 2014).9 The control condition, in which caste identity in the Zurich coin-flipping game were made more dis- was not revealed, demonstrated that low-caste boys honest by the context that made their criminal identity solve mazes just as well as high-caste boys. However, more salient in their minds. Significantly, the effect is publicly revealing caste in mixed-caste groups reduced the performance of the low-caste boys, which (control- ling for individual characteristics) created a 23 percent Figure 3.3 Cuing a stigmatized or entitled caste gap in total mazes solved in favor of the high identity can affect students’ performance castes (figure 3.3). Here, a context that primed a social identity (by revealing caste) affected performance. High-caste and low-caste boys from villages in India were When caste was revealed in segregated groups, the randomly assigned to groups that varied the salience of caste identity. When their caste was not revealed, high-caste and high-caste boys also underperformed. Why might this low-caste boys were statistically indistinguishable in solving be? If segregation, a mark of the privileges of the high mazes. Revealing caste in mixed classrooms decreased the caste, evokes a sense of entitlement in which the high- performance of low-caste boys. But publicly revealing caste in caste-segregated classrooms—a marker of high-caste caste boys feel less need to achieve, then the effect of entitlement—depressed the performance of both high-caste making caste salient may be to cause the high-caste and low-caste boys, and again their performance was students to feel, “Why try?” Meanwhile, the low-caste statistically indistinguishable. may feel, “I can’t, or don’t dare to, excel.”10 7 The staying power of mental models 6 The power and persistence of mental models are strik- ingly captured by a story Nelson Mandela told of a Number of mazes solved 5 time when he flew from Sudan to Ethiopia (Mandela 1995, 292). He started to worry when he noticed that the pilot was black: 4 We put down briefly in Khartoum, where we 3 changed to an Ethiopian Airways flight to Addis. Here I experienced a rather strange sensation. As I was boarding the plane I saw that the pilot 2 was black. I had never seen a black pilot before, and the instant I did I had to quell my panic. How 1 could a black man fly an airplane? Mental models may outlive their usefulness or, 0 indeed, may persist when they were never useful to Caste not revealed Caste revealed, Caste revealed, mixed-caste groups segregated groups begin with. We have a hard time abandoning mental models that are not serving us well. The Atlantic slave High caste Low caste trade, and the immense destruction of local institutions Source: Hoff and Pandey 2014. that it caused, ended over 100 years ago. Few people are THINKING WITH MENTAL MODELS 69 now exposed to the risk of being sold into slavery. Why Belief traps don’t people change their mental model and become The beliefs that people hold today may lead to choices more trusting, now that the danger has passed? that foreclose putting the belief to a test tomorrow. In standard economic models, many of the mis- It is easy to see how this can happen in the case of matches described here between mental models and trust. If people believe that trusting strangers or reality would not persist. Individuals make inferences putting money in a bank is risky, they will be reluc- from their experiences about what works and what tant to use financial intermediaries. If they did, they does not. If beliefs are inconsistent with outcomes, might discover that the trust they placed in them was then the decision maker in the standard economic warranted, and after enough good experiences, they model (see figure 1.2, panel a, in chapter 1) would revise would revise their belief. But the potential cost (as they his beliefs as he observes new information. In contrast, see it) of testing the belief may be too high. People what a “psychological, social, and cultural actor” sees who live in high-trust areas of Italy, for example, use and the inferences he draws from it are themselves banking services, such as checks, savings accounts, affected by his mental models. The result is that we financial instruments, and formal credit, rather than can live in a world that Hoff and Stiglitz (2010) describe cash (Guiso, Sapienza, and Zingales 2013). People in as an “equilibrium fiction.” A belief in a race-based or low-trust areas use banking services less—which could gender-based hierarchy can affect self-confidence in deprive them of the opportunity to update their beliefs ways that create productivity differences that sustain when, and if, banking services became secure. the beliefs, although no underlying differences exist. When the stakes are even higher, it is asking a great Four broad factors, discussed next, can explain the deal of people to challenge their mental model. For staying power of mental models that are not serving example, the practice of female genital cutting is sup- individuals well. ported by many social norms and beliefs, including, in some groups, a belief that it increases fertility or that Attention and perception contact with a woman’s genitals in their natural state Mental models influence what individuals per- can be harmful or even fatal (Mackie 1996; WHO 1999; ceive, pay attention to, and recall from memory. If, Gollaher 2000). When people hold such beliefs as these, for instance, people continually perceive leaders as it takes a brave soul to test them. In some countries, untrustworthy, the perceptions will reinforce the men- mothers bring their ill newborns to visit tooth extrac- tal model that they must be on guard against betrayal. tors who dig out the undeveloped baby tooth with a Without realizing it, people tend to fill in information sharp metal stick in order to avoid contamination from gaps based on default assumptions consistent with “false teeth,” which are believed to be dangerous or their mental models. Individuals may even reinterpret even deadly (Borrell 2012). It may be asking a great deal information about a person or object that seems incon- from a parent to test this belief. sistent with the category to which the person or object is assigned so that the information fits the category Ideology and confirmation bias (Baldwin 1992). Beliefs can lead people to ignore, suppress, or forget observations that would tend to undermine their The need to test some types of beliefs at beliefs. Confirmation bias, discussed in chapter 1, is the level of the society the tendency to search for and use information that Some beliefs—for example, that women cannot be supports one’s beliefs. If the bias is sufficiently strong, capable political leaders—cannot be tested at the indi- it is possible that the false hypothesis will never be vidual level. There must be a critical mass in society discarded, no matter how much evidence exists that that tests the belief together by experiencing the alter- favors the alternative hypothesis (Rabin and Schrag native world. Thus a test requires an event that leads 1999). The absence of a concept of a particular phenom- many people to question old beliefs. enon can leave individuals unable to discern actual For example, one of the factors that support trust patterns, or able to discern them only incompletely. in society is a norm against taking advantage of other For instance, a woman who suffers sexual harassment people’s trust. Beliefs about whether others are trust- before the concept exists in her society cannot prop- worthy influence whether parents teach children to erly comprehend her experience or communicate it trust others. As a result, mental models about whether intelligibly to others; the problem “stems from a gap others are trustworthy or not are transmitted across in shared tools of social interpretation” (Fricker 2007, generations, and initial beliefs are reinforced and 6). Chapter 10 presents experimental evidence that therefore never tested at a wide scale (Frank 1997; individuals, including development professionals, Tabellini 2008). who are otherwise quite capable of solving numerical 70 WORLD DEVELOPMENT REPORT 2015 problems, struggle to interpret data correctly when tions for their teenage daughters, raised the daughters’ the implications of the data conflict with their mental aspirations for themselves, and led to a slight narrow- models. The general point is that just as mental mod- ing of the gender gap in schooling (Beaman and others els are tools for constructing mental representations, 2009, 2012). The evidence suggests that the change in inappropriate mental models limit our ability to per- mental models caused the changes in behavior. ceive and communicate a pattern accurately. However, it is only under certain circumstances that The lack of an appropriate mental model can also changes in interactions—created by political affirma- impede learning. In experiments in Malawi, women tive action or other policies—lead to a positive change farmers who are trying to communicate new agricul- in attitudes. If negative stereotypes shape perceptions tural techniques are consistently rated as less knowl- strongly enough, interaction may simply reinforce edgeable than men, even though they have the same the negative stereotypes, undermining the hoped-for knowledge base (BenYishay and Mobarak 2014). In an effects of the policy. A study of political affirmation experiment in India, women who are in a dispute with for low-caste individuals in village government in men have less power of persuasion than their male India finds evidence that it increased absenteeism by counterparts, even when they are judged equally artic- high-caste teachers and lowered outcomes in primary ulate, wise, and otherwise credible (Hoff, Rahman, and schools, which were under the jurisdiction of the local Rao 2014). The findings provide an example of what village government (Pandey 2010). High-caste teachers Fricker (2007) calls “epistemic injustice”—discrimina- essentially boycotted the attempted change in the sta- tion against an individual as a source of knowledge tus of low-caste individuals. because of his or her social identity. The historical distribution of power and prestige between groups Changing mental models through the media affects perceptions of their credibility and thus can Exposure to fiction, such as a serial drama, can perpetuate these inequities. change mental models (see spotlight 2 on entertain- ment education). For example, when people who have Policies to improve the match of not been exposed to societies with low fertility were mental models with a decision exposed to engaging soap operas about families with context few children, fertility rates declined (Jensen and Oster The close relationship between mental models and 2009; La Ferrara, Chong, and Duryea 2012). The impact institutions such as caste and gender roles makes the of exposure to long-running serial dramas in Brazil, process of changing mental models difficult. Some pol- which deliberately crafted soap operas with small icies, such as self-help group programs, try to change families to bring about social change, was large and institutions and mental models at the same time: they immediate. The fertility decline across municipalities try to decrease economic dependence or other forms of in Brazil began after the first year the municipality dependence and to expand ways of understanding the had access to TV soap operas. The decline was espe- world. Other policies try to change only institutions, cially great for respondents close in age to the leading with the hope that the intervention will change mental female character in at least one of the soap operas models as an indirect effect. Still other policies target aired in a given year, which is consistent with a role mental models alone. This section considers the latter model effect. The decline was comparable to the effect two types of policy. of two additional years in women’s education. For women ages 35–44, the decrease was 11 percent of Changing institutions mean fertility. An example of the potential for policy that changes institutions to change mental models comes from a Changing mental models through education program of political affirmative action for women in methods and early childhood interventions West Bengal, India. The policy led some villages to A promising arena in which policy can affect mental have female leaders for the first time. Just seven years’ models is early education. Primary school is a for- exposure to women leaders reduced men’s bias in eval- mative experience for children. The experience can uating women in leadership positions. The men still shape the mental models that the individuals possess preferred male leaders to female leaders. However, in as adults. There is some evidence that “horizontal evaluating the performance of a given leader, gender teaching systems,” in which children interact with one was no longer a strong source of bias. Seven years’ another and engage in class discussions, are an impor- exposure to women leaders also raised parents’ aspira- tant learning tool and increase their level of trust. This THINKING WITH MENTAL MODELS 71 Insight into the long-term effects of interventions Figure 3.4 Changing disruptive children’s that aim to increase trust among children comes from mental models related to trust improved an experiment in Montreal that fostered self-control adult outcomes and social skills in disadvantaged and disruptive A role-playing intervention for disruptive boys ages 7–9 schoolchildren through a series of role-playing exer- reduced the gap in the rate of secondary school completion cises with more cooperative children (Algan, Beasley, between the disruptive and the nondisruptive boys by more than half. and others 2013). This program was targeted toward the most disruptive boys in kindergarten: those who a. Rate of obtaining a secondary were the most aggressive and had the most problems school diploma with self-control. The disruptive boys were randomly assigned to a treatment group or to a control group 60 whose members received no special help. Data were Percentage who obtained a secondary school diploma collected over 20 years on these two groups, as well as on a representative group of boys who were not dis- 40 ruptive as kindergarteners. As adolescents, the boys in the treatment group were found to be more trusting and to have more self-control. They also had sub- 20 stantially improved adult outcomes, including much 0 Policy interventions may be able Disruptive as a Disruptive, with Nondisruptive kindergartener, no intervention social skills and role-playing to expose people to alternative intervention experiences, ways of thinking, and role b. Levels of trust at ages 10–13 models that expand mental models Average level of trust at ages 10–13 0.60 and thereby broaden prototypes (such 0.40 as a woman leader), improve trust, encourage collective action, and increase 0.20 investments. 0 higher rates of completing secondary school, shown Disruptive as a Disruptive, with Nondisruptive in figure 3.4, panel a. Increased trust (not only self- kindergartener, social skills and no intervention role-playing control) seems to have been a factor in the improve- intervention ments in adult outcomes. Figure 3.4, panel b, shows the level of trust—calculated as an average of many Source: Algan, Beasley, and others 2013. questions that ask about levels of trust—in the control, Note: Trust is the average of the normalized responses to several different questions about trust, setting the control group mean to zero. treatment, and nondisruptive groups at ages 10–13, where the variable is normalized with the control group mean equal to zero. The gap between the non- body of evidence suggests new policy options. A shift in disruptive group and the disruptive group without teaching strategies—incorporating more group work the treatment is 0.29 standard deviations. Treatment projects, especially in education systems that have reduced that gap by about 50 percent. The evidence traditionally relied heavily on rote learning and mem- shows that higher levels of trust explain the narrowing orization—may be a promising avenue for improving of the achievement gap between the initially disrup- social capital (Algan, Cahuc, and Shleifer 2013). tive children and those who were nondisruptive. 72 WORLD DEVELOPMENT REPORT 2015 Conclusion 3. This effect of mental models gives rise to the availabil- People use mental models to make sense of the world ity heuristic. The heuristic entails basing judgments on around them. Most mental models emerge in a society information and scenarios that immediately come to through shared experiences, and they can be passed mind, rather than on using all information appropri- down across generations. They can persist even if ately. To take a simple example, most English-speaking people, when asked, will say that seven-letter words they have negative consequences for individuals and ending in ing are more common than seven-letter communities. Mental models influence both the legit- words whose sixth letter is n, even though the former imacy of predatory governance institutions and the category is a subset of the latter. The explanation for prospects for jointly beneficial collective action. Policy this common mistake is that English-speakers have a interventions may be able to expose people to experi- familiar category for ing words. ences that change their mental models. 4. This is called the prototype heuristic. It explains why, for Much evidence demonstrates the role of mental example, the median estimate of the annual number models in domains important for development. This of murders in the city of Detroit, Michigan, is twice as chapter discussed only a few. Later chapters will dis- high as the estimate of the number of murders in the cuss others, including mental models of child devel- state of Michigan (Kahneman and Frederick 2002). Detroit has a reputation for violence; Michigan as a opment (chapter 5), “mental accounting” for money whole does not. matters (chapter 6), mental models of productivity and 5. Over the years, there have been hundreds of defi- technology (chapter 7), mental models of health (chap- nitions of the term “culture.” Many economists use ter 8), and mental models of climate change (chapter 9). the term to mean individual values that are broadly Historians attribute the rise of the modern world shared within a group, but many anthropologists and to a change in the mental model of how the universe sociologists today do not accept the premise—which works. Shifting to believing we live under universal underlies that usage—that there are broadly shared, physical laws rather than divine caprice made it pos- uncontested values within a society. Instead, the sible for individuals to move from handicrafts to mass prevailing definition of culture in anthropology and production technologies (see, for instance, Mokyr sociology is the collection of mental models (or “sche- 2013). The Enlightenment represented a changed mas”) that are maintained and nurtured by rules and norms, actions, and rituals (Swidler 1986; DiMaggio mental model that gave rise to sweeping changes in 1997). Drawing on work in anthropology, sociology, economic structures, which in turn gave rise to mas- and cognitive science, many social scientists have sive changes in social patterns that helped create the gravitated toward a cognitive approach in which cul- modern world. ture is composed of mental tools (ways of interpreting Economic and political forces influence mental the world) instead of, or in addition to, values (ends models, but mental models can have an independent of action). See, for example, Rao and Walton (2004); influence on development by shaping attention, North (2005); Greif (2006); Rao (2008); Nunn (2012); perception, interpretation, and the associations that Gauri, Woolcock, and Desai (2013); and Mokyr (2013). automatically come to mind. This chapter has illus- 6. Nunn and Wantchekon (2011); Fernández and Fogli trated how a focus on mental models both gives policy (2009); Algan and Cahuc (2010). 7. See the vast literature that emerged from the work of makers new tools for promoting development and pro- Knack and Keefer (1997), reviewed in Algan and Cahuc vides new understandings for why policies based on (2013). standard economic assumptions can fail. 8. See, for example, Douglas (1986) and Fourcade (2011). 9. A feature of the caste system that makes it well suited Notes to identifying the effect of culture and identity on 1. In using the term mental models, the Report follows learning is that the meaning of the caste categories is the usage by Arthur Denzau and Douglass North not in doubt. The caste system is a closed order: status (1994) and Elinor Ostrom (2005). Psychology, sociology, is fixed by birth. High-caste individuals are consid- anthropology, and political science use related con- ered socially and intellectually superior in all respects cepts, including schemas or cognitive frames (Markus to low-caste individuals. Individuals at the bottom 1977; DiMaggio 1997). of the caste order were once called “untouchables.” 2. A simple example of a category—a mental model— Perhaps the most important fact about caste has been that has received a great deal of attention in behav- its emphasis on social segregation: “Segregation of ioral economics is a mental account. Economists use individual castes or groups of castes in the village the term to describe how individuals may bracket a was the most obvious mark of civic privileges and decision, using some information and discarding disabilities” (Jodhka 2002, 1815). Today, untouchability other information that affects the payoffs to the deci- is illegal, and evidence of a new social order is visible sion (Thaler 1999). to every schoolchild in the measures to encourage THINKING WITH MENTAL MODELS 73 school enrollment and in the broad participation of National Bureau of Economic Research, Cambridge, low-caste individuals in the political process. Yet chil- MA. dren are still likely to encounter the traditional order Berger, Peter L., and Thomas Luckmann. 1966. The Social of caste, segregation, and untouchability in their own Construction of Reality: A Treatise in the Sociology of Knowl- experiences, through the fables they learn and in edge. Garden City, NY: Anchor Books. the continued discrimination, insults, and atrocities Bernard, Tanguy, Stefan Dercon, Kate Orkin, and Alema- against upwardly mobile members of low castes. yehu Seyoum Taffesse. 2014. “The Future in Mind: 10. A seminal work on the ability of cues to an identity Aspirations and Forward-Looking Behaviour in Rural stereotyped as intellectually inferior to impair cog- Ethiopia.” Working Paper WPS/2014–16, Centre for the nitive performance is Steele and Aronson (1995). The Study of African Economies, Oxford, U.K. argument in this chapter is that identity cues may Bernard, Tanguy, Stefan Dercon, and Alemayehu Seyoum affect preferences to expend effort, as well as the abil- Taffesse. 2011. “Beyond Fatalism: An Empirical Explo- ity to perform, by triggering a sense of entitlement or ration of Self-Efficacy and Aspirations Failure in Ethio- a particular social role. pia.” IFPRI Discussion Paper 01101, International Food Policy Research Institute, Washington, DC. Bloom, Nicholas, Raffaella Sadun, and John Van Reenen. References 2012. “The Organization of Firms across Countries.” Acemoglu, Daron, Tristan Reed, and James A. Robinson. Quarterly Journal of Economics 127 (4): 1663–1705. doi: 2013. “Chiefs: Economic Development and Elite Con- 10.1093/qje/qje029. trol of Civil Society in Sierra Leone.” Working Paper Borrell, Brendan. 2012. “Magic or Medicine? If You Were 18691, National Bureau of Economic Research, Cam- a Poor Ugandan Mother with a Desperately Ill Baby, bridge, MA. Would You Turn to Western Medicine or the Village Acemoglu, Daron, and James A. Robinson. 2012. Why Healer?” Aeon. http://aeon.co/magazine/philosophy Nations Fail: The Origins of Power, Prosperity, and Poverty. /african-healers-and-western-medicine/. New York: Crown. Bregman, Elsie O. 1934. “An Attempt to Modify the Alesina, Alberto, Paola Giuliano, and Nathan Nunn. 2011. Emotional Attitudes of Infants by the Conditioned “Fertility and the Plough.” American Economic Review 101 Response Technique.” Pedagogical Seminary and Journal (3): 499–503. of Genetic Psychology 45 (1): 169–98. —— ——. 2013. “On the Origins of Gender Roles: Women Brooks, Benjamin, Karla Hoff, and Pryanka Pandey. and the Plough.” Quarterly Journal of Economics 128 (2): 2014. “Insult versus Accident: Caste Culture and the 469–530. Efficiency of Coordination.” Paper presented at the Algan, Yann, Elizabeth Beasley, Frank Vitaro, and Richard Latin American and Caribbean Economic Association Tremblay. 2013. “The Long-Term Impact of Social Skills Conference. Training at School Entry: A Randomized Controlled Cohen, Dov, Richard E. Nisbett, Brian F. Bowdle, and Nor- Trial.” Working Paper, Sciences Po, Paris. bert Schwarz. 1996. “Insult, Aggression, and the South- Algan, Yann, and Pierre Cahuc. 2010. “Inherited Trust and ern Culture of Honor: An ‘Experimental Ethnography’.” Growth.” American Economic Review 100 (5): 2060–92. Journal of Personality and Social Psychology 70 (5): 945–60. —— ——. 2013. “Trust and Growth.” Annual Review of Econom- doi: 10.1037/0022-3514.70.5. ics 5 (1): 521–49. Cohn, Alain, Michel André Maréchal, and Thomas Noll. Algan, Yann, Pierre Cahuc, and Andrei Shleifer. 2013. 2013. “Bad Boys: The Effect of Criminal Identity on “Teaching Practices and Social Capital.” American Eco- Dishonesty.” Working Paper 132, University of Zurich, nomic Journal: Applied Economics 5 (3): 189–210. Zurich. Anderson, Siwan, Patrick Francois, and Ashok Kotwal. Denzau, Arthur T., and Douglass C. North. 1994. “Shared 2014. “Clientelism in Indian Villages.” http://econ.arts Mental Models: Ideologies and Institutions.” Kyklos 47 .ubc.ca/fpatrick/documents/clientAERmay1313.pdf. (1): 3–31. Baldwin, Mark W. 1992. “Relational Schemas and the Pro- DiMaggio, Paul. 1997. “Culture and Cognition.” Annual cessing of Social Information.” Psychological Bulletin 112 Review of Sociology 23 (1): 263–87. (3): 461–84. Douglas, Mary. 1986. How Institutions Think. Syracuse, NY: Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, Syracuse University Press. Rohini Pande, and Petia Topalova. 2009. “Powerful Fernández, Raquel, and Alessandra Fogli. 2009. “Culture: Women: Does Exposure Reduce Bias?” Quarterly Journal An Empirical Investigation of Beliefs, Work, and Fer- of Economics 124 (4): 1497–1540. tility.” American Economic Journal: Macroeconomics 1 (1): Beaman, Lori, Esther Duflo, Rohini Pande, and Petia Topa- 146–77. lova. 2012. “Female Leadership Raises Aspirations and Fourcade, Marion. 2011. “Cents and Sensibility: Economic Educational Attainment for Girls: A Policy Experiment Valuation and the Nature of ‘Nature.’ ” American Journal in India.” Science 335 (6068): 582–86. of Sociology 116 (6): 1721–77. BenYishay, Ariel, and A. Mushfiq Mobarak. 2014. “Social Frank, Robert H. 1997. “The Frame of Reference as a Public Learning and Communication.” Working Paper 20139, Good.” The Economic Journal 107 (445): 1832–47. 74 WORLD DEVELOPMENT REPORT 2015 Fricker, Miranda. 2007. Epistemic Injustice: Power and the Eth- La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. ics of Knowing. Oxford, U.K.: Oxford University Press. 2012. “Soap Operas and Fertility: Evidence from Brazil.” Gauri, Varun, Michael Woolcock, and Deval Desai. 2013. American Economic Journal: Applied Economics 4 (4): 1–31. “Intersubjective Meaning and Collective Action in Mackie, Gerry. 1996. “Ending Footbinding and Infibu- Developing Societies: Theory, Evidence, and Policy lation: A Convention Account.” American Sociological Implications.” Journal of Development Studies 49 (1): Review 61 (6): 999–1017. 160–72. Mandela, Nelson. 1995. Long Walk to Freedom: The Autobiog- Gollaher, David L. 2000. Circumcision: A History of the World’s raphy of Nelson Mandela. New York: Little, Brown. Most Controversial Surgery. New York: Basic Books. Markus, Hazel. 1977. “Self-Schemata and Processing Infor- Greif, Avner. 2006. Institutions and the Path to the Modern mation about the Self.” Journal of Personality and Social Economy: Lessons from Medieval Trade. New York: Cam- Psychology 35 (2): 63–78. bridge University Press. Mokyr, Joel. 2013. “Cultural Entrepreneurs and the Origins Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2013. of Modern Economic Growth.” Scandinavian Economic “Long-Term Persistence.” Working Paper 23/13, Ein- History Review 61 (1): 1–33. audi Institute for Economics and Finance, Rome. North, Douglass C. 1990. Institutions, Institutional Change Haidt, Jonathan. 2012. The Righteous Mind: Why Good People and Economic Performance. Cambridge, U.K.: Cambridge Are Divided by Politics and Religion. New York: Random University Press. House. —— ——. 2005. Understanding the Process of Economic Change. Henrich, Joseph, Robert Boyd, Samuel Bowles, Colin Princeton, NJ: Princeton University Press. Nunn, Nathan. 2008. “The Long-Term Effects of Africa’s Camerer, Ernst Fehr, Herbert Gintis, and Richard Slave Trades.” Quarterly Journal of Economics 123 (1): McElreath. 2001. “In Search of Homo Economicus: 139–76. Behavioral Experiments in 15 Small-Scale Societies.” —— ——. 2012. “Culture and the Historical Process.” Eco- American Economic Review 91 (2): 73–78. nomic History of Developing Regions 27 (Supplement 1): Hoff, Karla, and Priyanka Pandey. 2006. “Discrimination, 108–26. Social Identity, and Durable Inequalities.” American Nunn, Nathan, and Leonard Wantchekon. 2011. “The Slave Economic Review 96 (2): 206–11. Trade and the Origins of Mistrust in Africa.” American — —— —. 2014. “Making Up People: The Effect of Identity Economic Review 101 (7): 3221–52. on Performance in a Modernizing Society.” Journal of Ostrom, Elinor. 2005. Understanding Institutional Diversity. Development Economics 106: 118–31. Princeton, NJ: Princeton University Press. Hoff, Karla, Tauhidur Rahman, and Vijayendra Rao. 2014. Pandey, Priyanka. 2010. “Service Delivery and Corruption “Epistemic Discrimination.” Unpublished manuscript. in Public Services: How Does History Matter?” Amer- Hoff, Karla, Rohini Somanathan, and Pontus Strimling. ican Economic Journal: Applied Economics 2 (3): 190–204. 2014. “Community Contracts: An Experimental Rabin, Matthew, and Joel L. Schrag. 1999. “First Impres- Investigation of Rule Formation in Indian Villages.” sions Matter: A Model of Confirmatory Bias.” Quarterly Working Paper, University of Virginia, Charlottesville. Journal of Economics 114 (1): 37–82. Hoff, Karla, and Joseph Stiglitz. 2010. “Equilibrium Fic- Rao, Vijayendra. 2008. “Symbolic Public Goods and the tions: A Cognitive Approach to Societal Rigidity.” Amer- Coordination of Collective Action: A Comparison of ican Economic Review 100: 141–46. Local Development in India and Indonesia.” In Con- Jensen, Robert, and Emily Oster. 2009. “The Power of TV: tested Commons: Conversations between Economists and Cable Television and Women’s Status in India.” Quar- Anthropologists, edited by Pranab Bardhan and Isha Ray, terly Journal of Economics 124 (3): 1057–94. chap. 10. Hoboken, NJ: Wiley-Blackwell. Jodhka, Surinder S. 2002. “Caste and Untouchability in Rao, Vijayendra, and Michael Walton. 2004. “Conclusion.” Rural Punjab.” Economic and Political Weekly 37 (19): In Culture and Public Action: A Cross-Disciplinary Dialogue 1813–23. on Development Policy, edited by Vijayendra Rao and Kahneman, Daniel. 2003. “Maps of Bounded Rationality: Michael Walton, chap. 16. Palo Alto, CA: Stanford Uni- Psychology for Behavioral Economics.” American Eco- versity Press. nomic Review 93 (5): 1449–75. Ribot, J. C. 2009. “Authority over Forests: Empowerment Kahneman, D., and S. Frederick. 2002. “Representative- and Subordination in Senegal’s Democratic Decentral- ness Revisited: Attribute Substitution in Intuitive ization.” Development and Change 40 (1): 105–29. Judgment.” In Heuristics and Biases: The Psychology of Ross, Lee, and Richard E. Nisbett. 1991. The Person and the Intuitive Judgment, edited by T. Gilovich, D. Griffin, and Situation: Perspectives of Social Psychology. New York: D. Kahneman, 49–81. New York: Cambridge University McGraw-Hill. Press. Steele, Claude M., and Joshua Aronson. 1995. “Stereotype Knack, Stephen, and Philip Keefer. 1997. “Does Social Cap- Threat and the Intellectual Test Performance of Afri- ital Have an Economic Payoff? A Cross-Country Inves- can Americans.” Journal of Personality and Social Psychol- tigation.” Quarterly Journal of Economics 112 (4): 1251–88. ogy 69 (5): 797–811. THINKING WITH MENTAL MODELS 75 Swidler, Ann. 1986. “Culture in Action: Symbols and Strate- Turner, John C. 1985. “Social Categorization and the gies.” American Sociological Review 51 (2): 273–86. Self-Concept: A Social Cognitive Theory of Group Tabellini, Guido. 2008. “The Scope of Cooperation: Values Behavior.” In Advances in Group Process: Theory and and Incentives.” Quarterly Journal of Economics 123 (3): Research, Vol. 2, edited by Edward J. Lawler, 77–122. 905–50. Greenwich, CT: JAI Press. Talhelm, T., X. Zhang, S. Oishi, C. Shimin, D. Duan, X. Lan, WHO (World Health Organization). 1999. “Female Genital and S. Kitayama. 2014. “Large-Scale Psychological Dif- Mutilation Programmes to Date: What Works and ferences within China Explained by Rice versus Wheat What Doesn’t.” Geneva: WHO. Agriculture.” Science 344 (6184): 603–08. Zerubavel, Eviatar. 1999. Social Mindscapes: An Invitation to Thaler, Richard H. 1999. “Mental Accounting Matters.” Cognitive Sociology. Cambridge, MA: Harvard Univer- Journal of Behavioral Decision Making 12 (3): 183–206. sity Press. Todd, Peter M., and Gerd Gigerenzer. 2000. “Précis of Simple Heuristics That Make Us Smart.” Behavioral and Brain Sciences 23 (5): 727–41. 76 WORLD DEVELOPMENT REPORT 2015 Entertainment education Can exposure to media provide a source of sustained the story as though they were one of the characters. change and a means of promoting development? The There is evidence that when individuals are absorbed in evidence to date is limited but encouraging. The use a narrative, they become less critical and defensive and of mass media for entertainment education creates an are more open to persuasion (Green and Brock 2000; opportunity to affect not only the mental models of Slater and Rouner 2002). Identification with a specific individual viewers but also the mental models accepted character works in a similar way; it involves a temporary by the wider society that create the context for collective loss of self and adoption of the character’s perspective. action. The links below provide examples of how enter- Because identification is not compatible with counter- tainment education works: arguing, persuasive messages are more easily accepted (Cohen 2001; Moyer-Gusé 2008). Evidence also suggests • Scandal!, a South African soap opera with financial mes- that people find entertainment more enjoyable when sages, including ones related to gambling: https://www they can be transported beyond themselves and identify .youtube.com/watch?v=ys5eSxTetF4&noredirect=1 with the situation of the character (de Wied, Zillmann, • 16 and Pregnant, a U.S. reality TV show on teen and Ordman 1994; Hall and Bracken 2011). pregnancy: http://www.mtv.com/shows/16_and_ A media program’s ability to persuade may also be pregnant/ affected by the social context in which it is consumed. • Shuga, an African drama on HIV/AIDS and gender Aspirational videos shown in rural Ethiopia were more violence: http://www.youtube.com/watch?v=pI8_P effective when more people in the community were Spotlight 2 _h89R8 exposed to the content (Bernard and others 2014). In the United States, teens who watched, with a parent or other The theory behind entertainment trusted adult, a comedy that included information on education contraception reported greater gains in knowledge (Col- The term entertainment education (EE) refers to entertain- lins and others 2003) because the program stimulated a ment media that incorporate an educational message discussion, which provided further information. or information of value to the audience to increase audience members’ knowledge about an issue, create Evidence of impact favorable attitudes, and change overt behavior (Singhal While there are many studies of entertainment educa- and Rogers 2002; Moyer-Gusé 2008). The theoretical tion, only a relatively small number employ rigorous underpinnings for entertainment education are usually quantitative methods, such as randomized controlled traced to the psychologist Albert Bandura, a pioneer in trials (RCTs).2 Among recent studies using RCTs, posi- social cognitive theory (Bandura 1977, 1986). Bandura tive impacts were found from an in-script partnership showed that children who viewed violent images on with a South African soap opera relating to financial television demonstrated more aggressive behavior than attitudes and behaviors, from videos shown in Ethiopia those who viewed neutral content (Bandura, Ross, and to induce future-oriented investments such as educa- Ross 1963). Observing the performance of others, people tion for children, and from a radio drama in Rwanda acquire not only patterns of behavior but also a cognitive that improved perceptions of social norms such as coop- framework about what the behaviors mean, according to eration and willingness to engage in dialogue, even on Bandura’s research. sensitive topics. Other RCTs, however, have not provided Most of the research on entertainment education significant evidence of impact, including a film shown has focused on narratives such as dramas and soap in Nigeria and a comic book in Kenya, both featuring operas. A narrative or story format can help motivate financial messages. change in the audience by showing positive role mod- Some of the most compelling evidence for EE comes els who experience “rewards” and negative role models from studies that use quasi-experimental methods to who are “punished” (Slater and Rouner 2002; Bandura evaluate the impact of entertainment education across 2004). A third type of role model—the transitional char- a society. For example, in Brazil, access to the TV Globo acter—who gradually moves from negative to positive network—which was dominated by soap operas with behaviors during the story—may also be important independent female characters with few, or even no (Sabido 2002).1 Narratives using these constructs from children—has been linked to the country’s rapid drop in EE can help guide audience members through a change fertility. Viewing the soap operas had an effect equal to process, including developing confidence in their own 1.6 years of additional education (La Ferrara, Chong, and abilities (self-efficacy) through association with desir- Duryea 2012). In India, access to cable television reduced able characters, and can facilitate behavior change. fertility and son preference and increased women’s Entertainment education may be especially effective autonomy (Jensen and Oster 2009). A radio program in when people are swept up in a narrative, or experience Tanzania was linked to a significant increase in condom ENTERTAINMENT EDUCATION 77 use and a reduction in the number of sexual partners Bernard, Tanguy, Stefan Dercon, Kate Orkin, and Alema- (Vaughan and others 2000). And in the United States, a yehu Seyoum Taffesse. 2014. “The Future in Mind: reality TV show was linked to a significant drop in teen Aspirations and Forward-Looking Behaviour in Rural pregnancy (Kearney and Levine 2014). Ethiopia.” Centre for the Study of African Economies, University of Oxford, Oxford. Bouman, Martine. 1999. “The Turtle and the Peacock: Col- Business models for entertainment laboration for Prosocial Change: The Entertainment- education Education Strategy on Television.” PhD dissertation, In Latin America, private television channels have been Landbouw Universiteit. producing a number of commercially successful tele- Cohen, Jonathan. 2001. “Defining Identification: A The- novelas with social content since the 1970s. In most other oretical Look at the Identification of Audiences with developing country markets, the main approach to enter- Media Characters.” Mass Communication & Society 4 (3): tainment education has been through the public sector 245–64. or donor-funded productions. Many successful examples Collins, Rebecca L., Marc N. Elliott, Sandra H. Berry, David of entertainment education have been produced in this E. Kanouse, and Sarah B. Hunter. 2003. “Entertainment way: Hum Log, Kalyani, and Taru in India; Meena in South Television as a Healthy Sex Educator: The Impact of Condom-Efficacy Information in an Episode of Asia; and Twende na Wakati in Tanzania, to name a few. Friends.” Pediatrics 112 (5): 1115–21. However, with many markets now saturated with de Wied, M., D. Zillmann, and V. Ordman. 1994. “The Role media, it is more challenging to break through and cre- of Empathic Distress in the  Enjoyment of Cinematic ate impact with a single show. New approaches to enter- Tragedy.” Poetics 23 (1): 91–106.  tainment education focus on partnerships between the Green, Melanie C., and Timothy C. Brock. 2000. “The Role public and private sectors and civil society to increase of Transportation in the Persuasiveness of Public Nar- audience size, overcome the high cost of media produc- ratives.” Journal of Personality and Social Psychology 79 (5): tion, and strengthen social impact (Miller 2011). In some 701. cases, this is happening at the firm level in media com- Hall, Alice E., and Cheryl C. Bracken. 2011. “ ‘I Really Liked Spotlight 2 panies committed to social action, such as Well Told That Movie’: Testing the Relationship between Trait Story in Kenya and Participant Media in the United Empathy, Transportation, Perceived Realism, and States. Organizations that seek to increase the system- Movie Enjoyment.” Journal of Media Psychology: Theories, Methods, and Applications 23 (2): 90. atic use of EE in the commercial entertainment industry Jensen, Robert, and Emily Oster. 2009. “The Power of TV: have also been formed in the United States, Europe, Cable Television and Women’s Status in India.” Quar- and recently in India. These nongovernmental organiza- terly Journal of Economics 124 (3): 1057–94. tions (NGOs) seek to bridge the gap that typically exists Kearney, Melissa S., and Phillip B. Levine. 2014. “Media between content experts in academia or government Influences on Social Outcomes: The Impact of MTV’s and media producers (Bouman 1999) through a variety ‘16 and Pregnant’ on Teen Childbearing.” National of methods, from intensive collaboration on scripts to Bureau of Economic Research, Cambridge, MA. soft-touch approaches such as after-hours “salons.” La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. 2012. “Soap Operas and Fertility: Evidence from Brazil.” Notes American Economic Journal: Applied Economics 4 (4): 1–31. 1. Miguel Sabido is a Mexican playwright and television Miller, Margaret. 2011. “Entertainment Education: A Solu- tion for Cost-Effective High-Impact Development producer who was the first to take Bandura’s social Communication.” World Bank, Washington, DC. learning theory and apply it to mass entertainment Moyer-Gusé, Emily. 2008. “Toward a Theory of Entertain- media in the 1970s and 1980s in Mexico. The resulting ment Persuasion: Explaining the Persuasive Effects telenovelas (Ven Conmigo, Acompañame, and others) were of Entertainment-Education Messages.” Communica- both extremely popular and credited with having an tion Theory 18 (3): 407–25. impact on key social issues such as adult literacy and Nariman, Heidi Noel, and Everett M. Rogers. 1993. Soap family planning (Nariman and Rogers 1993). Operas for Social Change: Toward a Methodology for 2. A World Development Report 2015 background paper, “The Entertainment-Education Television. Westport, CT: Praeger. Impact of Entertainment Education,” provides more Sabido, Miguel. 2002. El tono: Andanzas teóricas, aventuras detailed analysis of the entertainment education litera- prácticas, el entretenimiento con beneficio social. Mexico ture, including evidence of results. City: UNAM. Singhal, Arvind, and Everett M. Rogers. 2002. “A Theoreti- References cal Agenda for Entertainment-Education.” Communica- Bandura, Albert. 1977. “Self-Efficacy: Toward a Unifying tion Theory 12 (2): 117–35. Theory of Behavioral Change.” Psychological Review 84 Slater, Michael D., and Donna Rouner. 2002. “Entertainment- (2): 191. Education and Elaboration Likelihood: Understanding ——— —. 1986. Social Foundations of Thought and Action. Engle- the Processing of Narrative Persuasion.” Communica- wood Cliffs, NJ: Prentice Hall. tion Theory 12 (2): 173–91. ——— —. 2004. “Health Promotion by Social Cognitive Vaughan, Peter W., Everett M. Rogers, Arvind Singhal, Means.” Health Education & Behavior 31 (2): 143–64. and Ramadhan M. Swalehe. 2000. “Entertainment- Bandura, Albert, Dorothea Ross, and Sheila A. Ross. 1963. Education and HIV/AIDS Prevention: A Field Experi- “Imitation of Film-Mediated Aggressive Models.” Jour- ment in Tanzania.” Journal of Health Communication 5 nal of Abnormal and Social Psychology 66 (1): 3. (Supplement): 81–100. PART 2 Psychological and social perspectives on policy 4 CHAPTER Poverty At least once a year, hundreds of millions of parents they might pursue a strategy of persuasion to assist face a decision about school enrollment. Higher- someone like this father. Or if they assume that the income parents are probably choosing which school decision to keep a child out of school results solely their children will attend or which after-school activ- from a political and economic system that is inherently ities to sign them up for. For many parents in low- stacked against poor people, they might advocate quo- income settings, the choice is starker: whether or not tas or a large-scale redistribution of resources. to send their child to school at all. Imagine a poor Both these narratives of poverty offer an incomplete father who chooses not to enroll his son in secondary picture of decision making and choice. The first places school. The assumptions policy makers think underlie little emphasis on constraints beyond the control of this decision will likely affect the remedies they design the decision maker—such as the fees associated with to address low investment in education and other attending school or the absence of enforceable com- behaviors associated with poverty. pulsory education laws, which could coerce parents to send their child to school. The second narrative does not address the cognitive resources required to make Poverty is not simply a shortfall of money. a decision, especially when material resources are in short supply and when people’s willingness to act upon The constant, day-to-day hard choices their desires may be constrained (Mullainathan and Shafir 2013; Perova and Vakis 2013). associated with poverty in effect “tax” If this father lives in rural India, for example, he is most likely making his decision in May, nearly five an individual’s psychological and social months after the harvest—five months after he has earned most of his income for the year. While the resources. This type of “tax” can lead to returns to secondary education might be high and he might have been able to save funds for tuition, a num- economic decisions that perpetuate poverty. ber of other, more immediate concerns might be com- peting for his attention and his resources. He might have run out of kerosene the day before, or he might If policy makers assume that poverty results from need to find materials to patch a hole in his roof. It is poor people’s deviant values or character failings—as one month before the monsoon, so finding clean water did many antipoverty strategies of the United Kingdom requires extra effort. His neighbor might be expecting or the United States until well into the 19th century help with some medical bills, which should not be (Narayan, Pritchett, and Kapoor 2009; Ravallion, forth- ignored since this neighbor helped him pay for medi- coming)—or that poor people simply do not understand cine the year before. Even if a more affluent father feels the benefits of important investments like education, stress about a school enrollment decision, the choice is POVERTY 81 unlikely to trigger concerns about these kinds of basic, formal credit and insurance markets enable people to day-to-day trade-offs. rely less on social networks to weather shocks to their This chapter offers an alternative set of assumptions health or income. for thinking about decision making in contexts of pov- While these considerations may paint an even erty and for analyzing why poor people may engage in bleaker picture of poverty than is familiar to most peo- behaviors that ostensibly perpetuate poverty, such as ple, recent evidence suggests promising interventions borrowing too much and saving too little, underinvest- for reducing the cognitive, psychological, and social ing in health and education, and ignoring programs taxes of poverty. Some of these interventions need not and policies designed to assist them. Recent empirical entail complex interventions for influencing the psy- evidence suggests that these decisions do not arise chology or social environments of poor people. Instead, from deviant values or a culture of poverty particular modifications to the process of delivering products and to poor people. To the contrary: both poor people and services that take the cognitive taxes of poverty into people who are not poor are affected in the same fun- account could make existing interventions more effec- damental way by certain cognitive, psychological, and tive. Recognizing the cognitive and social dimensions social constraints on decision making. However, it is of poverty could also alter estimates of cost-benefit the context of poverty that modifies decision making in ratios of policy instruments, such as cash transfers and important ways. the development of the infrastructure, institutions, In particular, poverty is not simply a shortfall and markets that could serve to lessen the distractions of money. The constant, day-to-day hard choices and cognitive burdens of poverty. associated with poverty in effect tax an individual’s bandwidth, or mental resources. This cognitive tax, in Poverty consumes cognitive turn, can lead to economic decisions that perpetuate resources poverty. First, poverty generates an intense focus on “So if you want to understand the poor, imagine the present to the detriment of the future. When poor yourself with your mind elsewhere. You did not people must direct their mental resources toward deal- sleep much the night before. You find it hard to ing with the concerns of poverty—for example, paying think clearly. Self-control feels like a challenge. off debts or keeping their children safe—they have less You are distracted and easily perturbed. And attention to devote to other important tasks that may this happens every day. On top of the other be cognitively demanding, such as expending greater material challenges poverty brings, it also and more productive effort at work or making timely brings a mental one. . . . Under these conditions, investments in education and health (Mullainathan we all would have (and have!) failed.” and Shafir 2013). Second, poverty can also create poor frames —Mullainathan and Shafir, Scarcity: Why through which people see opportunities. Poverty can Having Too Little Means So Much (2013, 161) blunt the capacity to aspire (Appadurai 2004) and to take advantage of the opportunities that do present “She is worried about the future of her children themselves. and the struggles they have to face once they Third, the environments of people living in poverty grow up. Her immediate concern is to which make additional cognitive demands. The absence of house she should go for a loan of some food certain physical and social infrastructure that eases grains for their food that day.” cognitive burdens in high-income contexts—like —Narayan and others, description of a woman piped water, organized child care, and direct deposit in Pedda Kothapalli, India, in Voices of the and debit of earnings—encumbers those living in low- Poor: Crying Out for Change (2000, 37) income settings with a number of day-to-day decisions that deplete mental resources even further (Banerjee The material deprivation that accompanies poverty and Mullainathan 2008). In settings like the United has been well documented. The poor are more likely States, for instance, parents rarely need to actively to find themselves in situations in which they must weigh the costs and benefits of school attendance forgo meals or live in substandard housing. They may for their children. Birth registration systems and the have many debts to pay off. Their dwellings can be enforcement of truancy laws would counterbalance demolished by rain or expropriated by someone more any internal challenges that might steer parents away powerful. They might have to collect potable water from sending their children to school. Moreover, many times a day. Recent evidence suggests that these 82 WORLD DEVELOPMENT REPORT 2015 situations of scarcity—or a gap between the needs and This situation of scarcity need not apply solely the resources required to fulfill them—create addi- to those currently living below thresholds of $1.25 tional cognitive burdens that interfere with decision or $2.00 per day. It is one that many people in low- making in important ways beyond a person’s mone- income settings may find themselves in at one point tary constraints. In particular, the pressing financial or another, as shown in figure 4.1. Indeed, much of the concerns associated with poverty modify how people “middle class” in low-income countries lives on $2 to $6 allocate their attention and create an intense focus on a day and thus is still likely to face a number of trade- problems of the present to the neglect of others in the offs that can trigger a feeling of scarcity. future (Mullainathan and Shafir 2013). To return to the A real-world example of how situations of scarcity opening example of a father’s decision about investing can deplete mental resources comes from sugar cane in his son’s education, the many current claims on the farmers in India (figure 4.2). These farmers typically father’s attention and resources make the short-term receive their income once a year, at the time of har- costs of investment much more pressing than the vest. Thus just before the harvest (panel a), they may potentially high long-term returns of a secondary edu- feel poor, and just afterward (panel b), they may feel cation that are far off in the future. much more comfortable, having received most of the Figure 4.1 Poverty is a fluid state, not a stable condition In qualitative interviews around the world, community members were asked to rank everyone in the community on an economic ladder at the moment and 10 years earlier. They were also asked to indicate which rungs of the ladder should be equated with poverty. According to these community rankings, poverty is a fluid state rather than a stable characteristic. This finding is consistent with consumption-based estimates of chronic poverty from longitudinal data (Jalan and Ravallion 2000; Pritchett, Suryahadi, and Sumarto 2000; Dercon and Krishnan 2000). 1 Afghanistan 60 2 India (Assam) 3 Bangladesh 50 4 Colombia 40 5 India (Andhra Pradesh) 6 India (Uttar Pradesh) 30 7 India (West Bengal) 8 Indonesia 20 9 Malawi Percentage 10 Mexico 10 11 Morocco 12 Philippines (Bukidnon) 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 13 Philippines 14 Senegal 10 15 Sri Lanka 16 Tanzania 20 17 Thailand 30 18 Uganda 40 50 Proportion of initial poor who moved out of poverty Proportion of nonpoor falling into poverty Source: Narayan, Pritchett, and Kapoor 2009, table 3.2. POVERTY 83 earnings for the season. Indeed right before the har- Figure 4.2 Financial scarcity can vest, they are much more likely to be holding loans (99 consume cognitive resources versus 13 percent) and to have pawned some of their belongings (78 versus 4 percent) (Mani and others Sugar cane farmers in Tamil Nadu, India, receive most of their income once a year during the harvest. Immediately 2013). before receiving their income (panel a), the same farmers That these farmers are poorer before the harvest exhibit higher financial stress and lower cognitive scores, than after perhaps should shock no one. What is less relative to the postharvest period (panel b). This cannot be explained by a change in nutrition, physical exhaustion, obvious, however, is the toll this kind of financial biological stress, or a practice effect on the cognitive test. distress takes on their available cognitive resources right before the harvest. Before receiving their harvest a. Before harvest b. After harvest Percentage of people that pawned belongings income, farmers perform worse on a series of cognitive tests of executive function and fluid intelligence than when they take the same tests after receiving their earnings (for some examples of tasks that test execu- tive function and fluid intelligence, see figure 4.3). This gap cannot be explained by differences in nutrition before or after harvest, physical exhaustion, biolog- ical stress, or familiarity with the testing instrument after the harvest. The difference in scores translates to roughly 10 IQ points, which is approximately equal to three-quarters of a standard deviation and three- quarters of the cognitive deficit associated with losing an entire night of sleep (Mani and others 2013). This cognitive depletion induced by scarcity is not limited to poor farmers in India or to people living under some absolute poverty line. The poverty line of the United States, for example, is nearly seven times the poverty line of low-income countries ($13 versus $2 a day), but financial anxiety among low-income Percentage of people with loans individuals in the United States triggers a very sim- ilar effect. In an experiment in which people had to answer questions about how they would react to some hypothetical scenarios, such as financing an unfore- seen expense or an auto repair, some respondents received financially stressful scenarios (for example, a $2,000 expense), while others received less stress- ful variants (a $200 expense) (Mani and others 2013). As with the farmers, low-income respondents who had to think about a financially stressful situation performed worse on later cognitive tests by an equiv- alent of 13 IQ points, suggesting that simply thinking about the gap between needs and resources captures the mind. This diminishment of executive function might Cognitive test scores account for an intense focus on the present that is IQ IQ + 10 pts. beneficial in some ways but detrimental in others. In a laboratory experiment in the United States, research- ers induced “poverty” and “affluence” among relatively well-off subjects by endowing them with fewer or more items and paid them to perform certain tasks using those items. The experimentally poor tended to use their items more productively, earning more points for each task they attempted (Shah, Mullaina- than, and Shafir 2012). Scarcity focused the mind. Source: Mani and others 2013. 84 WORLD DEVELOPMENT REPORT 2015 started to neglect future rounds and overborrow. Their Figure 4.3 Measuring executive function and fluid overall performance fell, compared to a situation in intelligence which they could not borrow. In contrast, the option a. Executive function to borrow had no impact on the participants assigned State the color of each word. to the “affluent” group. Thus, when placed in a context of scarcity, however brief, otherwise well-off subjects Red Blue Green Purple Blue Red Purple Green exhibited decision-making patterns typically associ- Red Blue Green Purple Blue Red Purple Green ated with poverty. Together, these natural and labora- tory experiments suggest that financial concerns can This task is easier to do for the first set of words. More executive absorb considerable cognitive bandwidth and that sit- function is required to maintain accuracy in the second set of words. uations of scarcity can alter decision making in impor- This is called the Stroop effect. tant ways for both low- and high-income populations. b. Fluid intelligence Poverty creates poor frames Selecting from numbered options 1–8, find the symbol that completes the bottom right section in the box below. “When they assist you, they treat you like a beggar.” —Narayan and others, citing a participant in a discussion group of men and women in Vila Junqueira, Brazil, in Voices of the Poor: Crying Out for Change (2000, 2) Poverty may also generate an internal frame, or a way of interpreting the world and poor people’s role in it. Poor people may feel incompetent and disrespected, without hope that their lives can improve. If these kinds of frames prevent them from taking advantage of economic opportunities, then the poor could also 1 2 3 miss chances to escape poverty because of a deficit of aspirations (Appadurai 2004; Ray 2006; Duflo 2012). Indeed, avoiding the shame that arises from failing to meet social conventions has been described as a core capability (Sen 1983). 4 5 6 Recent empirical evidence suggests an association between poverty and low aspirations. Data from the World Values Surveys, for example, show that lower income—both within and across countries—is asso- 7 8 ciated with a higher tendency to report that life is meaningless, to agree that it is better to live day-to-day because of the uncertainty of the future, and to reject adventure and risk (Haushofer and Fehr 2014). Data from low-income populations in France suggest that This is an example of a Raven’s matrix, a set of puzzles commonly poor students have lower academic and employment used to measure fluid intelligence. (The correct answer is option 2.) aspirations than wealthier students who display the Source: Sample item similar to those found in the Raven’s Progressive Matrices, Standard same degree of academic achievement (Guyon and Progressive Matrices (Standard, Sets A–E). Copyright © 1998, 1976, 1958, 1938 NCS Pearson, Inc. Huillery 2014). Reproduced with permission. All rights reserved. This kind of empirical pattern, however, could Note: “Raven’s Progressive Matrices” is a trademark, in the United States and/or other countries, of Pearson Education, Inc., or its affiliate(s). suffer from problems of reverse causation. Perhaps these character traits are the root cause rather than a function of poverty. This sort of explanation would Then subjects were offered an option to borrow from be inconsistent with the movements in and out of future rounds, which forced them to make trade-offs poverty illustrated in figure 4.1. It is also the case that between the present and the future. This is the point other studies making use of external economic shocks at which the experimentally poor began to suffer. They (which cannot be driven by an individual’s aspiration) POVERTY 85 find a similar association between low income and atti- minutes to describe a personal experience that made tudes toward opportunities. A recent study finds that them feel successful and proud. Compared to other both within the United States and across 37 different groups that described only their daily meal routines or countries, experiencing a recession between the ages watched a funny video, the affirmed group performed of 18 and 25—the impressionable years of early adult- significantly better on cognitive tests of their executive hood—reduces the likelihood that a person believes control and fluid intelligence (Hall, Zhao, and Shafir that “people get ahead by their own hard work” as 2013). In contrast, self-affirmation did not increase the opposed to by “lucky breaks or help from other people” cognitive function of more affluent users of a public (Guiliano and Spilimbergo 2014). library. These results suggest that the intervention A similar change in attitudes arose in Argentina, helped alleviate the distracting stigma of poverty (after only in this instance in the direction of greater self- all, the poor were being tested in a soup kitchen), rather confidence. A land reform in the 1980s transferred than simply improving general feelings of confidence. titles to squatter families in the outskirts of Buenos Aires. The original owners of the land parcels legally contested the government’s expropriations, and many Poverty can blunt the capacity to of these suits were not resolved as of 2007, when a study of attitudes was conducted among the squat- aspire and to take advantage of the ters (Di Tella, Galiani, and Schargrodsky 2007). This situation created a natural experiment in which some opportunities that do present themselves. squatters acquired formal titles to their land, while others—sometimes right next door—did not. People with titles were 31 percent more likely to believe that it The impact of this simple five-minute intervention is possible to be successful alone, without a large group extended beyond an increase in abstract cognitive abil- in which everyone supports one another, and they ity. The researchers had also set up information booths were 34 percent more likely to believe that money is near the door of the soup kitchen that would have indispensable for happiness. They were also 17 percent appeared unrelated to the experiment. The affirmed more likely to report that other people in their country group was 31 percentage points, or 300 percent, more could be trusted. likely to pick up flyers about antipoverty programs for The effects of poor frames are not confined simply which they were eligible. to attitudes. Recent experimental evidence suggests that changing the frame through which poor people Social contexts of poverty can see themselves can alter school achievement among generate their own taxes poor children and improve interest in antipoverty In low-income settings, which often lack formal insti- programs among poor people. An intervention in tutions, informal institutions or social norms may fill the United States, for example, directed seventh grad- the gap. For example, poor households often benefit ers (12- to 13-year-olds) to use techniques of self- from a form of social insurance, tapping resources affirmation, which serve as reminders of sources of from friends, neighbors, family, and social groups, self-worth and pride. Throughout the school year, such as burial societies or rotating pools of credit, when students completed three to five structured writing their access to formal credit is limited and coverage by assignments that lasted 15 minutes each, writing about formal insurance is negligible. When they encounter values important to them, such as relationships with adverse shocks to their income, they can turn to such their family or their competence in art. This interven- social insurance to cushion their consumption, which tion helped narrow the achievement gap between tends not to plummet to the same extent as the shock at-risk minority students and other students. At the to income (Townsend 1995; Jalan and Ravallion 1999). end of eighth grade, more than a year after their last This means, however, that someone else in their social self-affirming writing assignment, African-American network is giving up resources to help. students sustained improvements in their grades and While this situation may very well be welfare decreases in grade repetition, particularly those who enhancing (especially if the development of formal were initially performing less well in school (Cohen insurance and credit markets is a long way in the and others 2006, 2009). future), investments in social capital carry their own set These results mirror the impact of a self-affirmation of costs, amounting to another kind of “tax.” According experiment among people who received lunch services to recent evidence, people in such situations want to in an inner-city soup kitchen in the United States. insulate some of their income from these types of social Some participants were asked to take three to five obligations. Nearly 20 percent of members in a micro- 86 WORLD DEVELOPMENT REPORT 2015 finance network in Cameroon, for example, appear to Sometimes, however, escaping these social obliga- take loans simply to signal that they have no cash to tions comes with a cost. In rural Paraguay, for exam- give to relatives and friends (Baland, Guirkinger, and ple, farmers who do not provide gifts to some people Mali 2011). These results were mirrored in a laboratory in their community risk theft of their crops (Schechter experiment in Kenya, in which women were willing to 2007). The diversion of assets to cover social obliga- pay a price to keep their earnings from a game hidden. tions like these may come at the expense of investment This tendency was more pronounced among women in private opportunities. whose relatives were also participating in the experi- ment (Jakiela and Ozier 2012). Implications for the design People in this situation may benefit from financial of antipoverty policies and products that allow them to insulate their income from programs social demands. A field experiment in Kenya demon- A number of constraints associated with poverty may strates that using a simple metal box with a padlock be difficult to observe and could extend beyond mate- and designating savings for a particular purpose can rial deprivation: a preoccupation with daily hassles help increase savings for people who must assist oth- and their associated depletion of cognitive resources ers in their social network. People were offered four required for important decisions; low self-image and its blunting of aspirations; and norms that may require investments in social capital to the detriment of private There are three promising ways to opportunities. Do these new insights into the decision- making contexts of poverty have any implications for ensure that poor people have adequate the design of policies and programs that target poor people? Much of the evidence is still new, and some cognitive space to make the best of the most intriguing results come from laboratory experiments that only simulate decision making in decisions: simplify procedures; target the real world. Nevertheless, some general lessons are emerging, along with some promising areas for assistance on the basis of bandwidth; and improvement. continue existing antipoverty strategies Minimizing cognitive taxes for poor people Previous chapters have demonstrated that everyone that aim to reduce income volatility and has limited “cognitive budgets,” which can make deci- sion making rather costly. This chapter makes clear improve infrastructure. that poverty often makes these budgets even tighter. While programs and policies rarely intend to make people poorer in a monetary sense, they sometimes types of savings products meant to increase spending impose cognitive taxes on poor people (Shah, Mul- on preventive health care and savings for health emer- lainathan, and Shafir 2012). There are three potentially gencies: a metal box with a padlock and a key; a locked promising ways to ensure that people living in poverty box without a key whose contents could be spent only have adequate cognitive space to make the best deci- on a preventive health care product; a health savings sions for themselves. The first is to simplify procedures account meant only for health emergencies; and mem- for accessing services and benefits. The second is to bership in a rotating savings and credit association, expand the criteria used for targeting assistance—in in which a group of individuals together make regu- particular, to target on the basis of bandwidth rather lar contributions and take turns receiving the funds. than wealth and expenditures alone. Finally, existing Sign-up for all these kinds of commitment devices was antipoverty policy instruments, such as cash transfers high: 66 percent 12 months after the program began, or the provision of infrastructure, may also generate and 39–53 percent three years later. Most notably, the positive impacts in the cognitive and psychological people in the community who gave out assistance to domains. others but received nothing in return benefited the most from these products (Dupas and Robinson 2013). Simplifying procedures Their savings for preventive health care increased For many programs around the world, in both low- more than did the savings of those who did not have to income and high-income settings, the procedures provide as much to their network. for accessing benefits—from filling out application POVERTY 87 forms to deciphering the rules of a program—can be was declared an underperforming school. Before 2004, daunting. While these might seem like minor transac- to find information about options, parents had to sift tion costs compared to the potentially large and often through a booklet that was more than 100 pages long long-term benefits of some programs, application and search a website for schools’ test scores to make forms have affected the take-up of many programs tar- school-by-school comparisons. After 2004, national geting low-income populations. In Morocco in 2007, regulations required that information about the test for example, a program was introduced that allowed scores of every school in the district be distributed in low-income households without piped water to buy a three-page spreadsheet. After the reform, parents on credit a connection to the water and sanitation in these situations chose higher-performing schools network in Tangier. To apply, these households had (Hastings and Weinstein 2008). to obtain authorization from their local authorities, In many contexts, however, governments and other provide photocopies of identification documents, agencies may want to limit participation in programs, and make a down payment at a local office. These especially if there is substantial leakage of benefits to procedures were sufficient to suppress participation; ineligible populations. A large cash transfer program in six months after the program was introduced, only 10 Indonesia (in which each household receives $130 per percent of households had signed up (Devoto and oth- year for six years) experimented with setting up small ers 2012). In an experiment, some households received hurdles to see if the number of ineligible households information about the program and assistance with benefiting from the program would decrease. Requir- the application procedures delivered right to their ing the poor to come to a centralized location in the door, including a visit by the local branch officer to village to be assessed for eligibility did improve the collect the down payment. Participation for this group efficiency of targeting, compared to a scheme in which reached 69 percent. government workers used the recommendations of In California, providing assistance to complete appli- village leaders and assessed the eligibility of families cations for health insurance for poor people (Medicaid) in their homes (Alatas and others 2013). These barriers, improved enrollment among the Hispanic population however, also prevented eligible households from ben- by 7 percent and among the Asian population by 27 per- efiting from the program. Among these households, cent. These impacts exceeded the results from advertis- average program take-up still reached only 15 percent, ing campaigns offered in Spanish and Asian languages and close to 40 percent of the poorest households did to reach those populations (Aizer 2007). Similarly, in the not even attempt to sign up. U.S. states of Ohio and North Carolina, the application How can development professionals be sure that rates of low-income students for financial aid or their program designs do indeed minimize, or at least avoid eventual attendance in college was not affected by maximizing, cognitive taxes on poor people? It should efforts to provide information alone about eligibility be fairly easy and quick to experiment with different and nearby colleges. In contrast, when low-income access procedures. What would be even easier, and per- parents who sought assistance in filing their federal haps more illuminating, would be for the designers of taxes were asked if they wanted to spend an additional programs to undergo the sign-up process themselves 10 minutes to use the tax information they had just before the program is launched (see the discussion finished providing to complete the federal form for in chapter 10 on “dogfooding,” the process by which financial aid for college, the college attendance of their product designers must try things out for themselves children increased by nearly 24 percent (Bettinger and before releasing their products to the market). others 2012). The extra 10 minutes of personal assis- tance in filling out the financial aid form made a big Targeting on the basis of bandwidth difference. It spurred beneficiaries to fill out the main While the poorest households—those falling below application forms of colleges and universities on their the threshold of $1.25 a day—are highly likely to incur own and gain admission to these institutions. the cognitive and social taxes described earlier, there This is not to suggest that information is unimpor- may be other easily identifiable populations that could tant or that poor people should be automatically signed benefit from assistance that helps them avoid errors in up for antipoverty programs. Indeed, the problem decision making when their bandwidth is low or when might just be that the information intended for them the bandwidth required to make a decision is fairly is too complex and too cognitively taxing to act upon. high (figure 4.4). In North Carolina, for example, parents could choose One such group includes people who work in a new school for their children when their current occupations where they receive earnings only once or school performed poorly on standardized tests and twice a year, such as cultivators or agricultural laborers. 88 WORLD DEVELOPMENT REPORT 2015 Programs to assist those living in poverty would ide- Bogotá that varied the structure of payments in a con- ally pay more attention to the timing of decisions and ditional cash transfer program that targeted families prevent them from coinciding with times when bene- with children in secondary school. Some households ficiaries’ cognitive resources may be heavily taxed. The received transfers every two months after meeting Indian sugar cane farmers described earlier, for exam- conditions related to the health and schooling of their ple, should best avoid making time-sensitive decisions children. Others received only two-thirds of the bene- about enrolling their children in school right before fit every two months, while the remaining third was harvest. saved in a bank account. These households were then Similarly, these kinds of investment decisions may given the savings in one lump sum in December, when be compromised if they happen to fall during months students are supposed to enroll for the next school year. when people may be particularly cash strapped because While both types of transfers were equally effective in of social obligations, as in the months coinciding with improving school attendance during the year, the sav- a festival or holiday, or because of a shock related to ings variant was more successful in increasing rates health or income. Indeed, farmers in Kenya whose of reenrollment for the next year (Barrera-Osorio and crops are dependent on rainfall exhibit higher stress others 2011). Similarly, as discussed in chapter 7, farm- (as measured by the hormone cortisol) when it does ers in Kenya increased their rates of adopting fertilizer not rain and their crops are therefore more likely to fail if they were given the opportunity to prepurchase it at (Chemin, De Laat, and Haushofer 2013). This kind of the time of harvest, when they would have more funds, stress has been associated with a bias toward the pres- rather than months later when they would be applying ent in laboratory environments. For example, when the fertilizer. subjects were asked to perform tasks that involved There are also important decisions that occur rel- deciding between smaller rewards sooner and larger atively infrequently and that inherently require con- rewards later, those who had been first administered siderable bandwidth. These might include applying hydrocortisone (which artificially elevated their corti- to a university or choosing a health insurance plan. In sol levels) showed a stronger tendency to opt for the the United States, for example, high school students earlier rewards (Cornelisse and others 2013). taking a popular college entrance exam can choose To see the benefits of altering the timing of an to have their scores sent directly to the universities intervention, consider some experiments in the city of to which they plan to apply. Before the fall of 1997, Figure 4.4 Targeting on the basis of bandwidth may help people make better decisions Bandwidth may be especially low at certain times, such as periods of higher expenditures during festivals, or when a mother is about to give birth. Key decisions, such as whether to enroll a child in school or whether to go to the hospital for a baby’s birth, would ideally be moved out of these periods. Some decisions, such as choosing a health insurance plan or applying to a university, may require high levels of bandwidth no matter when they fall. Policies that make these decisions easier could be targeted at the time of decision making. During a moment of bandwidth move decisions to another time or target assistance to the time need, policies should try to of the decision. Source: WDR 2015 team. POVERTY 89 students could send three reports for free, and each ments suggests that these types of interventions at additional score report would have cost $6 to send. least improve self-reported mental well-being. The When the number of reports they could send for free program in Morocco, discussed earlier, made it easier increased to four later that year, the number of test for households to obtain a connection for piped water; takers sending exactly four reports jumped from 3 this improvement reduced the time residents spent percent to 74 percent (Pallais, forthcoming). More fetching water by more than 80 percent (Devoto and important, this increase in score reports induced others 2012). Beneficiaries were more likely to perceive low-income students to apply to and eventually attend that their life had improved in the previous year and more selective universities. Because attending a more reported higher life satisfaction—despite a 500 percent selective university is associated with higher expected increase in their water expenditures and an absence future earnings, an effective subsidy of $6 improved of any improvements to their health. Similarly, a large, the expected earnings for low-income students by an one-time cash transfer to rural Kenyan households estimated $10,000. reduced symptoms of depression and stress approx- In Tanzania, promoters of community health insur- imately four months later (Haushofer and Shapiro ance took advantage of the disbursements of a condi- 2013). And a program in India that targeted the poorest tional cash transfer program to enroll more households of the poor suggests that antipoverty assistance can in the community health fund. They deliberately went have positive spillover effects beyond narrow program to the distribution points of the cash transfer program objectives (Banerjee and others 2011). The program to sign people up for the health insurance when they provided a livestock asset and a time-limited stipend had greater liquidity. This perhaps contributed to the for beneficiaries. Food consumption and nonlivestock nearly 20 percentage point (700 percent) increase in income increased beyond the monetary value of any- the use of health insurance to finance medical treat- thing provided. Program participants worked more ment among beneficiaries of the cash transfer pro- and reported improvements along many measures of gram (Evans and others 2014). mental well-being. Policy makers, however, cannot blindly target all situations in which income fluctuates, believing that Avoiding poor frames they have pinpointed contexts in which cognitive Poverty can contribute to a mindset that can make it bandwidth is likely to be low. It is important that these difficult for people to realize their own potential to fluctuations trigger financial stress. In the United take advantage of existing opportunities. It is impor- States, for example, cash-at-hand is typically higher tant to consider how the process of delivering services immediately after payday for low-income households or targeting poor people could be creating poor frames (it can be more than 20 percent lower immediately that further demotivate potential beneficiaries. A before payday). This predictable variation before and good place to start would be the names of programs after payday each month, however, is not associated and identification cards associated with them. “Needy with differences in cognitive function or risk taking families,” for example, could be replaced with “families (Carvalho, Meier, and Wang 2014). While this finding in action,” or “poor cards” with “opportunity cards.” may seem to conflict with the results from the sugar The distribution of productive assets and cash trans- cane farmers discussed earlier, people reported similar fers may help shift frames from despair to opportunity, levels of financial stress before and after payday, sug- as discussed. It may also be worth tackling aspirations gesting that temporary shortfalls, in this context, may more directly by paying attention to how poor people not further tax mental resources. regard themselves when deciding whether or not to apply for benefits. People working in social welfare Reducing economic volatility and offices or unemployment agencies, for example, can improving infrastructure be trained to avoid language and attitudes that could The natural and laboratory experiments discussed ear- be considered demeaning. In Peru, for example, focus lier suggest that monetary concerns absorb consider- group discussions revealed that beneficiaries often felt able cognitive capacity and blunt aspirations. Does this stigmatized when they went to health centers to fulfill mean that interventions that try to reduce economic the requirements for a cash transfer program (Perova volatility or directly decrease the cognitive demands of and Vakis 2013). Service providers would make them environments could also free up cognitive resources wait longer than other patients and stigmatize them or generate the confidence required to take advantage by overtly referring to the fact that they were receiving of economic opportunities? money from the government. While few programs currently monitor these Given all the design features of programs that can be kinds of impacts, some evidence from field experi- tweaked in this way, it might be difficult to predict how 90 WORLD DEVELOPMENT REPORT 2015 poor people will react and how transitory the effects programs can be more effective when the community of such manipulations might be. Experimentation can leaders of beneficiaries also participate in the program. be helpful, though, even on a small scale. Members of When several leaders in a community also received the Behavioural Insights Team of the U.K. government, conditional cash transfers, beneficiaries’ educational for example (see chapter 11), first worked in a single job investment and nutrition improved, as did the heights center to test whether interventions such as expressive and weights of their children (Macours and Vakis, writing or self-affirmation of strengths could move forthcoming). Social interactions between community job seekers off unemployment benefits and into a job leaders and the main beneficiaries amplified the effects more quickly. Based on their initial success, they have of the transfer program alone. Predicting exactly when set up a larger experiment in an entire region. these social relationships can help or hinder progress is still an open question and thus requires careful test- Incorporating social contexts into the ing of program design (see chapter 11). design of programs Designing programs that incorporate social contexts, Looking ahead however, poses a challenge. One extreme intervention More generally, this chapter has provided a new set of is to remove poor people from their current neighbor- diagnoses to explain decision making in contexts of hoods—although this is very expensive and not easily poverty and thus a new set of hypotheses to be tested scalable. For example, a large-scale experiment in the before designing a program or policy to assist poor United States, the Moving to Opportunity program, people. To return to the opening example of a father’s offered poor families a housing rental voucher that decision about whether to enroll his son in secondary could be redeemed only in neighborhoods with low school, it might be worth considering the cognitive, poverty. While adults reported better physical and psychological, and social barriers that might also mental health and higher subjective well-being 10–15 interfere with this particular investment decision, in years later, earnings, employment, and reliance on addition to testing the effectiveness of a scholarship, welfare payments did not change (Ludwig and others information campaign, or cash transfer program. For 2012). Moreover, the effects of the program on chil- example, if the decision falls during a period of partic- dren were mixed. The physical and mental health of ularly low income or high expenditure, a policy maker female youth improved and their engagement in risky could experiment with moving the decision to a less behavior declined, but the mental health of male youth financially stressful period or with offering prepur- declined, while their risky behaviors increased (Kling, chase opportunities when income is expected to be Liebman, and Katz 2007; Kessler and others 2014). high. Enrollment could also be made the default option An alternative approach to moving people out of so that parents would have to actively unenroll their their social environments would be to provide safe- child, as is now the case in Mexico’s main conditional guards that help mitigate the effects of demands from cash transfer program, which signs up beneficiaries others—for example, offering options that could help for automatic school enrollment. make savings harder to share. Savings accounts explic- If the father’s reluctance to enroll his son stemmed itly earmarked for certain purposes, for example, could from a deficit of aspirations, then programs that help stave off requests from friends and relatives, as directly tackle this lack of hope might also help. In they did in Kenya (Dupas and Robinson 2013). Chapter 7 Peru, for example, a financial literacy program first discusses a case in which illiquid transfers—such as an conducted a series of “self-esteem talks” among benefi- in-kind grant of equipment—can also insulate precious ciaries so that they understood that financial products funds from others. Whether or not these options can like savings accounts were real options for them (Per- ultimately improve welfare, however, is an empirical ova and Vakis 2013). question—especially in cases in which social networks If social demands left very few resources for the often substitute for more formal markets, such as the father to use for education, then financial products markets for credit and insurance, whose development that credibly earmarked savings for educational pur- may be far off in the future. poses might also help, just as they helped promote On the more positive side, as seen in chapters 6 and health savings in Kenya in the experiment described 7, social networks can also speed up the adoption of earlier (Dupas and Robinson 2013). certain financial products, such as crop insurance or Which one of these factors is the binding constraint microcredit, and foster social interactions and social in a particular context is very much an empirical learning that can improve earnings. Similarly, recent question that requires both good diagnosis and active evidence from Nicaragua suggests that antipoverty experimentation. While a great deal of empirical data POVERTY 91 exists describing the material deprivations that poor Evidence from Changes in Financial Resources at people experience, identifying metrics of the cognitive, Payday.” Unpublished. psychological, and social dimensions of poverty is still Chemin, Matthieu, Joost De Laat, and Johannes Haush- a new area of research (see spotlight 3). Similarly, the ofer. 2013. “Negative Rainfall Shocks Increase Levels of the Stress Hormone Cortisol among Poor Farmers in evidence base is still thin as to which program designs Kenya.” Unpublished. can directly open up the cognitive space required to Cohen, Geoffrey L., Julio Garcia, Nancy Apfel, and Alli- make complex decisions and increase the motivation son Master. 2006. “Reducing the Racial Achievement and aspiration required to take advantage of the oppor- Gap: A Social-Psychological Intervention.” Science 313: tunities that do arise. The potential for complementari- 1307–10. ties between programs that target income poverty and Cohen, Geoffrey L., Julio Garcia, Valerie Purdie-Vaughns, those that address cognitive bandwidth—such as access Nancy Apfel, and Patricia Brzustoski. 2009. “Recursive to finance or the development of infrastructure that Processes in Self-Affirmation: Intervening to Close the helps reduce the stresses of daily life—may be high but Minority Achievement Gap.” Science 324: 400–03. as yet has been largely undocumented. Ideally, more Cornelisse, S., V. A. van Ast, J. Haushofer, M. S. Seinstra, evidence will emerge as researchers and policy makers M. Kindt, and M. Joëls. 2013. “Time-Dependent Effect of Hydrocortisone Administration on Intertemporal experiment with programs that try to better align anti- Choice.” Unpublished. poverty interventions with the decision-making needs Dercon, Stefan, and Pramila Krishnan. 2000. “Vulnera- of those who find themselves in contexts of poverty. bility, Seasonality, and Poverty in Ethiopia.” Journal of Development Studies 36 (6): 25–53. References Devoto, Florencia, Esther Dulfo, Pascaline Dupas, William Aizer, Anna. 2007. “Public Health Insurance, Program Parienté, and Vincent Pons. 2012. “Happiness on Tap: Take-Up, and Child Health.” Review of Economics and Piped Water Adoption in Urban Morocco.” American Statistics 89 (3): 400–415. Economic Journal: Economic Policy 4 (4): 68–99. Alatas, Vivi, Abhijit Banerjee, Reema Hana, Benjamin J. Di Tella, Rafael, Sebastian Galiani, and Ernesto Schar- Olken, Ririn Purnamasari, and Matthew Wai-Poi. 2013. grodsky. 2007. “The Formation of Beliefs: Evidence “Self-Targeting: Evidence from a Field Experiment in from the Allocation of Land Titles to Squatters.” Quar- Indonesia.” Unpublished. terly Journal of Economics 121 (1): 209–41. Appadurai, Arjun. 2004. “The Capacity to Aspire: Culture Duflo, Esther. 2012. “Human Values and the Design of and the Terms of Recognition.” In Culture and Public the Fight against Poverty.” Lecture given at Harvard Action, edited by Vijayendra Rao and Michael Walton. University as part of the Tanner Lectures on Human Washington, DC: World Bank; Palo Alto, CA: Stanford Values. May. University Press. Dupas, Pascaline, and Jonathan Robinson. 2013. “Why Baland, Jean-Marie, Catherine Guirkinger, and Charlotte Don’t the Poor Save More? Evidence from Health Sav- Mali. 2011. “Pretending to Be Poor: Borrowing to Escape ings Experiments.” American Economic Review 103 (4): Forced Solidarity in Cameroon.” Economic Development 1138–71. and Cultural Change 60 (1): 1–16. Evans, David K., Stephanie Hausladen, Katrina Kosec, and Banerjee, Abhijit, Esther Duflo, Raghabendra Chatto- Natasha Reese. 2014. Community-Based Conditional Cash padhyay, and Jeremy Shapiro. 2011. “Targeting the Transfers in Tanzania: Results from a Randomized Trial. Hard-Core Poor: An Impact Assessment.” Unpublished. Washington, DC: World Bank. Banerjee, Abhijit, and Sendhil Mullainathan. 2008. “Lim- Guiliano, Paola, and Antonio Spilimbergo. 2014. “Grow- ited Attention and Income Distribution.” American ing Up in a Recession.” Review of Economic Studies 81: Economic Review 98 (2): 489–93. 787–817. Barrera-Osorio, Felipe, Marianne Bertrand, Leigh Linden, Guyon, Nina, and Elise Huillery. 2014. “The Aspiration- and Francisco Perez. 2011. “Improving the Design of Poverty Trap: Why Do Students from Low Social Back- Conditional Cash Transfer Programs: Evidence from ground Limit Their Ambition? Evidence from France.” a Randomized Education Experiment in Colombia.” Unpublished. American Economic Journal: Applied Economics 3 (2): Hall, Crystal, Jiaying Zhao, and Eldar Shafir. 2013. “Self- 167–95. Affirmation among the Poor: Cognitive and Behavioral Bettinger, Eric P., Bridget Terry Long, Philip Oreopoulos, Implications.” Psychological Science 25 (2): 619–25. and Lisa Sanbonmatsu. 2012. “The Role of Application Hastings, Justine S., and Jeffrey M. Weinstein. 2008. Assistance and Information in College Decisions: “Information, School Choice, and Academic Achieve- Results from the H&R Block FAFSA Experiment.” Quar- ment: Evidence from Two Experiments.” Quarterly terly Journal of Economics 127 (3): 1205–42. Journal of Economics 123 (4): 1373–1414. Carvalho, Leandro, Stephan Meier, and Stephanie W. Haushofer, Johannes, and Ernst Fehr. 2014. “On the Psy- Wang. 2014. “Poverty and Economic Decision-Making: chology of Poverty.” Science 344 (6186): 862–67. 92 WORLD DEVELOPMENT REPORT 2015 Haushofer, Johannes, and Jeremy Shapiro. 2013. “House- Mullainathan, Sendhil, and Eldar Shafir. 2013. Scarcity: hold Response to Income Changes: Evidence from Why Having Too Little Means So Much. New York: Times an Unconditional Cash Transfer Program in Kenya.” Books. Unpublished. Narayan, Deepa, Robert Chambers, Meera K. Shah, and Jakiela, Pamela, and Owen Ozier. 2012. “Does Africa Need Patt Petesch. 2000. Voices of the Poor: Crying Out for a Rotten Kin Theorem? Experimental Evidence from Change. Washington, DC: World Bank. Village Economies.” Policy Research Working Paper Narayan, Deepa, Lant Pritchett, and Soumya Kapoor. 2009. 6805, World Bank, Washington, DC. Moving Out of Poverty: Success from the Bottom Up. Wash- Jalan, Jyotsna, and Martin Ravallion. 1999. “Are the Poor ington, DC: World Bank and Palgrave Macmillan. Less Well Insured? Evidence on Vulnerability to Pallais, Amanda. Forthcoming. “Small Differences That Income Risk in Rural China.” Journal of Development Matter: Mistakes in Applying to College.” Journal of Economics 58 (1): 61–81. Labor Economics. — —— —. 2000. “Is Transient Poverty Different? Evidence for Perova, Elizaveta, and Renos Vakis. 2013. “Improving Gen- Rural China.” Journal of Development Studies 36 (6): 82–99. der and Development Outcomes through Agency: Pol- Kessler, Ronald C., Greg J. Duncan, Lisa A. Gennetian, icy Lessons from Three Peruvian Experiences.” Report Lawrence F. Katz, Jeffrey R. Kling, Nancy A. Samp- 79713. Washington, DC: World Bank. son, Lisa Sanbonmatsu, Alan M. Zaslavsky, and Jens Pritchett, Lant, Asep Suryahadi, and Sudarno Sumarto. Ludwig. 2014. “Associations of Housing Mobility 2000. “Quantifying Vulnerability to Poverty: A Pro- Interventions for Children in High-Poverty Neigh- posed Measure, Applied to Indonesia.” Policy Research borhoods with Subsequent Mental Disorders during Working Paper 2437, World Bank, Washington, DC. Adolescence.” Journal of the American Medical Association Ravallion, Martin. Forthcoming. The Economics of Poverty. 311 (9): 937–48. New York: Oxford University Press. Kling, Jeffrey R., Jeffrey B. Liebman, and Lawrence Ray, Debraj. 2006. “Aspirations, Poverty, and Economic Katz. 2007. “Experimental Analysis of Neighborhood Change.” In Understanding Poverty, edited by A. Baner- Effects.” Econometrica 75 (1): 83–119. jee, R. Bénabou, and D. Mookherjee. Oxford, U.K.: Ludwig, Jens, Greg J. Duncan, Lisa A. Gennetian, Oxford University Press. Lawrence F. Katz, Ronald C. Kessler, Jeffrey R. Kling, Schechter, Laura. 2007. “Theft, Gift-Giving, and Trust- and Lisa Sanbonmatsu. 2012. “Neighborhood Effects worthiness: Honesty Is Its Own Reward in Rural Para- on the Long-Term Well-Being of Low-Income Adults.” guay.” American Economic Review 97 (5): 1560–82. Science 337: 1505–10. Sen, Amartya. 1983. “Poor, Relatively Speaking.” Oxford Macours, Karen, and Renos Vakis. Forthcoming. “Chang- Economic Papers 35 (2): 153–69. ing Households’ Investment Behavior through Social Shah, Anuj, Sendhil Mullainathan, and Eldar Shafir. 2012. Interactions with Local Leaders: Evidence from a Ran- “Some Consequences of Having Too Little.” Science 338 domized Transfer Program.” Economic Journal. (6107): 682–85. Mani, Anandi, Sendhil Mullainathan, Eldar Shafir, and Townsend, Robert M. 1995. “Consumption Insurance: An Jiaying Zhao. 2013. “Poverty Impedes Cognitive Func- Evaluation of Risk-Bearing Systems in Low-Income tion.” Science 341: 976–80. Economies.” Journal of Economic Perspectives 9 (3): 83–102. 94 WORLD DEVELOPMENT REPORT 2015 How well do we understand the contexts of poverty? Behavioral economics has uncovered a number of ioral economics among representative samples surprising instances in which choices are influenced in three capital cities around the world (Jakarta, by factors that should seemingly be irrelevant, as Indonesia; Nairobi, Kenya; and Lima, Peru) and chapters 1–3 have discussed (see also Kahneman and among a sample of staff working at the World Spotlight 3 Tversky 1984; Kahneman 2010; Ariely 2008, 2010). Bank. These small inconsistencies have often been The results suggest that the choices made by revealed through people’s responses to vignettes or World Bank staff tend to replicate the choices made hypothetical situations. These vignettes have been by university and affluent samples. The choices of implemented mostly among samples of university people living in poor countries do not; their choices students attending elite universities. Do these tend to mirror the choices of a sample of poor peo- patterns reveal something universal about human ple in the United States. decision making, or could these choices perhaps be a function of wealth, just as susceptibility to some Responses of poor and affluent people visual illusions and preferences for fairness appear in New Jersey (United States) to be unique to certain societies (Henrich, Heine, In the United States, there is evidence that poor and and Norenzayan 2010)? affluent respondents do not use the same mental To find out, the World Development Report 2015 shortcuts (heuristics) when evaluating the bene- team implemented a classic vignette from behav- fit of a discount and that poorer respondents can Figure S3.1 How poor and affluent people in New Jersey view traveling for a discount on an appliance Appliance cost Poor $100 $500 $1,000 Affluent $100 $500 $1,000 0 10 20 30 40 50 60 70 80 90 100 Percentage that would travel for a discount Source: Hall 2008. Note: The discount was $50. HOW WELL DO WE UNDERSTAND THE CONTEXTS OF POVERTY? 95 make more consistent choices about the trade-off Responses of residents in Jakarta, between money (or the discount) and time. In a Nairobi, and Lima study in New Jersey, for example, three groups of In Jakarta, Nairobi, and Lima, residents from var- respondents were randomly assigned to read one ious wealth groups answered a similar question of three variants of the following vignette, which about a cell phone. The choices of respondents in differed solely in the total cost of an appliance that these cities much more closely resembled respon- could be purchased: dents’ choices in the New Jersey soup kitchen. In each city, respondents were stratified Imagine that a friend goes to buy an appliance across three wealth groups—lower, middle, and priced at $100 ($500, $1,000). Although the upper—which corresponded to terciles defined by store’s prices are good, the clerk informs your community averages for the poverty rate (Jakarta), friend that a store 45 minutes away offers assets (Nairobi), or consumption (Lima). Since these the same item on sale for $50 less. Would you wealth groups were defined within each country, it advise your friend to travel to the other store is possible that even respondents from the upper to save $50 on the $100 ($500, $1,000) item? groups correspond more closely to poorer popula- tions in more affluent countries. The total cost of the appliance was irrelevant Across all these wealth categories in Jakarta, for poor respondents in a New Jersey soup kitchen Nairobi, and Lima, the total price of the cell phone when deciding whether they would advise travel- rarely had a statistically significant bearing on ing for a discount (Hall 2008). Each group made whether a respondent would travel for a discount. the same choice as other groups that had randomly This finding contrasted with the more affluent received a different price. A sample of more affluent respondents in the United States and the World Spotlight 3 commuters at a train station, however, was signifi- Bank, where each increase in the total price of the cantly less likely to favor travel as the price of the product significantly diminished the attractiveness appliance rose, consistent with findings from uni- of traveling for a discount.1 (See figures S3.3, S3.4, versity students in the United States and Canada and S3.5.) (Tversky and Kahneman 1981). This suggests that they focused on relative savings, instead of absolute Implications savings. In every scenario, all respondents were Some have argued that differences like these contemplating the same trade-off: spending 45 between poor and wealthy respondents relate to dif- minutes to save $50. For the affluent sample, saving ferences in the degree to which monetary concerns $50 seemed like a better deal when the appliance are salient (Hall 2008; Mullainathan and Shafir was less expensive (see figure S3.1). 2013). Because even modest sums matter a great deal for poor people, they might focus on absolute Responses of World Bank staff savings. For more affluent people, these amounts For World Bank staff, the vignette was posed in do not trigger much concern; they may not imme- terms of deciding whether to travel for a $50 dis- diately think of alternative uses for the savings count on a watch. Staff exhibited a pattern similar and thus must focus on relative savings to gauge to the affluent samples of commuters and univer- whether or not the discount would be a good deal. sity students. Groups randomly receiving the more Regardless of the reasons, these results suggest expensive variant were significantly less likely to a divergence in preferences between people living say they would travel for a discount (see figure S3.2). in poor contexts and World Bank staff working to Figure S3.2 How World Bank staff view traveling for a discount on a watch Watch cost $150 $600 $1,250 0 10 20 30 40 50 60 70 80 90 100 Percentage that would travel for a discount Source: WDR 2015 team. Note: The discount was $50. 96 WORLD DEVELOPMENT REPORT 2015 Figure S3.3 How people in Jakarta, Indonesia, view traveling for a discount on a cell phone Cell phone cost Lower income Rp 500,000 Rp 1,000,000 Rp 5,000,000 Middle income Rp 500,000 Rp 1,000,000 Rp 5,000,000 Upper income Rp 500,000 Rp 1,000,000 Rp 5,000,000 0 10 20 30 40 50 60 70 80 90 100 Percentage that would travel for a discount Spotlight 3 Source: WDR 2015 team. Note: Rp = Indonesian rupiah. The discount was Rp 250,000. Figure S3.4 How people in Nairobi, Kenya, view traveling for a discount on a cell phone Cell phone cost Lower income K Sh 1,500 K Sh 3,000 K Sh 15,000 Middle income K Sh 1,500 K Sh 3,000 K Sh 15,000 Upper income K Sh 1,500 K Sh 3,000 K Sh 15,000 0 10 20 30 40 50 60 70 80 90 100 Percentage that would travel for a discount Source: WDR 2015 team. Note: K Sh = Kenyan shilling. The discount was K Sh 750. HOW WELL DO WE UNDERSTAND THE CONTEXTS OF POVERTY? 97 Figure S3.5 How people in Lima, Peru, view traveling for a discount on a cell phone Cell phone cost Lower income S/. 100 S/. 200 S/. 1,000 Middle income S/. 100 S/. 200 S/. 1,000 Upper income S/. 100 S/. 200 S/. 1,000 0 10 20 30 40 50 60 70 80 90 100 Percentage that would travel for a discount Spotlight 3 Source: WDR 2015 team. Note: S/. = Peruvian nuevo sol. The discount was S/. 50. design strategies to assist poor people. While there ————. 2010. The Upside of Irrationality: The Unexpected is no evidence that indicates these differences Benefits of Defying Logic at Work and at Home. New translate into ineffective antipoverty strategies, York: Harper. they should at least suggest caution when making Hall, Christel. 2008. “Decisions under Poverty: A assumptions about what motivates decision mak- Behavioral Perspective on the Decision Making ing in contexts of poverty. of the Poor.” PhD diss., Princeton University. Henrich, Joseph, Steve J. Heine, and Ara Note Norenzayan. 2010. “The Weirdest People in the 1. One exception is the case of respondents from World?” Behavioral and Brain Sciences 33: 61–135. the upper-wealth group in Lima, where limited Kahneman, Daniel. 2010. Thinking, Fast and Slow. willingness to participate in the survey severely New York: Farrar, Straus and Giroux. restricted the sample size of this population to Kahneman, Daniel, and Amos Tversky. 1984. 109 respondents across all question variants “Choices, Values, and Frames.” American and possibly introduced considerable noise in Psychologist 39 (4): 341–50. the data. Mullainathan, Sendhil, and Eldar Shafir. 2013. Scarcity: Why Having Too Little Means So Much. References New York: Times Books. Ariely, Daniel. 2008. Predictably Irrational: The Hidden Tversky, Amos, and Daniel Kahneman. 1981. “The Forces that Shape Our Decisions. New York: Harper Framing of Decisions and the Psychology of Perennial. Choice.” Science 211 (4481): 453–58. 5 CHAPTER Early childhood development Years before they set foot in school, children in poor the perspectives of other people, and focusing atten- families differ dramatically from children in richer tion—that are important for engaging effectively with families in their cognitive and noncognitive abilities. teachers and other children, paying attention in class, The differences have powerful and enduring conse- completing assignments, and behaving appropriately. quences for individuals’ health, well-being, education, The formative influence of poverty on family life and longevity, as well as for the societies in which chil- and developmental outcomes has been acknowl- dren grow into adulthood. edged for centuries, but only recently has it come to What is the source of these consequential dif- the forefront in economic thinking—in rich and poor ferences among children? It is well established that countries alike. Investing in policies that help dis- children living in poverty experience greater levels of advantaged families provide better support for their environmental and psychosocial stressors than their young children will have high rates of return (Heck- higher-income counterparts (Crockett and Haushofer man 2008). The emerging view of the potency of early 2014) and that stress and adversity in the first years experience in shaping both life outcomes and national outcomes is supported by new research in neurobiol- ogy, biopsychology, and developmental psychology. Children in poor families can differ The mental growth trajectories of children living in advantaged circumstances as compared with those dramatically from children in richer living in poverty begin to diverge very early in life. One goal of this chapter is to explore how experience families in their cognitive and beginning in infancy acts on biological mechanisms that cause these growth trajectories to diverge. The noncognitive abilities, resulting in robust differences among children in their cognitive and social competencies vary across contexts as well. enormous loss of human potential for Thus a second goal is to explore how differences in the mental models and parenting beliefs that motivate themselves and society. context-specific caregiving practices also contribute to the substantial gaps observed in children’s early lan- guage and cognitive development. Integrating research of life can permanently constrict the development of from the biological and sociocultural perspectives, this physical and mental capacities throughout adulthood chapter examines why millions of children fail to reach (Shonkoff and others 2012). Furthermore, children their developmental potential in the early years and from disadvantaged families are less likely to receive enter school without a strong foundation for learning, consistent support and guidance from responsive resulting in enormous loss of human potential. Finally, caregivers. They are also likely to have had less the chapter reports evidence that early childhood opportunity to develop the critical skills—including interventions can mitigate the effects of impoverished skills in controlling their impulses, understanding environments on children. The chapter demonstrates EARLY CHILDHOOD DEVELOPMENT 99 that social influences on the mind extend beyond their more complex reading and mathematical tasks at older effects on decision making, which part 1 highlighted, ages. The ability to recognize words was approximately and also include the long-term effects of the early social three-quarters of a standard deviation lower for chil- environment on cognitive and noncognitive skills. dren in the bottom wealth quintile than for children in the top wealth quintile. None of this difference can be Richer and poorer children differ explained by variation in maternal education because greatly in school readiness the estimates are already adjusted to account for differ- Gaps in children’s development between rich and poor ences in educational attainment on the part of mothers. households are substantial and emerge well before Recent work in five Latin American countries finds children enter school. In very low-income countries— further evidence of substantial wealth gradients in like Madagascar, where more than three out of four receptive language ability (Schady and others, forth- people live below $1.25 a day—children’s performance coming), the skill with the largest wealth gradient might be expected to be uniformly low. However, lan- in the Madagascar study and in research in devel- guage, cognitive abilities, and noncognitive skills of oped countries (Hackman and Farah 2009). Figure preschool children exhibit clear variations by wealth 5.2 presents the differences in receptive vocabulary (wealth gradients), as seen in figure 5.1. between the richest and the poorest wealth quartiles The wealth gradients were largest for receptive lan- in rural and urban areas in Chile, Colombia, Ecuador, guage (listening or reading comprehension), followed Nicaragua, and Peru. by executive function (sustained attention and work- Do the wealth gaps in children’s skills narrow over ing memory). Early language ability is predictive of time? The evidence to date indicates they do not. In later success in learning to read and to work with num- both the Madagascar and Latin American samples, by bers in the first years of school, as well as mastery of the time children enter school (age six), the differences Figure 5.1 Variations by wealth in basic learning skills are evident by age three in Madagascar There are socioeconomic gradients across a comprehensive set of child development measures in a population living in extreme poverty in an area of Madagascar. There are strong associations between wealth and child development outcomes among preschool children. Importantly, the largest gaps across socioeconomic groups are in receptive vocabulary, memory, and sustained attention, domains that are highly predictive of later success in school and adult outcomes. 108 106 104 Age-adjusted score 102 100 98 96 94 Receptive Working Memory of Visual-spatial Sustained Fluid vocabulary memory phrases processing attention reasoning 1st quintile (poorest) 2nd quintile 3rd quintile 4th quintile 5th quintile (wealthiest) Source: Fernald and others 2011. Note: Children between the ages of three and six were studied. The bars represent the average age-adjusted scores by wealth quintile (adjusted for maternal education) for each of the outcomes. The scores are normalized to have a mean of 100 and a standard deviation of 15. 100 WORLD DEVELOPMENT REPORT 2015 interaction and cognitive and socioemotional stimula- Figure 5.2 Abilities in receptive language for three- tion early in life, in addition to adequate nutrition and to six-year-olds vary widely by wealth in five Latin health care. The students more likely to flourish are American countries those who have established a foundation in multiple Wealth gradients in receptive language among preschool children across five skills that will affect their ability to perform well across Latin American countries are substantial in both rural and urban areas. The analysis adds systematic evidence that gaps in child development outcomes a wide range of domains. Noncognitive skills, and not appear early in the life cycle. just IQ, and related faculties such as working memory and cognitive processing, are very important. 0.8 Various disciplines characterize these soft skills in different ways. Some psychologists see them as related 0.6 to personality traits, while neurobiologists focus on the ability to control oneself (self-regulation) and related constructs. The cognitive components of self- 0.4 regulation, referred to as executive function, include Standardized scores the ability to direct attention, shift perspective, and adapt flexibly to changes (cognitive flexibility); retain 0.2 information (working memory); and inhibit automatic or impulsive responses in order to achieve a goal such as problem solving (impulse control) (Hughes 2011). For 0 U R U R U R R U R example, a child who ignores background noise in the classroom to focus on solving the math problems in front of him is relying heavily on these abilities. –0.2 Self-regulation also includes emotional components such as regulating one’s emotions, exhibiting self- control, and delaying gratification to enjoy a future –0.4 reward. Psychologists agree that skill in self-regulation should be considered a key component of school readi- –0.6 ness, just like emerging literacy (Blair and Diamond 2008; Ursache, Blair, and Raver 2012). Chile Colombia Ecuador Nicaragua Peru The rewards extend beyond the school years. Non- U Urban R Rural Richest quartile Poorest quartile cognitive skills are just as powerful as IQ and cognitive skills in predicting a wide range of life outcomes in Source: Based on table 2 in Schady and others, forthcoming. adulthood that are economically relevant and reinforce Note: The bars show the average age-standardized z-scores of receptive language for the richest and poorest quartiles of the distribution of wealth within each country, reported separately by urban (U) each other (Cunha and Heckman 2007, 2009).1 As Heck- and rural (R) areas. An important caveat is that data are generally representative of rural areas for all man (2008, 309) argues, “Skills beget skills.” Early suc- countries but are not representative for urban areas. No urban data are available for Nicaragua. cesses in learning make later investments more pro- ductive so that learning increases with higher levels of early skills (the self-productivity argument). In addition, in scores between children from the richest and the investments in skills at early stages increase the pro- poorest households widened beyond those seen at three ductivity of investments at later stages (the dynamic years of age and are virtually flat afterward. Similar complementarity argument). Cunha and Heckman (2007, gaps and patterns of persistence in academic test scores 2009) find strong evidence of self-productivity, espe- are observed between children from disadvantaged and cially for cognitive skills, and strong cross-productivity advantaged families in the United States (Farkas and effects of noncognitive skills on subsequent cogni- Beron 2004; Cunha and others 2006). tive skills, with important implications for the timing of policy. Children need multiple cognitive For children living in poverty, the development of and noncognitive skills to self-regulation skills can be disrupted by unpredictable succeed in school environments and sustained levels of stress. In addi- Where do these critical differences in children’s readi- tion, as discussed later in this chapter, disadvantaged ness for school come from? Although potential intelli- children are less likely to receive consistent support and gence appears to be partly inherited, adequate support guidance from responsive caregivers and are also likely from the environment is crucial for the development to have less opportunity to develop skills in impulse of children’s full potential. This support includes verbal control, perspective shifting, and focused attention. EARLY CHILDHOOD DEVELOPMENT 101 Poverty in infancy and early Figure 5.3 Unrelenting stress in early childhood can be childhood can impede early toxic to the developing brain brain development Toxic stress is the strong, unrelieved activation of the body’s stress management The new discoveries about the critical importance of system. This image depicts neurons in the brain areas most important for cognitive and noncognitive skills for success in school successful learning and behavior—the hippocampus and prefrontal cortex. and later life circle back to the central question: if The neuron shown in panel b, which has been subjected to toxic stress, clearly displays underdeveloped neural connections. children from rich and poor families differ substan- tially in core competencies when they enter school, a. Typical neuron: b. Neuron damaged by toxic when and how do these differences begin to emerge? stress: fewer connections many connections Such dramatic early differences are shaped by a multi- tude of environmental factors that can undermine the child’s development, including nutrition, health care, stress, and interactions between the child and care- givers. These factors can lead to a process of biological embedding. This occurs when differences in the quality of early environments provided to young children have direct effects on the sculpting and neurochemistry of the central nervous system in ways that impair later cognitive, social, and behavioral development. An infant frequently exposed to stressful events experiences persistent activation of a major part of the neuroendocrine system that controls reactions to Source: Shonkoff and others 2012. stress, the hypothalamic-pituitary-adrenal (HPA) axis. While responses to acute stress by the HPA axis can focus the body’s energy on the immediate task, and Weisleder 2013), disparities in the efficiency of lan- thus be helpful at the moment, prolonged and high guage processing and vocabulary by SES level were exposure to stress can result in chronically heightened evident at 18 months. By their second birthdays, there cortisol levels and maladaptive stress responses, even was a six-month gap between children from higher in young children. A child who reacts with extreme and lower SES families in processing skills known to anxiety to the small daily stresses in school can have be critical to language development. difficulty interacting with peers and can perform poorly on school assignments. Such experiences day Parents are crucial in supporting after day can reduce self-confidence and undermine the development of children’s academic achievement. The disappointments may capacities for learning continue to increase the child’s stress level in a feed- The discussion so far has focused on the negative side back loop that will further activate the HPA axis. of the biological embedding process. But there is also But that is just part of the story. Chronically ele- a positive side: the sensitive periods of early develop- vated stress in infancy affects the developing brain by ment represent a time of enormous growth if children damaging neurons in the areas involved in emotions are given sufficient positive support from the environ- and learning, as shown in figure 5.3, panel b. Elevated ment. Research in neurobiology now makes it clear stress can also impair the development of the prefron- that the consequences of early parenting, for better or tal cortex, which is the region of the brain crucial for worse, can also be biologically embedded. Supportive the emergence of the self-regulatory skills essential for parenting in early childhood is strongly predictive of success in school and adulthood (Shonkoff and others the development of brain structures, including the area 2012). Thus experiencing excessive stress and anxiety critical to the development of memory, the hippocam- in infancy impairs the early development of learning pus (Luby and others 2012). abilities and noncognitive skills, with cascading nega- tive consequences for later achievements. How parents support children’s language The neurocognitive systems or brain regions most learning vulnerable to the effects of adversity and differences Particular circumstances help infants learn their first in socioeconomic status (SES) in young children are words: those associated with language and executive func- tion (Noble, McCandliss, and Farah 2007). In a recent • Infants need to hear lots of words lots of times to study in the United States (Fernald, Marchman, and learn language, so repetition is valuable. 102 WORLD DEVELOPMENT REPORT 2015 • Parents can facilitate word learning by following the families in the United States revealed striking variabil- child’s interest and talking about what has engaged ity in the amount of adult speech addressed to the child the child’s attention. (Weisleder and Fernald 2013). One infant heard 100 • Children learn words best in meaningful contexts: words in five minutes, on average, while another heard knowledge is built by connecting words together only five words in five minutes. Infants who experi- in webs of meaning, not just by learning words in enced more child-directed speech at 18 months became isolation. more efficient in language-processing skill and had • Positive interactions support learning: asking ques- larger vocabularies by age 24 months. And it was only tions and elaborating on the child’s conversation are child-directed speech that mattered—speech that the more effective than giving commands that inhibit child simply overheard was unrelated to vocabulary curiosity. outcomes. These results revealed that even within a low-SES population of Spanish-speaking immigrants, However, caregivers vary considerably in their use caregiver speech had direct as well as indirect influ- of these supportive behaviors in interacting with an ences on language development. More exposure to infant. A landmark study found that families in differ- child-directed speech provides not only more exam- ent SES groups in the United States differed dramati- ples of words to learn but also more opportunities cally in the amount of child-directed speech that care- for practice, thus strengthening infants’ language- givers provided (Hart and Risley 1995). Children in the processing skills, with cascading benefits for vocabu- lary learning. Antipoverty programs and social policies How parents support children’s learning of executive function skills can have a powerful indirect effect on Given the robust relations between executive function skills at the time of preschool and children’s success child development by reducing key in later life, what is known about early precursors of these important noncognitive skills? As with language psychological stresses that prevent learning, the gradual development of children’s ability to resist impulsive responses, modulate their behavior, parents from attending to and engaging and plan ahead is strongly influenced by early expe- rience. Children in poverty are likely to have less well positively with their children. developed executive function skills than more advan- taged children. In families under stress, in which levels of harsh parenting are generally high, children often lowest SES group heard about 600 words per hour, have difficulties controlling inhibitions and regulating while children in the highest SES group heard more emotions (Lansford and Deater-Deckard 2012). This than 2,000 words per hour. By age four when they relation is consistent with neurobiological findings, entered preschool, the high-SES children had heard 30 discussed above, that early experience with high stress million more words directed to them than the low-SES has enduring effects on children’s reactivity to stress children. Caregivers’ speech varied in quality as well as (Shonkoff and others 2012) and that parents can play a in quantity. Parents in professional families were more critical role in protecting children against the negative likely to elaborate and use questions to encourage curi- effects of such stress. osity in the child, while parents in the low-SES families One aspect of parenting behavior important in nur- used more commands and prohibitions. turing executive function skills is scaffolding—a pro- If infants from advantaged and disadvantaged fam- cess by which a caregiver organizes and supports an ilies already differ in language-processing skills and activity to enable the child to succeed in a task beyond vocabulary at age 18 months—when many have barely his or her current level of ability. For example, a parent begun to speak—do differences in early language might scaffold a two-year-old’s effort to build a tower experience account for these disparities? Differences of blocks by helping the child choose blocks of the in caregivers’ speech to children do account for the right size and position them correctly. Scaffolding is a link between SES and the size of children’s vocabulary more complex skill than it might seem, since the adult (Hoff 2003). However, poverty in and of itself is not an must simplify the task to just the right level where the inevitable cause of the limited speech directed to chil- child can experience success, guide the child toward dren by caregivers. All-day recordings of parent-infant a particular goal, and manage the child’s frustration interactions at home in low-income Spanish-speaking if the task is difficult. The skillful caregiver must EARLY CHILDHOOD DEVELOPMENT 103 respond contingently to the child’s ongoing  activity, have not yet begun to talk (Kağıtçıbaşı 2007). For exam- while expanding that activity to direct it in more chal- ple, in rural villages in Kenya, Gusii mothers avoided lenging directions. Through incremental learning with eye contact with infants because of traditional beliefs emotional and cognitive support, scaffolding enables that direct gaze can be dangerous; thus the mothers children to gradually develop the abilities necessary to only rarely directed affectionate or social behaviors to solve tasks independently. Scaffolding can be strength- their babies, Dixon and others (1981) observed. While ened through appropriate interventions, as the final Gusii mothers were quick to protect, comfort, and feed section of this chapter will discuss. a crying infant, they tended to respond with touch and Since effective scaffolding frequently involves child- rarely with language. The use of contingent conversa- directed language, caregivers’ skill in using contin- tional communication strategies with young children, gent, situation-specific linguistic guidance also plays including the practices of verbal turn-taking and scaf- a role in the development of children’s emerging self- folding, were not observed. As children grew older, regulatory abilities. Mothers who provided more pos- parents spoke to them more often, but frequently used itive verbal stimulation with their toddlers at age two commands to direct the children to do something, had children who were better able to stay focused on a rather than using language to elaborate on their chil- task and to delay gratification at age six, Olson, Bates, dren’s interests. and Bayles (1990) found. Mothers who display a sensi- How can such variation in parenting behaviors tive and scaffolding parenting style also have children across different cultural groups be explained? The who have lower cortisol levels and better executive cultural psychologist Cigdem Kağıtçıbaşı (2007) pro- function skills (Blair and others 2011). Thus the quality vides functional explanations for cultural differences of parenting and verbal stimulation in infancy plays in parent-child relations by situating parental beliefs, a critical role in shaping the child’s stress-response values, and behaviors in their socioeconomic contexts. system and the development of critical noncognitive When a child is expected to make material contributions skills, as well as the development of language and cog- to the family, as in subsistence economies, then a utili- nitive skills more broadly. tarian value can be attributed to the child. Thus parents’ mental models of child rearing might be goal oriented, Parents’ beliefs and caregiving although not explicitly or consciously formulated practices differ across groups, in those terms. The child in a stable agrarian society with consequences for children’s whose future depends on mastering a traditional craft, developmental outcomes such as weaving, could be socialized to develop that competence through nonverbal observation of adult How different parenting styles evolve and weavers, with no need for extensive cognitive and lan- adapt differently to different economic guage stimulation early in life. However, as Kağıtçıbaşı contexts points out, “Teaching and learning limited to non-verbal Many studies in the United States have found dra- observational learning and non-inductive obedience- matic differences in caregiving behaviors among fam- oriented child socialization appear not to be optimal ilies. Parents with greater education and wealth tend for the promotion of high levels of cognitive and lin- to provide more cognitive and positive socioemotional guistic competence in the child” (2007, 83). In fact, she stimulation for their infants than do parents with less argues that these traditional socialization goals may education and fewer economic resources. Within SES be disadvantageous in contexts of social change—for groups, there is substantial variability as well. But how example, when uneducated parents must help their relevant are these findings to caregivers and children children prepare for formal education. To meet the new across the much broader range of contexts in develop- challenges of schooling, children need foundational ing countries? skills in language and executive function to acquire There has been little longitudinal research in devel- the higher-order cognitive abilities that are critical to oping countries that examines parenting behaviors in creative problem solving and success in school. relation to well-defined child outcomes. On the ques- tion of how parenting practices differ across societies, How parenting practices compare across however, ethnographers provide a rich literature of countries descriptive data. Many anthropological studies of How do parenting practices differ across high-, mother-child interaction in agrarian societies report middle-, and low-income countries? The association that parents are highly attentive to the safety and nutri- between two types of positive parenting practices— tional needs of infants, yet do not regularly engage in cognitive caregiving and socioemotional caregiving— social interaction or direct speech with children who and the country’s level of development, as measured 104 WORLD DEVELOPMENT REPORT 2015 by the Human Development Index (HDI), has been Another analysis of this data set focused on nega- examined in a recent study with comparable data from tive rather than positive parenting practices (Lansford 28 developing countries (Bornstein and Putnick 2012).2 and Deater-Deckard 2012). This study found a greater As discussed, cognitive caregiving—such as using prevalence of physical violence by parents toward child-directed language to stimulate the child’s under- children in countries with lower education, literacy, standing of the world—strongly predicts language and and income. Although these associations cannot be cognitive development. Socioemotional caregiving assumed to be causal, they corroborate the in-depth predicts the development of children’s interpersonal ethnographic studies that ground differences in paren- competencies and noncognitive skills. tal behaviors in their socioeconomic context. Mothers’ reports of the prevalence of these caregiv- ing practices differed substantially among countries, as Designing interventions that shown in figure 5.4. Mothers engaged more in socio- focus on and improve parental emotional than in cognitive caregiving overall, without competence much correlation with the level of the country’s devel- opment. While there were no consistent differences in Complementing direct antipoverty programs mean socioemotional caregiving by HDI level, mothers Antipoverty programs are often thought to affect child in each of the high-HDI countries engaged in more development through a traditional economic mech- cognitive caregiving activities than did mothers in the anism: alleviating income constraints during early low-HDI countries. Countries in the medium-HDI childhood enables parents to buy goods and services groups were split above and below the mean. that support child development. Can the wealth gaps Figure 5.4 There is greater variation across countries in cognitive caregiving than in socioemotional caregiving Cognitive caregiving activities, shown by the dark bars, tend to be much greater in countries with high Human Development Indexes (HDI) than in countries with low HDI, although there are only slight differences in socioemotional activities (light bars) across countries. The height of the bars with babies on them indicates the average number of cognitive caregiving activities reported by parents in low- and high-HDI countries. Low HDI Medium HDI High HDI Average number of caregiving activities 3 2 1 0 mb go nt Gu e d he Af ea ire R au rra blic e Be d Ar Jam ize Re ica Mo ublic Vi lia Ky zb am Re an Ta ublic me an p. ng a h Se o Ma B rbia nia s ia K Alba R d H kh a eg n ina do ru Ba Ghan an aza ni on r es erz sta an Y Re eg o an iss yz ist Ye ist T o Ga To l ab a ,F ov ce ela etn lad Sie epu ng ral in ’lv Le ail Cô ia, en p p rg ek jik ric -B n, Th nt Mo t U ian r Sy sn Bo Ce Cognitive caregiving activities Socioemotional caregiving activities Source: Bornstein and Putnick 2012. Note: The bar graphs show the number of caregiving activities reported by mothers in the past three days, based on comparable data from 25 developing countries ranked by the United Nations Human Development Index (HDI). The three categories of cognitive caregiving activities measured were reading books, telling stories, and naming/counting/drawing with the child. EARLY CHILDHOOD DEVELOPMENT 105 in child outcomes discussed at the beginning of this by lower salivary cortisol, a marker for stress system chapter be bridged by improving the socioeconomic activity—than children of nonparticipating mothers conditions of poor parents in the first place? How far (Fernald and Gunnar 2009). Significantly, the impact can structural antipoverty and cash programs go, and of Oportunidades on children’s stress and cortisol is through which pathways? concentrated mainly among children of mothers with Participation in conditional cash transfers (CCTs) depressive symptoms. may enhance children’s cognitive skills. For instance, Alleviating poverty alone in the short term does not in Mexico, children in households exposed longer to a automatically translate into increased positive parent- sizable and sustained CCT had improved motor skills ing practices. Direct interventions may be needed, in and higher cognitive development outcomes than which parents learn about child development (to pro- controls (Fernald, Gertler, and Neufeld 2008, 2009). mote certain types of positive adult-child interactions), However, in Ecuador, experimental evidence of the receive support in changing their beliefs and behaviors impact of a CCT showed only modest effects on child (to maintain higher levels of positive parenting), and development outcomes among the poorest children gain the opportunity to practice behaviors that support (Paxson and Schady 2010). Experimental evidence in the development of these competencies (to hone the Nicaragua on the impact of CCTs also showed mod- skills to engage more effectively with children). How est improvements in language and socioemotional can these skills be fostered? outcomes, which persisted two years after the cash program ended (Macours, Schady, and Vakis 2012). The persistence of these behavioral changes suggests Alleviating poverty alone does not that the programs have operated through mechanisms that go beyond the increase in material resources. automatically improve parenting The nutrition and parenting components of the CCT programs that directly target children and parenting practices. Direct interventions may skills are likely to be important pathways because they directly enhance the children’s environment. be needed. Another key pathway mediating the effects of family economic circumstances on the development of chil- Changing mindsets, underlying belief dren’s stress system is the mental health and mental systems, and mental models of parents’ role bandwidth of mothers and other caregivers. The stress Numerous barriers prevent parents from engaging associated with economic hardship and adversity may more fully with infants and young children. Barriers increase emotional distress and depression. There is may be due to their lack of knowledge about child devel- now evidence that having a predictable and stable source opment or lack of awareness that verbal interaction of income reduces parents’ mental stress and, through with children is important. Parents might also implic- that channel, the likelihood of inconsistent and unpre- itly believe that intelligence is fixed and immutable, dictable parenting behavior (Blair 2010). Some research- which undermines the motivation to change. Parents ers suggest that it can also affect the mental ability and might be held back by mental models based on tradi- capacity for attention that parents have for engaging tional beliefs that some practices can be harmful to the with their children (Mullainathan and Shafir 2013). child or by a fear of ridicule for violating a social norm As such, antipoverty programs and social policies that against talking to infants. How can parenting interven- provide income security could have a powerful indirect tions break these mental models and shift awareness effect on child development by reducing key psycho- about certain types of interactions that are beneficial to logical stressors that prevent parents from attending to their children? and engaging positively with their children. A study in urban Mexico demonstrated that mater- Making salient the link between parental nal depression can interfere with mothers’ capacity behavior and the consequences for to provide supportive and responsive care (Fernald, child outcomes Burke, and Gunnar 2008). The CCT program in Mexico, Many parenting programs emphasize the importance Oportunidades, has been associated with significant of communication and play and aim to encourage care- reductions in symptoms of maternal depression, par- givers to adopt sensitive and responsive care practices tially explained by mothers’ lower stress levels (Ozer (rather than negative harsh parenting). Qualitative and others 2011). In turn, children of mothers who par- work from group parenting programs highlights the ticipated in the program had less stress—as evidenced importance of shifting core parental beliefs about 106 WORLD DEVELOPMENT REPORT 2015 parenting, becoming aware that change in the way The Renforcement des Pratiques Parentales in Senegal parents engage with their children is possible, and is a promising example of a parenting intervention establishing a link between the parents’ own behavior designed to promote change in social norms. Caregiv- and the child’s behavior. ers may come from disadvantaged backgrounds with little social support; the group intervention provides Framing the link between the intervention a setting in which they can discuss and share experi- and future child outcomes ences with local facilitators in community meetings. Can mental models and beliefs be altered by framing The resulting women’s support network may improve the desired behavioral change and practices in terms parental effectiveness that can be sustained after the of future benefits for the child? A program in Senegal, intervention ends. Group parenting programs aim to Renforcement des Pratiques Parentales, aims to help promote participants’ self-confidence as parents and parents understand their crucial role in providing their to connect them emotionally through discussions infants with early verbal engagement. The facilitators, with other parents facing similar problems. Through from the nongovernmental organization Tostan, share group dynamics, parents recognize their strengths as simple techniques to enrich interactions between individuals, while discussing strategies that help them parents and their young children, such as speaking solve daily problems and reduce stress and avoiding to them using a rich and complex vocabulary, asking harsh self-judgment. the children questions and helping them respond, playfully copying their children, telling them stories, Changing mindsets (mental models) and describing objects in detail to them. The first few through “brief” interventions sessions of the group activities with mothers and other One potentially interesting approach focuses on caregivers introduce the important link between ver- psychological processes that are levers of change, bal engagement, the development of the child’s brain, with the expectation that they can set in motion self- and the future benefits of greater intelligence and reinforcing practices that sustain change in the long other positive outcomes (figure 5.5). The link relies on term. The core principle of mindset interventions has the mother’s aspirations for her child. The hypothesis been documented by the large body of work of psychol- behind this approach is that the emotional engage- ogists Carol Dweck and David Yeager in the context ment of the mothers might enable them to reexamine of schooling in the United States. The objective is to their uncritically held assumptions and beliefs. train participants with customized messages encour- aging the mindset that certain types of abilities are Mobilizing communities to change malleable rather than fixed and hence can be fostered. social norms Such brief interventions—as opposed to intensive Support for parents may also promote the develop- or repeated ones—have been effective in motivating ment of network support groups (Kağıtçıbaşı 2007). students from middle school to college age to change both their beliefs and their study habits, in contexts Figure 5.5 A program in rural Senegal encourages in which achieving higher grades provides a clear parents to engage verbally with their children measureable outcome of academic progress (Yeager The drawing is based on a poster used by facilitators in a parent education and others 2013). Extending this approach to changing program in rural Senegal to describe how speech from the mother stimulates mindsets to caregiver education, by teaching parents the infant’s brain. The facilitators meet with parents and village elders twice a that their children’s intelligence is malleable rather month. than fixed, might have beneficial effects. However, it is not yet clear whether such brief interventions would also be successful in motivating and enabling parents to change multiple behaviors over time as the child grows older or to develop and practice the much wider range of skills essential for improvement in pos- itive parenting. Providing parents with the opportunity to learn and practice new skills and improve their mental health Providing information about the benefits of positive parenting strategies and changing mental models and Source: WDR 2015 team, based on program material from the nongovernmental organization Tostan. caregiving goals may be necessary but not sufficient EARLY CHILDHOOD DEVELOPMENT 107 steps to change parents’ behavior. Equally critical is Figure 5.6 Early childhood stimulation in Jamaica providing parents with the opportunity to learn and resulted in long-term improvements in earnings practice new skills for effective interaction. Parents may need to learn strategies to reduce negative forms A program in Jamaica sought to develop cognitive, language, and socioemotional skills in disadvantaged toddlers. The program of home visits of discipline and engage in sensitive and effective to mothers and their toddlers in Kingston targeted stunted children in poor forms of caregiving in a sustained and consistent way. communities. Over two years, community health aides held one-hour play Challenges to behavioral change may stem from the sessions using a curriculum that promoted high-quality interactions between difficulty of dealing with temperamental differences mother and child. Twenty years later, a follow-up study found that the two-year program of home visits to the toddlers improved long-term outcomes; it closed among children, negotiating change with other house- the earnings gaps between the disadvantaged children and a better-off group. hold members, or simply being unprepared to find (There is no statistically significant difference between the earnings of the solutions to continuous developmental challenges that stunted group that received the program and a nonstunted comparison group.) For these disadvantaged children, the program broke the intergenerational arise at different ages. transmission of poverty. reference group (those not stunted as toddlers) (%) Building skills incrementally 120 Average adult earnings relative to earnings of the “The acquisition of skills requires a regular environ- ment, an adequate opportunity to practice, and rapid and unequivocal feedback about the correctness of 100 thoughts and actions” (Kahneman 2011, 416). Home vis- iting programs do this by helping mothers build these skills incrementally, providing a structured curriculum 80 that allows mothers to learn strategies for coping with each new challenge and learn ways to promote the 60 cognitive, language, and socioemotional development of their children. A seminal study in Jamaica provided home stimulation intervention to stunted children 40 aged 9–24 months in low-income communities for two years (Grantham-McGregor and others 1991). The curriculum included detailed structured activities that 20 promoted high-quality interactions between mother and child through role-play and homemade toys used to demonstrate new skills. The frequency and contin- 0 ued contact allowed plenty of opportunities to practice Not stunted Stunted, with Stunted, no as a toddler mother-toddler intervention the newly acquired skills over time. The study tracked home visits the children for 20 years. The early stimulation com- Source: Based on Gertler and others 2014. ponent resulted in important long-term labor market effects for the participants, as shown in figure 5.6. self-esteem of both the child and the mother incre- The study shows how an intensive early psychosocial mentally over time. Behavioral economists highlight intervention can effectively improve the long-term lack of mental energy and cognitive capacity among outcomes of disadvantaged children by closing their low-income parents as a barrier to engagement. There education and earning gaps relative to a better-off is potential for experimenting with approaches that group and break the intergenerational transmission of incorporate insights from behavioral science to improve poverty (Gertler and others 2014). parental focus, memory, mindful attention, and time management (Mullainathan and Shafir 2013; Kalil 2014). Targeting parents’ own mental well-being— not just their behavior toward their children Using complementary classroom-based Given the central role of parents’ psychosocial well- interventions to support parental being in enabling them to be consistently responsive competence and positive in their interactions with children, pro- For many children, interventions that focus on the grams that directly support parents’ own regulation quality of caregiving may not be sufficient if they of affect, stress, and cognition are likely to be useful do not also address children’s problems in regulating complements to programs that target only children themselves. Integrating parent training into preschool (Blair and Raver 2012). The Jamaica home visiting interventions with multipronged interventions—such program sought not only to improve the interactions as the Incredible Years program offered to Head Start between mothers and their children but also to build the parents in the United States (Webster-Stratton 1998) 108 WORLD DEVELOPMENT REPORT 2015 and the Chicago Heights Early Childhood Center critical role that parenting plays in shaping the child’s (Fryer and others 2014)—are promising approaches for early environment. improving children’s social skills and their understand- Traditional interventions generally alleviate the ing of their emotions. However, these improvements scarcity of resources in households with young chil- seem to reduce conduct problems in young children dren, as well as the scarcity of information about the and foster executive function skills only for the more child’s development. Going beyond these traditional disadvantaged and high-risk children. This finding interventions, many of the most successful programs highlights the importance of tailoring the design and provide parents with the tools they need for optimal intensity of interventions to the needs of the target parent-child interactions. The programs train local population (Morris and others 2014). community members to give  parents psychosocial Teachers in preschool can also play an important support, with the aim of changing the habitual ways role by enhancing early positive investments made by that parents interact with their young children. The parents and compensating for early deficiencies. Like programs also aim to change the implicit theories of parents, the ability of teachers to promote a warm and child development held broadly within the commu- positive emotional climate in the classroom is critical nity, providing children and parents with the oppor- for helping children develop their noncognitive skills, tunity to learn and practice new skills for an effective as well as their cognitive abilities. An analysis of teach- parent-child interaction. Results from a small number ers and learning outcomes in Ecuador (Araujo and of high-quality studies have shown that such carefully others 2014) documents substantial effects of the qual- designed interventions can pay lifelong returns for ity of preschool teachers (and of teacher practices) on individuals born in poverty. More experimentation both math and language outcomes, as well as on exec- and testing are needed to tailor interventions to the sit- utive function outcomes. Programs that help teachers uations that parents experience, harnessing insights define rules and build skills to discipline students and from neurobiology and the behavioral sciences to scaffold self-regulation reduce children’s stress and understand and tackle the psychological and cultural anxiety, thus lessening the need for teachers to impose barriers to effective parenting that arise from the con- discipline. Classroom curricula such as the Tools of the texts in which individuals live. Mind (Bodrova and Leong 2007) and Montessori focus directly on enhancing self-regulation, with a strong Notes emphasis on social pretend play, taking turns, and the 1. Two well-known preschool programs in the United child’s own planning of activities. There is some evi- States targeting very disadvantaged children—the Perry dence that these approaches may be effective in Preschool and the Abecedarian Project—demonstrated improving children’s executive functions, with sus- the sizable effects of enriching early environments (see, tained effects on reading and vocabulary into the first for example, Cunha and others 2006). grade (Blair and Raver 2012). Programs that supple- Both programs were evaluated through random assignment and with assessments that followed the ment classroom curricula—such as Promoting Alterna- children into adulthood. The two programs showed tive Thinking Strategies (PATHS), used in the Head that adults who had participated as young children Start REDI program in the United States—teach teach- in these interventions had stronger noncognitive ers to build children’s understanding of emotions, skills than those in the control group, who had not competencies in self-control, and interpersonal prob- participated in the interventions. While the early IQ lem solving (Bierman and others 2008). gains that emerged for participants in both programs As children grow older and progress through school, had faded by middle childhood, gains in noncognitive the scope for promoting learning, creativity, flexibility, skills persisted and were associated with positive and discipline and for strengthening both cognitive outcomes in adulthood, such as higher earnings, and noncognitive skills will be increased by curricula more stable relationships, and less criminal activity. that promote socioemotional competence alongside Both programs targeted very disadvantaged children. A similar outcome was obtained in the Montreal pro- cognitive skills. gram (Algan and others 2013) discussed in chapter 3, which focused on fostering the noncognitive skills Conclusion and the levels of trust of seven- to nine-year-old boys Beginning in infancy, experience acts on important “at-risk” with behavioral problems. biological and cultural mechanisms that cause the tra- 2. The data on caregiving practices were derived from the jectories of cognitive and socioemotional skills of chil- Multiple Indicator Cluster Survey, a nationally repre- dren living in poverty to diverge very early in life from sentative and internationally comparable household those of better-off children. This chapter described the survey of developing countries that provides informa- EARLY CHILDHOOD DEVELOPMENT 109 tion on protective and risk factors for children’s health on Life Cycle Skill Formation.” In Handbook of the Eco- and development (UNICEF 2006). More data and more nomics of Education, Vol. 1, edited by Eric A. Hanushek studies on caregiving practices across wealth levels and Finis Welch, 697–812. Amsterdam: Elsevier. within countries, in addition to across countries, are Dixon, Suzanne, Edward Tronick, Constance Keefer, and needed. T. Berry Brazelton. 1981. “Mother-Infant Interaction among the Gusii of Kenya.” In Culture and Early Inter- References actions, edited by Tiffany M. Field. Florence, KY: Psy- Algan, Yann, Elisabeth Beasley, Frank Vitaro, and Richard chology Press. E. Tremblay. 2013. “The Long-Term Impact of Social Farkas, George, and Kurt Beron. 2004. “The Detailed Age Skills Training at School Entry: A Randomized Con- Trajectory of Oral Vocabulary Knowledge: Differences trolled Trial.” Working Paper, Sciences Po, Paris. by Class and Race.” Social Science Research 33 (3): 464–97. Araujo, Maria Caridad, Ricardo Carneiro, Yyannu Cruz- Fernald, A., V. A. Marchman, and A. Weisleder. 2013. “SES Aguayo, and Norbert Schady. 2014. “A Helping Hand? Differences in Language Processing Skill and Vocab- Teacher Quality and Learning Outcomes in Kindergar- ulary Are Evident at 18 Months.” Developmental Science ten.” Unpublished paper, Inter-American Development 16: 234–48. Bank, Washington, DC. Fernald, Lia C. H., Heather M. Burke, and Megan R. Bierman, Karen L., Robert L. Nix, Mark T. Greenberg, Gunnar. 2008. “Salivary Cortisol Levels in Children of Clancy Blair, and Celene E. Domitrovich. 2008. “Exec- Low-Income Women with High Depressive Symptom- utive Functions and School Readiness Intervention: atology.” Development and Psychopathology 20 (2): 423–36. Impact, Moderation, and Mediation in the Head Start Fernald, Lia C. H., Paul J. Gertler, and Lynnette M. Neufeld. REDI Program.” Development and Psychopathology 20 (3): 2008. “Role of Cash in Conditional Cash Transfer Pro- 821–43. grammes for Child Health, Growth, and Development: Blair, Clancy. 2010. “Stress and the Development of An Analysis of Mexico’s Oportunidades.” Lancet 371 Self-Regulation in Context.” Child Development Perspec- (9615): 828–37. tives 4 (3): 181–88. ————. 2009. “10-Year Effect of Oportunidades, Mexico’s Blair, Clancy, and Adele Diamond. 2008. “Biological Conditional Cash Transfer Programme, on Child Processes in Prevention and Intervention: The Pro- Growth, Cognition, Language, and Behaviour: A Longi- motion of Self-Regulation as a Means of Preventing tudinal Follow-up Study.” Lancet 374 (9706): 1997–2005. School Failure.” Development and Psychopathology 20 (3): Fernald, Lia C. H., and Megan R. Gunnar. 2009. 899–911. “Poverty-Alleviation Program Participation and Sal- Blair, C., D. A. Granger, M. Willoughby, W. R. Mills- ivary Cortisol in Very Low-Income Children.” Social Koonce, M. J. Cox, M. T. Greenberg, K. Kivlighan, C. Science and Medicine 68 (12): 2180–89. Fortunato, and the Family Life Project Investigators. Fernald, Lia C. H., Ann Weber, Emanuela Galasso, and 2011. “Salivary Cortisol Mediates Effects of Poverty and Lisy Ratsifandrihamanana. 2011. “Socioeconomic Gra- Parenting on Executive Functions in Early Childhood.” dients and Child Development in a Very Low-Income Child Development 82 (6): 1970–84. Population: Evidence from Madagascar.” Developmental Blair, Clancy, and C. Cybele Raver. 2012. “Child Develop- Science 14 (4): 832–47. ment in the Context of Adversity: Experiential Canal- Fryer, Roland, Steven Levitt, John List, and Anya Savikhin ization of Brain and Behavior.” American Psychologist 67 Samek. 2014. “Chicago Heights Early Childhood Cen- (4): 309–18. Bodrova, Elena, and Deborah J. Leong. 2007. Tools of the ter: Early Results from a Field Experiment on Temporal Mind. 2nd ed. Upper Saddle River, NJ: Pearson. Allocation of Schooling.” Presentation by Anya Samek Bornstein, Marc H., and Diane L. Putnick. 2012. “Cognitive to the Institute for Research on Poverty, University of and Socioemotional Caregiving in Developing Coun- Wisconsin–Madison. tries.” Child Development 83 (1): 46–61. Gertler, Paul, Susan M. Chang, James Heckman, Rodrigo Crockett, Molly J., and Johannes Haushofer. 2014. “How Pinto, Sally Grantham-McGregor, Arianna Zanolini, Poverty Gets under the Skull.” Background paper Christel Vermeersch, and Susan Walker. 2014. “Labor prepared for the World Development Report 2015, World Market Returns to an Early Childhood Stimulation Bank, Washington, DC. Intervention in Jamaica.” Science 344 (6187): 998–1001. Cunha, Flavio, and James Heckman. 2007. “The Technol- Grantham-McGregor, Sally M., Charles A. Powell, Susan ogy of Skill Formation.” American Economic Review 97 P. Walker, and John H. Himes. 1991. “Nutritional Sup- (2): 31–47. plementation, Psychosocial Stimulation, and Mental ————. 2009. “The Economics and Psychology of Inequal- Development of Stunted Children: The Jamaican ity and Human Development.” Journal of the European Study.” Lancet 338 (8758): 1–5. Economic Association 7 (2/3): 320–64. Hackman, D. A., & M. J. Farah. 2009. “Socioeconomic Cunha, Flavio, James J. Heckman, Lance Lochner, and Status and the Developing Brain.” Trends in Cognitive Dimitriy V. Masterov. 2006. “Interpreting the Evidence Sciences 13 (2): 65–73. 110 WORLD DEVELOPMENT REPORT 2015 Hart, Betty, and Todd R. Risley. 1995. Meaningful Differences Olson, Sheryl L., John E. Bates, and Kathryn Bayles. 1990. in the Everyday Experience of Young American Children. Bal- “Early Antecedents of Childhood Impulsivity: The Role timore: Paul H. Brookes Publishing. of Parent-Child Interaction, Cognitive Competence, Heckman, James J. 2008. “Schools, Skills, and Synapses.” and Temperament.” Journal of Abnormal Child Psychology Economic Inquiry 46 (3): 289–324. 18 (3): 317–34. Hoff, Erika. 2003. “The Specificity of Environmental Influ- Ozer, Emily J., Lia C. H. Fernald, Ann Weber, Emily P. ence: Socioeconomic Status Affects Early Vocabulary Flynn, and Tyler J. VanderWeele. 2011. “Does Alleviat- Development via Maternal Speech.” Child Development ing Poverty Affect Mothers’ Depressive Symptoms? A 74 (5): 1368–78. Quasi-Experimental Investigation of Mexico’s Opor- Hughes, Claire. 2011. “Changes and Challenges in 20 Years tunidades Programme.” International Journal of Epidemi- of Research into the Development of Executive Func- ology 40 (6): 1565–76. tions.” Infant and Child Development 20 (3): 251–71. Paxson, Christina, and Norbert Schady. 2010. “Does Kağıtçıbaşı, Cigdem. 2007. Family, Self and Human Develop- Money Matter? The Effects of Cash Transfers on Child ment across Cultures: Theory and Applications. New York: Health and Development in Rural Ecuador.” Economic Psychology Press. Development and Cultural Change 50 (1): 147–81. Kahneman, Daniel. 2011. Thinking, Fast and Slow. New York: Schady, Norbert, Jere R. Behrman, Maria Caridad Araujo, Farrar Straus & Giroux. Rodrigo Azuero, Raquel Bernal, David Bravo, Florencia Kalil, Ariel. 2014. “Proposal 2: Addressing the Parenting Lopez-Boo, Karen Macours, Daniela Marshall, and Divide to Promote Early Childhood Development for Christina Paxson. Forthcoming. “Wealth Gradients in Disadvantaged Children.” Hamilton Project. Washing- Early Childhood Cognitive Development in Five Latin ton, DC: Brookings Institution. American Countries.” Journal of Human Resources. Lansford, Jennifer E., and Kirby Deater-Deckard. 2012. Shonkoff, Jack P., Andrew S. Garner, Benjamin S. Siegel, “Childrearing Discipline and Violence in Developing Mary I. Dobbins, Marian F. Earls, Laura McGuinn, Countries.” Child Development 83 (1): 62–75. John Pascoe, and David L. Wood. 2012. “The Lifelong Luby, Joan L., Deanna M. Barch, Andy Belden, Michael Effects of Early Childhood Adversity and Toxic Stress.” S. Gaffrey, Rebecca Tillman, Casey Babb, Tomoyuki Pediatrics 129 (1): e232–e46. Nishino, Hideo Suzuki, and Kelly N. Botteron. 2012. UNICEF (United Nations Children’s Fund). 2006. The State “Maternal Support in Early Childhood Predicts Larger of the World’s Children 2007: Women and Children: The Dou- Hippocampal Volumes at School Age.” Proceedings of the ble Dividend of Gender Equality, Vol. 7. New York: UNICEF. National Academy of Sciences 109 (8): 2854–59. Ursache, Alexandra, Clancy Blair, and C. Cybele Raver. Macours, Karen, Norbert Schady, and Renos Vakis. 2012. 2012. “The Promotion of Self-Regulation as a Means of “Cash Transfers, Behavioral Changes, and Cognitive Enhancing School Readiness and Early Achievement Development in Early Childhood: Evidence from a in Children at Risk for School Failure.” Child Develop- Randomized Experiment.” American Economic Journal: ment Perspectives 6 (2): 122–28. Applied Economics 4 (2): 247–73. Webster-Stratton, Carolyn. 1998. “Preventing Conduct Morris, Pamela, Shira K. Mattera, Nina Castells, Michael Problems in Head Start Children: Strengthening Par- Bangser, Karen Bierman, and Cybele Raver. 2014. enting Competencies.” Journal of Consulting and Clinical Impact Findings from the Head Start CARES Demonstra- Psychology 66 (5): 715. tion: National Evaluation of Three Approaches to Improving Weisleder, Adriana, and Anne Fernald. 2013. “Talking Preschoolers’ Social and Emotional Competence. OPRE to Children Matters: Early Language Experience Report 2014–44. Washington, DC: Office of Planning, Strengthens Processing and Builds Vocabulary.” Psy- Research and Evaluation, Administration for Children chological Science 24 (11): 2143–52. and Families, U.S. Department of Health and Human Yeager, David S., Dave Paunesku, Gregory M. Walton, and Services. Carol S. Dweck. 2013. “How Can We Instill Productive Mullainathan, Sendhil, and Eldar Shafir. 2013. Scarcity: Why Mindsets at Scale? A Review of the Evidence and an Having Too Little Means So Much. New York: Times Books. Initial R&D Agenda.” White Paper prepared for the Noble, Kimberly G., Bruce D. McCandliss, and Martha J. White House meeting “Excellence in Education: The Farah. 2007. “Socioeconomic Gradients Predict Indi- Importance of Academic Mindsets,” Washington, DC, vidual Differences in Neurocognitive Abilities.” Devel- May 10. opmental Science 10 (4): 464–80. 6 CHAPTER Household finance Financial decisions are difficult. They typically involve The market can provide commitment devices and great uncertainty about the future, whether about other mechanisms to help people overcome these future income, cash (liquidity) needs, or interest rates. biases, but it can also exacerbate them. In general, Much has been learned in recent years about how indi- the market will have weak or missing incentives for viduals actually make these decisions. More often than resolving these problems when borrowers are naïve not, financial decision making is influenced by impul- about their biases or underestimate their lack of under- sive judgments, emotions, temptation, loss aversion, standing. Moreover, organizations may deliberately and procrastination. misinform or underinform their customers about the The research discussed in chapter 1 revealed sys- terms of the contracts they are signing. tematic biases in decision making: that is, systematic The consequences of these biases can be profound departures from what individuals intend to do and for people in poverty, or on the edge of poverty, because say that they want to do and what they actually do. All they lack a margin for error. And because countries these biases apply particularly to financial decision may not have the institutional capacity and the safety nets to safeguard individuals against financial losses, poor people need to be even more attentive to financial The consequences of biases in financial decisions (Mullainathan and Shafir 2009). Poverty also heightens uncertainty about future costs and benefits decision making can be profound for people of different actions, magnifying the individual’s focus on the pressing and current scarcity of resources—and in poverty, or on the edge of poverty, further complicating decision making for individuals who are often overwhelmed with numerous important because they lack a margin for error. day-to-day decisions (see chapter 4). Given these considerations, providing an appro- priate institutional setting—that is, access to well- making, the topic of this chapter. Many factors drive functioning financial markets and a sound regulatory these biases. People strive for simplification when environment—may not be enough to improve people’s confronted with difficult decisions (they tend to use decisions. In developing countries, more proactive shortcuts, or heuristics). The way financial products policies may be necessary to address the behavioral and tools are presented can shape their decisions constraints on financial decision making. For exam- (framing effects). Their preferences can be affected ple, providing access to a new insurance instrument by acute aversion to uncertainty (loss aversion) and may not be sufficient to induce people to use it if they ambiguity. Emotions and the desire for immediate perceive the product as ambiguous or do not trust the gratification (present bias) often win out against fore- institution issuing it. sight. Even when people try to make careful financial This chapter suggests ways that policy makers decisions, the complexity of the decision environment can make institutions more responsive to the behav- often leads them astray. ioral factors driving people’s financial decisions. This HOUSEHOLD FINANCE 113 chapter also discusses better ways to design and For both Suresh and Novi, the negative experience implement policy goals, such as increasing savings or of their immediate losses has more impact on their access to and reliance on formal sources of credit. The decisions than the positive effects of potential long- chapter presents examples of interventions that have term gains. As a result, they make choices that can be been shown to help address behavioral constraints on described as economically suboptimal. financial decisions. A large number of experimental studies on human decision making have demonstrated that people inter- The human decision maker pret the outcomes of financial prospects in terms of in finance gains and losses in comparison to a reference point, Are people rational in their financial decision making? such as the status quo, and subsequently put more This question divides economists, as shown by the weight on potential losses than on gains in their deci- different views laid out in the 2013 Nobel acceptance sions (Kahneman and Tversky 1979; Wakker 2010).2 lectures by Eugene Fama and Robert Shiller.1 This This leads people to shy away from investment oppor- section presents examples of financial conduct that tunities that are profitable over time, on average, but is typical all over the world but that cannot easily be that might expose them to a loss at any given time. explained under the assumption that people carefully The importance of losses in financial decisions can be consider all costs and benefits before making a deci- finely observed in data on portfolio holdings showing sion. These patterns of financial conduct, however, can that people invest too little in risky assets relative be explained by findings from psychology about how to the level dictated by traditional views on risk and people make decisions. return. Many people hold no risky investments at all The discussion that follows presents a series of (see the review chapter by Guiso and Sodini 2013). This insights using stylized examples of individuals in pattern can be explained by loss aversion and a myopic developing countries, followed by supporting empir- short-term focus on fluctuations (Benartzi and Thaler ical evidence for each phenomenon, and the policy 1995; Gneezy and Potters 1997). In volatile equity mar- implications implied. kets, even a one-year investment horizon (rather than observing daily ups and downs as in Novi’s example) Losses loom larger than gains might lead to significant losses, thus inducing inves- Suresh is a farmer in rural India who grows cash crops. tors to favor portfolios with minimal risk. The land he farms has been handed down from gener- Moreover, people are unwilling to sell investments ation to generation, and his family has an established that turned out poorly (see the review by Barber and history of growing and selling a well-known crop that Odean 2013). By holding on to these investments, they yields a modest and low-risk return. In the past few avoid actually realizing losses, hoping to break even years, Suresh has noticed other farmers selling a dif- after future price increases. In comparison, people are ferent crop that is much more profitable. However, the often too eager to realize gains. The pattern of holding new crop is critically dependent on rainfall and thus on to “losers” and selling “winners” violates basic prin- carries greater risk. Suresh’s cousin, an accountant in ciples of learning about the quality of the investments: the nearby city, confirms that it would be more prof- while gains signal potentially good investments, losses itable in the long run for Suresh to invest in the new signal poor ones. Returns would be higher in the long crop, so Suresh devotes a small part of his land to the run by disposing of poor investments and keeping the new crop as a trial. Unfortunately, drought hits the good ones, but many people do not follow this precept region the next year, and the new crop does not do so because they are so averse to realizing losses. well. Suresh takes this loss to heart and abandons the Evidence from six Latin American countries new variety. He forgoes the potential for more learning suggests that the tendency to overvalue losses and and higher growth. undervalue gains can lead to economically significant Novi lives in urban Jakarta, Indonesia, and decides welfare losses: in an experimental survey with real to invest in the stock market. She closely follows the monetary payments, the more strongly an investor value of her investments on a financial website and was affected by superficial (economically irrelevant) worries as the value of her investments fluctuates. gain-loss framing, the worse the investor scored on Although her gains outweigh her losses, she feels a broad index of economic well-being (Cardenas and much more concerned about the losses, and after Carpenter 2013). some time she withdraws most of her funds from Policies that increase risk tolerance in the presence the stock market. She keeps a few stocks that have of losses and reduce investment short-sightedness fallen significantly in value, hoping to sell them when may be beneficial. They should provide a frame in prices recover. which losses become less salient, and information on 114 WORLD DEVELOPMENT REPORT 2015 long-term benefits becomes more salient (Keys and Savings, investment, and insurance are important Schwartz 2007). For example, people making financial development goals, yet people often face daunting decisions could be provided with aggregate informa- obstacles in pursuing them even when suitable finan- tion on volatile outcomes over time or over a cross- cial products are available and individuals have dis- section of risks, which makes the short-term losses posable income: that is, even when the basic supply less “visible” than the long-term benefits (Gneezy and and demand conditions are met. A major tendency Potters 1997; Thaler and others 1997). identified by the behavioral finance literature that accounts for the underutilization of financial products Present bias: Overweighting the present is present bias. This leads decision makers to shift good Sonja is a school teacher in Kampala, Uganda, and has experiences (consumption) toward the present and participated in a savings scheme in her neighborhood bad experiences (making difficult decisions about that specified monthly contributions. She accepted how much to save) toward the future, leading to over- these contribution amounts without further thought. consumption and procrastination. It also implies that The school in which she works is now offering a sub- people might be patient when weighing one future sidized savings account at the local bank for all its payoff against another but become very impatient employees. Sonja must decide how much to save and when making similar choices involving the present. put in the bank account. When the accounts were This pattern can lead them to reverse their prefer- offered, Sonja resolved to make her savings decisions ences—even if they have planned them carefully—and in the next few weeks. After a year and a half, she has prevent them from successfully implementing their still not invested the time to decide. financial plans (Laibson 1997; O’Donoghue and Rabin 1999). Temptation is an extreme form of time inconsis- tency: people may value some goods or payoffs only at People have a tendency to frame financial the moment of consumption, or on impulse, but not in the context of the past or the future (Banerjee and decisions in a narrow way, rather than Mullainathan 2010). The empirical evidence suggests that behavior considering their overall financial situation. and decisions driven by impatience, procrastination, and temptation are economically relevant. A strik- ing empirical example of the coexistence of strong Linda faces a similar problem. She recently bought impatience and procrastination comes from a study a house in Johannesburg, South Africa, and is consid- of University of Chicago business students (Reuben, ering insuring her property; she would feel better if Sapienza, and Zingales 2007). Students received pay- her property and valuables were covered. When she ment for participating in a survey and could choose finds time to delve into the details of the insurance, between receiving the payment immediately after the she discovers that there are many different contracts. survey or receiving a much larger payment two weeks Comprehensive insurance also involves significant later. Many students chose the immediate payment, monthly costs, biting into her budget. For some insur- indicating strong impatience. However, many did not ance, she would have to provide documentation on her cash their checks until four weeks after the experi- valuables, which will require more time. She decides to ment. Some procrastinators waited as long as 30 weeks wait a bit and spend more time thinking about what to cash their checks. Those who initially indicated a she should do. strong preference for immediate payment were also Financial decisions require difficult trade-offs. more likely to delay cashing their checks. The finding Although people like Sonja would like to save and pro- can be interpreted as an intention-action divide. vide for their future, current consumption needs loom Impatience is strongly correlated at the individual large. They may procrastinate and postpone decisions, level with low saving and imprudent financial plan- losing time in which they could be accumulating sav- ning (Moffitt and others 2011; Sutter and others 2013). ings. Similarly, people like Linda value the benefits of The flip side of saving is borrowing. A particularly security and the long-term benefits of financial pru- expensive way to borrow is maintaining revolving dence, but when they begin the process of obtaining balances on credit cards.3 Costly credit card borrowing insurance, they lose sight of these general benefits has been shown to be related to time-inconsistent, and get discouraged by the costs, the large number of present-biased preferences (Meier and Sprenger 2010), choices, and the unattractive details they must comb suggesting that people do not plan to incur costly fees through. Hence, they may remain uninsured. but are stuck in a vicious behavioral cycle. HOUSEHOLD FINANCE 115 Time in psychological terms also has a dimension preferences for saving on their future selves (Ashraf, of “distance.” Psychology research has shown that peo- Karlan, and Yin 2006; Bauer, Chytilová, and Morduch ple construe decisions differently when considering 2012; Gal and McShane 2012). them in general terms for the long run (“high distance”) from when they are delving into the details to imple- Cognitive overload and narrow framing ment them now or shortly (“low distance”) (Trope Ikram is a small business owner in Tangier, Morocco, and Liberman 2003; Trope, Liberman, and Wakslak and has a long-standing relationship with a local 2007; Fiedler 2007; Liberman and Trope 2008). Low dis- microfinance provider, having borrowed and repaid tance implies a focus on concrete and subordinate fea- funds many times. He does not earn very much, and a tures (the details), feasibility, and cost, while high dis- recent unexpected illness has left him with health fees tance implies a focus on abstract and superordinate that he cannot pay out of pocket. He approaches his features (general aspects), desirability, and benefits. trusted microfinance provider for funds, who agrees Because insurance and saving are beneficial in the long to provide him a loan based on his clean credit record. term (that is, under high distance) but require immedi- However, during this very stressful period, Ikram ate decisions and immediate monetary costs (that is, unintentionally neglects some of his other financial under low distance), differences between planning and responsibilities. He does not pay his rent on time and actually doing are exacerbated for these important forgets to pay the electricity bill. His landlord, who can- financial decisions. not reach him because Ikram is getting treatment in The traditional tool of providing information may the hospital, initiates an eviction order. His electricity not help overcome these problems. People may simply is cut off for nonpayment of the bill. Unintended neg- avoid information that makes them anxious or uncom- ligence worsens the monetary burdens and anxiety of fortable. Policy measures that neglect these effects Ikram’s already tenuous situation. may backfire. For example, without complementary People have limited attentional and mental support or individualized counseling, informing peo- resources. Poverty leads to situations that impose a ple that their savings balances may be too low may not high cognitive tax so that these resources are used up be effective or might even be discouraging (Caplin and quickly; the resulting behavior leads to financial costs Leahy 2003; Carpena and others 2013). that add even more strain, possibly initiating a vicious The behavioral obstacles to financial decisions cycle of poverty (see chapter 4). discussed here are likely to have much larger detri- Willpower and attention are limited cognitive mental effects in low-income countries than in higher- resources. In times of acute scarcity, financial decisions income countries. Behavior and choices from one time place strong demands on these resources, using them period to another are influenced by the psychological up quickly. When cognitive resources are overtaxed, resource of willpower, which has been likened to a mus- decision quality typically suffers, as decisions are cle: it can be depleted by the exertion of free will and driven by emotional impulses and a narrow short-term requires time and resources to replenish (Baumeister focus (Baumeister, Vohs, and Tice 2007; Shah, Mullaina- and others 1998; Baumeister, Vohs, and Tice 2007). Sig- than, and Shafir 2012). Moreover, in such settings, small nificantly, from the perspective of development policy, situational factors such as an exasperating bus ride to the pressing demands of poverty can make it more a bank are often a compelling hindrance to implement- difficult for the poor to exert and replenish willpower ing prudent financial choices (Bertrand, Mullainathan, (Spears 2011),4 worsening the effects of time inconsis- and Shafir 2004; Mullainathan and Shafir 2009). tency and self-control. People also have a tendency to frame financial deci- While sophisticated financial products such as sions in a narrow way, rather than considering their automatic deposits to savings, mandatory retirement overall financial situation (Thaler 1990; Choi, Laibson, contributions, or default insurance programs are com- and Madrian 2009; Rabin and Weizsäcker 2009; Soman monplace in advanced economies, the poor in devel- and Ahn 2010; Hastings and Shapiro 2012). Narrow oping countries do not typically have access to such framing can lead individuals to compartmentalize instruments (Collins and others 2009). The resulting funds into mental categories. They may treat funds cash-based economy is highly susceptible to tempta- for food purchases as distinct from funds for school tion, procrastination, and other behavioral diversions fees, for instance, and neglect the overall financial to saving. This latter aspect provides a strong rationale situation. In Ikram’s example, it is conceivable that he for policy interventions, especially in developing coun- has put some funds aside for family events like a wed- tries, to provide specific institutions that help people ding; but because he mentally tagged these funds for a overcome willpower deficits and impose their current “wedding,” during his recent period of strain, he might 116 WORLD DEVELOPMENT REPORT 2015 not have considered using them to pay his health bill, to—the person being advised (the advisee), and how housing, or similar expenses unrelated to the tag. the nonexpert advisee uses this information and the According to a well-documented example in which advice to come to a decision. money is not treated as fully fungible, people often Disclosure requirements can have perverse effects have some low-interest savings, while at the same time on the products agents recommend, since agents could they are borrowing at much higher rates (Gross and shift their recommendations from those products for Souleles 2000, 2002; Stango and Zinman 2009). A holis- which disclosure has been made more stringent to tic view of their finances, though, would allow them to other products for which commissions remain opaque avoid high credit costs by using their savings to repay (Anagol, Cole, and Sarkar 2013). In particular, even their expensive loans. if firms are required to offer basic, affordable, and How people categorize funds depends on how transparent products, they may not provide sufficient and why they received them, on the social rules and information about them. Instead, they may offer more rituals directing their circulation, and on socially and opaque alternatives with hidden and complex fees and culturally supported mental models. For instance, life costs (Giné, Martinez Cuellar, and Mazer 2014). insurance in the United States was once considered Psychological research into advisers’ reactions to a gross breach of mental categories—human life was disclosure requirements shows that when conflicts incommensurable and sacred, and the monetary world of interest cannot be avoided (for example, because was profane. Over the course of the 19th century, life agents are paid based on commissions), then advisers insurance became acceptable, but only because life often give even more biased advice (Sah and Loewen- insurance itself was changed into a kind of sacred rit- stein 2013). This finding supports the importance of ual, when prudential planning became part of a “good having an institutional framework that allows for death” and the social basis for a new mental account independent, unbiased intermediaries in markets was established. The same was true of life insurance where financial advice is essential. for children, which was once viewed with great suspi- Even when the agent aims to provide the best cion, but eventually came to be a way to value the love advice possible for the customer, agents may mis- and affection children provided to families. More gen- judge the risk tolerance of their clients and recom- erally, Zelizer (2010, 100) notes that “mental accounting mend inappropriate products as a result. Judgments cannot be fully understood without a model of ‘socio- about other people’s attitudes toward risk are central to logical accounting.’ ” virtually all financial products and decisions. However, Providing individuals with a holistic view of their there is a well-documented tendency to judge people finances would be a useful policy goal in developing who are risk averse as less risk averse than they truly countries. In addition, timely reminders about upcom- are and people who are risk loving as less risk loving ing payments or savings can have substantial influ- than they are (Hsee and Weber 1997; Faro and Rotten- ence on improving financial outcomes, as discussed streich 2006). later in the policy solutions section of this chapter. The agent’s problem in assessing the risk prefer- ences of his or her client is compounded by framing The social psychology of the advice effects. Different formats for presenting risk typically relationship lead clients to reveal different attitudes toward risk. Victor is the sole provider for his family members in Which of these formats leads to the best decision, in Buenos Aires. He worries about what would happen the sense that it maximizes returns over time? Some to them if he was injured and could not work. He also studies have developed computerized simulation wants to save and invest for the future. He goes to a techniques that allow decision makers to “experience” branch of the local bank to meet an adviser, who offers the risk and volatility of different investments before him a range of life insurance and investment products. deciding which to choose (Goldstein, Johnson, and Victor does not have much understanding of or inter- Sharpe 2008; Donkers and others 2013; Kaufmann, est in financial issues, but he follows the advice of the Weber, and Haisley 2013). The evidence suggests that bank’s agent and buys a broad insurance product with these techniques lead to decisions that are most stable a conservative savings component. over time (and therefore to “buy and hold” strategies, Financial advice is offered by multiple people who increasing returns) (Kaufmann, Weber, and Haisley often have diverging incentives and differing informa- 2013). Even after experiencing a bad outcome, deci- tion. Structuring policy for financial advice therefore sion makers more often stick with their investment requires taking account of the possible self-interests strategies following a decision aided by a simulation of the agents who give advice, the content and quality technique than when they made a decision based on of the information that is collected from—and given other presentation formats. HOUSEHOLD FINANCE 117 These insights and the evidence suggest that the should be provided (in the case of a regulator). There measurement and communication of the clients’ risk are two important insights on framing interventions. tolerance, and the presentation of the financial prod- First, alternatives can be presented to financial deci- uct, are important considerations for financial agencies sion makers in various ways that address the biases in designing, implementing, and enforcing regulations. described earlier, without affecting the economic While these findings point in clear directions for the essence of the information. Second, financial products regulation of financial advice, at a deeper level it can be can be described either in simple and clear ways or in asked whether research provides a genuine rationale complex and opaque ways, with direct impacts on how for consumer protection in the advice relationship. decisions are made. One might predict that clients anticipate the motives Many studies have demonstrated the power of of self-interested agents and thus interpret the advice framing effects. A study on payday borrowers in the given by agents in light of their incentives. Empirical United States, for example, illustrates the effective- research has shown, however, that clients often fol- ness of framing in an experiment where repayments low advice blindly, literally shutting down their own were presented either in dollar amounts or as inter- thinking about the decision problem (Engelmann and est rates (figure 6.1) (Bertrand and Morse 2011). This others 2009). Clients may not understand, or perhaps very simple reframing of information significantly even perceive, the strategic aspects in the advice rela- discouraged costly repeat borrowing. The study makes tionship. Changes in disclosure rules on conflicts of an important point: an information format that seems interest do not change investors’ behavior in an experi- most informative and thus most useful from the per- mental agency setting (Ismayilov and Potters 2013; see spective of a financial professional or an economist is converging evidence in Sah and Loewenstein 2013). not necessarily suited to help nonexperts make good Careful regulation of financial advice therefore seems decisions. Interest rates can be confusing to decision warranted. makers and may mask the magnitude and frequency of repayment obligations. Similar effects have also Policies to improve the quality been observed for percentages versus frequencies, of household financial decisions especially when relating to conditional probabilities This section presents examples of several policies (Gigerenzer and others 2007). For example, the claim shown to improve financial decisions. It begins that “the number of successful investments increased with how choices are presented (framing) and then by 150 percent” conveys very different information describes several policies that actually change the from the claim that “the number of successful invest- choices that people are offered. ments increased from two in a thousand to five in a thousand.” Framing choices effectively Products or investments are typically presented to Decisions and financial outcomes often can be consumers in groups or categories. The categorization improved at virtually zero cost by choosing the can be arbitrary and can have strong effects on choices. description carefully (in the case of an institution that For example, when offered different investment cat- aims to help people make good financial decisions) egories, people sometimes tend to split investment or by stipulating requirements for how information amounts roughly equally across categories, irrespective Figure 6.1 Simplifying information can help reduce take-up of payday loans Simple changes in how repayment information is presented can have meaningful impacts on financial behavior. In this study, payday borrowers were provided with repayment in APR terms and in terms of dollar amounts. Presenting information in dollar amounts led to significant reductions in repeat borrowing from payday lenders. Repayments presented as: Standard APR in % terms Accumulated fees in $ terms 0 20 40 60 80 100 Loan take-up (%) Source: Bertrand and Morse 2011. Note: APR = annual percentage rate. 118 WORLD DEVELOPMENT REPORT 2015 of the nature of the categories. Thus when presented classical view that more information is always better with the two categories of stocks from North America than less information. (Canada and the United States) and South America (including Argentina, Brazil, Chile, Uruguay, and Changing the default República Bolivariana de Venezuela), individuals are One of the best-established findings in the behavioral likely to invest more in U.S. stocks than when presented finance literature concerns the power of defaults with the five categories of stocks from Argentina, (Madrian and Shea 2001). Defaults are ubiquitous in Brazil, Canada, Chile, the United States, Uruguay, and the administration of financial choices: newcomers República Bolivariana de Venezuela. These effects have to a job, for instance, are presented with a multitude been demonstrated in various studies, including those of forms requesting their choices on pension contri- with experienced managers and those with significant butions, health insurance plans, tax-favored savings stakes in market environments (Bardolet, Fox, and opportunities, and much more. Except in cases in Lovallo 2011; Sonnemann and others 2013). which legal restrictions make participation in certain Another example of effective framing is choice schemes mandatory, the natural default has long been simplification, particularly with respect to the number perceived to be no participation and no contribution; of alternatives presented. For most products, an agent, yet in some circumstances this default assumption adviser, or bank can present the decision maker with may not be the best policy. In many situations, positive only a limited set of alternatives. If people had unlim- contributions imply a higher net income discounted ited bandwidth, more information would always be at all reasonable market discount rates. This is partic- better for decision makers, assuming that they could ularly true for all schemes in which employers match freely choose the number of alternatives they want to contributions or the government provides favorable consider, given some search cost. In practice, however, tax treatment. A positive contribution default there- people are often overwhelmed by a large number of fore often means higher income; for those who have alternatives and end up postponing decisions or using strong reasons for significantly smaller, but immedi- simple heuristics or rules of thumb (Johnson and oth- ate payouts, it is typically sufficient to tick a box. ers 2012; Drexler, Fischer, and Schoar 2014). Reducing Various studies have demonstrated that nonpar- the number of alternatives can therefore be an effec- ticipation in highly profitable schemes is driven to tive intervention. It has been shown that procrastina- a large extent by procrastination and passivity. For tion is less severe as the choice set becomes smaller example, studies that examine the effects of a switch (Tversky and Shafir 1992). A study of consumer credit to automatic enrollment in 401(k) pension plans for in South Africa finds that more loans were made when employees of large U.S. firms find that both enroll- a smaller number of combinations of interest rates and ment and contribution amounts are strongly driven loan amounts were suggested to customers (Bertrand by the defaults provided by employers (Madrian and and others 2010). The effect of this simple framing Shea 2001; Beshears and others 2008). The effects of manipulation was equivalent to a 2.3 percent reduc- defaults can often be amplified when combined with tion in the loan interest rate. Similarly, in their current the framing interventions discussed above: reducing work, Giné, Martinez Cuellar, and Mazer (in progress) a complex choice of a retirement savings plan into a are finding a significant improvement in the ability simple binary choice between the status quo and a of respondents in Mexico to identify the optimal loan preselected default alternative dramatically increases and savings products when they were presented with participation in the plan (Beshears and others 2013). succinct summary information about savings rates People often find it easier to make decisions that and loan costs, as compared to a finer breakdown of require trade-offs between only future outcomes, as commissions, fees, and returns. discussed. Choosing between different savings rates From the perspective of regulation, it is important in the future does not involve the short-term focus to keep in mind that individuals are very sensitive to and immediate financial consequences of decisions the framing of alternatives and that there is typically for today. A clever intervention uses these insights to no “neutral” or “natural” frame: should eight differ- have people choose their own defaults for the future. In ent insurance products be offered or nine? Should this method, known as SMarT (Save More Tomorrow), they be presented in two categories or three? To this employees stipulate increases in savings out of future end, policy makers need to take into account the pay raises (Thaler and Benartzi 2004; Benartzi and behavioral consequences of different presentation Thaler 2013). No current payoffs need to be considered; formats and choose the format that maximizes con- no reductions in disposable income are experienced, sumer welfare. The evidence discussed here shows which could be perceived as losses and therefore that the optimal format will often deviate from the weigh heavily in the decision; future increases occur HOUSEHOLD FINANCE 119 Figure 6.2 Changing default choices can improve savings rates The Save More Tomorrow (SMarT) plan allows employees to allocate a percentage of future pay raises toward retirement savings. By committing to save more in the future through automatic payroll deductions, participants increased savings without sacrificing current disposable income. 13.6 14 11.6 12 Average savings rates (%) 10 9.4 8 6.5 6 6.1 6.3 6.2 6.1 5.9 4 3.5 2 0 Pre-advice First pay raise Second pay raise Third pay raise Fourth pay raise Participants who joined the SMarT plan Participants who declined the SMarT plan Source: Thaler and Benartzi 2004. automatically, by default, allowing savings to accumu- Serving clients who have few assets, MFIs extend late as long as the person remains passive (figure 6.2). noncollateralized loans to the poor. MFIs rely on One possible explanation for the effectiveness screening, monitoring, and contract enforcement of the changes in default is that it is easier to choose within borrower groups and generally have high repay- the default. To make a choice not to comply with the ment rates (Giné, Krishnaswamy, and Ponce 2011). option provided entails a cognitive cost: people must However, recent work draws attention to the influence stop and reflect and may need to determine their pref- of social factors on the high repayment rates. Ties of erences for options they had never considered before loyalty among group members due to social norms, as (Stutzer, Goette, and Zehnder 2011). There may also be well as the fear of the stigma of default, deter high risk an endorsement effect, where individuals interpret the taking and encourage repayment of group-based loans default as a form of advice coming from a knowledge- (Bauer, Chytilová, and Morduch 2012; Cassar, Crowley, able party (Madrian and Shea 2001; Atkinson and oth- and Wydick 2007). In contrast, exposure to informa- ers 2013). In both cases, a policy that is aimed at setting tion on defaults by unrelated people can reduce indi- defaults in a psychologically informed way will exert viduals’ propensity to repay (Guiso, Sapienza, and Zin- little influence on people with strong preferences. gales 2013), and solvent borrowers may have a higher However, these same defaults will have considerable inclination to adopt adverse behavior if they perceive influence on those people who would otherwise not that the lender is not financially strong (Trautmann ponder the decision carefully. and Vlahu 2013). These findings suggest that trust and confidence among group members, as well as views Making microfinance more effective of the lender, serve as an important foundation for Abundant evidence indicates that access to financial ser- successful microcredit lending and that the design of vices for households with limited income is an impor- information-sharing mechanisms may be guided with tant factor in reducing poverty and inequality (Karlan this insight in mind. and Morduch 2010; World Bank 2008; Imai and Azam 2012; Mullainathan and Shafir 2013). Furthermore, a Using nudges and reminders large body of evidence shows that by extending beyond A recurring insight from research on behavioral conventional reaches of markets, microfinance institu- finance is that simple interventions that account for tions (MFIs) enable the poor to smooth income shocks or remove psychological constraints, such as social (see the review in Armendáriz and Morduch 2010). nudges and reminders, can go a long way toward 120 WORLD DEVELOPMENT REPORT 2015 improving financial behavior. One aspect of human here highlight the potential role of simple and often behavior where reminders can be particularly effective inexpensive nudges that can help improve financial is overcoming lack of attention. behaviors. These nudges may even play on the behav- A series of experimental studies in Bolivia, Peru, ioral patterns and use them in smart ways. and the Philippines show that simple, timely text mes- sages reminding people to save improve savings rates Fighting temptation through commitment in line with earlier established goals (Karlan, Morten, Lack of self-control is a leading explanation for lack of and Zinman 2012). The studies find that reminders savings, and the absence of default savings plans for that emphasize a specific goal, such as saving for a pur- most people in developing countries makes the prob- chase of a consumer durable like a television, are twice lem worse. While individuals tend to put off important as effective as generic reminders; this finding suggests financial decisions to the future, often the same indi- that individuals treat money differently depending on vidual recognizes the importance of difficult financial the intended purpose and are more likely to be willing choices—as long as the decision point occurs in the to save for a specified purchase than more generally. future. Perhaps policy makers can design and offer Likewise, reminders about late fees on loans have been products that allow individuals to commit to certain shown to significantly improve timely repayment savings goals but do not allow them to renege without behaviors up to two years after the reminder (Stango significant penalty. and Zinman 2011). The most basic form of such commitment comes People’s tendency to mentally structure income from the experience of rotating savings and credit and spending in different accounts can be turned associations (ROSCAs). Such neighborhood savings into a tool for policy. In a recent study of employees in schemes are very popular in developing countries India, a simple nudge was used to establish different and allow people to invest in goods that require large up-front payments. The mechanism of ROSCAs cen- ters on the illiquid nature of contributions and funds. There is typically no “neutral” or “natural” Each ROSCA member contributes a fixed monthly amount to the central pot, and a randomly chosen frame. Policy makers need to take into individual gets the entire pot each month. By making saving a public act, these schemes exploit the value of account the behavioral consequences social pressure from other ROSCA members to commit them to their desired level of savings (Ardener and of different presentation formats and Burman 1996). This arrangement is similar to the group lending model in microfinance. Traditional choose the format that maximizes savings arrangements like ROSCAs may provide not only savings opportunities where access to financial consumer welfare. markets is missing but also a commitment device in circumstances in which the cultural or social envi- ronment makes individual implementation of a strict accounts for spending and savings among workers savings schedule difficult or impossible. with very low savings rates (Soman and Cheema 2011). Evidence from developing countries shows that Weekly salaries were artificially partitioned into two substantial demand for savings exists and that com- separate envelopes: one labeled “for consumption” and mitment devices are likely to have strong and positive another labeled “for saving.” Although there was no impacts on behavior. When savings accounts were binding restriction on spending from the “for saving” offered in the Philippines without the option of with- envelope, this simple manipulation led to an improve- drawal for six months, there was a large demand for ment in saving over the usual method of single lump- such accounts and a take-up rate of nearly 30 percent sum remuneration. (Ashraf, Karlan, and Yin 2006). After one year, individ- The policy lesson from these examples is clear: uals who had been offered and had used the accounts while policy makers may not be able to solve indi- increased savings by 82 percent more than a control viduals’ behavioral constraints, they can certainly group that was not offered such accounts. A recent recognize those constraints and design policy to study in Kenya finds that providing people with a account for them. The silver lining is that this need not lockable metal box, padlock, and passbook increased involve monumental changes in policy making or even investment in health products by 66–75 percent increases in budgets. Rather, the examples discussed (Dupas and Robinson 2013). HOUSEHOLD FINANCE 121 Figure 6.3 Commitment savings accounts can improve agricultural investment and profit Smallholder cash crop farmers in Malawi were provided formal savings accounts. A randomly selected number were also offered commitment accounts. Those with access to commitment accounts were able to save more for the planting season and generated larger farm profits from the next harvest. No access to 59.3 92.6 savings accounts Access to ordinary 67.7 93.7 savings accounts Access to commitment 74.8 112.0 savings accounts 0 25 50 75 100 125 150 175 200 Malawian kwacha (thousands per year) Total value of farming inputs Farm profit Source: Brune and others 2013. A study among farmers in Malawi randomized efits of simplicity by comparing the benefits of a full- access to ordinary savings accounts and commitment fledged financial education module to those of a mod- savings accounts. The results show higher demand for ule based on simple rules of thumb (Drexler, Fischer, commitment accounts and find suggestive evidence of and Schoar 2014). The simpler training yielded signifi- relatively larger welfare gains from such accounts in cant effects on knowledge and behavior, while the tra- the form of crop output and other farming outcomes, ditional financial education had only limited impact. as well as household expenditures (Brune and others These results suggest that financial education policy 2013). Figure 6.3 shows the expansion in the size of can be designed to highlight key heuristics, especially smallholder cash crop farms as farmers gain access to in poor populations that may have no prior financial commitment savings devices in a randomized evalua- training. tion. Although the experiment did not identify the pre- Another important psychological aspect of making cise channel of productivity improvements (resisting financial decisions is salience, or relevance. People are borrowing from social networks, enabling higher risk more likely to pay attention to financial education if taking with savings buffers, or committing against it is specifically targeted to their needs, rather than pure self-control problems), given the high take-up rate provided in general terms. In a study of microfinance and usage among local farmers, commitment mecha- clients in India, when researchers offered assistance nisms have the potential for increasing farm profits as in setting financial goals and individualized financial financial access is broadened. One concern, however, counseling, they found that both interventions led to about binding commitment devices is that they may, significant improvements in savings and budgeting at least initially, crowd out existing social or cultural behavior (Carpena and others 2013). In contrast, the mechanisms for the accumulation of resources. study found that financial education without the addi- tion of either goal setting or counseling had no impact Simplifying and targeting financial on informal or formal savings, opening bank accounts, education or purchasing financial products such as insurance. Increasingly financial education programs are becom- ing an integral part of development reform. Whereas Utilizing emotional persuasion earlier programs focused on providing basic knowl- People often make important choices based on emo- edge, more recent research also tries to remove psycho- tions rather than on careful thought. Economists and logical barriers to changing financial behavior. psychologists have long studied dual-process decision One of the most compelling findings in the realm of models in which decision making is essentially a pro- financial education is to keep it simple. Limited cogni- cess of negotiation between a “hot” and fast emotional tive and computational ability leads people to econo- system and a more deliberative and “cool” cognitive mize on cognition while making decisions (Datta and system (Metcalfe and Mischel 1999), or an interaction Mullainathan 2012). In a study of business owners between two systems of intuitive and deliberative in the Dominican Republic, researchers tested the ben- responses (fast and slow thinking) (Kahneman 2003). 122 WORLD DEVELOPMENT REPORT 2015 Previous studies clearly show that the internal nego- South Africa shows that television programming can tiation process can be influenced by external appeals. be harnessed to improve financial decisions, as well The most obvious example comes from the field of (Berg and Zia 2013). The authors studied the effects of advertising, which often relies on emotional appeals incorporating messages on debt management into a to attract customers. Such appeals often resonate more nationally televised and popular soap opera in South deeply than logical messages. If advertising can be Africa and found significant improvements in content- persuasive for commercial reasons, perhaps the power specific financial knowledge, greater likelihood of bor- of media can be used to influence welfare-enhancing rowing formally and for productive purposes, reduc- choices as well. One of the most widespread and influ- tion in borrowing through expensive shop credit, and ential media for conveying such messages is television. lower propensity to gamble—all messages that were conveyed in the soap opera story line (figure 6.4). The study employed a mixture of quantitative and qual- A psychologically informed understanding itative analytical tools to identify conformity to the messages delivered by the leading character. The study of decision making can help policy makers found that financial messages delivered by a peripheral character were largely ignored. This disparity in results improve the match between intended suggests that emotional connections are an important pathway for retention of educational messages and and actual effects of a financial policy that these connections can be built even in entertain- ment media aimed at large groups of consumers. and can help individuals achieve their Shaping intertemporal preferences at an financial goals. early age Habits and preferences formed in early life tend to stay As discussed in spotlight 2, entertainment pro- with people into adulthood and can have profound gramming on television that presents characters with effects on how they make socioeconomic decisions. A whom the audience can identify has been shown to compelling example is a long-term longitudinal experi- influence important social outcomes such as fertility ment conducted in the United States. Young children and demand for health screenings. A recent study in were invited into a room and offered a marshmallow to Figure 6.4 Popular media can improve financial decisions Financial education messages on debt management and gambling were incorporated into the story line of a two-month-long popular soap opera in South Africa. Providing messages in this way led to higher financial literacy and better financial decision making. Score on content-specific financial literacy test Borrowed money from a formal bank Borrowed for investment Household member used shop credit Household member has gambled money 0 10 20 30 40 50 Percentage Watched soap opera without financial messages Watched soap opera with financial messages Source: Berg and Zia 2013. HOUSEHOLD FINANCE 123 eat, but with one catch: if they resisted the temptation to financial decision making hard. Policy interventions eat the marshmallow right away and instead waited for to address these tendencies include changing default a few minutes, they would be rewarded with two marsh- options, using social networks in microfinance, mallows! Some kids resisted and some did not. (One lit- employing nudges and reminders, offering commit- tle girl licked the marshmallow and then quickly put it ment devices, simplifying financial education, and back—literally having her sweet and eating it, too.) The using emotional persuasion. The evidence shows that researchers tracked these children into adulthood and a psychologically informed understanding of decision found that the children who exhibited more patience making can help financial policy makers improve the and self-control achieved better educational and socio- match between intended and actual effects of a policy economic outcomes (Mischel, Shoda, and Rodriguez and can help individuals achieve their financial goals. 1989). Other research has verified that the ability to con- trol temptation and delay gratification among youth is Notes an important determinant of lifetime academic, eco- 1. See http://www.nobelprize.org/nobel_prizes/economic nomic, and social outcomes (Duckworth and Seligman -sciences/laureates/2013/. 2005; Moffitt and others 2011; Golsteyn, Grönqvist, and 2. For empirical estimates, see Abdellaoui, Bleichrodt, and Lindahl, forthcoming; Sutter and others 2013).5 Paraschiv (2007, table 1). While financial policy and products can be shaped 3. For evidence for the United States, see Ausubel (1991) to account for behavioral constraints among adults, a and Stango and Zinman (2009). complementary policy goal may be to try to improve 4. See also Shah, Mullainathan, and Shafir (2012). 5. See also the discussion in the section on overweighting such preferences in early life. A number of studies the present. have argued that willpower and self-control resemble a muscle that requires time and resources to replen- References ish and that becomes stronger with repeated practice Abdellaoui, Mohammed, Han Bleichrodt, and Corina (Baumeister and Heatherton 1996; Baumeister and Paraschiv. 2007. “Loss Aversion under Prospect Theory: others 1998; Muraven and Baumeister 2000; Baumeis- A Parameter-Free Measurement.” Management Science ter, Vohs, and Tice 2007). Indeed, research into devel- 53 (10): 1659–74. opmental psychology and education has shown that Anagol, Santosh, Shawn Cole, and Shayak Sarkar. cognitive control can be exercised and improved in 2013. “Understanding the Advice of Commissions- children through early life interventions in preschool Motivated Agents: Evidence from the Indian Life (see chapter 5). Training can improve their learning Insurance Market.” Working Paper, Wharton School capacity through such skills as absorbing, recalling, of Business, University of Pennsylvania, Philadelphia. Ardener, Shirley, and Sandra Burman. 1996. Money-Go- and applying concepts learned in class. Rounds: The Importance of ROSCAs for Women. New York: A recent study among high school students in Bra- Bloombury. zil shows that financial education that offers repeated Armendáriz, Beatriz, and Jonathan Morduch. 2010. The instruction and opportunities to practice responsible Economics of Microfinance. Cambridge, MA: MIT Press. intertemporal choices (such as saving for purchases Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying rather than buying on credit, comparison shopping, Odysseus to the Mast: Evidence from a Commitment negotiating prices, and keeping track of expenses) can Savings Product in the Philippines.” Quarterly Journal of have important influences on student financial pref- Economics 121 (2): 635–72. erences and outcomes (Bruhn and others 2013). The Atkinson, Jesse, Alain de Janvry, Craig McIntosh, and study, which involved nearly 900 schools and 20,000 Elisabeth Sadoulet. 2013. “Prompting Microfinance high school students, finds significant improvements Borrowers to Save: A Field Experiment from Guate- in student financial knowledge and attitudes, savings mala.” Economic Development and Cultural Change 62 (1): rates, and spending behavior and important improve- 21–64. Ausubel, Lawrence M. 1991. “The Failure of Competition ments in intertemporal financial preferences, as mea- in the Credit Card Market.” American Economic Review sured by indexes of financial autonomy and intention 81 (1): 50–81. to save. Banerjee, Abhijit V., Esther Duflo, Rachel Glennerster, and Cynthia Kinnan. 2013. “The Miracle of Microfinance? Conclusion Evidence from a Randomized Evaluation.” Working This chapter presents key insights into the social and Paper 1309, Massachusetts Institute of Technology, behavioral influences on financial decision making. Cambridge, MA. It shows that loss aversion, present bias, cognitive Banerjee, Abhijit, and Sendhil Mullainathan. 2010. “The overload, and the social psychology of advice make Shape of Temptation: Implications for the Economic 124 WORLD DEVELOPMENT REPORT 2015 Lives of the Poor.” Working Paper 15973, National in Rural Malawi.” Working Paper 5748, World Bank, Bureau of Economic Research, Cambridge, MA. Washington, DC. Barber, Brad, and Terrance Odean. 2013. “The Behavior Caplin, Andrew, and John Leahy. 2003. “Behavioral Policy.” of Individual Investors.” In Handbook of the Economics In The Psychology of Economic Decisions. Vol. I of Rational- of Finance, edited by Milton Harris, George M. Con- ity and Well-Being, edited by Isabelle Brocas and Juan D. stantinides, and Rene M. Stulz, vol. 2, part B, 1533–70. Carrillo, 73–87. New York: Oxford University Press. Amsterdam: Elsevier. Cardenas, Juan Camilo, and Jeffrey Carpenter. 2013. “Risk Bardolet, David, Craig R. Fox, and Dan Lovallo. 2011. “Cor- Attitudes and Economic Well-Being in Latin America.” porate Capital Allocation: A Behavioral Perspective.” Journal of Development Economics 103 (July): 52–61. Strategic Management Journal 32 (13): 1465–83. Carpena, Frank, Shawn Cole, Jeremy Shapiro, and Bilal Bauer, Michal, Julie Chytilová, and Jonathan Morduch. Zia. 2013. “The ABC’s of Financial Literacy: Experi- 2012. “Behavioral Foundations of Microcredit: Exper- mental Evidence on Attitudes, Behavior and Cognitive imental and Survey Evidence from Rural India.” Ameri- Biases.” Working Paper. can Economic Review 102 (2): 1118–39. Cassar, Alessandra, Luke Crowley, and Bruce Wydick. Baumeister, Roy F., Ellen Bratslavsky, Mark Muraven, and 2007. “The Effect of Social Capital on Group Loan Dianne M. Tice. 1998. “Ego Depletion: Is the Active Self Repayment: Evidence from Field Experiments.” Eco- a Limited Resource?” Journal of Personality and Social nomic Journal 117 (517): F85–F106. Psychology 74 (5): 1252–65. Choi, James J., David Laibson, and Brigitte C. Madrian. Baumeister, Roy F., and Todd F. Heatherton. 1996. 2009. “Mental Accounting in Portfolio Choice: Evi- “Self-Regulation Failure: An Overview.” Psychological dence from a Flypaper Effect.” American Economic Inquiry 7 (1): 1–15. Review 99 (5): 2085–95. doi: 10.1257/aer.99.5.2085. Baumeister, Roy F., Kathleen D. Vohs, and Dianne M. Tice. Collins, Daryl, Jonathan Murdoch, Stuart Rutherford, and Orlanda Ruthven. 2009. Portfolios of the Poor: How the 2007. “The Strength Model of Self-Control.” Current World’s Poor Live on $2 a Day. Princeton, NJ: Princeton Directions in Psychological Science 16 (6): 351–55. University Press. Benartzi, Shlomo, and Richard Thaler. 1995. “Myopic Loss Datta, Saugato, and Sendhil Mullainathan. 2012. “Behav- Aversion and the Equity Premium Puzzle.” Quarterly ioral Design: A New Approach to Development Policy.” Journal of Economics 110 (1): 73–92. Policy Paper 016, Center for Global Development, ——— —. 2013. “Behavioral Economics and the Retirement Washington, DC. Savings Crisis.” Science 339 (March 8): 1152–53. Donkers, Bas, Carlos Lourenço, Daniel Goldstein, and Ben- Berg, Gunhild, and Bilal Zia. 2013. “Harnessing Emotional edict Dellaert. 2013. “Building a Distribution Builder: Connections to Improve Financial Decisions: Evaluat- Design Considerations for Financial Investment and ing the Impact of Financial Education in Mainstream Pension Decisions.” Netspar Design Paper 20, Network Media.” Policy Research Working Paper 6407, World for Studies on Pensions, Aging, and Retirement, Til- Bank, Washington, DC. bery University, the Netherlands. Bertrand, Marianne, Dean Karlan, Sendhil Mullainathan, Drexler, Alejandro, Greg Fischer, and Antoinette Schoar. Eldar Shafir, and Jonathan Zinman. 2010. “What’s 2014. “Keeping It Simple: Financial Literacy and Rules Advertising Content Worth? Evidence from a Con- of Thumb.” American Economic Review 62 (2): 1–31. sumer Credit Marketing Field Experiment.” Quarterly Duckworth, Angela L., and Martin E. P. Seligman. 2005. Journal of Economics 125 (1): 263–306. “Self-Discipline Outdoes IQ in Predicting Academic Bertrand, Marianne, and Adair Morse. 2011. “Information Performance of Adolescents.” Psychological Science 16 Disclosure, Cognitive Biases, and Payday Borrowing.” (12): 939–44. Journal of Finance 66 (6): 1865–93. Dupas, Pascaline, and Jonathan Robinson. 2013. “Why Don’t Bertrand, Marianne, Sendhil Mullainathan, and Eldar the Poor Save More? Evidence from Health Savings Shafir. 2004. “A Behavioral-Economics View of Pov- Experiments.” American Economic Review 103 (4): 1138–71. erty.” American Economic Review 94 (2): 419–23. Engelmann, Jan B., C. Monica Capra, Charles Noussair, Beshears, John, James J. Choi, David Laibson, and Brigitte and Gregory S. Berns. 2009. “Expert Financial Advice C. Madrian. 2008. “How Are Preferences Revealed?” Neurobiologically ‘Offloads’ Financial Decision- Journal of Public Economics 92 (8): 1787–94. Making under Risk.” PLoS ONE 4 (3): 4957. ————. 2013. “Simplification and Saving.” Journal of Eco- Faro, David, and Yuval Rottenstreich. 2006. “Affect, Empa- nomic Behavior and Organization 95: 130–45. thy, and Regressive Mispredictions of Others’ Prefer- Bruhn, Miriam, Luciana de Souza Leao, Arianna Legovini, ences under Risk.” Management Science 52 (4): 529–41. Rogelio Marchetti, and Bilal Zia. 2013. “The Impact Fiedler, Klaus. 2007. “Construal Level Theory as an Inte- of High School Financial Education: Experimental grative Framework for Behavioral Decision-Making Evidence from Brazil.” Policy Research Working Paper Research and Consumer Psychology.” Journal of Con- 6723, World Bank, Washington, DC. sumer Psychology 17 (2): 101–106. Brune, Lasse, Xavier Giné, Jessica Goldberg, and Dean Gal, David, and Blakeley B. McShane. 2012. “Can Small Yang. 2013. “Commitments to Save: A Field Experiment Victories Help Win the War? Evidence from Consumer HOUSEHOLD FINANCE 125 Debt Management.” Journal of Marketing Research 49 (4): Johnson, Eric J., Suzanne B. Shu, Benedict G. C. Dellaert, 487–501. Craig Fox, Daniel G. Goldstein, Gerald Häubl, Rich- Gigerenzer, Gerd, Wolfgang Gaissmaier, Elke Kurz- ard P. Larrick, John W. Payne, Ellen Peters, and David Milcke, Lisa M. Schwartz, and Steven Woloshin. 2007. Schkade. 2012. “Beyond Nudges: Tools of a Choice “Helping Doctors and Patients Make Sense of Health Architecture.” Marketing Letters 23 (2): 487–504. Statistics.” Psychological Science in the Public Interest 8 (2): Kahneman, Daniel. 2003. “A Perspective on Judgment and 53–96. Choice: Mapping Bounded Rationality.” American Psy- Giné, Xavier, Cristina Martinez Cuellar, and Rafael chologist 58 (9): 697–720. Keenan Mazer. 2014. “Financial (Dis-)Information: Evi- Kahneman, Daniel, and Amos Tversky. 1979. “Prospect dence from an Audit Study in Mexico.” Policy Research Theory: An Analysis of Decision under Risk.” Economet- Working Paper 6902, World Bank, Washington, DC. rica 47 (2): 263–91. — —— —. In progress. “Disclosure and Demand Elasticity Karlan, Dean, and Jonathan Morduch. 2010. “Access to of Financial Products: Evidence from Mexico.” World Finance.” In Handbook of Development Economics, edited Bank, Washington, DC. by Dani Rodrik and Mark Rosenzweig, vol. 5, chap. 71, Giné, Xavier, Karuna Krishnaswamy, and Alejandro Ponce. 4703–84. Amsterdam: Elsevier. 2011. “Strategic Default in Joint Liability Groups: Evi- Karlan, Dean, Melanie Morten, and Jonathan Zinman. dence from a Natural Experiment in India.” Unpub- 2012. “A Personal Touch: Text Messaging for Loan lished manuscript. Repayment.” Working Paper 17952, National Bureau of Gneezy, Uri, and Jan Potters. 1997. “An Experiment on Risk Economic Research, Cambridge, MA. Taking and Evaluation Periods.” Quarterly Journal of Kaufmann, Christine, Martin Weber, and Emily Haisley. Economics 112 (2): 631–45. 2013. “The Role of Experience Sampling and Graphical Goldstein, Daniel G., Eric J. Johnson, and William F. Displays on One’s Investment Risk Appetite.” Manage- Sharpe. 2008. “Choosing Outcomes versus Choosing ment Science 59 (2): 323–40. Products: Consumer-Focused Retirement Investment Keys, Daniel J., and Barry Schwartz. 2007. “ ‘Leaky’ Ratio- Advice.” Journal of Consumer Research 35 (3): 440–56. nality: How Research on Behavioral Decision Making Golsteyn, Bart H. H., Hans Grönqvist, and Lena Lindahl. Challenges Normative Standards of Rationality.” Per- Forthcoming. “Adolescent Time Preferences Predict spectives on Psychological Science 2 (2): 162–80. Lifetime Outcomes.” Economic Journal. Laibson, David. 1997. “Golden Eggs and Hyperbolic Dis- Gross, David B., and Nicholas Souleles. 2000. “Consumer counting.” Quarterly Journal of Economics 112 (2): 443–78. Response to Changes in Credit Supply: Evidence from Liberman, Nira, and Yaacov Trope. 2008. “The Psychology Credit Card Data.” Unpublished manuscript, Univer- of Transcending the Here and Now.” Science 322 (5905): sity of Chicago and University of Pennsylvania. 1201–05. — —— —. 2002. “Do Liquidity Constraints and Interest Rates Madrian, Brigitte C., and Dennis F. Shea. 2001. “The Power Matter for Consumer Behavior? Evidence from Credit of Suggestion: Inertia in 401(k) Participation and Sav- Card Data.” Quarterly Journal of Economics 117 (1): 149–85. ings Behavior.” Quarterly Journal of Economics 116 (4): Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2013. “The 1149–87. Determinants of Attitudes toward Strategic Default on Meier, Stephan, and Charles Sprenger. 2010. “Present- Mortgages.” Journal of Finance 68 (4): 1473–1515. Biased Preferences and Credit Card Borrowing.” Ameri- Guiso, Luigi, and Paolo Sodini. 2013. “Household Finance: can Economic Journal 2 (1): 193–210. An Emerging Field.” In Handbook of the Economics of Metcalfe, Janet, and Walter Mischel. 1999. “A Hot/Cool- Finance, edited by Milton Harris, George M. Constan- System Analysis of Delay of Gratification: Dynamics tinides, and Rene M. Stulz, vol. 2, part B, 1349–1532. of Willpower.” Psychological Review 106 (1): 3. Amsterdam: Elsevier. Mischel, Walter, Yuichi Shoda, and Monica I. Rodriguez. Hastings, Justine, and Jesse M. Shapiro. 2012. “Mental 1989. “Delay of Gratification in Children.” Science 244 Accounting and Consumer Choice: Evidence from Com- (4907): 933–38. modity Price Shocks.” Working Paper 18248, National Moffitt, Terrie E., Louise Arseneault, Daniel Belsky, Nigel Bureau of Economic Research, Cambridge, MA. Dickson, Robert J. Hancox, Hona Lee Harrington, Hsee, Christopher K., and Elke U. Weber. 1997. “A Funda- Renate Houts, Richie Poulton, Brent W. Roberts, and mental Prediction Error: Self-Others Discrepancies Stephen Ross. 2011. “A Gradient of Childhood Self- in Risk Preference.” Journal of Experimental Psychology: Control Predicts Health, Wealth, and Public Safety.” General 126 (1): 45. Proceedings of the National Academy of Sciences 108 (7): Imai, Katsushi S., and M. D. Shafiul Azam. 2012. “Does 2693–98. Microfinance Reduce Poverty in Bangladesh? New Mullainathan, Sendhil, and Eldar Shafir. 2009. “Savings Evidence from Household Panel Data.” Journal of Policy and Decision-Making in Low-Income House- Development Studies 48 (5): 633–53. holds.” In Insufficient Funds: Savings, Assets, Credit, Ismayilov, Huseyn, and Jan Potters. 2013. “Disclosing Advi- and Banking among Low-Income Households, edited by sor’s Interests neither Hurts nor Helps.” Journal of Eco- Rebecca Blank and Michael Barr, 121–46. New York: nomic Behavior and Organization 93 (September): 314–20. Russell Sage Foundation. 126 WORLD DEVELOPMENT REPORT 2015 —— ——. 2013. Scarcity: Why Having Too Little Means So ————. 2011. “Fuzzy Math, Disclosure Regulation, and Much. New York: Times Books. Market Outcomes: Evidence from Truth-in-Lending Muraven, Mark, and Roy F. Baumeister. 2000. “Self- Reform.” Review of Financial Studies 24 (2): 506–34. Regulation and Depletion of Limited Resources: Does Stutzer, Alois, Lorenz Goette, and Michael Zehnder. 2011. Self-Control Resemble a Muscle?” Psychological Bulletin “Active Decisions and Prosocial Behaviour: A Field 126 (2): 247–59. Experiment on Blood Donation.” Economic Journal 121 O’Donoghue, Ted, and Matthew Rabin. 1999. “Doing It (556): F476–F93. Now or Later.” American Economic Review 89 (1): 103–124. Sutter, Matthias, Martin Kocher, Daniela Rützler, and Rabin, Matthew, and Georg Weizsäcker. 2009. “Narrow Stefan Trautmann. 2013. “Impatience and Uncertainty: Bracketing and Dominated Choices.” American Eco- Experimental Decisions Predict Adolescents’ Field nomic Review 99 (4): 1508–43. doi: 10.1257/aer.99.4.1508. Behavior.” American Economic Review 103 (1): 510–31. Reuben, Ernesto, Paola Sapienza, and Luigi Zingales. 2007. Thaler, Richard H. 1990. “Anomalies: Saving, Fungibility, “Procrastination and Impatience.” Working Paper and Mental Accounts.” Journal of Economic Perspectives 13713, National Bureau of Economic Research, Cam- 4 (1): 193–205. bridge, MA. Thaler, Richard H., and Shlomo Benartzi. 2004. “Save Sah, Sunita, and George Loewenstein. 2013. “Nothing to More Tomorrow™: Using Behavioral Economics to Declare: Mandatory and Voluntary Disclosure Leads Increase Employee Saving.” Journal of Political Economy Advisors to Avoid Conflicts of Interest.” Psychological 112 (S1): S164–S87. Science 25 (2): 575–84. Thaler, Richard H., Amos Tversky, Daniel Kahneman, and Shah, Anuj K., Sendhil Mullainathan, and Eldar Shafir. Alan Schwartz. 1997. “The Effect of Myopia and Loss 2012. “Some Consequences of Having Too Little.” Sci- Aversion on Risk Taking: An Experimental Test.” Quar- ence 338 (6107): 682–85. terly Journal of Economics 112 (2): 647–61. Soman, Dilip, and Hee-Kyung Ahn. 2010. “Framing, Men- Trautmann, Stefan T., and Razvan Vlahu. 2013. “Strategic tal Accounting and Individual Welfare.” In Perspectives Loan Defaults and Coordination: An Experimental on Framing, edited by Gideon Keren, 65–92. Hove, Analysis.” Journal of Banking and Finance 37 (3): 747–60. United Kingdom: Psychology Press. Trope, Yaacov, and Nira Liberman. 2003. “Temporal Con- Soman, Dilip, and Amar Cheema. 2011. “Earmarking strual.” Psychological Review 110 (3): 403–21. and Partitioning: Increasing Saving by Low-Income Trope, Yaacov, Nira Liberman, and Cheryl Wakslak. 2007. Households.” Journal of Marketing Research 48 (Special “Construal Levels and Psychological Distance: Effects Issue): S14–S22. on Representation, Prediction, Evaluation, and Behav- Sonnemann, Ulrich, Colin F. Camerer, Craig R. Fox, and ior.” Journal of Consumer Psychology 17 (2): 83–95. Thomas Langer. 2013. “How Psychological Framing Tversky, Amos, and Eldar Shafir. 1992. “Choice under Con- Affects Economic Market Prices in the Lab and Field.” flict: The Dynamics of Deferred Decision.” Psychological Proceedings of the National Academy of Sciences 110 (29): Science 3 (6): 358–61. 11779–84. Wakker, Peter P. 2010. Prospect Theory: For Risk and Ambigu- Spears, Dean E. 2011. “Economic Decision-Making in ity. Cambridge, U.K.: Cambridge University Press. Poverty Depletes Behavioral Control.” B. E. Journal of World Bank. 2008. Finance for All? Policies and Pitfalls in Economic Analysis and Policy 11 (1). http://riceinstitute.org Expanding Access. Policy Research Report. Washington, /wordpress/wp-content/uploads/downloads/2011/12 DC: World Bank. /depletion-published.pdf. Zelizer, Viviana A. 2010. Economic Lives: How Culture Shapes Stango, Victor, and Jonathan Zinman. 2009. “What Do the Economy. Princeton, NJ: Princeton University Press. Consumers Really Pay on Their Checking and Credit Card Accounts? Explicit, Implicit, and Avoidable Costs.” American Economic Review 99 (2): 424–29. 7 CHAPTER Productivity In 2008, a bank in Colombia realized that it faced loans in the beginning of the month increased by 18 a problem: loan officers across its branches were percent, with no changes in the total number of new postponing their registration of new clients and col- loans per month or credit quality. Loan officers earned lection of credit to the last two weeks of the month, the bonuses they had been missing earlier—increasing just before their monthly performance bonuses were their monthly earnings by 25 percent—and at the same calculated, even though they had weekly targets and time reported less stress (Cadena and others 2011). their monthly bonuses were reduced when they failed Why did these bank officers require weekly remind- to meet them. These practices made it harder to man- ers to earn more money? This chapter makes the case age cash flows and also added to the stress of the loan that a number of the cognitive, psychological, and officers. social barriers described in earlier chapters affect how much effort employees may exert on the job or how much entrepreneurs and farmers may invest in Understanding motivation and behavior new technologies. Increasing productivity is central to raising living standards, and productivity growth at work requires us not only to zoom in can arise either from augmenting the factors of pro- duction—human capital, physical capital, and technol- to examine cognitive and psychological ogy—or from making better use of existing factors. This chapter focuses on the latter. The productivity of barriers that individuals face and the labor tends to be low in both the agricultural and the nonagricultural sectors in low-income settings (Caselli frames that work environments create but 2005), as is the adoption of business and farming practices that have proven effective elsewhere (Bloom also to zoom out to examine the social and others 2010). Insufficient motivation in those who provide public services is also common in developing contexts in which work takes place. countries and has been well documented in the past decade, ranging from absenteeism of school teachers to negligence among doctors.1 The bank experimented with decreasing the time To increase worker motivation, employers in both between effort and rewards and with making the the private and the public sectors typically turn to rewards more salient. They gave loan officers small monetary incentives: performance pay, bonuses, or the weekly prizes like movie tickets or restaurant coupons threat of dismissal. Underlying these strategies is an if they met their goals in the first half of the month and assumption that effort responds primarily to these sent weekly reminders about targets. In the branches kinds of incentives. Similarly, to address the lack that implemented these changes, the sourcing of new of productive investment among entrepreneurs and PRODUCTIVITY 129 farmers, a policy maker may rely on subsidies (under what constitutes adequate effort from those working the assumption that careful cost-benefit calculations around them. underlie investment decisions) or training services (under the assumption that these workers lack infor- Overcoming procrastination mation about the benefits of a technology). In India, for example, data entry clerks are primarily While these assumptions may indeed capture paid weekly through a piece rate; they earn a small fee important relationships between monetary incentives for showing up and then an amount for every accurate and effort and between the distribution of returns and field (piece) they enter. This kind of contract, however, investment, recent evidence suggests additional diag- still failed to motivate some workers in a large data noses for these problems. As the chapters in this Report entry firm in the city of Mysore to exert as much have shown, other cognitive, psychological, and social effort as they would have liked. They tended to work barriers—sometimes more difficult to observe—could less hard until rewards or needs became more salient. also be interfering with the productivity of employees, Their output increased by 8 percent on paydays, for entrepreneurs, and farmers and could also shape the example, over that at the start of the week—an increase effectiveness of monetary incentives. Individuals may in productivity equivalent to a 24 percent increase in face challenges in translating their intentions to work the piece rate (Kaur, Kremer, and Mullainathan, forth- harder or to increase their investment into concrete coming). Output would also spike by 15 percent in the action. Because of the many competing demands week before major festivals, which typically entail on their attention, they may miss opportunities to higher expenditures. improve their productivity and earnings. Individuals While it could be the case that the workers preferred may seek meaning in their work and may care about to increase effort only at these times, their responses how their employers treat them. They may also care to survey questions suggested that they struggled to about what their peers are doing. translate their intentions to work harder into action. Understanding these aspects of motivation and More than three-quarters agreed with the statement, behavior requires us not only to zoom in to examine “Some days I don’t work as hard as I would like to.” the cognitive and psychological barriers that individ- Likewise, nearly three-quarters concurred that “it uals face and the frames that work environments may would be good if there were rules against being absent create but also to zoom out and examine the broader because it would help [me] come to work more often.” social contexts in which work takes place (see spot- In a field experiment, workers were offered an alter- light 4). This chapter reviews evidence on the role that native contract that could help them commit to higher various cognitive, psychological, and social factors effort; they could set their own target for the number may play in the effort employees exert on the job, in of accurate entries for the week. If they reached the recruitment, in the performance of small businesses, target, they would be paid their usual piece rate; if and in the adoption of technology in agriculture. It they did not reach their target, they would be paid a concludes with some general lessons that could be use- lower rate. ful in designing interventions to improve productivity. More than one-third chose this kind of commitment contract, even though it increased their risk of being Improving effort among paid less if they did not meet their own goal. Their employees output increased by 6 percent, an effect equivalent To maximize incentives for employee effort, an to increasing the usual piece rate by 18 percent. The employer may design a contract that ties pay to out- workers who showed the greatest tendency to increase put: when employees produce more, they earn more. effort just before payday were 50 percent more likely While recent evidence suggests that this is a useful to opt for the commitment contract, and they also starting point, sometimes these financial incentives increased their output by much more: 28 percent. are not sufficient. People may want to exert more While certainly this was a more cost-effective effort tomorrow than today, and this procrastination alternative to a blanket increase in wage rates, one can happen even in the presence of performance might wonder whether the effects of a commitment contracts, as the case of the Colombian bank officers contract would persist. For example, did they occur demonstrated. People are also sensitive to how tasks simply because the scheme was novel? Would workers are framed and how they understand their relationship who tended to procrastinate self-diagnose and choose with their employer—whether they are being treated the appropriate kind of contracts? Because these data fairly, for example. People may also take cues about entry clerks were paid weekly and the experiment took 130 WORLD DEVELOPMENT REPORT 2015 place over 13 months, they had ample time to learn new recruits and found that emphasizing their value about the scheme and adjust their behavior. Demand as individuals substantially decreased turnover and for these contracts persisted over time. Workers who improved satisfaction among company clients using tended to increase productivity closer to payday were their services (Cable, Gino, and Staats 2013). One more likely to take up the commitment contract over group received the standard orientation that focused time, suggesting that they could realize that they on skills training and general facts about the firm. required the additional motivation and could adopt an Another group received the same training plus an option that helped them improve their productivity. additional one-hour session in which they participated in self-reflection and group exercises that focused on Framing tasks and compensation their unique attributes that lead to personal happiness Ample evidence also suggests that employees’ produc- and high performance at work and encouraged them tivity can depend on how they perceive the value of to think of ways that they could replicate such behav- their work or how they perceive their treatment as an ior in their current job. During the training, they also employee, not simply on their financial compensation. wore sweatshirts and badges with their names printed That is, their productivity depends on the way their on them. For a third group, the additional one-hour ses- tasks and rewards are framed. Financial incentives sion focused on organizational identity. Senior workers may also function differently when the rewards of discussed the firm’s values and why it was successful. working are framed as lost opportunities versus poten- Workers were directed to reflect on and discuss quali- tial gains or when work environments are competitive. ties of the firm that made the workers feel proud; their sweatshirts and badges bore only the company’s name. The significance of tasks and the value After six months, the employees who had gone of employees through the standard orientation and the variant Most contracts are incomplete. They do not specify that stressed the organization’s identity had turnover every possible task a worker may be assigned, the rates that were 216 and 300 percent higher than that performance expectations for each task, or the impli- of the employees whose individual identities had been cations of every possible contingency on employment emphasized. While it could be the case that the best and compensation. The terms of such an explicit employees are the most likely to depart the firm (if, for contract would be difficult to verify and enforce, and example, they are highly sought after by other firms), the costs of monitoring worker performance could be this does not appear to explain these large differences prohibitively expensive. Moreover, workers may come in turnover: the clients of employees who had under- to the job with a certain amount of intrinsic motiva- gone achievement training were more satisfied than tion or inherent enjoyment or satisfaction from doing clients of employees who had not. a task that is not based on external rewards, which could obviate the need for explicit links between per- Reciprocity in the workplace formance and compensation. A number of field experiments also demonstrate that When employees first enter an organization, they rewards and compensation can alter how employees typically undergo some training or orientation, how- perceive they are being treated, which in turn can ever brief, to acquaint them with their new position. affect their productivity. These findings are consis- Some evidence from field experiments suggests that tent with models of jobs as a form of gift exchange, in the frames created during this stage of the employ- which workers reciprocate perceived acts of employer ment relationship can influence later productivity. generosity by increasing effort and punish treatment Emphasizing the significance of a task, for example, they consider to be unfair (Akerlof and Yellen 1990; motivated fundraisers for a university in the United Fehr, Kirchsteiger, and Riedl 1993). States. During their training, some fundraisers first In China, a consumer electronics company offered read inspirational stories about how their job could a one-time bonus equivalent to 20 percent of average make a difference in the lives of students who received weekly pay that was not tied to worker performance. scholarships, while others read stories about how the The bonuses improved hourly productivity by 3–5 skills they acquired through fundraising could help percent (Hossain and List 2012). The improvement their future careers (Grant 2008). The group that had lasted several weeks after the bonus was discontinued read the inspirational stories collected 69 percent more and was statistically indistinguishable from another donations while fundraising. incentive scheme they tried in parallel, in which work- In India, a division of a large software company ers could earn the same bonus only if they met certain experimented with multiple ways of orienting their production targets for four straight weeks. PRODUCTIVITY 131 Something similar happened in Tanzania. Health correctly entered 20 percent more items (Gilchrist, care workers who received a gift of a biography of Luca, and Malhotra 2013) (figure 7.1). an American doctor working in low-income settings Over time, employees may begin to think of an inscribed with a thank-you message from the research increase in their earnings as a permanent part of their team improved their adherence to medical protocols compensation—that is, they may rescale their expec- for many weeks afterward (Brock, Lange, and Leonard tations. Some evidence comes from an evaluation of a 2014). The process of providing the gift mattered and pay equalization reform in southern India that affected generated differential effects over time. When the gift tea plantations. Employees who pluck tea leaves were was given immediately and without conditions, it trig- typically paid a fixed daily wage and a piece rate after gered a large response within three weeks of receipt, surpassing certain output thresholds (Jayaraman, Ray, equivalent to 0.25 standard deviations in protocol and De Vericourt 2014). One month after unions and adherence. After 10 weeks, on average, however, per- tea plantations negotiated a contract revision that formance returned to the level of health workers who increased the daily wage by 30 percent to be in line with had received no gift. When the gift was made condi- minimum wages mandated by state legislation, output tional on observed performance, it triggered a smaller per worker increased by an average of 34–37 percent immediate improvement in protocol adherence—equal over that of the same plantations the year before and of to about 60 percent of the effect of the unconditional plantations whose contracts had not been revised. By gift—which also disappeared in the long run. the fourth month, however, this productivity improve- Both these methods were outperformed by one in ment had declined to 10 percent. which the book was promised but delivered later— which triggered both an immediate response at the Figure 7.1 Unexpected wage increases can time of the promise, equivalent to 64 percent of the trigger a productivity dividend effect of the unconditional gift—and a larger additional In an online experiment, data entry workers were offered three response when the gift was delivered, which persisted different wage rates. Two groups were offered $3 per hour or even one month later. After 10 weeks, these health care $4 per hour. A third group was offered $3 per hour, but after workers demonstrated protocol adherence that was as accepting the offer, group members were told they would actually be getting $4 per hour due to an unexpected increase high as the immediate effect of the unconditional gift. in budget. This last group correctly entered 20 percent more While these experiences from China and Tanzania items than the other groups. demonstrate that improvements in productivity in response to gifts can persist for several weeks, the 1,000 extent to which such effects persist is an empirical question and is likely to depend on a number of factors, Number of correctly entered items such as the nature of the employment relationship, the 950 type of task, and possibly the wages in the external labor market. Much more transient improvements in productivity in response to monetary gifts have been 900 observed among tree planters in Canada and tem- porary workers in the United States hired for several days for fundraising and data entry work in a library 850 (Bellemare and Shearer 2009; Gneezy and List 2006). It might be the unexpected nature of a gift that generates reciprocity. In an online experiment hiring 800 freelance data entry workers whose online profiles listed an asking wage below $3 per hour, workers faced one of three wage structures. One group was simply 750 hired at $3 per hour, while another was hired at $4 per hour. A third group was hired at $3 per hour but right before they started their work, the workers learned that they would earn $4 per hour due to an unexpected 700 increase in the employer’s budget (call this the $3+1 $3 per hour $4 per hour $3 per hour, group). At the end of the task, the $3 and $4 groups had wage wage + $1 more per hour in unexpected wages performed identically. Paying a higher wage did not generate higher productivity. The $3+1 group, however, Source: Gilchrist, Luca, and Malhotra 2013. 132 WORLD DEVELOPMENT REPORT 2015 If employees reward employers’ generosity with an In China, for example, an experiment in a high-tech increase in effort, to what extent is the converse true? manufacturing factory explored this tendency. Some Would they reduce effort in response to perceptions of workers were informed that they would receive a unfair treatment or compensation that deviates from bonus after their group’s output reached a certain their expectations? Some evidence from high-income target (the bonus was framed in terms of a gain). settings suggests that this response is possible. For Others were told that they would be given a bonus but nearly a 20-year period, when police officers in New that it would be rescinded if they failed to meet the Jersey did not receive the wage they requested in bind- target (the bonus was framed in terms of a loss). While ing arbitration, crime reports increased in the months both types of bonuses increased worker productivity, following arbitration, and arrest rates declined (Mas total productivity was 1 percent higher under the 2006). The greater the gap between their requested loss framing (Hossain and List 2012). While this may wage and what they received, the less effort the police seem like a small difference, it is important to note officers expended on the job. Similarly, workers in a tire that it resulted solely from a change in the wording factory in Illinois produced defective tires when they of the contract. had to make wage concessions and work alongside Would similar results occur outside a factory? strikebreakers, and the company’s tires were linked to In particular, in an application very important to more than 270 fatalities and 800 injuries (Krueger and low-income countries, could this reframing of awards Mas 2004). improve the performance of civil service workers like These deviations in expectations can have very health care workers or teachers, who in many places adverse effects in a public health system. In the United are not penalized with lower salaries or the threat of Kingdom, as in many countries, nurses’ wages in the dismissal for underperformance? public hospital system are set by centralized pay reg- A number of studies from low-income settings ulation. There is very limited regional variation. Thus, have revealed substantial increases in students’ test there are some regions and times when nurses’ pay scores or the quantity of health services in response may be close to the wage prevailing in the local private to standard performance pay bonuses framed as gains sector market, but sometimes they diverge. According (Glewwe, Ilias, and Kremer 2010; Muralidharan and to an analysis of nine years of data from the public hos- Sundararaman 2011; Basinga and others 2011). In the pital system, in regions where the nurses earned much United States, in low-income neighborhoods near less than the wage that prevailed in the external labor Chicago, an alternative loss-framed variant generated market, a 10 percent increase in the outside wage was improvements where the standard gain-framed bonus associated with a 15 percent increase in the fatality rate had proven unsuccessful (Fryer and others 2012). for patients admitted for heart attacks (Propper and Some teachers in these Chicago schools were offered Van Reenen 2010). In contrast, in regions where there the standard bonus at the end of the school year; the was only a small pay differential between the central- bonus would be determined by the test score gains ized wage and outside wages, changes in the outside their students achieved. Another group of teachers wage did not affect patient survival. was given the amount that administrators expected to The upshot of all of this is not to institute a policy be the average bonus ($4,000) at the beginning of the of gift giving in the workplace or to regulate increas- school year. If their students’ performance turned out ingly higher wages. Rather, this evidence suggests that to be above average, they would receive an additional workers’ effort is sensitive to their expectations of how payment at the end of the school year. If it was below they should be compensated and that it is possible to average, however, they would have to return the differ- improve productivity at least temporarily by exceeding ence between what they received in the beginning and these expectations. In some settings, a one-off surge in the final bonus they should have received. output may be required—such as one in tandem with This loss-frame manipulation really mattered. a public health campaign or during a particularly busy Math scores of students taught by teachers who faced time due to business cycle effects. Exceeding worker loss-framed bonuses were 0.2–0.4 standard deviations expectations during times like these could have big higher than the scores of students of teachers paid payoffs in productivity. their regular salaries without any kind of bonus. Loss versus gain frames Competitive work environments As previous chapters have discussed, people some- Recent field experiments also suggest that the orga- times put more weight on potential losses than on nization of the workplace—particularly whether it potential gains. This tendency can also affect people’s is competitive—may have an independent effect on level of effort in response to monetary incentives. productivity. People often do not work in isolation and PRODUCTIVITY 133 tend to compare themselves to others doing similar Figure 7.2 Public recognition can improve work, which can be a powerful way to motivate or performance more than financial demotivate people. Consider rankings or social com- incentives can parisons, where employees learn about their relative performance in a firm or organization. If individuals A public health campaign in Zambia experimented with different ways of motivating hairdressers to distribute thrive on competition, they may exert more effort. female condoms to their clients. Some hairdressers worked Or, somewhat perversely, they may decrease effort if as volunteers, while others received either a 10 percent or a they believe that they have relatively high ability but 90 percent commission for every packet of condoms sold. A do not want to see it tested empirically. Decreasing fourth group received a star that was displayed on a poster in their salon for each packet sold. This last group sold twice as effort allows them to maintain their self-image and many condoms as the other groups. tell themselves that the reason for their relatively poor performance was that they were not really try- 15 ing. Existing empirical evidence is consistent with both possibilities, which underscores the importance Number of condoms sold of experimentation and adaptation to local contexts 12 (chapter 11). Once a firm in Germany began to include employ- ees’ ranks in the distribution of productivity on their 9 paychecks, productivity increased by 7 percent, even though the firm did not use these rankings to adjust wages (Blanes i Vidal and Nossol 2011). Similarly, when a small retail chain in the Netherlands organized tour- 6 naments in which groups of stores competed against one another to achieve the highest sales growth, sales growth increased, regardless of whether winners of 3 the tournament earned any monetary rewards (Delf- gaauw and others 2013). In Zambia, recognition proved to be more effective 0 than performance pay among hairdressers tasked by No reward Financial Large Recognition a public health organization with selling female con- reward financial doms to their clients. Hairdressers who earned a star reward for every packet of condoms sold, which was stuck on Source: Ashraf, Bandiera, and Jack, forthcoming. a poster in their salon, sold more than twice as many condoms as hairdressers who received commissions. Considering social relations in the This impact was strengthened as the number of workplace other salons in the neighborhood also earning stars Peers in the workplace can also exert a strong influ- increased. Meanwhile, hairdressers who received a 90 ence on an individual’s effort by enforcing social percent commission on each condom did not sell more norms, whether that enforcement is intentional or not. condoms, on average, than those who earned nothing If coworkers see others slacking off, they may do the and essentially sold the condoms as volunteers (Ashraf, same, even if this means their earnings may decrease; Bandiera, and Jack, forthcoming) (figure 7.2). conversely, people may work harder if others are work- In another field experiment in Zambia, however, ing harder. This could have implications for how teams introducing a competitive element into training back- should be formed. fired among trainees preparing to work as community The experiment in India with data entry clerks, for health workers. When they learned that their relative example, suggests that peers may help bridge the gap rankings from exam scores would be revealed, their between intentions and actions. Even though their exam performance dropped by more than a third of a earnings depended solely on their own output, when standard deviation (Ashraf, Bandiera, and Lee 2014a)— employees were assigned seats near colleagues who an effect that was more pronounced among trainees displayed above-average productivity, their own output with previously low test scores. Similarly, a firm in increased by 5 percent (Kaur, Kremer, and Mullainathan the United States found that removing feedback on 2010), mainly because they increased their work hours, employee rankings among their furniture sales staff rather than their efficiency. When seated next to actually increased its sales performance by 11 percent above-average peers, these workers were also less likely (Barankay 2012). to opt for the commitment contract described earlier. 134 WORLD DEVELOPMENT REPORT 2015 Proximity to more productive workers can also lead publicly minded job applicants. In 2011, the federal to increases in efficiency. Cashiers in a national super- government of Mexico began a program to increase market chain in the United States, for example, were the presence of the state in marginalized and conflict- compensated primarily through a fixed wage that was affected communities through community develop- not sensitive to their productivity (Mas and Moretti ment agents who could identify the needs of the com- 2009). When they worked on a shift with a worker who munity and report directly to the federal government. was more productive, however, their own productivity The government experimented with the monthly wage improved. This improvement in productivity occurred offers used to recruit agents. In some areas, it offered only among cashiers when they could see the more 3,750 pesos, while in others, it offered 5,000 pesos productive worker, and the effect declined with dis- (corresponding to the 65th and 80th percentiles of the tance. Thus cashiers were truly calibrating their effort wage distributions in program areas, respectively). to what they could see around them. Less productive The higher wage offer attracted applicants who workers did not exert a similarly negative effect, so were more qualified (Dal Bó, Finan, and Rossi 2013). the supermarket could have sold the same number of Their previous earnings were 22 percent higher, they items in fewer hours if it had rearranged its shifts in were more than 50 percent more likely to be employed a way that maximized skill diversity on a team at any at the time of application, and they were nearly 30 per- given time. cent more likely to have worked in a white-collar posi- This might not always be the case, however; some- tion in their previous job. They also scored higher on a times only certain peers matter for these kinds of cognitive test. This increase in qualifications, however, productivity spillovers. Despite being compensated did not come at the expense of prosocial motivation. through individual piece rates, farmworkers on a fruit The higher wage also attracted applicants with a higher farm in the United Kingdom picked more or less fruit inclination toward public service, as measured by a depending on the productivity of team members who standard public service motivation index. These appli- were their friends (Bandiera, Barankay, and Rasul cants, for example, found policy making more attrac- 2010). Compared to when they had no friends on their tive and reported a stronger belief in social justice. team, workers who were generally more productive In Zambia, researchers collaborated with the gov- than their friends picked less fruit and sacrificed ernment to test two methods of recruiting candidates around 10 percent of their earnings when assigned to for a new community health worker position. The sole teams with their friends; likewise, workers who were difference was whether the posters that advertised the less productive than their friends increased their earn- positions emphasized career benefits or social bene- ings by 10 percent when assigned to teams composed fits. In some districts, the posters called on applicants of their friends. to “become a highly trained member of Zambia’s health care system, interact with experts in medical fields, and Recruiting high-performance access future career opportunities including: clinical employees officer, nurse, and environmental health technologist.” If effort on the job can be influenced by the framing of In other districts, applicants were called to “learn about tasks and compensation and by social relations among the most important health issues in [their] community, employees and if employees themselves demonstrate gain the skills [they] need to prevent illness and pro- considerable heterogeneity, could these factors also mote health for [their] family and neighbors, work affect the types of employees that apply for a job at the closely with [their] local health post and health center, recruitment stage? For example, could high wages for and become a respected leader in [their] community.” work that has prosocial benefits, such as jobs in the As was the case in Mexico, emphasizing career- public sector, attract applicants who care solely about related incentives did not attract applicants with lower their own career advancement and who exhibit little to measures of social motivation (Ashraf, Bandiera, and no prosocial orientation? Lee 2014b). It did, however, attract more qualified can- A number of laboratory experiments suggest that didates as measured by their past academic achieve- financial incentives may crowd out intrinsic motiva- ment, and workers recruited through this method tion, or the inherent enjoyment or satisfaction from performed better once employed. Workers recruited doing a task that is not based on external rewards.2 through career incentives made 29 percent more visits Two recent field experiments, however, found that to households (for environmental inspections, health stressing financial incentives during recruitment counseling, and referring sick cases to health posts) drives for public sector positions did not attract less and organized 100 percent more community meetings. PRODUCTIVITY 135 They were also no more likely to leave their positions fluctuate due to both shocks such as illnesses and pre- than workers who had been recruited through mes- dictable expenses such as school fees (Dupas and Rob- sages that stressed the social benefits of the job. inson 2014). As a result, they forgo some 5–8 percent of their potential income. Fishermen in India also fish Improving the performance of less in response to recent increases in the value of their small businesses catches (Giné, Martinez-Bravo, and Vidal-Fernandez Many of the barriers that affect job performance 2010). among employees also affect decision making by the Owners of small businesses in Kenya also failed self-employed. Self-employment accounts for nearly to notice an opportunity to increase their business 60 percent of the world’s labor force,3 and even in income. These businesses are typically ventures such low-income countries, the self-employed account for as fruit and vegetable vending, retail shops, restaurants, one-third of the nonagricultural labor force (de Mel, tailoring shops, and barbershops, and their transactions McKenzie, and Woodruff 2010). Divides between take place almost entirely in cash. To complete their intentions and actions and the neglect of potential transactions, owners must be able to make change. This opportunities may loom even larger for the self- requires that they come to work each day with enough employed because they do not have contracts with an cash in small denominations. The majority of owners, employer interested in their level of effort or explicit work arrangements that dictate what is expected of them. The near absence of certain markets in many Divides between intentions and low-income settings—in particular the markets for insurance and credit—may also create narrower mar- actions and the neglect of potential gins for error for the self-employed. In Ghana, for example, a test between two different opportunities may loom even larger for methods of providing support to small-scale entrepre- neurs suggests that difficulties in translating inten- the self-employed because they do not tions into action could prevent them from making profitable investments. Entrepreneurs who received have explicit work arrangements that in-kind grants, which came in the form of business equipment, generated 24 percent more profits than dictate what is expected of them. those who received no support (Fafchamps and others 2014). Entrepreneurs who received support in the form of cash grants, however, did not increase their profits; however, report losing a sale in the previous week the grants ended up partially financing household because they did not have change readily available and needs and requests from relatives. The difference spending about an hour and a half searching for change was especially large for entrepreneurs who also had from nearby vendors (Beaman, Magruder, and Robin- difficulties in other areas, such as saving, that require son 2014). translating intention into actions. Pointing out the problem, even indirectly, did If losses also loom larger than gains for the self- improve things. Simply asking the owners about the employed, then individuals might be expected not only ways they managed their change once a week for two to avoid losses but also to neglect potential gains—and or three weeks led to a 32 percent reduction in the thus miss opportunities to increase earnings. There number of lost sales. Taking a few minutes to go over is evidence that taxi drivers and bike messengers a calculation of the lost profits attributable to poor in high-income settings like the United States and change management led to a similar reduction, which Switzerland work with target earnings or target hours translated into an increase in profits of 12 percent. in mind. They do not take advantage of temporary Even managers of larger firms may fail to notice increases in their compensation per ride or per mes- what seem to be obvious ways of improving produc- sage they could receive by working more. Instead, they tivity. Many large textile plants in India, for example, either reduce their hours or reduce their effort per had piles of garbage, tools, and other obstructions hour (Camerer and others 1997; Fehr and Goette 2007). that slowed the flow of workers on production floors This phenomenon occurs in low-income settings, and unlabeled and unsorted yarn inventories that as well. Bicycle taxi drivers in Kenya appear to work increased the probability of defects in quality (Bloom just enough to meet their daily cash needs, which and others 2013). Because their firms were profitable, 136 WORLD DEVELOPMENT REPORT 2015 many managers believed that they did not need a qual- their village leaders received the program. When the ity control process. village leaders also participated in the program, benefi- One might ask why these firms failed to notice ciaries’ income from nonagricultural self-employment these opportunities. Why are they not driven out of the increased by more than 160 percent and the value of market? While there is little empirical evidence that animal stock by 94 percent, while agricultural wages can address these questions, it is possible to speculate. went down by 60 percent. Social interactions also Many of these businesses may face little competition. increased, consistent with these impacts. Beneficiaries Or when choosing a small shop, customers may also of the business grant were more than four times as put less weight on prices and more on their relation- likely to report that they had talked to someone in the ship with the owner. It is also possible that managing community about their business (Macours and Vakis a business and making all production and sales deci- 2014). sions alone tax a person’s “bandwidth,” or cognitive resources, and capture attention that could otherwise Increasing technology adoption be directed toward improving the business. in agriculture While these failures to notice opportunities can be Macroeconomic and microeconomic data suggest that addressed directly with information or business train- differences in agricultural labor productivity across ing, the ideal programs would take the finite bandwidth countries are much larger than aggregate productivity of busy entrepreneurs into account. A program in the differences.4 One possible reason underlying these Dominican Republic, for example, offered an account- differences in agricultural productivity may be the ing curriculum based on rules of thumb that taught low adoption of simple technologies, such as the use basic heuristics, such as maintaining two different of fertilizer or reduced tillage planting techniques. In drawers, one for business and one for personal income, 2011, for example, farmers used an average of 13.2 kilo- and a system of IOU notes for any transfers across grams of fertilizer per hectare of arable land in Sub- the two drawers. This strategy was more successful Saharan Africa, compared to 118.3 in OECD (Organi- than a curriculum that taught the fundamentals of sation for Economic Co-operation and Development) accounting. Microentrepreneurs who received rule-of- member states (WDI database). thumb training improved the way they managed their Much of this underinvestment may be explained by finances—their sales during bad weeks improved by 30 the underdevelopment of certain markets, such as the percent—and they were 6 percent more likely to have markets for insurance or credit. In Ghana, for example, any personal savings. In contrast, a standard training an offer of insurance indexed to rainfall led farmers to package did not achieve any of these benefits (Drexler, apply chemicals that were 24 percent more expensive, Fischer, and Schoar 2014). and they also spent 14 percent more on land prepa- Even though entrepreneurs work primarily alone, ration (Karlan and others 2014). Nonetheless, just as it may also be possible to take advantage of their rela- factors other than financial incentives determine the tionships within their social networks when designing productivity of employees and the self-employed, the interventions aimed at increasing their productive expected distribution of returns to investment may be potential. In Nicaragua, for example, access to a busi- only one component that a farmer considers in decid- ness grant program was randomly allocated in such a ing whether to adopt a new technology. way that community leaders received the same pro- gram as beneficiaries in some villages, while in other Working around procrastination and villages, community leaders did not. The program scarcity of attention consisted of a $200 grant that was conditional on the One potentially important factor for farmers is the creation of a business development plan, technical need to translate intentions into action, since crop assistance, some follow-up visits by a professional, and cycles require specific investments at specific times. an invitation to participate in training workshops on Missing these timely investments could throw off business skills organized within the communities. farm income for an entire season. The grants did not generate any significant Certain fertilizers for maize, for example, need to improvements in income for beneficiaries whose be applied when the maize is knee-high, at the time of village leaders did not also participate in the pro- top dressing, which is roughly two months after plant- gram. However, beneficiaries experienced substantial ing and nearly four months after the harvest. When increases in income and began to rely much less on farmers apply fertilizer at this time, they can increase agriculture for their livelihoods when three or four of income by 11–17 percent, according to experimental PRODUCTIVITY 137 evidence from Western Province in Kenya (Duflo, was statistically indistinguishable from a 50 percent Kremer, and Robinson 2008). However, fewer than 30 subsidy offered later in the season when fertilizer was percent of farmers sampled in this area reported using needed. These results were not driven by free delivery. fertilizer as of 2009; they attributed their lack of use to When the NGO offered some farmers free delivery by a lack of money, even though they could buy fertilizer itself later in the season, fertilizer use did not improve in small quantities and apply it to only part of their significantly (Duflo, Kremer, and Robinson 2011). land at a time. If financial resources were indeed a key Moreover, the increase in fertilizer use disappeared in constraint, then one policy response would be to pro- subsequent seasons when the NGO stopped offering vide subsidies to lower the cost of fertilizer for farmers. the intervention, which suggests that farmers found it Lack of money, however, may not be the main difficult to commit on their own to purchasing fertil- barrier to fertilizer use. The problem could be the dif- izer early in the season when they had cash. ference between the timing of income at harvest and While these interventions suggest alternatives to the timing of fertilizer needs. Farm household income subsidies for increasing the adoption of productive typically fluctuates, increasing after harvest and technologies, the extent to which fertilizer decisions tapering off afterward, and that income must compete were suboptimal to begin with is not known. The with many other demands both inside and outside the demonstration trials indicated considerable variation household. Another obstacle could be the effort—both in farmers’ profits after they started applying fertil- monetary and cognitive—required to buy fertilizer. izer. If the farmers who were induced to purchase fer- Most farmers in the area would have had to walk for tilizer through the prepayment option are also likely 30 minutes to the nearest town center and, once there, to have trouble translating intentions into actions for decide what type and how much fertilizer to buy. other parts of the agricultural production cycle, such Recent interventions in this area experimented as for weeding, then the intervention may have served with ways of overcoming these types of obstacles (fig- only to increase purchases among a population that ure 7.3). When a nongovernmental organization (NGO) stands to gain the least from fertilizer. Nevertheless, offered free delivery and the opportunity to prepur- these results suggest that increasing take-up need chase fertilizer at the time of harvest, fertilizer adop- not require subsidies in all cases; paying attention to tion increased by 64 percent—an improvement that potential disconnects between the timing of income Figure 7.3 Altering the timing of purchases can be as effective as a subsidy for improving investment Farmers in a region in rural Kenya typically purchase fertilizer just before they apply it, not right after the harvest when they have the most cash in hand. Without any intervention, 26 percent of farmers purchase fertilizer. Providing free home delivery right after the harvest increases the amount of fertilizer purchased much more than free delivery provided just before fertilizer is to be applied. Its impact is equivalent to offering a 50 percent subsidy at the time of fertilizer application. 42.5% 26% 35.6% 40.2% Harvest Fertilizer application February March April May June July No intervention Free delivery 50% subsidy Source: Duflo, Kremer, and Robinson 2011. 138 WORLD DEVELOPMENT REPORT 2015 and the timing of uptake decisions could yield clues (figure 7.4). Even though farmers can experiment and for designing strategies that help make these deci- test their assumptions about the importance of certain sions easier. aspects of production, they must first notice that they The neglect of potential gains can also be especially are indeed making a decision. serious for farmers, who must always juggle multiple Seaweed farmers in Indonesia, for example, had tasks at any given time. Consider seaweed farming. no problem noticing that the spacing between pods While seaweed may be one of the simplest life forms— determined the amount of seaweed they could grow, an algae—farming it is quite complex. Farmers attach and they could accurately report the spacing on their strands (or pods) of seaweed to lines submerged in the own lines. They failed to notice, however, that the ocean. They must decide where to locate their plots, length of the pod also mattered; they did not even how long the lines should be, how far to space the lines, know the lengths of the pods that they used, even what kind of seaweed to use, the spacing between though farmers had an average of 18 years of experi- pods, the length of pods, how tightly to attach their ence and harvested multiple crop cycles per year and pods to the lines, and when to harvest the seaweed thus had plenty of opportunities for learning by doing Figure 7.4 Not noticing a decision can hurt productivity Seaweed farming entails many decisions (examples are presented in 1 through 9). Even experienced seaweed farmers in Indonesia overlooked a crucial factor in the growth of their crop—the length of the pods—until researchers presented the missing information in a highly salient and individualized way. r ate n of w io ect 4 Dir of ot ess f pl htn ment tiono ig T ach Loc a att 1 3 s pod tween e gb cin 2 Spa 9 eed fs eaw eo Typ 8 t ves to har 5 en r Wh ate fw ho Dept d f po ho Le ngt 6 g cin Spa lines 7 en we gth bet n e le Lin Source: Hanna, Mullainathan, and Schwartzstein 2014. PRODUCTIVITY 139 (Hanna, Mullainathan, and Schwartzstein 2014). Even improved awareness and adoption of new technolo- when randomized controlled trials on their own plots gies much more than similar activities implemented demonstrated the importance of both length and spac- solely through the government’s extension agents ing—at least for researchers analyzing the data—the (BenYishay and Mobarak 2014). farmers did not notice the relationship between length and yields simply from looking at their yields in the Using these insights in policy experimental plots. Only after researchers presented design them with data from the trials on their own plots that The evidence reviewed in this chapter suggests some explicitly pointed out the relationship between pod general lessons for diagnosing problems of productiv- size and revenues did farmers begin to change their ity and designing effective solutions. First, there are production method and vary the length of the pods. many nonremunerative aspects of work that influence the effort that employees exert on the job. The time lag Harnessing the power of social networks between effort and rewards, for example, may induce While one way of overcoming farmers’ failure to notice employees to procrastinate and concentrate their would be to provide individual farmers with person- effort only at certain times. Perceptions of generous alized data that make it difficult to ignore whichever or unfair treatment can lead employees to increase aspect of production they are neglecting, this may be or decrease their performance, as can ideas about the an expensive service to deliver. An alternative would value of a person’s work or the competitive nature of be to exploit the fact that farmers may look to their the work environment. Even when production does peers for information about what they should be doing. not directly depend on teamwork, peers can serve as Evidence from the adoption of high-yield varieties of an important reference group and can have an impact seeds during the Green Revolution in India suggests on an employee’s productivity. that friends and neighbors played a role in their adop- Changing many of these nonremunerative attrib- tion (Foster and Rosenzweig 1995; Munshi 2004). utes could be relatively inexpensive because they Similarly, pineapple growers in Ghana calibrate their do not affect employees’ financial compensation or fertilizer use to what others in their network are doing require any new technologies. Simply recognizing (Conley and Udry 2010). good performance, for example, would be virtually In Uganda, these types of links played a role in costless, as would emphasizing the meaning of a task diffusing information about growing cotton. A field or the importance of an employee in an organization. experiment tested two different methods for teaching female cotton farmers proper growing techniques. In some villages, both male and female cotton farm- Not only the content of the interventions, ers attended a standard training program. In other villages, the training program targeted women and but also the process of delivering them, is focused on social networking. Each woman was ran- domly assigned a partner whom she did not know important. Design matters greatly. before. They played games that gave rewards for remembering cotton farming facts, and they received pictures of their partners and a reminder to talk to Similarly, for the self-employed and those work- them throughout the season. Thus each female farmer ing in agriculture, factors beyond the returns to gained at least one new link in her social network. investment can affect the adoption of productivity- This additional link proved to be more valuable enhancing practices and technologies. Competing than what was offered during the standard training demands may make it difficult to save enough to make (Vasilaky and Leonard 2013). For farmers who were timely investments, and the absence of institutions not among the highest producers, yields increased by that could compensate for such tendencies, such as 98 kilograms per acre, an increase of over 60 percent, markets for credit or insurance, could worsen the while the standard training increased it by 67 kilo- impacts of these tendencies in low-income settings. grams per acre (42 percent). The networking interven- The sheer number of decisions that the self-employed tion also seems to have generated higher impacts for must make may increase the likelihood that they fail to the woman in the pair that had lower yields at the start. notice opportunities. Findings from a field experiment in Malawi A second lesson that emerges from field experiments mirrored these results. Agricultural extension activi- around the world is that not only the content of the ties implemented through incentivized peer farmers interventions, but also the process of delivering them, is 140 WORLD DEVELOPMENT REPORT 2015 important. While pay-for-performance contracts, subsi- on Incentives for Public Service Delivery.” Journal of dies, and training are promising instruments for tack- Public Economics. ling low productivity among employees, entrepreneurs, Ashraf, Nava, Oriana Bandiera, and Scott Lee. 2014a. and farmers, the design of these approaches matters “Awards Unbundled: Evidence from a Natural Field Experiment.” Journal of Economic Behavior and Organiza- greatly. Discounts for fertilizer in Kenya, for example, tion 100 (April): 44–63. were more effective in improving farmers’ purchases ————. 2014b. “Do-Gooders and Go-Getters: Career Incen- when they were delivered right after harvest, when tives, Selection, and Performance in Public Service farmers had more cash on hand, than months later at Delivery.” 2014. Working Paper, Harvard Business the time when the fertilizer was needed. In Malawi and School, Cambridge, MA. Uganda, information about new farming technologies Bandiera, Oriana, Iwan Barankay, and Imran Rasul. 2010. had greater impact when it came from peers than “Social Incentives in the Workplace.” Review of Economic through standard channels, such as extension agents. Studies 77 (2): 417–58. In the Dominican Republic, financial training was more Barankay, Iwan. 2012. “Rank Incentives: Evidence from a Randomized Workplace Experiment.” Working effective when converted into simple rules of thumb. Paper, Wharton School, University of Pennsylvania, Third, people are heterogeneous. Different groups Philadelphia. may be more or less affected by intention-action Basinga, Paulin, Paul J. Gertler, Agnes Soucat, and Jennifer divides and what their peers are doing, and the inter- Sturdy. 2011. “Effect on Maternal and Child Health Ser- pretation of tasks and rewards is likely to vary sub- vices in Rwanda of Payment to Primary Health-Care stantially from person to person, and even from task Providers for Performance: An Impact Evaluation.” to task. Close to one-third of data entry clerks in India Lancet 377 (9775): 1421–28. and maize farmers in Kenya responded to the commit- Beaman, Lori, Jeremy Magruder, and Jonathan Robinson. ment devices that were offered to them. The others 2014. “Minding Small Change among Small Firms in perhaps required a different intervention. Kenya.” Journal of Development Economics 108: 69–86. Bellemare, Charles, and Bruce Shearer. 2009. “Gift Giv- This importance of both process—the small details ing and Worker Productivity: Evidence from a Firm- of implementing an intervention—and heterogeneity Level Experiment.” Games and Economic Behavior 67 (1): suggests that finding the most effective interventions 233–44. for a population will require an inherently experimen- BenYishay, Ariel, and A. Mushfiq Mobarak. 2014. “Social tal approach, including testing multiple approaches at Learning and Communication.” Working Paper 20139, the same time or in sequence (chapter 11). The low costs National Bureau of Economic Research, Cambridge, MA. of some of these new designs and the potential for high Blanes i Vidal, Jordi, and Mareike Nossol. 2011. “Tour- payoffs to otherwise difficult or intractable problems, naments without Prizes: Evidence from Personnel however, should justify the experimentation required Records.” Management Science 57 (10): 1721–36. Bloom, Nicholas, Benn Eifert, Aprajit Mahajan, David to find out. McKenzie, and John Roberts. 2013. “Does Management Matter? Evidence from India.” Quarterly Journal of Eco- Notes nomics 128 (1): 1–51. 1. For teacher absenteeism, see Chaudhury and others Bloom, Nicholas, Aprajit Mahajan, David McKenzie, and 2006. For doctors’ negligence, see Leonard and Masatu John Roberts. 2010. “Why Do Firms in Developing 2005; Das and Hammer 2007; Das and others 2012. Countries Have Low Productivity?” American Economic 2. See, for example, Gneezy and Rustichini 2000; Heyman Review: Papers and Proceedings 100 (2): 619–23. and Ariely 2004. Brock, J. Michelle, Andreas Lange, and Kenneth L. 3. WDR 2015 team estimate based on the International Leonard. 2014. “Giving and Promising Gifts: Experi- Income Distribution Database (I2D2). mental Evidence on Reciprocity from the Field.” Work- 4. The labor productivity of agriculture in the 90th and ing Paper 165, European Bank for Reconstruction and 10th percentiles of countries, for example, differs by Development, London. a factor of 45–50, compared to a factor of 22 for total Cable, Daniel M., Francesca Gino, and Bradley R. Staats. labor productivity (Caselli 2005; Gollin, Lagakos, and 2013. “Breaking Them In or Eliciting Their Best? Waugh 2014). Reframing Socialization around Newcomers’ Authen- tic Self-Expression.” Administrative Science Quarterly 58 References (1): 1–36. Akerlof, George A., and Janet L. Yellen. 1990. “The Fair Cadena, Ximena, Antoinette Schoar, Alexandra Cristea, Wage-Effort Hypothesis and Unemployment.” Quar- and Héber M. Delgado-Medrano. 2011. “Fighting terly Journal of Economics 105 (2): 255–83. Procrastination in the Workplace: An Experiment.” Ashraf, Nava, Oriana Bandiera, and B. Kelsey Jack. Forth- Working Paper 16944, National Bureau of Economic coming. “No Margin, No Mission? A Field Experiment Research, Cambridge, MA. PRODUCTIVITY 141 Camerer, Colin, Linda Babcock, George Loewenstein, and Fehr, Ernst, and Lorenz Goette. 2007. “Do Workers Work Richard Thaler. 1997. “Labor Supply of New York City More if Wages Are High? Evidence from a Random- Cabdrivers: One Day at a Time.” Quarterly Journal of ized Field Experiment.” American Economic Review 97 (1): Economics 112 (2): 407–41. 298–317. Caselli, Francesco. 2005. “Chapter 9, Accounting for Fehr, Ernst, Georg Kirchsteiger, and Arno Riedl. 1993. Cross-Country Income Differences.” In Handbook of “Does Fairness Prevent Market Clearing? An Experi- Economic Growth, Vol. 1, Part A, edited by Aghion Phi- mental Investigation.” Quarterly Journal of Economics 108 lippe and N. Durlauf Steven, 679–741. Amsterdam: (2): 437–59. Elsevier. Foster, Andrew D., and Mark R. Rosenzweig. 1995. “Learn- Chaudhury, Nazmul, Jeffrey Hammer, Michael Kremer, ing by Doing and Learning from Others: Human Karthik Muralidharan, and F. Halsey Rogers. 2006. Capital and Technical Change in Agriculture.” Journal “Missing in Action: Teacher and Health Worker of Political Economy 103 (6): 1176–1209. Absence in Developing Countries.” Journal of Economic Fryer Jr., Roland G., Steven D. Levitt, John List, and Sally Perspectives 20 (1): 91–116. Sadoff. 2012. “Enhancing the Efficacy of Teacher Incen- Conley, Timothy G., and Christopher R. Udry. 2010. “Learn- tives through Loss Aversion: A Field Experiment.” ing about a New Technology: Pineapple in Ghana.” Working Paper 18237, National Bureau of Economic American Economic Review 100 (1): 35–69. Research, Cambridge, MA. Dal Bó, Ernesto, Frederico Finan, and Martín A. Rossi. Gilchrist, Duncan, Michael Luca, and Deepak Malhotra. 2013. “Strengthening State Capabilities: The Role of 2013. “When 3+1>4: Gift Structure and Reciprocity in Financial Incentives in the Call to Public Service.” the Field.” Harvard Business School Working Paper, Quarterly Journal of Economics 128 (3): 1169–1218. Cambridge, MA. Das, Jishnu, and Jeffrey Hammer. 2007. “Money for Noth- Giné, Xavier, M. Martinez-Bravo, and Marian Vidal- ing: The Dire Straits of Medical Practice in Delhi, Fernandez. 2010. “Intertemporal Substitution, Weekly India.” Journal of Development Economics 83 (1): 1–36. Target Earnings or Both? Evidence from Daily Labor Das, Jishnu, Alaka Holla, Veena Das, Manoj Mohanan, Supply of Southern Indian Fishermen.” Unpublished Diana Tabak, and Brian Chan. 2012. “In Urban and manuscript, World Bank. Rural India, a Standardized Patient Study Showed Low Glewwe, Paul, Nauman Ilias, and Michael Kremer. 2010. Levels of Provider Training and Huge Quality Gaps.” “Teacher Incentives.” American Economic Journal: Applied Health Affairs 31 (12): 2774–84. Economics 2 (3): 205–27. De Mel, Suresh, David McKenzie, and Christopher Wood- Gneezy, Uri, and John A. List. 2006. “Putting Behavioral ruff. 2010. “Who Are the Microenterprise Owners? Economics to Work: Testing for Gift Exchange in Evidence from Sri Lanka on Tokman Versus De Soto.” Labor Markets Using Field Experiments.” Econometrica In International Differences in Entrepreneurship, edited by 74 (5): 1365–84. Josh Lerner and Antoinette Schoar, 63–87. Chicago: Gneezy, Uri, and Aldo Rustichini. 2000. “Pay Enough or University of Chicago Press. Don’t Pay at All.” Quarterly Journal of Economics 115 (3): Delfgaauw, Josse, Robert Dur, Joeri Sol, and Willem 791–810. Verbeke. 2013. “Tournament Incentives in the Field: Gollin, Douglas, David Lagakos, and Michael E. Waugh. Gender Differences in the Workplace.” Journal of Labor 2014. “Agricultural Productivity Differences across Economics 31 (2): 305–26. Countries.” American Economic Review 104 (5): 165–70. Drexler, Alejandro, Greg Fischer, and Antoinette Schoar. Grant, Adam M. 2008. “Does Intrinsic Motivation Fuel 2014. “Keeping It Simple: Financial Literacy and Rules the Prosocial Fire? Motivational Synergy in Predicting of Thumb.” American Economic Journal: Applied Economics Persistence, Performance, and Productivity.” Journal of 6 (2): 1–31. Applied Psychology 93 (1): 48–58. Duflo, Esther, Michael Kremer, and Jonathan Robinson. Hanna, Rema, Sendhil Mullainathan, and Joshua 2008. “How High Are Rates of Return to Fertilizer? Schwartzstein. 2014. “Learning through Noticing: The- Evidence from Field Experiments in Kenya.” American ory and Experimental Evidence in Farming.” Quarterly Economic Review 98 (2): 482–88. Journal of Economics 129 (529): 1311–53. —— — —. 2011. “Nudging Farmers to Use Fertilizer: Theory Heyman, James, and Dan Ariely. 2004. “Effort for Payment: and Experimental Evidence from Kenya.” American Eco- A Tale of Two Markets.” Psychological Science 15 (11): nomic Review 101 (6): 2350–90. 787–93. Dupas, Pascaline, and Jonathan Robinson. 2014. “The Hossain, Tanjim, and John A. List. 2012. “The Behavioral- Daily Grind: Cash Needs, Labor Supply and Self- ist Visits the Factory: Increasing Productivity Using Control.” Unpublished paper. Simple Framing Manipulations.” Management Science Fafchamps, Marcel, David McKenzie, Simon Quinn, and 58 (12): 2151–67. Christopher Woodruff. 2014. “Microenterprise Growth Jayaraman, Rajshri, Debraj Ray, and Francis De Vericourt. and the Flypaper Effect: Evidence from a Randomized 2014. “Productivity Response to a Contract Change.” Experiment in Ghana.” Journal of Development Economics Working Paper 19849, National Bureau of Economic 106 (January): 211–26. Research, Cambridge, MA. 142 WORLD DEVELOPMENT REPORT 2015 Karlan, Dean, Robert Darko Osei, Isaac Osei-Akoto, and Mas, Alexandre, and Enrico Moretti. 2009. “Peers at Christopher Udry. 2014. “Agricultural Decisions after Work.” American Economic Review 99 (1): 112–45. Relaxing Credit and Risk Constraints.” Quarterly Journal Munshi, Kaivan. 2004. “Social Learning in a Heteroge- of Economics 129 (2): 597–652. neous Population: Technology Diffusion in the Indian Kaur, Supreet, Michael Kremer, and Sendhil Mullaina- Green Revolution.” Journal of Development Economics 73 than. 2010. “Self-Control and the Development of (1): 185–213. Work Arrangements.” American Economic Review 100 (2): Muralidharan, Karthik, and Venkatesh Sundararaman. 624–28. 2011. “Teacher Performance Pay: Experimental Evi- ——— —. Forthcoming. “Self-Control at Work.” Journal of dence from India.” Journal of Political Economy 119 (1): Political Economy. 39–77. Krueger, Alan B., and Alexandre Mas. 2004. “Strikes, Scabs Propper, Carol, and John Van Reenen. 2010. “Can Pay Reg- and Tread Separations: Labor Strife and the Produc- ulation Kill? Panel Data Evidence on the Effect of Labor tion of Defective Bridgestone/Firestone Tires.” Journal Markets on Hospital Performance.” Journal of Political of Political Economy 112 (2): 253–89. Economy 118 (2): 222–73. Leonard, Kenneth L., and Melkiory C. Masatu. 2005. “The Vasilaky, Kathryn, and Kenneth L. Leonard. 2013. “As Good Use of Direct Clinician Observation and Vignettes as the Networks They Keep? Improving Farmers’ Social for Health Services Quality Evaluation in Developing Networks via Randomized Information Exchange in Countries.” Social Science and Medicine 61 (9): 1944–51. Rural Uganda.” Working Paper, University of Mary- Macours, Karen, and Renos Vakis. 2014. “Changing House- land, College Park. holds’ Investments and Aspirations through Social WDI (World Development Indicators) (database). World Interactions: Evidence from a Randomized Transfer Bank, Washington, DC. http:/data.worldbank.org Program.” Economic Journal 124 (576): 607–33. /data-catalog/world-development-indicators. Mas, Alexandre. 2006. “Pay, Reference Points, and Police Performance.” Quarterly Journal of Economics 121 (3): 783–821. 144 WORLD DEVELOPMENT REPORT 2015 Using ethnography to understand the workplace Understanding the social and cultural framework within a large organization, simi- context of formalistic procedures in lar to what can occur in a smaller organization African utility companies through case-by-case agreement with the superior. Researchers studying the Water Authority of Togo Eventually, other utility companies in Africa fol- Spotlight 4 in the early 1990s, a high-performing company at lowed suit, adopting similar manuals. Manuals of the time, found that most employees welcomed the detailed procedures seemed to improve workplace fact that there was a voluminous manual of proce- performance, observers believed. dures (Henry 1991). Employees agreed with man- Why were these manuals—which might be seen agement that these detailed procedures improved to be intrusive in other environments—valuable for relations between colleagues and between superi- the companies? As this Report argues, context mat- ors and subordinates. ters. The manuals correspond to the written rules A short time later, the chief executive officer that are used in traditional associations in many of the Cameroon Electricity Company decided West and Central African communities, the tontines that his company should draft similar procedures (Henry, Tchenté, and Guillerme-Dieumegard 1991). to address a long-standing issue of lack of staff They prescribe, with the same sense of minutia, the empowerment (d’Iribarne and Henry 2007). Feeling conduct to be observed for everything from dealing apprehensive, employees were constantly coming with lateness, to the right to make jokes, to the orga- to their superiors to obtain authorization for what nization of meals. they were going to do. To address this situation, an In Cameroon and Togo, as elsewhere in the impressive manual, comprising a dozen large bind- world, the success of collective enterprises depends ers, was written in just a few months. The manual on managing tensions between personal interests described what everyone should do and how it and group goals. Observation of the particular should be done (detailed questions to be asked, culturally informed strategies for managing these rules of good behavior, the procedures and content conflicts helped shape the business manuals. for management checks, and so on). On-the-ground investigation found that employees Some foreign experts were puzzled: they thought constantly and subtly sounded out the underlying these procedures amounted to micromanagement. intentions and interests of the people around them However, employees strongly backed the detailed (Smith 2008; Godong 2011). People feared greed and manuals: “They put them at ease,” explained a super- “bad faith guided by personal interests.” Conversely, visor. Detailed procedures provide a comprehensive each person was examined to see if he or she was acting as a “true friend.” In that context, acting as This spotlight is based on a background paper a true friend meant participating in the duty of prepared by Agence Française de Développement. mutual aid. Refusals could be viewed as a sign of USING ETHNOGRAPHY TO UNDERSTAND THE WORKPLACE 145 underlying nastiness of character. Many people description can also sometimes treat individual were questioning whether business decisions were lives as abstractions, almost like characters in lit- motivated by duties of mutual assistance or by the erary texts (Clifford and Marcus 1986). But wielded disinterested application of a rule. Professional appropriately, ethnography can be a powerful tool situations were reexamined in light of the personal for understanding the ways in which social and relationships among the parties involved. At the cultural context shapes decision making, choices, same time, people feared acting in ways that might and interpersonal relations. elicit suspicion. “People are afraid of anyone saying, ‘There’s the nasty guy,’ ” explained a director. “They References think that it might bring trouble down on their own Asad, Talal. 1993. Genealogies of Religion: Discipline head or on the family.” and Reasons of Power in Christianity and Islam. The approach of formalizing procedures, Baltimore: Johns Hopkins University Press. enforced by a regular audit, was seen as a way to Booth, David, and Diana Rose Cammack. 2013. reassure others that what each person does was Governance for Development in Africa: Solving not motivated by his or her own personal inter- Collective Action Problems. London: Zed Books. ests, their friends’ interests, or bad intentions, but Clifford, James, and George E. Marcus. 1986. Writing by what the company expects. Formal procedures Culture: The Poetics and Politics of Ethnography. reassured people and made them more responsible. Berkeley: University of California Press. d’Iribarne, Philippe. 2002. “Motivating Workers in Emerging Countries: Universal Tools and Local Ethnography can be a powerful Adaptations.” Journal of Organizational Behavior 23 (3): 243–56. tool for understanding the ways in Spotlight 4 d’Iribarne, Philippe, and Alain Henry. 2007. Successful Companies in the Developing World: Managing in which social and cultural context Synergy with Cultures. Paris: Agence Française de Développement. shapes decision making, choices, Geertz, Clifford. 1994. “Thick Description: Toward an Interpretive Theory of Culture.” In Readings in and interpersonal relations. the Philosophy of Social Science, edited by Michael Martin and Lee C. McIntyre, 213–31. Cambridge, This brief account shows the value of careful MA: MIT Press. ethnographic observation. In the words of anthro- Godong, Serge Alain. 2011. Implanter le capitalisme en pologist Clifford Geertz (1994), “thick descrip- Afrique: Bonne gouvernance et meilleures pratiques tion”—or a detailed understanding of the social de gestion face aux cultures locales. Paris: Karthala and cultural context surrounding decisions and Editions. actions—was necessary to understanding how Henry, Alain. 1991. “Vers un modèle du manage- employees interpreted their interpersonal relations ment Africain.” Cahiers d’études Africaines 447–73. and organizational procedures (d’Iribarne 2002; Henry, Alain, Guy-Honoré Tchenté, and Philippe Booth and Cammack 2013). Guillerme-Dieumegard. 1991. Tontines et banques Although valuable, thick descriptions have au Cameroun: Les principes de la société des amis. limitations. A danger with some forms of thick Paris: Karthala Editions. description is that they can leave out the ways in Smith, James Howard. 2008. Bewitching Development: which political and economic power, in addition Witchcraft and the Reinvention of Development in to cultural meanings, also shape individual choice Neoliberal Kenya. Chicago: University of Chicago and behavior (Asad 1993). Approaches to thick Press. 8 CHAPTER Health Every day, people get sick, stay sick, or even die Changing health behaviors in the because of missed opportunities.1 Each year, 7.6 mil- face of psychological biases and lion children under the age of five die from avoidable social influences causes (Liu and others 2012). In countries that suffer Telling people that there is a way to improve their the greatest share of these deaths, the most effective health is rarely sufficient to change behavior. In gen- interventions are almost all preventive or therapeutic eral, successful health promotion campaigns engage measures that should be within the reach of most people emotionally and activate or change social norms households and communities, including breastfeed- as much as they provide information. The message ing, vaccinations, assisted deliveries, oral rehydra- disseminated should be that others will support you or tion therapy, water sanitation measures that do not even applaud you if you do it, not just that something require major investments in infrastructure, and insec- is good for you. Successful campaigns address many ticide-treated mosquito nets (Jones and others 2003). or most of the following: information, performance, problem solving, social support, materials, and media (Briscoe and Aboud 2012). A campaign should tell Telling people that there is a way to people that a behavior will improve their health (infor- mation), demonstrate and model the behavior (perfor- improve their health is rarely sufficient to mance), reduce barriers to its adoption (problem solv- ing), create a system for supporting people who choose change behavior. Successful information to adopt it (social support), provide the materials necessary to begin adoption (materials), and provide a campaigns are as much about social background of support through in-person, print, radio, television, and other approaches (media). norms as they are about information. An example of a campaign that pulled together these elements occurred in Bangladesh. In 2006, more than 75 percent of urban dwellers and 60 percent of Health outcomes can be improved by applying the rural residents used oral rehydration salts (ORS) as a insights from behavioral economics and related fields: treatment for diarrhea, thanks to a prior public health individuals have limited attention and act on the basis campaign (Larson, Saha, and Nazrul 2009). But in addi- of what is salient (chapter 1); individuals intrinsically tion, public health officials wanted people to use zinc value social approval and adherence to social norms (which was widely available and cheap) together with (chapter 2); and individuals have many frames (or ORS, and a major campaign was introduced to increase mental models) through which they can interpret a the use of zinc as a supplement for infants, which situation (chapter 3). greatly increases the rate of survival in cases of severe HEALTH 147 diarrhea. Officials mounted a campaign that included behaviors and prevent negative behaviors only when direct marketing (painted dinner plates), community the campaigns are paired with local efforts to support engagement and social support (courtyard meetings), the desired behavior change (Wakefield, Loken, and and role modeling (plays, radio dramas, and television Hornik 2010). Most campaigns are too short in dura- serials), as well as public displays like branded rick- tion, and some even backfire. For instance, a recent U.S. shaws. As a result of this campaign, knowledge about antidrug campaign targeting youth may have uninten- the use of zinc increased from almost zero to more tionally increased drug use by suggesting that it was than 75 percent. commonplace. Teens took this message to mean that it was acceptable among their peers (Wakefield, Loken, Enhancing the use of mass media and Hornik 2010). The health information was ignored, Three examples of mass media illustrate the dual chal- but not the information about the social norm. lenge of changing individuals’ beliefs and their health Mass media campaigns on health do not appear to behaviors. The examples relate to breastfeeding, smok- be useful in changing mistaken mental models of ill- ing, and HIV testing. ness because the message is filtered through the model Breastfeeding is one of the least expensive strate- itself. For example, over a third of poor women in India gies for improving the health of young children. Many believe that increasing fluid intake for children with mass media campaigns have encouraged breastfeeding. Evaluation of seven campaigns in developed countries found that they increased rates of initiating breast- Policy makers can make major feeding among poor women (Dyson, McCormick, and Renfrew 2006). No mass media campaigns in devel- strides in improving health outcomes oping countries have been systematically evaluated, but the available information suggests that they can by understanding that people think work when paired with local efforts that involve direct and proactive interactions with women and their automatically, interpret the world social networks (Renfrew and others 2012; Naugle and Hornik 2014). based on implicit mental models, Mass media campaigns have frequently been used to reduce smoking rates. Such campaigns have been and think socially. extensively studied and evaluated in developed coun- tries, mostly in the United States, where variation in campaigns across states can be used to measure diarrhea makes them sicker. They follow a model in impact. These campaigns have been most effective which diarrhea is interpreted as leaking; since more at preventing young people from taking up smoking fluid means more leaking, it must be bad (Datta and and in supporting individuals who have already quit Mullainathan 2014). With such a mental model, the smoking (see, for example, Bala and others 2013). Using message that ORS helps children survive diarrhea may community members such as teachers and parents to fall on deaf ears, since, according to that model, ORS deliver messages and extending the campaigns over a only increases leaking—it does not decrease it. long period (at least 12 months) increase their success. One opportunity for tackling mental models can However, the reviews find no evidence that the cam- come from the juxtaposition of well-known “moral” or paigns lead smokers to quit smoking or change the “valuable” members of society and misunderstood ill- social norms of smoking. nesses or stigmatized individuals. For example, media Similarly, a review of over 20 mass media cam- coverage of celebrity medical diagnoses increases paigns to encourage HIV testing finds no long-term screening and can stimulate interest in behavior effects after the campaigns ended (Vidanapathirana change (Ayers and others 2014). In 2011, for example, and others 2005). In many cases, however, there are former Brazilian president Lula da Silva publicly dis- short-term effects. In the case of HIV testing, even cussed his throat cancer, which he attributed to his short-term effects are socially important. long-held smoking habit. His frank discussion of the ill- A review of the published literature evaluating all ness and his own role in causing it was widely covered types of mass media campaigns for health echoes these in the media (photo 8.1). Following his announcement, findings. The campaigns effectively promote positive interest in quitting smoking reached unprecedented 148 WORLD DEVELOPMENT REPORT 2015 Social learning about health care quality Photo 8.1 Former Brazilian president People learn about the quality of health care from each Lula da Silva’s battle with throat cancer other. Typically, if an individual visits a new doctor and was widely covered in the media is cured, the word spreads and the doctor’s reputation improves. But what happens when the individual visits a new doctor and does not get the medicine he sought (antibiotics or steroids, for example)? Some- times households will take the event as evidence that the doctor is not responsive to patients’ needs or does not stock the necessary medicines, rather than that the doctor knows what is best for the patient and is deter- mined to provide the best possible care. When people learn from one another, they may all end up holding the correct beliefs, or they may all end up mistaken. For example, if a person receives a referral by one doctor to Credit: Ricardo Stuckert/Instituto Lula. visit another doctor, households take that as a signal to avoid the referring practitioner and visit only the referred practitioner (Leonard, Adelman, and Essam levels, and Brazil passed new antismoking laws. 2009). This behavior prevents people from learning the Figure 8.1 shows one indicator of interest—Google underlying relationships between the practice of refer- searches related to quitting smoking. In Brazil, these ral and health outcomes. Because households avoid searches were 71 percent higher even four weeks after doctors who refer their patients, they do not learn that the announcement, long after the media had stopped those providers are actually better than the ones who covering Lula’s diagnosis. According to Ayers, “Lula’s refuse to refer their patients. announced cancer diagnosis, though tragic, was Evidence from multiple studies of rural African potentially the greatest smoking cessation–promoting households (reviewed in Leonard 2014) shows that peo- event in Brazilian history” (Price 2013). ple seek to match their illness to the most appropriate Figure 8.1 If a well-known person has a disease, the public might think more seriously about ways to prevent it After former Brazilian president Lula da Silva publicly discussed his throat cancer, which he attributed to smoking, Brazilians became much more receptive to information about smoking. 100 Relative search volume 80 60 40 20 0 Oct. 1 Oct. 15 Nov. 1 Nov. 15 Dec. 1 2011 2008–10 Oct. 29, 2011—public learns of Lula’s diagnosis Source: Ayers and others 2014. Note: The blue line shows time trends for daily Internet searches related to quitting smoking around the time of former Brazilian president Lula da Silva’s cancer diagnosis. The green lines show time trends for the same period from prior years. Search volume is measured as relative search volume (RSV), where RSV = 100 is the day with the highest search proportion, and RSV = 50 is a day with 50 percent of that highest proportion. HEALTH 149 health care provider. When a new option or a new doc- Inducing people to take more preventive care is tor becomes available, they are particularly interested difficult, but a deeper understanding of the way people to hear about others’ experiences. They are more likely think can help. One possibility, for example, is to reduce to visit a doctor when someone in their close commu- barriers to the desired behavior by making the exact nity has recently visited that provider and had a good steps needed for the preventive care more salient or by outcome (Leonard, Adelman, and Essam 2009). By fol- providing a small material incentive. It is also possible lowing this simple process of updating expectations in to alter the way people weigh the benefits of action the face of the unknown, households in Tanzania made by using nudges and other behavioral tools to alter better decisions over time and visited better doctors, the choice architecture (Thaler and Sunstein 2008). as objectively measured by medical experts (Leonard, In addition, it may be possible to change behavior by Mliga, and Haile Mariam 2002). The process of social changing the beliefs a person holds that are not related learning, though, even when it is useful, can be very to the private benefits and costs of a given health mea- slow. For example, it took between three and four years sure, including beliefs that others would approve of for communities to learn whether new doctors in their the behavior, beliefs that others engage in the behav- area gave good or bad advice (Leonard 2007). ior, and beliefs in one’s ability to perform the behavior Unlike information about how to improve their (self-efficacy). In addition, people may be more willing health (which is often ignored), information that aids to engage in the behavior if they know they will receive households in seeking the best available care, based support, reinforcement, feedback, or reminders. on realistic assessments of the capabilities and quality The discussion that follows gives several examples of the facilities from which households can choose, is of the first two methods: presenting advice in ways likely to be very useful because households are already that recognize how people make decisions, and reduc- seeking this information. Better information could ing the barriers to changing behaviors. The second two help them make better decisions more quickly. methods—community-level models of behavior and the use of support, reinforcement, and feedback—are Psychological and social covered in the upcoming section on follow-through approaches to changing health and habit formation. behavior Even after people accept information, they do not Framing information about vaccinations always act on it. The zinc campaign discussed earlier and HIV testing succeeded in educating 75 percent of Bangladeshis, There is a world of difference between these two but two years after the program, zinc was used in only statements—“If you get the flu vaccine, you will be 35 percent of the indicated cases. Although there are less likely to get the flu” versus “If you do not get the many models of health behavior,2 an assumption com- flu vaccine, you are more likely to get the flu”—even mon to most is that people carefully weigh the benefits though they contain the same information. In a and barriers to adoption against their susceptibility to, review of 94 studies comparing gain-framed to loss- and the likely severity of, bad outcomes if they do not framed messages, gain-framed messages consistently adopt. That is, the standard models assume that indi- improved adoption of preventive behaviors (such as viduals think deliberatively, not automatically. But as vaccinations) when compared to loss-framed messages chapter 1 showed, the reverse is actually true. with the same objective information (Gallagher and Imagine someone who considers getting tested Updegraff 2012). Interestingly, people who hear one or for tuberculosis: she knows she has a chance of being the other of the two messages are equally likely to say infected and that the illness is severe. But at the same that they want to seek preventive care, but people who time, taking the test would require her to leave work hear the first message are much more likely to follow early and stand in a line at a clinic. Studies about how through and actually get the vaccine. In general, the people make decisions about health care have consis- same information can be presented in different ways tently found that people tend to consider the benefits to improve actual behavior. and barriers, while ignoring susceptibility and severity Chapter 1 described how raising the number of free (Zimmerman and Vernberg 1994; Carpenter 2010). Thus test reports, from three to four, that a testing agency people will often forgo preventive medicine because of routinely sent to colleges had the effect of increasing small obstacles, even when they know that they are the number of low-income students attending selective highly susceptible and face potentially severe conse- colleges. More generally, many program choices entail quences. Individuals frame the problem too narrowly. a default condition in which people either can choose 150 WORLD DEVELOPMENT REPORT 2015 to do something when asked (opt in) or are automat- ated a shift in its approach to counseling and testing ically enrolled but given the choice to withdraw (opt for HIV by declaring an opt-out approach to be ethically out). When using a preventive service is believed to acceptable for certain populations (specifically includ- make all or most people better off, and when, even for ing people with tuberculosis). Reviews of programs those who do not gain from it, the cost is small, then that compare the opt-in to the opt-out default gener- wherever possible, it is better to present preventive ally find increased testing rates, but they also find low services on an opt-out basis: the default should be set levels of testing in either program (see, for example, to the behavior that would make most people better Baisley and others 2012). Most often, the reason is that off. Health care is full of examples of opt-out activities. the health services that provide counseling and testing The doctor takes patients’ pulse and temperature with- have shortcomings of infrastructure, incentives, or out asking them if they think it would be a good idea, governance (Roura and others 2013). In addition, some for example. Surgical consent forms are designed to of the studies find that the increased rate of testing did present the doctors’ recommendations as the default, not result in more detection of HIV/AIDS, suggesting forcing the patient to find an alternative. that the additional people tested because of the new What is the potential for increasing the number of defaults were from populations that had not been defaulted behaviors? HIV testing is an area in which exposed to HIV/AIDS. a change from opt in to opt out has been extensively Opt-out defaults are likely to increase the use of studied. In 2004, the World Health Organization initi- preventive services when health systems are able to provide them. However—as in the case of HIV-negative patients not opting out of testing—this improvement Figure 8.2 Take-up of health products drops may occur at the cost of bringing in people for whom precipitously in response to very small fees these services are less useful. Policies often set the prices of preventive health care products low to promote access while also providing a revenue stream to providers. But if access is Prices as a source of meaning important, it makes sense to bring the price all the way down to zero. A series of Because the most obvious barrier to adopting new evaluations finds that even small price increases above zero lead to large drops behavior is cost, lowering prices should be the best way in the number of people who choose to buy health products. to improve adoption. But prices have many meanings besides a value in exchange. Prices at or near zero may 100 constitute a special threshold, according to a review by Kremer and Glennerster (2011). People are willing to Percentage of people taking up product 90 adopt many health goods at a price of zero (or almost 80 zero) but almost completely unwilling to adopt it at prices just slightly above zero (see figure 8.2). The 70 study demonstrates this effect for deworming medi- cine, mosquito nets, water disinfectants, and soap. 60 It appears that prices contain at least two differ- ent signals for people. First, low prices make things 50 more affordable. But free means something special. 40 When prices fall toward zero, free may convey a social norm: we all should be doing this. Free allows people to 30 experiment with a product when they are uncertain of its value, and free can have an affective influence (an 20 individual is excited to have won the opportunity to get something free). Households given free mosquito nets 10 may use them differently from the way households 0 that purchase subsidized mosquito nets use them and may be more likely to use them for their children, 0 0.10 0.20 0.30 0.40 0.50 0.60 0.70 0.80 0.90 1.00 possibly responding to a social signal in the price Price of product (2009 U.S. dollars) (Hoffmann 2009). Perhaps households adopt new tech- nologies that are free in the short run, and then after Deworming, Kenya Bed net vouchers, Kenya Water disinfectant, Kenya they have experienced their value, they become willing Bed nets in clinics, Kenya Water disinfectant, Zambia Soap, India to pay positive prices for them later, as Dupas (2014) Source: Abdul Latif Jameel Poverty Action Lab 2011. suggests. People are less likely to adopt a free option if HEALTH 151 they have been asked to reflect carefully on its value in money can help people focus on completing a full comparison to a positively priced item, as Shampanier, course of action. Mazar, and Ariely (2007) show. This finding suggests Commitment devices can help people follow that the immediate response to free items is not fully through on intentions to change behavior. In a case rational. At least in some cases, it is based on an auto- involving smokers in the Philippines (Giné, Karlan, matic, not a reasoned, response. and Zinman 2010), people voluntarily deposited their When things are free, however, people may over- own money in accounts that would be forfeited if they consume or waste the product. Positive prices may also did not quit smoking; participating in this experiment help target goods to where they can do the most good. did indeed help smokers quit (and not resume) smok- And for curative medicine, the willingness to pay can ing. The randomly selected individuals offered this be high (Ashraf, Jack, and Kamenica 2013). opportunity were 3 percent more likely to quit (as mea- The cognitive effect of free or minimal pricing is sured one year later). Eleven percent of people offered a new area of research in developing countries, and the opportunity chose to commit their own funds, and much will be learned over the coming decade. How- 34 percent of them made good on their intentions. ever, policy makers can already begin to think about how to signal the excitement that is contained in the Asking people why they don’t seek care word free without incurring the costs of offering a zero is not useful price. If the good is important to health and has posi- Asking people why they forgo care that would seem to tive externalities, if demand for the good is otherwise make them better off is generally not helpful for form- low, and if waste is not a large concern, then reducing ing policy. The studies often appear to have great pre- the price to zero should be considered for the sake of dictive power, in that what people say matches what the affective response it can invoke. Vaccinations, for they do. However, this is deceptive because people example, meet these conditions. adapt their beliefs to match their behavior (Harrison, In contrast, if waste is a large concern, prices should Mullen, and Green 1992), and thus while the studies be kept above zero, and social norms should be invoked tell us that people did choose a certain behavior, they to increase demand for the good. The positive prices do not tell us why. These studies do a better job of are a targeting mechanism: they help ensure that the explaining and predicting intentions than actions. right people are buying the good or service. Coupons, Thus surveys of knowledge, attitudes, and practices prizes, public celebrations, and media can all be used (so-called KAP studies) fail to identify explicit ways to to create or strengthen a social norm or generate an change behavior. affective response, even if prices are not zero. Improving follow-through and Conditional cash transfers and commitment habit formation mechanisms As discussed, sometimes people form intentions to In some cases, goods and services are free and people adopt preventive actions but do not follow through. still do not use them. Many preventive services, such They intend to change, but before an activity becomes as antenatal services, are provided free but are under- a habit, it is difficult for them to maintain the energy utilized. One well-documented way to increase use is and focus to carry out their good intentions. The key to to create conditional cash transfers (CCTs) where, for behavioral interventions is to make the long-term ben- example, women receive payments for going to the efits of adherence salient in the short term. Individuals antenatal clinic but forfeit them if they fail to go. On the often do not need information about distant benefits; surface, these programs do not appear to use anything they need to experience immediate benefits. A good but standard economic incentives to improve behavior, example of how immediate benefits can help improve but, considered more broadly, some incentive programs adherence is HIV/AIDS treatment in Africa. Despite reveal interesting behaviors. Banerjee and others (2010) significant additional difficulties in access, education, examine a program in which women received free and information, Mills and others (2006) found that lentils and plates as an incentive to immunize their baseline adherence to antiretroviral therapy (ART) children. Many parents were taking their children to was much higher among African patients than among receive at least one vaccination, but were not following patients in developed countries like the United States, through to complete the entire series. The incentive primarily because the African patients were sicker helped increase the rates of full immunization. Thus, when they first received care and therefore felt the even when people value services (many parents made benefits of ART more immediately. Adherence is easier sure that their children were partially immunized), when the benefits are salient on a day-to-day basis. 152 WORLD DEVELOPMENT REPORT 2015 Using reminders to increase adherence to Figure 8.3 Text message reminders can medical regimens improve adherence to lifesaving drugs One of the most rapidly expanding tools in health care is the use of mobile phones to communicate regularly Text message reminders improved adherence to antiretroviral therapy in a study among HIV/AIDS patients in Kenya. with populations that were previously difficult to Although reminders have been used for other behaviors reach. There has been positive experience in multiple as well, especially saving, not all reminders work the same settings with reminders, now easily sent through text way. In this case, people seemed to tune out reminders that arrived daily. messaging. In developed countries, there is robust evidence of the effectiveness of using mobile technol- 100 ogies to remind people to attend health appointments Percentage achieving 90% adherence (Tomlinson and others 2013). 90 The evidence in developing countries is more 80 mixed—not because the technology is not effective but because few studies have been carefully evaluated. Sys- 70 tematic reviews of the existing evidence in developing 60 countries recommend implementation and scaling up but caution that little evidence points to what works 50 best in different situations (Cole-Lewis and Kershaw 40 2010). They suggest that mobile messages are more likely to be effective when there is follow-up, when 30 the message is personally tailored to the recipient, and 20 when the frequency, wording, and content are highly relevant to the patient. Blasting text messages to large 10 portions of the population reminding them of all the things they can do to improve their health is likely to 0 No reminder Daily reminder, Weekly reminder, be a waste of resources: the messages are not salient “This is your “This is your or tailored. Indeed, Pop-Eleches and others (2011) find reminder.” reminder.” that daily messages about adherence to ART for HIV/ Source: Pop-Eleches and others 2011. AIDS are not effective but that weekly messages are, suggesting that people are not forgetting to take their resulting from inadequate and unsafe water, poor san- medicine (taken daily) but rather need a reinforcing itation, and unhygienic practices (UNICEF and WHO message on a less frequent basis (see figure 8.3). 2009). New evidence discussed by Spears, Ghosh, and Cumming (2013) links open defecation to stunting. By Triggering community-level responses reducing normal nutrient absorption, diarrheal dis- Patients are more likely to adopt a new health practice eases lead to impaired physical growth and cognitive when their experience with the provider has been pos- development. itive (Peltzer and others 2002) or if they have positive The traditional approach to ending open defecation responses from their community. A good experience was to provide information to communities about the with the provider gives patients a sense of immediate transmission of disease and to subsidize the construc- satisfaction when they follow through, similar to the tion of toilets. An alternative approach, Community- sense of satisfaction from conforming to commu- Led Total Sanitation (CLTS),3 aims to generate demand nity norms. In the latter case, community feedback for a community free of open defecation and to elicit becomes the benefit. Thus even when there are no from the community itself an increase in the supply immediate benefits to adherence or adoption, commu- of sanitation products. It does this by raising collective nity reinforcement can be generated by encouraging awareness of the sanitation problem. Facilitators are adoption at the community level. sent to the community to initiate discussions, which Consider one of the biggest causes of health prob- are held in public places and involve a “walk of shame,” lems in the world, open defecation. Globally, 2.5 billion during which groups walk to places that have been people have inadequate sanitation; 1.2 billion defecate used for open defecation, collect some of the feces, in the open. Lack of sanitation causes a tremendous place it on the ground next to a bowl of rice, and watch disease burden among the poor, especially poor as flies move between the feces and the rice. Then the infants and young children. Each year, more than 1.5 CLTS facilitator asks community members, “Would million children under the age of five die from diarrhea you like to eat the rice?” Although people know that HEALTH 153 flies travel, the image of food and feces next to each Figure 8.4 Changing social norms is other triggers an emotional response (disgust) that important but not sufficient for ending makes it difficult for them to forget their own intention open defecation to change behavior. The program stimulates a desire by the villagers to end open defecation and to forge their Community-Led Total Sanitation (CLTS) is a methodology for own plan for achieving it, with limited follow-up sup- engaging communities in eliminating open defecation. It tries port. Communities that become open-defecation free to trigger collective shame and disgust for the implications of receive recognition by local governments. open defecation. Until recently, the available evidence on the success of CLTS was from small-scale interventions. In 2007, local and national governments in rural India and Randomized controlled trials have tested the effectiveness of Indonesia,4 with technical support from an interna- CLTS implemented at scale in Madhya Pradesh, India, and tional sanitation program, began implementing the East Java, Indonesia. first large-scale CLTS programs to be experimentally evaluated (Cameron, Shah, and Olivia 2013). Some communities were randomly selected to receive the Open defecation decreased. treatment, while others were randomly selected to Percentage reporting open defecation serve as controls and not to receive the treatment 100 within the period of the evaluation. As shown in figure 90 8.4, the CLTS programs were found to decrease open 80 defecation by 7 percent and 11 percent from very high 70 levels in Indonesia and India, respectively, compared to the control villages. Additional findings suggest 60 that CLTS can complement, but perhaps not substitute 50 for, resources for building toilets. In India, the CLTS 40 program was combined with a subsidy for toilet con- 30 struction, and the impact on toilet construction—20 20 percentage points—was greater than that in Indonesia. In summary, a comparison of outcomes in treatment 10 and control communities shows declines in open def- 0 India Indonesia ecation and increases in toilet construction. A program to change social norms about sanitation in these two No CLTS With CLTS intervention countries was important but not sufficient to end open defecation. The program increased toilet construction, with a particularly Encouraging health care large effect in India, where the CLTS program was combined providers to do the right with a subsidy for toilet construction. things for others Health is co-created by patients, doctors, nurses, other Percentage of households with a toilet experts, community health workers, and household India members.5 As Ashraf (2013) has noted, “Health isn’t something that can be handed to people; it is a state that they must produce themselves by interacting 24.2 44.1 with a health care system . . . providers and recipients co-create health” (120–23). A key element in the pro- Indonesia duction of health is the trust that patients have in their providers: trust to seek care, trust to follow through on the prescribed treatment, and trust to understand 13 messages about what is good for them. Such trust is 15.9 not possible in a system that provides low-quality care. No CLTS With CLTS Why do health care providers sometimes provide Sources: Patil and others 2014; Cameron, Shah, and Olivia 2013. low-quality care? It is not sufficient to focus only Note: The study in Indonesia measured the presence of a toilet that was on material incentives for providing quality care. constructed in the two years prior to the survey. 154 WORLD DEVELOPMENT REPORT 2015 Empirical work points to many additional factors. should. In Delhi, India, Das and Hammer (2007) found Even with the best training, health care providers suf- that some of the most qualified doctors were the least fer from the same biases as everyone else. They cannot likely to follow through on their knowledge, implying consider all possible symptoms, conditions, diagnoses, that the doctors with the highest qualifications were and treatments. They must use simplifying rules and not providing the best medicine. Many other studies heuristics to do their job, and these heuristics can have also found that, although knowledge could be lead to systematic mistakes (Croskerry 2002). Many of higher, doctors do not use the knowledge they already these biases have been discussed in previous chapters: possess (Das and others 2012). As a result, there has anchoring, the availability and representativeness of been a shift from a focus on competence to a focus heuristics, framing effects, overconfidence bias, and on the “know-do” gap, the difference between compe- confirmation bias. In addition, some biases are more tence and performance (Rowe and others 2005; Das, specific to health care and to the relationship between Hammer, and Leonard 2008; Das and others 2012). providers and their patients. For example, “diagnosis Given the existence of that gap, increasing spending momentum” occurs when changing a diagnosis feels on training will not improve quality, and it is time to harder than keeping it, despite new evidence that runs focus on ways to get doctors to put into practice what counter to the original diagnosis. “Fundamental attri- they already know. bution error” occurs when health workers blame their patients for their symptoms rather than looking for Reminders for adhering to protocols other causes. “Gender bias” occurs when health work- Often, simply reminding health workers of the social ers assume that gender is a factor in an illness even expectations of their performance can improve it. Evi- when the evidence is not supportive. “Outcome bias” dence from almost 100 studies on the impact of peer results when health workers choose the diagnosis that visits to remind health workers about best practices has the best possible outcome (essentially hoping for finds that these visits have an impact—but not because the best), despite evidence that points to a different they introduced a financial incentive to improve qual- diagnosis. Health workers suffer from “premature ity (Jamtvedt and others 2007). For example, clinicians closure,” ending their careful consideration of a case as in urban Tanzania significantly increased their effort soon as they have a plausible diagnosis but before they when a visiting peer simply asked them to improve can be sure. their care (Brock, Lange, and Leonard, forthcoming). Health workers also suffer from a “visceral bias,” in In that study, there was no new information or change which liking or disliking the patient causes them to in incentives or material consequences from the rule out certain outcomes too soon (Croskerry 2002). visit. Health workers already have the competence to This bias is especially likely when a patient suffers improve quality and will respond to visits by their peers from a stigmatized illness or is a member of a stigma- that set new expectations or remind them of existing tized population. Such a patient is less likely to seek expectations. Indeed, supportive supervision (regular care in the first place, and when he does, he is much contact with medical peers who provide reminders less likely to receive the type of care he needs. The bias of expectations, not enforcement of rules or regula- can be subtle, in the sense of premature diagnoses, or tions) is necessary for sustained success, a review of severe. In its worst manifestation, the health system the literature on successful community health worker assigns a low priority to illnesses suffered by an entire programs concludes (Jaskiewicz and Tulenko 2012). population (Gauri and Lieberman 2006; Lieberman Even for community health workers, who are serving 2009), health care providers refuse to provide service, their neighbors and should be the most likely to be and afflicted individuals are reluctant to seek treat- motivated by prosocial preferences, exposure to peers ment for even life-threatening health problems. and supervisors (not neighbors) is necessary to sustain Thus, there is also a gap between knowledge and norms of professional behavior. actions. Recent studies in Tanzania (Leonard, Masatu, and Vialou 2007; Leonard and Masatu 2010) found that Professional and service norms in in sessions with a patient, almost half the doctors did recruitment and quality assurance not touch the patient and therefore did not know the Reminding health workers about the impact of their patient’s temperature, respiratory rate, pulse, and the actions on the welfare of their patients and on their like. The research found that the doctors knew much reputation among peers can improve service quality. of what they were supposed to do and were even will- Most health workers are responsive to the norms ing to demonstrate all the proper steps to the research set by their peers, which is a type of professionalism team. They decided not to do what they knew they (Freidson 2001; Akerlof and Kranton 2005). How can HEALTH 155 policy makers create or activate a professional norm providing information is not sufficient to get people in settings where quality is low? The literature offers to change behaviors that undermine health. Fram- many examples in which leadership transformed an ing effects that make social expectations and social underperforming health service into a high-quality approval more salient can sometimes encourage service with motivated providers at all levels (see, for individuals to seek preventive care and adhere to treat- example, Tendler 1997; Wasi 2000; Hall and Lamont ment when they otherwise would not, even though the 2009; Wibulpolprasert and others 2011). The success individual benefits exceed the individual costs. Indi- stories point to the fact that the transformation is pos- viduals can suffer from an intention-action divide and sible, but they cannot isolate the elements of a solution that would work in all places. Programs that pay providers to improve quality or Whereas increasing information or quantity of services (pay for performance or results- based financing) have gained attention recently, in part knowledge is often not helpful or based on the success documented in Rwanda (Basinga and others 2011). Providers in that study responded dra- sufficient, simply reminding health matically to a change in financing from input-based (paying for what is needed) to reward-based (paying workers of social expectations about staff bonuses if certain targets are met with respect to assisted deliveries, vaccinations, or well-baby visits, their performance can improve it. for example). This might suggest that monetary incen- tives are the solution and that behavioral interventions are not important. However, a careful examination so can health care providers, and commitment devices of pay-for-performance incentive programs such as and reminders can narrow those divides. Appealing that in Rwanda reveals that the programs not only to social expectations and professional standards can use monetary incentives but also expand autonomy, lead to significant improvements in the actions of pro- accountability, team-based recognition of effort, and viders. When providers act in the best interests of their exposure to external peers. All these aspects could patients, their patients are likely to notice and increase increase quality by activating professional norms. their trust in the advice provided by these same pro- Growing evidence indicates that health workers viders, which should lead to further improvements in respond well to social cues in the form of recognition health outcomes.6 and gifts (for more on this, see chapter 7). When health workers are given small gifts like a book or a pen, Notes they will respond by improving the quality of care 1. This chapter benefited from a number of recent they provide, in some cases, for significant periods of review pieces, notably, Frederick, Loewenstein, and time (Currie, Lin, and Meng 2013; Brock, Lange, and O’Donoghue (2002); DellaVigna (2009); Dupas (2011); Leonard, forthcoming). In addition, health workers Kremer and Glennerster (2011); Lawless, Nayga, and Drichoutis (2013); and Kessler and Zhang (forthcoming). respond to the recognition that comes from awards 2. See Noar and Zimmerman (2005) for a survey of and token prizes like stars to display in the workplace elements in health behavior models, including the and congratulatory plaques (Ashraf, Bandiera, and Jack, Health Belief Model, the Theory of Reasoned Action, forthcoming). This response to gifts and tokens makes the Theory of Planned Behavior, the Social Cognition little sense in standard economic models but can be Theory, and the Transtheoretical Model. easily understood in the terms laid out in part 1: in this 3. Kamal Kar developed CLTS in Bangladesh in 2000. broader view, gifts can be understood as a way of mak- Since then, CLTS has been used in over 60 countries ing social ties and connections more salient, activating and has become national policy in at least 20 countries. a frame of gift giving, and signaling social approval. 4. The program was also launched in a third country, Tan- zania, but results from this case are not yet available. Conclusion 5. Recognizing the economic implications of co-created health has been proposed as one of the unique features Understanding that people think automatically, inter- of traditional medicine in Africa (Leonard and Graff pret the world based on implicit mental models, and Zivin 2005); it may be that this indigenous institution think socially allows policy makers to make major represents the first foray into behavioral health. strides in improving health outcomes. Individuals 6. This chapter was based on a systematic literature sometimes value information highly (for example, review using the following methods. In October 2013, when seeking curative care), but at other times we conducted keyword searches of the following 156 WORLD DEVELOPMENT REPORT 2015 databases: Academic Search Premier, Econlit, PsycINFO, and Testing among Young People Participating in PsycARTICLES, and PsycAbstracts. The search was an HIV Prevention Trial: Comparison of Opt-out and restricted to academic articles published after 1990. We Opt-in Strategies.” PLoS ONE 7 (7): e42108. doi: 10.1371 used search terms related to the following categories: /journal.pone.0042108. social norms, present bias, status quo bias, and trust/ Bala, Malgorzata M., Lukasz Strzeszynski, Roman Topor- persuasion intersecting with the following diseases: Madry, and Kate Cahill. 2013. “Mass Media Interven- pneumonia, measles, diarrhea, malaria, and tubercu- tions for Smoking Cessation in Adults.” Cochrane Data- losis, HIV/AIDS, smoking, and obesity. Due to the large base of Systematic Reviews January 23 (1): CD004704. doi: number of hits for HIV/AIDS, smoking, and obesity, 10.1002/14651858.CD004704.pub2. the evidence presented for these three fields is based Banerjee, Abhijit Vinayak, Esther Duflo, Rachel Glenner- on Cochrane Reviews of behavioral interventions. The ster, and Dhruva Kothari. 2010. “Improving Immuni- abstracts of the extracted articles were reviewed and sation Coverage in Rural India: Clustered Randomised only papers meeting the following criteria were con- Controlled Evaluation of Immunisation Campaigns sidered. First, only studies in English and studies with with and without Incentives.” BMJ 2010 (340): c2220. human subjects were included. Second, studies that doi: http://dx.doi.org/10.1136/bmj.c2220. were not directly relevant to one of the five diseases Basinga, Paulin, Paul J. Gertler, Agnes Soucat, and Jennifer and one of the keyword categories were excluded. This Sturdy. 2011. “Effect on Maternal and Child Health Ser- restriction accounts for the fact that some of the arti- vices in Rwanda of Payment to Primary Health-Care cles that were identified in the keyword search used Providers for Performance: An Impact Evaluation.” the keywords to illustrate a different concept than the Lancet 377 (9775): 1421–28. behavioral biases we are interested in. Other articles Briscoe, Ciara, and Frances Aboud. 2012. “Behaviour Change focused on a different disease but mentioned the dis- Communication Targeting Four Health Behaviours ease of interest in passing, for example, as a side effect. in Developing Countries: A Review of Change Tech- The third selection criterion was that the study had to niques.” Social Science and Medicine 75 (4): 612–21. contain an evaluation of an intervention (a randomized Brock, J. Michelle, Andreas Lange, and Kenneth L. Leo- controlled trial, pre-post study, natural experiment, or nard. 2014. “Giving and Promising Gifts: Experimental the like) or a lab or field economic experiment. Finally, Evidence on Reciprocity from the Field.” Working studies that contained only qualitative data from inter- Paper 165, European Bank for Reconstruction and views and opinions about different interventions were Development, London. also excluded. ————. Forthcoming. “Generosity and Prosocial Behavior in Health Care Provision: Evidence from the Labora- References tory and Field.” Journal of Human Resources. Abdul Latif Jameel Poverty Action Lab. 2011. “The Price Is Cameron, Lisa, Manisha Shah, and Susan Olivia. 2013. Wrong: Charging Small Fees Dramatically Reduces “Impact Evaluation of a Large-Scale Rural Sanitation Access to Important Products for the Poor.” April Project in Indonesia.” Policy Research Working Paper Bulletin. http://www.povertyactionlab.org/publication 6360, World Bank, Washington, DC. /the-price-is-wrong. Carpenter, Christopher J. 2010. “A Meta-Analysis of the Akerlof, George A., and Rachel E. Kranton. 2005. “Identity Effectiveness of Health Belief Model Variables in and the Economics of Organizations.” Journal of Eco- Predicting Behavior.” Health Communication 25 (8): nomic Perspectives 19 (1): 9–32. 661–69. Ashraf, Nava. 2013. “Rx: Human Nature.” Harvard Business Cole-Lewis, Heather, and Trace Kershaw. 2010. “Text Review 91 (4): 119–25. Messaging as a Tool for Behavior Change in Disease Ashraf, Nava, Oriana Bandiera, and B. Kelsey Jack. Forth- Prevention and Management.” Epidemiologic Reviews coming. “No Margin, No Mission? A Field Experiment 32: 56–69. on Incentives for Public Service Delivery.” Journal of Croskerry, Pat. 2002. “Achieving Quality in Clinical Deci- Public Economics. sion Making: Cognitive Strategies and Detection of Ashraf, Nava, B. Kelsey Jack, and Emir Kamenica. 2013. Bias.” Academic Emergency Medicine 9 (11): 1184–1204. “Information and Subsidies: Complements or Substi- Currie, Janet, Wanchuan Lin, and Juanjuan Meng. tutes?” Journal of Economic Behavior and Organization 88: 2013. “Social Networks and Externalities from Gift 133–39. Exchange: Evidence from a Field Experiment.” Journal Ayers, John W., Benjamin M. Althouse, Seth M. Noar, and of Public Economics 107 (November): 19–30. Joanna E. Cohen. 2014. “Do Celebrity Cancer Diagnoses Das, Jishnu, and Jeffrey Hammer. 2007. “Money for Noth- Promote Primary Cancer Prevention?” Preventive Medi- ing: The Dire Straits of Medical Practice in Delhi, cine 58: 81–84. India.” Journal of Development Economics 83 (1): 1–36. Baisley, Kathy, Aoife M. Doyle, John Changalucha, Kaballa Das, Jishnu, Jeffrey S. Hammer, and Kenneth L. Leonard. Maganja, Deborah Watson-Jones, Richard Hayes, and 2008. “The Quality of Medical Advice in Low-Income David Ross. 2012. “Uptake of Voluntary Counselling Countries.” Journal of Economic Perspectives 22 (2): 93–114. HEALTH 157 Das, Jishnu, Alaka Holla, Veena Das, Manoj Mohanan, Jones, Gareth, Richard W. Steketee, Robert E. Black, Diana Tabak, and Brian Chan. 2012. “In Urban and Zulfiqar A. Bhutta, and Saul S. Morris. 2003. “How Rural India, a Standardized Patient Study Showed Low Many Child Deaths Can We Prevent This Year?” Lancet Levels of Provider Training and Huge Quality Gaps.” 362 (9377): 65–71. Health Affairs 31 (12): 2774–84. Kessler, Judd B., and C. Yiwei Zhang. Forthcoming. “Behav- Datta, Saugato, and Sendhil Mullainathan. 2014. “Behav- ioral Economics and Health.” The Oxford Textbook of ioral Design: A New Approach to Development Policy.” Public Health. http://assets.wharton.upenn.edu/~czhan Review of Income and Wealth 60 (1): 7–35. /KesslerZhang_BehavioralEconomicsHealth.pdf. DellaVigna, Stefano. 2009. “Psychology and Economics: Kremer, Michael, and Rachel Glennerster. 2011. “Improv- Evidence from the Field.” Journal of Economic Literature ing Health in Developing Countries: Evidence from 47 (2): 315–72. Randomized Evaluations.” In Handbook of Health Dupas, Pascaline. 2011. “Health Behavior in Developing Economics, Vol. 2, edited by Thomas G. McGuire, Mark Countries.” Annual Review of Economics 3 (1): 425–49. V. Pauly, and Pedro P. Barros, 201–315. Amsterdam: — —— —. 2014. “Short-Run Subsidies and Long-Run Adop- Elsevier. tion of New Health Products: Evidence from a Field Larson, Charles P., Unnati Rani Saha, and Hazera Nazrul. Experiment.” Econometrica 82 (1): 197–228. 2009. “Impact Monitoring of the National Scale Up Dyson, Lisa, F. McCormick, and Mary J. Renfrew. 2006. of Zinc Treatment for Childhood Diarrhea in Ban- “Interventions for Promoting the Initiation of Breast- gladesh: Repeat Ecologic Surveys.” PLoS Med 6 (11): feeding.” Evidence-Based Child Health: A Cochrane Review e1000175. Journal 1 (2): 592–616. Lawless, Lydia J. R., Rodolfo Nayga, and Andreas Dri- Frederick, Shane, George Loewenstein, and Ted choutis. 2013. “Time Preference and Health Behavior: O’Donoghue. 2002. “Time Discounting and Time Pref- A Review.” Agricultural and Food Economics 1: 17. doi: 10.1186/2193-7532-1-17. erence: A Critical Review.” Journal of Economic Literature Leonard, K. L. 2007. “Learning in Health Care: Evidence 40 (2): 351–401. of Learning about Clinician Quality in Tanzania.” Eco- Freidson, Eliot. 2001. Professionalism: The Third Logic. Chi- nomic Development and Cultural Change 55 (3): 531–55. cago: University of Chicago Press. ————. 2014. “ ‘Active Patients’ in Rural African Health Gallagher, Kristel M., and John A. Updegraff. 2012. “Health Care: Implications for Research and Policy.” Health Pol- Message Framing Effects on Attitudes, Intentions, and icy and Planning 29 (1): 85–95. Behavior: A Meta-Analytic Review.” Annals of Behavioral Leonard, Kenneth L., Sarah W. Adelman, and Timothy Medicine 43 (1): 101–16. Essam. 2009. “Idle Chatter or Learning? Evidence of Gauri, Varun, and Evan S. Lieberman. 2006. “Boundary Social Learning about Clinicians and the Health Sys- Institutions and HIV/AIDS Policy in Brazil and South tem from Rural Tanzania.” Social Science and Medicine Africa.” Studies in Comparative International Development 69 (2): 183–90. 41 (3): 47–73. Leonard, Kenneth L., and Melkiory C. Masatu. 2010. Giné, Xavier, Dean Karlan, and Jonathan Zinman. 2010. “Using the Hawthorne Effect to Examine the Gap “Put Your Money Where Your Butt Is: A Commitment between a Doctor’s Best Possible Practice and Actual Contract for Smoking Cessation.” American Economic Performance.” Journal of Development Economics 93 (2): Journal: Applied Economics 2 (4): 213–35. 226–34. Hall, Peter A., and Michèle Lamont, eds. 2009. Successful Leonard, Kenneth L., Melkiory C. Masatu, and Alexandre Societies: How Institutions and Culture Affect Health. New Vialou. 2007. “Getting Doctors to Do Their Best: The York: Cambridge University Press. Roles of Ability and Motivation in Health Care Qual- Harrison, J. A., P. D. Mullen, and L. W. Green. 1992. “A ity.” Journal of Human Resources 42 (3): 682–700. Meta-Analysis of Studies of the Health Belief Model Leonard, Kenneth L., Gilbert R. Mliga, and Damen Haile with Adults.” Health Education Research 7 (1): 107–16. Mariam. 2002. “Bypassing Health Centres in Tanzania: Hoffmann, Vivian. 2009. “Intrahousehold Allocation of Revealed Preferences for Quality.” Journal of African Free and Purchased Mosquito Nets.” American Economic Economies 11 (4): 441–71. Review 99 (2): 236–41. Leonard, Kenneth L., and Joshua Graff Zivin. 2005. Jamtvedt, G., J. M. Young, D. T. Kristoffersen, M. A. O’Brien, “Outcome versus Service Based Payments in Health and A. D. Oxman. 2007. “Audit and Feedback: Effects Care: Lessons from African Traditional Healers.” Health on Professional Practice and Health Care Outcomes.” Economics 14 (6): 575–93. Cochrane Library. http://onlinelibrary.wiley.com/doi Lieberman, Evan S. 2009. Boundaries of Contagion: How /10.1002/14651858.CD000259.pub2/abstract. Ethnic Politics Have Shaped Government Responses to AIDS. Jaskiewicz, Wanda, and Kate Tulenko. 2012. “Increas- Princeton, NJ: Princeton University Press. ing Community Health Worker Productivity and Liu, Li, Hope L. Johnson, Simon Cousens, Jamie Perin, Effectiveness: A Review of the Influence of the Work Susana Scott, Joy E. Lawn, Igor Rudan, Harry Camp- Environment.” Human Resources for Health 10 (1): 38. doi: bell, Richard Cibulskis, and Mengying Li. 2012. “Global, 10.1186/1478-4491-10-38. Regional, and National Causes of Child Mortality: 158 WORLD DEVELOPMENT REPORT 2015 An Updated Systematic Analysis for 2010 with Time Quality Performance of Health Workers in Low- Trends since 2000.” Lancet 379 (9832): 2151–61. Resource Settings?” Lancet 366: 1026–35. Mills, Edward J., Jean B. Nachega, Iain Buchan, James Shampanier, Kristina, Nina Mazar, and Dan Ariely. 2007. Orbinski, Amir Attaran, Sonal Singh, Beth Rachlis, “Zero as a Special Price: The True Value of Free Prod- Ping Wu, Curtis Cooper, and Lehana Thabane. 2006. ucts.” Marketing Science 26 (6): 742–57. “Adherence to Antiretroviral Therapy in Sub-Saharan Spears, Dean, Arabinda Ghosh, and Oliver Cumming. 2013. Africa and North America: A Meta-Analysis.” JAMA 296 “Correction: Open Defecation and Childhood Stunting (6): 679–90. in India: An Ecological Analysis of New Data from 112 Naugle, Danielle A., and Robert C. Hornik. 2014. “Sys- Districts.” PLoS ONE 8 (9): September 23. http://www tematic Review of the Effectiveness of Mass Media .plosone.org/article/info%3Adoi%2F10.1371%2Fjournal Interventions for Child Survival in Low- and Middle- .pone.0073784. Income Countries.” Journal of Health Communication: Tendler, Judith. 1997. Good Government in the Tropics. Balti- International Perspectives 19 (Supplement 1): 190–215. more: Johns Hopkins University Press. Noar, Seth M., and Rick S. Zimmerman. 2005. “Health Thaler, Richard H., and Cass R. Sunstein. 2008. Nudge: Behavior Theory and Cumulative Knowledge Regard- Improving Decisions about Health, Wealth, and Happiness. ing Health Behaviors: Are We Moving in the Right New Haven: Yale University Press. Direction?” Health Education Research 20 (3): 275–90. Tomlinson, M., M. J. Rotheram-Borus, L. Swartz, and Patil, S. R., B. F. Arnold, A. L. Salvatore, B. Briceno, S. Gan- A. C. Tsai. 2013. “Scaling Up mHealth: Where Is the Evi- guly, J. Colford Jr., and P. J. Gertler. 2014. “The Effect dence?” PLoS Med 10 (2): e1001382. doi: 10.1371/journal of India’s Total Sanitation Campaign on Defecation .pmed.1001382. Behaviors and Child Health in Rural Madhya Pradesh: Topp, Stephanie M., Julien M. Chipukuma, Matimba M. A Cluster Randomized Controlled Trial.” PLoS Med 11(8): Chiko, Chibesa S. Wamulume, Carolyn Bolton-Moore, e1001709. doi: 10.1371/journal.pmed.1001709. and Stewart E. Reid. 2011. “Opt-Out Provider-Initiated Peltzer, K., H. Onya, P. Seoka, F. M. Tladi, and R. N. Malema. HIV Testing and Counselling in Primary Care Out- 2002. “Factors at First Diagnosis of Tuberculosis Asso- patient Clinics in Zambia.” Bulletin of the World Health ciated with Compliance with the Directly Observed Organization 89 (5): 328–35. Therapy (DOT) in the Limpopo Province, South Africa.” UNICEF (United Nations Children’s Fund) and WHO Curationis 25 (3): 55–67. (World Health Organization). 2009. Diarrhoea: Why Pop-Eleches, Cristian, Harsha Thirumurthy, James Hab- Children Are Still Dying and What Can Be Done. New York: yarimana, Joshua Graff Zivin, Markus Goldstein, UNICEF and WHO. Damien de Walque, Leslie Mackeen, Jessica Haberer, Vidanapathirana, J., M. J. Abramson, A. Forbes, and C. Fair- John Sidle, and Duncan Ngare. 2011. “Mobile Phone ley. 2005. “Mass Media Interventions for Promoting Technologies Improve Adherence to Antiretroviral HIV Testing.” Cochrane Database of Systematic Reviews Treatment in Resource-Limited Settings: A Random- July 20 (3): CD004775. ized Controlled Trial of Text Message Reminders.” Wakefield, Melanie A., Barbara Loken, and Robert C. AIDS 25 (6): 825–34. Hornik. 2010. “Use of Mass Media Campaigns to Price, Michael. 2013. “Stars Offer Incentive to Quit.” San Change Health Behaviour.” Lancet 376 (9748): 1261–71. Diego State University News Center. http://newscenter Wasi, Prawase. 2000. “Triangle That Moves the Mountain .sdsu.edu/sdsu_newscenter/news.aspx?s=74629. and Health Systems Reform Movement in Thailand.” Renfrew, Mary J., Felicia M. McCormick, Angela Wade, Human Resources for Health Development Journal 4 (2): Beverley Quinn, and Therese Dowswell. 2012. “Support 106–10. for Healthy Breastfeeding Mothers with Healthy-Term Wibulpolprasert, Suwit, Vichai Chokevivat, Cecilia Oh, Babies.” Cochrane Database of Systematic Reviews May 16 and Inthira Yamabhai. 2011. “Government Use Licenses (5): CD001141. doi: 10.1002/14651858.CD001141.pub4. in Thailand: The Power of Evidence, Civil Movement Roura, M., D. Watson-Jones, T. M. Kahawita, L. Ferguson, and Political Leadership.” Globalization and Health 7 and D. A. Ross. 2013. “Provider-Initiated Testing and (1): 1–8. Counselling Programmes in Sub-Saharan Africa: A Zimmerman, Rick S., and Dee Vernberg. 1994. “Models of Systematic Review of Their Operational Implementa- Preventative Health Behavior: Comparison, Critique, tion.” AIDS 27 (4): 617–26. and Meta-Analysis.” Advances in Medical Sociology 4: Rowe, A. K., D. de Savigny, C. F. Lanata, and C. G. Victora. 45–67. 2005. “How Can We Achieve and Maintain High- 9 CHAPTER Climate change English settlers to the New World believed that the land were anomalously cold, including the absence climate of Newfoundland would be moderate, New of people with good character. “He is wretched that England would be warm, and Virginia would be like believes himself wretched,” scoffed one writer (quoted southern Spain. They based these beliefs on the seem- in Kupperman 1982, 1283). ingly commonsense view that climate is much the Eventually, the English settlers did adjust their same at any given latitude around the globe. mental models about North American climate. The What is striking is that these views persisted despite accumulation of scientific data, combined with per- mounting evidence to the contrary. As late as 1620, sonal experience, was undeniable. But the adjustment after 13 years in the settlement, residents in James- was slow and costly, both in money and in lives lost. town, Virginia, were still trying to import olive trees Mental models about climate do not change easily. and other tropical plants, perhaps inspired by Father Responding to climate change is one of the defin- ing challenges of our time. There is massive scientific evidence that human activity is changing the earth’s An important role for psychological climate, with consequences that may be disruptive— potentially even catastrophic.1 Evidence on climate and social insights is to identify ways to change and its risks are reported in the technical summary from the 2014 Assessment Report of the Inter- convince populations to support, and governmental Panel on Climate Change (Stocker and others 2013; Field and others 2014). This material is governments to adopt, effective economic widely considered to be the most authoritative review of scientific knowledge on climate change. To briefly tools, such as carbon pricing, to curb paraphrase, in the history of modern civilization, the earth’s surface has never been so hot. Glaciers are greenhouse gas emissions. already disappearing, and the ice masses of Green- land are melting. Depending on how much carbon is released into the air, sea levels will rise significantly Andrew White, who had assured them that it was in this century, potentially flooding coastal cities and “probable that the soil will prove to be adapted to all submerging low-lying islands. Temperatures will rise the fruits of Italy, figs, pomegranates, oranges, olives, and likely become more variable. Rainfall patterns also etc.” Captain John Smith, whose books and maps will change, with more and heavier rains in some areas helped encourage English colonization of the Ameri- and more intense and longer droughts in others. cas, predicted that the crops of all the richest parts of The causes of climate change are clear. Greenhouse the world would be grown in New England. Settlers gases trap heat from the sun that would otherwise continued to arrive in Newfoundland despite early escape Earth. The levels of greenhouse gases (such failures. Investors and settlers resorted to ever more as carbon dioxide, methane, and nitrous oxide) are complex explanations for why winters in Newfound- too high. Carbon dioxide is released largely from CLIMATE CHANGE 161 burning fossil fuels and deforestation, while methane thinking, cognitive illusions, mental models, and social and nitrous oxide—which are more powerful green- norms contribute to behavior. It is clear that taxes on house gases than carbon dioxide—are released from carbon emissions, property rights in carbon abate- agriculture (growing crops and raising grass-eating ment, redistributive transfers, or other changes in eco- and grain-eating livestock). Never in the past 800,000 nomic incentives will be required to address climate years have concentrations of greenhouse gases in the change adequately. This chapter argues, however, that atmosphere been so high. These phenomena explain economic incentives are not the whole story and that changes in weather patterns, ice melts, the already vis- inertia in behavior arises from psychological and ideo- ible rise in sea levels, and many other factors, such as logical sources as well. At the same time, the chapter increases in seawater acidity. examines the prospects for invoking social norms and Changes in climate create a variety of risks affecting other communication strategies both to change behav- human well-being (Stocker and others 2013; Field and ior and to generate support for various policies—such others 2014). The likelihood and severity of these risks as carbon prices, cap-and-trade systems, and financial will depend on the amount of additional greenhouse transfers for lower emissions—that would be needed gases added to the atmosphere and on the extent to to overcome the inertia embodied in physical capital. which individuals and organizations take steps to mit- In other words, an important role for psychological and igate and adapt to the risks. While climate change is a social insights is to identify ways to convince popula- global threat, it is of most danger to developing coun- tions to support, and governments to adopt, what are tries, which are both more exposed to its impact and known to be effective economic tools, such as carbon less well equipped to deal with it (World Bank 2012). pricing, to curb greenhouse gas emissions. The World Development Report 2010 (World Bank 2009) describes three sources of inertia that make Cognitive obstacles inhibit responding to climate change such a pressing and dif- action on climate change ficult challenge. The first is inertia in the environment itself. Even if greenhouse gas emissions are stabilized Biases affect how people process complex over the next 100 years, biological and geophysical information feedback loops will cause increases in temperatures Climate is usually understood as the weather condi- and sea levels and other climatic changes to continue tions prevailing in an area over a long period. It is a for centuries—in some cases even millennia. Second, long-term pattern of variations among meteorological inertia embodied in physical capital, as well as in cur- variables, including average temperature and variabil- rent research and development streams, dramatically ity across time in rainfall. Grasping climate change affects the cost of reducing emissions. Retiring, retro- and its socioeconomic impacts requires a shift from fitting, and replacing power plants and other machin- automatic and associative to deliberative and analytic ery using high-carbon energy sources the world over thinking. The paradigmatic time period for identifying will require significant investments and substantial variations in climate—a 30-year window—is much social and technological coordination. more easily examined with long-term data sets and Finally, there is inertia in the behavior of individuals computer modeling techniques than with personal and organizations. In the industrialized world, people memories and conversations. Because analytic think- have grown accustomed to driving particular kinds ing is hard and attention is costly, people tend to use of automobiles, living and working in comfortable mental shortcuts to evaluate the evidence on climate indoor temperatures, and raising and eating methane- change and its risks. emitting animals. Many people in developing coun- Typically, how people think about climate change tries also engage in “high-carbon behaviors,” or they is subject to the availability heuristic (Marx and Weber aspire to do so. In addition, farmers around the world 2012). The term refers to the human tendency to judge grow crops that may be unsuited to a changing climate, an event by the ease with which examples of the event households settle in vulnerable zones, and builders use can be retrieved from memory or constructed anew. construction methods not designed to optimize energy A number of studies present strong evidence that a efficiency. Finally, political parties in many countries recent pattern of warm weather affects beliefs in cli- depend on fuel subsidies to garner support, and gov- mate change. For each 3.1 degrees Fahrenheit increase ernments fear the implications, for the economy or tax in local temperatures above normal in the week before revenues, of changes in energy policies. being surveyed, Americans become one percentage This chapter presents ideas related to that last cate- point more likely to agree that there is “solid evidence” gory of inertia—the behavior of individuals and organi- that the earth is getting warmer—an effect size com- zations. For the most part, it focuses on how automatic parable to that of age and education but less than the 162 WORLD DEVELOPMENT REPORT 2015 influence of political party identification and ideol- “expressive rationality.” One explanation for inaction ogy on assessments of scientific evidence (Egan and on climate change is that it is a complex problem and Mullin 2012). People typically do not systematically that more information better explained will raise update their views over months and years but rather concern and trigger action. On that view, which the express views based on what they have experienced authors call the Scientific Communication Thesis, recently. Eventually, memories of personal experiences perceptions of risk should increase as numeracy could become a reliable indicator that the climate has and scientific literacy increase. Figure 9.1, panel a, changed, but this adjustment may be slow, given the presents this prediction visually. Risk perceptions are inertia of the climate system and the nature of people’s based on responses to the question, How much risk beliefs. Assuming that adjusting a mental model of do you believe climate change poses to human health, climate requires three consecutive years in which the safety, or prosperity? The study uses a battery of stan- maximum temperature is a full standard deviation or dard questions to measure numeracy and scientific more above the historical high, Szafran, Williams, and literacy. The authors assess political worldviews along Roth (2013) calculate, using a simulation based on U.S. two standard dimensions: individualism (a belief weather station data from 1946 to 2005, that it will take that government should avoid affecting individual the majority of people up to 86 years to adjust their choice) and egalitarianism (support for equality and mental models—too late for policies aiming to forestall nondiscrimination). climate disruption. What they find is that perceptions of climate Generally speaking, grasping climate change is chal- change risk actually decline as scientific literacy and lenging because it requires understanding complex numeracy increase (figure 9.1, panel b). That decline aspects of both mathematics and atmospheric chemis- is attributable to the decline in risk perceptions try, including probabilities, recognizing the difference among a subset of people who support individualism between the flow of greenhouse gases and the existing and oppose egalitarianism (whom the authors call stock in the air, and appreciating feedback loops and “hierarchical individualists”). The authors argue that time lags. As in preventive health, the immediate and people may use their scientific knowledge defensively, direct effects of risky behaviors are often invisible. In identifying and resisting efforts to convince them to light of this, literature reviews in science communica- go against their allegiances. tion emphasize that “mere transmission of informa- The way people respond to scientific communica- tion in reports and presentations is not enough” and tion about climate change seems to depend on whether, that interactive, transparent simulations of the climate and how, messages trigger group identities and use may be more valuable (Sterman 2011, 821). charged language. For instance, the use of the word “tax” leads more individuals to focus on cheap options Cultural worldviews and social networks with lower environmental benefits, but the term “off- inform opinions set” does not have that effect. Moreover, when people Crucially, however, people interpret scientific infor- choose between otherwise identical products or ser- mation in light of their cultural worldviews, obtain vices, whether a surcharge for emitted carbon dioxide information through social networks and favored is framed as a tax or as an offset changes preferences media channels, and rely on trusted messengers to for some political groups but not for others (Hardisty, make sense of complex information. A number of Johnson, and Weber 2010). studies show that many people interpret evidence of This means that even more information, however climate change in the light of their worldviews and beautifully presented, might fail to move climate social networks. An individual’s level of support for change opinion in a politicized environment. Indeed— social hierarchy and equality is a better predictor of in a related fashion but on another topic—a recent his or her perceptions of changes in temperature over survey experiment found that presenting information, the past few years than actual temperature changes, as scientifically ratified data, images, and personal nar- Goebbert and others (2012) demonstrate, drawing on ratives all failed to convince people that the measles, an account developed by Douglas and Wildavsky (1983) mumps, and rubella vaccine is safe. Parents who were of how worldviews affect risk perceptions. already anxious about vaccine safety became less It may be that people use their rational facul- likely to have their children vaccinated after receiving ties not primarily to understand the world but any of those four modes of intervention (Nyhan and to express solidarity with their group, Kahan, others 2014). Similarly, a recent study observed that Jenkins-Smith, and Braman (2011) and Kahan and in the United States, politically conservative individ- others (2012) argue, putting forward an account of uals were less likely to purchase a more expensive CLIMATE CHANGE 163 energy-efficient lightbulb labeled as environmentally Figure 9.1 Worldviews can affect perceptions of the risk friendly than to buy the identical product when it was posed by climate change unlabeled (Gromet, Kunreuther, and Larrick 2013). In general, scientific communication needs to be mindful While the Scientific Communication Thesis (panel a) predicts that perceptions of climate change risks increase as scientific literacy and numeracy improve, in of a potential boomerang effect, in which arguments trig- actuality risk perceptions remain unchanged or even decline (panel b), especially ger antagonistic responses by threatening the attach- for people with particular cultural worldviews. Individualism refers to a belief ment of individuals to their social groups (Dillard and that government should avoid affecting individual choice; communitarianism is its opposite. Egalitarianism refers to support for equality and nondiscrimination; Shen 2005) or lead to unexpected and worse outcomes hierarchy is the opposite. by highlighting low levels of support for what people believed to be a common social behavior (Schultz and a. Prediction according to the Scientific Communication Thesis others 2007). Higher perceived risk How the media portray a social problem can also have powerful effects. Assessing how frames affect support for altruistic policies in another domain, Iyen- gar (1990) shows that media presentations influence support for antipoverty policies. For example, epi- sodic coverage of poverty, usually focused on specific individuals, led people to blame individuals for being poor, but thematic coverage of antipoverty policies led people to think that the government was primarily Low science High science responsible for poverty. Similarly, stating that human literacy/numeracy literacy/numeracy activity is responsible for climate change dramatically increases support for actions that address it (Pew Research Center 2009). Again, although this line of work is suggestive, it is preliminary, and more work is needed to understand how normative frames affect support for action on climate change.2 It is also likely to be the case that frames need to be tailored to spe- Lower perceived risk cific audiences. For example, while students respond b. Actual response to messages about energy consumption presented in terms of carbon emissions (Spence and others 2014), Higher perceived risk middle-class families are more attuned to messages 1.00 emphasizing the financial cost of energy consumption (Simcock and others 2014). 0.75 Egalitarian Communication about climate change can draw on communitarian 0.50 local narratives. In parts of Brazil, India, Melanesia, Standard deviations and the Sahel, some residents believe that weather is a 0.25 reward for good human behavior or a punishment for bad human behavior. While these rewards and pun- 0 Low science High science ishments are believed to be channeled through a deity, literacy/numeracy literacy/numeracy other groups, like the Kalahari San, the Inuit, and the –0.25 indigenous Siberian, share similar beliefs without a reli- gious connection. These narratives of human influence –0.50 Hierarchical on the weather may provide foundations for present- individualist ing contemporary accounts of anthropogenic climate –0.75 change and informing dialogues among citizens and –1.00 scientists in different settings (Rudiak-Gould 2013). Lower perceived risk Source: Kahan and others 2012. Automatic cognitive processes affect how people interpret probabilities farmers in the Central Plateau of Burkina Faso correctly Sometimes strong ties to specific places or landscapes, perceived that “big rains” in their region have become and incentives to pay attention, may help people assess less frequent and dry years more common over the past changes in climate. For example, elders and subsistence 20 years. As one elderly man said, 164 WORLD DEVELOPMENT REPORT 2015 Now is not like before. It is the drought. Before As chapter 1 discussed, there are two distinct “sys- Independence, we could count on rain until tems” involved in cognitive processing: the automatic October and grow long-cycle millet that we system and the deliberative system. Human beings would harvest and leave in granaries out in our rely on both when processing probabilities. Tversky fields. Since then, there is not enough rain and and Kahneman (1982) showed that although most we can’t grow that kind of millet anymore. It people neglect information about background fre- has vanished from here. There is less rain now quencies, such as the fact that “15% of the taxis in a city and we grow different crops. (West, Roncoli, and are operated by the Blue Cab company,” they notice Ouattara 2008, 296) information that is case specific and information that is part of a narrative, such as the fact that “15% of the In that instance and locale, farmers were able, after taxi accidents in a city involve the Blue Cab company.” training, to adapt to variation in climate. But this The reason is that the automatic system (System 1) is response is far from uniform. In a study of whether highly attuned to situations of cause and effect; it is farmers in Zimbabwe shifted from maize to millet deployed when processing information about taxi acci- when seasonal rain forecasts changed, Grothmann dents but not for unadorned statements concerning and Patt (2005) found that farmers typically ignored relative frequencies. rainfall forecasts unless they were at the extremes— Researchers have used this insight to help individ- rainfall far above normal or far below normal. Figure uals make sense of forecasts about climate change. 9.2 depicts this tendency graphically. In particular, Analogies to how an injury to a star player would affect farmers’ forecasts of seasonal rainfall in the D range the odds of winning a football match between Argen- were indistinguishable from their forecasts of rainfall tina and Zimbabwe helped Zimbabwean farmers grasp in the B range. Farmers focused on the fact that in how the El Niño phenomenon might affect the odds of both instances both maize and millet harvests could a rainy season (Suarez and Patt 2004). Comparisons to be grown. They did not focus on the fact that millet the familiar task of predicting the gender of an unborn could be grown successfully under nearly all rainfall child helped Ugandan farmers understand the proba- amounts in the B range but under only a portion of bility distribution that underlies government-issued those in the D range, and vice versa for maize. Some weather forecasts (Orlove and Kabugo 2005). Concrete farmers also felt that the B and D ranges were equal images and comparisons to familiar experiences help because the forecasts expressed estimates using terms make concepts such as relative frequencies and condi- such as might or likely, rather than certainties. tional probabilities easier to absorb. Figure 9.2 Predicting the effect of rainfall forecasts on the success of growing familiar crops was difficult for farmers in Zimbabwe In a series of workshops, subsistence farmers in Zimbabwe were asked what crops they grew, given seasonal rainfall forecasts. Farmers said they grew maize when forecasts were in the D range, and they did not switch to millet when the forecast was in the B range, even though millet was more likely to be successful. Forecast rainfall (mm) 0 Below normal 300 About normal 625 Above normal A B C D E Unlikely success for Likely success for millet; Likely success for Likely success for maize; Unlikely success for maize and millet unlikely success for maize maize and millet unlikely success for millet maize and millet Success range for millet Success range for maize Source: Grothmann and Patt 2005. Note: mm = millimeters. CLIMATE CHANGE 165 The future is far off, and risk is emotional The common framing of climate change as an A key obstacle to action on climate change is the fact unsolvable global tragedy may also be contributing to that human beings focus intensely on the present and a sense of uncertainty and a lack of self-efficacy that discount concerns perceived to be in the far-off future, together disempower local action. Ostrom (2014, 107) such as climate change risks (see the discussions of argues that “the ‘problem’ has been framed so often present bias and psychological “distance” in chapter 6). as a global issue that local politicians and citizens But research indicates that the extent to which people sometimes cannot see that there are things that can undertake future-oriented actions depends not only be done at a local level that are important steps in the on cognitive processes but also on emotional ones; right direction.” While admittedly many such steps are furthermore, risk is not constant across activities but needed to deal with the truly global-scale challenge of rather is contextual. People process risk as a feeling mitigating greenhouse gas emissions, local action has rather than as a probability (Loewenstein and others considerable potential for reducing vulnerability to 2001). Because perception of risk and support for pol- climate change risks. icy are strongly influenced by experiences, emotions, imagery, and values (Leiserowitz 2006), climate change messaging might be more effective if it tugged at the emotions more often. Communication strategies can draw on However, too much doom and gloom may lower an individual’s sense of self-efficacy and reduce the local mental models. The presentation of motivation to act. People may have a “finite pool of worry” available to handle problems. For example, the climate forecasts can be more intuitive. proportion of Americans who viewed climate change as a “very serious” problem dropped from a two-year Institutions can take advantage of steady level of 47 percent to 35 percent during the global financial crisis (Pew Research Center 2009). In cooperative tendencies and social the domain of adaptation, a study of Argentine farm- ers showed that steps to cue more worry about global networks among policy makers and firms. warming decreased concern about the political situa- tion in Argentina (Hansen, Marx, and Weber 2004). Relatedly, Argentine farmers who were worried about Film and entertainment education can global warming were more likely to change some change opinions, but the effects may not aspect of their production practices (such as insurance last long or irrigation) but hardly ever undertook more than one Hoping for traction with busy people on a seemingly change. It was as if the farmers were eager to dismiss remote and global topic, some climate change cam- climate change worries in their own minds, believing paigners have turned to art and imagery. This may be that with one action they had addressed their problems useful. An experiment was done to coincide with the (Weber 1997). release of a movie called The Day after Tomorrow, which The ambiguous, difficult-to-quantify risks sur- depicts the impact and aftermath of catastrophic rounding climate change may also pose challenges. storms hitting major U.S. cities, including Los Angeles It has been argued that when people face risks of and New York—storms caused by a climate shift that unknown magnitude (ambiguous risks), they tend to ultimately brings on an ice age. The film had a sig- avoid making decisions (Ellsberg 1961; Shogren 2012). nificant impact on people’s belief in climate change, However, for some individuals, ambiguity can increase despite the fact that the climate shift shown in the the likelihood of taking precautionary measures. A movie is scientifically fallacious. Forty-nine percent of recent framed field experiment documented high lev- viewers surveyed said that seeing the movie increased els of risk aversion among coffee farmers in Costa Rica. their worry about global warming, while only 1 percent The study also found that, among farmers with clearly said it made them less worried (Leiserowitz 2004). identifiable preferences regarding ambiguity, twice as More generally, narrative communication structures many chose to adapt to the risk than not to adapt when may also play a key role in influencing an individual’s confronting ambiguous climate change risks (Alpizar, perception of risk and policy preferences, especially Carlsson, and Naranjo 2011). In other words, the fact through the vehicle of a “hero” character. In a 2013 that the risk was unknown induced more adaptation study, respondents exposed to climate change informa- than the corresponding situation with known risk. tion presented in a narrative structure—complete with 166 WORLD DEVELOPMENT REPORT 2015 a setting, characters (heroes and villains), a plot, and a (these are nicely summarized in Bernauer 2013). Every moral—were more inclined to view the hero and the country might want a global agreement to reduce car- hero’s preferred policy solution favorably (Jones 2014). bon emissions, but what it might desire even more is Three carefully tailored narratives were each designed for every other country to comply with the agreement to appeal to one particular worldview—egalitarian, and make the requisite economic sacrifices, while it hierarchical, and individualist—with a control group does not. Recognizing this, some countries may decide receiving objective climate facts in a bulleted list. Those to focus just on adapting to climate change, rather exposed to narrative structures were found to have than also taking steps to mitigate it; resources spent on retained more information from the story and were adaptation will benefit the country, whereas resources better able to draw emotional conclusions about groups spent on mitigation may provide little gain if other portrayed as either heroes or villains than the respon- countries do not live up to their end of the bargain. A dents in the control group. These results suggest that second barrier to an international agreement is that overt value statements, cultural symbolism, and strong the costs and benefits of reducing carbon emissions connections to individual or group “heroes” may be are not distributed equally. Poor countries and com- more effective forms of climate messaging than objec- munities are generally more vulnerable to the effects tive scientific communication strategies currently used of climate disruption and also bear significant costs in the mainstream media. during a transition to a low-carbon economy. Finally, just as countries cannot easily coordinate with one another, different political generations cannot coordi- Social norms and comparisons can be nate effectively. Even if people made sacrifices today, future political leaders might reverse course. used to reduce energy consumption. In addition, nations need to converge on a work- ing agreement, or at least an overlapping consensus, Information campaigns can be made more regarding fairness. Principles of fairness are the sub- ject of intense competition and controversy among effective and clear. Default settings can be nations and social groups. There are many ways to distribute the burdens of mitigating and adapting to used more widely. climate change; and there are several principles of dis- tributive justice underlying those distributions, from It is unclear, however, how long the effect of the idea that the people and countries with the most watching such a movie persists and whether people’s emissions should contribute the most to abating green- increased concern translates into action. A recent house gases (“polluter pays”), to strict egalitarianism of study of U.K. viewers of the climate change movie emissions rights on a per capita basis, to contributions The Age of Stupid found that people reported increased linked to income levels, to equal percentage reductions concern about climate change after viewing the movie, for each country. Thus finding a shared view of fair- as well as a greater sense of agency and motivation to ness that promotes climate action is a major obstacle act. When the moviegoers were surveyed again several (see, for instance, Gardiner and others 2010). weeks later, however, these effects had disappeared Moreover, efforts to identify an international stan- (Howell 2014). dard of fairness are complicated by the widespread One problem with movies and media campaigns human tendency to select principles of fairness that is that people often experience them individually, not happen to coincide with one’s interests (self-serving as political actors or in social groups. Only “organiza- bias). Drawing on a survey of participants in workshops tionally mobilized public opinion matters,” as Skocpol’s sponsored by the Intergovernmental Panel on Climate political history of climate change legislation suggests Change (IPCC), Lange and others (2010) show that (2013, 118). What is needed is not messaging with “sub- there is a strong correlation between the principles of liminal” appeal but a focus on networks and organi- distributive justice that negotiators endorse and their zations, which are the “real stuff” of politics, Skocpol national self-interests. Taking this a step further, Kriss argues. and others (2011) show that Chinese and U.S. students can agree how burdens for environmental challenges People understand fairness in should be distributed between two anonymous coun- self-serving ways tries but stake out very different views as soon as the International negotiations on climate are hampered countries are named as China and the United States. In by well-known problems related to collective action other words, people may be able to agree on a fairness CLIMATE CHANGE 167 principle, but their social allegiances and mental mod- els affect their moral reasoning. What psychological Figure 9.3 Democratic rules can achieve high levels of resource sustainability and social factors underlie individuals’ allegiances to fellow nationals, most of whom they will never meet? In a laboratory experiment, individuals made contributions to shared resources This is an intriguing topic on which more research is on behalf of future generations of players. Decisions of a small minority resulted in very few pools being sustained. When binding votes were used to make needed. One possibility is that prioritizing the interests decisions, all resource pools were sustained (panel a). Voting rules led more of fellow nationals is a social norm; in other words, peo- individuals to contribute because the rules reassured conditional cooperators ple prioritize conationals not from reasoned choice but that others also would have contributed (panel b). because that is what most people around them do and a. Resource pools sustained believe they should do (Baron, Ritov, and Greene 2013). No voting Democratic rules and laws can promote conditional cooperation Chapter 2 argued that most individuals are conditional Voting cooperators. In the context of global warming, this means that people would be more willing to take action 0 20 40 60 80 100 Percentage to address climate change if they could be assured that others will do the same. Hauser and others (2014) b. Individuals who contributed conduct a laboratory experiment in which individu- als make contributions to combat climate change on No voting behalf of “future generations” of players. They find, pessimistically, that even if most people are prepared to conserve public resources on behalf of future gen- Voting erations, those resources can nevertheless be ruined 0 20 40 60 80 100 by a small minority of people within a population who Percentage do not conserve resources. More optimistically, they Contributors Noncontributors find that conditional cooperation, in the form of binding democratic votes, can make a difference: by introduc- Source: Hauser and others 2014. ing democratic principles, the contributing majority can force the “selfish” minority to conserve. And even tic actors, such as civil society and courts, which then more interestingly, players increase contributions may impose costs on the state (Simmons 2009; Gauri to shared resources when they are assured that their 2011; Bernauer 2013). Participation may itself affect good behavior is being reciprocated by others; in other choices. According to Spilker’s findings (2012), devel- words, contributions increase because democratic vot- oping countries with higher levels of membership in ing brings conditional cooperators on board. As figure international organizations have lower levels of green- 9.3 shows, voting measures dramatically increased house gas emissions, controlling for time trends and the sustainability of resource pools in the laboratory a range of economic and political variables, although experiment. The implications are that many individu- issues of selection and causality remain to be worked als are indeed ready to sacrifice for the greater good if out in such analyses. The research in this area is pre- institutions can be crafted to take advantage of condi- liminary and suggestive, but intriguing. tional cooperation. Conditional cooperation can also be promoted by Psychological and social insights international law and international organizations. for motivating conservation Even if a body of law has weak enforcement mecha- nisms, as is the case in various domains of interna- Invoking social norms can reduce tional law, it can affect behavior when it expresses and consumption concentrates social meanings (Sunstein 1996). If inter- There have been several pioneering efforts to use social national climate agreements were entirely ineffective, norms to cut energy consumption and encourage countries would not hesitate to sign them; that many adoption of energy-conserving practices and technol- countries do avoid signing indicates that countries ogies. In a series of large-scale programs run in the regard noncompliance as potentially costly (Bernauer United States in partnership with the energy company 2013). International multilateral and bilateral agree- Opower, “home energy reports” were mailed to residen- ments can serve as a focal point for mobilizing domes- tial utility customers, providing them with feedback on 168 WORLD DEVELOPMENT REPORT 2015 how their own energy use compared to that of their readily lends itself to a story line (Kunreuther, Meyer, neighbors (as well as providing simple information and Michel-Kerjan 2013). This suggests that opinion about energy consumption). On average, this inter- leaders might be used to push individuals toward more vention reduced energy consumption by 2 percent— adaptive behaviors. equivalent to the effect of a short-run increase in While it appears that social norm-based policy electricity prices of about 11–20 percent (Allcott 2011). interventions can be cost effective and have a lasting Numerous other projects have found similar effects impact, careful attention to their design is critical. First, (see, for example, Ayres, Raseman, and Shih 2013; Dolan it is necessary to identify the relevant social norm. Evi- and Metcalfe 2013). dence from a study of the participation of hotel guests While these interventions elicit immediate energy in an environmental conservation program suggests conservation and behavior change in the short term, that messages appealing to social norms (such as “The consumers’ initial efforts to conserve tend to decrease majority of guests reuse their towels”) are more effec- over time. The longest-running study sites of the tive in encouraging conservation behavior than mes- Opower energy conservation program, for example, sages focusing on environmental protection. The most showed that consumers’ initial efforts began to decline effective messages (resulting in 49 percent reuse) are in less than two weeks (Allcott and Rogers 2014). those that refer to circumstances that are most closely However, as the interventions have been repeated and related to the current situation (such as “The majority more reports have been delivered, customers seem of guests in this room reuse their towels”) (Goldstein, to develop new consumption habits or acquire a new Cialdini, and Griskevicius 2008). stock of physical capital (purchasing more energy- Messages about social norms can also have unin- efficient lightbulbs, for example). Long-term impacts tended consequences; they can normalize undesirable persist. Overall, the intervention costs between 1.4 as well as desirable behaviors. Information campaigns and 1.8 cents per kilowatt hour of electricity saved. aimed at reducing undesirable behavior sometimes Commonly used energy conservation programs typi- unwittingly draw attention to the fact that a specific cally cost between 1.6 and 3.3 cents per kilowatt hour undesirable behavior is actually widespread (Cialdini (Allcott and Rogers 2014). A similar intervention found 2003). In an environmental context, it has been shown that impacts of an intervention centered on changing that visitors to Arizona’s Petrified Forest National Park social norms on residential water use could be detected who receive empirical information (“Many past vis- more than two years after residents received a message itors have removed the petrified wood from the park, (Ferraro, Miranda, and Price 2011). Spotlight 5 docu- changing the state of the Petrified Forest”) were likely ments how the city of Bogotá drew on social norms to to steal more petrified wood, whereas normative mes- reduce consumption during a water supply crisis. sages (‘‘Please don’t remove the petrified wood from the Social norms might also be used to motivate indi- park’’) helped reduce theft (Cialdini and others 2006). viduals to adapt to environmental risks. In a laboratory The use of messages expressing certain social norms simulation, individuals were asked to make improve- has also been shown to have a boomerang effect: mes- ments to homes to reduce their exposure to the risk sages about average neighborhood energy use have led of earthquakes. At the end of the experiment, each to energy savings among households with high levels person was paid the difference between the value of of energy consumption but have increased consump- his or her home and the amount of interest earned tion among those households already consuming at on money they did not invest in home improvements low rates. Adding a message about normative expec- minus the cost of repairs and the cost of damage. No tations was found to eliminate this boomerang effect one knew whether home repairs to reduce earthquake (Schultz and others 2007). Furthermore, there may be risk were cost effective or not, but each person could important complementarities between social norms observe the choices others made. Half the subjects and financial incentives; social comparison messages were placed in a world where repairs were cost effec- related to water consumption were found to be most tive and half were not. The major driver of individual effective in reducing consumption among the least decisions was the average level of investment made price-sensitive users, such as those consuming large by neighbors. Even players who were told that invest- amounts of water before the intervention (Ferraro and ments were 100 percent effective started copying their Price 2013). neighbors and investing less—probably because, as To be most effective, interventions such as these mentioned, unadorned probabilities may mean less to also benefit from careful targeting. Peer comparisons people than narratives, and the behavior of neighbors targeting energy conservation through the means CLIMATE CHANGE 169 of home electricity reports, for example, are two to tional but smaller impacts from information about four times more effective when sent to political lib- energy use or carbon emissions (Newell and Siikamäki erals than to conservatives (Costa and Kahn 2013). In 2013). However, evidence on the effectiveness of such contexts where environmental social norms are inef- interventions is mixed (Kallbekken, Sælen, and Her- fective, focusing on health-based messages related to mansen 2013). the dangers of climate change could provide a useful While disclosing information can have a significant alternative. impact on people’s behavior, it is important to consider Depending on the context, it may be useful to com- how that information is conveyed. If information is too plement private information with public information, abstract or vague, too detailed and complex, or poorly if feasible. Providing college students in residence halls framed, disclosing that information may be ineffective in the United States with private information on their in bringing about behavior change. As people’s atten- real-time energy use for appliances compared to their tion is a scarce resource, vivid and novel ways of pre- peers was ineffective in reducing energy consumption. senting information can capture the attention in ways However, students who also received an individual that abstract or familiar ones cannot (Sunstein 2013). conservation rating that was publicly available sig- Without careful design, information disclosure can be nificantly reduced their use of heating and cooling, not only ineffective and confusing but also potentially leading to a 20 percent drop in electricity consumption misleading and counterproductive. The widely used (Delmas and Lessem 2012). measure of fuel efficiency, “miles per gallon,” for exam- Finally, behavioral “barriers” to investments in ple, is generally not well understood and leads people energy-saving technologies apply to firms as well as to to undervalue the fuel and cost savings of replacing the individuals. A systematic literature review found that most inefficient vehicles (Larrick and Soll 2008). business investments in energy efficiency in coun- People often struggle to make decisions in situa- tries in the Organisation for Economic Co-operation tions of risk and uncertainty. Even when people do and Development (OECD) require very high rates of understand the risks and benefits of different actions, return—higher than for other investments with com- they are more likely to act on the basis of this informa- parable risks (Centre for Sustainable Energy 2012). tion if they are also provided with information about The review attributes this finding to organizational how to proceed (Nickerson and Rogers 2010; Milkman norms and to the lack of salience, for many firms, and others 2011). Identifying a specific plan of action of energy efficiency. To motivate firms, it advocates can thus have a significant impact on bringing about reframing energy efficiency and climate policy as social outcomes, as complex or vague information can a strategic benefit, rather than as a short-term cost lead to inaction, even when people understand the decision. risks and benefits associated with different choices. In practice, some informational campaigns may be Psychological and social insights can framed around climate change only indirectly. A recent make information campaigns and large-scale randomized controlled trial found that indicators more effective messages emphasizing the health-related impacts of Disclosing information is often viewed as a useful energy consumption were more effective in motivating policy tool in many different areas, including finance, energy conservation than similar messages focusing health, and the environment. A recent meta-analysis on potential cost savings (Asensio and Delmas 2014). of information-based energy conservation experi- Given the fact that both social norms and the extent ments quantifies the effectiveness of interventions, to which people try to conform differ according to evaluating evidence from 156 published field trials and social context, both the effectiveness and the particu- 525,479 study subjects between 1975 and 2012 (Delmas, lar features of such policies will vary. Similarly, efforts Fischlein, and Asensio 2013). It finds that average to replace fuel subsidies with social transfers, often electricity consumption is reduced by 7.4 percent in couched as reforms for efficiency or equity, would have the studies but also finds that this effect decreases significant effects on greenhouse gas emissions as well with increasing rigor of the study. A recent study of (Stocker and others 2013; Field and others 2014). The energy-efficiency labeling attempts to disentangle the IPCC Working Group 3 on Mitigation notes the polit- relative importance of different kinds of information. ical importance of emphasizing policies to “integrate Simple information on the economic value of saving multiple objectives” and produce “co-benefits.” energy was found to be most important in guiding Policies requiring carbon disclosure for companies, investments in energy-efficient technology, with addi- and then benchmarking company emissions, can 170 WORLD DEVELOPMENT REPORT 2015 capitalize on social motivations. The Carbon Disclo- missioned by the French government, Stiglitz, Sen, and sure Project (CDP) and the associated Climate Perfor- Fitoussi (2009) examined a wider variety of possible mance Leadership Index work in that manner. These economic indicators, with the goal of developing indi- kinds of public pressure may be effective: combined cators that better incorporate well-being metrics and with shareholder activism, participation in the CDP environmental sustainability. These are much more can increase shareholder value if the external business likely to be widely adopted if major economies make a environment is climate conscious (Kim and Lyon 2011). collective decision, perhaps through a body such as the Why a firm may choose to join a carbon disclo- OECD, to begin reporting them as part of their standard sure initiative is an intriguing area of study and one economic statistics. Once statistics like these become closely related to the establishment and emergence more readily available, peer comparison will follow. of social norms. One recent analysis of 394 European and Latin American corporations that chose to join Setting the default the United Nations Global Compact looked at three Default rules can help overcome procrastination behavioral influences on their institutions and stake- and inertia, promoting social goals while preserving holders: coercive, normative, and mimetic behaviors. Coer- people’s freedom of choice. “Green defaults” have been cive behavior—in this case, government regulation— tested for a number of policy interventions, including exerted little effective pressure on firm participation. choosing an electricity provider, conserving energy, Rather, it was the normative pressure from academia, and reducing food waste. Three related mechanisms as well as the mimetic pressure to imitate peer corpora- appear to contribute to the effectiveness of default tions listed on the New York Stock Exchange, that had rules: people’s inertia and tendency to procrastinate, a perceived implicit endorsement of the default rule, and the establishment of a reference point relative to which Climate change is such a large problem changes may be evaluated (Sunstein and Reisch 2013). In southern Germany, for example, the power com- that multiple, coordinated approaches will pany Energiedienst GmbH offered three separate tar- iffs: a default “green” tariff (which was also 8 percent be needed to address it. Psychological, cheaper than the previous tariff), a cheaper but less green tariff, and a greener but more expensive tariff. social, and cultural insights can make Almost everyone (94 percent of consumers) remained with the default tariff; only 4.3 percent switched to significant contributions. the cheaper option, less than 1 percent switched to the greener tariff, and the remainder switched to a differ- ent supplier (Pichert and Katsikopoulos 2008). While the strongest effects (Perez-Batres, Miller, and Pisani many people in Germany stated a preference for green 2011). These types of pressure may already be driving energy, the national average percentage of consumers new norms for social sustainability. actually choosing this kind of energy provider, under Social norms also operate on policy makers them- circumstances in which the “green” tariff was not the selves, who appear responsive not just to their con- default, was less than 1 percent for a long time. Defaults stituents but also to one another. What neighboring thus appear to have a powerful effect on social choices. jurisdictions do influences policy choice, as shown in Similar results have been reported in the United a number of policy domains, including the adoption of States, where more customers enroll in time-based rate vaccines, Washington Consensus policies, and carbon programs (designed to encourage smarter energy use) taxes (Gauri and Khaleghian 2002; Dobbin, Simmons, when these are offered on an opt-out rather than on and Garrett 2007; Krause 2011). The insight that coun- an opt-in basis. Participation rates among customers tries, companies, and localities care about their relative recruited using an opt-out approach were 84 percent, performance can be leveraged to generate political while only 11 percent of customers joined the program support. This is an instance of what has become known when recruitment was done on an opt-in basis (U.S. as “governance by indicators”—using metrics to create Department of Energy 2013). new forms of peer pressure to induce better gover- In developing such policy interventions, the ques- nance policies.3 At the macro level, alternatives to the tion arises, Which default should be chosen? Choos- measure of gross domestic product can offer countries ing an overly ambitious default might lead to greater clearer economic indicators of their stewardship of opt-out rates. A randomized controlled experiment of core resource stocks. In a wide-ranging report com- thermostat default settings for heating found that CLIMATE CHANGE 171 relatively small decreases in the default setting (1°C) Notes led to a greater reduction in the average setting chosen 1. A 2013 study on the evolution of the scientific con- than did large decreases in the default setting (2°C) sensus on man-made (anthropogenic) climate change (Brown and others 2013). (ACC) analyzed 11,944 peer-reviewed papers studying Defaults can be used to improve outcomes when “global climate change” or “global warming” from 2001 to 2011. Of the abstracts that took a stance on people faced with certain decisions choose not to make ACC, more than 97 percent agreed with the scientific an active choice. The power of defaults arises from the consensus, including more than 97 percent of authors fact that people’s behavior may not be determined by when asked. The authors concluded that “the number active choice most of the time. Evidence from a study of papers rejecting [ACC] is a minuscule proportion of of an eight-month period of compulsory electricity the published research, with the percentage slightly rationing in Brazil shows that the policy led to a persis- decreasing over time” (Cook and others 2013, 1). Clearly, tent reduction of electricity use, with consumption 14 for misperceptions about the occurrence of climate percent lower even 10 years after the period of ration- change and its potential threats to persist in light of the ing. Household data on ownership of appliances and body of evidence, there is more at work here; psycho- consumption habits indicate that habits have been the logical, cultural, and political factors are likely at play main source of the persistent reduction in electricity (Norgaard 2009). consumption (Costa 2013). 2. Indeed, a number of studies (notably Small, Loewen- stein, and Slovic 2005) find that when people are shown that specific individuals are suffering, they are Conclusion more likely to be generous, seemingly contradicting Dan Ariely (2010, 251) notes that “if we tried to manu- the findings by Iyengar (1990). The frames that moti- facture an exemplary problem that would inspire gen- vate personal generosity may be distinct from those eral indifference,” it would probably be climate change. that motivate support for public action. This is because climate change implicates several 3. For more on the concept, see Davis and others 2012. cognitive illusions. Climate changes slowly, whereas individuals’ judgments about the climate are based References on what they have perceived recently. Ideological and Allcott, H. 2011. “Social Norms and Energy Conservation.” social allegiances affect how communication about cli- Journal of Public Economics 95 (9): 1082–95. doi: DOI mate change is received. People tend to ignore or under- 10.1016/j.jpubeco.2011.03.003. appreciate information presented in probabilities, such Allcott, Hunt, and Todd Rogers. 2012. “The Short-Run and as forecasts for seasonal rainfall and other climate- Long-Run Effects of Behavioral Interventions: Experi- related variables. Human beings are far more con- mental Evidence from Energy Conservation.” Working cerned with the present than with the future, whereas Paper 18492, National Bureau of Economic Research, Cambridge, MA. many of the worst impacts of climate change could ————. 2014. “The Short-Run and Long-Run Effects of take place many years from now. Some of those risks Behavioral Interventions: Experimental Evidence remain ambiguous, and some people avoid action in from Energy Conservation.” American Economic Review the face of the unknown. When deciding how to share 104 (10): 3003–37. the burdens of responding to climate change, indi- Alpizar, Francisco, Fredrik Carlsson, and Maria A. Naranjo. viduals and organizations usually adopt principles of 2011. “The Effect of Ambiguous Risk, and Coordination fairness that serve their own interests. on Farmers’ Adaptation to Climate Change: A Framed At the same time, promising approaches to action on Field Experiment.” Ecological Economics 70 (12): 2317–26. climate change also draw on psychological and social Ariely, Dan. 2010. The Upside of Irrationality: The Unexpected insights. Communication strategies can incorporate Benefits of Defying Logic. New York: Harper Perennial. local mental models and narratives. The presentation Asensio, Omar, and Magali A. Delmas. 2014. “The Dynam- of climate forecasts can be more intuitive. Institutions ics of Information Framing: The Case of Energy Con- can be crafted to take advantage of conditional coop- serving Behavior.” UCLA Working Paper, University of California, Los Angeles. eration and social networks. Social norms and com- Ayres, Ian, Sophie Raseman, and Alice Shih. 2013. “Evi- parisons can be used to reduce energy consumption. dence from Two Large Field Experiments That Peer Information campaigns can be made more effective Comparison Feedback Can Reduce Residential Energy and clear. Default settings can be used more widely. Usage.” Journal of Law, Economics, and Organization 29 (5): Climate change is such a large problem that multiple, 992–1022. coordinated approaches will be needed to address it. Baron, Jonathan, Ilana Ritov, and Joshua D. Greene. 2013. Psychological, social, and cultural insights can make “The Duty to Support Nationalistic Policies.” Journal of significant contributions. Behavioral Decision Making 26 (2): 128–38. 172 WORLD DEVELOPMENT REPORT 2015 Bernauer, Thomas. 2013. “Climate Change Politics.” Annual ments on the Role of Incentives on Energy Conserva- Review of Political Science 16: 421–48. tion.” Discussion Paper, Centre for Economic Perfor- Brown, Zachary, Nick Johnstone, Ivan Haščič, Laura Vong, mance, London School of Economics, London. and Francis Barascud. 2013. “Testing the Effect of Douglas, Mary, and Aaron Wildavsky. 1983. Risk and Defaults on the Thermostat Settings of OECD Employ- Culture: An Essay on the Selection of Technological and ees.” Energy Economics 39: 128–34. Environmental Dangers. Berkeley: University of Califor- Centre for Sustainable Energy. 2012. “What Are the Factors nia Press. Influencing Energy Behaviours and Decision-Making Egan, Patrick J., and Megan Mullin. 2012. “Turning Per- in the Non-Domestic Sector? A Rapid Evidence Assess- sonal Experience into Political Attitudes: The Effect of ment.” U.K. Government Policy Paper, Department of Local Weather on Americans’ Perceptions about Global Energy and Climate Change, London. Warming.” Journal of Politics 74 (3): 796–809. Cialdini, Robert B. 2003. “Crafting Normative Messages to Ellsberg, Daniel. 1961. “Risk, Ambiguity, and the Savage Protect the Environment.” Current Directions in Psycho- Axioms.” Quarterly Journal of Economics 75 (4): 643–69. logical Science 12 (4): 105–09. Ferraro, Paul J., Juan Jose Miranda, and Michael K. Price. Cialdini, Robert B., Linda J. Demaine, Brad J. Sagarin, 2011. “The Persistence of Treatment Effects with Daniel W. Barrett, Kelton Rhoads, and Patricia L. Norm-Based Policy Instruments: Evidence from a Winter. 2006. “Managing Social Norms for Persuasive Randomized Environmental Policy Experiment.” Amer- Impact.” Social Influence 1 (1): 3–15. ican Economic Review 101 (3): 318. Cook, John, Dana Nuccitelli, Sarah A. Green, Mark Ferraro, Paul J., and Michael K. Price. 2013. “Using Non- Richardson, Bärbel Winkler, Rob Painting, Robert Way, pecuniary Strategies to Influence Behavior: Evidence Peter Jacobs, and Andrew Skuce. 2013. “Quantifying from a Large-Scale Field Experiment.” Review of Eco- the Consensus on Anthropogenic Global Warming in nomics and Statistics 95 (1): 64–73. the Scientific Literature.” Environmental Research Letters Field, C. B., and others. 2014. “Technical Summary.” In 8 (2): 024 (7 pp.). Climate Change 2014: Impacts, Adaptation, and Vulner- Costa, Dora L., and Matthew E. Kahn. 2013. “Energy Con- ability. Part A: Global and Sectoral Aspects. Contribution servation ‘Nudges’ and Environmentalist Ideology: of Working Group II to the Fifth Assessment Report of the Evidence from a Randomized Residential Electricity Intergovernmental Panel on Climate Change, edited by Field Experiment.” Journal of the European Economic C. B. Field, V. R. Barros, D. J. Dokken, K. J. Mach, M. D. Association 11 (3): 680–702. Mastrandrea, T. E. Bilir, M. Chatterjee, K. L. Ebi, Y. O. Costa, Francisco J. M. 2013. “Can Rationing Affect Long- Estrada, R. C. Genova, B. Girma, E. S. Kissel, A. N. Levy, Run Behavior? Evidence from Brazil.” Working Paper, S. MacCracken, P. R. Mastrandrea, and L. L. White. London School of Economics, London. Available at Cambridge, U.K., and New York: Cambridge University http://ssrn.com/abstract=2028684 or http://dx.doi.org Press. http://ipcc-wg2.gov/AR5/images/uploads/WGI /10.2139/ssrn.2028684. IAR5-TS_FGDall.pdf. Davis, Kevin, Angelina Fisher, Benedict Kingsbury, and Gardiner, Stephen, Simon Caney, Dale Jamieson, and Sally Engle Merry. 2012. Governance by Indicators: Global Power through Classification and Rankings. Vol. 1. Oxford, Henry Shue. 2010. Climate Ethics: Essential Readings. U.K.: Oxford University Press. New York: Oxford University Press. Delmas, Magali A., Miriam Fischlein, and Omar I. Asen- Gauri, Varun. 2011. “The Cost of Complying with Human sio. 2013. “Information Strategies and Energy Con- Rights Treaties: The Convention on the Rights of the servation Behavior: A Meta-Analysis of Experimental Child and Basic Immunization.” The Review of Interna- Studies from 1975–2012.” Working Paper, University tional Organizations 6 (1): 33–56. of California Center for Energy and Environmental Gauri, Varun, and Peyvand Khaleghian. 2002. “Immu- Economics. nization in Developing Countries: Its Political and Delmas, Magali, and Neil Lessem. 2012. “Saving Power to Organizational Determinants.” World Development 30 Conserve Your Reputation?” Working Paper, Institute (12): 2109–32. of the Environment and Sustainability, University of Goebbert, Kevin, Hank C. Jenkins-Smith, Kim Klockow, California, Los Angeles. Matthew C. Nowlin, and Carol L. Silva. 2012. “Weather, Dillard, James Price, and Lijiang Shen. 2005. “On the Climate, and Worldviews: The Sources and Conse- Nature of Reactance and Its Role in Persuasive Health quences of Public Perceptions of Changes in Local Communication.” Communication Monographs 72 (2): Weather Patterns.” Weather, Climate and Society 4 (2): 144–68. 132–44. Dobbin, Frank, Beth Simmons, and Geoffrey Garrett. 2007. Goldstein, Noah J., Robert B. Cialdini, and Vladas Griskevi- “The Global Diffusion of Public Policies: Social Con- cius. 2008. “A Room with a Viewpoint: Using Social struction, Coercion, Competition, or Learning?” Annual Norms to Motivate Environmental Conservation in Review of Sociology 33: 449–72. Hotels.” Journal of Consumer Research 35 (3): 472–82. Dolan, Paul, and Robert Metcalfe. 2013. “Neighbors, Gromet, Dena M., Howard Kunreuther, and Richard Knowledge, and Nuggets: Two Natural Field Experi- P. Larrick. 2013. “Political Ideology Affects Energy- CLIMATE CHANGE 173 Efficiency Attitudes and Choices.” Proceedings of the Reduce Losses from Natural Catastrophes.” In Behav- National Academy of Sciences 110 (23): 9314–19. ioral Foundations of Policy, edited by E. Shafir. Princeton, Grothmann, Torsten, and Anthony Patt. 2005. “Adaptive NJ: Princeton University Press. Capacity and Human Cognition: The Process of Indi- Kupperman, Karen Ordahl. 1982. “The Puzzle of the Amer- vidual Adaptation to Climate Change.” Global Environ- ican Climate in the Early Colonial Period.” American mental Change 15 (3): 199–213. Historical Review 87 (5): 1262–89. Hansen, James W., Sabine M. Marx, and Elke U. Weber. Lange, Andreas, Andreas Löschel, Carsten Vogt, and 2004. “The Role of Climate Perceptions, Expecta- Andreas Ziegler. 2010. “On the Self-Interested Use of tions, and Forecasts in Farmer Decision Making: The Equity in International Climate Negotiations.” Euro- Argentine Pampas and South Florida.” International pean Economic Review 54 (3): 359–75. Research Institute for Climate Prediction Technical Larrick, Richard P., and Jack B. Soll. 2008. “The MPG Illu- Report 04-01, Earth Institute, Columbia University, sion.” Science 320 (5883): 1593–94. New York. Leiserowitz, Anthony A. 2004. “Day after Tomorrow: Study Hardisty, David J., Eric J. Johnson, and Elke U. Weber. 2010. of Climate Change Risk Perception.” Environment: Sci- “A Dirty Word or a Dirty World? Attribute Framing, ence and Policy for Sustainable Development 46 (9): 22–39. Political Affiliation, and Query Theory.” Psychological ————. 2006. “Climate Change Risk Perception and Policy Science 21 (1): 86–92. Preferences: The Role of Affect, Imagery, and Values.” Hauser, Oliver P., David G. Rand, Alexander Peysakhovich, Climatic Change 77: 45–72. and Martin A. Nowak. 2014. “Cooperating with the Loewenstein, George F., Elke U. Weber, Christopher K. Future.” Nature 511 (7508): 220–23. Hsee, and Ned Welch. 2001. “Risk as Feelings.” Psycho- Howell, Rachel A. 2014. “Investigating the Long-Term logical Bulletin 127 (2): 267. Impacts of Climate Change Communications on Marx, Sabine M., and Elke U. Weber. 2012. “Decision Mak- Individuals’ Attitudes and Behavior.” Environment and ing under Climate Uncertainty: The Power of Under- Behavior 46 (1): 70–101. standing Judgment and Decision Processes.” In Climate Iyengar, Shanto. 1990. “Framing Responsibility for Politi- Change in the Great Lakes Region: Navigating an Uncertain cal Issues: The Case of Poverty.” Political Behavior 12 (1): Future, edited by Thomas Dietz and David Bidwell, 19–40. 99–128. East Lansing: Michigan State University Press. Jones, M. D. 2014. “Cultural Characters and Climate Milkman, K. L., J. Beshears, J. J. Choi, D. Laibson, and Change: How Heroes Shape Our Perception of Climate B. C. Madrian. 2011. “Using Implementation Inten- Science.” Social Science Quarterly 95:  1–39. doi:  10.1111 tions Prompts to Enhance Influenza Vaccination /ssqu.12043. Rates.” Proceedings of the National Academy of Sciences Kahan, Dan M., Hank Jenkins-Smith, and Donald Braman. 108 (26): 10415–20. 2011. “Cultural Cognition of Scientific Consensus.” Newell, Richard G., and Juha V. Siikamäki. 2013. “Nudging Journal of Risk Research (2) 14: 147–74. Energy Efficiency Behavior: The Role of Information Kahan, Dan M., Ellen Peters, Maggie Wittlin, Paul Slovic, Labels.” Working Paper 19224, National Bureau of Eco- Lisa Larrimore Ouellette, Donald Braman, and Gre- nomic Research, Cambridge, MA. gory Mandel. 2012. “The Polarizing Impact of Science Nickerson, David W., and Todd Rogers. 2010. “Do You Literacy and Numeracy on Perceived Climate Change Have a Voting Plan? Implementation Intentions, Voter Risks.” Nature Climate Change 2: 732–35. Turnout, and Organic Plan Making.” Psychological Sci- Kallbekken, Steffen, Håkon Sælen, and Erlend A. T. Her- ence 21 (2): 194–99. mansen. 2013. “Bridging the Energy Efficiency Gap: A Norgaard, Kari Marie. 2009. “Cognitive and Behavioral Field Experiment on Lifetime Energy Costs and House- Challenges in Responding to Climate Change.” Policy hold Appliances.” Journal of Consumer Policy 36 (1): 1–16. Research Working Paper 4940, World Bank, Washing- Kim, Eun-Hee, and Thomas Lyon. 2011. “When Does ton, DC. Institutional Investor Activism Increase Shareholder Nyhan, Brendan, Jason Reifler, Sean Richey, and Gary L. Value? The Carbon Disclosure Project.” B. E. Journal of Freed. 2014. “Effective Messages in Vaccine Promotion: Economic Analysis and Policy 11 (1). http://www.bepress A Randomized Trial.” Pediatrics 133 (4): e835–42. .com/bejeap/vol11/iss1/art50. Orlove, B. S., and Merit Kabugo. 2005. “Signs and Sight in Krause, Rachel M. 2011. “Policy Innovation, Intergovern- Southern Uganda: Representing Perception in Ordi- mental Relations, and the Adoption of Climate Protec- nary Conversation.” Etnofoor 18 (1): 124–41. tion Initiatives by U.S. Cities.” Journal of Urban Affairs 33 Ostrom, Elinor. 2014. “A Polycentric Approach for Coping (1): 45–60. with Climate Change.” Annals of Economics and Finance Kriss, Peter H., George Loewenstein, Xianghong Wang, 15 (1): 97–134. and Roberto A. Weber. 2011. “Behind the Veil of Igno- Perez-Batres, Luis A., Van V. Miller, and Michael J. Pisani. rance: Self-Serving Bias in Climate Change Negotia- 2011. “Institutionalizing Sustainability: An Empirical tions.” Judgment and Decision Making 6 (7): 602–15. Study of Corporate Registration and Commitment to Kunreuther, Howard, Robert Meyer, and Erwann the United Nations Global Compact Guidelines.” Jour- Michel-Kerjan. 2013. “Overcoming Decision Biases to nal of Cleaner Production 19 (8): 843–51. 174 WORLD DEVELOPMENT REPORT 2015 Pew Research Center. 2009. Modest Support for “Cap and Social Progress Revisited.” Reflections and Overview. and Trade” Policy: Fewer Americans See Solid Evidence Paris: Commission on the Measurement of Economic of Global Warming. Washington, DC: Pew Research Performance and Social Progress. Center. Stocker, T. F., and others. 2013. “Technical Summary.” In Pichert, Daniel, and Konstantinos V. Katsikopoulos. Climate Change 2013: The Physical Science Basis. Contribu- 2008. “Green Defaults: Information Presentation and tion of Working Group I to the Fifth Assessment Report of Pro-Environmental Behaviour.” Journal of Environmental the Intergovernmental Panel on Climate Change, edited by Psychology 28 (1): 63–73. T. F. Stocker, D. Qin, G.-K. Plattner, M. Tignor, S. K. Rudiak-Gould, Peter. 2013. Climate Change and Tradition in Allen, J. Bosc Hung, A. Nauels, Y. Xia, V. Bex, and P. M. a Small Island State: The Rising Tide. London: Routledge. Midgley, 33–115. Cambridge, U.K., and New York: Cam- Schultz, P. Wesley, Jessica M. Nolan, Robert B. Cialdini, bridge University Press. Noah J. Goldstein, and Vladas Griskevicius. 2007. “The Suarez, Pablo, and Anthony Patt. 2004. “Cognition, Cau- Constructive, Destructive, and Reconstructive Power tion, and Credibility: The Risks of Climate Forecast of Social Norms.” Psychological Science 18 (5): 429–34. Application.” Risk Decision and Policy 9 (1): 75–89. Shogren, Jason. 2012. “Behavioural Economics and Envi- Sunstein, Cass R. 1996. “On the Expressive Function of ronmental Incentives.” Environment Working Paper Law.” University of Pennsylvania Law Review 144 (5): 49, OECD, Paris. 2021–53. Simcock, Neil, Sherilyn MacGregor, Philip Catney, ————. 2013. “Behavioral Economics, Consumption, and Andrew Dobson, Mark Ormerod, Zoe Robinson, Simon Environmental Protection.” Regulatory Policy Pro- Ross, Sarah Royston, and Sarah Marie Hall. 2014. gram Working Paper RPP-2013-19, Mossavar-Rahmani “Factors Influencing Perceptions of Domestic Energy Center for Business and Government, Harvard Ken- Information: Content, Source and Process.” Energy nedy School, Harvard University, Cambridge, MA. Policy 65 (Feb.): 455–64. Sunstein, Cass R., and Lucia A. Reisch. 2013. “Green by Simmons, Beth A. 2009. Mobilizing for Human Rights: Inter- Default.” Kyklos 66 (3): 398–402. national Law in Domestic Politics. Cambridge, U.K.: Cam- Szafran, Robert F., Jerry L. Williams, and Jeffery E. Roth. bridge University Press. 2013. “If Local Weather Was Our Only Indicator: Skocpol, Theda. 2013. “Naming the Problem: What It Will Modeling Length of Time to Majority Belief in Climate Take to Counter Extremism and Engage Americans in Change.” Simulation and Gaming 44 (2/3): 409–26. the Fight against Global Warming.” Paper presented Tversky, A., and D. Kahneman. 1982. “Evidential Impact at the Harvard University symposium on the politics of Base Rates.” In Judgment under Uncertainty: Heuristics of America’s fight against global warming, Cambridge, and Biases, edited by D. Kahneman, P. Slovic, and A. MA, February 14. Tversky, 153–62. Cambridge, U.K.: Cambridge Univer- Small, Deborah A., George Loewenstein, and Paul Slovic. sity Press. 2005. “Can Insight Breed Callousness? The Impact of U.S. Department of Energy. 2013. “Assessment of Demand Learning about the Identifiable Victim Effect on Sym- Response and Advanced Metering.” Staff Report, Fed- pathy.” Unpublished manuscript. eral Energy Regulatory Commission, U.S. Department Spence, A., C. Leygue, B. Bedwell, and C. O’Malley. 2014. of Energy, Washington, DC. “Engaging with Energy Reduction: Does a Climate Weber, Elke U. 1997. “Perception and Expectation of Change Frame Have the Potential for Achieving Climate Change.” In Psychological Perspectives to Envi- Broader Sustainable Behaviour?” Journal of Environmen- ronmental and Ethical Issues in Management, edited tal Psychology 38: 17–28. by D. Messick, M. Bazerman, A. Tenbrunsel, and Spilker, Gabriele. 2012. “Helpful Organizations: Member- K. Wade-Benzoni, 314–41. San Francisco: Jossey-Bass. ship in Inter-Governmental Organizations and Envi- West, Collin Thor, Carla Roncoli, and Frederic Ouattara. ronmental Quality in Developing Countries.” British 2008. “Local Perceptions and Regional Climate Trends Journal of Political Science 42 (2): 345–70. on the Central Plateau of Burkina Faso.” Land Degrada- Sterman, John D. 2011. “Communicating Climate Change tion and Development 19 (3): 289–304. Risks in a Skeptical World.” Climatic Change 108 (4): World Bank. 2009. World Development Report 2010: Develop- 811–26. ment and Climate Change. Washington, DC: World Bank. Stiglitz, Joseph, Amartya Sen, and Jean-Paul Fitoussi. ————. 2012. Turn Down the Heat: Why a 4°C Warmer World 2009. “The Measurement of Economic Performance Must Be Avoided. Washington, DC: World Bank. 176 WORLD DEVELOPMENT REPORT 2015 Promoting water conservation in Colombia Behind every intervention lies an assumption about Second, the city government launched engaging human motivation and behavior. When a tunnel and entertaining campaigns to teach individuals providing water to the city of Bogotá, Colombia, par- the most effective techniques for household water tially collapsed in 1997, triggering a water shortage, conservation. The campaigns contained memora- Spotlight 5 the city government declared a public emergency ble slogans and organized 4,000 youth volunteers and initiated a communication program to warn to go throughout the city to inform people about the inhabitants of the threat of a crisis: 70 percent of the emergency and teach them effective strategies to city would be left without water if current water use reduce consumption (Formar Ciudad [city develop- was not reduced. ment plan], 1995–97). The mayor himself appeared The city’s strategy was based on the assump- in a TV ad taking a shower with his wife, explaining tion that if individuals were informed of the how the tap could be turned off while soaping and situation, they would adjust their behavior and suggesting taking showers in pairs. Catholic priests reduce usage—after all, no one wants to be without were explicitly asked to invite their communities water. But the assumption was wrong. In fact, the to join the cooperative efforts, which, in a religious city’s strategy increased water consumption. Many country, proved to be particularly effective. people did not change their behavior because they did not think they could make a difference and did A change in strategy, building on not know which steps were most important. Some people even started to stockpile water. conditional cooperation, helped Recognizing the mistake in its assumptions, the city government changed its strategy (Guillot create a new social norm to 2014). First, the government reminded people to take action by conserving water at times when they conserve. were most likely to overuse it. Stickers featuring a picture of a statue of San Rafael—which was the Third, the city government publicized informa- name of the emergency reservoir the city was rely- tion about who was cooperating and who was not. ing on after the tunnel collapse—were distributed The chief executive officer of the water company throughout the city. People were asked to place a personally awarded households with exceptional sticker by the faucet that a particular household, water savings a poster of San Rafael with the leg- office, or school used most frequently. The stick- end, “Here we follow a rational plan for using the ers made the need to conserve water at all times precious liquid.” These awards were made visible salient. Daily reports of the city’s water consump- in the media. Three months later, when a second tion were prominently published in the country’s tunnel collapsed in the reservoir, the city imposed major newspapers. The reports became a part of sanctions for despilfarradores (squanderers), those public discussions about the emergency. with the highest levels of overconsumption. While PROMOTING WATER CONSERVATION IN COLOMBIA 177 Figure S5.1 The story of Bogotá’s 1997 the sanctions were minor—squanderers had to water supply crisis participate in a water-saving workshop and were subject to an extra day of water cuts—they were In January 1997, a tunnel connecting Bogotá, the capital nevertheless effective because they targeted highly of Colombia, to its main supply of fresh water partially visible actors. Car-washing businesses, although collapsed, leaving the city dependent on a small collectively not a major source of water waste, were emergency reservoir. the primary targets. The assumption underlying the new strategy An emergency was declared. was that conservation would improve if the city cre- ated a greater scope for social rewards and punish- Water consumption ments that helped to reassure people that achieving Water demand in Bogotá (cubic meters/second) at first increased as 16 the public good—continued access to water—was citizens stockpiled likely (see chapter 2 for a fuller discussion of the water. Eventually, the 15 dynamic of conditional cooperation, which may city’s strategy led to have undergirded the success of the city’s revised a decrease in water 14 ’97 ’01 ’05 ’09 strategy). This time, the assumption was correct. consumption. Water Year The change in strategy helped to create a social consumption then norm of water conservation. By the eighth week of stayed low for several years. the campaign, citywide water savings had signifi- cantly exceeded even the most optimistic technical Mayor Antanas Mockus launched measures to change predictions. Moreover, the reductions in water use conservation norms among citizens. persisted long after the tunnel was repaired and the Spotlight 5 emergency had been addressed (see figure S5.1). Daily reports in newspapers became references for public This case study from Bogotá provides a real- discussion and featured personal experiences of citizens’ world example of how interventions that take into conservation efforts. The mayor even showered with his account conditional cooperation may be useful for wife in a TV ad to demonstrate a water-reduction strategy. achieving policy goals. Reference Guillot, Javier. 2014. “Achieving Long-Term Citywide Cooperation in Water Consumption Reduction: The Story of Bogotá’s 1997 Water Supply Crisis.” Source: El Espectador (Bogotá), February 28, 1997, A6. Background note prepared for the World Develop- ment Report 2015. –13.8% Citywide water savings peaked at 13.8% after 8 weeks. Per capita water usage remained lower than precrisis levels for more than a decade even when water cuts were implemented after a second tunnel collapsed. This suggests that the new social norms around conservation persisted over time. Avg. savings (%) 20 15 10 5 0 –5 Feb. Apr. June Sep. Nov. Jan. 1997 1998 Cooperation Water cuts PART 3 Improving the work of development professionals 10 CHAPTER The biases of development professionals Experts, policy makers, and development professionals and the beliefs and attitudes that shape their lives. A are also subject to the biases, mental shortcuts (heu- deeper understanding of the context yields policies ristics), and social and cultural influences described that more accurately “fit” local conditions and thus elsewhere in this Report. Because the decisions of devel- have a higher probability of succeeding. To put this opment professionals often can have large effects on assumption to the test, the 2015 team for the World other people’s lives, it is especially important that mech- Development Report (WDR 2015 team) collected data anisms be in place to check and correct for those biases examining how World Bank staff perceived the beliefs and influences. Dedicated, well-meaning professionals and attitudes of the poor across several measures and in the field of development—including government pol- compared their findings against the actual beliefs and icy makers, agency officials, technical consultants, and attitudes of a representative sample of individuals in developing countries. It is perhaps uncontroversial to suggest that World Development professionals can be Bank staff have a different worldview from others. World Bank staff are highly educated and relatively susceptible to a host of cognitive wealthier than a large proportion of the world. How- ever, it is interesting to note that while the goal of biases, can be influenced by their social development is to end poverty, development profes- sionals are not always good at predicting how poverty tendencies and social environments, and shapes mindsets. For example, although 42 percent of Bank staff predicted that most poor people in Nairobi, can use deeply ingrained mindsets when Kenya, would agree with the statement that “vaccines are risky because they can cause sterilization,” only 11 making choices. percent of the poor people sampled in Nairobi actually agreed with that statement. Overall, immunization coverage rates in Kenya are over 80 percent. There were frontline practitioners in the public, private, and non- also no significant differences in the responses of Bank profit sectors—can fail to help, or even inadvertently staff in country offices and those in headquarters or in harm, the very people they seek to assist if their choices responses of staff working directly on poverty relative are subtly and unconsciously influenced by their social to staff working on other issues. This finding suggests environment, the mental models they have of the poor, the presence of a shared mental model, not tempered and the limits of their cognitive bandwidth. They, too, by direct exposure to poverty. The disparity represents rely on automatic thinking and fall into decision traps. not simply knowledge gaps on the part of development Perhaps the most pressing concern is whether devel- professionals but a mistaken set of mental models for opment professionals understand the circumstances in how poverty can shape the mindsets of poor people. which the beneficiaries of their policies actually live This is crucially important since how development THE BIASES OF DEVELOPMENT PROFESSIONALS 181 professionals perceive the poor affects how develop- 2. Just referral. Refer to orthopedics and do not ment policy is created, implemented, and assessed. start any new medication. This chapter focuses on the kinds of automatic thinking that can compromise the effectiveness of The second half of doctors received a scenario that development professionals in light of the three main differed in just one respect. The doctor learned, just insights discussed throughout this Report. It argues before sending the patient to an orthopedic consul- that development professionals are susceptible to a tant, that there are two drug treatments (ibuprofen host of cognitive biases, are influenced by their social and piroxicam) that the patient has not yet tried. The tendencies and social environments, and use deeply respondent’s task in the second version was to choose ingrained mindsets when making choices. This chap- among three options:  ter reviews four kinds of challenges and the associated decision traps that affect them: the use of shortcuts 1. Ibuprofen + referral. As above. (heuristics) in the face of complexity; confirmation 2. Piroxicam + referral. Refer to orthopedics bias and motivated reasoning; sunk cost bias; and the and also start piroxicam. effects of context and the social environment on group 3. Just referral. As above. decision making. The challenge that development organizations face is how to develop better decision- More physicians chose the simplest option—“just making procedures and policy processes to mitigate referral”—in the second, more complicated version these problems. Improving these decision-support than in the basic version (72 percent versus 53 percent). environments can have a direct impact on policy out- Increasing the complexity of the problem may have comes simply by eliminating errors of reasoning. led physicians to skip over possibly effective medi- cines altogether. This happened to highly educated Complexity and experienced professionals who are dedicated to Development is a complex, messy, conflict-ridden their patients’ health. Development professionals who process. Its complexity may affect the kinds of deci- design and implement development projects grapple sions made by development professionals. The more with highly complex problems, too. That very com- complex a decision is, the more difficult it is to make. plexity gives rise to a special set of challenges (Rama- However, even the decisions in areas in which people lingam 2013). have expertise can be affected by the complexity of Many situations offer not only several options but the decision-making environment. As the number of also multiple ways of understanding those options. options increases, people’s ability to accurately evalu- How policy options are framed has a large effect on ate the different options declines. behavior. This is known as the framing effect (see chap- This point is demonstrated in an experiment by ters 1 and 3). One of the most famous demonstrations Redelmeier and Shafir (1995). Family physicians were of the framing effect was done by Tversky and Kahne- mailed a survey outlining a medical situation: a patient man (1981). They posed the threat of an epidemic to stu- suffered with chronic hip pain, and doctors were asked dents in two different frames, each time offering them to decide whether to put their patient on new medica- two options. In the first frame, respondents could tion. In the case received by the first half of doctors in definitely save one-third of the population or take a the sample, all prior drug treatments had failed. The gamble, where there was a 33 percent chance of saving problem was described in roughly this way (some lan- everyone and a 66 percent chance of saving no one. In guage is paraphrased for brevity, and labels are added the second frame, they could choose between a policy for emphasis): in which two-thirds of the population definitely would die or take a gamble, where there was a 33 percent You decide to refer the patient to an orthopedic chance that no one would die and a 66 percent chance consultant for consideration for hip replacement that everyone would die. Although the first and sec- surgery. He agrees to this plan. However, before ond conditions frame outcomes differently—the first sending him away, you learn that there is one in terms of gains, the second in terms of losses—the medication (ibuprofen) that the patient has not policy choices are identical. However, the frames yet tried.  Your task is to choose between two affected the choices students made. Presented with alternatives: the gain frame, respondents chose certainty; presented with a loss frame, they preferred to take their chances. 1. Ibuprofen + referral. Refer to orthopedics and The WDR 2015 team replicated the study with World also start ibuprofen. Bank staff and found the same effect. In the gain 182 WORLD DEVELOPMENT REPORT 2015 frame, 75 percent of World Bank staff respondents Wetherick 1962). Confirmation bias may arise from a chose certainty; in the loss frame, only 34 percent did. fundamental tendency of human beings to use reason Despite the fact that the policy choices are equivalent, for the purposes of persuading others and winning how they were framed resulted in drastically different arguments (Mercier and Sperber 2011). responses. Recent research has shown that cultural and polit- Faced with complex challenges, development ical outlooks affect how individuals interpret data. agencies seek to bring a measure of uniformity and Kahan and others (2013) present respondents with two order through the widespread application of standard versions of identical data—one framed in the context management tools—a process Scott (1998) calls “thin of a study on the effectiveness of a skin cream, the simplification.” This approach brings its own potential other on the effectiveness of gun control laws. Respon- for error in the opposite direction, as discussed later in dents are randomly assigned to one of the two frames. this chapter. The study assesses the numeracy of respondents, as One promising strategy for constructively address- well as their cultural and ideological outlooks. The ing complexity stems from the work of Weick (1984), authors find that for the case of skin cream, as might who proposes breaking down seemingly intractable be expected, the likelihood of correctly identifying the issues into more discrete problems, thereby gen- answer supported by the data goes up as numeracy erating an incremental set of “small wins.” Argyris increases and is not affected by cultural and political (1991) extends this insight to stress the importance, outlooks. However, in the case of gun control laws, for organizations, of a kind of learning in which not respondents are more likely to get the right answer only the means used but also the ends sought and when that answer corresponds to their cultural views strategies employed are reexamined critically; that than when it does not. Moreover, when the answer in effort entails learning not only from success but also the gun control law framing is consistent with ideol- from failure. More recent work by Andrews, Pritchett, ogy, numeracy helps (by boosting the odds of getting and Woolcock (2013) proposes incorporating such an the answer right), but when the answer is inconsistent approach more systematically into development oper- with ideology, numeracy has minimal impact. On top- ations. Rather than trying to grapple with problems at ics that are important for social and political identity, higher orders of abstraction or defining problems as individuals tend to engage in motivated reasoning, the the absence of a solution (for example, inadequately tendency to arrive at conclusions they like. trained teachers), decision makers instead are urged to To see if cultural cognition of this kind affects pursue a concerted process of problem identification: development experts, and not only the general pop- the most basic step is to identify the problem correctly. ulation used in the study by Kahan and others (2013), Then development professionals can work incremen- the WDR 2015 team implemented a very similar test by tally with counterparts to define a problem such that surveying World Bank staff.1 The team replicated the it becomes both an agreed-upon binding constraint to skin cream (neutral) frame, but replaced the gun con- reaching a certain set of goals and a manageable chal- trol law frame with one about the impact of minimum lenge that allows for some initial progress (for exam- wage laws on poverty rates—a controversial topic ple, enhancing student learning in the classroom). among development economists, whose views on the issue appear to be related to broader disciplinary and Confirmation bias political identities. When development professionals engage with proj- Using a sample of professional-level World Bank ects and other development problems, they bring with staff, stationed both in country offices and the Wash- them disciplinary, cultural, and ideological priors, leav- ington, D.C., headquarters, the team found that respon- ing them susceptible to confirmation bias. Confirmation dents are significantly less accurate when interpreting bias refers to the selective gathering of (or the giving of data on minimum wage laws than when interpreting undue weight to) information in order to support a pre- data on skin cream (figure 10.1), even though the data viously held belief (Nickerson 1998) and to the neglect presented are identical in each scenario. The differ- (or discounting) of information that does not support ences in accuracy are not explained by differences in those previously held beliefs. It arises when individu- cognitive ability or seniority. As in the study by Kahan als restrict their attention to a single hypothesis and and others (2013), there is, however, evidence of a rela- fail to actively consider alternatives (Fischhoff and tionship between ideology and accuracy. Respondents Beyth-Marom 1983). Once a particular hypothesis has were asked whether they were more likely to support been accepted, individuals selectively look for informa- the statement “Incomes should be made more equal” tion to support it (see, among others, Wason 1960, 1977; or the statement “We need larger income differences THE BIASES OF DEVELOPMENT PROFESSIONALS 183 as incentives for individual effort.” Respondents sup- cially, to evaluate critically the arguments that others porting income equality were significantly less accu- make. By creating argumentative and deliberative rate when the data presented showed that minimum environments, organizations can reduce the risk of wage laws raise poverty rates than they were when confirmation bias. Crucially, these processes require minimum wage laws were shown to lower poverty exposing people to individuals with different view- rates. This study illustrates that ideological outlooks points. Discussions among people who share similar affect the reasoning of highly educated development views can lead them to become more extreme in their professionals. Like most people, they tend to come up positions, as Schkade, Sunstein, and Hastie (2010) have with reasons for why the evidence supports their own shown. In those circumstances, hearing from others ideological commitments. only confirms the biases that people hold. The failure What can be done to overcome confirmation bias? to confront individuals with differing views can lead One of the best ways is to expose people to opposing to consistently biased decision making (box 10.1). views and invite them to defend their own. Individuals In short, group deliberation among people who readily argue and defend their views when exposed to disagree but who have a common interest in the truth opposition, but in the absence of a social setting that can harness confirmation bias to create “an efficient forces them to argue, individuals usually fall back on division of cognitive labor” (Mercier and Sperber 2011). their prior intuitions. Social settings can motivate In these settings, people are motivated to produce the people to produce more effective arguments and, espe- best argument for their own positions, as well as to critically evaluate the views of others. There is sub- stantial laboratory evidence that groups make more Figure 10.1 How development professionals consistent and rational decisions than individuals and interpreted data subjectively Identical sets of data were presented to World Bank staff, but in different frames. In one frame, staff were asked which of two Because the decisions of development skin creams was more effective in reducing a rash. In the other, they were asked whether or not minimum wage laws reduce poverty. Even though the data were identical, World Bank professionals often can have large effects respondents were significantly less accurate when considering the data for minimum wage laws than for skin cream. Views on on other people’s lives, it is especially whether minimum wage laws lower poverty tend to be related to cultural and political outlooks. Respondents supporting income equality were significantly less accurate when the data important that mechanisms be in place presented conflicted with their outlooks (and showed that minimum wage laws raise poverty rates) than they were when the data corresponded to their outlooks (and showed that to check and correct for their biases and minimum wage laws lower poverty rates). blind spots. 100 Percentage reporting correct choice are less “likely to be influenced by biases, cognitive 80 limitations, and social considerations” (Charness and Sutter 2012, 158). When asked to solve complex reason- ing tasks, groups succeed 80 percent of the time, com- pared to 10 percent when individuals are asked to solve 60 those tasks on their own (Evans 1989). By contrast, efforts to debias people on an individual basis run up against several obstacles, including the problem that 40 critical thinking skills appear to be domain specific and may not generalize beyond the particular exam- ples supplied in the debiasing efforts (Willingham 20 2007; Lilienfeld, Ammirati, and Landfeld 2009). Indeed, when individuals are asked to read studies whose conclusions go against their own views, they find so 0 many flaws and counterarguments that their initial Skin cream vignette Minimum wage vignette attitudes are sometimes strengthened, not weakened Source: WDR 2015 team survey of World Bank staff. (Lord, Ross, and Lepper 1979). 184 WORLD DEVELOPMENT REPORT 2015 Red teaming is an approach to fighting confirma- plans can be made more effective before resources are tion bias that has been a standard feature of modern wasted. Red teams are institutionally distinct from military planning. In red teaming, an outside team the policy makers themselves, which creates space for challenges the plans, procedures, capabilities, and more candor and critique. This approach has already assumptions of commanders in the context of partic- moved beyond military planning and into general ular operational environments, with the goal of taking government use, particularly for vulnerability analy- the perspective of partners or adversaries. This process sis. Red teaming encourages a culture of perspective is institutionalized in some military organizations.2 taking and independent adversarial analysis as part of Teams specialize in challenging assumptions. The goal a stakeholder assessment. is to avoid “groupthink,” uncover weaknesses in exist- This approach is broadly similar to the long-standing ing plans and procedures, and ensure that attention work of Fishkin (2009), who has sought to use open is paid to the context. It draws on the idea that indi- deliberative forums (citizens’ juries) to help citizens viduals argue more effectively when placed in social come to greater agreement (if not consensus) on oth- settings that encourage them to challenge one another. erwise polarizing issues. In his forums, citizens with In a development context, while there may not be different initial views on controversial issues, such adversaries, there are often a variety of stakeholders, as migration and regional trade agreements, are ran- each of whom comes with a different set of mental domly assigned to groups where they receive presenta- models and potentially different goals and incentives. tions by leading researchers on the empirical evidence Institutionalizing teams that review plans in an explic- in support of varying policy positions. Participants itly argumentative manner offers a greater chance that are encouraged to pose questions to presenters and to Box 10.1 The home team advantage: Why experts are consistently biased Even the best-trained, most experienced, and seemingly impartial pro- This dynamic goes a long way toward explaining the failure of other- fessionals can make systematically biased decisions. In a comprehensive wise competent and experienced regulatory officials in public finance to empirical analysis of major sports leagues, with important implications adequately anticipate and respond to the global financial crisis of 2008, for other professional arenas, Moskowitz and Wertheim (2011) find that Barth, Caprio, and Levine (2013) argue. In this case, the “home team” in all such sports, and especially during critical moments (for example, is—or became, over time—the private sector banks and allied financial at the end of close championship games), referees consistently favor industries, whose senior officials move in a “revolving door” between the the home team. Even though the referees in such games are the best highest levels of the public and private sectors (for example, between available—and, significantly, sincerely believe themselves to be utterly the U.S. Federal Reserve and Goldman Sachs). In social circles and pro- impartial in performing their duties in all circumstances—they nonethe- fessional gatherings, the people public officials thus most frequently less make decisions that give the home team a clear advantage. At the encountered—the people whose opinions were most proximate and end of soccer games, for example, referees have discretionary authority salient to these officials—were those from the private sector. Without to add a few extra minutes corresponding to the amount of time lost needing to question the professional integrity or competence of financial due to injuries and substitutions; they routinely add more time when the sector regulators, the public interest—and in particular, ordinary citizens home team is behind and less time when it is ahead. Similarly, in the final whose transactions depend on the security and solvency of the banks innings of championship baseball games, marginal calls on whether par- holding their deposits and mortgages—became, in effect, the perpet- ticular pitches are called as strikes or balls are made in favor of the home ual “away team,” with no one adequately voicing and protecting their team. Under pressure, in other words, even the best professionals make interests. When the pressure intensified—when the system started to demonstrably biased decisions. Why is this? Does this process play out in implode—only the home team continued to get the key calls. public policy? If so, what can be done about it? These types of problems cannot be adequately addressed by providing Notionally independent experts make consistently biased decisions at “more training” or “capacity building” for individuals, since this research decisive moments because they want to appease the passions—most shows compellingly that even the “best and brightest” favor the “home especially, to avoid the wrath—of those closest to them, Moskowitz and team,” however that team comes to manifest itself. A partial solution in Wertheim (2011) conclude. Put differently, the home team advantage professional sports, at least, has been the introduction of instant replay, stems not so much from home team players being more familiar with which has been shown empirically to improve the objective decision mak- the idiosyncrasies of their environment or the extra effort players make ing of referees: when referees know their actions are subject to instant in response to being cheered on by their more numerous and vocal and public scrutiny, often from multiple angles, their bias for the home supporters, but from those same supporters exerting pressure on other- team markedly declines. This chapter later presents approaches in which wise impartial officials to make fine, but deeply consequential, judg- development professionals might learn to view topics from multiple ment calls in their favor. No one wants to incur the displeasure of those angles and in which they, as well as others, examine and observe one around them. another, thus exposing and mitigating ingrained biases. THE BIASES OF DEVELOPMENT PROFESSIONALS 185 explore the finer points through discussion with one been made (Arkes and Blumer 1985). To stop a project another. Fishkin’s approach, which has been carried is to acknowledge that past efforts and resources have out in dozens of different country contexts on differ- been wasted; thus the bias may arise from the cultural ent policy issues, has been used to help citizens arrive admonition not to appear wasteful (even though, par- at more informed and reasoned views and to reduce adoxically, continuing a project that is questionable the degree of polarization between competing policy may incur needless costs). Actors less concerned with viewpoints. appearing wasteful, such as children and nonhuman Note that these approaches differ from standard animals, do not exhibit sunk cost bias (Arkes and peer review processes in development organizations. Ayton 1999). Examples in the field of engineering illus- For the most part, those who prepare concept notes, trate particularly well an escalating commitment to a appraisal documents, or program assessments are “failing course of action,” where individuals continue allowed to nominate their peer reviewers, thereby to support a project and cite sunk costs as the major infusing the entire process with a susceptibility to reason for doing so (Keil, Truex, and Mixon 1995). The confirmation bias. Authors will inevitably select sym- implications of this line of research are that policy mak- pathetic like-minded colleagues to review their work, ers are particularly sensitive to policies already put in who in turn not only are likely to assess the work action. Being held politically accountable for risk tak- through a similar lens but also know that, in time, the ing explains some of the sunk cost effects, particularly roles are likely to be reversed. The risk of confirmation the reluctance to experiment and try new ideas. bias could be reduced by including at least one “double- The WDR 2015 team investigated the susceptibility blind” peer reviewer in the assessment process: that is, of World Bank staff to sunk cost bias. Surveyed staff a person drawn at random from an appropriate pool were randomly assigned to scenarios in which they of “knowledgeable enough” reviewers, whose iden- assumed the role of task team leader managing a five- tity would remain anonymous and who (in principle) year, $500 million land management, conservation, would not know the name(s) of the author(s) of the and biodiversity program focusing on the forests of work he or she is assessing. a small country. The program has been active for four A final and related option is to require a stronger years. A new provincial government comes into office empirical case to be made up front about the likely and announces a plan to develop hydropower on the impacts of the proposed intervention, following from main river of the forest, requiring major resettlement. a clearly stated theory of change. Such a process would However, the government still wants the original proj- need to make a serious effort to integrate—and where ect completed, despite the inconsistency of goals. The necessary reconcile—evidence pointing in different difference between the scenarios was the proportion of directions (see Ravallion 2011). Agencies and devel- funds already committed to the project. For example, opment institutions like the World Bank should be in one scenario, staff were told that only 30 percent exercising due diligence in this domain by engaging ($150 million) of the funds had been spent, while in in a more robust debate with scholarly findings, where another scenario staff were told that 70 percent ($350 such findings exist. However, this approach should not million) of the funds had been spent. Staff saw only imply that the only proposals allowed to go forward one of the four scenarios. World Bank staff were asked are those formally and unambiguously verified by whether they would continue the doomed project by elite research. In addition to questions concerning the committing additional funds. external validity of studies, this approach would bias While the exercise was rather simplistic and clearly development projects toward areas in which it is easier did not provide all the information necessary to make a to conduct high-impact research. It would also stifle decision, it highlighted the differences among groups innovation (which by definition has uncertain impacts randomly assigned to different levels of sunk cost. As initially) and set unreasonable standards for function- levels of sunk cost increased, so did the propensity ing in the contexts in which most development work of the staff to continue. The data show a statistically takes place. Nor should this approach imply that par- significant linear trend in the increase in likelihood ticular methodologies are inherently privileged over of committing remaining funds. Staff also perceived others when determining “what works” (or is likely to their colleagues as being significantly more likely to work in a novel context or at a larger scale). continue to commit the remaining funds to the dying project (figure 10.2). This divergence between what Sunk cost bias individual staff say about their own choices and what Policy makers can also be influenced by the sunk cost they say about how other staff will behave is consis- bias, which is the tendency of individuals to continue tent with the existence of a social norm for disbursing a project once an initial investment of resources has funds for a dying project. 186 WORLD DEVELOPMENT REPORT 2015 choosing between welfare-oriented and labor-oriented Figure 10.2 How World Bank staff viewed sunk costs approaches to combating poverty. Most fundamentally, World Bank staff were asked if they would commit remaining funds to a dying to take a policy stance, policy makers must have some project. Staff were more likely to commit additional funds as the sunk costs knowledge about the decision context that exists in the increased. They also perceived their colleagues as being more likely to commit funds than they themselves were, which is consistent with the existence of a population. In the absence of knowledge or objective social norm for disbursing funds for a dying project. interpretation of that knowledge, automatic thinking, as well as thinking unduly influenced by social context 60 and cultural mental models, may prevail. Percentage committing funds In this regard, designing and implementing poli- 55 cies combating poverty are difficult in three respects. First, most policy makers have never been poor and 50 thus have never personally experienced the psycho- logical and social contexts of poverty or scarcity (see chapter 4); as a result, their decision-making processes 45 may differ from those of people living in poverty. An example of this gap is how development profession- 40 als, like other well-off people, think about trade-offs between time and money. The poor often exhibit more 35 classically rational behavior when it comes to making 30 50 70 90 such trade-offs, as Mullainathan and Shafir (2013) have Sunk costs (percentage of budget) argued. When presented with an option to save $50 on a $150 purchase by driving 45 minutes, a poor per- Likelihood of others committing funds son would take the option. He or she would also take Likelihood of self committing funds the option for a $50 savings on higher-priced goods. Source: WDR 2015 team survey of World Bank staff. Wealthier people, however, tend to be less inclined to save $50 as the base price goes up. Although the deal is always the same—$50 for 45 minutes—the percentage How might organizations mitigate sunk cost effects? discount goes down. The wealthy respond to the dis- The basic principle is to avoid the judgment that to cut count rate, whereas the poor respond to the absolute off a dying project is to waste resources. When indi- value of the monetary savings. viduals can justify why they have “wasted” resources, The WDR 2015 team replicated this result with they are less likely to be trapped by sunk costs (Soman World Bank staff. In this experiment, respondents were and Cheema 2001). It can be easier to justify cutting randomly assigned to one of three different prompts. off a project when there are no untoward career con- In each prompt, the basic setup was identical: a $50 sequences for doing so and when criteria for ending a savings in exchange for a 45-minute drive. However, project are clear and public. For development organiza- the only piece of information that changed was the tions, there are important implications from recogniz- price of the object (in this case, a watch). As the price of ing that development is complex, that many projects the watch increased (that is, the discount rate dropped), will fail, and that learning is as important as investing. World Bank staff were significantly less likely to report traveling to the store. Staff valued time and money dif- The effects of context on ferently from the way the people whose lives they were judgment and decision making working to improve valued them. No income groups The biases policy makers themselves may hold about in Nairobi, Kenya, who were asked this question the population they are intending to serve are also changed their answers when the price of the object very important. When designing policies appropriate (in this case, a cell phone) increased (see spotlight 3). for a target group, policy makers must make some Second, even the most well-intentioned and assumptions about this group. At a basic level, know- empathic policy maker is a representative of an organi- ing whether the group’s literacy rate is low or high will zation and a professional community that deploy par- guide the design of policies (for example, road safety ticular language, assumptions, norms, and resources. signs may use numbers and pictures rather than These may be so familiar to policy makers that they are letters if some drivers in the group cannot read). Less unaware of how alien they may appear to outsiders intuitively, knowing how poor people’s labor supply and those they are ostensibly trying to serve. Develop- would change in response to a transfer is useful in ment initiatives and discourse are replete with phrases THE BIASES OF DEVELOPMENT PROFESSIONALS 187 espousing the virtues of “participation,” “empower- the coherence, effectiveness, and legitimacy of those ment,” and “accountability,” for example, but as articu- prevailing systems. lated by development practitioners, these concepts What can be done to close the gap between the largely reflect the sensibilities of donor agencies and mental models of development professionals and the urban elites (Gauri, Woolcock, and Desai 2013), who “beneficiaries” of their “interventions”? Lessons from tend to use them in confined ways. These may be dif- the private sector may be useful. Consider the high- ferent from how prevailing systems of order and technology sector, where experts attempt to create com- change are experienced in, say, a given village in rural plex products for “typical” consumers. Since designers Ghana or Indonesia (Barron, Diprose, and Woolcock in this industry have very specific training and are 2011). Even among professionals, academic researchers constantly immersed in the world of product design, take it as a given that development policy should be the lens through which they view the world is often “evidence-based,” and on this basis they proceed to quite different from that of a common user who lacks frame arguments around the importance of conduct- knowledge of the theoretical principles and necessary ing “rigorous evaluation” to assess the “effectiveness” trade-offs guiding design processes. Moreover, design- of particular interventions. In contrast, seasoned prac- ers spend countless hours with their products, while titioners tend to regard evidence as one factor among users encounter them only when they are trying to sat- many shaping what policies become politically sup- isfy some particular need. The result can be substantial portable and implementable and thus, on the basis of underutilization of otherwise highly capable products these latter criteria, are deemed “effective” (box 10.2). and programs (such as all the buttons on remote control Third, development policy makers and profession- devices to operate televisions) or, at worst, abandon- als usually are not familiar with the mental models and ment in the face of a futile, frustrating experience. mindsets that poor people use. Policy makers are likely One approach to meeting this challenge is known in to live in different places from the poor, to send their the software industry as dogfooding. This expression children to different schools, to receive medical treat- comes from the colloquialism, “Eat your own dog food”; ment in different hospitals, to travel on different modes it refers to the practice in which company employees of transport, and to have much stronger incentives to socialize with and listen to those who are more likely to be able to support their policy agenda and political Box 10.2 A clash of values between development professionals and the local populace: Agricultural career. One constructive response to this problem has reform in Lesotho been “village immersion” programs, in which senior officials commit to living the lives of their constituents An agricultural modernization program initiative in Lesotho provides an illustra- for a week, working alongside them and eating in their tion of widely divergent views of value between development professionals and houses, the better to experience firsthand what spe- the local populace. In this landlocked nation, development professionals saw the cific problems they encounter (Patel, Isa, and Vagneron country’s grasslands as one of the few potentially exploitable natural resources 2007). In a broader sense, the widening inequality in and its herds of grazing animals as a “traditional” practice ripe for transformation society makes it less likely that people from different by a “new” modern economy. Necessary changes, planners believed, included walks of life will encounter one another, even inhabit controlled grassland use, new marketing outlets for surplus animals, and more the same “moral universe” (Skocpol 1991; World Bank productive breeds. This seems straightforward enough from an economic point 2005), rendering the preferences and aspirations of of view. But within a year of the land being fenced off for the exclusive use of marginalized groups even more marginal. The result- more “commercially minded” farmers, the fence was cut, the gates had been ing difference in mindsets between rich and poor can stolen, and the land was being freely grazed by all. Moreover, the association manifest itself in very concrete ways (box 10.3). manager’s office had been burned down, and the program officer in charge was Development professionals usually interpret the said to be in fear for his life. What happened here? local context as something that resides “out there” in The mental models of the development professionals regarding the “value” developing countries—as something that policy mak- of various agricultural practices failed to take account of unique but critical ers and practitioners should “understand” if they are to features of the Lesotho economy. Planners viewed animals as simple commod- be effective. Taking local contexts seriously is crucial ities. But community members saw them very differently. Grazing animals were (Rao and Walton 2004). Development professionals excluded from the otherwise modern and highly monetized economy, carrying must be constantly aware that development program- an intrinsic value of their own that was embedded within a very different set of ing cannot begin from scratch. Every human group has rules—sometimes referred to as “the bovine mystique”—that prioritized owning a system of some kind already in place for addressing cattle over cash. its prevailing challenges and opportunities. The intro- duction of development projects can bolster or disrupt Source: Ferguson 1994. 188 WORLD DEVELOPMENT REPORT 2015 Box 10.3 It may be difficult for development professionals to accurately predict the views of poor people As part of the research for this Report, data were collected both from b. Helplessness in dealing with life’s problems development professionals within the World Bank and from individuals in Survey question: I feel helpless in dealing with the problems of life. the bottom, middle, and top thirds of the wealth distribution in the capital 100 Percentage reporting “agree”/“strongly agree” cities of selected developing countries (Jakarta, Indonesia; Nairobi, Kenya; and Lima, Peru). The data reveal a large gap between how development 90 professionals perceive the context of poverty and how the bottom third 80 views it. In the three figures that follow, this difference can be seen clearly 70 in three distinct areas crucial to development: whether the bottom third thinks of themselves as having control over their lives (figure B10.3.1, 60 panel a), how helpless they feel in dealing with the problems in their 50 life (panel b), and their knowledge about health services (their attitudes 40 toward vaccinations, for example) (panel c). 30 Figure B10.3.1 How World Bank staff predicted the views 20 of poor people 10 a. Control of the future Survey question: What happens to me in the future mostly depends 0 Self Predictions Actual Predictions Actual on me. for the responses for the responses poorest of poor middle and of the middle Percentage reporting “agree”/“strongly agree” 100 individuals individuals top thirds and top thirds (bottom third) 90 Jakarta Nairobi Lima 80 Source: WDR 2015 team survey data. 70 60 c. The dangers of vaccines Survey question: Vaccines are risky because they can cause sterilization. 50 40 100 Percentage reporting “most chose true” 30 90 20 80 70 10 60 0 Self Predictions Actual Predictions Actual 50 for the responses for the responses poorest of poor middle and of the middle 40 individuals individuals top thirds and top thirds (bottom third) 30 Jakarta Nairobi Lima 20 Source: WDR 2015 team survey data. 10 0 Predictions Actual Predictions Actual Panels a and b reveal a large disparity between how development for the responses for the responses professionals believe poor individuals (bottom third) will answer these poorest of poor middle and of the middle questions and how poor individuals in fact answered them. Development individuals individuals top thirds and top thirds (bottom third) professionals imagine that poor individuals are very different from themselves in their self-perceptions, but in fact they are not. In all cases, Jakarta Nairobi Lima responses by the bottom and by the middle and top thirds of the income Source: WDR 2015 team survey data. distribution are similar. However, development professionals believe there is a large disparity between the poor and the rest and see themselves as those of the rest of the population. This finding suggests that develop- closer to the upper-level groups than to poor individuals. ment professionals assume that poor individuals are less autonomous, In another area, development professionals imagine poor individuals less responsible, less hopeful, and less knowledgeable than they in fact to be much more suspicious of vaccines than they actually are (panel c). are. These beliefs about the context of poverty shape policy choices. It is In each instance, the responses of poor individuals are very close to important to check these beliefs against reality. THE BIASES OF DEVELOPMENT PROFESSIONALS 189 themselves use a product (“eat the dog food”) to work Conclusion out the kinks before releasing it to the marketplace. The This chapter has sought to explain why good people can approach turns on the belief that the product (whether make bad decisions. More specifically, it has sought to dog food, an iPad, or an electronic toothbrush) should document four different ways in which development be good enough and user-friendly enough for everyone professionals can make consequential mistakes even in the company to operate or consume before they when they are diligent, sincere, technically competent, expect customers to do so. The key idea driving the and experienced. Largely because of the organizational dogfooding process is that while a product’s designers imperatives within which they and their counterparts are often blind to how user-friendly the product is, operate and the primary reference groups with which other employees—although not exactly typical users— they associate most frequently—and thus whose at least bring fresh eyes to the product and are therefore approval they covet (or whose opprobrium they seek to more easily able to spot trouble points, nonintuitive avoid)—such professionals can consistently contribute steps, and other impediments. It is very easy to assume to outcomes biased against those on whose behalf they that your client is just like you; dogfooding helps bring are working. implicit (and often flawed) assumptions to the surface, In this sense, development professionals, like to check them against the facts, discover unexpected professional people everywhere, are likely to make uses, and identify opportunities for redesigns that bet- decisions that favor certain groups over others. In ter serve everyday customers’ needs. Dogfooding forces the development context—where knowledge, status, developers to reconcile their general, abstract knowl- and power differentials are rife—this often means edge with local or “situational” practical knowledge and that disadvantaged groups face additional hurdles thus raises the odds of generating a successful product. to getting their voices heard, their concerns heeded, This approach is related to the importance of piloting and their aspirations realized. Although these biases and testing before wide-scale implementation, ideally cannot be fully eliminated, being aware of their pres- with the designers themselves or with a subset of the ence, their consequences, and the mechanisms and users to ensure that the product (policy) has maximized incentives underpinning them is the first step toward effectiveness and efficiency. Chapter 11 discusses this addressing them. process of piloting and testing in more detail. Some activities analogous to dogfooding already exist in the public sector.3 Governments can use “green While the goal of development is to end public purchasing” as a means of testing environmen- tal policies and greener products. By experimenting poverty, development professionals with policies and products within their own organi- zations, governments might make more informed are not always good at predicting how decisions about how regulations might affect broader public and private markets (see OECD 2003). When poverty shapes mindsets. the Behavioural Insights Team in the United Kingdom initiated an effort to improve services at employment centers, team members went through the centers The second step is to put in place measures that themselves to get a better sense of the user expe- might plausibly help counteract them. This chapter rience. Sometimes, however, it is necessary for an has identified four sources of bad decision making organization to go directly to its customers or users to on the part of development professionals: complexity, understand how they will behave in certain situations. confirmation bias, sunk cost bias, and the influence In such cases, well-structured focus groups can be a of context on judgment and decision making. Each of useful research tool for giving planners access to the these can be addressed, at least in part, through organi- experiential worlds they do not otherwise encounter. zational measures. They allow policy makers and designers to see the As this Report has shown, because the determinants mental models of other people in action so that they of behavior are often subtle and hard to detect, better can better understand their preferences, attitudes, means of detection, starting with asking the right ques- expectations, and abilities—in the process generating tions, are needed (see chapter 11). This would suggest a useful insights at reasonable cost. Bringing “experts” more prominent place for investing more extensively into direct contact with “users” enables both parties in analyses of local social and political economies (to to gain practical knowledge of how and why the other better understand the nature of changing contextual group behaves in the way it does. idiosyncrasies). 190 WORLD DEVELOPMENT REPORT 2015 In the case of confirmation bias, it is crucial to Notes expose individuals to social contexts in which indi- 1. The WDR team invited 4,797 World Bank staff (exclud- viduals disagree with each other’s views but share a ing consultants) from all sectors of the World Bank to common interest in identifying the best policies and participate in a survey designed to measure percep- programs. This can be done through red teaming major tions. The sample was representative of staff working decisions: that is, subjecting the key assumptions and in World Bank headquarters in Washington, D.C., and arguments underlying policies to a critical and adver- of country offices across the world. The final number sarial process. Other approaches take the form of of respondents was 1,850 staff (900 from headquarters double-blind peer review and more intense engage- and 950 from country offices, yielding a response rate ment with the scholarly community. of 38.6 percent), which is well above the 1,079 needed for For sunk cost bias, the key is to change the interpre- representativeness. tation of a canceled program or project. This involves 2. The U.S. military (University of Foreign Military and recognizing that “failure” is sometimes unavoidable Cultural Studies 2012) and the U.K. government (United in development and encouraging individuals to learn, Kingdom, Ministry of Defense 2013) both have guides to rather than hide, from it. Indeed, it is often unclear red teaming. IBM contracts out red teams as part of its whether apparent futility is really a product of a funda- consulting services—essentially to break into people’s mentally flawed strategy that no manner of persistence information technology infrastructure. They brand or tinkering can fix (and thus should be abandoned) or them “tiger teams,” and the teams are seen as a model a product of a strategy that is otherwise fundamentally to be emulated. Grayman, Ostfeld, and Salomons (2006) sound confronting a deeply ingrained problem—like describe using red teams to determine locations for dowry systems or child marriage—that requires cour- water quality monitors in a water distribution system. age and commitment for success even to be possible. 3. Roman emperors allegedly used similar techniques to Crucially, development professionals need to recognize ensure the reliability of bridges: after a given bridge that even failures are opportunities to learn and adapt. was completed, those involved in its construction were The more failures are treated as somewhat expected required to sleep under it for several days. This practice and as opportunities to learn, the easier it can be to let ensured that all involved had the strongest incentive to go of a failing project. build infrastructure that actually functioned reliably, Finally, this chapter has also shown how giving rather than merely looking impressive or being com- inadequate attention to context can bias key decisions. pleted on time. The decision-making processes, languages, norms, and mental models of development professionals, whether foreign or domestic, differ from those of their clients References Andrews, Matt, Lant Pritchett, and Michael Woolcock. and counterparts. To address these differences, devel- 2013. “Escaping Capability Traps through Problem- opment professionals can engage in more systematic Driven Iterative Adaptation (PDIA).” World Development efforts to understand the mindsets of those they are 51 (Nov.): 234–44. trying to help. For project and program design, devel- Argyris, Chris. 1991. “Teaching Smart People How to Learn.” opment professionals should “eat their own dog food”: Harvard Business Review 69 (3). http://hbr.org/1991/05 that is, they should try to experience firsthand the pro- /teaching-smart-people-how-to-learn/ar/1. grams and projects they design. Arkes, Hal R., and Peter Ayton. 1999. “The Sunk Cost and If the prevalence and effects of these four errors— Concorde Effects: Are Humans Less Rational than and the many others discussed in preceding chapters— Lower Animals?” Psychological Bulletin 125 (5): 591. are as important as this Report suggests, development Arkes, Hal R., and Catherine Blumer. 1985. “The Psychol- organizations face the stark choice of “paying now or ogy of Sunk Cost.” Organizational Behavior and Human paying later”: they can choose to make considered, Decision Processes 35 (1): 124–40. Barron, Patrick, Rachael Diprose, and Michael J. V. Wool- strategic investments in minimizing these errors up cock. 2011. Contesting Development: Participatory Projects front, or they can choose to deal with all manner of and Local Conflict Dynamics in Indonesia. New Haven: legal, ethical, political, financial, and public relations Yale University Press. disasters that may emerge after the fact. (Neglecting to Barth, James, Gerard Caprio, and Ross Levine. 2013. “Bank choose is its own form of choice.) Good social science, Regulation and Supervision in 180 Countries from hard-won experience, basic professional ethics, and 1999 to 2011.” Journal of Financial Economic Policy 5 (2): 2. everyday common sense suggest that “an ounce of pre- Charness, Gary, and Matthias Sutter. 2012. “Groups Make vention” is a far preferable course of action for deliv- Better Self-Interested Decisions.” Journal of Economic ering on the World Bank’s core agenda and mandate. Perspectives 26 (3): 157–76. THE BIASES OF DEVELOPMENT PROFESSIONALS 191 Evans, Jonathan St. B. T. 1989. “Problem Solving, Rea- Ramalingam, Ben. 2013. Aid on the Edge of Chaos: Rethinking soning and Decision Making.” Cognitive Psychology: International Cooperation in a Complex World. Oxford, Research Directions in Cognitive Science. European Perspec- U.K.: Oxford University Press. tives 1: 85–102. Rao, Vijayendra, and Michael Walton. 2004. Culture and Ferguson, James. 1994. The Anti-Politics Machine. Minneapo- Public Action: A Cross-Disciplinary Dialogue on Develop- lis: University of Minnesota Press. ment Policy. Palo Alto, CA: Stanford University Press. Fischhoff, Baruch, and Ruth Beyth-Marom. 1983. “Hypo- Ravallion, Martin. 2011. “Development Impact Calls for thesis Evaluation from a Bayesian Perspective.” Psycho- Knowledgeable Development Practitioners.” Develop- logical Review 90 (3): 239. ment Impact (blog), World Bank. http://blogs.world Fishkin, James. 2009. When the People Speak: Deliberative bank.org/impactevaluations/development-impact Democracy and Public Consultation. Oxford, U.K.: Oxford -calls-for-knowledgeable-development-practitioners. University Press. Redelmeier, Donald A., and Eldar Shafir. 1995. “Medical Gauri, Varun, Michael Woolcock, and Deval Desai. 2013. Decision Making in Situations That Offer Multiple “Intersubjective Meaning and Collective Action in Alternatives.” JAMA-Journal of the American Medical Developing Societies: Theory, Evidence and Policy Impli- Association-U.S. Edition 273 (4): 302–05. cations.” Journal of Development Studies 49 (1): 160–72. Schkade, David, Cass R. Sunstein, and Reid Hastie. 2010. Grayman, Walter M., Avi Ostfeld, and Elad Salomons. “When Deliberation Produces Extremism.” Critical 2006. “Locating Monitors in Water Distribution Sys- Review 22 (2/3): 227–52. tems: Red Team-Blue Team Exercise.” Journal of Water Scott, James C. 1998. Seeing Like a State: How Certain Schemes Resources Planning and Management 132 (4): 300–04. to Improve the Human Condition Have Failed. New Haven, Kahan, Dan M., Ellen Peters, Erica Cantrell Dawson, and CT: Yale University Press. Paul Slovic. 2013. “Motivated Numeracy and Enlight- Skocpol, Theda. 1991. “Targeting within Universalism: ened Self-Government.” Working Paper, Cultural Cog- Politically Viable Policies to Combat Poverty in the nition Project, Yale Law School, New Haven, CT. United States.” In The Urban Underclass, edited by Keil, Mark, Duane P. Truex III, and Richard Mixon. 1995. Jencks C. Petersen, 411–36. Washington, DC: The “The Effects of Sunk Cost and Project Completion on Brookings Institution. Information Technology Project Escalation.” Engineer- Soman, Dilip, and Amar Cheema. 2001. “The Effect of ing Management 42 (4): 372–81. Windfall Gains on the Sunk-Cost Effect.” Marketing Lilienfeld, Scott O., Rachael Ammirati, and Kristin Land- Letters 12 (1): 51–62. feld. 2009. “Giving Debiasing Away: Can Psychological Tversky, Amos, and Daniel Kahneman. 1981. “The Framing Research on Correcting Cognitive Errors Promote of Decisions and the Psychology of Choice.” Science 211 Human Welfare?” Perspectives on Psychological Science 4 (4481): 453–58. (4): 390–98. United Kingdom, Ministry of Defense. 2013. Red Team- Lord, Charles G., Lee Ross, and Mark R. Lepper. 1979. ing Guide. 2nd ed. https://www.gov.uk/Government “Biased Assimilation and Attitude Polarization: The /Uploads/System/Uploads/Attachment_Data/File Effects of Prior Theories on Subsequently Considered /142533/20130301_Red_Teaming_Ed2.Pdf. Evidence.” Journal of Personality and Social Psychology 37 University of Foreign Military and Cultural Studies. 2012. (11): 2098. Red Team Handbook. http://Usacac.Army.Mil/Cac2 Mercier, Hugo, and Dan Sperber. 2011. “Why Do Humans /Ufmcs/Repository/Rt_Handbook_V6.Pdf. Reason? Arguments for an Argumentative Theory.” Wason, Peter C. 1960. “On the Failure to Eliminate Behavioral and Brain Sciences 34 (2): 57–74. Hypotheses in a Conceptual Task.” Quarterly Journal of Moskowitz, Tobias, and L. Jon Wertheim. 2011. Scorecast- Experimental Psychology 12 (3): 129–40. ing: The Hidden Influences behind How Sports Are Played ————. 1977. “Self-Contradictions.” In Thinking: Readings and Games Are Won. New York: Crown Publishing. in Cognitive Science, edited by P. N. Johnson-Laird and Mullainathan, Sendhil, and Eldar Shafir. 2013. Scarcity: Why P. C. Wason, 114–28. Cambridge, U.K.: Cambridge Uni- Having Too Little Means So Much. New York: Times Books. versity Press. Nickerson, Raymond S. 1998. “Confirmation Bias: A Ubiq- Weick, Karl E. 1984. “Small Wins: Redefining the Scale of uitous Phenomenon in Many Guises.” Review of General Social Problems.” American Psychologist 39 (1): 40. Psychology 2 (2): 175. Wetherick, N. E. 1962. “Eliminative and Enumerative OECD (Organisation for Economic Co-operation and Behaviour in a Conceptual Task.” Quarterly Journal of Development). 2003. The Environmental Performance of Experimental Psychology 14 (4): 246–49. Public Procurement. London: OECD. Willingham, D. T. 2007. “Critical Thinking: Why Is It So Patel, Praful, Qazi Azmat Isa, and Caroline Vagneron. 2007. Hard to Teach?” American Educator (Summer): 8–19. “The World Bank’s Village Immersion Programme in World Bank. 2005. World Development Report 2006: Equity South Asia.” Participatory Learning and Action 57 (1): and Development. Washington, DC: World Bank. 34–35. 11 CHAPTER Adaptive design, adaptive interventions Behind every policy lie assumptions about why peo- subsidy at the time fertilizer was applied (Duflo, Kre- ple behave the way they do. A policy that subsidizes mer, and Robinson 2011). In Malawi, allowing farmers fertilizer, for example, assumes that farmers find the to direct some of their harvest profits into commit- price too high; that they can easily learn about price ment savings accounts, which held the money until reductions once a subsidy is enacted; that they would the following planting season, increased investment benefit from using fertilizer and are aware of those back into crops and significantly increased the value benefits; that they are willing to invest some of their of the subsequent harvest (Brune and others 2013). own money today and accept the associated risk to get Recognizing that individuals think automatically, payoffs at the end of the farming cycle; and that they think socially, and think with mental models expands have time to go purchase the product. But assumptions the set of assumptions policy makers can use to analyze may often be incorrect, and solutions based on the a given policy problem and suggests three main ways wrong assumptions can lead to ineffective policies. for improving the intervention cycle and development effectiveness. First, concentrating more on the defini- tion and diagnosis of problems, and expending more Concentrating more on the definition and cognitive and financial investments at that stage, can lead to better-designed interventions. For example, diagnosis of problems, and expending taking the time to figure out that application forms for financial aid for college might be the obstacle that more cognitive and financial resources at depresses college attendance rates for low-income populations could lead to strategies that help students that stage, can lead to better-designed and their families fill out those applications—and could spare investments in an expensive and possibly ineffec- interventions. tive information campaign (Bettinger and others 2012). Second, an experimental approach that incorporates testing during the implementation phase and tolerates For instance, as chapter 7 showed, farmers might failure can help identify cost-effective interventions find it difficult to translate their intentions to invest in (Glennerster and Takavarasha 2013; Duflo and Kremer fertilizer into concrete action at the time they need to 2005). As many of the studies cited throughout this purchase the fertilizer. The divide between intentions Report indicate, the process of delivering products and actions may arise from the fact that farmers have matters as much as the product that is being delivered, cash in hand after harvest but do not need fertilizer and it can be difficult to predict what will matter in until a few months later during the planting season. In which context and for which population. For example, Kenya, allowing farmers to prepay for fertilizer during who could have predicted that weekly text-message the harvest and get it delivered during the next plant- reminders would improve adherence to a critical drug ing season proved as effective as offering a 50 percent regimen for treating HIV/AIDS in Kenya better than ADAPTIVE DESIGN, ADAPTIVE INTERVENTIONS 193 daily reminders (Pop-Eleches and others 2011) (see know what should be done without having made their chapter 8). An experiment was required to learn that assumptions explicit and without having diagnosed financial incentives were not effective in motivating the actual problem and its causes. While many devel- the distributors of female condoms in Zambia (Ashraf, opment practitioners would agree that they often do Bandiera, and Jack, forthcoming) (see chapter 7). not know what will work in a given context, their orga- Third, since development practitioners themselves nizational environment may not allow them to admit face cognitive constraints, abide by social norms, and as much (Pritchett, Samji, and Hammer 2013). use mental models in their work, development organi- Delving deeper into the subject may lead to a better zations may need to change their incentive structures, understanding of the underlying causes of an observed budget processes, and institutional culture to promote behavior and to identifying ways to intervene effec- better diagnosis and experimentation so that evidence tively. In a complex and iterative process (figure 11.1), can feed back into midcourse adaptations and future problems may need to be redefined and rediagnosed, intervention designs. Development practitioners must and multiple interventions may need to be piloted often act quickly and may thus feel compelled to skip simultaneously—some of which will fail—before an a careful diagnosis and immediately apply “best prac- effective intervention can be designed. tice.” Indeed, the intervention cycle typically allows This chapter builds on the work by Datta and Mul- neither the time nor the space to collect the data and lainathan (2014), who discuss how to design develop- perform the analysis needed to identify the problem ment programs and policies in ways that are cognizant properly, diagnose its determinants, and assess the of and informed by the insights from the behavioral fit between program and context or to make needed sciences, an approach that has been applied to design midcourse changes. As spotlight 3 and chapter 10 interventions for low-income populations across the demonstrate, the mindsets of development practition- United States (CFED and ideas42 2013). ers can also differ substantially from those that prevail To see how diagnoses and program design can among low-income populations for whom they may be evolve in the process of finding a solution to a chal- designing programs. Because development practition- lenge, consider the problem of ensuring access to clean ers often have preconceived notions about a problem water in rural Kenya and a series of field experiments and its potential solutions, they may believe that they that tested the effectiveness of different methods of Figure 11.1 Understanding behavior and identifying effective interventions are complex and iterative processes In an approach that incorporates the psychological and social aspects of decision making, the intervention cycle looks different. The resources devoted to definition and diagnosis, as well as to design, are greater. The implementation period tests several interventions, each based on different assumptions about choice and behavior. One of the interventions is adapted and fed into a new round of definition, diagnosis, design, implementation, and testing. The process of refinement continues after the intervention is scaled up. & diagnose fine & rediagnose Define Rede t ap Ad Design nt & evaluate leme Imp Source: WDR 2015 team. 194 WORLD DEVELOPMENT REPORT 2015 averting the incidence of diarrhea among children the type of environmental or institutional setting (low- (Ahuja, Kremer, and Zwane 2010). Lack of access to or high-income, low- or high-capacity). clean water was diagnosed as a problem, and thus an This chapter discusses the components of the more early intervention aimed to improve infrastructure at complex and more iterative intervention cycle pro- households’ water sources, naturally occurring springs, posed in figure 11.1: (1) diagnosing and rediagnosing which were susceptible to contamination from the sur- psychological and social obstacles; (2) designing an rounding environment. In particular, the springs were intervention; (3) experimenting during implementa- covered with concrete so that water flowed from a pipe tion; and (4) learning from these previous steps and rather than seeping from the ground. While this con- adapting future interventions accordingly. siderably improved water quality at the source, it had only moderate effects on water quality in households Diagnosing psychological and because the water could easily be recontaminated dur- social obstacles ing transport or storage (Kremer and others 2011). While it goes without saying that identifying prob- Thus the problem was not simply access to clean lems or obstacles must precede the design of solutions, water; instead, it could be redefined as a problem of there is less clarity on just how one should go about inadequate water treatment within the home. Another this process of diagnosis. Measuring an individual’s iteration of experiments demonstrated that providing lack of material resources or information, for example, free home delivery of chlorine or discount coupons is relatively straightforward, and countless household that could be redeemed in local shops elicited very surveys provide data on these sorts of obstacles. In high take-up of the water treatment product at first but contrast, identifying the presence of psychological ultimately failed to generate sustained results. People biases, cognitive burdens, social norms, and mental needed to remember to chlorinate their water when models may require more in-depth investigations. they returned home from the water source, and they Thick description, for example, and other forms of needed to continue to go to the store to purchase the ethnography (spotlight 4) can be used to understand product. decision-making contexts. In traditional anthropology, These results in turn suggested yet another diag- ethnographic fieldwork consists of extensive partic- nosis of the problem: households found it difficult ipant observations, interviews, and surveys. More to sustain the use of water treatment over time. This problem-driven forms of the ethnographic approach insight led to the design of free chlorine dispensers can be used to help development practitioners refine next to the water source, which made water treatment their hypotheses about what drives specific behaviors, salient (the dispenser served as a reminder just when as well as to monitor newly emerging behaviors. In people were thinking about water) and convenient Denmark, for example, a ban on indoor smoking shifted (there was no need to make a trip to the store, and the smokers to the areas just outside the doors of buildings. necessary agitation and wait time for the chlorine to This posed a problem for Copenhagen Airport, since the work automatically occurred during the walk home). secondhand smoke could easily find its way back into It also made water treatment a public act. This proved the building through doors and air vents. Simply creat- to be the most cost-effective method for increasing ing a no-smoking zone around entrances did not help. water treatment and averting the incidence of diarrhea Careful “fieldwork,” however, in which the habits of (Abdul Latif Jameel Poverty Action Lab 2012). those smoking at the airport were closely observed and As this example and other chapters demonstrate, mapped, was instrumental in finding solutions that context matters in particular ways. Seemingly small cut smoking near entrances by more than 50 percent. details of design and implementation of policies and Since smokers tended to come from inside the building programs can have disproportionate effects on indi- and reach for their cigarettes as they were exiting the vidual choices and actions. Similar challenges can have building, stickers with an icon of a lit cigarette and the different underlying causes. An approach that works distance to the smoking zone were placed on the floors in one country may not necessarily work in another. right before the doors. Benches and trash cans, which Indeed, evidence on the policy implications of a psy- tended to attract smokers, were placed farther from chological and social perspective on development chal- airport entrances in zones especially designated for lenges is just now coming into view. smoking (iNudgeyou 2014). Along with “thick descrip- This Report does not advocate specific interven- tion” like this, another useful way to characterize tions. Instead, it argues for the need to change the pro- decision-making contexts is the “Reality Check” (box 11.1). cess of arriving at solutions, regardless of the nature of More quantitative methods, such as surveys, can the problem (acute, chronic, last mile, and so forth) or also be informative at this stage of the intervention ADAPTIVE DESIGN, ADAPTIVE INTERVENTIONS 195 cycle. A number of measurement techniques can help reduce courtesy bias (where respondents provide Box 11.1 Taking the perspective of program beneficiaries through the Reality Check approach answers they think the questioner wants to hear) and measure psychological patterns that respondents The everyday experiences, awareness, and aspirations of people living in poverty themselves may not be aware of (box 11.2). are often unmeasured and may in fact be dynamic. This challenges development Finally, there may be nothing as illuminating as the professionals to keep in touch and up to date. An immersion program called the technique of dogfooding, discussed in chapter 10. In this Reality Check approach has been used by donors, governments, and nongovern- practice, company employees themselves use a prod- mental organizations (NGOs) to understand how poor people make decisions. uct they have designed to work out its kinks before Social science researchers live for several days and nights with a poor family, not releasing it to the marketplace. Policy designers could as an important visitor but as an ordinary person, aiming to observe and build try to sign up for their own programs or access exist- relationships, trust, and respect. This qualitative approach has uncovered impor- ing services to diagnose problems firsthand. tant findings that might have been missed with more quantitative surveys. For example, in Bangladesh and Nepal, government health providers felt pressured Designing an intervention every day to provide free medicine to people who they knew were not ill but who Once key obstacles have been identified, the task were selling it to others or who wanted it for their livestock. In northern Ghana, becomes designing an intervention that incorporates researchers learned that at certain times of year, the heat made it unreasonable these insights. Sometimes the diagnosis phase of an to expect people to get inside a mosquito net. intervention may reveal multiple obstacles but not their relative importance, and each of these would Source: www.reality-check-approach.com. imply different designs for tackling the larger prob- lem at hand. Consider again the example of home water chlori- who would use the product only if she saw someone in nation. Table 11.1 lists a number of different obstacles her peer group using it. that could interfere with home water treatment and Table 11.2 presents a list of designs and related inter- the corresponding interventions that could overcome ventions that have been experimentally evaluated to them. An intervention designed for someone who identify effective interventions across a wide class of knows the benefits of chlorine and can afford to pur- problems (Richburg-Hayes and others 2014). In light chase it but simply forgets to would look somewhat of this growing body of work, it has been argued different from an intervention designed for someone that a science of design is emerging in which the Box 11.2 Measurement techniques that can help uncover psychological and social obstacles Techniques for eliciting sensitive information assigned one of two questions and asked to report the number of • Introduce personal distance. Sometimes, answers are best elicited items that they agree with or that apply to them. The lists differ solely through questions that are asked indirectly. For instance, rather than in the presence of the sensitive item or topic. asking an official whether he has ever accepted a bribe, the researcher can ask whether a person in his position typically accepts bribes. Measuring attitudes and social norms Eliciting information through vignettes or hypothetical situations • Implicit association tests. These tests measure automatic associa- about fictional people allows respondents to think about a situation in tions between concepts (such as the home or a career) and attrib- a way that is more emotionally removed from their personal concerns utes (male and female) (Greenwald, McGhee, and Schwartz 1998; but that tends to reveal social expectations. Banaji 2001; Beaman and others 2009; Banaji and Greenwald 2013). • Allow a cover of randomness. For instance, when asked a sensitive They are easy to administer and can be adapted for nonliterate question that should have a yes/no answer, a respondent can be asked populations. Demonstration tests can be found at www.implicit to privately flip a coin and say yes if it comes up heads or answer truth- .harvard.edu. fully if it comes up tails. This can allow the person to answer truthfully • Identifying social norms. Survey questions in household surveys or and still allow the researcher to learn about the share of the population ethnographic work can uncover perceptions about expected and that engages in a potentially shameful behavior, even if she would not prescribed behaviors. For example, questions like “Out of 10 of your know about the behavior of any given individual. List experiments neighbors, how many exclusively breastfeed their children?” can help (Blair and Imai 2012; Droitcour and others 2011; Holbrook and Krosnick reveal what people expect others to be doing. Questions like “If you 2009; Karlan and Zinman 2012) are another method for measuring the decided to exclusively breastfeed your child, would you worry about share of a population that engages in a taboo behavior or holds an anyone disapproving?” can help reveal the relevant network to which opinion that may not be freely admitted. Respondents are randomly the social norm applies. 196 WORLD DEVELOPMENT REPORT 2015 psychological and social sciences can play a key role domly divided into five groups that received a visit (Datta and Mullainathan 2014). from a community health worker at different inter- Many of the quantitative and qualitative methods vals after the distribution of the nets (1–3 days, 5–7 useful for diagnosing obstacles can also assist in the days, 10–12 days, 15–17 days, and six weeks later). Self- design phase—particularly in narrowing down options installation and retention rates were then compared that could be tested at a larger scale. Two experiences across the five groups. These household visits revealed from Zambia demonstrate this approach. A “mama that bed nets were hung by recipients within the first kit” is a package provided to an expectant mother that 10 days; that nets that were not hung after 10 days contains all the materials she would need to ensure were unlikely to be hung at all; and that retention was the clean and safe delivery of her child. The kits are stable for the two months or so following distribution. typically used to encourage delivery in a health facility. These results provided the government a clear path to Semi-structured interviews with women and a survey designing an optimal visit frequency, and it crafted of local wholesale prices helped determine the kit guidelines specifying the optimal time to visit house- contents that mothers would find desirable and that holds and hang up the remaining bed nets as 10 days would be feasible to provide. This up-front qualitative after distribution. work to optimize the content of the mama kit paid off. Mechanism experiments are another useful technique Ultimately, a randomized controlled trial found that for narrowing down candidate policies for experi- the mama kits increased facility delivery rates by 44 mentation. Consider how such an experiment could percent (IDinsight 2014a). be used to design a strategy for tackling the problem Similarly, the Zambian government quickly exper- of obesity in low-income neighborhoods (Ludwig, imented with different frequencies for household Kling, and Mullainathan 2011). Suppose that policy visits by community health workers to ensure that makers were concerned about “food deserts”: that subsidized antimalarial bed nets were actually being is, neighborhoods where there is plenty of food but used (IDinsight 2014b). Households that received bed none of it is healthy. One possible policy option would nets through fixed-point distribution sites were ran- be to experiment with offering incentives for green Table 11.1. Different obstacles may require different intervention design (Case study: increasing home water chlorination) Free bottles Discount coupon Using promoters Chlorine dispenser Design of intervention delivered at redeemable at Detailed Improved Persuasion from social at point of home local shop instructions storage messages network collection Potential obstacles People do not understand how to use chlorine. Procrastination may cause individuals to postpone visits to the store where the chlorine is sold. People are not motivated to use chlorine because the effect on health is delayed. People forget to chlorinate the water. People are affected by what others in the community do. Product may be too expensive. Some people are not convinced about the importance of clean water. Source: WDR 2015 team. ADAPTIVE DESIGN, ADAPTIVE INTERVENTIONS 197 grocers to locate in these areas. However, this would consume any of the free produce, it is worth asking if be an extremely costly intervention. An alternative they would take a more active step and purchase fruits experimental design would involve sampling low- and vegetables at a local store, even if the prices were income families and randomly selecting some of them heavily subsidized. If people did consume the produce to receive free weekly home delivery of fresh fruits and but did not lose weight, then policy makers should vegetables. This would be a much cheaper experiment perhaps not focus on the problem of food deserts as and would provide valuable information before trying a priority in solving the problem of obesity in low- a more expensive variant. If, for example, people do not income neighborhoods. Table 11.2 Experimental evidence is accumulating on the effectiveness of many psychologically and socially informed designs Type Strength of the evidence Examples Reminders 73 papers, appearing in 6 domains A regular text-message reminder to save money increased savings balances by 6 percent (Karlan and others 2010). Social influence 69 papers, appearing in all 8 domains Homeowners received mailers that compared their electricity consumption with that of neighbors and rated their household as great, good, or below average. This led to a reduction in power consumption equivalent to what would have happened if energy prices had been raised 11–20 percent (Allcott 2011). Feedback 60 papers, appearing in 5 domains A field experiment provided individualized feedback about participation in a curbside recycling program. Households that were receiving feedback increased their participation by 7 percentage points, while participation among the control group members did not increase at all (Schultz 1999). Channel and hassle 43 papers, appearing in 8 domains Providing personalized assistance in completing the Free Application for Federal Student Aid factors (FAFSA) led to a 29 percent increase in two consecutive years of college enrollment among high school seniors in the program group of a randomized controlled trial, relative to the control group (Bettinger and others 2012). Micro-incentives 41 papers, appearing in 5 domains Small incentives to read books can have a stronger effect on grades than incentives to get high grades (Fryer Jr. 2010). Identity cues and 31 papers, appearing in 5 domains When a picture of a woman appeared on a math test, female students were reminded to recall identity priming their gender and performed worse on the test (Shih, Pittinsky, and Ambady 1999). Social proof 26 papers, appearing in 5 domains Phone calls to voters with a “high turnout” message—emphasizing how many people were voting and that the number was likely to increase—were more effective at increasing voter turnout than a “low turnout” message, which emphasized that election turnout was low last time and likely to be lower this time (Gerber and Rogers 2009). Physical environment 25 papers, appearing in 5 domains Individuals poured and consumed more juice when using short, wide glasses than when using cues tall, slender glasses. Cafeterias can increase fruit consumption by increasing the visibility of the fruit with more prominent displays or by making fruit easier to reach than unhealthful alternatives (Wansink and van Ittersum 2003). Anchoring 24 papers, appearing in 3 domains In New York City, credit card systems in taxis automatically suggested a 30, 25, or 20 percent tip. This caused passengers to think of 20 percent as the low tip—even though it was double the previous average. Since the installation of the credit card systems, average tips have risen to 22 percent (Grynbaum 2009). Default rules and 18 papers, appearing in 7 domains Automatically enrolling people in savings plans dramatically increased participation and retention automation (Thaler and Benartzi 2004). Loss aversion 12 papers, appearing in 7 domains In a randomized controlled experiment, half the sample received a free mug and half did not. The groups were then given the option of selling the mug or buying a mug, respectively, if a determined price was acceptable to them. Those who had received a free mug were willing to sell only at a price that was twice the amount the potential buyers were willing to pay (Kahneman, Knetsch, and Thaler 1990). Public/private 11 papers, appearing in 4 domains When people promised to perform a task, they often completed it. People imagine themselves commitments to be consistent and will go to lengths to keep up this appearance in public and private (Bryan, Karlan, and Nelson 2010). Source: Richburg-Hayes and others 2014. Note: The eight domains covered were charitable giving, consumer finance, energy and the environment, health, marketing, nutrition, voting, and workplace productivity. 198 WORLD DEVELOPMENT REPORT 2015 Experimenting during A first set of field experiments in Kenya showed that implementation investment in fertilizer is surprisingly low despite high Sometimes practitioners might not have the luxury of returns (Duflo, Kremer, and Robinson 2007, 2008, 2011). time for all the possible qualitative and quantitative Diagnosis suggested that several factors, some psycho- work before implementation. Immediate action may be logical and social and some market related, could help required. In such cases, it will still be important to embed explain this puzzle: credit constraints, information experimentation during the implementation phase. constraints, absent-mindedness, and intention-action Experimentation during the implementation process divides. A second set of experiments tested these can still test psychological and social predictions and proposed theories by implementing several different optimize impact within a particular intervention cycle. interventions simultaneously and found that inter- Moreover, while using evidence from elsewhere may be ventions that provided a way for farmers to commit to very useful at the preparation stage, it will not replace fertilizer purchases (by paying for them when they had generating and using evidence from within the very cash on hand) were the most successful. Similar com- policy intervention as it is being carried out. mitment products were tested in Malawi with tobacco One way to test the importance of implementation producers with large positive effects (Brune and others details, for example, would be to experiment with dif- 2013). The findings were then taken to scale and evalu- ferent modes of implementation. In 2009, the Kenyan ated by the World Bank in Rwanda in the context of a government announced a nationwide contract teacher government intervention with a typical population of program that would eventually employ 18,000 teach- subsistence farmers (Kondylis, Jones, and Stein 2013). ers. In the pilot area, some schools were randomly As these examples show, a more adaptive, empir- chosen to receive contract teachers as part of the ically agile approach to the intervention cycle can government program, while others received a contract help identify effective ways to improve development teacher under the coordination of a local NGO. The outcomes. Has anyone succeeded in systematically evaluation showed how the implementation by the implementing this more psychological and socially NGO improved students’ test scores across diverse informed and experimental approach at a scale beyond contexts, while government implementation had no field experiments with NGOs? The Behavioural effect at all (Bold and others 2013). Insights Team in the United Kingdom has dedicated The series of experiments on commitment devices itself to bringing psychological and social insights for farmers discussed earlier also illustrates how into government policy and service delivery and tests experimental implementation can be used iteratively policy alternatives through experimentation (Haynes, to learn how to adapt policies before scaling them up. Goldacre, and Torgerson 2012 ) (box 11.3). Box 11.3 Using psychological and social insights and active experimentation in the United Kingdom The Behavioural Insights Team (BIT, also known as the “Nudge Unit”) tax on time.” Citing social norms that referred to others in the tax- was created in 2010 with the objective of applying insights from aca- payer’s own town led to a 15 percentage point increase in the fraction of demic research in behavioral economics and psychology to public policy taxpayers responding with payment in the following three months (BIT and services. It was created at a time of economic and financial crisis 2012). and resource scarcity, when psychological and socially informed inter- In another experiment, team members from BIT embedded them- ventions seemed a viable alternative to legislation. selves in an unemployment center to see what obstacles the unemployed BIT uses a four-part methodology to identify what works and faced in moving off unemployment benefits and into a job. They identi- can be scaled up and what does not: (1) define the desired outcome; fied a cumbersome process that involved considerable paperwork and (2) use ethnography to understand better how individuals experience that failed to motivate job seekers. They then designed a pilot program the service or situation in question; (3) build new interventions to that asked job seekers to make commitments for future job search activi- improve outcomes; and (4) test and try out the interventions, often using ties (as opposed to reporting on past activities) and to identify their per- randomized controlled trials. sonal strengths. These changes increased transitions away from benefits The unit tried to harness the power of social norms to encourage by nearly 20 percent (Bennhold 2013). timely tax payments. They tested various messages in letters sent to Until January 2014, BIT was funded by the public. It is now a company taxpayers, which either invoked no social norm or contained messages owned by its employees, the U.K. government, and Nesta (the leading like “9 out of 10 people in [Britain/your postcode/your town] pay their innovation charity in the United Kindgom). ADAPTIVE DESIGN, ADAPTIVE INTERVENTIONS 199 Conclusion: Learning and The findings in this Report bring another very large adapting and complex source of uncertainty to development As these and countless other examples throughout projects: the role of psychological and social factors in the Report have demonstrated, finding effective solu- the decision making and behavior of end users, imple- tions requires continual research and development menters, and development practitioners themselves. (R&D). Although time and resource constraints might interfere with efforts to adopt more systematic diag- noses and experimental implementation, the biggest To account for the fact that development challenge may be overcoming the psychological and social obstacles within development organizations practitioners themselves face cognitive themselves. Measures are needed to ensure that devel- opment practitioners account for their own automatic constraints, abide by social norms, thinking, mental models, and the social influences on their own choices. To do so, they may need to rethink and use mental models in their work, the process of research and development. First, R&D is not meant to yield immediate profits development organizations may need to or immediate improvements. It delivers uncertain ben- efits in the future. Time-pressed and risk-averse orga- change their incentive structure, budget nizations in search of immediate but certain results might thus underinvest in R&D. They may require processes, and institutional culture. commitment devices or risk-mitigating measures that can help them set aside the time and resources required for adequate diagnosis and experimental implementa- This uncertainty is not insurmountable for develop- tion. Over the years, development practitioners have ment practice. Indeed, one purpose of this Report has become familiar with the risks related to financing been to synthesize some of the most compelling scien- and political economy and to technical uncertainty. But tific research on the topic. It is hoped that this Report they need to pay more attention to another set of risks: can inspire development practitioners who are ready those associated with the development and implemen- to take up the challenge. tation of new products, services, and modes of delivery. What matters are not just the high-level policies that References governments and development actors adopt but also Abdul Latif Jameel Poverty Action Lab. 2012. “Cleaner how those policies are implemented and delivered. Just Water at the Source.” J-PAL Policy Briefcase (Septem- as a science of designing development interventions ber). http://www.povertyactionlab.org/publication may be emerging, so too might an art and science of /cleaner-water-source. Ahuja, Amrita, Michael Kremer, and Alix Peterson Zwane. service delivery in development. 2010. “Providing Safe Water: Evidence from Random- Second, R&D entails failure. Tendencies to continue ized Evaluations.” Annual Review of Resource Economics paying sunk costs or to pay attention only to evidence 2 (1): 237–56. that confirms their own biases (confirmation bias) can Allcott, Hunt. 2011. “Social Norms and Energy Conserva- prevent development practitioners from admitting to tion.” Journal of Public Economics 95 (9): 1082–95. and learning from failures. However, it is often through Ashraf, Nava, Oriana Bandiera, and B. Kelsey Jack. Forth- the process of experimenting, failing, and learning coming. “No Margin, No Mission? A Field Exper- from those failures that effective, evidence-based diag- iment on Incentives for Public Service Delivery.” noses and intervention strategies emerge. Journal of Public Economics. http://dx.doi.org/10.1016/j Creating the mechanisms for accommodating the .jpubeco.2014.06.014. Banaji, Mahzarin R. 2001. “Implicit Attitudes Can Be Mea- optimal levels of diagnosis, risk taking, and failure sured.” In The Nature of Remembering: Essays in Honor is an organizational challenge. Ideally, development of Robert G. Crowder, edited by James S. Nairne, Henry practitioners would be accountable for the quality of L. Roediger III, Ian Neath, and Aimee M. Surprenant, the learning and experimentation strategy all along chap. 7. Washington, DC: American Psychological the intervention cycle rather than just for compliance Association. with the plans defined before the start of the interven- Banaji, Mahzarin R., and Anthony G. Greenwald. 2013. tion—a situation that cripples innovation, hampers Blindspot: Hidden Biases of Good People. New York: midcourse corrections, and stirs fears of failure. Delacorte. 200 WORLD DEVELOPMENT REPORT 2015 Beaman, Lori, Raghabendra Chattopadhyay, Esther Duflo, ————. 2008. “How High Are Rates of Return to Fertilizer? Rohini Pande, and Petia Topalova. 2009. “Powerful Evidence from Field Experiments in Kenya.” American Women: Does Exposure Reduce Bias?” Quarterly Journal Economic Review 98 (2): 482–88. of Economics 124 (4): 1497–1540. ————. 2011. “Nudging Farmers to Use Fertilizer: Theory Bennhold, Katrin. 2013. “Britain’s Ministry of Nudges.” and Experimental Evidence from Kenya.” American New York Times, December 7. http://www.nytimes Economic Review 101 (6): 2350–90. .com/2013/12/08/business/international/britains Fryer, Jr., Roland G. 2010. “Financial Incentives and Stu- -ministry-of-nudges.html?pagewanted=all. dent Achievement: Evidence from Randomized Trials.” Bettinger, Eric P., Bridget Terry Long, Philip Oreopoulos, Working Paper 15898, National Bureau of Economic and Lisa Sanbonmatsu. 2012. “The Role of Application Research, Cambridge, MA. Assistance and Information in College Decisions: Gerber, Alan S., and Todd Rogers. 2009. “Descriptive Social Results from the H&R Block FAFSA Experiment.” Quar- Norms and Motivation to Vote: Everybody’s Voting terly Journal of Economics 127 (3): 1205–42. and So Should You.” Journal of Politics 71 (1): 178–91. BIT (Behavioural Insights Team). 2012. Applying Behavioural Glennerster, Rachel, and Kudzai Takavarasha. 2013. Run- Insights to Reduce Fraud, Error and Debt. London: Cabinet ning Randomized Evaluations: A Practical Guide. Prince- Office, Behavioural Insights Team. ton, NJ: Princeton University Press. Blair, Graeme, and Kosuke Imai. 2012. “Statistical Analysis Greenwald, Anthony G., Debbie E. McGhee, and Jordan L. of List Experiments.” Political Analysis 20 (1): 47–77. K. Schwartz. 1998. “Measuring Individual Differences Bold, Tessa, Mwangi Kimenyi, Germano Mwabu, Alice in Implicit Cognition: The Implicit Association Test.” Ng’ang’a, and Justin Sandefur. 2013. “Scaling Up What Journal of Personality and Social Psychology 74 (6): 1464–80. Works: Experimental Evidence on External Validity Grynbaum, Michael. 2009. “New York’s Cabbies Like Credit in Kenyan Education.” Working Paper 321, Center for Cards? Go Figure.” New York Times, November 7, A1. Haynes, Laura, Ben Goldacre, and David Torgerson. 2012. Global Development, Washington, DC. “Test, Learn, Adapt: Developing Public Policy with Brune, Lasse, Xavier Giné, Jessica Goldberg, and Dean Randomised Controlled Trials.” Policy Paper, Cabinet Yang. 2013. “Commitments to Save: A Field Experiment Office, Behavioural Insights Team, London. in Rural Malawi.” Policy Research Working Paper 5748, Holbrook, Allyson L., and Jon A. Krosnick. 2009. “Social World Bank, Washington, DC. Desirability Bias in Voter Turnout Reports: Tests Using Bryan, Gharad, Dean Karlan, and Scott Nelson. 2010. the Item Count Technique.” Public Opinion Quarterly 74 “Commitment Devices.” Annual Review of Economics 2 (1): 37–67. (1): 671–98. IDinsight. 2014a. “Measuring the Impact of Mama Kits on CFED and ideas42. 2013. “Small Changes, Real Impact: Facility Service Delivery Rates in Rural Chadize and Applying Behavioral Economics in Asset-Building Serenje Districts in Zambia.” Technical Report for the Programs.” http://www.ideas42.org/publication/view Government of Zambia. Available at http://idinsight /small-changes-real-impact-applying-behavioral .org/wp-content/uploads/2014/10/Mama-Kits-Report -economics-in-asset-building-programs/. -Aug-2014.pdf. Datta, Saugato, and Sendhil Mullainathan. 2014. “Behav- IDinsight. 2014b. “Measuring Community-Level Point ioral Design: A New Approach to Development Policy.” ITN Distribution Dynamics and the Impact of CHW Review of Income and Wealth 60 (1): 7–35. Hang-Up in Rural Rufunsa District in Zambia.” Tech- Droitcour, Judith, Rachel A. Caspar, Michael L. Hubbard, nical Report for the Government of Zambia. Available and Trena M. Ezzati. 2011. “The Item Count Technique at http://idinsight.org/wp-content/uploads/2014/10/ITN as a Method of Indirect Questioning: A Review of Its -in-progress-report-October-2014.pdf. Development and a Case Study Application.” In Mea- iNudgeyou. 2014. “When Prescription Works Better than surement Errors in Surveys, edited by P. B. Beimer, R. M. Proscription.” Blog, February 10. http://inudgeyou.com Groves, L. E. Lvberg, N. A. Mathiowetz, and S. Sudman, /when-prescription-works-better-than-proscription/. 185–210. Hoboken, NJ: John Wiley & Sons. Kahneman, Daniel, Jack L. Knetsch, and Richard H. Thaler. Duflo, Esther, and Michael Kremer. 2005. “Use of Ran- 1990. “Experimental Tests of the Endowment Effect domization in the Evaluation of Development Effec- and the Coase Theorem.” Journal of Political Economy 98 tiveness.” In Evaluating Development Effectiveness, Vol. 7 (6): 1325–48. of World Bank Series on Evaluation and Development, Karlan, Dean, Margaret McConnell, Sendhil Mullaina- edited by Osvaldo N. Feinstein, George Keith Pitman, than, and Jonathan Zinman. 2010. “Getting to the Top and Gregory K. Ingram. New Brunswick, NJ: Transac- of Mind: How Reminders Increase Saving.” Working tion Publishers. Paper 16205, National Bureau of Economic Research, Duflo, Esther, Michael Kremer, and Jonathan Robinson. Cambridge, MA. 2007. “Why Don’t Farmers Use Fertilizer? Experimen- Karlan, Dean, and Jonathan Zinman. 2012. “List Random- tal Evidence from Kenya.” Working Paper, Massachu- ization for Sensitive Behavior: An Application for Mea- setts Institute of Technology and Harvard University, suring Use of Loan Proceeds.” Journal of Development Cambridge, MA. Economics 98(1): 71–75. ADAPTIVE DESIGN, ADAPTIVE INTERVENTIONS 201 Kondylis, Florence, Maria Jones, and Daniel Stein. 2013. Richburg-Hayes, Lashawn, Caitlin Anzelone, Nadine “Rwanda: Program of Impact Evaluations in Agri- Dechausay, Saugato Datta, Alexandra Fiorillo, Louis culture.” Development Impact Evaluation Initiative Potok, Matthew Darling, and John Balz. 2014. “Behav- (DIME) Concept Note, March, World Bank, Washing- ioral Economics and Social Policy: Designing Innova- ton, DC. https://www.gafspfund.org/sites/gafspfund tive Solutions to Programs Supported by the Admin- .org/files/Documents/LWH%20Impact%20Evaluation istration for Children and Families.” OPRE Report No. %20Concept%20Note_final.pdf. 2014 16a, Office of Planning, Research, and Evaluation Kremer, Michael, Jessica Leino, Edward Miguel, and Alix (OPRE), U.S. Department of Health and Human Ser- Peterson Zwane. 2011. “Spring Cleaning: Rural Water vices, Washington, DC. Impacts, Valuation, and Property Rights Institutions.” Schultz, P. Wesley. 1999. “Changing Behavior with Norma- Quarterly Journal of Economics 126 (1): 145–205. tive Feedback Interventions: A Field Experiment on Ludwig, Jens, Jeffrey R. Kling, and Sendhil Mullainathan. Curbside Recycling.” Basic and Applied Social Psychology 2011. “Mechanism Experiments and Policy Evalua- 21 (1): 25–36. tions.” Journal of Economic Perspectives 25 (3): 17–38. Shih, Margaret, Todd L. Pittinsky, and Nalini Ambady. Pop-Eleches, Cristian, Harsha Thirumurthy, James Habyar- 1999. “Stereotype Susceptibility: Identity Salience and imana, Joshua Graff Zivin, Markus Goldstein, Damien Shifts in Quantitative Performance.” Psychological Sci- de Walque, Leslie MacKeen, Jessica Haberer, John Sidle, ence 10 (1): 80–83. and Duncan Ngare. 2011. “Mobile Phone Technologies Thaler, Richard H., and Shlomo Benartzi. 2004. “Save Improve Adherence to Antiretroviral Treatment in More Tomorrow™: Using Behavioral Economics to Resource-Limited Settings: A Randomized Controlled Increase Employee Saving.” Journal of Political Economy Trial of Text Message Reminders.” AIDS 25 (6): 825–34. 112 (S1): S164–S87. Pritchett, Lant, Salimah Samji, and Jeffrey Hammer. 2013. Wansink, Brian, and Koert van Ittersum. 2003. “Bottoms “It’s All About MeE: Using Structured Experiential Up! The Influence of Elongation on Pouring and Con- Learning (‘e’) to Crawl the Design Space.” Working sumption Volume.” Journal of Consumer Research 30 (3): Paper 322, Center for Global Development, Washing- 455–63. ton, DC. 202 WORLD DEVELOPMENT REPORT 2015 Why should governments shape individual choices? All people—rich and poor alike—sometimes make anyone else saw a person attempting to cross a choices that do not promote their own well-being. bridge which has been ascertained to be unsafe, Although mistakes can arise even after careful delib- and there were no time to warn him of his dan- eration, people are especially prone to make choices ger, they might seize him and turn him back, that do not reflect their long-term interests when they without any real infringement of his liberty; for think automatically. Automatic thinking means not liberty consists in doing what one desires, and he bringing to bear full knowledge about the dimensions does not desire to fall into the river. (Mill 1859, 95) and consequences of choices. People may also get stuck in habits, succumb to inertia, and repeatedly pro- Just as that man did not desire to fall into the river, crastinate despite intentions to do otherwise. Mental most of us do not want to be forgetful, to procrastinate, models filter the information that people receive and or to miss out on important opportunities. pay attention to and shape their interpretations. Social Second, because decision making is often based on pressures and social norms can function like taxes or only the most accessible and salient information and is subsidies on behavior, making some choices easier also influenced by subtle social pressures and received and others harder (Sunstein 1996); when internalized, mental models, individuals’ preferences and immedi- social norms shape cognitions, emotions, and even ate aims do not always advance their own interests. Spotlight 6 physiological reactions. Individuals might choose differently, in ways more consistent with their highest aspirations, if they had Using psychological and social insights more time and scope for reflection. The assumption to promote freedom and well-being that individuals always make choices that promote This Report provides evidence that these phenom- their own interests—often a fundamental benchmark ena are widespread and significantly affect choices, for policy analysis—is misguided. But if decision mak- behaviors, well-being, and important development ers do at times require assistance, what guidelines outcomes. What should development actors—whether are to be used for the policy interventions aimed at development professionals, nongovernmental organi- shaping choice? Development actors should focus on zations, governments, or international agencies—do the most important freedoms. In the development with this knowledge? There are three compelling rea- context, these include freedom from poverty, disease, sons to use this knowledge to promote both freedom and oppression. and well-being. Although older accounts described liberty, as Mill First, doing so helps people obtain their own goals. does above, as “doing what one desires,” and argued Reminders to save money or take medicine help that the only legitimate limitations on desire involve people who are otherwise caught up in life achieve interpersonal harm, more contemporary accounts objectives that they themselves have set. Commitment distinguish between desires of greater and lesser sig- contracts, which markets underprovide, can reinforce nificance. The freedoms to express one’s thoughts and decisions to adopt beneficial behaviors. Matching the feelings in speech and to live a long and healthy life timing of social transfers to the timing of charges for are highly valued. By contrast, the “freedom” to forget school enrollment, or making it easier to buy fertilizer to sign up for a savings plan is less important. Most of at harvest time when cash is at hand, helps overcome us do not prize the freedom to purchase a genuinely intention-to-action divides for people who may be dangerous medicine from a pharmacy and prefer that forgetful or possess insufficient willpower (that is to government place at least some limits on the kinds of say, all of us). Many development policies that operate medicine we can buy. at the boundary of economics and psychology can be The philosopher Charles Taylor (1985) compares understood in these terms. John Stuart Mill, the great two countries. One has limited freedom of conscience. champion of personal liberty, acknowledged a legiti- The other ensures freedom of conscience but has mate role for government in providing both protection many, many more traffic lights. The country with all and information. He put it this way: the traffic lights, in sheer quantitative terms, restricts many more choices, but most would agree that people [It] is a proper office of public authority to guard live more freely in it. The example demonstrates that against accidents. If either a public officer or it matters which choices are constrained and which are WHY SHOULD GOVERNMENTS SHAPE INDIVIDUAL CHOICES? 203 encouraged; and most people agree that when govern- those that aim to shape individual choice. Moreover, ments shape crucial choices, such as those involving some of the recent findings reviewed in this Report the escape from poverty, they are casting development warrant less government intervention, not more— as a kind of freedom (Sen 1999) and making a trade-off sometimes local social norms can resolve collective that is appropriate. action problems more effectively than regulation and Third, socially reinforced practices can block choices taxation can. that enhance agency and promote well-being and In most instances, governments are only one prevent individuals from even conceiving of certain among many players who seek to influence the choices courses of action, as when discrimination and inequal- that people make. Moneylenders and banks frame ity sometimes lead people, understandably, to adopt the complexity of the loans they offer. Firms tempt low aspirations. This Report argues that social interde- individuals with tasty but unhealthy foods and easy pendence and shared mental models affect significant money. Elites of all types enforce informal rules and choices, sometimes creating traps for communities and shape public opinion in ways that benefit themselves individuals, such as low trust, ethnic prejudice, and gen- as a group. Any number of interested parties exploit der discrimination. The social practice of female genital people’s tendency to think automatically (Akerlof and cutting is one example; tax compliance, corruption, road Shiller, forthcoming). safety, outdoor defecation, and environmental conser- vation also hinge on interdependent choice. These are situations in which public action targeting shared men- Governments should act when tal models, social norms, and other collective goods, inadequate engagement, situational both physical and symbolic, may change outcomes in ways that make some better off but others worse off. framing, and social practices In these situations, government action on behalf of agency can be justified, as well. Although development undermine agency and create or Spotlight 6 actors have legitimate differences concerning some of these issues and place different weights on individual perpetuate poverty. freedoms and collective goals, widely shared and rati- fied human rights constitute a guiding principle for addressing these trade-offs. With these other forces at work, government should not play the role of a neutral referee. When it is widely understood that private actors can and should pursue An additional justification for their self-interest, private sector encroachment on government action agency is to be anticipated. It will be uncommon for the The standard justifications for government action influences on decision making to be evenly balanced. In in market economies are monopolies, externalities, that context, governments that do not restrain or coun- public goods, asymmetric information, redistribution, terbalance concerted efforts to influence choice, such and macroeconomic stabilization. This Report adds as deceptive framing and misleading advertising, may another. Governments should act when inadequate be seen not only to permit but even to encourage them. engagement, situational framing, and social practices John Stuart Mill was also receptive to government inter- undermine agency and create or perpetuate poverty. vention when third parties with vested interests, such As noted, these efforts should themselves be guided by as liquor houses, were the ones providing individuals a healthy respect for individual dignity and welfare— with information because, as he put it, “sellers have a for the freedom of individuals to articulate and imple- pecuniary interest in promoting excess.” Governmental ment their own vision of a good life and for a respect inaction does not necessarily leave space for individual for human rights. freedom; rather, government inaction may amount to In this approach, the identification of market fail- an indifference to the loss of freedom. ures remains a useful criterion for public action in mar- kets in which one can reasonably assume that behavior References is indicative of individual preferences. However, one Akerlof, G. A., and R. Shiller. Forthcoming. “Phishing for cannot assume that this is always or even mostly the Phools.” Unpublished manuscript. case, particularly in nonmarket settings. Policy makers Mill, J. S. 1859. “On Liberty.” Indianapolis: Hackett. themselves, moreover, are subject to cognitive errors, Sen, Amartya K. 1999. Development as Freedom. New York: including confirmation bias and the use of possibly Knopf. inappropriate mental models (as discussed in chapter Sunstein, C. R. 1996. “Social Norms and Social Roles.” 10). As a consequence, they should search for and rely Columbia Law Review 96 (4): 903–68. on sound evidence that their interventions have their Taylor, C. 1985. Philosophy and the Human Sciences. Vol. 2 intended effects and allow the public to review and of Philosophical Papers. Cambridge, U.K.: Cambridge scrutinize their policies and interventions, especially University Press. Index Boxes, figures, notes, and tables are indicated by “b,” “f,” “n,” and “t,” respectively, following page numbers. A seasonal earnings and stress, 14, 27, 82–83, 83f, 88 Abecedarian Project (U.S.), 108n1 social networks’ and peers’ role in adoption of new accessibility of information, 30 technology, 139 Acemoglu, Daron, 66 subsistence farmers, government intervention for, adaptive design/adaptive interventions, 19–21, 192–201 198 design of intervention, 195–97, 196t technology adoption and, 136–39 diagnosis of psychological and social obstacles, weather insurance and, 51 194–95 AIDS. See HIV/AIDS experimental approach to identify cost-effective Alesina, Alberto, 65 interventions, 192, 197t, 198, 198b altruism, 44–46 measurement techniques to reveal biases and ambiguous risk and climate change, 165 obstacles to, 195b anchoring, 30–31, 197t mechanism experiments, 196 Andrews, Matt, 182 need to change process of arriving at solutions, 194 antipoverty policies and programs. See poverty recognizing thinking patterns and, 192 Argentina research and development (R&D), continual need for, climate change views in, 165 199 squatters’ vs. landowners’ attitudes in, 85 advertising, persuasiveness of, 122 Argyris, Chris, 182 Africa. See also specific countries Ariely, Dan, 151, 171 agriculture and technology adoption in, 136 Aronson, Joshua, 73n10 behavioral health in, 155n5 Ashraf, Nava, 153 female genital cutting in, 53, 69 Assessment Report of the Intergovernmental Panel on Climate slave trade in, 65 Change (2014), 160 social learning about health in, 148–49 automatic enrollment in savings and retirement plans, 118, utility companies providing procedures manuals to 197t employees in, 144 automatic thinking, 3, 5–6, 6t, 25, 26–40 The Age of Stupid (film), 166 accessibility of information and, 30 agriculture association tests to determine, 195b agricultural extension activities involving peer biases in. See biases farmers, 10 choice architecture and, 6, 36–37, 37f climate change views and, 163–65, 164f climate change, effect on understanding, 163–64 cognitive function and fluid intelligence affected by default options and, 34–35, 35f financial distress of farmers, 14, 83, 83f detail creating believability in, 30, 38nn4–5 development professionals on agricultural reform, of development professionals, 20, 181 187b financial decision making and, 14, 31–32 farmer-managed irrigation systems as superior to framing in. See framing effect government-managed systems, 47–48, 55n3 intention-action divides and, 37–38, 114 mental models influenced by production modes, 65 loss aversion and, 29, 35–36 prepurchase of fertilizer by farmers, 88, 137, 137f negatives of, 202 productivity of farmworkers, 134 nudges and reminders and, 36, 38, 39n10 seaweed farming and, 138–39, 138f partial worldview from, 7f 205 206 INDEX automatic thinking (cont.) Caprio, Gerard, 184b “peanuts effect” and, 32 carbon dioxide release. See climate change perception and, 63 Carbon Disclosure Project (CDP), 170 in two systems of thinking, 5–6, 6t, 26–29, 27t, 164 cash transfer programs. See conditional cash transfers (CCTs) availability heuristic, 72n3, 161 CCTs. See conditional cash transfers Ayers, John W., 148 celebrities environmental conservation messages from, 10, 176 B health information shared by, effect of, 147–48, 148f Bandura, Albert, 76 Chicago Heights Early Childhood Center, 108 Banerjee, Abhijit Vinayak, 32 child abuse, 104 Bangladesh child development. See early childhood development government health providers providing free medicine child mortality related to lack of hygiene, 152 to drug sellers in, 195b China health promotion campaigns in, 17, 146–47, 149 corruption in, 60 purdah limits on women’s activities, effect in, 54 employee productivity in, 130, 132 Rural Advancement Committee, 17 mental models influenced by agricultural production sanitation needs, CLTS in, 155n3 modes in, 65 Barth, James, 184b weather insurance in, 51 Bates, John E., 103 choice architecture, 6, 36–37, 37f Bayles, Kathryn, 103 civic capacity, building of, 50–51 behavioral economics, 26 civil conflicts, 45–46, 45f Behavioural Insights Team (U.K.), 90, 189, 198, 198b religious violence and, 49–50 belief traps, 69 climate change, xi, 17–18, 160–74 biases. See also gender; social norms; women automatic cognitive processes, effect on in assessing information, 29–33 understanding, 163–64 in assessing value, 34–36 biases’ effect on understanding of, 161–62, 171n1 in climate change response, 161–62, 171n1 cultural worldviews and social networks, 162–63 confirmation bias, 18, 27–28, 69–70, 182–85, 190 default setting, 170–71 courtesy bias, 195 democratic rules and laws, effect of, 167, 167f of development professionals/policy makers, xii, entertainment education, effect of, 165–66 180–91, 184b. See also development professionals/ fairness and, 166–67 policy makers farmers’ views on, 163–64, 164f financial decision making and, 112 future risks, discounting of, 165 of health care providers, 154 future vs. present concerns, 165 health choices, effect on, 146–49 greenhouse gases, effect of, 160–61 home team advantage, 184b inertia as response to, 161 measurement techniques to reveal, 195b information campaigns on, 163, 167–68, 169–70 present bias. See present bias prestige as motivator to change behavior for, 44 self-serving bias, 18, 166 social norms to reduce consumption, 167–69 Bolivia, savings reminders in, 15, 120 sustainability of resources, 167, 167f boomerang effect, 163 water conservation, promotion of, 176–77, 177f Botswana’s safe-sex programs, reminders to overcome Climate Performance Leadership Index, 170 psychological resistance or cognitive biases in, 39n10 coercive behavior, 170 brain development, 101, 101f cognitive capacity, 3, 5b, 100. See also executive function Braman, Donald, 162 cognitive flexibility, 100 Brazil cognitive frames. See framing effect; mental models anticorruption campaigns in, 61 cognitive overload, 5b, 115–16, 136 antismoking information and laws in, 147–48, 148f cognitive tax of poverty, 14, 81, 86–89, 115 climate change views in, 163 collective action energy rationing in, 18, 171 to overcome corruption, 61 entertainment education in, 70, 76 shared mental models for, 62 financial education in, 123 Colombia simplification of voting procedures in, 36–37, 37f, 39n9 bank compensation practices and productivity in, 128 breastfeeding, effect on health of children, 147 conditional cash transfer program for school bribery. See corruption attendance in, 38 Burkina Faso, climate change views in, 163–64 school readiness differences between rich and poor families in, 99, 100f C water conservation in, 10, 176–77, 177f California, access to health insurance for minority poor in, 87 commitment devices, 15, 20, 38, 50, 120, 121, 121f, 129–30, 140, Cameroon 151, 197t, 199, 202 Cameroon Electricity Company, 144 common sense, 12, 190 income protection in, 86 community-driven development (CDD) programs, 50–51 Canada community involvement productivity of employees in, 131 bringing consensus to polarizing issues, 184 trust development in children in, 71, 71f, 108n1 climate change risks, reduction of, 165 INDEX 207 community development agents’ role in marginalized in health care, 149–50 and conflict-affected communities, 134 for household finance, 118–19, 197t community leaders participating in social programs, deliberative thinking, 6, 6t, 26–27, 27t effect of, 90 Denmark, indoor smoking in, 194 Community-Led Total Sanitation (CLTS), 17, 152–53, design of policies. See adaptive design/adaptive interventions; 153f, 155n3 policy design early childhood development, support for parents to detail creating believability, 30, 38nn4–5 change from social norms, 106 development professionals/policy makers, 18–19. See also health care support, 50, 152–53 adaptive design/adaptive interventions school scorecards, monitoring of, 48 automatic thinking of, 20, 181 conditional cash transfers (CCTs) biases of, xii, 20, 180–91, 184b community leaders participating in, 90 confirmation bias, 18, 27–28, 69–70, 182–85, criminal behavior changed through, 46 190 early childhood development and, 105 sunk cost bias, 185–86, 186f health care, commitment mechanisms to, 151 complexity of work of, 18–19, 21f, 181–82, 193f health insurance enrollment and, 89 data interpretation by, 18, 182–83, 183f school attendance and enrollment tied to, 13t, 38 difficulty in predicting views of poor people, 5, 18, 94, conditional cooperation, 7, 10f, 47, 47–48f, 52, 167 188b confirmation bias, 18, 27–28, 69–70, 182–85, 190 errors of, 3, 69–70, 190 red teaming to overcome, 19, 184, 190n2 home team advantage and, 184b consumer decisions in credit markets, 31–32 judgment and decision making, effects of context on, context. See framing effect 20, 186–89 cooperation knowledge gaps of, 180–81 conditional cooperation, 7, 10f, 47, 47–48f, 52, 167 personal distance needed, 195b “crowd in” vs. “crowd out” cooperation, 48–49 randomness, introduction of, 195b culture of honor and, 67 red teaming to overcome confirmation bias, 19, 184, economic yields based on, effect on mental models, 190n2 65 dictator game, 45 inclusive societies and, 66–67 discrimination, 9. See also fairness; gender; women corruption, 2, 9, 60–61 caste system and, 12, 12f, 72–73n9 Costa Rica, climate change views in, 165 epistemic injustice, 70 cost-effective behavioral interventions, 13t, 197t Dixon, Suzanne, 103 country differences dogfooding, 19, 87, 187–89, 195 in early childhood development, 14 Dominican Republic, financial education in, 121, 140 in financial decision making, 112 Douglas, Mary, 162 in parenting practices, 103–4, 104f Duflo, Esther, 32 similarities in ways of thinking, 4 Dupas, Pascaline, 150 credit card borrowing, 114, 116 Dweck, Carol, 106 criminal identity, effect of increasing salience of, 67–68, 67f dynamic complementarity argument, 100 culture. See also worldviews caste system and, 72n9 E defined, 72n5 early childhood development, 14, 98–110 norm breaking in oppositional cultures, 55n6 brain development affected by poverty, 101 parenting behavior and, 103 cognitive and noncognitive skills needed in, 100 productivity enhanced by understanding of, 144–45 country differences in parenting practices, 103–4, 104f shared schemas of, 12, 62, 72n5 design of interventions to improve parental culture of honor and cooperation, 67 competence, 104–8 Cunha, Flavio, 100 changing mindsets and underlying belief systems, 105–6 D complementary classroom-based Das, Jishnu, 154 interventions, 107–8 Datta, Saugato, 193 opportunity for parents to learn and practice The Day after Tomorrow (film), 165 new skills and improve their mental decision making. See also default options; thinking health, 106–7 assumptions involved in, 3 educational success tied to multiple cognitive and dual-process, 121–22 noncognitive skills, 100 economic theory about, 5b financial decision-making skills learned in, 122–23 principles of, 5–13 language learning, 99, 101–2 standard economic theory and, 29, 29f mental models changed through, 70–71, 71f “thick descriptions” and understanding social and parents’ role in, 14, 15f, 101–3 cultural context for, 145, 194 executive function skills acquisition, 102–3 default options, 34–35, 35f, 197t improving parental skills incrementally, 107, for climate change, 170–71 107f college applicants’ test scores offered free, effect of improving parents’ own mental well-being, default of, 34, 35f, 39n7, 88–89, 149 107 208 INDEX early childhood development, parents’ role in (cont.) low psychological agency of disadvantaged persons language learning, 101–2 in, 4 parenting style, differences in, 103–4 savings rates in, 4 scaffolding, 102–3 ethnographic approach, 144–45, 194 school readiness differences between rich and poor executive function, 83, 84f, 85 families, 99–100 skills acquisition in early childhood and, 99, 101, Ebola, xi–xii 102–3, 108 economic theory about human decision making, 5b F conditional cooperation and, 47, 48f fairness decision making and, 29, 29f climate change and, 166–67 dictator game and, 45 tax policies and, 52 mental accounting, 4, 72n2 Fama, Eugene, 113 mental models of cooperation and, 65 family planning, 54 Ecuador farming. See agriculture conditional cash transfers and early childhood feedback use, 197t development in, 105 females. See women preschool teachers’ role in, 108 fertility school readiness differences between rich and poor interventions for, 13, 54, 70, 76 families in, 99, 100f mental models as belief traps about, 69 education. See also early childhood development; 5th Pillar (NGO), 60 entertainment education financial advice, 116–17 automatic thinking and, 28 financial decision making. See household finance civic education leading to wider gender gap, 51–52 financial education, 4, 121 conditional cash transfer program for school Fishkin, James, 184–85 attendance, 13t, 38 Fitoussi, Jean-Paul, 170 financial education, 4, 121 fluid intelligence, 14, 83, 84f, 85 girls’ education constraints, 52 framing effect, 4, 27, 28f, 181 health education, 146–47 biases and, 30 higher education climate change and, 165 college applicants’ test scores offered free, default options and, 34–35 effect of default of, 34, 35f, 39n7, 88–89, 149 description and presentation, 27 low-income students applying for financial early childhood development and, 106 aid, 87, 197t household finance choices, 115–16, 117–18 mental models changed through, 70–71, 71f mental editing and interpretation, 27, 64, 64f parental aspirations for children and, 4 payday loans and, 6, 8f, 32, 33f the poor and, 80, 85 poverty, understanding contexts of, 94–97 returns to education, 14, 32, 89 reframing decisions, 8f, 33f school scorecards, community monitoring of, 48 vaccinations and HIV testing, 149–50 student performance work tasks and compensation, 130–33 caste identity and, 68, 68f freedom and well-being, promotion of, 202–3 micro-incentives for improving, 197t free riders, 10f, 46, 47, 47–48f, 49 teacher bonuses based on, 16, 132 Friedman, Milton, 5b emotional persuasion, 121–22 fuel efficiency, 169 employment. See productivity; wage increases future risks, discounting of, 165. See also present bias Energiedienst GmbH (German power company), 170 energy efficiency, 168–69 G energy rationing, 18, 171 Geertz, Clifford, 145 Engel, Stefanie, 47 gender. See also women entertainment education, 13, 76–77 division of labor by, 65 business models for, 77 social norms and, 51–52 climate change, effect on, 165–66 Georgia defined, 76 corruption in, 61 evidence of impact, 76–77 war-related violence, effect on children and young financial decision making, effect on, 122, 122f adults in, 45–46, 45f mental models changed through, 70 Germany soap operas, effect on social norms, 4, 13, 53–54, 70, employee productivity in, 133 122, 122f green energy in, 170 theory behind, 76 Ghana entrepreneurs, productivity of, 135–36 agriculture and technology adoption in, 136, 139 epistemic injustice, 70 anticorruption campaign in, 61 equilibrium behaviors, 51 entrepreneur support in, 135 Ethiopia mosquito net use in, 195b forest users groups, conditional cooperation in, 47 Giné, Xavier, 118 INDEX 209 Giuliano, Paola, 65 home team advantage, 184b Gneezy, Ayelet, 38 home visiting programs for maternal skills acquisition, 107 Goebbert, Kevin, 162 household finance, 14–16, 112–26 governance by indicators, 170 automatic thinking and, 14, 31–32 government interventions to promote individual choice, 20, cognitive overload and narrow framing, 115–16 202–3 investment opportunities, risk tolerance for, 113–14 Grayman, Walter M., 190n2 policies to improve decision making, 117–23 green energy, 170–71 changing default choices, 118–19, 119f greenhouse gases. See climate change early childhood development of preferences, Grothmann, Torsten, 164 122–23 group attachment, 45–46, 45f emotional persuasion, use of, 121–22 group parenting programs, 106 entertainment education as part of, 122, 122f Guiso, Luigi, 67 framing choices, 117–18 microfinance and, 50, 119 H nudges and reminders, 119–20 Hammer, Jeffrey, 154 overcoming temptation through commitment, Hastie, Reid, 183 120–21 Hayek, F. A., 5b simplifying and targeting financial education, Head Start (U.S.), 107–8 121 health, 17, 146–58. See also HIV/AIDS; mosquito nets present bias, 112, 114–15 child mortality from lack of hygiene, 152 social psychology of advice relationship, 116–17 community-level responses, 50, 152–53 human behavior. See also decision making; psychological and conditional cash transfers and commitment social approaches; thinking mechanisms, 151 cost-effective behavioral interventions, 13t, 197t follow-through and habit formation, 151–53 improving, xi government health providers providing free medicine Human Development Index, 14, 15f, 104, 104f to drug sellers in, 195b knowledge, attitudes, and practices (KAP) surveys I on, 151 identity mass media, use of, 147–48 group dynamics and, 44–46 medical regimens, adherence to, 38, 150–52, 152f mental models’ influence on, 67–68, 67f, 73n10, 197t reminders to increase, 13t, 152, 152f, 154 immigrants’ mental models, 64–65 mosquito nets, free provision of, 150 impulse control, 100 oral rehydration therapy (ORT), 17, 146 India preventive care and opt-out basis, 149–50 anticorruption campaigns in, 61 price decreases, effect on adoption of new behavior, antipoverty program’s spillover effects in, 89 17, 150–51, 150f caste system in, 72–73n9 psychological and social approaches to changing clientelism in, 66 health behavior, 149–51 psychological biases and social influences, effect of, climate change views in, 163 146–49 corruption in, 60–61 quality of health care providers, 17, 153–55 employee productivity in, 16, 130, 133, 135, 140 recruitment of health care workers, 134, 154–55 wage increases, effect on, 131 sanitation policies and disease burden, 17, 152–53, 153f, entertainment education in, 76, 77 155n3 epistemic injustice against women in, 70 social learning about health care quality, 148–49 female sex workers, building self-esteem of, 46 vaccinations and HIV testing, framing information Green Revolution in, 139 about, 17, 18, 149–50 health care professionals in, 154 Heckman, James J., 100 health information campaigns, failure when Henrich, Joseph, 65 conflicting mental model involved, 17, 147 heuristics Hindus and Muslims both participating in voluntary availability heuristic, 72n3, 161 associations in, 49–50 health care providers relying on, 154 household finance decision making in, 120 prototype heuristic, 72n4 microfinance in, 50, 121 Hirschman, Albert, 5b national employment programs in, 32 HIV/AIDS, xi open defecation, effects of, 17, 152–53, 153f adherence to antiretroviral therapy (ART), 151, 192 “peanuts effect” and fruit vendors in, 32 HIV testing political affirmative action information campaigns about, 147, 149 for low-caste individuals in, 70 opt-in vs. opt-out approach to, 150 for women in, 70 intention-action divides and, 37–38 procrastination in, 129 safe-sex precautions, 44, 133, 133f, 193 self-employment in, 135 sexual behavior and, misperceptions about, 52 student performance and caste identity in, 12, 12f, 68, Hoff, Karla, 69 68f 210 INDEX India (cont.) poverty in sugar cane farmers’ mental stress in, 14, 27, 82–83, wealthy vs. poor respondents to determine 83f, 88 decision-making skills, 95, 96f individual choices, governments shaping, 202–3 World Bank staff predictions about views held Indonesia by the poor, 18, 94 open defecation, effects of, 17, 152–53, 153f productivity in self-employed and small businesses seaweed farming and productivity issues, 138–39, 138f in, 135 wealthy vs. poor respondents to determine decision- traffic fatalities in, 52–53, 53f making skills in, 95, 96f World Bank staff predictions about views held by the World Bank staff predictions about views held by the poor in, 188b poor in, 18, 94, 188b Keynes, John Maynard, 5b inequality of wealth, persistence of, 65 Kosfeld, Michael, 47 inertia and climate change response, 161 Kriss, Peter H., 166 informal social insurance, 85 information campaigns L avoidance of, 115 Lange, Andreas, 166 on bus safety, 13t, 53, 53f language learning. See early childhood development on climate change, 163, 167–68, 169–70 Latin America. See also specific countries for farmers, 51 entertainment education in, 77 on health risks, 32, 146–47, 149–50 investment options, risk tolerance in, 113 on returns to higher education, 32 school readiness differences between rich and poor tax policies and, 52 families in, 99, 100f on water conservation, 176 law’s ability to change social norms, 53 institutions, relationship with mental models, 65–67, 70 in climate change views, 167, 167f instrumental reciprocity, 46 learning. See also social learning intention-action divides, 16, 37–38, 114, 135, 140, 192 mental models and, 64–65 Intergovernmental Panel on Climate Change (IPCC), 160, 166, Lebanon, community support for breastfeeding program in, 169 50 international climate agreements, 167 Lesotho, agricultural reform in, 187b Internet, anticorruption campaigns on, 61 Levine, Ross, 184b intrinsic reciprocity, 46 Liberia, social interventions to change behavior of criminals investment opportunities, risk tolerance for, 113–14 in, 46 ipaidabribe.com, 61 life insurance, 116 Israel, crowding-out phenomenon in day-care centers in, 49 loans and consumer decision making, 14–15, 31–32, 33f Italy, belief traps in, 69 loss aversion, 29, 35–36, 113, 197t Iyengar, Shanto, 171n2 Lula da Silva, 147–48, 148f J Jacoby, Hanan G., 52 M Jamaica, home stimulation intervention for stunted children Madagascar, school readiness differences between rich and in, 14, 107, 107f poor families in, 99, 99f Jenkins-Smith, Hank, 162 malarial prevention. See mosquito nets Malawi K agriculture in Kağıtçıbaşı, Cigdem, 103 agricultural extension activities, 10 Kahan, Dan, 19, 162, 182 female farmers, view of, 70 Kahneman, Daniel, 26, 38n2, 63, 164, 181 harvest profits, use of, 192 Kamenica, Emir, 55n2 seed technology in, 51 Kar, Kamal, 155n3 technology adoption and, 139 Kenya savings rates in, 121, 121f agriculture in Mali, civic education leading to wider gender gap in, 51–52 commitment devices for farmers, 140 Mandela, Nelson, 68 prepurchase of fertilizer by farmers, 88, 137, Mansuri, Ghazala, 52 137f, 140, 198 Martinez Cuellar, Cristina, 118 education and parenting styles in, 103 mass media. See also entertainment education; information health issues in campaigns access to clean water, 13t, 19, 193–94 aspirations increased due to, 13t HIV/AIDS drug regimens, 38 development interventions involving, 13 poor people’s view on vaccines, 180 financial decisions affected by, 122, 122f household finance in health education through, 146–48 cash transfer to rural households, 89 mental models changed through, 70 investment rates, 120 social norms changed through, 53–54 savings and income protection, 4, 86, 90 maternal depression and early childhood development, information campaigns and entertainment education 105 in, 76 Mazar, Nina, 151 INDEX 211 Mazer, Rafael Keenan, 118 government health providers providing free medicine mental accounting, 4, 72n2, 116 to drug sellers in, 195b mental models, 3, 11–13, 11f, 25, 62–75 Netherlands, employee productivity in, 133 attention and perception, 63, 64f, 69, 169 networks. See social networks beliefs tested across society, 69 Newfoundland settlement, 160 belief traps, 69 New Jersey of climate change, 160 police officers’ compensation in, 132 context and, 64, 64f, 67, 68f poor vs. affluent thinking processes in, 94–95, 94f corruption perpetuated by, 60 Nicaragua defined, 11, 62, 72n1 antipoverty programs in, 90 development professionals not aware of poor people’s, entrepreneurs, access to business grant program in, 136 187 school readiness differences between rich and poor education methods and early childhood interventions families in, 99, 100f making changes in, 70–71, 71f, 105–6. See also early Nigeria, entertainment education in, 76 childhood development Nisbett, Richard E., 63 framing and, 27. See also framing effect nongovernmental organizations (NGOs) how mental models work, 63–65 anticorruption campaigns of, 60 ideology and confirmation bias, 69–70 in entertainment education, 77 importance of, 13, 63 experimenting during implementation by, 198 influence on identity, 67–68 prepurchase of fertilizer offered by, 137 information campaigns, effect on, 147 norm entrepreneurs, 54 institutional relationships with, 65–67 North Carolina, education programs aimed at low-income intergenerational transfer of, 64–65 families in, 87 matching to decision context, 70–71 nudges and reminders media making changes in, 13, 70 to change behavior, 13t, 36, 39n10 persistence of, 68–70 to health care workers, 154 poverty and, 14, 63 HIV/AIDS drug regimens and, 38, 152, 152f as shared schemas of a culture, 62, 72n5 of household finance decision making, 119–20, 197t social influences on, 55n1 Nunn, Nathan, 65 sources of, 63, 65–67 use of, 63–65 O Mexico observation’s effect on behavior, 47, 49f community development agents in, 134 Olson, Sheryl L., 103 conditional cash transfer program in, 90 open defecation, effects of, 17, 152–53, 153f early childhood development and, 105 Oportunidades (Mexican CCT program), 105 entertainment education in, 77n1 Opower, 18, 167–68 financial decision making of low-income populations oral rehydration therapy (ORT), 17, 146 in, 15, 16f, 31, 32, 34f Ostfeld, Avi, 190n2 financial products offered in, 118 Ostrom, Elinor, 165 microfinance, 50, 119, 121 micro-incentives, 13t, 197t P Mill, John Stuart, 202, 203 Pakistan, educational opportunities for girls in, 52, 55n7 Mills, Edward J., 151 parents’ role. See early childhood development Minnesota tax policies, 52 patron-client relationships, 66 “money illusion,” 35 Patt, Anthony, 164 Montessori, 108 payday loans, 6, 8f, 31–32, 33f, 117, 117f Morocco, water connections for low-income households in, “peanuts effect,” 32 87, 89 peer pressure, 10, 133–34, 168–69, 170, 197t Moskowitz, Tobias, 184b perception and mental models, 63, 64f, 69 mosquito nets, free provision of, 150, 150f, 195b, 196 Perry Preschool (U.S.), 108n1 motivation personal distance, 195b monetary incentives for, 16, 128–29 persuasion, 121–22 nonmonetary incentives for, 13t Peru prestige as motivator, 44 health care for welfare recipients in, 89 rewards and prizes as motivators, 128 savings reminders in, 15, 120 Moving to Opportunity program (U.S.), 90 school readiness differences between rich and poor Mullainathan, Sendhil, 81, 186, 193 families in, 99, 100f Multiple Indicator Cluster Survey, 108–9n2 wealthy vs. poor respondents to determine decision- Myrdal, Gunnar, 5b making skills in, 95, 97f, 97n1 World Bank staff predictions about views held by the N poor in, 18, 94, 188b Nepal Philippines, savings rates in farmer-managed irrigation systems as superior to commitment devices for, 15, 38, 120 government-managed systems in, 47–48 reminders for, 15, 120 212 INDEX policy design. See also adaptive design/adaptive interventions; procrastination, ways to overcome, 129–30, 136–39 simplification reciprocity in workplace, 130–32 antipoverty policies and programs, 27, 86–90, 95–97, recruiting high-performance employees, 134–35 104–5 reward programs for employees, 128 for climate change views, 18, 169–71 small businesses’ performance, 135–36 early childhood development interventions, 104–8 social networks’ power and, 139 for household finance decisions’ quality social relations in workplace, 133–34 improvement, 117–23 wage increases triggering, 130–31, 131f human behavior and decision making’s effect on, 29 Promoting Alternative Thinking Strategies (PATHS), 108 mental models and, 63, 70–71 prototype heuristic, 72n4 multi-armed interventions, 20 psychological and social approaches. See also specific for productivity improvement, 139–40 development problem areas social norms and, 52 to adaptive design/adaptive interventions, 194–95, tax policies, 52 197t poverty, 13–14, 80–92 advocating use in development communities, 2, 3 cognitive resources consumed by, 81–84, 83–84f behavioral model of decision making and, 29, 29f cognitive taxes on the poor, 81, 86–89, 115 to climate change information campaigns, 169–70 contexts of decision-making skills in, 81, 94–97, to health behavior changes, 149–51 94–97f to household finance advice, 116–17 design of antipoverty policies and programs, 27, to individual choices, 202–3 86–90 public goods games, 7, 45, 45f, 46–48, 47f, 48f avoiding frames of poverty, 89–90 public-private partnerships in entertainment education, 77 incorporating early childhood development, 104–5 R incorporating social contexts, 90 randomness, 195b simplifying procedures, 86–87 ranking schemes as motivators to change state actions, 44 targeting on basis of bandwidth, 14, 87–89, 88f Reality Check approach, 194, 195b understanding context of poverty, 94–97, recessions, effect on attitudes of young adults in, 85 94–97f reciprocity development professionals’ difficulty in predicting attainment of collective goods and, 46–48 views of poor people, 5, 18, 94, 188b instrumental reciprocity, 46 early childhood development and, 98, 101, 102 intrinsic reciprocity, 46 education and, 80, 85, 87 in workplace, 130–32 financial decision making and, 31, 32, 34f, 112 recycling programs, 53, 197t as fluid status, 82, 82f Redelmeier, Donald A., 181 frames created by, 84–85 red teaming, 19, 184, 190n2 mental models and, 14, 63 Reed, Tristan, 66 scarcity and, 82–83, 82–83f religious violence, 49–50 social contexts generating taxes and, 85–86 reminders. See nudges and reminders willpower and, 115 Renforcement des Pratiques Parentales (Senegal), 106, 106f preschools. See early childhood development research and development (R&D), continual need for, 3, 199 present bias, 37–38 retirement savings plans, automatic enrollment in, 118 climate change and, 165 returns to education, 14, 32, 89 financial decision making and, 112, 114–15 reward programs for employees, 128 stress and, 88 Ribot, J. C., 66 prestige as motivator, 44 risk tolerance for investment opportunities, 113–14, 116–17 Pritchett, Lant, 182 Robinson, James A., 66 private sector’s understanding of customer behavior, 3. Ross, Lee, 63 See also dogfooding rotating savings and credit associations (ROSCAs), 120 development professionals borrowing from, 19 Rustagi, Devesh, 47 procrastination, 114, 118, 129–30, 136–39 Rwanda productivity, 16, 128–42 entertainment education in, 76 agriculture and technology adoption, 136–39 health care quality in, 155 attention deficits, ways to overcome, 136–39 social norms changed through mass media in, 53–54 commitment contracts, use of, 129–30 subsistence farmers, government intervention for, competitive work environments, 132–33, 133f 198 design of policy for, 139–40 ethnography used to enhance, 144–45 S framing tasks and compensation, 130–33 Sabido, Miguel, 77n1 improving employee efforts, 129–32 safe-sex precautions, 44, 133, 133f, 193 incentives, use of, 16, 128–29. See also motivation salience of information, 30, 121 loss vs. gain frames, 132 Salomons, Elad, 190n2 pay-for-performance contracts, 16, 140 Samuelson, Paul, 5b procedures manuals and, 144 Sapienza, Paola, 67 INDEX 213 savings rates, 3–4, 15, 86, 90, 120–21, 121f, 197t intrinsic reciprocity and attainment of collective scarcity, 82–83, 82–83f goods, 46–48 Schkade, David, 183 observation’s effect on behavior, 47, 49f schools. See education social recognition and power of social incentives, Scientific Communication Thesis, 162, 163f 43–44 self-affirmation programs, 12, 85 ultimatum game and, 46 self-employment, 135, 139 war-related violence, effect on children and young self-productivity argument, 100 adults, 45–46, 45f self-regulation, 100, 102, 103 social referents, effect on changing social norms, 54 self-serving bias, 18, 166 social rewards, 43–44 Sen, Amartya, 170 social thinking, 3, 6–11, 9f, 25, 42–58. See also social Senegal’s Renforcement des Pratiques Parentales, 106, 106f preferences service delivery, xi, 89–90, 199 interactions to support new behaviors and build civic Shafir, Eldar, 81, 181, 186 capacity, 50–51 Shampanier, Kristina, 151 social networks’ influence on individual decision Shiller, Robert, 113 making, 42, 49–51 Shu, Suzanne B., 38 social norms’ role, 51–54. See also social norms Sierra Leone targeting specific individuals to lead and amplify governance in, 66 social change, 51 war-related violence, effect on children and young social unrest, 45 adults in, 45–46, 45f South Africa Simon, Herbert, 5b entertainment education in, 76, 122, 122f simplification financial education in, 4 of antipoverty policies and programs, 86–87 savings rates in, 4 choice architecture and, 36–37, 37f sexual behavior and HIV risk in, 52 of financial education, 32, 121 South Asia. See also specific countries of financial products or investments, 117–18 entertainment education in, 77 of health care program, 181 health promotion campaigns in, 17 Skocpol, Theda, 166 Spilker, Gabriele, 167 slavery, 65, 68–69 Steele, Claude M., 73n10 small businesses’ performance, 135–36 stereotypes, 12 SMarT (Save More Tomorrow), 118 Stiglitz, Joseph, 69, 170 Smith, Adam, 5b sunk cost bias, 185–86, 186f Smith, John, 160 Sunstein, Cass R., 36, 183 sociality Switzerland defined, 6, 42 salience of criminal identity in, 67–68, 67f social change and, 6–7, 42–43, 43f, 55 self-employed in, 135 social learning social rewards in, 44 about health care quality, 148–49 effect on earnings, 90 T social networks. See also peer pressure Tanzania defined, 49 communal grass cutting of schoolyard, crowding-out influence on individual decision making, 10, 42, phenomenon in, 49 49–51, 197t conditional cash transfer program for health social norms, 9, 9f, 51–54 insurance enrollment in, 89 activating existing norms to change behavior, 52–53 employee productivity in, 131 changing existing norms to change behavior, 53–54 entertainment education in, 76–77 corruption as, 9, 60–61 health care professionals in, 17, 154 defined, 51 tax payment as social norm, 198b division of labor by, 65 tax policies, 52 early childhood development benefiting from Taylor, Charles, 202 changes in, 106 temptation, 114, 120–21, 123 environmental consumption and, 167–69 Thaler, Richard H., 36 identification of, 195b “thick descriptions” and understanding social and cultural marketing existing norms to change behavior, 52 context for decision making, 145, 194 norm breaking in oppositional cultures, 55n6 thinking. See also framing effect policy designed to work around behavioral effects automatically, 5–6, 6t, 25, 26–40. See also automatic of, 52 thinking tax payment and, 198b deliberative, 6t, 26–27 trust and, 69 with mental models, 3, 11–13, 11f, 25, 62–75. See also social preferences, 3, 9, 9f, 43–49. See also social norms mental models altruism, identity, and group dynamics, 44–46 socially, 3, 6–11, 9f, 25, 42–58. See also social thinking conditional cooperation and, 47, 47–48f, 52 systems of, 5–6, 6t, 26–29, 27t, 164 “crowd in” vs. “crowd out” cooperation, 48–49 Togo, Water Authority of, 144 214 INDEX Tools of the Mind, 108 recessions, effect on attitudes of young adults traffic accidents and fatalities, 13t, 52–53, 53f in, 85 “Trafficking in Persons Report” (U.S.), 44 research on scarcity’s effect on cognitive trust, 2, 65, 68–69, 70–71, 71f processes in, 83–84 Tversky, Amos, 26, 164, 181 self-affirmation techniques used at inner-city soup kitchen, 85 U recycling programs in, 53 Uganda retirement savings plans, automatic enrollment in, 118 agricultural extension activities in, 10 self-employment in, 135 agriculture and technology adoption in, 139 social rewards vs. wage increases in, 44 coffee-producer cooperative members playing “Trafficking in Persons Report,” 44 dictator game in, 45 University of Chicago business students demonstrating corruption in, 61 procrastination, 114 safe-sex programs, reminders to overcome psychological resistance or cognitive biases in, V 39n10 vaccinations, framing information about, 17, 18, 149–50 school scorecards, community monitoring in, 48 Voices of the Poor: Crying Out for Change, 81 ultimatum game, 46 voting procedures, 36–37, 37f, 39n9, 197t United Kingdom Behavioural Insights Team, 90, 189, 198, 198b W employee productivity in, 134 wage increases experimentation and use of psychological and social “money illusion” and, 35 insights in policy and program design in, 198b productivity triggered by, 130–31, 131f tax policies in, 52 social rewards vs., 44 United Nations war-related violence, effect on children and young adults, Gender Empowerment Measure, 44 45–46, 45f Global Compact, 170 Washington Consensus, 170 Human Development Index, 14, 15f, 104, 104f water connections for low-income households, 87, 89 United States water conservation, 10, 176–77, 177f antidrug campaign, ineffectiveness of, 147 water sources, access to clean water, 13t, 19, 193–94 climate change views in, 162–63, 165, 167–68 water treatment for home use, 19–20, 194, 195, 196t energy conservation and, 18, 169 Wertheim, L. Jon, 184b green energy and, 170 White, Andrew, 160 early childhood development in Wikipedia contributors, 44 Abecedarian Project, 108n1 Wildavsky, Aaron, 162 Incredible Years program for Head Start willpower, 115 parents, 107–8 women mindset interventions for, 106 breastfeeding, education about benefits of, 147 Perry Preschool, 108n1 epistemic injustice and, 70 Promoting Alternative Thinking Strategies female genital cutting in Africa, 53, 69 (PATHS), use of, 108 political affirmative action for, 70 school readiness differences between rich and purdah limits on activities of, 54 poor families, 100 social networks of women, role in adoption of new socioeconomic status differences and agricultural technology, 139 language acquisition, 101, 102 social norms and, 51–52, 55n7 education in Woolcock, Michael, 182 bullying in schools, status of program working memory, 100 participants to address, 54 workplace. See productivity college applicants’ test scores offered free in, World Bank. See also development professionals/policy effect of default of, 34, 35f, 39n7, 88–89, 149 makers loss aversion and teachers’ bonuses, 35, 132 civic participation in Sudan program of, 50 school attendance decisions, 81 Doing Business rankings, 44 self-affirmation techniques used with at-risk staff survey responses, 4, 18, 190n1 minority students, 12, 85 compared to wealthy and poor respondents to student performance, teacher bonuses based determine decision-making skills, 95, 95f on, 35, 132 to complexity, 21f, 181–82, 193f employee productivity in, 130, 131, 134 confirmation bias and, 182–83 life insurance in, 116 (mis)perceptions on beliefs of the poor, 180–81, payday loans in, 6, 8f, 117, 117f 188b, 188f poverty in sunk cost bias and, 185–86, 186f Moving to Opportunity program, 90 World Development Report poor vs. affluent thinking processes in, 31, 2004: Making Services Work for Poor People, 36 94–95, 94f 2010: Development and Climate Change, 161 poverty line in, 83 World Values Surveys, 84 INDEX 215 worldviews condom sales by hairdressers and barbers in, 44, 133, automatic thinking, partial worldview from, 7f 133f climate change in, 162–63, 163f prenatal care kits distributed in, 196 Zelizer, Viviana A., 116 Y Zimbabwe Yeager, David, 106 climate change views in, 164, 164f sanitation policies in, 17 Z Zingales, Luigi, 67 Zambia community health workers, recruitment of, 134 ECO-AUDIT Environmental Benefits Statement The World Bank is committed to preserving Saved: endangered forests and natural resources. • 116 trees The Publishing and Knowledge Division • 52 million Btu of total has chosen to print the World Development energy Report 2015: Mind, Society, and Behavior • 9,988 pounds of net on recycled paper with 50 percent postcon- greenhouse gases sumer fiber in accordance with the recom- • 54,169 gallons of mended standards for paper usage set by the wastewater Green Press Initiative, a nonprofit program • 3,627 pounds of solid supporting publishers in using fiber that is waste not sourced from endangered forests. For more information, visit www.greenpress initiative.org.