89033 Enhancing financial capability and behavior in low- and middle-income countries Enhancing financial capability and behavior in low- and middle-income countries Edited by MATTIAS LUNDBERG FLORENTINA MULAJ © 2014 International Bank for Reconstruction and Development / The World Bank 1818 H Street, NW Washington, DC 20433 Telephone: 202-473-1000 Internet: www.worldbank.org The findings, interpretations, and conclusions expressed here do not necessarily reflect the views of the Executive Directors of The World Bank 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. Rights and Permissions The material in this publication is copyrighted. Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law. The International Bank for Reconstruction and Development / The World Bank encourages dissemination of its work and will normally grant permission to reproduce portions of the work promptly. For permission to photocopy or reprint any part of this work, please send a request with complete information to the Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, USA, telephone 978-750-8400, fax 978-750-4470, http://www.copyright.com/. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mail pubrights@worldbank.org. Cover design/layout: Nita Congress Contents FOREWORD................................................................................ xiii ACKNOWLEDGMENTS................................................................xv CONTRIBUTORS........................................................................ xvii ABBREVIATIONS........................................................................ xix 0. OVERVIEW.............................................................................1 Mattias Lundberg and Florentina Mulaj 1 THE IMPACT OF FINANCIAL LITERACY TRAINING FOR MIGRANTS: EVIDENCE FROM AUSTRALIA AND NEW ZEALAND...................................................................33 John Gibson, David McKenzie, and Bilal Zia 2 FINANCIAL EDUCATION AND BEHAVIOR FORMATION: LARGE-SCALE EXPERIMENTAL EVIDENCE FROM BRAZIL...59 Miriam Bruhn, Luciana de Souza Leão, Arianna Legovini, Rogelio Marchetti, and Bilal Zia 3 UNDERSTANDING AND IMPROVING HOUSEHOLD INVESTMENT BEHAVIOR IN BRAZIL: A STOCK MARKET SIMULATOR AS A LEARNING-BY-DOING FINANCIAL LITERACY TOOL................................................................105 Deniz Anginer, Aidan Coville, Vincenzo Di Maro, Martin Kanz, Arianna Legovini, Caio Piza, and Astrid Zwager 4 DOES FINANCIAL EDUCATION AFFECT SAVINGS BEHAVIOR? EXPERIMENTAL EVIDENCE FROM INDIA.......139 Margherita Calderone, Nathan Fiala, Florentina Mulaj, Santadarshan Sadhu, and Leopold Sarr   ◾  vii viii  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 5 COMIC FX IN KENYA: CAN CARTOONS IMPROVE THE EFFECTIVENESS OF FINANCIAL EDUCATION?..................179 Nada Eissa, James Habyarimana, and William Jack 6 SOCIAL NETWORKS, FINANCIAL LITERACY, AND INDEX INSURANCE: EVIDENCE FROM A RANDOMIZED EXPERIMENT IN KENYA....................................................195 Xavier Giné, Dean Karlan, and Muthoni Ngatia 7 WHY IS VOLUNTARY FINANCIAL EDUCATION SO UNPOPULAR? EXPERIMENTAL EVIDENCE FROM MEXICO............................................................................209 Miriam Bruhn, Gabriel Lara Ibarra, and David McKenzie 8 FINANCIAL (DIS-)INFORMATION: EVIDENCE FROM AN AUDIT STUDY IN MEXICO.................................................247 Xavier Giné, Cristina Martínez Cuellar, and Rafael Keenan Mazer 9 LEARNING BY DOING? USING SAVINGS LOTTERIES TO PROMOTE FINANCIAL INCLUSION IN NIGERIA.................279 Martin Kanz, Alex Kaufman, Kelly Shue, and Benedikt Wahler 10 NIGERIA’S NOLLYWOOD NUDGE: AN ENTERTAINING APPROACH TO SAVING.....................................................315 Aidan Coville, Vincenzo Di Maro, Siegfried Zottel, and Felipe Alexander Dunsch 11 HARNESSING EMOTIONAL CONNECTIONS TO IMPROVE FINANCIAL DECISIONS: USING SAVINGS LOTTERIES TO PROMOTE FINANCIAL INCLUSION IN SOUTH AFRICA...... 361 Gunhild Berg and Bilal Zia 12 THE IMPACT OF FINANCIAL EDUCATION ON FINANCIAL SERVICE USE: EVIDENCE FROM A FINANCIAL DIARIES STUDY IN UGANDA...........................397 Guy Stuart 13 INCREASING THE IMPACT OF CONDITIONAL CASH TRANSFERS THROUGH FINANCIAL LITERACY: EVALUATION PILOT IN THE DOMINICAN REPUBLIC..........423 Xavier Giné, Dean Karlan, and Greg Fischer   ◾  ix 14 DIRECT DEPOSIT AND COMMITMENTS TO SAVE IN MALAWI.......................................................................437 Xavier Giné 15 EVALUTION OF OLD MUTUAL’S ON THE MONEY PROGRAM: FINANCIAL LITERACY IN SOUTH AFRICA.......451 Shawn Cole, Bilal Zia, Martin Abel, Lucas Crowley, Christian Salas Pauliac, and Veronica Postal 16 FINANCIAL MANAGEMENT AND VOCATIONAL TRAINING IN UGANDA: IMPACT AND NETWORK EFFECTS IN INFORMAL INDUSTRIAL CLUSTERS...............495 Francisco Campos, Markus Goldstein, Obert Pimhidzai, Mattea Stein, and Bilal Zia BOXES 0.1 Evaluations of interventions featuring traditional financial education....12 0.2 Evaluations of interventions featuring nontraditional financial education...................................................................................................16 0.3 Evaluations of interventions featuring combined methods.....................19 12.1 A case study in group savings discipline................................................ 412 12.2 A case study in increased savings.......................................................... 415 FIGURES 1.1 Examples of materials used in the financial literacy training..................44 2.1 Impacts of financial education program on students..............................82 2.2 Distribution shift in financial proficiency scores......................................83 3.1 Rate of participation by wealth and asset class..................................... 110 3.2 First-order stochastic dominance for the number of trades in the simulator....................................................................................... 119 3.3 First-order stochastic dominance for the length of use of the simulator (in days)................................................................................... 119 3.4 First-order stochastic dominance for the number of days traded in the simulator....................................................................................... 119 3.5 Distribution of the Sharpe ratio measure of simulator users................122 3.6 Distribution of adjusted returns of simulator users...............................122 4A.1 Kernel densities....................................................................................... 174 5.1 Location of schools with JAK clubs.........................................................183 6.1 Study chronology in 2011........................................................................198 6.2 Randomization.........................................................................................200 7.1 Financial literacy training take-up rates by incentive group..................226 7.2 Median savings account balance............................................................240 7.3 Median credit card balance....................................................................241 x  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 7.4 Average percentage of credit card debt paid off each month..............241 8.1 Sample marketing materials for consumer credit collected during auditor visits.................................................................................267 9.1 Social media campaign and celebrity endorsements............................294 9.2 Sequence of prize drawings by state.....................................................297 9.3 Branch network of treatment and control banks, January 2011.......... 300 9.4 Geographical distribution of bank customers by state..........................301 9.5 Timeline of the ISIW savings promotion.................................................304 9.6 Mean account balance, by month..........................................................306 9.7 Fraction of accounts qualifying for lottery, by month............................308 10.1 Sample invitation.....................................................................................327 11.1 Study timeline.......................................................................................... 374 11.2 NDMA daily call frequency......................................................................382 12.1 Home deposit amounts as a share of income by week and group, when financial inflows are zero.............................................................. 415 12.2 Formal and semiformal deposit amounts as a share of income by week and group, when financial inflows were zero.......................... 417 13.1 Training delivery......................................................................................428 13.2 Project timeline.......................................................................................436 15A.1 Study timeline..........................................................................................469 15A.2 Study map................................................................................................470 TABLES 0.1 Priority areas guiding the selection of Trust Fund field evaluations........10 1.1 Characteristics of sample by treatment status .......................................41 1.2 Attrition rates by survey round.................................................................45 1.3 Impact on financial knowledge.................................................................47 1.4 Impact on financial knowledge.................................................................49 1.5 Impact on remittance outcomes..............................................................50 1.6 Impact on the likelihood of switching remittance methods....................51 1.7 Impact on Pacific Island migrants having different financial products six months after treatment.......................................................52 2.1 Number of schools in sample by state.....................................................72 2.2 Survey participation..................................................................................75 2.3 Preprogram summary statistics...............................................................77 2.4 Reported program take-up and usage in treatment group schools........78 2A.1 Impact on student financial proficiency...................................................91 2A.2 Impact on student saving and spending behavior...................................91 2A.3 Impact on student attitudes and future behavior....................................92 2A.4 Impact on student participation in household finance............................92 2A.5 Trickle-up impact on parent financial knowledge....................................93 2A.6 Trickle-up impact on parent saving and spending behavior....................93 2A.7 Impact of parent financial education workshop......................................94 3.1 Descriptive statistics for simulator users............................................... 118   ◾  xi 3.2 Descriptive statistics for participation indicators and return variables..................................................................................................121 3.3 Effect of the usage of the simulator on disposition effect....................124 3.4 Effect of the usage of the simulator on diversification effect...............125 3.5 Effect of the usage of the simulator and behavioral biases on adjusted returns......................................................................................126 3.6 Effect of the usage of the simulator and behavioral biases on adjusted returns (include previous experience as controls)..................127 3.7 Effect of the usage of the simulator and behavioral biases on the Sharpe ratio..................................................................................128 3.8 Effect of the usage of the simulator and behavioral biases on adjusted returns (include previous experience as controls)..................129 3.9 Participation equation: is the simulator an effective nudging tool?— unrestricted sample................................................................................130 3.10 Participation equation: is the simulator an effective nudging tool?— individuals with more than five trades...................................................131 3A.1 Barriers to entry and investor trading biases.........................................135 4.1 The content of the financial education training.....................................143 4.2 Descriptive statistics on household savings..........................................146 4.3 Descriptive statistics on budgeting quality and interest in financial matters.....................................................................................148 4.4 Descriptive statistics and comparability of our measures of financial knowledge................................................................................149 4.5 Sample characteristics and balance test............................................... 151 4.6 Pre- and post-intervention differences..................................................153 4.7 Average impacts on savings...................................................................155 4.8 Average impacts of treatment and postharvest reminders on savings.....................................................................................................156 4.9 Average impacts on financial literacy.....................................................157 4.10 Average impacts on financial literacy for the subsample of clients who answered both the baseline and the endline survey.........158 4.11 Average impacts on financial literacy using the standardized indicators.................................................................................................159 4.12 Heterogeneity of impacts on savings for client education and baseline financial literacy........................................................................ 161 4.13 Heterogeneity of impacts on financial literacy for client education and baseline financial literacy................................................................. 161 4.14 Heterogeneity of impacts on savings for client gender and time preferences.............................................................................................162 4.15 Heterogeneity of impacts on financial literacy for client gender and time preferences..............................................................................162 4.16 Heterogeneity of impacts on savings for baseline per capita total expenditures and whether client had non-FINO formal savings account.......................................................................................164 4.17 Heterogeneity of impacts on financial literacy for baseline per capita total expenditures and whether client had non-FINO formal savings account.......................................................................................164 xii  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 4.18 Average impacts on consumption..........................................................165 4.19 Average impacts on loans and assets....................................................165 4.20 Average impacts on saving controlling for quality of FINO services.....167 4.21 Heterogeneity of impacts on FINO saving for FINO agents’ presence..................................................................................................167 4A.1 Questions on financial knowledge.......................................................... 171 4A.2 Attendance..............................................................................................172 4A.3 Nonresponse rates for the outcome measures showed in table 4.6....173 4A.4 Attrition....................................................................................................173 4B.1 Heterogeneity of impacts on financial literacy for client education and baseline financial literacy.................................................................175 4B.2 Heterogeneity of impacts on financial literacy for client gender and time preferences.............................................................................. 176 4B.3 Heterogeneity of impacts on financial literacy for baseline per capita total expenditures and whether client had a non-FINO formal savings account...........................................................................177 5.1 School distribution by region..................................................................184 5.2 School characteristics.............................................................................184 5.3 Experimental design...............................................................................185 5.4 Geographic distribution of schools.........................................................189 5.5 JAK and non-JAK competition winners and winnings............................190 5.6 Impact of the Interventions on financial literacy...................................191 5.7 Impact of the interventions on stated savings for all students and actual savings for competition winners..........................................193 5.8 Impact of the interventions on changing student goals........................193 6.1 Summary statistics by intervention........................................................199 6.2 Impact of comics and vouchers on whether respondents purchased insurance...............................................................................201 6.3 Ordinary least squares regressions of whether respondents purchased insurance...............................................................................203 6.4 Correlates with attrition in follow-up data.............................................204 6.5 Knowledge score on index insurance.....................................................205 6.6 Outcomes from follow-up.......................................................................206 7.1 Confirming randomization using baseline data......................................223 7.2 Impact of incentive treatments on take-up in treatment group............227 7.3 Determinants of program take-up in treatment group..........................229 7.4 Impact on financial knowledge...............................................................232 7.5 Impact on savings behavior and outcomes............................................234 7.6 Impact on retirement savings behavior..................................................235 7.7 Impact on credit card behavior and outcomes......................................237 7.8 Impact on loan behavior and outcomes.................................................238 7.9 Heterogeneous treatment effects (intention to treat)...........................239 8.1 Cost and return of financial products.....................................................254 8.2A Interview time, number of products offered, and product alignment with needs for savings products............................................................264   ◾  xiii 8.2B Interview time, number of visits, and approval rate for credit products..................................................................................................266 8.3A Printed and oral information for savings products................................268 8.3B Printed and oral information for credit products...................................269 8.4A Quality of information and return on savings.........................................272 8.4B Quality of information and cost of credit...............................................273 8.5 Credit auditor’s assessment of staff......................................................275 9.1 Prize drawings.........................................................................................297 9.2 Geographical distribution of customers, by bank and state..................299 9.3 Basic demographic statistics for ICB and Access Bank customers (%)..........................................................................................302 9.4 Pre-post and differences-in-differences estimation, net balance changes...................................................................................................307 9.5 Differences-in-differences estimation, product usage after ISIW launch..............................................................................................309 9.6 Treatment effects, lottery-induced variation.........................................310 10.1 Baseline balance.....................................................................................330 10.2 Selection into screenings........................................................................331 10.3 Balance across screening participants...................................................333 10.4 Attrition in endline survey.......................................................................334 10.5 Item nonresponse across screening participants (%)............................335 10.6 Compliance table (%)...............................................................................338 10.7 Self-reported exposure to interventions................................................339 10.8 Financial literacy indexes........................................................................340 10.9 Perceptions of microfinance banks........................................................342 10.10 Perception of female financial performance at endline.........................343 10.11 Intentions.................................................................................................344 10.12 Savings behavior.....................................................................................346 10.13 Savings account sign-up rates................................................................347 10.14 Borrowing behavior.................................................................................348 10A.1 Descriptive statistics (female) ................................................................356 10A.2 Descriptive statistics (male) ...................................................................357 10A.3 Balance across screening participants (female) ....................................358 10A.4 Balance across screening participants (male) .......................................359 11.1 Summary statistics and tests of randomization....................................373 11.2 Sample attrition.......................................................................................375 11.3 Sample crossovers (%)............................................................................ 376 11.4 Financial knowledge................................................................................377 11.5 Debt management...................................................................................378 11.6 Reasons for borrowing............................................................................379 11.7 Hire purchase and gambling...................................................................380 11.8 Savings and well-being............................................................................381 11.9 Seeking financial advice..........................................................................383 12.1 Financial education training syllabus..................................................... 400 xiv  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 12.2 Financial capability concepts and indicators.........................................401 12.3 Number of financial transactions per week...........................................407 12.4 Mean amount of financial transactions (U Sh).......................................407 12.5 Quotes about the benefits of different savings locations...................... 410 12.6 Number of savings transactions per week............................................. 411 12.7 Mean amount of savings transactions................................................... 411 12.8 Model of home savings deposit behavior ............................................. 414 12.9 Model of semiformal and formal deposit behavior................................ 416 12.10 Cash gifts received and given: transactions per week.......................... 418 12.11 Cash gifts received and given: amount per transaction........................ 418 15.1 Impact on financial awareness...............................................................463 15.2 Impact on financial attitudes and perceptions......................................464 15.3 Impact on savings behavior....................................................................465 15.4 Impact on borrowing behavior...............................................................465 15.5 Impact on gambling bahavior.................................................................466 15.6 Impact on financial planning...................................................................466 15.7 Impact on well-being...............................................................................467 15.8 Impact on numeracy skills......................................................................467 15A.1 Baseline survey records..........................................................................471 15A.2 Endline survey records............................................................................473 15A.3 Balance checks........................................................................................ 476 15A.4 Heterogeneous effects by education level: financial awareness..........477 15A.5 Heterogeneous effects by education level: financial attitudes and perceptions..............................................................................................477 15A.6 Heterogeneous effects by education level: savings behavior...............478 15A.7 Heterogeneous effects by education level: borrowing behavior...........478 15A.8 Heterogeneous effects by education level: gambling behavior............478 15A.9 Heterogeneous effects by education level: financial planning..............479 15A.10 Heterogeneous effects by education level: well-being..........................479 15A.11 Heterogeneous effects by education level: numeracy skills.................479 15A.12 Heterogeneous effects by education level: financial awareness..........480 15A.13 Heterogeneous effects by survey group: financial awareness..............481 15A.14 Heterogeneous effects by survey group: numeracy skills.....................482 Foreword I n the past 10 years, the global community has come to understand that all citizens must be equipped with the ability to negotiate an increasingly complex world of personal and business finance. The growth of markets for insurance and risk, the expansion of banking, and the increasing need to make one’s own employment all heighten the importance of financial literacy and capability. This is true not only for people in middle- and high-in- come countries, but also for those in poor countries. As Daryl Collins and her coauthors show in their groundbreaking study, Portfolios of the Poor: How the World’s Poor Live on $2 a Day (Princeton University Press, 2009), poor families lead astonishingly complex financial lives. They manage to keep food on the table, children in school, and plan for the future—all on small and unstable incomes. They maintain elaborate financial portfolios, although with uncertain success. They save and borrow small amounts often at high cost and low return, which inhibits their ability to grow out of poverty. The complexity, high cost, and modest returns that characterize the financial arrangements of poor families are a serious cause for public concern, since unproductive assets lead to lower growth and to poverty. These concerns have only increased since the recent global financial crisis, which reinforced the recognition that financial capability is an essential individual life skill. It is generally true that individuals with better financial knowledge make better financial decisions. This simple correlation has prompted the develop- ment of a wide range of financial education initiatives by governments, regu- lators, and various private and civil stakeholders, sometimes combined with financial consumer protection measures. The results of these early efforts were not encouraging. Prior to the work supported under the Russia Trust Fund for Financial Literacy and Education, research had found little evidence that improving financial capability yields better financial behavior. Moreover, the vast majority of studies had come from developed countries in North America and Western Europe. Would these interventions work differently among different populations in different environments? Is the relative lack of success of financial capacity-building interventions due to poor design or poor implementation, or is it possible that financial education of any kind will actually do very little for financial capability? Happily, this previous research, together with the literature on behav- ioral economics and finance, offered some insights into why education alone   ◾  xv xvi  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES may have limited effects on improving financial behavior, and demonstrated that improving behavioral outcomes would require a major revision in the design of policies and programs for financial capability. Partly this is already happening: pension plans are changing default options to take advantage of inertia and the status quo bias, and reducing the number of options to prevent the cognitive overload that creates indecision. Previous experience also pointed the way to new avenues for research—innovative interventions and policies that might prove successful where previous attempts had failed, for example, employing tools such as social marketing to influence behavior. The Russia Trust Fund for Financial Literacy and Education was set up to address some of these concerns, and to support the advancement of financial literacy and capability in low- and middle‑income countries. Established in 2008 with funding provided by the Ministry of Finance of the Russian Federation, the Trust Fund enabled the World Bank and the Organisation for Economic Co-oper- ation and Development (OECD) to conduct methodological, analytical, and policy work on financial literacy, capability, and education. The program has supported work on the definition and measurement of financial capability, methods, and tools to understand the impact of financial capability programs; and field research on interventions designed to enhance financial capability. These and other resources may be found at http://www.finlitedu.org/. This volume presents the results of pilot projects financed by the Trust Fund. These include some truly pioneering research into new areas for financial capacity building, including the use of diaries to increase financial awareness, innovative methods for information provision, and new ways to deliver messages that encourage safer financial behavior, such as feature films, TV soap operas, and comic books. These methods open financial literacy to a much broader audience, and with poten- tially greater impact, than has been achieved through more conventional means. This work is the culmination of the efforts of a very large number of people, not only from within the World Bank and OECD, but also dozens of researchers and academics, implementing partners, financial service providers, and policy makers spanning the globe from Latin America to Africa to Asia. The dedication of these diverse stakeholders to sharing information, testing new ideas, and learning from experience is already helping us realize our goal of greater financial security for all. Above all, this work would not have been possible without the vision and guid- ance of Robert Holzmann and Richard Hinz. Robert was the originator and architect of the Russia Trust Fund program, and Richard managed the program following Robert’s retirement from the Bank. Their strong leadership made it possible to undertake the innovative research presented in this volume and in the many other books and research reports produced under the aegis of the Russia Trust Fund. Arup Banerji Director, Social Protection and Labor Global Practice The World Bank Group Acknowledgments T his is the final volume to come from a multiyear, multi-institution project, and it owes an enormous debt of gratitude to the more than 100 people who generously gave their time, effort, and creativity to making this such a success. Above all, we want to thank the staff of the Ministry of Finance of the Russian Federation who brought the theme of financial literacy to the agenda of the Russia G8 Presidency, and who charged the World Bank– Organisation for Economic Co-operation and Development (OECD) team with producing this body of knowledge. Special thanks are due to the Vice Minister of Finance, Sergey Storchak, who initiated the enterprise; and to Andrey Bokarev, Director of the Department of International Relations within the Ministry of Finance, who provided guidance throughout the process. Others in the ministry who deserve recognition include Anna Valkova, Elena Ilina, and Anna Zelentsova. We must also acknowledge the critical role played by our World Bank colleague Andrei Markov as interlocutor and facilitator throughout the project. The effort was launched with the intellectual contributions of partici- pants at a November 2009 workshop in Washington, D.C. We would like to thank the participants and contributors at this and subsequent conferences in Cape Town, Montevideo, Paris, Washington, Saint Petersburg, Nairobi, and New Delhi, who challenged and validated the methods and results presented by the extended Trust Fund team. Some of the participants at these confer- ences contributed in other ways as well, providing guidance and helping shape the research. These include Gerrit Antonides, Sharon Collard, Billy Jack, Olga Kuzina, Annamaria Lusardi, Lew Mandell, Christian Poppe, Robert Walker, Alina Wyatt, and especially Elaine Kempson, our lead external advisor. From the International Network on Financial Education, we would like to acknowledge Diana Crossan, Jason Fichtner, Sue Lewis, and José Alexandre Cavalcanti Vasco. The staff of the RAND Corporation, which developed the Toolkit of Evaluation Methods—in particular, Angela Hung, Arie Kapteyn, and Joanne Yoong—made a significant contribution to the overall effort. Last but not least, we would like to thank our OECD colleagues and co-managers of the Trust Fund program, particularly Adele Atkinson, Andre Laboul, and Flore- Anne Messy, for the collaboration and intellectual stimulation they provided throughout the process.   ◾  xvii xviii  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Most importantly, this work could not have been completed without the dedicated contributions of the teams represented by the papers in this volume. This was an outstanding example of a collective effort by the impact evaluation teams comprising regional World Bank staff, the staff of the Bank’s Development Economics Research Department, and the Social Protection and Financial Devel- opment Departments. These include Gunhild Berg, Miriam Bruhn, Aline Coudouel, Aidan Coville, Xavier Giné, Mary Hallward-Driemeier, Martin Kanz, Leora Klapper, Arianna Legovini, Vincenzo Di Maro, Rogelio Marchetti, David McKenzie, Margaret Miller, Douglas Pearce, Leopold Sarr, Peer Stein, and Bilal Zia. A small part of their contributions can be found in this volume and in the other publications available on the program’s web site (http://www.finlitedu.org/); but our gratitude extends far beyond the tangible outputs listed there. Finally, we would like to acknowledge the heroic efforts of the team respon- sible for the administration of the Trust Fund in Washington. Above all, this includes Robert Holzmann and Richard Hinz, who created and managed the Trust Fund for most of its life. In addition, we are enormously grateful to Arup Banerji, Anush Bezhanyan, Nita Congress, Raiden Dillard, Inas Ellaham, Anya Maria Mayans, Amira Nikolas, Valeria Perotti, Laura Rawlings, Kinnon Scott, and Sophie Warlop for their input, support, patience, and guidance throughout the project. Without them, this report would not exist. Contributors Martin Abel, Kennedy School of Government, Harvard University Deniz Anginer, Pamplin College of Business, Virginia Tech Gunhild Berg, World Bank Miriam Bruhn, World Bank Margherita Calderone, German Institute for Development Research Francisco Campos, World Bank Shawn Cole, Harvard Business School Aidan Coville, World Bank Lucas Crowley, Innovations for Poverty Action (IPA) Felipe Alexander Dunsch, World Bank Nada Eissa, Georgetown University Nathan Fiala, German Institute for Development Research Greg Fischer, London School of Economics John Gibson, Waikato Management School, University of Waikato Xavier Giné, World Bank Markus Goldstein, World Bank James Habyarimana, Georgetown University William Jack, Georgetown University Martin Kanz, World Bank Dean Karlan, Yale University Alex Kaufman, Harvard University Rafael Keenan Mazer, World Bank   ◾  xix xx  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Gabriel Lara Ibarra, World Bank Luciana de Souza Leão, Columbia University Arianna Legovini, World Bank Mattias Lundberg, World Bank Rogelio Marchetti, World Bank Vincenzo Di Maro, World Bank Cristina Martínez Cuellar, Harris School of Public Policy, University of Chicago David McKenzie, World Bank Florentina Mulaj, World Bank Muthoni Ngatia, Abdul Latif Jameel Poverty Action Lab, Massachusetts Institute of Technology Obert Pimhidzai, World Bank Caio Piza, University of Sussex Veronica Postal, International Monetary Fund Santadarshan Sadhu, Centre for Microfinance, Institute for Financial Management Research Christian Salas Pauliac, World Bank Leopold Sarr, World Bank Kelly Shue, Booth School of Business, University of Chicago Mattea Stein, Paris School of Economics Guy Stuart, Microfinance Opportunities Benedikt Wahler, Roland Berger Strategy Consultants Bilal Zia, World Bank Siegfried Zottel, World Bank Astrid Zwager, World Bank Abbreviations ANBIMA Brazilian Association of Financial and Capital Markets (Associação Brasileira das Entidades dos Mercados Finan- ceiro e de Capitais) APR annual percentage rate ATM automated teller machine BM&FBOVESPA Brazilian Securities, Commodities, and Futures Exchange BSSP burial society support plan (South Africa) CAEd Center for Public Policy and Education Evaluation (Centro de Políticas Públicas e Avaliação da Educação) (Brazil) CVM Securities and Exchange Commission (Comissão de Valores Mobiliários) (Brazil) CCT conditional cash transfer CONDUSEF National Commission for the Protection of Users of Finan- cial Services (Comisión Nacional para la Protección y Defensa de los Usuarios de Servicios Financieros) (Mexico) FEBRABAN Federation of Brazilian Banks (Federação Brasileira de Bancos) FSA Financial Services Authority (United Kingdom) GAP Pedagogical Support Group (Grupo de Apojo Pedagógico) (Brazil) GDP gross domestic product GPS global positioning system HFHU Habitat for Humanity Uganda ICB InterContinental Bank (Nigeria) INR item nonresponse ISIW I-Save I-Win (Nigeria) ITT intention to treat IV instrumental variable JAK Junior Achievement Kenya KASSIDA Katwe Small Scale Industries Association (Uganda) KSAs knowledge, skills, and attitudes M&E monitoring and evaluation MaMA Million-a-Month Account (South Africa)   ◾  xxi xxii  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES MFB microfinance bank NDMA National Debt Mediation Association (South Africa) NIDS National Income Dynamics Study (South Africa) OECD Organization for Economic Co-operation and Development OLS ordinary least squares ROSCA rotating savings and credit association SACCO savings and credit cooperative organization (Uganda) SCAP cooperative society (sociedad cooperativa de ahorro y prés- tamo) (Kenya) SOFIPO popular financial society (sociedade financiera popular) (Mexico) SOFOME multipurpose society (sociedade financiera de objeto múltiple) (Mexico) ToT treatment effect on the treated WDB Women’s Development Businesses (South Africa) Overview MATTIAS LUNDBERG AND FLORENTINA MULAJ 0.1 BACKGROUND AND MOTIVATION Among the main lessons of the 2008 global financial crisis were the low level of financial literacy of populations around the world, and the links between financial capability and financial well-being. Klapper, Lusardi, and Panos (2013) found that individuals with greater financial capability were significantly less likely to report a negative income shock during 2009. Since then, governments and policy makers in both developed and developing countries increased the attention and resources dedicated to financial literacy programs as a way to empower consumers in their interactions with financial markets. Many coun- tries even developed national strategies for financial literacy and education, putting the topic high on the respective government’s agenda (Grifoni and Messy 2012). However, despite growing interest and investment in the topic, the empirical evidence and understanding of what drives consumer financial behavior has been limited. Policy makers have had no clear guidance on what types of programs needed to be scaled-up to achieve results. In other words, while there was general understanding that many people lacked knowledge and skills about personal finance, there was no conclusive evidence on the link between financial literacy interventions and improved consumer deci- sion making. The field had an initial narrow focus on financial literacy defined mainly in terms of knowledge and numeracy, as opposed to a broader definition that includes the manifestation of this knowledge in skills, attitudes, and behavior. The overly narrow definition led policy makers and researchers to ignore more innovative interventions and to focus on school-based programs. Under researched were interventions that utilize social marketing, technology, and media; and interventions that employ behavioral treatments and multi-armed   ◾ 1 2  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES approaches. Moreover, there was rather little investment in monitoring and eval- uation (M&E) to inform a rigorous understanding of the real impact of these programs in the field. As of 2010, few financial education interventions had undergone rigorous impact evaluation. A realistic figure would put the share of rigorously monitored and evaluated interventions in the developed world below 1 percent—and in the developing world, close to zero. Furthermore, the few evaluations that did exist suffered a range of shortcom- ings, including the following: ◾◾ A lack of understanding among program providers of the importance of rigorous M&E for the design, implementation, and improvement of inter- ventions ◾◾ A lack of standard and relevant methods for M&E that go beyond generic instructions on research methodology and address the concepts and issues within a financial capability framework ◾◾ The perceived high budgetary costs of rigorous M&E for new interventions ◾◾ The public good nature of rigorous M&E, which produces knowledge that can be consumed by others, but whose cost is borne by the provider of the individual intervention ◾◾ Resistance from program providers to an evaluation, as there is always the risk that it may reveal unfavorable results and/or limited impact, conse- quently threatening funding for program activities Against this background, the Russia Financial Literacy and Education Trust Fund was established in the World Bank in October 2008. Its main objectives included the following: ◾◾ The financing of field impact evaluations of competitively selected inter- ventions across regions and income levels that explore causality between financial literacy interventions and consumer financial behavior ◾◾ The development of an M&E toolkit geared toward financial capability interventions informed by Trust Fund–supported impact evaluations ◾◾ The financing of evaluations for a set of interventions that conducted comparative testing of effectiveness; i.e., comparing the effectiveness of a specific intervention across countries or alternative interventions under comparable settings in one country ◾◾ The local capacity building in developing countries for the design and implementation of M&E by linking World Bank researchers to local govern- ment officials, policy makers, and other partners and program providers This chapter focuses on the Trust Fund–supported impact evaluations, including the programs evaluated, research methods, and the broader lessons learned (for the M&E toolkit and capacity-building activities, refer to the Trust Overview  ◾  3 Fund website at http://www.finlitedu.org/). The chapter concludes with a discus- sion of policy implications and suggests direction for future research. 0.2 CONCEPTUAL DEVELOPMENT Earlier policy responses to help individuals make better financial decisions predominantly focused on financial literacy, defined as knowledge and numeracy (Lusardi and Mitchell 2006; Mandell 1997), which were driven by and led to a focus on education programs to improve literacy levels. This was primarily motivated by survey evidence documenting the correlation between financial literacy and household well-being. For example, lower levels of financial literacy were found to be negatively related with engagement in saving, credit, and investment (Hilgert, Hogarth, and Beverly 2003), with planning for retirement (Lusardi and Mitchell 2007), borrowing at high interest rates (Stango and Zinman 2009), and with the use of informal sources of borrowing (Klapper, Lusardi, and Panos 2012). However, while some financial education programs may have been effective in improving literacy levels, evidence of their impact on actual behavior has been inconclusive, if not exclusively bleak (Braunstein and Welch 2002; Cole and Shastry 2008; Gale, Harris, and Levine 2012; Mandell and Klein 2009). To investigate this phenomenon and determine whether the relationship between financial educa- tion, literacy, and desired outcomes is in fact causal, a number of scientific experi- ments began in recent years, and even before the Russia Trust Fund program. The results of these studies were mixed (Bertrand and Morse 2010; Cole, Sampson, and Zia 2011). For example, Duflo and Saez (2003) conducted a randomized study to measure the impact of a benefit fair on retirement plan enrollment, and found small effects on take-up. Focusing on developing countries, Cole, Sampson, and Zia (2009) conducted a randomized study in Indonesia to measure the impact of financial education on savings; they also found no substantial impact. Similar find- ings reporting on the limitations of financial education on behavioral outcomes were found by Karlan and Valdivia (2011) in Peru and by McKenzie and Weber (2009) in Uganda. Recent developments in the field of behavioral economics have led to substantial changes in our understanding of financial and economic decision making. Drawing on lessons from psychology, this field has produced a number of experiments that help understand some of the psychological barriers that prevent people from moving from knowledge and intentions to actions. This literature generally maintains that, due to a number of cognitive, emotional, and social factors, individuals often fail to make optimal decisions even in the pres- ence of information. To list a few examples: there is evidence that people are loss averse—that is, they attribute greater value to losses than gains (Kahneman and Tversky 1979); have status quo preference—i.e., in the presence of many options, they tend to avoid changing their course of action (Samuelson and Zeckhauser 4  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 1988); discount the long term in comparison to the present (Loewenstein and Prelec 1992); and are highly influenced by emotions and peers (Andrade and Ariely 2009; Kahneman, Slovic, and Tversky 1982). Additional developments in decision making have come from a greater under- standing of financial behavior in the context of poverty. Experimental work in this area has found that many of these behavioral biases are even more pronounced among the poor (Banerjee 2000; Duflo 2006; Mullainathan and Thaler 2000). One key observation from this work is that poverty creates myopia, which refers to a focus on solving short-term problems without the ability or mental space to make decisions that take into account long-term implications. As a consequence, many decisions reached under poverty are detrimental for the future (e.g., high interest rate loans). In retrospect, what this literature suggests is that there is more to changing behavior than the traditional microeconomic model would predict. Simply providing individuals with financial education and access to instruments is insufficient, and additional or alternative behavioral treatment mechanisms might be necessary to improve outcomes. These results from field experiments and behavioral research have led to a shift away from financial literacy toward a wider concept that includes “A combination of awareness, knowledge, skill, attitude and behavior necessary to make sound financial decisions and ultimately achieve individual financial wellbeing” (Atkinson and Messy 2012). The expansion of the definition in broad terms represents a shift in conceptualization and measurement from an explicit knowledge-based to an outcome-driven approach, focusing on the mechanisms that drive behavior. This approach guided the selection of the Trust Fund–supported impact evaluations. Yet even with this broader definition, problems in identifying effective interven- tions persist. In fact, a key challenge in adopting a more holistic definition is that the interventions become more diverse in their design and may follow different theories of change. As the range of interventions expands, careful thinking is required to develop a standard results framework and methods for evaluation that allow for accurate comparison of findings across projects and countries. 0.3 IMPLICATIONS FOR PROGRAM DESIGN AND EVALUATION Financial literacy programs are designed to achieve outcomes through formal education, and assume a causal chain reaction from knowledge to skills to behavior. While these interventions may also include other delivery mechanisms to transfer education, the basic premise is that improving knowledge leads to desired outcomes. Financial capability programs, on the other hand, do not assume a particular relationship that derives in a linear fashion from education to behavior, and hence may either follow a noncognitive approach or combine financial education with other mechanisms. Overview  ◾  5 In addition to programs that employ standard classroom and workshop models, interventions more widely encountered in recent studies include edutainment, social marketing, personal counseling, consumer protection, and behavioral treatments. Edutainment (educational entertainment) is a form of entertainment designed to educate the audience through stories rather than explicit didactic messages. It usually involves delivery of educational content through amusing and emotional tales in TV films, soap operas, theater, computer-based games, and others. Unique features of this intervention are that it can reach a wide audience, appeal to emotions, and disseminate messages in a way that sticks to memory. The field of public health has extensively and successfully used edutainment to influence behavioral change in a number of areas, including nutrition, tobacco and alcohol use, safe sex, physical exercise, and immunization. The application of this nethod to consumer finance is still in the early stages of experimentation. To the authors’ knowledge, prior to the Trust Fund efforts, the only edutain- ment program for personal finance was Makutano Junction in Kenya, a TV soap opera funded by the U.K. Department for International Development (DFID).1 The program, however, has never been evaluated. More recent examples that involve evaluation include Trust Fund–financed experiments in South Africa, Nigeria, and Kenya, where content-specific messages are integrated in a soap opera, a feature film, and comics, respectively (discussed in section 0.6). Social marketing refers to the systematic application of standard commer- cial marketing to alter preferences and influence a behavioral change. Though similar to edutainment, social marketing is less focused on education and more on persuasion through short messages. These include anything from print handouts and billboards, to public service announcements linked to TV shows and films, to celebrity endorsements, street theater, and formal in-school and in-workplace presentations. Among many examples are initiatives involving health (anti-to- bacco, safe sex), road safety (buckle up, wear bike helmets), the environment (recycling, energy saving), and politics (voter participation in elections). In promoting responsible financial management, this mechanism is also at its early stages of application. In a paper prepared for the Consultative Group to Assist the Poor (CGAP), Lee and Miller (2012) reviewed 100 case studies in which social marketing has influenced financial behavior; however, none of these programs have been rigorously evaluated. To the authors’ knowledge, the only intervention to date that employs social marketing coupled with a rigorous eval- uation in the field of personal finance is the Trust Fund–financed experiment in Kenya, which involves a national marketing campaign to promote savings (discussed in section 0.6). 1  For more information, see http://www.makutanojunction.org.uk/ (accessed May 4, 2014). 6  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Personal counseling involves individualized and one-on-one training or assistance at a particular point in time. This method is best known from credit counseling designed to either avoid debt or to help individuals develop debt management and repayment plans. It differs from traditional financial education aimed at teaching more general concepts. Personal counseling is customized and targeted to a specific individual and topic, and is usually applied at the time when relevant. It may include interventions where an individual is directed to budget properly and develop expense plans, save for a particular goal, or establish and improve credit history, or to guide product selection. It is mostly useful in light of major life events, such as loss of job, unexpected reduction in income, domestic violence, etc. For example, following the results of a baseline survey of financial capability, the United Kingdom’s Financial Services Authority (FSA) developed a pathfinder service offering individual information and guidance on a wide range of financial matters (such as budgeting, retirement planning, and choosing appropriate finan- cial products). This was delivered face to face, by telephone, and through an inter- active website, and was found to be effective (Kempson and Collard 2010). A more recent example includes a Trust Fund–financed experiment in India that involves a combination of standard classroom training with reminders sent through mobile phones and personal counseling through home visits (discussed in section 0.6). Consumer protection refers to institutions, laws, and policies mandated by governments and organizations to ensure the rights of consumers, including the free disclosure of information. These rules are designed to protect consumers against predatory activities of businesses and financial service providers. Inter- ventions that involve consumer protection are considered a form of financial capability because they often involve mechanisms to direct individuals toward “better” decisions or provide them with information that will affect their behavior. For example, in response to a survey that revealed low levels of capability for choosing from among a wide variety of financial products, policy makers in the United Kingdom developed a set of “decision trees” designed to assist consumers through the optimum pathway for selecting the appropriate products.2 A more recent example is a Trust Fund–financed experiment in Mexico that tests the impact of a program that modified product-specific disclosure statements about interest rates and other loan characteristics to help consumers better understand their choices (discussed in section 0.6). Behavioral treatments are motivated by recent developments in psychology and behavioral economics. These are interventions that either help people over- come or harness behavioral biases. The most well-known interventions in this category are those involving “nudging,” which refers to programs designed to More information on this initiative may be found at http://www.pensionsnetwork.com/ 2  pensions_calculator/pdfs/trees.pdf (accessed May 4, 2014). Overview  ◾  7 push people toward better decisions by altering the choices they face (Thaler and Sunstein 2008). This may take the form of a choice architecture, sending reminders, providing incentives, or a combination of these, most commonly by changing default options for retirement or savings plan enrollments, which take advantage of the common inertia that people exhibit when confronted by finan- cial decisions. Other types of behavioral interventions that have received attention lately include programs that aim at influencing actions through “learning by doing,” peer pressure, and social networks. The premise behind learning by doing is that once individuals become users of formal financial systems, their interactions with banking may provide the basis for learning and skills development. The assump- tion behind programs that explore social networks is that both learning and deci- sions may be spread from certain participants who receive a treatment to others who do not but who are in contact or proximity with the treated individuals. More recent examples include Trust Fund–financed experiments in Kenya, Uganda, and Malawi, where researchers are investigating, respectively, the impact of learning by doing through participation in automated savings accounts, social networks on transferring practices, and activating mental accounting to improve savings (discussed in section 0.6). Financial capability programs may come in different forms and follow different theories of change. Understanding the differences among interventions is important for proper evaluation and accurate interpretation of results. For evalu- ation, having a clear understanding of the program’s objectives and its conceptual framework is critical to developing a proper hypothesis, mapping variables along the causal chain, and identifying indicators from inputs to outputs and outcomes. 0.4 PAUCITY OF THE EXISTING EVIDENCE BASE Prior to the Trust Fund, there was little rigorous empirical research designed to identify the impact of financial capability interventions. While there was some research on these issues, earlier studies can generally be characterized by a lack of rigor. A comprehensive review of financial capability initiatives commissioned by the FSA in 2008 concluded that “not only has there been relatively little work in the past on financial capability in the United Kingdom or other countries, but also that rigorous, credible policy evaluation showing the incremental impact of finan- cial capability work is difficult to find” (Atkinson 2008). This was later confirmed by another stocktaking exercise conducted by the Organization for Economic Co-operation and Development (OECD) in 2009, the conclusions of which were presented at the 2009 OECD/International Network on Financial Education (INFE) Conference in Brazil. In 2010, the World Bank and the RAND Corporation undertook another stock- taking exercise under the Trust Fund program, built on the previous reviews by the 8  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FSA and the OECD and aimed at capturing more recent studies. The review also aimed to expand the scope of the reviewed evaluations to account for alternative and broader financial capability interventions. The results confirmed the paucity of evaluations around financial capability programs in general and revealed the following: ◾◾ Most of the evaluations lacked rigor and did not address the key princi- ples of good evaluations related to internal and external validity, such as confounding factors, selection bias, spillover effects, contamination, and heterogeneity. ◾◾ From the ones rigorously conducted, the majority applied experimental methods, which are effective in determining cause and effect, but may fall short in explaining how and why impacts were (or were not) observed and may have uncertain external validity. ◾◾ The majority of the evaluations lacked information about the intervention. They also lacked details about the impact measures (most reported that there was no change in financial literacy) or even the use of inappropriate measures given the nature of the interventions. ◾◾ Linked to the above points, the majority reached sweeping conclusions, such as that financial education has no impact on financial literacy/capa- bility. A closer look at these studies, however, reveals that their results apply to that particular intervention, delivered in a specific way, to a partic- ular audience, at a specific time, and by a particular agency; they do not necessarily generate lessons that apply to other interventions and circum- stances. ◾◾ There was an overemphasis on measuring changes in knowledge, as opposed to other outcomes, such as changes in attitudes, behavior, etc. ◾◾ Most of the programs evaluated were workshop based. The lack of evidence is common to many areas of public policy. Evaluation is undersupplied in large part because it is considered a public good and faces the classical free-rider problem. It creates knowledge that can benefit the public, but the costs are usually incurred by the provider. Even when mandated, however, evaluation may not be welcomed, because it is perceived to be costly and technically challenging, and can pose reputational risks. Considering that most programs, especially in developing countries, face budget constraints, diverting resources to research is a difficult decision for managers. Further, a well-de- signed evaluation is also challenging to implement. To be able to answer both the “what” and the “why,” evaluation often needs to employ a combination of quantitative, process, and theory-based approaches, and follow a well-defined results framework. Also, when using randomized experiments, planning of the research design is required before the program begins. Evaluation may also be Overview  ◾  9 politically controversial, especially if it reveals that a program has not achieved its intended impact. Lastly, as with any field in its infancy, there are a number of narrowly defined research questions that need to be addressed. These questions often determine program characteristics that are required to permit proper testing. Therefore, in addition to evaluating existing programs, it is often important to work with program developers to design innovative interventions in the understudied areas. 0.5 SELECTED APPROACH: COMPETITIVELY SELECTED FIELD EVALUATIONS The discussion in the previous sections highlights that the field of financial capa- bility is relatively new and characterized by limited empirical evidence. Estab- lished to tackle essential gaps in research, the Trust Fund evaluation program developed three parallel initiatives: ◾◾ Funding of a number of individual and cross-country field evaluations that allow experimentation with different program designs and research tech- niques ◾◾ Development of a methodological and operational toolkit for M&E of finan- cial capability programs in low- and middle-income countries ◾◾ Capacity building for design and implementation of rigorous evaluations in client countries through a series of clinical evaluation workshops3 To better understand the state of knowledge of program effectiveness, and to identify gaps in literature to guide the funding of evaluations, the Trust Fund undertook a stocktaking exercise in the early stages of program development (briefly mentioned earlier in the chapter). This review was led by the World Bank in collaboration with the RAND Corporation and other experts in the field. While other reviews of financial capability programs had been conducted in the past, they were limited in terms of the information provided about the reviewed evalu- ations. The Trust Fund team focused on gathering information related to program characteristics and performance, as well as the rigor of the evaluations. Based on a carefully selected sample of 129 evaluations, a number of under-researched areas were identified. First, in terms of thematic coverage, budgeting and saving were found to be covered most widely. Other topics that are highly relevant were found to be underrepresented, such as financial capa- bility related to credit, loans, investments, and the usage of insurance products. Again, this chapter describes the field evaluations which are presented in this volume; 3  for more on the Trust Fund’s M&E toolkit and capacity-building activities, refer to the Trust Fund website at http://www.finlitedu.org/. 10  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Second, the majority of the evaluations were found to focus on the delivery of education through workshops and seminars, followed by classroom and coun- seling. Alternative means of delivery, such as the use of media, marketing, behav- ioral treatments, or even mixed interventions, were found to be understudied. Third, the majority of evaluations either used process or randomized controlled trials. Very few employed mixed methods or examined spillover effects. The review also revealed a gap in studies that compare the impact of inter- ventions across countries, contexts, and populations. Lastly, a large number of choice architecture interventions appeared to be under way, particularly in the field of commitment devices. However, studies appeared to treat behavioral economics methods as a distinct alternative to financial capability training, rather than adopting behavioral economics insights to improving financial capability. This exercise helped the Trust Fund establish priority areas to guide invita- tions of proposals and selection of evaluations (table 0.1). TABLE 0.1  Priority areas guiding the selection of Trust Fund field evaluations Thematic focus ƒƒBudgeting and savings ƒƒPlanning for known and unexpected events (savings for old age, participation in pension schemes, precautionary savings, formal and informal insurance) ƒƒPlanning and making provisions for known needs and changes in circumstances (investments, loans and debt, education planning, changes in finances due to job loss, sickness, disability, or seasonal variations in income or crops) Program ƒƒBroadcast media or new media, such as radio, television, print delivery material, SMS-mobile technology, Internet-based communication, and computer games ƒƒPrograms that link financial education to government cash transfers Special ƒƒComparing the relative impact of different methods of delivering research financial education, including relative cost-effectiveness—e.g., eval- questions uation of the impact of school-based models or workshops versus nontraditional learning, such as radio, games, theater, TV, films, print media ƒƒExamination of the importance of the intensity and duration of exposure to education or information—e.g., measuring how much information is optimal and how often it needs to be reinforced to impact behavior Evaluation ƒƒPreference for mixed methods and underrepresented approaches; methodology these may include, e.g., regression discontinuity design, matching estimators, and difference-in-difference techniques Regional ƒƒPreference for underrepresented regions: Eastern Europe, Central coverage Asia, and the Middle East and North Africa Overview  ◾  11 0.6 FILLING THE GAPS THROUGH FIELD EVALUATIONS The 17 field evaluations financed by the Trust Fund (16 of which are discussed in this book, as the results from the remaining study were not yet available at time of press) represent a highly coordinated approach across themes and countries and involve a large number of clients and technical experts. All evaluations aimed at addressing critical gaps in research by contributing both methodologically and conceptually. This section provides a brief description of the individual designs and their methodological characteristics. It discusses them in three broad cate- gories: traditional financial education through schools and workshops, nontra- ditional education through entertainment media and marketing, and financial education through mixed interventions. For each set of studies, a box summa- rizing the key aspects of each evaluation is provided; these summaries indicate the corresponding chapter of this volume in which the evaluation is discussed at length. On methodology, while the Trust Fund adopted an open view on research design, all of the studies funded aimed to determine both causality and attribu- tion. In most cases, this was satisfied through randomization, but other methods for identifying counterfactuals were also adopted, especially in cases of media and social marketing interventions where treatment is not easily excludable. In addition, some studies supplemented the methodologies with qualitative and process evaluation to explain how and why change occurs. Conceptually, the evaluations intended to answer a diverse set of questions; while individually they were designed to test specific theories, collectively they intended to contribute to the same broader policy questions. 0.6.1 Traditional financial education: schools, workshops, and training seminars Does formal financial education delivered through school curricula, workshops, and trainings work in improving knowledge and desired outcomes? The six evalu- ations described in box 0.1 explore this question.4 As with most efforts to study the impact of policies, the main challenge in esti- mating the effect of financial education on individuals’ knowledge and behavior is establishing counterfactuals—that is, determining what would have happened to those individuals in the absence of the intervention. Almost all design discussions are concerned with this question; and in general, the more attention that can be dedicated to the design of the project ex ante, the simpler and more robust is the estimation of results ex post. All six studies discussed in this subsection achieve 4  Throughout the boxes presented in this section, the reported results on impact are statis- tically significant with at least 90 percent confidence. 12  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES BOX 0.1  Evaluations of interventions featuring traditional financial education THE IMPACT OF FINANCIAL LITERACY TRAINING FOR MIGRANTS: EVIDENCE FROM AUSTRALIA AND NEW ZEALAND (CHAPTER 1) ƒƒ Pathways to change: Group-based financial literacy seminar ƒƒ Thematic focus: Remittances, credit and financial product selection ƒƒ Target groups: Migrant workers ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys ƒƒ Results: Impact on increasing knowledge and information-seeking behavior and reducing the risk of switching to costlier remittance products. No impact on changing frequency or levels of remittances. The study examined the impact of financial education targeting migrant workers and their remitting behavior in Australia and New Zealand. Training consisted of a two-hour session on reasons to remit, strategies for comparing costs, and information about different remittance products. Results show that the training led to increases in financial knowledge; migrants were more likely to know it is cheaper to send one large transfer than individual smaller ones, and more likely to know cheaper methods of remitting. The study also found that migrants changed behavior in response to knowledge gained; however, training was not found to change frequency of remitting, amount remitted, or the take-up of products. FINANCIAL EDUCATION AND BEHAVIOR FORMATION: LARGE-SCALE EXPERIMENTAL EVIDENCE FROM BRAZIL (CHAPTER 2) ƒƒ Pathways to change: Classroom financial education ƒƒ Thematic focus: Budgeting, savings, and general financial management ƒƒ Target groups: High school students ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Self-administered questionnaires ƒƒ Results: Impact on improving knowledge, attitudes, and behavior The study assesses the impact of high school financial education in Brazil. It includes nearly 900 schools and 26,000 students. Administration of the program through schools allowed for a broad coverage of content in the curriculum. To control for quality of content, the educational material was developed by experts. Separate training was provided to a group of parents of the students to examine whether inside-the-household interactions influenced behavior. Results found that the program increased student financial knowledge by a quarter of a standard deviation, which led to a 1.4 percentage point increase in savings—a relatively large and economically relevant effect. A complementary workshop for parents induced children to save even more. Both current attitudes and forward-looking intentions to save improved. (continued) Overview  ◾  13 DOES FINANCIAL EDUCATION AFFECT SAVINGS BEHAVIOR? EXPERIMENTAL EVIDENCE FROM INDIA (CHAPTER 4) ƒƒ Pathways to change: Classroom training seminar ƒƒ Thematic focus: Budgeting, savings, and general financial management ƒƒ Target groups: Low-income households ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys ƒƒ Results: Impact on savings and attitudes related to financial management. Financial literacy levels did not improve. The study measured the impact of a classroom financial literacy training on the take-up of branchless banking and on savings behavior. The intervention consisted of a two-day training that covered the role of formal banking in people’s lives and responsible borrowing, spending, saving, and cash management. The experiment was conducted on a random sample of 3,000 clients of branchless banking across two adjacent districts in the state of Uttar Pradesh. The results reveal that the intervention had impact on savings and that attitudes related to financial management improved, but overall financial literacy did not. This suggests that a causal chain reaction from knowledge to behavior might not necessarily be required in such order to achieve desired outcomes. THE ROLE OF FINANCIAL ACCESS, KNOWLEDGE, AND SERVICE DELIVERY IN SAVINGS BEHAVIOR: EVIDENCE FROM A RANDOMIZED EXPERIMENT IN INDIA ƒƒ Pathways to change: Classroom seminar, counseling, and reminders through phones ƒƒ Thematic focus: Savings, budgeting, and selection of financial products ƒƒ Target groups: Low-income households ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys ƒƒ Results: No results reported at this stage. The intervention is ongoing. The study measures the impact of a mixed and multilayered intervention on savings behavior. The program consists of three treatments: a classroom financial education training, followed by reminders through mobile SMS and voice messages over a period of several months, and personal counseling through physical visits to participants’ homes. It aims to measure the overall impact but also to disentangle the separate effects from the individual treatments. The novelty of this study is that it experiments with combining traditional financial education with behavioral treatments and explores a multilayered intervention over a longer period of time as opposed to a single and one- time treatment. (continued) 14  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES WHY IS VOLUNTARY FINANCIAL EDUCATION SO UNPOPULAR? EXPERIMENTAL EVIDENCE FROM MEXICO (CHAPTER 7) ƒƒ Pathways to change: Classroom training seminar ƒƒ Thematic focus: Saving, retirement, use of credit ƒƒ Target groups: Bank credit card customers ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys ƒƒ Results: Impact on improving knowledge and savings rates, but no impact on credit card usage The study tested the impact of financial literacy training on savings and borrowing behavior and credit card usage patterns of credit card customers in Mexico City. It involved approximately 40,000 bank consumers. The training course lasted for about four hours and consisted of four modules on savings, retirement, credit cards, and responsible use of credit. Results show a 9 percentage point increase in financial knowledge, and a 9 percentage point increase in saving outcomes, but no impact on credit card behavior, retirement savings, or borrowing. Moreover, administrative data suggest that the savings impact is relatively short-lived. The results point to the limits of using general-purpose workshops to improve financial literacy and decision-making patterns for the general population. EVALUTION OF OLD MUTUAL’S ON THE MONEY PROGRAM: FINANCIAL LITERACY IN SOUTH AFRICA (CHAPTER 15) ƒƒ Pathways to change: Group-based interactive financial literacy seminar ƒƒ Thematic focus: Remittances, credit, and financial product selection ƒƒ Target groups: Members of burial societies ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys and self-administered questionnaires ƒƒ Results: The study is still ongoing. Based on early observations, the intervention was found to improve budgeting and savings, reduce gambling, and decrease risk aversion. The study examines the impact of one-day financial education training on savings, financial planning, budgeting, and debt management. The purpose is to encourage individuals to seek more efficient financial vehicles, as well as to save and use credit wisely. The target population consists of members of burial societies and women’s development groups in the Eastern Cape area of South Africa. It uses a randomized experiment involving approximately 1,300 individuals. Outcome measures are obtained from a variety of sources, including individual surveys and administrative data. While the study is still ongoing, preliminary results show that the intervention had impact on budgeting skills and savings behavior, as well as on reducing gambling and decreasing risk aversion. Overview  ◾  15 this through random assignment of beneficiaries to treatment and control groups. Data are then collected to measure outcomes, in some cases combined with qualitative and process evaluations. Random assignment requires that the evaluation objectives and methodolog- ical development for data collection and analysis be in place before the program is rolled out, and this requires collaboration with program providers early in the process. Close collaboration among the learning and program staff is especially important to address ethical issues, ensure compliance and tracking among both treatment and control groups, and minimize attrition and spillover. The projects discussed here adopt innovative methods to address these concerns. Furthermore, these projects are designed to provide additional insight on whether different aspects of program design, such as content, delivery mecha- nism, context, and duration, affect results. For example, the novelty of the Brazil project (discussed at length in chapter 2) is that it evaluates financial education incorporated across a number of subjects in the standard school curriculum and delivered during three consecutive academic semesters. Although financial education in schools has been studied before, this is the first time it has been integrated so extensively in the school program of study. The two India studies (one of which is the subject of chapter 4) explore the impact of classroom educa- tion when combined and reinforced over time through reminders and personal counseling. This, too, is different from evaluating single and one-time interven- tions usually delivered in much shorter periods of time. 0.6.2 Nontraditional financial education: use of mass media and social marketing As discussed earlier, lessons from psychology suggest that behavioral treatments can be effective in achieving outcomes. The five Trust Fund evaluations described in box 0.2 represent the first experiments that incorporate lessons from behav- ioral psychology into financial capability programs. All of the studies summarized in box 0.2 experiment with different ways of incorporating financial education into alluring stories or short messages trans- mitted via mass media and social marketing campaigns or innovative technol- ogies. Conceptually, the studies are designed to shed light on whether media and marketing tools can improve individuals’ financial capabilities and improve decision making, or whether school-based financial education material is more effective when presented in a more entertaining way. Also, they provide valuable insight into the extent to which these interventions, by appealing to emotions and sticking to memory, lead to more effective decision making in the future. Conversely, “learning-by-doing” experiments test the reverse hypothesis; that by “nudging” and otherwise affecting behavior directly (e.g., enrolling in automatic savings), individuals will become more interested in interacting with financial institutions and a broader range of financial products. 16  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES BOX 0.2  Evaluations of interventions featuring nontraditional financial education UNDERSTANDING AND IMPROVING HOUSEHOLD INVESTMENT BEHAVIOR IN BRAZIL: A STOCK MARKET SIMULATOR AS A LEARNING-BY-DOING FINANCIAL LITERACY TOOL (CHAPTER 3) ƒƒ Pathways to change: Online stock market simulator ƒƒ Thematic focus: Stock market participation and investments ƒƒ Target groups: Stock market simulator users and participants ƒƒ Evaluation method: Randomized controlled trial design and regression methods ƒƒ Data collection: Stock market simulator data and individual stock market data ƒƒ Results: No results reported at this stage. The intervention is ongoing. The study measures the impact of financial education on stock market participation and investment in Brazil. It involves an online stock market simulator that serves as a platform through which financial education is passed to participants and as a mechanism through which participants develop practical experience. A combination of education, warning, and reminders about good practices and learning- by-doing is expected to improve decisions over time. The study uses data of approximately 600,000 investors. Understanding why people make the investments they do can help identify interventions to improve consumer protection and support development of capital markets. COMIC FX IN KENYA: CAN CARTOONS IMPROVE THE EFFECTIVENESS OF FINANCIAL EDUCATION? (CHAPTER 5) ƒƒ Pathways to change: Comic books ƒƒ Thematic focus: Financial management and saving ƒƒ Target groups: Schoolchildren ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys ƒƒ Results: No impact on literacy levels and on savings behavior. However, it found impact on the likelihood that students want to start a business in the future. The study tests the absolute and relative impact of different program delivery mechanisms on the financial capability and behavior of Kenyan youth. It compares delivering education though classroom with that through comic books and radio shows. It uses a sample of 220 high schools, randomly assigned to two main treatment groups, a placebo group and a control. One novelty of this study is that in addition to completing both baseline and endline surveys, students are also asked to make financial decisions using real resources. This allows recording how actions differ from stated intentions, and how both stated intentions and actions change over time. Results show little evidence that the interventions improved financial literacy. Similarly they find no effect on stated and actual savings behavior. However, the study finds impacts on the likelihood that students want to start a business in the future. (continued) Overview  ◾  17 LEARNING BY DOING? USING SAVINGS LOTTERIES TO PROMOTE FINANCIAL INCLUSION IN NIGERIA (CHAPTER 9) ƒƒ Pathways to change: National marketing campaign and savings lottery ƒƒ Thematic focus: Savings and use of banking services ƒƒ Target groups: Existing and new bank users ƒƒ Evaluation method: Regression; discontinuity design ƒƒ Data collection: Microdata collected from the banks on daily account balances ƒƒ Results: Increased savings and use of additional financial products within a week of intervention. However, no evidence on persistent changes after incentives removed. The study measures the impact of a national marketing campaign launched by one of the largest banks in Nigeria to encourage savings. It entails lottery prizes for individuals who open a savings account and maintain a threshold amount in the account for 90 days during a period of three months. The lottery is publicized through advertisements with celebrity endorsements and media releases. The research assesses the extent to which the different components of the campaign affect the take-up rate. It also measures the impact of “learning-by-doing”—the extent to which interactions with banking motivate people to continue to save. Results report that during the intervention there was improvement on savings behavior and on the usage of the bank’s other financial products. However, there was no evidence that the incentive program led to persistent changes after explicit incentives were removed. NIGERIA’S NOLLYWOOD NUDGE: AN ENTERTAINING APPROACH TO SAVING (CHAPTER 10) ƒƒ Pathways to change: Entertainment media (feature film) ƒƒ Thematic focus: Savings and credit ƒƒ Target groups: Low-income households and small business owners ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Surveys and self-administered questionnaires ƒƒ Results: The study is still ongoing. Early observations report on the impact on perceptions and trust in microfinance institutions and increasing the take-up of savings accounts in the short run. Limited evidence of an impact on longer-term behavioral change. The study involves a sample of 3,000 individuals in Nigeria to assess the extent to which a feature film can promote responsible borrowing and improve savings. A basic premise on which the study is developed is that emotions have an influence on actions, and while the emotional state might be transient and short-lived, the decisions reached under the emotional state could potentially provide the basis for future actions. To capture immediate actions, the experiment includes the presence of microfinance institutions at the movie screening venues. Furthermore, to control for spillover effects—that is, the extent to which individuals in the treatment and control groups might talk to each other and share information—the experiment adopts a mix of individual and cluster randomization. (continued) 18  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES HARNESSING EMOTIONAL CONNECTIONS TO IMPROVE FINANCIAL DECISIONS: USING SAVINGS LOTTERIES TO PROMOTE FINANCIAL INCLUSION IN SOUTH AFRICA (CHAPTER 11) ƒƒ Pathways to change: Entertainment media (television soap opera) ƒƒ Thematic focus: Debt management ƒƒ Target groups: Low-income households with and without existing consumer debt ƒƒ Evaluation method: Randomized encouragement design ƒƒ Data collection: Phone and face-to-face surveys and qualitative focus groups ƒƒ Results: Improvement of knowledge and borrowing behavior The study investigates whether debt management may be improved through TV soap operas. To control for content quality, the project develops a soap opera storyline through focus groups. It involves around 1,000 randomly selected individuals divided into treatment and control groups. The treatment watches a soap opera with financial literacy messages, called Scandal!; the control watches a different show aired at the same time but with no financial literacy messages. Financial incentives are provided to ensure compliance. Results report that individuals assigned to watch Scandal! had higher financial knowledge on issues highlighted in the storyline. Scandal! viewers were more likely to borrow from formal sources, less likely to engage in gambling, and less prone to enter hire purchase agreements. Mass media campaigns present a unique evaluation challenge and require innovative evaluation methods. First, because it is not possible to prevent people from watching television (it is not excludable), it is difficult to distinguish between treatment and control groups and to minimize spillovers. Second, because these interventions tend to be multidimensional and contain several levels of treat- ment, it is difficult to identify the specific program characteristics responsible for impact. The teams have developed novel mechanisms to maximize the sepa- ration between treatment and control, to motivate participation, and to design placebo treatments as falsification tests. These methods are explained in more detail in the relevant chapters. 0.6.3 Financial education through combined interventions Five Trust Fund evaluations explore the impact of mixed or combined interven- tions on knowledge and behavioral outcomes, as described in box 0.3. In addition to adopting rigorous methods to measure impact on behavior and outcomes and addressing the methodological concerns discussed throughout this chapter, the five studies addressed in this subsection all explore a number of questions that have been peripheral to the field until recently. These topics include providing product-specific information disclosure, improving mental accounting, and taking advantage of network and peer effects on knowledge and Overview  ◾  19 BOX 0.3  Evaluations of interventions featuring combined methods SOCIAL NETWORKS, FINANCIAL LITERACY, AND INDEX INSURANCE: EVIDENCE FROM A RANDOMIZED EXPERIMENT IN KENYA (CHAPTER 6) ƒƒ Pathways to change: Comic books ƒƒ Thematic focus: Long-term planning; index-based weather insurance ƒƒ Target groups: Rural, small-scale farmers ƒƒ Evaluation method: Randomized controlled trial design and regression methods ƒƒ Data collection: Face-to-face and phone surveys ƒƒ Results: Impact on encouraging the take-up of index-based drought insurance The study presents a randomized field experiment measuring the direct impact and social network spillovers of providing financial literacy and discount vouchers on farmers’ decision to purchase index-based drought insurance in Kenya. The experiment covers around 14 villages and uses comic books as a delivery mechanism of financial education; the comic details the index-insurance product and how it can help families protect themselves from the risk of drought. The study finds social network spillovers to the provision of financial literacy materials, but no spillovers to the provision of discount vouchers on farmers’ decision to purchase insurance. It further finds that financial materials have spillover effects on farmers’ attitudes toward insurance but limited effects on understanding as narrowly measured in the survey. These results provide suggestive evidence that financial literacy materials are efficacious in encouraging take-up when farmers’ social contacts similarly receive access to financial literacy materials. FINANCIAL (DIS-)INFORMATION: EVIDENCE FROM AN AUDIT STUDY IN MEXICO (CHAPTER 8) ƒƒ Pathways to change: Product disclosure formats, and mobile SMS and telephone counseling ƒƒ Thematic focus: Savings and credit ƒƒ Target groups: Low-income consumers of credit ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Surveys and credit reports ƒƒ Results: Preliminary results suggest that disclosure and transparency improve the ability of consumers to select the best product out of several options and to identify the least expensive of several credit products offered. The study measured the impact of product-specific information disclosure on financial decisions. It assumes that the more transparent and relevant the information, the better consumer decisions will be with regard to product selection. It involved development and testing of a series of alternate product-specific disclosure formats, which were then used by low-income consumers in Mexico to choose between a series of credit or savings products. The testing of formats was complemented with financial education information delivered prior to the exercise to some participants, either (continued) 20  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES via SMS messages or by phone consultation. Preliminary results suggest that disclosure and transparency improve the ability of consumers to select the best product out of several options and the ability to identify the least expensive of several credit products offered. These findings point to potential benefits of focusing on product-specific information disclosure and consumer education. THE IMPACT OF FINANCIAL EDUCATION ON FINANCIAL SERVICE USE: EVIDENCE FROM A FINANCIAL DIARIES STUDY IN UGANDA (CHAPTER 12) ƒƒ Pathways to change: Classroom training sessions ƒƒ Thematic focus: General financial education topics ƒƒ Target groups: Low-income households ƒƒ Evaluation method: Qualitative (financial diaries) ƒƒ Data collection: Face-to-face surveys ƒƒ Results: Qualitative findings suggest changes in knowledge, skills, and attitudes; however, they also indicate that they do not always translate directly into behavior change, at least not within a short time frame. The study uses financial diaries in combination with in-depth interviews in Uganda to understand and measure the financial capabilities of low-income households. It compares changes in knowledge, skills, attitudes, and behaviors of respondents in the treatment and comparison groups, highlighting situations where the former underwent a change that might be the result of the impact of the financial education. Results suggest that financial education affects knowledge, skills, and attitudes. Nevertheless, they also indicate that they do not always translate into behavioral change and affect decision making, at least not within a short time frame. However, there is some suggestion of change in savings behavior in terms of saving at home. INCREASING THE IMPACT OF CONDITIONAL CASH TRANSFERS THROUGH FINANCIAL LITERACY: EVALUATION PILOT IN THE DOMINICAN REPUBLIC (CHAPTER 13) ƒƒ Pathways to change: Professional and peer trainings ƒƒ Thematic focus: Household and business financial management ƒƒ Target groups: Conditional cash transfer beneficiaries ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Surveys ƒƒ Results: No results reported at this stage. The intervention is ongoing. The study assesses whether conditional cash transfer (CCT) programs can be leveraged to deliver financial education and affect both knowledge and behavior. Evaluations of CCT programs have shown that they can be successful in increasing usage of health care and education services. This project explores the extent to which CCTs can improve financial capabilities. Working with around 60 beneficiaries of the Solidaridad cash transfer program in the Dominican Republic, the study randomly selects one group of beneficiaries to participate in the training and another group that does not to serve as a control. The experiment is further divided into subtreatments to test whether (continued) Overview  ◾  21 training delivered by professionals versus peers has different effects. It also measures the relative impact of business training versus soft job skills training on decreasing unemployment among beneficiaries. DIRECT DEPOSIT AND COMMITMENTS TO SAVE IN MALAWI (CHAPTER 14) ƒƒ Pathways to change: One-on-one training; labeled banking accounts ƒƒ Thematic focus: Household and business financial management ƒƒ Target groups: Low-income agricultural wage earners and smallholder farmers ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Surveys ƒƒ Results: No results reported at this stage. The intervention is ongoing. The study investigates innovative ways to address low levels of formal savings by leveraging psychological mechanisms. The target population consists of low-income agricultural wage earners and smallholder farmers in Malawi. It examines whether direct deposit of wages as opposed to cash payments can help individuals match desired savings and expenditure patterns with actual behavior. It also tests whether labeling particular bank savings accounts with particular expenditures and labels (for example, college fund, car purchase, etc.) reinforces commitments to save. FINANCIAL MANAGEMENT AND VOCATIONAL TRAINING IN UGANDA: IMPACT AND NETWORK EFFECTS IN INFORMAL INDUSTRIAL CLUSTERS (CHAPTER 16) ƒƒ Pathways to change: Practice-based vocational training and classroom-based business training ƒƒ Thematic focus: Vocational and business training, and business network effects ƒƒ Target groups: Small-scale industries and business owners ƒƒ Evaluation method: Randomized controlled trial design ƒƒ Data collection: Face-to-face surveys ƒƒ Results: Very preliminary results suggest some short-term effects of training on financial literacy and technical knowledge, optimism, and adherence to technical standards, but not on core business outcomes. The intervention is ongoing. The study explores the extent to which personal financial choices are affected by peers. It consists of a randomized evaluation in Uganda to identify the impact of a comprehensive financial management and vocational training program for small-scale industries, focusing on network effects. The study identifies business networks and examines whether the enhanced knowledge received through the training program spreads to other businesses and across networks, influencing certain behavior among the untreated population. Potential positive spillovers would constitute efficient ways to scale the impact of trainings and provide a natural source of leverage for these programs. 22  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES practice transfer. One study (discussed in more detail in chapter 13) also made use of the conditionality imposed by a cash transfer program to achieve changes in financial behavior among program beneficiaries. Among other things, these studies expand the range of programs that can be used to improve financial capabilities. The evaluation approaches adopted encompass a range of methods, predominantly through random assignment to treatment and control groups, with closed-form data collection, although some studies combine these methods with qualitative and process techniques. These methods are presented in more detail in the respective chapters. 0.7 LESSONS AND POLICY IMPLICATIONS This section outlines some general lessons that can be inferred from the Trust Fund–supported evaluations. Some of these lessons and suggestions for future work have to do with the design and orientation of further research, and some concern the design and implementation of policies to enhance financial capability and behavior. 0.7.1 Process of implementation Close collaboration with the measurement component of the Trust Fund activities strengthened the orientation and impact of learning. The Trust Fund consisted of two main components—the first to define and measure what it means to be financially capable (through survey development), and the second to learn how one can achieve financial capability (through evaluation). The learning half benefited enormously from the measurement half: the goals of programs to enhance financial decision making have been shaped by a better understanding of what financial capability really entails. This experience suggests that future activities in this field ought to be built on both measurement and evaluation simul- taneously. The measurement work defined financial capability and identified the levels of financial capability in a population and among different target groups; program evaluation identified the interventions needed to improve financial capa- bility. Policy makers need information on both identification of targets and appro- priate policy measures to address them. Rigorous evaluation of existing programs is important; however, the knowl- edge generated may not address critical research gaps. To test new and refined theories, evaluations must encourage and support the development of innovative interventions. The objective of the Trust Fund was to finance the impact evaluation efforts, not the interventions. It was assumed that programs existed and that through systematic solicitation, interventions that permitted different types of hypothesis testing would be identified. On the contrary, it was a challenge to identify programs that responded to specific research constraints. Overview  ◾  23 The theory to be tested often determines the design of the program. For example, to compare whether a program is more effective when delivered through comics than through traditional classroom textbooks, an existing program would have to be identified that entailed the delivery of the same financial education content through these two mechanisms while targeting the same audience. Such programs are often absent in practice. Therefore, after gaps in the literature are identified, it is often necessary to partner with researchers and program providers to either design new or modify existing interventions to respond to the necessary research characteristics. It is critical to clearly identify knowledge gaps to target financing in priority areas. A systematic review of the existing and ongoing literature including both survey-based and experimental work helped identify the existing knowledge and knowledge gaps and describe how the proposed work is related to or builds on prior efforts. While this may sound obvious, when the Trust Fund joined the inter- national efforts, there had been no systemic review of evaluations in the field—at least not sufficient to provide detailed information related to the characteristics of program design and delivery, or methodological approaches. To address this void, the Trust Fund undertook a stocktaking effort based on which a number of gaps emerged, both on the methodological and conceptual fronts. Building on this exercise and maintaining a database of these studies is a valuable resource to minimize duplication of effort and to ensure that resources are invested in priority areas, that the results from different studies can be combined to inform the broader policy questions, and that future work builds on the existing studies to explore their implications fully. Collaboration among policy makers, program staff, service providers, and those responsible for learning helps enhance validity. The design and imple- mentation of innovative, appropriate, and sustainable interventions to advance financial capability require coordination among the various key stakeholders in the process. This is especially true of the private sector, whose participation is vital, and who will not be encouraged to participate in activities that are inim- ical to or even peripheral to their interests. Thus, programs must be designed to be “incentive compatible”: that is, the interests and objectives of all stake- holders, including both consumers and private service providers, should be taken into account when designing programs. Programs must provide or enhance the services consumers believe best suit their needs, as well as encourage consumer take-up. This is often a challenge for researchers, especially when conducting studies specifically designed to test hypotheses and generate findings. Strong collaboration with stakeholders during project development is crucial for appro- priate research design, proper implementation, and adequate inferences made from the data collected. 24  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 0.7.2 Research methodology When feasible, mixed methods for evaluation should be adopted to answer both whether X causes Y and to understand why and how. In choosing impact evaluation methods, it is important to think about causality, as well as why and how a program works or does not work. Otherwise, even after a highly rigorous approach is adopted, results could be inadequate—or worse, could misguide policy. Take, for example, a randomized assessment to measure the impact on savings of a financial education seminar delivered to employees of a company that finds no significant impact. Such a result could be interpreted to suggest that the program does not warrant replication. In contrast, it could be that minor aspects of the intervention affected the outcome, such as the time of day the seminar was delivered, the motivation of the particular teachers, or a number of other unobserved factors. In other words, knowing whether a program works or does not is only half the battle. Understanding why it does not work helps improve program design. In this view, employing mixed-method approaches, such as relying on randomized controlled trials when feasible but also supplementing them with qualitative and process techniques, helps tell a more complete and useful story. When evaluating multidimensional interventions, it is important to isolate the specific aspects of programs that are responsible for change. This is especially relevant for media-based and marketing interventions that often entail a number of treatments administered at the same time. For example, one of the key aspects of these interventions is relying on alluring stories to appeal to people’s emotions. In this case, it is difficult to precisely estimate whether the outcome of interest is driven from the actual quality of the content delivered, the channel of delivery, the extent to which the audience favors the actors involved, or other factors. One solution is to include placebo groups, “sham treatments” designed to have no effect; however, the more complex interventions are, the more difficult and costly it becomes to develop multiple and layered treatments. The evaluation of comics as a delivery mechanism in the Kenya example (chapter 5) employs a similar approach. For these interventions, combining qualitative and quantitative methods can be particularly valuable for identification of outcomes. When feasible, it is useful to include a “Latin square”–type design, in which different combinations of interventions or components can be tested. These evaluations require larger samples, because they aim to answer many questions simultaneously. Conduct comparative examinations to measure the relative impact of programs and projects delivered in different settings or across different target groups. External validity is often ignored. Programs may be carefully evalu- ated to establish causality, but it is important to understand whether the program will work elsewhere, or among a different group of participants. Moreover, finding that a program works does not make it the ideal program. One should evaluate not Overview  ◾  25 only whether a particular program works, but whether it works better or worse compared to other programs. It is especially important to experiment with nontra- ditional modes of delivering financial education and information, including utiliza- tion of technology, marketing, and behavioral treatments. Of equal importance is comparing delivery methods or alternative versions of the same delivery method against one another, and testing one delivery method in different contexts. This will help in understanding whether a program works across settings, and whether it works better or worse when disseminated through different delivery channels. The relevance of quality and intensity or dosage of the service (e.g., educa- tion and information) delivered should be explored. What is the quality of the financial education and information delivered? Research to date has not established the extent to which the outcomes are driven by content as opposed to other aspects of the program structure. Most studies, though not all, do not include content development or content testing as part of overall program devel- opment and evaluation. Having relevant and engaging financial education mate- rial significantly explains differences in financial literacy (Mandell and Klein 2007). As such, the quality and relevance of the content to the target audience should be an integral part of the research design. Related is the examination of the intensity and duration of exposure to education and information. How much education and information is optimal, and how often should it be reinforced to affect behavior? Is a program with fewer but more intensive contact more or less effective than a program with more contacts for shorter duration? Is a program of six months’ duration twice as effective as one that lasts three months? Studies should also look at changes in impact over time—i.e., at what point do marginal benefits from exposure to education and information start to decrease, and how long do effects last? Insights from psychology and behavioral economics continue to yield important lessons for intervention design and implementation that should be incorporated and tested. A behavioral perspective suggests explanations for behavioral biases that affect financial decisions. Psychological factors such as overconfidence, loss aversion, status quo preference, and hyperbolic discounting, among others, have been found to influence financial decision making. Deriving from Kahneman and Tversky’s Prospect Theory (1979), which was among the first acknowledgments of deviation from the classical full-information utility maximizing model, this literature is indisputably important to consumer financial decision making. However, aside from a couple of studies funded by the Trust Fund, not much has been done elsewhere in incorporating these insights in developing and testing new theories and models. It is therefore of crucial impor- tance to further study this area and examine the types of behavioral biases that affect decisions, to develop a theoretical framework to collect data around the psychology of financial decision making, and to experiment with designing inter- ventions that help overcome such biases. 26  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES There is insufficient information about costs and cost-effectiveness to assess the feasibility of scaling up. It is one thing to experiment with inter- ventions and determine whether a program works, and another to determine the extent to which that particular intervention is scalable, especially from a budget point of view. Information about cost-effectiveness is essential to help policy makers in developing countries allocate scarce resources, and should be inte- gral in rigorous impact evaluation. Interventions can either be compared with a single intervention or a small set of similar interventions or with an agreed-upon benchmark representing the assumed willingness of policy makers and program providers to invest. 0.7.3 Identification of impact Evidence suggests that financial capability has an impact on consumer financial decision making, but results are somewhat mixed, pointing to certain areas that require further investigation. Of the 17 Trust Fund projects, 12 are complete and have reported results as of the date of this volume; these are reported on in chapters 1–12. The remaining five are still undergoing implementa- tion or endline analysis, although most have reported on preliminary findings and observations; four of these are discussed in chapters 13–16. Overall, the results from this set of projects suggest that financial capa- bility enhancement—whether achieved through traditional classroom models or through innovative ways that utilize media, mobile technology, entertain- ment, or behavioral treatments—can be effective in both improving knowledge and awareness, and in changing behavior. Yet generalizing from each individual impact study to other settings and populations is a challenge, as is the case with impact studies in general. As with any randomized experiment, the more rigorous the experiment, the more likely the study is to sacrifice external validity. However, the results from the completed studies combined with preliminary observations from the ongoing studies seem to suggest that, in general, financial capability enhancement programs work better when: ◾◾ financial education content is targeted and relevant; ◾◾ they address consumers at “teachable moments”; ◾◾ they are entertaining and appeal to emotions; and ◾◾ exposure to information is longer term. Evidence strongly suggests that one-time interventions (such as short courses or workshops) can have an impact in the short term, but effects tend to fade over time. This has important policy implications in the field of financial capability, because programs are generally aimed at altering long-term behavior. Overview  ◾  27 0.8 CONCLUSION AND PROPOSALS FOR NEXT STEPS The Trust Fund financing of impact evaluations for a diverse set of interventions has yielded some insights concerning the presence and alleviation of constraints to improved financial decision making. But it has also clarified that rigorous research is necessary for impact evaluations to generate useful lessons and help improve programs in the longer term. This requires the engagement of technical experts at early stages of project development, including at the level of hypoth- esis creation and intervention design. For many of the Trust Fund–financed evalu- ations, the provision of technical expertise was among the most valuable aspects of the program. The Trust Fund program component that funded the 17 evaluations had three goals: (1) to assess the effectiveness of interventions designed to improve finan- cial capability and to ultimately yield better financial decisions; (2) to move beyond traditional financial education and explore whether other alternative mechanisms are effective in achieving desired outcomes; and (3) to address specific gaps in literature identified by researchers as crucial for generating a more conclu- sive body of knowledge to guide policy. For all these dimensions, the research targeted poor segments of populations in resource-scarce environments. The program made its main progress in conducting a systematic review of existing and technically sound evaluations to identify research gaps both theo- retically and methodologically. This stocktaking exercise represents one of the first efforts in the field to generate a database of evaluations that reports on both intervention and research characteristics. Furthermore, the program was successful in promoting evaluations of new and innovative interventions beyond traditional financial education. This was triggered by an outcome-based and agnostic approach that conceptually did not hold predispositions on the path- ways needed to achieve good outcomes but instead explored different theories of change. Research scope was thus expanded to include a variety of innovative programs that had not been addressed by previous studies, such as the utilization of media, marketing, and behavioral treatments in program design. Final results and preliminary findings from the 12 completed and 5 ongoing Trust Fund–supported research projects provide valuable insights that need to be further explored by the research community to be able to offer more conclusive evidence. The following lessons are especially important: Financial capability may be achieved through different interventions; programs that utilize mass media and social marketing tools promise to be especially effective. Results suggest that financial capability programs, whether delivered through schools and workshops or through more innovative methods such as mass media and marketing, can be effective in changing both knowledge and behavior. Alternative interventions, such as using TV soap operas, films, and promotion campaigns, prove to be especially effective. For these interventions, 28  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES two features are presumed to affect the outcome. First, they can be more enter- taining and have the capacity to transmit messages through appealing stories that stick to the memory; and second, they can serve as mechanisms through which messages can be repeated and reinforced over time, keeping the audience engaged in treatment for longer than traditional financial education mechanisms do (e.g., soap operas can last for months or years). The quality of the content delivered affects outcomes. Results tend to suggest that financial capability programs work better when the content is rele- vant, targeted at the right audience, and delivered at teachable moments. While this might seem an obvious point, most financial capability programs, devel- oped by both program providers and researchers, do not tend to focus much on developing the financial education content. The stocktaking exercise conducted by the Trust Fund finds that most of the programs in the field either use similar financial education material or develop generic curriculum that encompasses key concepts on financial management without tailoring the subject to the targeted audience or the specific setting. Financial behavior improves even in the absence of improvements in finan- cial literacy. Trust Fund research finds that programs can be effective in changing individual decision making even in cases where the financial literacy of partic- ipating individuals does not improve. This speaks to the importance of paying more attention to the mechanisms that drive behavior, whether knowledge or behavioral treatments. If the policy objective of a program is to improve savings rates, for example, providing financial education might not be necessary, and in certain cases might not be cost-effective. Interventions that employ media and marketing tools could improve savings rates without following a cognitive route. This does not suggest that financial education should not be promoted, but rather that interventions other than traditional financial education are available to policy makers to help people make better financial choices. Impacts from one-time interventions fade over time, and reinforcement is required to sustain behavioral change. Results suggest that improvement of financial capability among the poor is extremely difficult, especially when the improvement in question refers to longer-term behavior change and habit forma- tion. The results from Trust Fund and other studies suggest that while interventions may be successful in influencing immediate post-intervention behavior, the effects tend to fade over time. This observation is in line with research in other areas that focuses on mechanisms to change decision making—e.g, in health (promoting a healthy lifestyle) or energy (changing consumption patterns). This is not a demon- stration that the financial capability interventions are not effective, but rather that treatments might need to be repeated over time to maintain their effect. The Trust Fund supported a wide range of interventions in many different contexts that have yielded information of great value to policy makers and Overview  ◾  29 program staff alike with respect to the orientation, content, and effective delivery of financial capability information across the developing world. Nowhere before have the research and policy communities joined forces in such a coordinated manner to understand these lessons so comprehensively and rigorously. But even this comprehensive effort leaves many questions unanswered. In moving forward, the focus areas for the next generation of research need to be identified. This chapter recommends investing in the following priority research areas: 1. Explore interventions other than financial education. The limitation of traditional financial education in changing consumer financial behavior is well documented by research and is confirmed by the results from Trust Fund projects. And even if they work, they typically have high treatment costs per individual. In contrast to traditional and classroom-based financial education, interventions that utilize mass media or social media (tele- vision programs, radio commercials, mobile phones and other new tech- nology, etc.), social marketing techniques (promotion campaigns, etc.), and behavioral treatments (reminders, choice framing, peer pressure, etc.) have often shown promise to be effective in influencing consumer choices, have mostly low treatment costs per individual, and can be taken rapidly to national scale. A main condition is high technology penetration which is met nowadays in most low- and middle-income countries. This suggests that future research needs to further explore the scope and limitations of these interventions to better understand what specific mechanisms work better and in which settings. 2. Explore the difference between financial literacy and financial capa- bility and conduct comparative research to measure the relative impact on consumer decision making. These two terms, while to an extent used interchangeably in the literature in recent years, mean different things and may have different objectives and outcomes. Achieving financial literacy means improving knowledge that may improve decisions for some financial outcomes—e.g., budgeting or specific investments. Achieving financial capability means improving financial decisions for the same or, perhaps, other outcomes—e.g., increasing the savings rate or asking for financial advice. This conjecture emerges strongly from both the financial measurement work as well as from impact pilots, but the differences and overlaps with regard to policy interventions (say, classroom intervention versus mobile phone reminders) and their effectiveness are little known. 3. Identify interventions with long-term impacts, such as the Trust Fund Brazil project to improve financial education in school. The Brazil pilot (discussed in chapter 2) has focused on increasing financial literacy as well as financial outcomes (saving) and may have long-term effects. These effects require (1) relevant high-quality material/textbook developed by experts, (2) well-trained and highly motivated staff (through incentives), 30  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES (3) multiple treatments over a longer period (three consecutive academic semesters), and (4) the involvement of parents. 4. Explore the importance of intensity or dosage of the service deliv- ered. There is little evidence on how much education and information is optimal or cost-effective and how often it should be reinforced to achieve desired outcomes. Studies should be developed to look at optimal dosages for different interventions and examine timing and repetition effects—i.e., whether impacts diminish as the time spent in classes or exposure to messages declines and how long effects last. 5. Focus on the sustainable long-term behavior change. Results from the Trust Fund suggest that most interventions that find impact usually refer to short-term impact. Individuals generally tend to regress to former (pre-in- tervention) behavior. It is not enough for researchers to identify interven- tions that succeed in controlled clinical trials; they must also show that the impact detected can continue over time, otherwise an intervention cannot be deemed successful and suggested for replication. Future research should focus on the maintenance stage: i.e., ways (or additional treatments needed over time) to help individuals maintain behavior in the longer term. 6. Apply behavioral treatments to address the limitations and biases of individual decision-making capacity to ensure effective policy design. The recent revolution in behavioral economics research has provided a range of empirical evidence that contradicts the rational agent model, suggesting that people often behave in contrast to their best interest and that they are very sensitive to the way choices are framed. This is especially relevant in the field of decision making and personal finance. Future research should therefore explore this area further and examine how behavioral models can inform personal financial decision making. Specifically, it should examine what types of behavioral biases affect choices, in what contexts, and how to collect data around behavioral characteristics. Furthermore, it should examine the interaction between individual behavioral characteristics and external environmental factors that might influence behavior. 7. Integrate cost-effectiveness analysis in research design to inform scale and sustainability. Scientific research must report on both the impact of an intervention on desired outcomes and on the cost of devel- oping and delivering the intervention. This is especially important for low- and middle-income countries, because if a program is deemed effective but expensive, countries might not have the means to deliver at scale. It is understandable that in the early stages of research, when “successful interventions” have yet to be identified, it might not be necessary to invest in cost-effectiveness analysis (especially if no impact is detected). However, as the second-generation research in this area advances, it is crucial that studies that find impact report on benefits and costs. They must present Overview  ◾  31 cost-effectiveness ratios in terms of treatment cost of desired outcome per individual (or similar measures). Studies are also encouraged to focus on cost comparisons between alternative interventions, such as the cost of traditional financial education programs versus alternative methods. REFERENCES Andrade, Eduardo B., and Dan Ariely. 2009. “The Enduring Impact of Transient Emotions on Decision Making.” Organizational Behavior and Human Decision Processes 109 (May): 1–8. Atkinson, Adele. 2008. “Evidence of Impact: An Overview of Financial Education Evaluations.” Consumer Research No. 68. London: Financial Services Authority. Atkinson, A., and F.-A. Messy. 2012. “Measuring Financial Literacy: Results of the OECD/ International Network on Financial Education (INFE) Pilot Study.” OECD Working Paper on Finance, Insurance and Private Pensions No. 15, Organization for Economic Co-operation and Development, Paris. http://dx.doi.org/10.1787/5k9csfs90fr4-en. Banerjee, Abhijit. 2000. “The Two Poverties.” Nordic Journal of Political Economy 26 (3): 129–41. Bertrand, Marianne, and Adair Morse. 2010. “Information Disclosure, Cognitive Biases and Payday Borrowing.” Chicago Booth Research Paper No. 10-01, University of Chicago Booth School of Business, Chicago. Braunstein, Sandra, and Carolyn Welch. 2002. “Financial Literacy: An Overview of Practice, Research, and Policy.” Federal Reserve Bulletin 88 (11): 445–57. Cole, Shawn, Thomas Sampson, and Bilal Zia. 2009. “Financial Literacy, Financial Decisions, and the Demand for Financial Services: Evidence from India and Indonesia.” Working Paper 09-117, Harvard Business School, Cambridge, MA. —. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Cole, Shawn, and Gauri Kartini Shastry. 2008. “If You Are So Smart, Why Aren’t You Rich? The Effects of Education, Financial Literacy and Cognitive Ability on Financial Market Participation.” Harvard Business School, Cambridge, MA. Duflo, Esther. 2006. “Poor but Rational?” in Understanding Poverty, edited by Abhijt Banerjee, Dilip Mookherjee, and Roland Benabou, 367–78. New York: Oxford University Press. Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Gale, William, Benjamin  Harris, and Ruth Levine. 2012. “Raising Household Saving: Does Financial Education Work?” Social Security Bulletin 72 (2). www.ssa.gov/policy/docs/ ssb/v72n2/v72n2p39.html. Grifoni, A., and F.-A. Messy. 2012. “Current Status of National Strategies for Financial Education: A Comparative Analysis and Relevant Practices.” OECD Working Paper 16 on Finance, Insurance and Private Pensions, Organization for Economic Co-operation and Development, Paris. Hilgert, Marianne A., Jeanne M. Hogarth, and Sondra G. Beverly. 2003. “Household Financial Management: The Connection between Knowledge and Behavior.” Federal Reserve Bulletin 89: 309–22. 32  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Kahneman, Daniel, Paul Slovic, and Amos Tversky. 1982. Judgment Under Uncertainty: Heuristics and Biases. New York: Cambridge University Press. Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decisions under Risk.” Econometrica 47 (2): 263–91. Karlan, D., and M. Valdivia. 2011. “Teaching Entrepreneurship: Impact of Business Training on Microfinance Clients and Institutions.” Review of Economics and Statistics 93 (2): 510–27. Kempson, Elaine, and Sharon Collard. 2010. “Money Guidance Pathfinder: A Report to the Financial Services Authority.” Evaluation Report No. 1, Consumer Financial Education Body. http://www.bristol.ac.uk/geography/research/pfrc/themes/advice/pfrc1002.pdf. Klapper, Leora F., Annamaria Lusardi, and Georgios A. Panos. 2012. “Financial Literacy and the Financial Crisis: Evidence from Russia.” Policy Research Working Paper 5980, World Bank, Washington, DC. —. 2013. “Financial Literacy and Its Consequences: Evidence from Russia during the Financial Crisis.” Journal of Banking and Finance 37 (10): 3904–23. Lee, Nancy R., and Margaret Miller. 2012. “Influencing Positive Financial Behaviors: The Social Marketing Solution.” Journal of Social Marketing 2 (1): 70–86. Loewenstein, George, and Drazen Prelec. 1992. “Anomalies in Intertemporal Choice: Evidence and an interpretation.” Quarterly Journal of Economics 107 (2): 573–97. Lusardi, Annamaria, and Olivia S. Mitchell. 2006. “Financial Literacy and Planning: Implications for Retirement Wellbeing.” Working Paper No. 1, Pension Research Council. http://www.dartmouth.edu/~alusardi/Papers/FinancialLiteracy.pdf. —. 2007. “Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education.” Business Economics 42 (1): 35–44. Mandell, Lewis. 1997. “Personal Financial Survey of High School Seniors.” Jump Start Coalition for Personal Financial Literacy, Washington, DC. Mandell, Lewis, and L. S. Klein. 2007. “Motivation and Financial Literacy.” Financial Services Review 16 (2): 106–16. —. 2009. “The Impact of Financial Literacy Education on Subsequent Financial Behavior.” Journal of Financial Counseling and Planning Education 20 (1): 15–24. McKenzie, David, and Michaela Weber. 2009. “The Results of a Pilot Financial Literacy and Business Planning Training Program for Women in Uganda.” Finance and Private Sector Development Impact Note 8, World Bank, Washington, DC. Mullainathan, S., and R. Thaler. 2000. “Behavioral Economics.” NBER Working Paper 7948, National Bureau of Economic Research, Cambridge, MA. Samuelson, William, and Richard Zeckhauser. 1988. “Status Quo Bias in Decision Making.” Journal of Risk and Uncertainty 1 (1): 7–59. Stango, V., and J. Zinman. 2009. “Exponential Growth Bias and Household Finance.” Journal of Finance 64 (6): 2807–49. Thaler, Richard, and Cass Sunstein. 2008. Nudge: Improving Decisions about Health, Wealth, and Happiness. New Haven: Yale University Press. CHAPTER 1 T he impact of financial literacy training for migrants Evidence from Australia and New Zealand JOHN GIBSON, DAVID MCKENZIE, AND BILAL ZIA ABSTRACT Remittances are a major source of external finance for many developing countries, but the cost of sending remittances remains high for many migra- tion corridors. International efforts to lower costs by facilitating the entry of new financial products and new cost comparison information sources rely heavily on the financial literacy of migrants. This chapter presents the results of a randomized experiment designed to measure the impact of providing financial literacy training to migrants. Training appears to increase finan- cial knowledge and information-seeking behavior and reduces the risk of switching to costlier remittance products, but does not change either the frequency or level of remittances. 1.1 INTRODUCTION International migration from a poor to a rich country is perhaps the single act most likely to succeed in dramatically increasing the income of an individual (see, e.g., Clemens, Montenegro, and Pritchett 2008; McKenzie, Gibson, and Stillman 2010) as well as that of family members remaining behind (see, e.g., The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work; Wendy Li, Wasana Karunarathne, and Halahingano Rohorua for leading the training sessions and survey interviews; Kim Hailwood and the New Zealand Ministry of Pacific Island Affairs for their work in helping to develop the training content with us; and all the survey participants and community groups that participated in this study. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 33 34  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Gibson, McKenzie, and Stillman 2013; Yang 2008; Yang and Martínez 2006). The most direct channel through which international migration can lower poverty for household members remaining in a developing country is through remit- tances. However, high costs of sending remittances limit the amount received by remaining household members from a given remittance transfer, as well as the incentives of migrants to send remittances if such transfers are effectively taxed by these high transaction costs. Lowering the cost of sending remittances has thus become one of the most discussed areas for policy intervention in recent years (see, e.g., World Bank 2006), in part because doing so is viewed as politically uncontroversial compared to efforts to increase opportunities for migration. Two of the main policies to lower the costs of remittances have been regulatory reform to allow the introduction of new financial products, and efforts to increase the disclosure of the costs of remitting via each product, pioneered by Mexico (www.remesamex.gob.mx) and the United Kingdom (www.sendmoneyhome.org1). However the efficacy of policies to reduce the cost of remitting and spur competition by allowing new product entry and increasing disclosure of costs relies heavily on the ability of migrants to understand how to use the different methods available for remitting and the costs implied by each method. While systematic evidence on the financial literacy of migrants is scarce, the data available suggest migrants often lack knowledge of the components of a remittance cost, the methods available, or how to compare methods (Gibson, McKenzie, and Rohorua 2006; Gibson et al. 2007). There therefore seems to be promising scope for financial literacy training to change remitting behavior. There is also growing interest from policy makers in providing financial literacy training in this area. Much of the focus on financial literacy training for migrants and their families has traditionally been on either integrating immigrants into the financial system in the destination country through content that focuses on building knowledge of banking services and covering basic household budgeting and savings topics (LIRS 2006), or content focused on encouraging remittance receivers to better use the money they receive, as is the focus in the content of the Microfinance Opportunities/Freedom from Hunger Global Financial Education Program.2 However, a number of countries have also started focusing on teaching the migrants themselves more about the costs and details of different methods of remitting; these include migrant-sending countries such as Indonesia and the Philippines, and pilot programs for seasonal migrants from the Pacific Islands working in Australia and New Zealand. However, to date, there is no rigorous evidence on the effectiveness of such programs. 1  This website has since expanded and changed its name to www.fxcompared.com. The World Bank has also launched a remittance prices database (http://remittanceprices. worldbank.org) covering costs of remittances in 165 corridors. 2  http://www.globalfinancialeducation.org/future.html#remittance. 1.  The impact of financial literacy training for migrants  ◾ 35 This chapter presents the results of a randomized experiment designed to measure the impact of providing financial literacy training to migrants in Australia and New Zealand—countries that had recently launched a remittance cost comparison website (www.sendmoneypacific.org) for sending money to the Pacific Islands, and, in the case of New Zealand, where regulatory reform had led to the introduction of new remitting methods. The training taught migrants the different elements that make up the cost of sending remittances and how to compare costs across methods; explained how different methods of remitting work, including alerting them to the presence of new methods; and also covered content on comparing costs of different methods of short-term credit financing for immigrants. The experiment was carried out on three different groups that had differing levels of existing education and financial knowledge, and differing intensities of remitting. The first group was Pacific Island migrants in New Zealand, who remitted relatively frequently and had relatively low education and financial literacy at baseline. The second group was East Asian migrants in New Zealand, who had low frequencies of remitting but relatively high education and finan- cial literacy; and the final group was Sri Lankan migrants in Melbourne, Australia, who remitted relatively frequently and had relatively high education and financial literacy levels. We find the training led to increases in financial knowledge of the Pacific Island and East Asian migrants, but not of the Sri Lankans, which is consistent with such training being most important for those with either low knowledge or low experience. This increased knowledge was coupled with changes in behavior, with Pacific Island and East Asian migrants being more likely to use information to compare the costs of remitting across different methods, and the Pacific Island sample being less likely to switch remitting channels to methods that were not obviously better. However, we find no changes in the frequency of remitting, nor in the amount remitted. The Pacific Island training also contained information on the costs of different forms of credit. This did not succeed in increasing use of credit cards from a low base, but did lead to an increase in hire purchase loans during a period when they were a relatively good deal, and to individuals setting up rotating savings and credit associations (ROSCAs) to avoid high-interest payday loans. Unfortunately, the comparison of the Pacific Island group with the other two groups is hampered by the fact that few of the East Asian sample regularly sent remittances, while attrition was high and unbalanced by treatment status for the Sri Lankan sample. The results are thus cleanest for the sample the content was developed for in the first place, the Pacific Island group. One implication of these results is that simply informing remitters about remittance costs, which is a relatively cheap and uncontroversial intervention, will not necessarily lower average costs from remitters switching to cheaper methods. Instead, governments targeting reduced average money transfer costs may need to address other barriers, which may include excessive regulation and exclusive arrangements made by state-owned entities that deter new entry into 36  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES remittance corridors.3 Another implication is that the case for providing financial literacy training for migrants needs to rest on criteria other than the financial savings from cheaper remittances—such as the improvements in their capabil- ities from being more informed customers and the potential savings from other aspects of financial management, such as choice of debt levels and instruments. The remainder of the chapter is organized as follows: section 1.2 provides the background to the randomized experiment, in terms of the surveys and financial literacy training that were carried out and the context of the remittance corridors studied. In section 1.3, the results of the experiment are described, focusing on financial knowledge, information seeking, remittance frequency, amounts and methods, credit use, and qualitative evaluations from the study participants. Section 1.4 contains the discussion and section 1.5 the conclusions. 1.2 BACKGROUND CONTEXT, THE SAMPLE, AND THE FINANCIAL LITERACY INTERVENTION The cost of remitting money has fallen dramatically in a number of migration corri- dors over the past 15 years due to increased competition, new product offerings, and the advent of price-comparison websites. For example, Profeco, Mexico’s consumer protection agency, started reporting weekly the cost of sending money from several cities in the United States to Mexico in 1998, and Hernández-Coss (2005) reports that the cost of sending US$300 fell from approximately US$32 in 1999 to US$12 by 2003; by September 2011, one could send US$300 for US$3.60 using Bank of America’s account-to-account or cash-to-cash products.4 Nevertheless, costs of sending money are still high when sending along other migration corridors, with transfer costs between several African countries costing 15–20 percent on a US$200 transaction in 2011.5 This was also the context in work that we did examining remittance costs in the Pacific in the mid-2000s, where we found the costs of sending money from Australia or New Zealand to several Pacific Islands was in the range of 15–20 percent on a typical $NZ 200 transaction (Gibson, McKenzie, and Rohorua 2006; Gibson et al. 2007; McKenzie 2007). This work also revealed that while costs were high on average, there were lower-cost possibilities available—such as the use of debit cards to make automated teller 3  For example, the New Zealand government–run bank with a mandate to serve low-income customers (KiwiBank) is not active in providing a remittance product for migrants. One likely reason is that KiwiBank branches are all in post offices, which already act as agents for an existing money transfer operator, Western Union. 4  Remittance Prices Worldwide (database), World Bank, Washington, DC (accessed March 6, 2012), http://remittanceprices.worldbank.org/Country-Corridors/United-States/Mexico/. 5  http://remittanceprices.worldbank.org. 1.  The impact of financial literacy training for migrants  ◾ 37 machine (ATM) withdrawals—that were not being used, and that few migrants had heard of such methods. Moreover, although a typical remittance transaction incurs both a fixed fee and an exchange rate commission, the latter component was often opaque, leading to migrants often comparing methods of remitting purely on the basis of the fixed-fee component. Spurred by these research findings, the Australian and New Zealand govern- ments and their aid agencies, along with the World Bank, worked to try and lower the costs of remitting in the region. In New Zealand, this resulted in a change in excessive anti–money laundering regulations, thereby allowing banks to give migrants an ATM card for themselves and one for their family back home without the bank having to verify the identity of the second cardholder in person. Westpac Bank was the first to release a new product under these revised regulations, with the Westpac Express prepaid debit card targeted at migrants launched to positive reviews (Stock 2009). Also, these organizations launched a new website for both Australia and New Zealand (www.sendmoneypacific.org), based on the successful sendmoneyhome website in the United Kingdom. This website provides detailed information on the cost of sending remittances from Australia and New Zealand to the Pacific Islands by various channels and is updated regularly. However, despite the introduction of new products and a new information source, the take-up of the Westpac Express product and the volume of trans- fers using it have not been as high as hoped for (Pacific Islands Forum Secre- tariat 2011). One plausible reason suggested for this was lack of financial literacy. Only 12 percent of Pacific Island migrants in our sample had heard of this card at baseline, and less than half of them had ever used any source of information to compare the costs of sending money across different methods. Coupled with increasing policy interest in providing financial education to migrants, we there- fore decided to conduct a randomized experiment to measure the impacts of doing so on financial knowledge and remitting behavior of migrants. 1.2.1 The sample The Westpac Express card and www.sendmoneypacific were both designed for Pacific Island migrants in New Zealand. However, to examine whether training that focuses on understanding how to remit and to compare prices is also effec- tive for other migrant groups, we decided to also consider other migrant groups. Since migrants are a rare population, especially when focusing on migrants from specific countries, obtaining a representative sample can be prohibitively expensive (McKenzie and Mistiaen 2009). We therefore decided to recruit study participants through intercept points where migrant populations are known to congregate, mimicking the approach that would typically be used by policy makers and financial institutions trying to reach migrant populations. This has the advantage of making our results relevant for the population most likely to be the subject of financial literacy efforts, even if it does not allow measurement 38  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES of the impacts on migrants not found in these locations, who are typically less connected to their home countries (and less likely to remit). Our first group consists of Pacific Islanders living in urban areas in the upper North Island of New Zealand. Approximately one-third of the recruitment was from attendees at a Pacific cultural festival in Hamilton, which drew participants from up to 60 miles away (including South Auckland, which has the largest concentra- tion of Pacific immigrants). The remainder was recruited from the main Pacific outdoor market (which operates every Saturday morning) in South Auckland, and from churches in Auckland and Hamilton. The church-based recruitment tended to bring in older participants, whereas the cultural festival participants were typi- cally in their 20s. The Pacific Islanders were predominantly (three-quarters) from Tonga; with the remainder born in Samoa, the Solomon Islands, Fiji, Australia, and New Zealand. In the Pacific Island community, even second-generation migrants send remittances due to the ongoing linkage with extended families in the Islands (Lee 2003); thus, we did not rule out any Australian- or New Zealand–born partic- ipants. The second group was chosen to be the other main immigrant group in New Zealand, East Asians. Chinese and Korean participants from Auckland (four-fifths of the total for the East Asian group) were recruited from five different churches located in north, west, and central Auckland; and from a tai chi group and a Chinese health organization, both located in northern Auckland. The remaining Chinese participants were from Hamilton (60 miles south of Auckland), where they were recruited from several churches and from preexisting research networks of the Chinese team leader (who was based in Hamilton). In no case did any one church or locality contribute more than one-seventh of the sample. This sample was restricted to first-generation migrants. The final group in our study consists of first-generation Sri Lankan migrants in Melbourne, Australia.6 They were recruited through snowball sampling. Initially, 20 people were selected from various Sri Lankan organizations (both formal and informal) in Melbourne. These organizations were selected to represent different demographic and economic groups, in terms of length of time residing in Australia, method of migration (skilled, family reunification, and student), education level, ethnicity (predominantly Sinhalese), and location in the greater Melbourne urban area. Each individual from these organizations was asked to provide names and contact details for five individuals who could be interviewed; out of the 100 potential participants identified this way, 80 on the seed list agreed to partici- pate in the baseline survey. In turn, when the interviews were conducted with these 80 people, they were asked to provide further referrals, leading to another 6  We also planned a sample of Pacific Island migrants in Sydney, Australia, but the field leader in charge of this process experienced health problems partway through the recruit- ment and training which led to this sample strata being dropped from the study. 1.  The impact of financial literacy training for migrants  ◾ 39 129 people who were interviewed. Some of the participants who were obtained through the second round of referrals were uncomfortable with the questions on financial information and remittances, and refused to provide any contact addresses—which ruled them out from being invited to the training or partici- pating in the four follow-up surveys. In order to ensure that the surveys and financial literacy training were conducted in the most effective and culturally appropriate way, we recruited indi- viduals from these same migrant populations to lead the fieldwork for each of the component studies. In fact, each of these team leaders had a Ph.D., two in economics and one in psychology (but specializing in field studies of migrants); so the level of training and skill for the providers of the financial literacy intervention is likely to be atypically high. Each of the team leaders recruited local assistants who were individuals drawn from the same population groups that were being studied. The questionnaires, PowerPoint presentations, and any written material handed out was available in English, Korean, and Mandarin for the participants in the East Asian group, and in English for the Pacific and Sri Lankan groups (English is the language of schooling throughout the Pacific, and the Sri Lankan group members were highly educated, even if English was not their first language). 1.2.2 Baseline survey, randomization, and financial literacy levels Respondents were recruited December 2010–January 2011 (Pacific Island sample), December 2010–February 2011 (East Asian sample), and January–March 2011 (Sri Lankan sample). The resulting sample sizes were 349 Pacific Islanders, 352 East Asians, and 209 Sri Lankans. A baseline questionnaire collected information on their use and awareness of different remittance methods; their financial literacy, with specific emphasis on knowledge relevant to remittances and use of financial instruments; and their background characteristics. Within each of the three samples, we formed eight strata, based on three baseline characteristics: (1) reported frequency of remitting (remitting at least every three months or not), (2) knowledge of the saving in transaction costs from bundling two remittances of US$100 into a single remittance of US$200, and (3) knowledge of which credit card user would face the highest finance charges based on different repayment patterns. Individuals were then randomized by computer into a treatment group, which was invited to financial literacy training; and a control group, which was not. Table 1.1 displays the baseline characteristics of each of the three samples by treatment status. For each sample, an F-test cannot reject joint orthogonality, confirming that we did not get an unlucky draw and that randomization succeeded in achieving balance on baseline characteristics. We see the three migrant groups differ from one another in a number of important ways. 40  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES ◾◾ The Pacific Island migrants are younger and less educated than the other groups, with almost half aged under 35 and only 9 percent having a univer- sity degree. Thirty-nine percent are male, and just under half have a parent or child in a Pacific Island home country. They are relatively frequent remit- ters, with 59 percent remitting at least once every three months, with a mean remittance amount of $NZ 299 and a median of $NZ 200. ◾◾ The East Asian migrants are older, with only one-third under age 35; and more settled, with only one-quarter having immigrated in the past five years. Forty-three percent are male, and 57  percent have university degrees. They are infrequent remitters, with only 6.5  percent remitting within the past three months, despite 66 percent having a parent or child remaining in the home country. The few remittances that do occur are for relatively large amounts, with a mean (median) of $NZ 4,235 ($NZ 1,000). ◾◾ The Sri Lankan migrants differ in being majority male (73  percent); they have the highest education and employment rates, with 59  percent having a university education. They are also relatively frequent remitters, with 55  percent remitting at least every three months, a mean (median) remittance of $NZ 1,525 ($NZ 675), and 75 percent having a parent or child remaining in Sri Lanka. These differences across groups likely in part reflect the different immigra- tion categories of migrant entry: Pacific Islanders tend to immigrate to New Zealand through family reunification and special concessionary migration quotas (McKenzie, Gibson, and Stillman 2010), whereas the Asian migrants to New Zealand typically entered through points systems that reward skills and wealth. The Sri Lankan migrants are typically individuals who entered Australia either as students or as a result of civil conflict in Sri Lanka, qualifying under Australia’s points-based migration system. The baseline survey asked three questions to measure remittance-specific financial literacy, as well as two questions on broader financial literacy related to credit (see annex). Baseline financial literacy was lowest among the Pacific Island migrants: only 49 percent knew it was cheaper to bundle remittances as a single transaction than to send separately (and pay the fixed fee twice); only 5.7 percent knew that the prepaid ATM card was the cheapest method of remitting; and only 3 percent knew that the remittance fee consists of an exchange rate commission and a fixed fee. Knowledge of the available methods of remitting was also rela- tively low, as migrants were asked whether they had heard of each of 10 different methods of sending money (Western Union, Melie Mei Langi, Travelers Cheque, etc.), with the mean respondent having heard of only 3 such methods. Financial literacy related to credit was also relatively low, with only 41  percent knowing that someone who pays only the minimum payment would pay the most on credit card fees; only 3 percent were able to correctly calculate the annual percentage rate (APR) on a two-week payday loan. 1.  The impact of financial literacy training for migrants  ◾ 41 TABLE 1.1  Characteristics of sample by treatment status PACIFIC ISLANDERS EAST ASIANS IN SRI LANKANS IN IN NEW ZEALAND NEW ZEALAND AUSTRALIA TREAT- CON- TREAT- CON- TREAT- CON- MENT TROL MENT TROL MENT TROL Variables stratified on Remit at least every 3 months 0.59 0.59 0.08 0.05 0.54 0.56 Knows it is cheaper to bundle 0.49 0.49 0.64 0.65 0.78 0.78 remittances into large transaction Knows only paying minimum on 0.41 0.41 0.55 0.54 0.44 0.44 credit card costs the most Personal characteristics Male 0.36 0.42 0.45 0.41 0.76 0.69 Age is under 35 0.47 0.49 0.33 0.33 0.28 0.30 First-generation migrant 0.81 0.81 0.99 0.98 1.00 0.99 Migrated within last 5 years 0.34 0.35 0.25 0.27 0.47 0.41 Has a parent or child in the origin 0.47 0.42 0.69 0.64 0.75 0.76 country Married 0.69 0.65 0.69 0.70 0.84 0.93 Education of 5th form (10th grade) 0.46 0.39 0.09 0.11 0.01 0.02 or less University degree 0.10 0.08 0.59 0.54 0.64 0.54 Employed 0.63 0.59 0.42 0.47 0.81 0.80 Uses e-mail at least weekly 0.31 0.33 0.59 0.56 0.73 0.79 Financial characteristics Ever compared costs of sending 0.48 0.47 0.40 0.41 0.62 0.59 remittances Has a checking account 0.35 0.33 0.56 0.58 0.79 0.75 Has an ATM card 0.80 0.76 0.80 0.76 0.88 0.85 Has a credit card 0.15 0.16 0.64 0.60 0.76 0.72 Last amount remitted conditional 288 310 4,235 4,234 1,200 1,884 on remitting ($NZ) Gets APR on 2-week loan correct 0.02 0.03 0.44 0.47 0.42 0.41 Knows components of a remit- 0.03 0.02 0.13 0.11 0.20 0.15 tance fee Number of methods for sending 3.00 2.83 2.63 2.69 3.68 3.64 remittances known Sample size 177 172 179 173 107 102 p-value for test of joint orthogonality 0.913 0.978 0.356 42  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Baseline financial literacy rates were higher among the Sri Lankans and East Asians, reflecting their much higher education levels and greater use of credit cards and checking accounts. Sixty-five percent of the East Asians and 78 percent of the Sri Lankans knew it was cheaper to bundle remittance transactions, and over 40  percent of both groups were able to correctly calculate the APR on a two-week payday loan. However, knowledge of the components of a remittance fee was still low, with only 12 percent of the East Asians and 18 percent of the Sri Lankans knowing the correct answer to this question. As with the Pacific Islanders, these migrant groups only claimed to have heard of 3 or 4 possible ways of sending remittances out of a list of 10–12 methods. 1.2.3 Potential savings from greater financial literacy The remittance methods available to the participants had transaction costs that ranged from almost zero to over 15 percent for a typical transaction at the start of the intervention. The greatest potential gains appear to be for the Pacific Island migrants in New Zealand, as a result of substantial heterogeneity in costs and lower typical remittance amounts. For example, at baseline, spending $NZ 200 (the median transaction) to send money to Tonga (the main destination) would attract transaction costs of 15  percent using a bank transfer, 11–12  percent using either Western Union or MoneyGram, 8 percent using the major indigenous money transfer operator (Melie Mei Langi), or just 5 percent using the Westpac Express prepaid remittance card. Moreover, one Internet-only provider (KlickEx) had transaction costs of less than 1 percent, although no participants had ever used this method. Since the most common methods used were Western Union and Melie Mei Langi, bundling two transactions into one would save the fixed fee of $NZ 8–14, while switching from one of these methods to the Westpac Express card would save $NZ 6–12 per $NZ 200 transaction. There was less variation in costs for the East Asians in New Zealand, and, with few migrants remitting, less potential gains to be had. For money transfers to China, the transaction costs for sending $NZ 200 varied between 14 percent for both Western Union and the most expensive indigenous money transfer operator (Global FX) and 10 percent using the cheapest money transfer operator (Conver- gence Group). Some of the Chinese money transfer operators would only transfer a minimum of $NZ 1,000, for which the transaction costs were as low as 3 percent. Since the cost of remitting falls as a percentage of the amount remitted due to the fixed-fee component, and the median amount remitted for the few East Asian migrants who were remitting was $NZ 1,000 at baseline, the percentage cost of remitting ranged from 3 to 6 percent. For the Sri Lankan participants in Australia, spending $A 200 on a remittance would attract transaction costs of 16  percent using a bank transfer, 9  percent using Western Union, 5 percent using MoneyGram, and just 3 percent using any of the cheapest indigenous money transfer operators (FastCash, Remittance 1.  The impact of financial literacy training for migrants  ◾ 43 Plus, or Serandib). For their median remittance amount of $A 500 ($NZ 675), the transaction costs would range from 11 percent using a bank transfer, 5–6 percent using MoneyGram or Western Union, 4  percent using Kapruka, and 3  percent using FastCash, Remittance Plus, or Serandib. One money transfer operator that began operations after the intervention started, Cash Express, had costs for a $A 500 remittance of just 2 percent. Since the most common methods at baseline were Kapruka and FastCash, there were relatively limited gains to be had from switching providers for this sample. 1.2.4 The financial literacy intervention The financial literacy training content was originally developed by the authors in collaboration with the Ministry of Pacific Island Affairs and piloted on a Pacific Island population in 2009. The material begins with a discussion of the different reasons people remit and the different factors that enter into the choice of method of remitting, such as cost, speed, convenience to the sender and to the receiver, familiarity, trust, and other services offered by the financial provider. The main focus is then on understanding the components of remittance costs, teaching strategies for reducing these costs, and highlighting sources of infor- mation for comparing costs and learning about new remittance products. This included explaining the fixed-fee and exchange rate commission components of the cost and illustrating how much they vary across different providers, showing how the transaction costs fall with the amount sent so that bundling smaller transactions into one large transaction saves costs, and providing information about the sendmoneypacific website for comparing costs and about the Westpac direct debit card as a new product. The remittance material was then adapted for the East Asian and Sri Lankan populations. Since sendmoneypacific does not cover remittance transactions for these remittance corridors, both groups were given instructions and shown how to obtain rates and the expected amount received on the Western Union website, plus ANZ Bank online and fxcompared.com for the Sri Lankans, and MoneyBookers for the East Asians in New Zealand. Initially we had planned that the financial literacy training would focus just on remittances, but inspection of the baseline survey data from the Pacific Island group indicated that very few of the participants had credit cards, and a lack of awareness of this credit channel was corroborated by qualitative discussions with participants. For the Pacific Island group only, we therefore expanded the material covered in the financial literacy training to also include comparisons of sources of credit, especially in terms of their annual percentage rate of interest. The teaching material was based on examples of advertisements for payday lenders and other short-term finance companies that used prominent sports stars to target Pacific Islanders (e.g., figure 1.1a). The participants were taught how to calculate credit interest rates so they could compare the annual cost of a payday or short-term 44  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES loan or hire purchase with the cost of obtaining the same funds either as a cash advance or a purchase with a credit card. The randomly chosen survey respondents were then invited to the financial literacy training sessions, which were held at multiple times and venues to ensure maximum participation from those who were invited. These were typically held at churches, community centers, and sports clubs and were usually for groups of about 30 at a time and took about two hours. In addition to a presentation of about 25 PowerPoint slides, written material was handed out and there were worked examples and continuous discussion with the community. As noted above, the presenters were all members of the immigrant groups themselves to break down any cross-cultural communication barriers. It was emphasized that the aim of the training was not to advocate for any one particular remittance provider but instead aimed to help the participants become more informed consumers who could shop around for better remittance deals (see figure 1.1b as an example). The attendance rate for the training session was 148 out of 177 for the Pacific Island treatment group (84 percent); 3 members of the control group also attended (1.7  percent), accompanying friends in the treatment group. For the East Asian migrants, attendance was 112 out of 179 (63 percent) in the treatment group, with 26 of 173 (15 percent) in the control group attending; these latter were friends from the same churches as the treatment group participants. Among the Sri Lankans, the attendance rate was 60 out of 107 (56 percent) for the treatment group, and none from the control group. All analysis will be based on intent-to- treat effects, using the random assignment to be invited to training. FIGURE 1.1  Examples of materials used in the financial literacy training a. Ad featuring celebrity b. Illustration of remittance cost differences 1.  The impact of financial literacy training for migrants  ◾ 45 1.2.5 Follow-up surveys At the end of each of the following three months after the financial literacy training, all respondents from the baseline survey were given a short follow-up survey on their remittance activity during the past month along with questions on major financial actions taken during the previous month, such as applying for a credit card. In addition, the one-month survey asked several financial literacy questions in order to measure whether financial knowledge had increased with the literacy training. Six months after training, all participants from both the treat- ment and control groups were invited back to community forums. A final round follow-up survey was conducted at the start of this forum, after which both treat- ment and control groups were given information on the main messages of the training course and information on new products and developments in the market that had occurred since the original intervention. Table 1.2 shows the attrition rates by survey round and ethnic group sample. Attrition rates are lowest for the Pacific Island sample, averaging 5 percent at one month, 9–10  percent at two and three months, and 14  percent at six months. In no survey round can we reject balance between treatment and control groups. The East Asian sample has attrition rates of 9  percent at one month, 17 percent at two months, 23 percent at three months, and then 75 percent at six months, again balanced by treatment status. The Sri Lankan group had the highest attrition, with 45  percent of the treatment group already attriting by month 1 compared to 29 percent of the control group, and 76 percent attrition in the six-month follow-up. The high attrition was attributed by the survey leader to refusals to give contact details by some participants, along with discomfort TABLE 1.2  Attrition rates by survey round 1 MONTH 2 MONTHS 3 MONTHS 6 MONTHS Pacific Islanders in New Zealand Treatment group 0.06 0.09 0.09 0.16 Control group 0.05 0.09 0.10 0.13 p -value of equality 0.673 0.886 0.814 0.415 Asians in New Zealand Treatment group 0.10 0.18 0.25 0.74 Control group 0.08 0.16 0.20 0.76 p -value of equality 0.359 0.526 0.228 0.547 Sri Lankans in Australia Treatment group 0.45 0.45 0.44 0.79 Control group 0.29 0.30 0.32 0.72 p -value of equality 0.017 0.025 0.073 0.194 46  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES discussing financial matters by some of the sample, which the invitation to the training may have exacerbated. Sri Lankans with lower education levels and with lower baseline financial literacy were more likely to attrit, although this differen- tial attrition did not vary with treatment status. Despite the use of door prizes and gifts to their community groups, the six-month attrition was high in the East Asian and Sri Lankan groups since these surveys were done in community events, which had very low attendance for these groups.7 We therefore do not use the six-month survey data for these two samples. Given these attrition rates, we view the evidence as most reliable for the Pacific Island group and for the short-term outcomes for the East Asians. 1.3 RESULTS To estimate the impact of financial literacy training on different outcomes of interest, we estimate the following equation: s Outcomei,t = a + b * Treati + c * Outcomei,0 + ∑s=1 δsdi,s + εi,s where Treati is a dummy variable indicating assignment to treatment, and we control for the lagged outcome variable where possible (McKenzie 2012) and dummy variables, di,s for randomization strata (Bruhn and McKenzie 2009) in order to maximize power. Robust (white-corrected) standard errors are reported in parentheses under the coefficients in the tables. 1.3.1 Impact on financial knowledge Table 1.3 examines whether the financial literacy training succeeds in increasing the knowledge migrants have about the costs of remitting and of using credit. We see large short-term impacts on financial knowledge for the Pacific Island sample—they are 16  percentage points more likely to know it is cheaper to bundle remittances into a larger transaction, 52  percentage points more likely to know the ATM/prepaid debit card is the cheapest method of remitting among the options asked about, 29  percentage points more likely to know that only paying the minimum on a credit card is more expensive than paying more than the minimum, and 29 percentage points more likely to know that payday loans are more expensive than credit cards or hire purchase. All of these impacts are signif- icant at the 1 percent level; as a result, the average knowledge score, which is a mean of these four questions, also shows a positive and significant effect. There 7  The average value of gifts given to individuals (or the groups they represented) as a thank you for being involved and as incentives such as door prizes was US$40 per participant. This incentive design plus all aspects of the study had received prior ethical approval from the Waikato Management School human ethics committee. 1.  The impact of financial literacy training for migrants  ◾ 47 TABLE 1.3  Impact on financial knowledge EXCHANGE RATE APR CORRECTLY KNOWS PAYDAY CREDIT CARD IS KNOWS ABOUT UNDERSTANDS REMITTANCES MINIMUM ON COMMISSION PAYING ONLY CALCULATES KNOWLEDGE COMPOUND IS CHEAPER TO BUNDLE EXPENSIVE EXPENSIVE CHEAPEST KNOWS IT INTEREST AVERAGE METHOD LOAN IS KNOWS KNOWS SCORE TIME AFTER INTERVENTION 1 MO 6 MO 1 MO 1 MO 6 MO 1 MO 6 MO 1 MO 6 MO 6 MO 6 MO Panel A: Pacific Island migrants in New Zealand Assigned to 0.160*** 0.0682 0.532*** 0.289*** 0.0503 0.289*** 0.159*** 0.320*** 0.214*** 0.0114 −0.0241 treatment (0.0487) (0.0484) (0.0440) (0.0531) (0.0508) (0.053) (0.0567) (0.0377) (0.0546) (0.0277) (0.0332) Observations 328 302 323 329 299 329 300 330 296 299 301 Control 0.53 0.54 0.13 0.31 0.32 0.31 0.35 0.32 0.25 0.06 0.11 group mean Panel B: Asian migrants in New Zealand Assigned to 0.125*** — 0.0988** −0.00592 — −0.056 — 0.051* — — — treatment (0.0477) (0.0401) (0.0438) (0.057) (0.0268) Observations 321 — 308 304 — 288 — 323 — — — Control 0.68 — 0.10 0.19 — 0.65 — 0.42 — — — group mean Panel C: Sri Lankan migrants in Australia Assigned to −0.00802 — 0.184*** — — — — — — — — treatment (0.0522) (0.066) Observations 131 — 131.00 — — — — — — — — Control 0.86111 — 0.67 — — — — — — — — group mean Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. All regressions include controls for strata dummies, and for baseline outcome where available. Six-month follow-up results only shown for the Pacific Island sample due to extreme attrition in other samples. Sri Lankan one-month fol- low-up survey did not ask questions on credit card or payday loan knowledge. is some suggestion of a decline in this knowledge on several questions at the six-month follow-up survey, but even at six months migrants who were assigned to training are more likely to understand the exchange rate commission and to know that payday loans are more expensive than credit cards or hire purchase. Consistent with other studies of financial literacy (Carpena et al. 2011), we find no impact on computational measures of financial literacy such as the ability to correctly calculate the APR on a payday loan or to understand compound interest (which was not taught in the course). We also see some increases in financial knowledge about remittances for the other two groups: East Asian migrants are 12 percentage points more likely to know that it is cheaper to bundle remittances into a larger transaction, and 10  percentage points more likely to know the cheapest method for remitting. Sri Lankan migrants saw an increase in knowledge of the cheapest method, but no increase (from a high base) in knowledge on remittance bundling. The Sri 48  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Lankan results should be caveated by the high and differential attrition rates. Recall that credit issues were not covered in the financial literacy training for the other two groups. Consistent with this, we find no increase in knowledge about credit among the East Asians (and the questions were not asked in the Sri Lankan sample). 1.3.2 Impact on information seeking and budgeting Our follow-up surveys asked each month whether respondents had used any source of information to compare the costs of remitting across methods or prod- ucts, and if so, what information source they had used. Since our intervention focused on the use of several Internet comparison sites, we are particularly inter- ested in whether participants use the Internet more to compare remitting costs as a result of the intervention. Finally, our surveys at one month and six months also asked whether individuals always keep track of how much they spend each month. The financial literacy course made no mention of doing this, and we would thus not expect to see effects. It therefore serves as a check on reporting bias, to ensure that individuals who attended training are not just reporting that they do more of every perceived desirable financial behavior. Table 1.4 examines whether the increase in knowledge about the costs of remitting leads to changes in behavior. We see that in the short term, assignment to training leads to the Pacific Island and East Asian migrants being more likely to use information to compare remitting costs and to be more likely to use the Internet to compare costs, with no impact on keeping track of monthly expenses. However, the strongest impacts are found in the month right after the training, with no impacts on the use of the Internet at three or six months. The impacts are positive, but not statistically significant for the Sri Lankan sample, which may reflect imprecision due to the smaller sample size, or the impacts of differential attrition by treatment status, or perhaps that the effects are weaker for this group since the majority already used one of the cheapest methods anyway. 1.3.3 Impact on remitting frequency and amount Next we examine whether the financial literacy training had any impact on either the likelihood of sending remittances or on the total amount remitted. Ex ante it is not clear what direction we should expect the effect of financial literacy training to be—the content on bundling transactions together into a few, less frequent, larger transactions would be expected to reduce the frequency of remitting while having no impact on the total amount sent; whereas content that stresses cheaper methods of remitting may lead individuals to be more willing to make smaller transactions and therefore increase frequency and potentially also lead to individuals sending more remittances (Aycinena, Martínez, and Yang 2010; Gibson, McKenzie, and Rohorua 2006). 1.  The impact of financial literacy training for migrants  ◾ 49 TABLE 1.4  Impact on financial knowledge KEEPS TRACK USED INFORMATION TO COMPARE USED INTERNET TO COMPARE REMITTING OF MONTHLY REMITTING COSTS COSTS SPENDING 1 MO 2 MO 3 MO 6 MO AVG 1 MO 2 MO 3 MO 6 MO AVG 1 MO 6 MO Panel A: Pacific Island migrants in New Zealand Assigned to 0.206*** 0.0289 −0.0576 0.0975* 0.0720** 0.188*** 0.155*** −0.00212 −0.0350 0.0822** −0.0466 0.000522 treatment (0.0428) (0.0490) (0.0460) (0.0513) (0.0357) (0.0430) (0.0501) (0.0536) (0.0484) (0.0365) (0.0414) (0.0478) Observ. 329 318 316 302 332 330 321 317 302 332 308 283 Control 0.60 0.68 0.80 0.58 0.66 0.20 0.31 0.40 0.31 0.30 0.61 0.56 group mean Panel B: Asian migrants in New Zealand Assigned to 0.0977** 0.0372 0.0827* — 0.104*** 0.0396 0.0515 0.0269 — 0.0582** −0.0915 — treatment (0.0480) (0.0474) (0.0467) — (0.0350) (0.0442) (0.0325) (0.0330) — (0.0255) (0.0564) — Observ. 318 291 272 — 320 321 293 274 — 323 282 — Control 0.23 0.20 0.14 — 0.18 0.18 0.06 0.06 — 0.10 0.63 — group mean Panel C: Sri Lankan migrants in Australia Assigned to 0.0528 0.0387 0.0137 — 0.0344 −­0.0119 −0.0385 −0.0105 — −0.00400 — — treatment (0.0796) (0.0757) (0.0690) (0.0583) (0.0529) (0.0407) (0.0368) — (0.0331) — — Observ. 129 128 127 — 130 131 130 129 — 132 — — Control 0.24 0.20 0.16 — 0.20 0.10 0.07 0.04 — 0.07 — — group mean Note: See table 1.3. Table 1.5 shows that the financial literacy training had no significant impacts on either the likelihood of remitting or on the total amounts remitted for any of the three groups. Moreover, in some cases, these are relatively precise zero effects. For example, averaging observations over all four follow-up surveys for the Pacific Island sample, the point estimate on the monthly frequency of remit- ting is −0.02, with a 95 percent confidence interval of (−0.077, +0.036). That is, we can rule out that the training had large positive effects on remitting frequency, and also rule out large negative impacts. Remittance amounts have more vari- ation, but even so, we have a point estimate for the Pacific Island sample of an increase of $NZ 4, with a 95 percent confidence interval of (−$NZ 63, +$NZ 71) for total remittances over six months—which is small both in absolute terms and when compared to the median annual income of $NZ 20,000–30,000 for Pacific Islanders in our sample. 1.3.4 Impact on remitting channel Even if immigrants do not change their frequency of remitting or the amount remitted, they may still benefit from the training if it causes them to change the method they use to remit. In table 1.6, we examine whether migrants in 50  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 1.5  Impact on remittance outcomes MADE A REMITTANCE IN PAST MONTH TOTAL AMOUNT REMITTED (UNCONDITIONAL) 1 MO 2 MO 3 MO 6 MO AVG 1 MO 2 MO 3 MO 6 MO AVG Panel A: Pacific Island migrants in New Zealand Assigned to −0.0110 −0.0585 −0.0550 0.0435 −0.0210 12.33 −2.839 −0.903 −6.185 3.946 treatment (0.0383) (0.0377) (0.0350) (0.0443) (0.0288) (9.534) (6.038) (5.549) (16.89) (34.09) Observations 328 316 317 299 332 321 310 308 292 278 Control 0.16 0.17 0.14 0.18 0.16 20 22 16 53 115 group mean Panel B: Asian migrants in New Zealand Assigned to 0.00224 0.0104 −0.0143 0.00127 −172.7 292.2 −172.1 81.21 — — treatment (0.0309) (0.0230) (0.0185) (0.0182) (160.9) (224.6) (143.5) (314.4) Observations 321 293 274 — 323 316 290 271 — 269 Control 0.088 0.041 0.029 — 0.054 352 62 193 — 523 group mean Panel C: Sri Lankan Migrants in Australia Assigned to −0.0352 0.0949 −0.0764 −0.00451 −456.7 −163.2 −60.10 −469.1 — — treatment (0.0777) (0.0796) (0.0782) (0.0610) (329.0) (135.1) (51.23) (371.1) Observations 131 130 129 — 132 130 129 128 — 126 Control 0.361 0.282 0.319 — 0.326 630 361 144 — 868 group mean Note: See table 1.3. our samples use a different method of remitting in any of the follow-up surveys compared to that which they had used in the 12 months prior to the baseline survey. We see that in the Pacific Island sample, 16 percent of the control group use a different method at least once during the four follow-up surveys, and finan- cial literacy training leads to a significant reduction of this, halving the rate of switching to new products. Recall that the training introduced these migrants to a new product, the Westpac Express prepaid card, but we see no increase in usage of this among those treated. Instead, the main effect appears to be fewer individuals switching from Western Union or Melie Mei Langi to use either Mana or Epokifo’ou, two less used methods. Mana is a church-based money transfer method with a low fixed fee ($NZ 5–8) but an unfavorable exchange rate, so that for the median remittance it is the most expensive of the money transfer opera- tors. Moreover, Mana is not included on the sendmoneypacific website. Epokifo’ou is a money transfer organization that had very similar costs to Melie Mei Langi for much of our sample period, and that was slightly cheaper than Western Union. So the main effect seems to be to stop people from switching to methods that provide little or no benefit to switching, or for which it is less easy to track costs. We find no significant impact on switching methods for the other two groups. For the East Asians, this in part reflects the low frequency of remitting overall, meaning that there are a small number of transactions to look at switching over. For the Sri Lankans, most immigrants were using a relatively cheap method at 1.  The impact of financial literacy training for migrants  ◾ 51 TABLE 1.6  Impact on the likelihood of switching remittance methods EVER USE EVER USE EVER MANA/ WESTPAC 1 MO 2 MO 3 MO 6 MO SWITCH EPOKIFO’OU CARD Panel A: Pacific Island migrants in New Zealand Assigned to −0.0380 −0.0645** −0.0675** −0.0118 −0.0865** −0.0499** 0.00598 treatment (0.0264) (0.0280) (0.0271) (0.0200) (0.0340) (0.0245) (0.0161) Observations 329 318 316 299 332 332 332 Control group mean 0.08 0.10 0.10 0.04 0.16 0.079 0.018 Panel B: Asian migrants in New Zealand Assigned to −0.0123 −0.00563 0.000445 — −0.0240 — — treatment (0.0154) (0.0110) (0.0114) (0.0195) Observations 321 293 274 — 323 — — Control group mean 0.03 0.01 0.01 — 0.04 — — Panel C: Sri Lankan migrants in Australia Assigned to 0.00714 0.0849 −0.00737 — 0.0997 — — treatment (0.0494) (0.0589) (0.0502) (0.0711) Observations 131 130 129 — 132 — — Control group mean 0.09722 0.070422 0.0869 — 0.15277 — — Note: See table 1.3. baseline, implying relatively little to be gained from switching (and also we have the caveat of high and unbalanced attrition). 1.3.5 Impacts on credit Finally, we examine whether the credit portion of the financial literacy training led to individuals being more or less likely to have certain forms of credit in our final survey round. We restrict our analysis here to the Pacific Island sample, since the other two groups did not have credit covered in their financial literacy training, and because these other two groups had such high attrition in the final round survey. Table 1.7 shows the credit impacts of the financial literacy training. We do not see any significant impacts on individuals having checking accounts, savings accounts, or ATM cards. Despite the emphasis in training on the Westpac prepaid card being the cheapest method to send, and credit cards being a cheaper source of financing, we do not see significant increases in usage of either. Despite (or perhaps as a result of) media coverage of the high cost of payday loans (e.g., Waikato Times 2012), we find no one in our sample who said they had applied for a payday loan. This may be genuine, or reflect reluctance to admit use of this form of credit (Karlan and Zinman 2008). What we do see are positive and significant impacts on obtaining hire purchase loans and on other loans. Qualitative discus- sions reveal that the hire purchase loans were the result of special promotions 52  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 1.7  Impact on Pacific Island migrants having different financial products six months after treatment CHECKING SAVINGS ATM WESTPAC CREDIT HIRE PUR- PAYDAY OTHER ACCOUNT ACCOUNT CARD CARD CARD CHASE LOAN LOAN LOAN Assigned to −0.0273 0.0674 0.000609 0.0179 −0.0336 0.0867*** 0 0.168*** treatment (0.0551) (0.0469) (0.00894) (0.0189) (0.0413) (0.0315) (0.0376) Observations 317 317 317 317 317 317 317 317 Control group 0.397 0.737 0.994 0.019 0.179 0.045 0.000 0.051 mean Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. All regressions include controls for strata dummies. before Christmas, in which special “zero interest rate” deals were coupled with free Christmas hams, possibly making these loans a good deal. The main other loans were extended family ROSCAs, which respondents said they had set up to avoid the high costs of payday loans. Taken together, this suggests some potential positive impact of the training, although since we do not measure any reduction in use of other, more expensive forms of credit, this use of cheaper forms of credit may lead to an increase in credit usage—which is of uncertain overall welfare effect. 1.3.6 What do participants say the main benefits have been? Individuals who attended training were asked in the one-month follow-up whether it had been useful, whether they would recommend it to friends and family, what the most useful topic was, and whether they had changed behavior as a result. Among the Pacific Island group, all participants said it was very useful and that they would recommend it to others. The topic rated most useful was on the different costs and methods of sending money, with a number also mentioning information on credit cards as useful. Sixty percent of respondents said they had changed behavior as a result, with the main change being the use of the website to compare costs, and asking around for better rates. For the East Asians, 80  percent of those attending said it was useful and 75  percent would recommend it to others, with the most useful knowledge being about remit- tance fees. Only 21 percent said they had changed behavior, mostly in terms of examining the costs of sending money. Among the Sri Lankan sample, 91 percent of those answering the follow-up survey said the training had been useful, but only 2 people in the Sri Lankan sample who had attended training said they had changed their behavior as a result. These direct reports are therefore consis- tent with the empirical results, suggesting that there was a knowledge increase, and some changes in information-seeking behavior for the Pacific Island group in particular, but no big changes in remitting behavior. 1.  The impact of financial literacy training for migrants  ◾ 53 1.4 DISCUSSION The training did succeed in increasing financial knowledge about the components of remittance costs and in getting people to search for more information about the costs of sending money. It was fairly cheap to deliver—courses were typi- cally taught in churches or other community spaces; and once the content was developed, the main costs were the time of the trainer, and snacks and refresh- ments for those attending—for an approximate cost of $NZ 20–30 per attendee. Given these low costs, the benefits observed in knowledge and behavior may be enough to justify providing this course. Nonetheless, despite the emphasis on the Westpac prepaid debit card as the lowest cost of remitting, and credit cards as a low cost of obtaining credit for the Pacific Island sample, we do not see any impact on these outcomes. The final round survey asked people why they were not using the cheapest method of remitting; 41 percent replied that another method was more convenient for them, and 55 percent that another method was more convenient for the receiver. The latter is consistent with Gibson et al. (2007), who show the geographic spread of ATM facilities in Tonga covers a lower share of the population than Western Union offices. Further evidence of the lack of take-up of low-cost technologies in the Pacific Islands comes from two other methods: the Internet-only money transfer operator KlickEx had the lowest costs overall but was unused by participants throughout our study. In October 2011, KlickEx linked with a major mobile phone provider, Digicel, to offer remittance transfers into a mobile wallet in Tonga, Samoa, or Fiji which could then be withdrawn as cash, again at very low transac- tion costs. To date, this new method seems to have very low take-up. One reason that convenience may win out is that the amounts saved through better financial literacy may just be too trivial to warrant action. This is especially the case for the East Asians (who tend not to remit that much, and when they do, remit large amounts), and the Sri Lankans (who were using cheap methods to start with). But even for the Pacific Island sample, the savings from switching to one of these cheapest methods might amount to $NZ 6–12 per $NZ 200 trans- action, which may only merit the costs of learning a new method and trying it out if people remit very frequently. Given that the baseline survey revealed an average frequency of remitting for the Pacific Island sample of five times per year, the annual saving from switching to the cheapest methods is just $NZ 30–60. The scope for changes in ultimate outcomes may be greater for financial literacy transactions that focus more on saving and budgeting behavior, or those that allow people at risk of obtaining expensive credit to avoid very expensive loans (e.g., Bertrand and Morse 2011). 54  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 1.5 CONCLUSIONS Our results show that simple financial education training for migrants can change their knowledge about the costs of remitting and lead them to look around more regarding prices. There is also some suggestive evidence from the Pacific Island sample that coupling this with information on different sources of credit may help migrants avoid some of the most expensive forms of credit. Nevertheless, we find no big changes in ultimate outcomes—migrants avoid switching to more expen- sive or less transparent remittance channels, but do not change the amount or frequency of remitting. Thus, despite simply informing remitters about remittance costs being a relatively cheap and uncontroversial intervention, it will not necessarily lower average costs from remitters switching to cheaper methods. Instead, govern- ments targeting reduced average money transfer costs may need to address other barriers, which may include excessive regulation and exclusive arrangements made by state-owned entities that deter new entry into remittance corridors, and financial access on the receiving country side. It is also possible that the mere process of providing transparent information on the costs of remitting by different methods will lower costs through competition, without migrants needing to switch providers. For example, transfer fees from New Zealand to the Pacific Islands have fallen since the launch of sendmoneypacific—although how much can be attribut- able to the website rather than other market events is an open research question. As far as financial education itself, though, a further implication of this work is that the case for providing financial literacy training for migrants needs to rest on other criteria than the financial savings from cheaper remittances, such as the improvements in their capabilities from being more informed customers, and the potential savings from other aspects of financial management such as choice of debt levels and instruments. Experimenting further with additional content on budgeting, saving, and debt management seems a fruitful area for policy refine- ment. REFERENCES Aycinena, Diego, Claudia Martínez, and Dean Yang. 2010. “The Impact of Remittance Fees on Remittance Flows: Evidence from a Field Experiment among Salvadoran Migrants.” http://www.researchgate.net/publication/43015233_The_Impact_of_Remittance_ Fees _on _ Remit t ance_ Flows _ Evidence_from _ a _ Field _ E xperiment _ Among _ Salvadoran_Migrants. Bertrand, Marianne, and Adair Morse. 2011. “Information Disclosure, Cognitive Biases, and Payday Borrowing.” Journal of Finance 66 (6): 1865–93. Bruhn, Miriam, and David McKenzie. 2009. “In Pursuit of Balance: Randomization in Practice in Development Field Experiments.” American Economic Journal: Applied Economics 1 (4): 200–232. 1.  The impact of financial literacy training for migrants  ◾ 55 Carpena, Fenella, Shawn Cole, Jeremy Shapiro, and Bilal Zia. 2011. ”Unpacking the Causal Chain of Financial Literacy.” Policy Research Working Paper 5798, World Bank, Washington, DC. Clemens, Michael A., Claudio E. Montenegro, and Lant Pritchett. 2008. “The Place Premium: Wage Differences for Identical Workers across the U.S. Border.” Working Paper 148, Center for Global Development, Washington, DC. Gibson, John, Geua Boe-Gibson, David J. McKenzie, and Halahingano Rohorua. 2007. “Efficient Remittance Services for Development in the Pacific.” Asia-Pacific Development Journal 14 (2): 55–74. Gibson, John, David J. McKenzie, and Halahingano Rohorua. 2006. “How Cost Elastic Are Remittances? Evidence from Tongan Migrants in New Zealand.” Pacific Economic Bulletin 21 (1): 112–28. Gibson, John, David McKenzie, and Steven Stillman. 2013. “Accounting for Selectivity and Duration-Dependent Heterogeneity When Estimating the Impact of Emigration on Incomes and Poverty in Sending Areas.” Economic Development and Cultural Change 61 (2): 247–80. Hernández-Coss, Raúl. 2005. “The U.S.-Mexico Remittance Corridor: Lessons on Shifting from Informal to Formal Transfer Systems.” Working Paper No. 47, World Bank, Washington, DC. Karlan, Dean, and Jonathan Zinman. 2008. “Lying about Borrowing.” Journal of the European Economic Association 6 (2–3): 510–21. Lee, Helen. 2003. Tongans Overseas: Between Two Shores. Honolulu: University of Hawaii Press. LIRS (Lutheran Immigration and Refugee Service). 2006. Financial Literacy for Newcomers: Weaving Immigrant Needs into Financial Education. http://archive.lirs.org/InfoRes/ PDFs/FinancialLiteracy.pdf. McKenzie, David. 2007. “Remittances in the Pacific.” In Immigrants and Their International Money Flows, edited by Susan Pozo, 99–121. W.  E. Upjohn Institute for Employment Research: Kalamazoo, MI. —. 2012. “Beyond Baseline and Follow-up: The Case for More T in Experiments.” Journal of Development Economics 99 (2): 210–21. McKenzie, David, John Gibson, and Steven Stillman. 2010. “How Important Is Selection? Experimental vs Non-experimental Measures of the Income Gains from Migration.” Journal of the European Economic Association 8 (4): 913–45. McKenzie, David, and Johan Mistiaen. 2009. “Surveying Migrant Households: A Comparison of Census-Based, Snowball, and Intercept Surveys.” Journal of the Royal Statistical Society Series A 172 (2): 339–60. Pacific Islands Forum Secretariat. 2011. “Regional Remittance Issues.” Paper 4 for the Forum Economic Minister’s Meeting in Apia, Samoa, July 20–21. http://www.forumsec. org/resources/uploads/attachments/documents/2011FEMM_FEMS.06.pdf. Stock, Rob. 2009. “Remittance Card Shows Bank’s Other Face.” Sunday Star Times, August 9, A6. Waikato Times. 2012. “Payday Loans Cost Kiwis Big Premium.” March 5. http://www.stuff. co.nz/waikato-times/news/6523093/Payday-loans-cost-Kiwis-big-premium. World Bank. 2006. Global Economic Prospects 2006: Economic Implications of Remittances and Migration. World Bank: Washington, DC. Yang, Dean. 2008. “International Migration, Remittances, and Household Investment: Evidence from Philippine Migrants’ Exchange Rate Shocks.” Economic Journal 118: 591–630. Yang, Dean, and Claudia A. Martínez. 2006. “Remittances and Poverty in Migrants’ Home Areas: Evidence from the Philippines.” In International Migration, Remittances, and the Brain Drain, edited by Çaglar Özden and Maurice Schiff, 81–121. Washington, DC: World Bank and Palgrave Macmillan. 56  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES ANNEX:  BASELINE FINANCIAL LITERACY QUESTIONS (PACIFIC ISLAND VERSION) Remittance-specific measures In your opinion, which of the following methods is the cheapest way of sending $NZ 200 to someone in the Pacific Islands? a. Western Union b. Bank transfer through ANZ or Westpac c. Melie Mei Langi or Samoa Money Transfer d. ATM card or Visa prepaid card e. Other (specify) When money is sent by someone in New Zealand to people in the Pacific Islands, what are the various costs that the bank or money transfer operator charges? (tick all that apply) a. A fixed fee imposed on the sender b. A fixed fee imposed on the recipient c. An exchange rate commission d. All of the above e. None of the above f. Don’t know Suppose you want to send $NZ 200 to someone in the Pacific Islands. Which would cost you more, sending it all at once as $200, or sending it at two different times of $100 each time, or is the cost the same either way? a. Cheapest to send $200 all at once b. Cheapest to send $100 two times c. The same cost either way d. I don’t know General financial literacy In your opinion, which one of the following credit card users is likely to pay the GREATEST dollar amount in finance charges per year, if they all charge the same amount per year on their cards? a. Semisi, who pays at least the minimum amount each month, and more when he has the money b. Samisoni, who only pays the minimum amount each month 1.  The impact of financial literacy training for migrants  ◾ 57 c. Sione, who always pays off his credit card in full shortly after he receives it d. Tevita, who generally pays off his credit card in full, but occasionally will pay the minimum when he is short of cash. A consumer takes out a payday loan for $100 which has a $15 fee. After 2 weeks, the consumer pays back the full $115. What do you think is the annual percentage rate (APR) charged on this loan? a. 15 percent b. 115 percent c. 315 percent d. 390 percent CHAPTER 2 F inancial education and behavior formation Large-scale experimental evidence from Brazil MIRIAM BRUHN, LUCIANA DE SOUZA LEÃO, ARIANNA LEGOVINI, ROGELIO MARCHETTI, AND BILAL ZIA ABSTRACT This chapter demonstrates that a high-quality financial education program targeted at youth can improve financial knowledge, attitudes, and behavior. We collaborate with a national partnership of major financial and educa- tional actors in Brazil to conduct a large-scale randomized evaluation of a comprehensive financial education program for high school students. The study spans 6 states, 868 schools, and approximately 20,000 students aged 15–17. The program increases student financial knowledge by a quarter of a standard deviation and shifts the distribution of financial proficiency scores rightward. The change in knowledge leads to a 1.4 percentage point increase in savings—a large and economically significant effect. A complementary workshop for parents induces their children to save even more. Both current attitudes and new forward-looking indexes of intentions to save and financial We are deeply indebted to all in-country partners without whom the project would not have been possible, including the Center for Public Policy and Educational Eval- uation of the Federal University of Juiz de Fora; the CVM; the Central Bank of Brazil; the Complementary Pensions Secretary and the Superintendent of Private Insurance; BM&FBOVESPA; ANBIMA; FEBRABAN; Instituto Unibanco; the Ministry of Education; and the state education departments of Rio de Janeiro, São Paulo, the Federal District, Ceará, Tocantìns, and Minas Gerais. Special thanks to José Alexandre Vasco of CVM for leading the design and implementation of the financial literacy program, and to Alzira Silva and Célia Bittencourt for their outstanding dedication and support. We also thank Christian Salas for research assistance. Finally, the authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 59 60  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES autonomy improve significantly. “Trickle-up” impacts on parents are also significant, with improvements in parent financial knowledge, savings, and spending behavior. The scale and geographical scope lend strong external validity to the findings. 2.1 INTRODUCTION Navigating today’s financial markets can be a difficult task. Financial systems have grown in complexity and sophistication, often outpacing the capacity of individ- uals and families to make informed financial choices (Lusardi and Mitchell 2007; Lusardi, Mitchell, and Curto 2010). Personal financial decisions are further compli- cated by the rapid infusion of cleverly marketed consumer products that are often coupled with expensive credit and installment plan offers. These advancements have heightened the risk of misplaced or misinformed spending decisions that can lead to real and long-lasting financial burdens on household budgets. Indeed, the personal bankruptcy rates even in developed countries such as the United States have skyrocketed, increasing by as much as 20  percent annually during the recent financial crisis (Administrative Office of the United States Courts 2012). The policy response to these troubling trends has been to introduce proac- tive measures on both the supply and demand sides. On the supply side, many countries now have consumer protection bureaus that are tasked with ensuring financial providers adopt and adhere to transparent consumer disclosure laws. On the demand side, there has been a significant push to educate the public on financial matters through various financial literacy programs, though rigorous evidence on their effectiveness remains scarce (Cole, Sampson, and Zia 2011). This chapter makes a significant contribution to the literature by experimen- tally testing a comprehensive approach to financial education for youth in Brazil. We collaborate with a national partnership of major financial and educational actors in Brazil and study the causal impact of their state-of-the-art financial education program for high school students. We measure changes in financial knowledge, preferences, and attitudes toward current and future financial deci- sions, and also saving and spending behavior. We focus our study on youth for a number of important reasons. First, good financial habits formed at an early age are likely to benefit schooling, employ- ment, and standards of living throughout adulthood. Second, well-informed students have the opportunity to affect not only their own financial choices, but also to act as agents of change in their households’ financial decisions. Third, behavioral evidence suggests that individuals display anomalies in their prefer- ences and that these anomalies may be related to variations in cognitive ability. In a study of Chilean high school students, Benjamin, Brown, and Shapiro (2012) find evidence that small-stakes risk aversion and short-run discounting are less common among those with higher standardized test scores. This raises the question of whether improving knowledge and understanding of intertemporal 2.  Financial education and behavior formation   ◾ 61 choices for youth can lower the demands on cognitive resources and improve current and future financial decisions and attitudes. Further, the focus on youth leverages their learning capacity as students who are primed to absorb, recall, and apply learning in school. Finally, despite the potential benefits of financial education for youth, we simply do not know what works; and the few existing studies either show conflicting results, are too narrowly focused, or suffer from important identification concerns.1 To date, our study is the largest randomized evaluation in the financial educa- tion literature, with a sample of six Brazilian states—São Paulo, Rio de Janeiro, Ceará, Tocantìns, Minas Gerais, and the Federal District; 868 public high schools; and approximately 20,000 students. As part of the study design, schools were first stratified by state, pair-wise matched by school and community character- istics, and then randomly assigned to treatment and control groups. The study spanned three academic semesters, from August 2010 to December 2011, and data were collected in three rounds consisting of baseline (August 2010), follow-up 1 (December 2010), and follow-up 2 (December 2011). The financial education program included study materials, teacher training, monitoring, and participation awards. Unlike typical financial education programs that involve external instructors providing one-off classes on financial education, our program spanned 18 months; was delivered by regular teachers; and was integrated in classroom mathematics, science, history, and Portuguese curricula. The instruction used new textbooks with interactive classroom exercises on financial education themes, take-home exercises such as creating household budgets with parents, and role-playing assignments such as visiting local markets and organizing class graduation parties. The curriculum was complemented by extensive teacher training, web learning tools, and instructor handbooks. The textbooks and implementation strategy were developed in collaboration with representatives of the financial and education sectors. Finally, schools with high levels of survey participation received awards and public recognition. As such, the program’s intensity of treatment was much stronger than for other financial education interventions. The results show significant improvements in financial knowledge, attitudes, and behavior. Specifically, we find a quarter of a standard deviation increase in the financial knowledge of students, as measured by an SAT-like financial profi- ciency test. Importantly, the entire distribution of scores shifts to the right with students at all levels of capability showing marked improvements in testing. The knowledge gains help students improve their current financial behavior, with a statistically significant increase of 1.4  percentage points in savings and signifi- cant improvements in the likelihood of making budgets and negotiating prices and payment methods. 1  See section 2.2 of this chapter for a detailed discussion of the existing literature. 62  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES While these changes in current behavior are important, as many students in our sample have part-time jobs and hence the option to save, of equal importance are their preferences and attitudes toward future financial decisions. To better understand this potential, a local education survey firm specializing in school-based evaluations helped develop and test two new forward-looking indicators of finan- cial behavior in our sample—student financial autonomy and intention to save. The financial autonomy index aggregates a series of questions designed to measure whether students feel empowered, confident, and capable of making independent financial decisions and influencing the financial decisions of their households. The intention to save index includes a series of questions that identify preferences over future and hypothetical savings and spending scenarios. The analysis finds strong and statistically significant treatment effects on both these measures, with an effect size ranging between 0.08 and 0.12 of a standard deviation. Next, we turn to the household and investigate treatment effects on parents of high school students in two distinct ways. First, we measure “trickle-up” impacts of our financial education treatment for students. Many of the take-home exercises involved interaction with parents, such as making household budgets or researching and comparing interest rates on savings accounts. We survey parents and identify several important findings. As proof of concept, we find that parents in treatment schools are significantly more likely to report that their children discuss financial matters with them at home and that they volunteer to help orga- nize household budgets. In addition, we detect improvements in parental finan- cial knowledge on standard financial literacy questions used in the literature. And finally, we find significant improvements in parental financial behaviors, with an increase of 0.67 percentage points in savings and improvements in the likelihood of keeping household budgets. We also study household-level effects through a standard adult workshop on financial education for parents. This involved a DVD-based intervention where parents in treated schools were randomly assigned to either a financial educa- tion screening or a health education screening. Although the attendance at these workshops was low, we detect further improvements in the savings rates of students from families that attended the financial education workshops. Hence, these workshops helped parents reinforce the messages taught to the students. Overall, our study shows that financial education can be an effective tool in improving financial outcomes of students when delivered in a comprehensive manner and over a significant period of time. Also, key complementary benefits can be derived by involving the entire household—students and parents—as indi- cated by our trickle-up and parents’ workshop impacts. Finally, it is important to distinguish our study from others in the literature based on scope and scale. Specifically, the large sample size and breadth of coverage through six of the most populated states in Brazil provides strong support for the external validity of our findings and the generalizability of the results. Moreover, our careful eval- uation has led to significant and real policy impact as the Ministry of Education 2.  Financial education and behavior formation   ◾ 63 in Brazil has recently agreed to expand the financial education program to 3,000 new schools across the country after reviewing evidence from our study. Other countries in the region have approached us and our country partners so that they can learn and adopt a similar program in their respective education systems. This chapter proceeds as follows. Section 2.2 summarizes the literature on financial education for youth and details the Brazilian context. Section  2.3 discusses the financial education curriculum, and section 2.4 presents the research and sampling methodology as well as the study timeline. Section 2.5 describes the data collection, and section 2.6 presents summary statistics and analysis on process evaluation. Section 2.7 discusses the main results, section 2.8 discusses the impact of our study, and section 2.9 concludes. 2.2 BACKGROUND AND CONTEXT 2.2.1 Financial education for youth There is a growing literature on financial education and its determinants for youth, but much of the evidence comes from developed rather than developing coun- tries. A consistent finding among these studies is that financial literacy tends to peak among adults in the middle of the life cycle, and is significantly lower among youth. In the United States, for example, those in the prime age group (25–65) tend to perform about 5 percent better on financial literacy questions than those under 25 (Lusardi and Mitchell 2011). Strikingly, Lusardi, Mitchell, and Curto (2009) find that less than a third of American teenagers (ages 12–17) possess basic knowl- edge of interest rates, inflation, and risk diversification. Mandell (2006) notes that there is even evidence that youth financial literacy has been declining in the United States since the late 1990s. Similar evidence comes from Australia, where Beal and Delpachitra (2003) identify low levels of financial literacy among youth. Perhaps in response to such trends, policy makers around the world have made financial education for youth a priority, with many school-based initiatives now part of education reform. Yet the impacts of such school programs on finan- cial knowledge and behavior are still not well understood and even contradictory. For example, Bernheim, Garrett, and Maki (2001) employ a difference-in-difference approach to analyze the impact of state high school financial education mandates on savings behavior in the United States; they find that mandates appear to effec- tively increase exposure to financial education and have a significant subsequent effect on future saving. However, Cole and Shastry (2009) replicate and extend the analysis using a much larger sample from U.S. census data and find, in contrast, no significant impact of high school financial education on future saving. Other studies have also shown mixed results. Carlin and Robinson (2010) study the effects of a financial literacy course for high school students in the United States that involved role-playing fictitious budget situations following 19 hours 64  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES of financial training that included credit card management, taxes, budgeting, and simple investments (Junior Achievement Finance Park). The simple pre-post anal- ysis shows that training raised completion rates (successfully crafting a balanced budget) from 5 percent to over 50 percent, and savings increased fourfold from pretreatment levels. At the same time, however, treated students were worse at judging a health insurance plan that had higher monthly premiums but lower out-of-pocket costs. Similarly, Walstad, Rebeck, and MacDonald (2010) study the effects of a DVD-based curriculum for high school students on financial knowledge (Financing Your Future). The five video segments cover such topics as savings, money management, banking, credit and debt, and investing, and comprise six hours of instruction. They find that students who participated in the education program showed a significant gain in financial knowledge (as measured by pre- and post-test scores) compared to students in a matched control group, but they do not study subsequent behavior change. Randomized evaluations of school-based financial education programs are scarce; in fact, we are aware of only two other such studies. Berry, Karlan, and Pradhan (2012) conduct an evaluation of a program offering voluntary after- school clubs in Ghana for primary and junior high students in 135 schools over a 10-month period. The study randomly assigns a group of fifth and seventh graders to a social and financial treatment and another to basic training. It is not clear how exposed students are to financial decision making at such a young age, and eliciting reliable and consistent responses in this age group is generally difficult; hence there are some measurement concerns. The treatment is also short and participation voluntary. The findings are unsurprisingly muted—while the study identifies some effects on saving, there are no improvements in financial knowl- edge, test scores, or social and psychological measures. Another experiment is a study among 17- to 19-year-old high school students in Italy. Becchetti, Caiazza, and Coviello (2011) offer a 16-hour-long course on finance over three months and find positive effects on financial knowledge. However, a follow-up study (Becchetti and Pisani 2012) cautions that these increases may be due to students adapting to repeated financial literacy tests rather than conceptual learning. One other recent study is an evaluation of financial education in German high schools. Lührmann, Serra-Garcia, and Winter (2012) evaluate 90-minute financial education sessions delivered to 14–16-year-old students in lower-stream high schools. While their findings show significant improvements in financial knowl- edge and a hypothetical savings scenario, their follow-up period is very short, only one to three weeks after the intervention. Furthermore, schools are chosen to receive financial education sessions based on how busy teachers feel they are with students prior to the end of the academic year; this raises serious selection concerns since teachers in control schools may also be dealing with relatively poorly performing students or other unobserved underlying student, class, or school characteristics. Finally, their sample size is extremely small—less than 2.  Financial education and behavior formation   ◾ 65 50 schools in the entire sample, some of which they had to drop. Such a small sample raises concerns about the external validity of the findings. While the lack of strong evidence on the impact of financial education for youth is striking, one key lesson that emerges from this work is that financial education tends to be more effective when it is targeted to the specific needs and desires of the audience. For instance, Varcoe et al. (2005) evaluate a program (Money Talks) in which teenagers in various settings (including juvenile halls, migrant education programs, pregnancy and parenting programs, public high schools, and youth groups) were solicited with regard to topics, format, and when and where to receive financial education information. The curriculum took the form of four newsletters targeted to 13–18-year-olds which covered different topics, including savings habits, shopping tips, car costs, and money values. The authors find positive changes in both knowledge and behavior, although self-selection and the requirement of a parent-signed permission slip raise some important identification concerns. The apparent success of the targeting strategy behind this program, however, seems to be corroborated by Mandell and Klein (2007), who highlight the importance of motivation and goal setting in increasing the relevance of financial education for youth. The importance of targeting has been borne out in studies outside the youth segment as well. For example, Duflo and Saez (2003) find positive impacts of a financial education program focused on retirement savings for university employees. Similarly, the only currently completed randomized evaluation of a financial education program for households in developing countries (Cole, Sampson, and Zia 2011) finds significantly greater effects of financial education on the likelihood of opening a bank account for household heads with low education and below-median baseline levels of financial literacy. Against the state of the existing literature on financial education, our study fills an important gap by focusing on financial education in high schools and using rigorous methodology. Specifically, our study brings together: (1) a randomized evaluation methodology, (2) a comprehensive financial education intervention that lasted three full semesters over 18 months, (3) a very large sample size (868 schools and nearly 20,000 students), (4) widespread coverage over six states in Brazil, and (5) a unique set of forward-looking outcome measures developed specifically for the youth segment of the population. As such, our study offers new insights into the impact of financial education for youth. 2.2.2 The Brazilian context Brazil’s low saving and investment rates and low levels of awareness on financial matters among the population have triggered a national policy response led by the financial sector. National savings are around 16  percent of gross domestic product (GDP) overall, and levels of financial awareness are low. A survey conducted by the Instituto Data Popular in 2008 shows that 82 percent of Brazilian 66  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES consumers are unaware of the interest rate when borrowing money. Against this background, the financial sector recently introduced the idea of financial literacy as a mechanism to improve awareness and understanding, build personal finance skills, stimulate savings, and broaden the use of financial services, with the goal of improving consumers’ ability to make decisions that are beneficial to their financial well-being. The early steps in the establishment of a national policy of financial educa- tion were taken in the mid-2000s. In 2007, the Supervisory and Regulatory Committee of Financial Systems, Capital Markets, Private Insurance, and Social Welfare (COREMEC) approved the creation of a working group to develop and propose—under the coordination of the Brazilian Securities and Exchange Commission (CVM [Comissão de Valores Mobiliários])—a National Strategy for Financial Education. This group included representatives from public financial institutions including CVM, the Central Bank, and pensions and insurance regu- latory agencies (Superintendência Nacional de Previdência Complementar and Superintendência de Seguros Privados). The group also included civil society insti- tutions and select private sector institutions such as the Brazilian Association of Financial and Capital Markets (ANBIMA [Associação Brasileira das Entidades dos Mercados Financeiro e de Capitais]), Brazilian Securities, the Brazilian Securi- ties, Commodities, and Futures Exchange (BM&FBOVESPA), and the Federation of Brazilian Banks (FEBRABAN [Federação Brasileira de Bancos]). After the working group fulfilled its purpose, a new committee, the National Committee on Financial Education (CONFEF), which includes ministers of education, justice, finance, and social security, was created to spearhead the financial education agenda. The National Strategy was launched in 2010 with the goal of fostering a culture of financial education in the country, enabling citizens to make sound financial decisions, and contributing to the efficiency of financial markets. Its scope is national, targeting children, youth, and adults; and its goals ambitious, involving a large sector of actors and a multiplicity of delivery mechanisms for financial education. In parallel with the National Strategy, a Pedagogical Support Group (GAP [Grupo de Apojo Pedagógico]) was established in 2008 to bring in the federal Ministry of Education and the municipal and state secretariats of education to work together with these financial institutions. The GAP also included the National Council of Education Secretaries (CONSED [Conselho Nacional de Secretários de Educação]) and the National Union of Municipal Directors of Education, (UNDIME [União Nacional dos Dirigentes Municipais de Educação]). The GAP’s creation was a critical step in mobilizing the development of a school curriculum for basic and high school education and planning for its introduction in public school educa- tion. The priority for the collaboration between top educators and financial sector experts was the design of the high school program, the first program that would be piloted to ascertain its value in youth’s behavior formation and their ability to serve as a vehicle for social and household change. The plan was for financial 2.  Financial education and behavior formation   ◾ 67 education to be introduced not as a separate subject, but as material teachers could use to supplement their regular school curriculum. In 2009, the World Bank was asked to evaluate the impact of the high school pilot program, with the objective of applying rigorous economic research methods to establish the causal impact of financial education on student and household financial knowledge, attitudes, and behavior. The results would be used as an independent assessment to inform improvements to, and the eventual national scale-up of, the program. CVM coordinated the pilot project, with collaboration and support from BM&FBOVESPA, ANBIMA, FEBRABAN, Instituto Unibanco, and Citi Foundation. The World Bank worked with the GAP to agree on study proto- cols, including scale of the pilot to ensure statistical power, geographical scope to ensure external validity, experimental design to ensure unbiased results, and roll-out of the pilot to ensure feasibility. The World Bank partnered with the Center for Public Policy and Education Evaluation (CAEd [Centro de Políticas Públicas e Avaliação da Educação]) for its experience and expertise in the design and implementation of student education surveys in Brazilian schools. The pilot was launched in 2010 according to the agreed protocols. 2.3 FINANCIAL EDUCATION CURRICULUM The financial education curriculum consists of didactically innovative material designed to capture the interest of young adults and be relevant to their lives. It consists of 72 case studies that can be integrated into regular school subjects such as mathematics, Portuguese, science, geography, and history. The decision to incorporate financial education into other subjects instead of forming a sepa- rate class was taken in collaboration with the federal ministry of education. In contrast to typical seminar-based financial literacy programs that are delivered in one shot and vary in length from 90 minutes to a few hours, the case study–based program in our study provides material for between 72 and 144 hours of teaching (1–2 hours per case study), spread out over one and a half school years (three semesters).2 The material is interactive and includes exer- cises that the students complete with their parents (e.g., household budgeting and planning). The program trained teachers in advance through an introductory seminar that lasted one to two days and a reference DVD, and made available a training website with distance education material which remains active throughout the program. Teacher training was also important for generating interest and support for the program, which ultimately led to a large fraction of teachers applying the financial education material in class (see section 2.6.1). The intensity and scope of The material was developed by the GAP; the complete outline of the curriculum is in 2  annex B to this chapter. 68  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES the intervention is significantly stronger than is the case in other financial educa- tion programs or evaluations. The student textbook is divided into three blocks and covers nine themes: family life, social life, personal property, work, entrepreneurship, large expen- ditures, public goods, country economy, and world economy. These themes were chosen to provide a solid foundation of knowledge on financial issues and to address the specific financial education needs of high school students. More specifically, the program aimed at providing students with tools to over- come financial challenges they were facing or were about to face, such as family finances, entry into the labor market or further studies, and preparing for financial independence. The themes are taught through 72 didactic situations that include theoretical and applied content, exercises and activities, as well as self-evaluative questions. These case studies make use of texts, stories, images, and tables to convey the material in an accessible way. They also contain “experiment” sections designed to make the material relevant to students’ daily life. For example, when discussing the international economy, students are asked to identify the imported products they use in everyday life. Each case study concludes with a short outline of what students are expected to learn. The first three themes (family life, social life, and personal property) introduce students to the concepts of budgeting, separating personal and family expendi- tures, and differentiating between fixed and variable spending. These concepts are similar to those used in standard financial education programs for adults, but adapted and simplified for the youth audience. For example, instead of making a household budget for which the student may not yet be responsible, students are instead asked to involve their parents in discussing and writing out a household budget. Not only do such exercises promote student financial education, they also promote dialogue on financial matters between students and their family financial decision makers. The work and entrepreneurship themes discuss professional growth and practical issues in starting and running a business. Students are asked to think about their professional ambitions and talents and relate them to different sorts of jobs or business opportunities. Concepts such as gross and net income, struc- tural unemployment, retirement planning, and the main aspects of the creation and execution of a good business idea are analyzed. In the large expenditures theme, students experience situations that involve large financial outlays, such as home purchases or planning a school graduation party. One of the exercises involves students in fundraising, planning, and orga- nizing an end-of-year class celebration. The public goods theme addresses several issues related to the use and financing of public goods and services, among them the payment of taxes, government efficiency, corruption, and basic economic notions of the nature of public goods. 2.  Financial education and behavior formation   ◾ 69 Finally, the country economy and world economy themes expose students to aspects of the national and international economies that are relevant to their personal lives. Some of the important concepts discussed with students are infla- tion, the law of supply and demand, the concept of minimum wage, supervisors of the national financial system, imports and exports, international economic blocs, and measures of a country’s well-being. Teacher guidelines explain how to integrate the financial education case studies in the regular curriculum. These can be used in any order at the discretion of the teacher, but with strict guidelines on exposure of conceptual material. 2.4 RESEARCH DESIGN, SAMPLING, AND TIMELINE This section discusses our research design, sampling methodology, and project timeline. 2.4.1 Research design The study was designed in advance of program implementation as a randomized control trial to ensure that the causal effects of the program could be measured accurately and precisely. The scale and scope were defined to ensure that the study would have enough statistical power to measure and detect impacts on all dimensions of interest and enough coverage to have strong external validity. Finally, variation in treatment was introduced to better understand household dynamics. The study evaluates two interventions in public high schools, one targeted to high school students and one to their parents. Treatment for the student interven- tion was assigned at the school level; treatment for the parent intervention was assigned at the individual level within student treatment schools.3 Student treatment schools received financial education material and teacher training. Control schools did not receive any material or training, but participated in surveys and testing in the same manner as the treatment schools. One 11th grade class in each school participated in the pilot program and study. These classes moved on to the 12th grade by the end of the study, the last year of high school. Students in the sample were between 15 and 18 years of age at the start of the intervention, with high repetition rates explaining the age variation. In schools that had more than one section in the 11th grade at the start of the pilot, the school chose which section would participate in the study. In year two, parents of students in treatment schools were invited to partici- pate in a school workshop. At the workshop, they were directed to watch either a Due to political constraints, we were unable to extend the parent intervention to control 3  schools. 70  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES financial education video (workshop treatment group) or a health education video (workshop control group) through random assignment. Parents learned which video they had been assigned to watch only after they arrived at the workshop. The purpose of the parents’ workshop was to reinforce the messages taught to students in class, and to measure the combined impact of parent and student interventions on household-level financial outcomes and decision making. The reason for showing a health education video to the control group was that our counterparts did not think it was logistically and politically feasible to invite only certain parents within a school to a workshop but not others. We do not expect the health video to have an impact on financial outcomes. The figure below summarizes the study design: Group 1 schools Group 2 schools Student intervention Control Treatment Parent intervention — ½ treatment, ½ control 2.4.2 Sample selection and randomization for student intervention Brazil is divided into 26 states and a Federal District. The Federal District and five states were part of the study, including three of the most populous and developed states (São Paulo, Minas Gerais, and Rio de Janeiro), and two states (Ceará and Tocantins) that represent the less developed areas of Brazil. In 2009, the Federal District had the highest GDP per capita in Brazil (US$28,951). Minas Gerais, Rio de Janeiro, and São Paulo are located in the southeast, with GDP per capita of US$8,289, US$12,524, and US$14,872, respectively.4 Ceará and Tocantins are located in the northeast and north of Brazil with GDP per capita of US$4,399 and US$5,960, respectively. In April and May 2010, the education secretariat in each state assembled a list of public high schools that volunteered to participate in the financial educa- tion pilot program, totaling 815 schools. We divided these schools into treatment and control groups through stratified and matched randomization as soon as we received each list, so that the teacher training for treatment schools could be organized and conducted before the midyear school break in July. São Paulo, the state with the largest number of schools, sent four separate lists on different dates. Also, one of the project’s partner institutions (Instituto Unibanco) provided us with an additional list of 101 public schools that it partnered with in Rio de Janeiro, São Paulo, Minas Gerais, and the Federal District. 4  Currency rates as of June 2013 are used: R$1 = US$0.416916; US$1 = R$2.39883. 2.  Financial education and behavior formation   ◾ 71 For each list, we first stratified schools by whether they were located in a municipality with above or below the median number of financial institutions per capita. Within strata, we formed matched pairs of schools and randomly assigned one school in each pair to be in the treatment group and the other to be in the control group. We matched on the following school and municipal variables to improve balance on these characteristics across the treatment and control groups: GDP per capita of the municipality where the school is located, savings volume per capita of the municipality where the school is located, number of students in the school, number of teachers in the school, school drop-out rate, and school continuation rate (percentage of students moving on to the next grade).5 We chose these variables since they may be correlated with the impact of the material on financial knowledge, attitudes, and behavior. We were not able to match on variables collected through our surveys because the randomization had to take place before the baseline survey to enable the program to train the teachers on time. Municipal-level variables used in the randomization come from the Brazilian Statistical Institute (IBGE [Instituto Brasileiro de Geografia e Estatís- tica]) and refer to 2009. School-level variables used in the randomization are for 2008 and were provided by the Federal Ministry of Education (2009 data were not yet available at the time of randomization). Randomization was done by the authors by computer, implying that any differences across the treatment and control groups are due to pure chance. After the randomization was completed, we had to move three control schools to the treatment group manually, since some states requested that at least one school in each school district participate in the program. We chose these schools at random among the schools in the school district and drop them and their pairs from the analysis. Also, after the randomization was completed, but before the program was implemented, we discovered that 12 schools indicated by the Insti- tuto Unibanco decided not to participate in the pilot. We drop these schools and their pairs from the analysis. In addition, five schools had accidentally been listed twice (three were on both the state ministry and Instituto Unibanco lists, and two were duplicated on the state ministry lists). We randomly chose which entry to drop (along with its pair). Another six treatment group schools from São Paulo did not participate in any of our surveys for unknown reasons. We drop these schools 5  For Ceará, we did not have numbers of teachers, drop-out rates, and continuation rates, so we matched on the remaining variables only. For Tocantins, we did not have drop-out rates and continuation rates, so we matched schools on the remaining variables only. Also, since all schools in the Federal District and in Minas Gerais (Juiz de Fora) were located in the same municipality, we only matched on school-level variables in these states. For the Rio de Janeiro and São Paulo schools from the Instituto Unibanco list, we stratified by financial institutions per capita, but only matched on school-level variables not municipality-level variables, since many of these schools were located in the same municipalities, restricting the possible matches for the remaining schools. 72  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES from our sample along with their pairs. In the end, we are left with 868 schools for the analysis—432 in the treatment group and 436 in the control group. Table 2.1 lists the number of schools in each state. Most schools are located in São Paulo, Rio de Janeiro, and Ceará. TABLE 2.1  Number of schools in sample by state STATE TREATMENT CONTROL São Paulo 180 182 Rio de Janeiro 132 133 Ceará 59 60 Federal District 30 29 Tocantins 17 17 Minas Gerais 14 15 Total 432 436 2.4.3 Sample selection for parents’ intervention For the implementation of the parents’ workshops, treatment schools sent in a list of current students enrolled in the class receiving financial education during the spring of 2011. Only schools that sent the list were included in the study, as this was taken as a signal that they were willing to implement the workshop. We used these lists instead of relying on information from earlier surveys because of potential student turnover from one school year to the next.6 A total of 264 treat- ment schools provided a list, totaling 8,534 students. We matched the lists with data from the baseline parent survey on the basis of the student’s name, and stratified and randomly assigned parents in each school into treatment and control groups. We used the following strata: (1) no baseline information on parents, (2) parent had low baseline financial literacy, and (3) parent had high baseline financial literacy. We defined the level of base- line financial literacy based on the number of correct answers to the two finan- cial literacy questions asked on the survey. About 41 percent of parents did not answer any question correctly. We classify these parents as having low baseline financial literacy. Parents who answered one or two questions correctly are clas- sified as having high financial literacy. Each school that provided a list of their students received two separate lists in return: one with the names of the students whose parents would watch the financial education video during the school workshop (treatment group) and 6  According to official statistics for school year 2009 from the National Institute for Educa- tion Studies and Surveys, the average repetition rate was close to 25  percent and the drop-out rate was almost 10 percent in our sample. 2.  Financial education and behavior formation   ◾ 73 one with the names of students whose parents would watch the health video (control group). Schools were provided the financial education and health videos and parent exit questionnaires. Each school was responsible for organizing and implementing the workshop at a time of its choosing and administering the ques- tionnaire at the end of the workshop. 2.4.4 Study timeline The study was designed and agreed on in December 2009 and launched in the spring of 2010. The sample selection and randomization for the student inter- vention occurred in April–May 2010. The baseline survey was conducted in early August 2010 among students and parents in both treatment and control schools. The financial education program was rolled out immediately after. By mid-August 2010, teachers began using the financial education materials in the classroom. The program continued until November 2011 for a total of three school semesters. Over the course of this study, two rounds of follow-up surveys were conducted. The first follow-up survey was implemented in early December 2010, four months after the program started. The results of this survey measure the short-term effects of the program. A second follow-up survey was implemented in December 2011 to assess the longer-term impacts. The parent intervention was introduced in May 2011, and parent outcomes recorded through an exit survey and the December 2011 follow-up survey. 2.5 DATA COLLECTION AND OUTCOME MEASURES For the data collection, we partnered with CAEd of the Federal University of Juiz de Fora, a research institute and survey firm with extensive experience in designing and implementing knowledge tests in Brazilian schools. CAEd has a wide network of supervisors and surveyors across Brazil, and was able to imple- ment the survey across the schools in our sample at the same time. Survey implementation took three days in each school during each survey round (baseline and both follow-ups). On the first day, CAEd staff administered a financial knowledge test and distributed parent questionnaires to students. The students were asked to take the parent questionnaire home, ask one of their parents to fill out the questionnaire, and return it on one of the following days. On the second day, students filled out a self-administered questionnaire measuring financial attitudes and behavior. The third day provided an opportunity for any student who had missed one or both of the previous days to fill out the test and/ or questionnaire. The student tests and questionnaires were administered in the classroom as a regular school exam—i.e., distributed to students, supervised by the surveyor, and collected by the surveyor at the end of the allocated time. 74  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 2.5.1 Outcome measures The surveys included a financial knowledge test and behavioral questionnaire, which were the main instruments to collect data on outcome measures. The test was tailored to the program’s material and objectives.7 Teachers were not privy to it at any point prior, after, or even the day of the tests; and CAEd proctors were present in classrooms for the entire duration. The financial proficiency of students was then calculated based on performance on this knowledge test, with scores ranging from 0 to 100. To assess the impact of the financial education program on future financial attitudes, new measures of financial autonomy and intention to save for students were developed. These measures are relevant to the age group of interest to signal current as well as future financial decision-making potential. Although high school students may not be exposed to a full range of financial decisions, a large proportion of them are in the labor market or have other type of income. Many have the opportunity to use mobile phones and credit cards, and to make purchases in installments. They also must decide to stay in school, and plan for their future studies and employment. Furthermore, financial education provides students with a knowledge base that will allow them to make more informed financial decisions in the future. The financial autonomy measure captures student confidence, indepen- dence, and willingness to participate and influence household financial decisions. The survey asked students the extent to which they agree or disagree with state- ments on (1) reflexive autonomy, such as “I like to think carefully before deciding to buy something”; (2) emotional autonomy, such as “I am prepared to talk about money with my parents”; and (3) functional autonomy, such as “I always try to save some money to do things I really like.” Their responses were then aggregated into an index on financial autonomy (Micarello, Palacios, and Burgos 2012). The intention to save measure aggregated responses to questions on (1) atti- tudes toward future behavior, such as “In my opinion, saving money every month is extremely beneficial”; (2) subjective norms and expectations, such as “My family has the habit of saving money every month”; and (3) perceptions of controlling one’s behavior, such as “I believe I can save a little money every month.” 7  Identical tests were administered in treatment and control schools. Tests were constructed using item response theory (which is also used to construct the GRE and SAT in the United States). This implies that a series of equivalent questions were used to test the same concept, leading to different combinations of questions on each test. The ques- tions that each student received were thus likely to be different in each survey round, and different students within the same class received different questions. This minimizes the risk that students simply remember the question and correct answer without truly under- standing the question; it also reduces the scope for cheating. Item response theory ensures that tests results are comparable across students and across time. 2.  Financial education and behavior formation   ◾ 75 The student surveys also included a series of questions on current financial behavior, asking for example whether students keep track of expenses, whether they make a budget, and how much they save. The parent questionnaires were kept short (about two pages) to increase the probability that parents would be willing to fill them out. These surveys included questions on sociodemographic characteristics, measured financial literacy through standard questions used in the literature, and elicited parents’ finan- cial behavior regarding budgeting and savings. The surveys also asked whether parents talk to students about financial matters and whether students help orga- nize the household budget. 2.5.2 Survey participation Table 2.2 shows the number of schools, students, and parents who participated in the surveys at baseline and at follow-up. At baseline, 866 out of 868 schools in our study sample participated in the survey, although two of these schools did not implement the financial literacy test and the parent questionnaire. In each follow-up survey, about 40 schools did not implement the survey. Reasons for non-implementation varied and were mostly related to scheduling difficulties. The schools that did not participate in the follow-up surveys were different in each round, so we have follow-up data for most schools from either follow-up 1 or follow-up 2. TABLE 2.2  Survey participation   NUMBER OF SCHOOLS NUMBER OF STUDENTS BASELINE FOLLOW-UP 1 FOLLOW-UP 2 BASELINE FOLLOW-UP 1 FOLLOW-UP 2 Financial knowledge test (Day 1) Baseline 864 827 820 23,290 15,554 10,941 Follow-up 1 829 793 17,831 9,305 Follow-up 2 824 18,420 All rounds 791 8,410 Student questionnaire (Day 2) Baseline 866 826 823 24,473 15,076 11,013 Follow-up 1 826 793 16,512 8,602 Follow-up 2 824 176,80 All rounds 793 8,092 Parent questionnaire (take-home) Baseline 864 808 814 21,603 11,603 8,376 Follow-up 1 810 771 13,532 6,281 Follow-up 2 818 14,080 All rounds 769 5,689 76  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES After realizing that survey participation had dropped between baseline and follow-up 1, we provided incentives for survey completion during follow-up 2. Both treatment and control schools where more than 75 percent of the students completed at least 80 percent of the survey questions were entered into a lottery for 1 of 25 computers. Despite this incentive, we did not have a higher number of schools participate in follow-up 2 than in follow-up 1. However, a greater number of students within the participating schools answered the surveys in follow-up 2 than in follow-up 1. The number of students surveyed per school in follow-up 2 is still lower than in the baseline (about 22 versus 28 students per school). One reason for this decline in the number of students per school is that drop-out rates are quite high in our sample (about 10 percent per year). The numbers in table 2.2 also illustrate the fact that student rotation is quite high in our study sample. The majority of students in follow-up 1 were also present in baseline (follow-up 1 was implemented in the same semester as the baseline, about four months apart). However, only about 60 percent of students in follow-up 2 were present at baseline (follow-up 2 took place about 16 months after baseline). With respect to the financial education program, this high rotation implies that more than a third of the sample was not exposed to the material for a full three semesters, but rather for only one or two semesters.8 The participation in the parent questionnaire was quite high, considering that this questionnaire was self-administered at home. About 88 percent of students returned parent questionnaires at baseline. In the follow-up surveys, the partici- pation rates in the parent questionnaire dropped to about 76 percent. 2.6 SUMMARY STATISTICS, ATTRITION, AND PROCESS EVALUATION Table 2.3 displays preprogram summary statistics of school and student char- acteristics for treatment and control schools. The school-level variables are for 2008 and were provided by the Federal Ministry of Education. As a result of the school matching and randomization procedures described in section 2.4, the preprogram school characteristics are the same on average in the treatment and This high turnover of students in our sample needs to be interpreted in light of a 8  systematic problem of school retention that characterizes secondary education in Brazil. Schwartzman (2010), compiling data from 1998 to 2008, shows that 21.3 percent of high school students repeat at least one year, either for lack of achievement or because they drop out. Comparing the same figure to countries with similar socioeconomic levels— such as Chile with 3.18  percent repeaters in secondary education and Argentina with 13.43  percent—and even to countries with lower socioeconomic levels—such as Peru with 5.62 percent of repeaters and Colombia with 2.70 percent—shows how complex the student retention problem is in Brazil. 2.  Financial education and behavior formation   ◾ 77 TABLE 2.3  Preprogram summary statistics NUMBER MEAN SCHOOL-LEVEL VARIABLES USED TREATMENT CONTROL TO FORM MATCHED PAIRS FOR SCHOOLS STUDENTS GROUP GROUP P-VALUE RANDOMIZATION (1) (2) (3) (4) (5) Number of students in school 867 669.86 640.20 0.366 Number of teachers in school 748 38.11 37.41 0.702 School drop-out rate 698 9.90 9.83 0.925 Schoolwide class passing rate 698 74.26 74.24 0.981 Student background characteristics Student is female 864 26,131 0.56 0.55 0.069* Mother’s education: < high school 863 22,474 0.43 0.45 0.392 Father’s education: < high school 863 22,357 0.40 0.41 0.449 Student has failed at least 1 school 863 22,435 0.32 0.30 0.100 year Receives Bolsa Familia cash transfer 863 22,662 0.33 0.32 0.281 Student has computer with Internet 863 22,502 0.53 0.52 0.542 at home Student has income 866 24,319 0.67 0.66 0.111 Student works 866 24,303 0.35 0.35 0.902 Student financial knowledge and behavior Financial proficiency index 864 23,255 50.15 49.80 0.461 Saves more than zero 866 24,089 0.46 0.45 0.075* Makes a list of all expenses every 866 24,080 0.11 0.10 0.680 month Negotiates price or payment method 866 23,867 0.76 0.75 0.462 Financial autonomy index 866 2,789 49.04 49.11 0.844 Intention to save index 866 22,797 48.19 48.29 0.772 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. School characteristics are for 2008 and come from administrative data from the Federal Ministry of Education. Student characteristics were collected through the baseline survey conducted in August 2010. Financial knowledge and behavior indexes are scaled to lie between 0 and 100. Column 5 displays the p -value, i.e., statistical significance level, of the difference in the treatment and control group means for each variable. control groups. For example, the drop-out rate is about 10 percent, and the class passing rate is 74 percent. The remaining variables in table 2.3 were collected through the baseline survey conducted in August 2010. Student background characteristics show that 56  percent of students participating in the study were female, 67  percent had some form of income (from work or from parents), and about 35  percent were working at baseline. Additionally, 33 percent were beneficiaries of the Bolsa Família government cash transfer program, indicating that they belonged to low-income households. On the parent side, about 60  percent of the students’ parents had less than a high school education. In terms of financial behavior, about 78  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 45 percent of students reported that they saved at least some of their income, only 11 percent made a list of their expenses every month, but 75 percent nego- tiated the price or payment method when making purchases. Overall, the data in table 2.3 indicate that students’ background characteristics, as well as financial knowledge and behavior, were the same across the treatment and control groups at baseline—as expected, since treatment status was randomly assigned. 2.6.1 Take-up and process evaluation Following school assignment to treatment and control groups, the program distributed textbooks to treatment schools and organized training sessions for the teachers. The vast majority of treatment schools received the financial educa- tion textbooks and distributed them to students. In both rounds of the follow-up survey, we asked school principals as well as teachers and students a series of questions regarding implementation and usage of the financial education program. Follow-up 1 spans the first semester of the program, while follow-up 2 covers the second and third semesters. Table 2.4 shows that over 95 percent of treatment school principals report that they received the textbooks for the first semester, and 93  percent report receiving them for semesters 2 and/or 3. The large majority of teachers also say that students received the textbooks (i.e., the books were actually distributed to students): 94  percent and 92  percent in follow-up 1 and 2, respectively. On training, 78  percent of teachers report that they received training on how to use the financial education material in the first semester, and 65 percent report receiving it in the second/third semesters. In terms of usage, 87 percent of students report that teachers actively used the financial education textbooks in classrooms in the first semester, though a drop-off occurs in semesters 2 and 3, with 74 percent of students reporting usage. The  percentage of principals reporting that financial education was taught in school remained high throughout the study period: 93 percent for all semesters. Schools in the control group did not receive textbooks or teacher training through the financial education program studied in this chapter. However, they TABLE 2.4  Reported program take-up and usage in treatment group schools FOLLOW-UP FOLLOW-UP 1 2 Principal: School received financial education textbooks 0.955 0.927 Teacher: Students received financial education textbook 0.938 0.921 Teacher: Received training 0.783 0.647 Student: Teacher used textbook in classroom 0.871 0.744 Principal: Financial education was taught in school 0.928 0.928 2.  Financial education and behavior formation   ◾ 79 may have implemented other types of financial education. The principal and teacher questionnaires were only applied in treatment schools during the first follow-up survey; but during the second follow-up survey, control schools also answered these questionnaires.9 In this survey, 16.6  percent of control group principals reported that the school had a financial education program, and 11 percent of control group teachers reported receiving some training related to financial education. We do not have detailed information on the financial educa- tion program implemented in control schools. However, only 5 percent of control group principals reported that the school received a textbook with financial education material, suggesting that these programs may be less intensive than the one studied in this chapter. We supplemented this quantitative analysis with qualitative work in the form of teacher focus groups, which were organized with our Brazilian counterparts. The purpose of these meetings was to learn how teachers implemented the program, how material was integrated into the regular curriculum, and how the contents and/or delivery could be improved for future implementation. Six such meetings were held in September and October 2011, with teachers and educators from all six states. All meetings were conducted in facilities provided by the state education departments and meals were provided. The meetings were generally well attended (more than 200 educators attended in São Paulo), and lasted an average of four hours each. In all states, the teachers and educators greatly approved of the textbooks and the financial education program. They said that the material allowed students to learn by themselves, that students loved the case studies and felt that they connected well with their daily life situations. The teachers also complimented the clarity with which concepts were conveyed in the books. One teacher commented, “What motivates me the most about this project are the books. The content is directly related to students’ lives and helps to insert them in a highly competitive market society. I learned a lot from the material, and similarly to what happened to my students, it has helped me to plan better for the future.” In terms of implementation, most teachers employed work group strategies to teach the material, where students were divided into small groups and asked to work on different tasks. Examples of such tasks varied from identifying steps for opening a small firm to creating a school market. Several teachers reported assigning projects to students related to fundraising and organizing their own graduation party. Other teachers reported undertaking field trips to local markets, universities, and companies to learn how they operate. Others used various forms of media to explore the topics in the textbook, such as computer simulations and videos. The student questionnaire did not include questions on program implementation in 9  control group schools in either follow-up survey. 80  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES While these qualitative responses are incredibly encouraging for the program, the teachers also reported some difficulties in implementation. For instance, some teachers expressed concern at the lack of resources to complement the financial education textbooks. Many felt they did not have enough time to learn the material themselves before it was due to be taught to students. Some even felt they had to give up some core curriculum time so they could learn the mate- rials through training. Other concerns were more logistical. Teachers from all states felt that it was difficult to obtain timeline information about the project. Since the project delivery was decentralized, many teachers and educators reported not being able to identify a central resource for logistical inquiries. Similarly, some teachers said they were only told of the follow-up surveys a few days prior and had little time to accommodate the surveys in the class schedule. Finally, some teachers expressed concern about the continuity of the curriculum given the high turnover rate among students. Overall, the teachers and educators felt the financial education program was valuable, that the textbooks were extremely relevant, and that they as well as their students enjoyed the new learning opportunities afforded to them. 2.6.2 Take-up of parent workshops For the parent workshops, we asked schools to mail the filled-in parent exit questionnaires back to us after the workshop, and we determine which schools implemented the workshops based on our receipt of these questionnaires and on our random assignment. We received 1,553 filled-in parent questionnaires from 109 schools, implying that parent workshops did not take place in the other 153 schools (the school either did not organize the workshop or parents did not attend the workshop). When planning the parent intervention, our counterparts had cautioned that attendance rates may be low, since parents of public high school students in Brazil tend not to participate in school events. Comparing the lists of current students we received from schools before the workshop to the filled-in exit questionnaires gives an average attendance rate of 46 percent across the 109 schools that returned questionnaires. The attendance rate does not differ across parents who were randomized into the treatment and control groups. Parents did not know whether they had been assigned to watch a financial education or health video until the video was screened at the work- shop—i.e., the decision to attend was independent of treatment status. Hence, in our impact analysis of the parent workshop, we only keep students and parents if the parent attended a workshop. The interpretation of the results is applicable to families that in general are more interested in school events and respond to school invitations. 2.  Financial education and behavior formation   ◾ 81 2.7 RESULTS AND DISCUSSION Based on the random assignment, the impact of the program is measured as the difference in average outcomes in the treatment and control groups. The results presented in figures 2.1 and 2.2 are statistically significant at the 1 percent level (unless otherwise stated), with standard errors clustered at the school level. Annex A presents full regression tables. We include school pair dummies in all specifications and control for baseline values of available dependent variables, as per Bruhn and McKenzie (2009) and McKenzie (2011). When baseline values have missing observations, we replace these with 0 and include a dummy variable indi- cating that the observation was missing.10 Finally, note that results from the first and second follow-up surveys are not fully comparable due to changes in class composition from one year to the next. 2.7.1 Student financial proficiency The impact of the financial education program on student financial proficiency is economically and statistically significant. Test scores indicate that the average level of financial proficiency is significantly higher in the treatment group than in the control group in both follow-up 1 and follow-up 2, as shown in figure 2.1a. The differences are of 4 points and 3 points, respectively, or a 5–7 percent increase in financial knowledge. This is equivalent to a quarter of a standard deviation increase, which is a substantial effect size. Further, test scores improve across the distribution, benefiting low- and high-achieving students. Specifically, the proportion of students who perform exceptionally well increases by 28  percent, and that of students performing exceptionally poor decreases by 26 percent. This effect represents a rightward shift in the distribution of test scores for treated schools compared to control schools, as shown in figure 2.2. Hence, the financial education program helps poorly performing students improve significantly, and high-performing students to do even better. These distributional effects are important and show that the program benefited students along a broad performance spectrum rather than being driven by any one category, and that the curriculum speaks to the learning needs and interests of all types of students. To account for the parent workshops that were held prior to follow-up 2, we complement 10  our analysis by running regressions with two independent treatment variables: the first indicating treated students and parents who were assigned to the financial literacy work- shop, and the second indicating treated students and parents who were assigned to the health literacy workshop. These variables are compared to the control group in the regres- sion analysis, and the coefficients remain statistically significant for both treatments. The F-tests for differences in coefficients between the two treatments are not significant for our outcomes of interest (except one parent financial literacy question). 82  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 2.1  Impacts of financial education program on students a. Financial proficiency b. Saving more than zero c. Income saved Score Percent Percent 65 100 100 62 61 60 80 80 59 57 56 60 60 49 46 44 40 53 40 40 49 20 20 13 14 45 0 0 Follow-up 1 Follow-up 2 Follow-up 1 Follow-up 2 Follow-up 1 Follow-up 2 d. Listing expenses in a budget e. Negotiating prices f. Financial autonomy Percent Percent Score 50 100 53 52 40 80 78 77 74 74 51 51 51 30 60 20 17 40 49 16 14 49 13 10 20 0 0 47 Follow-up 1 Follow-up 2 Follow-up 1 Follow-up 2 Follow-up 1 Follow-up 2 g. Intention to save h. Participation in HH finances i. Participation in HH budgeting Score Percent Percent 54 100 100 53 52 80 71 70 74 80 51 51 67 60 60 54 52 56 50 49 49 40 40 48 20 20 46 0 0 Follow-up 1 Follow-up 2 Follow-up 1 Follow-up 2 Follow-up 1 Follow-up 2 2.7.2 Saving and spending behavior Next, we investigate saving and spending behavior and identify significant improvements in both. Note first that in our sample almost two-thirds of students had some type of income and about 38–42  percent were in the labor market. Hence, these students faced decisions on how much to save and what to spend their money on, whether to make a monthly list of expenses, and whether to negotiate price and/or payment methods when making a purchase. Their choices offer key insights into the spending behavior of youth whose purchases of mobile phones, clothes, and fashion accessories may be driven by factors other than financial planning and foresight. 2.  Financial education and behavior formation   ◾ 83 FIGURE 2.2  Distribution shift in financial proficiency scores a. Follow-up 1 a. Follow-up 2 Percent Percent 0.025 0.025 Control Control 0.015 0.015 Treatment Treatment 0.005 0.005 0 0 20 40 60 80 100 120 20 40 60 80 100 120 Financial proficiency score Financial proficiency score The analysis indicates strong treatment effects on all these dimensions. Specif- ically, students in the treatment group increase savings and exhibit significantly improved spending behavior relative to the control group. A higher proportion of students in the treatment group saved at least some of their income (49 percent in treatment compared to 44  percent in control at follow-up 1; and 46  percent compared to 40 percent, respectively, at follow-up 2). Not only does the propor- tion of students saving increase, but the actual savings amounts per student increase significantly also, by 1.4 percentage points in follow-up 2 (14.3 percent of income saved in the treatment group as compared to 12.9 percent of income saved in the control group).11 These results are shown in figures 2.1b and 2.1c. We find significant treatment effects on spending behavior as well. The results show that 16  percent of students in treatment schools make a list of monthly expenses as part of a budgeting exercise, compared to 13 percent in the control group in follow-up 1. These numbers are 17 percent and 14 percent, respectively, in follow-up 2, as shown in figure 2.1d. Additionally, a higher percentage of students in the treatment group say they often negotiate prices or payment methods when making a purchase (78 percent in treatment compared to 74 percent in control in follow-up 1; and 77 percent and 74 percent, respectively, in follow-up 2). These results are plotted in figure 2.1e. 11  This question was not asked in follow-up 1. 84  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 2.7.3 Student financial autonomy and intention to save In addition to improvements in current saving and spending behavior, we iden- tify positive treatment effects on forward-looking indexes of financial autonomy and intention to save. Specifically, the average financial autonomy score for students in the treatment group is 51 compared to 49 in the control group in follow-up 1; and 52 compared to 51, respectively, in follow-up 2. In addition, we find that students in the treatment group have a higher measured intentioned to save (51) than those in control group (49) at follow-up 1, and 53 compared to 51 in follow-up 2. These results are presented in figures 2.1f and 2.1g. Normalizing these effect sizes, they range from 8 percent to 12 percent of a standard deviation. 2.7.4 Student participation in household finance Finally, we investigate whether financial education has any impact on students’ current participation in household finances using data collected through the parent questionnaires described in section 2.5. We find that a significantly larger percentage of students in the treatment group talk to their parents about finances and participate in organizing the household budget. First, figure 2.1h shows that 71 percent of students in treatment schools participate in household financial decisions compared to 67  percent in control schools in follow-up 1, and 74 percent compared to 70 percent in follow-up 2. Figure 2.1i shows similar positive treatment effects for students helping orga- nize household budgets. Overall, we find strong and statistically significant evidence that the financial education program was effective in improving financial knowledge, current and future financial behavior, and participation in household finances and budgeting. 2.7.5 Impact of student intervention on parent outcomes Because the student financial education program included take-home exercises, such as making a budget that required the participation of parents, we examine whether the program has any trickle-up effects on parents’ financial knowledge and behavior. Our results show no impact of the student financial education on parents’ financial knowledge in follow-up 1. Financial knowledge is measured through two standard questions on inflation and interest rates. However, in follow-up 2, parents of treatment school students are significantly more likely to answer these questions correctly than parents of students in control schools (see table 2A.5 in annex A). These effects are statistically significant at the 10 percent level. In follow-up 2, we also detect a positive and statistically significant differ- ence in financial knowledge of budgeting between the two groups: Compared to 68 percent of parents in the control group, parents of students in the treatment schools are 6.3 percent more likely to understand the composition of a budget. 2.  Financial education and behavior formation   ◾ 85 This is a promising result, as budgeting was keenly taught in the student curric- ulum, including a take-home exercise where students were asked to make a household budget in consultation with their parents. When examining the impact of student financial education on parents’ finan- cial behavior, we detect no effects in follow-up 1, but again see several posi- tive impacts in follow-up 2. The percentage of parents who save more than zero increases from 76 percent in control schools to 78 percent in treatment schools. The average percentage of income saved increases from 12  percent in control schools to close to 13  percent in treatment schools. Both of these effects are statistically significant at the 5 percent level. Parents in student treatment schools are also more likely to list monthly expenses in a budget, with an increase from 37 percent of parents in the control schools to 39 percent of parents in the treat- ment schools. These results indicate that the student financial education has a significant trickle-up impact, or spillovers, to parents. The next section discusses whether the parent workshops reinforced this effect. 2.7.6 Impact of parent workshops Some parents of treatment school students participated in a financial education workshop that took place between the first and second follow-up surveys. This workshop was designed to raise awareness among parents about the importance of financial education and to encourage them to interact more with their children on financial matters—thereby leveraging and reinforcing the material students were taught through the program. Of the treatment schools, 109 volunteered to hold a parent workshop, and 46  percent of parents in these schools attended. These parents were randomly divided into two groups: a treatment group that watched a financial literacy video on the benefits of saving and budgeting, and a control group that watched a health video on the benefits of adopting preventa- tive measures for a number of diseases. To measure the impact of the parent workshop, we compare outcomes of parents who attended and also the outcomes of their children across the groups that watched the financial literacy video and the health video. We only include parents who actually attended a workshop in this analysis, excluding parents in the same schools who were invited but did not come to the workshop. Parents who went to the workshop have different baseline characteristics from those who did not attend. They are more likely to be recipients of the Bolsa Família cash transfer program, and fathers are less likely to have completed at least some secondary education—suggesting that they are from relatively more disadvan- taged households. Drop-out rates are also higher in schools that held a parent workshop. These differences imply that the parent workshop results are not representative of the full sample, although they are valid for the group of parents who attended a workshop (and the corresponding students). 86  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES We find no significant improvements in parents’ financial behavior as a result of watching the financial literacy video—parents who watched this video are no more likely to make a budget or improve savings behavior compared to those who watched the health video. The lack of significant impact of the parent workshop may be due to the relatively low intensity of the treatment or other constraints parents face in responding to the information. The exposure to financial education material through the DVD was relatively short, and the workshops included no supple- mentary discussions. This is consistent with the existing literature on financial education (see, e.g., Cole, Sampson, and Zia 2011). The remarkable result from this intervention, however, is the impact of the parents’ workshop on student behavior. The parent workshops significantly increase the savings rate among students by 2.5  percentage points. Specifi- cally, students whose parents participate in the health literacy workshops save on average 13.5 percent of their income. In comparison, students whose parents participate in the financial literacy workshops save 16  percent of their income. This difference is statistically significant at the 5 percent level. Hence, parents are able to use their improved knowledge to reinforce the school messages with their children. 2.8 IMPACT OF THE STUDY The findings of this study are being used to guide policy discussions on the impact of financial education in schools. The audience includes financial institutions and education sectors in Brazil and several other interested countries. In fact, the Brazilian Ministry of Education has recently approved a continuation of the program. The program will be scaled up to 3,000 public high schools as part of two federal programs: Ensino Médio Inovador and Mais Educação. Brazilian stake- holders also decided to develop financial education textbooks and a pilot project for primary schools, as well as a financial education program targeted toward female beneficiates of Bolsa Família and retirees. Several other countries in the region have expressed interest in the Brazilian experience to learn and adapt the program to their respective environments and school systems. As an initial indicator of policy impact, the results of the study were widely covered in the national media of Brazil—in newspapers, radio, and television. Journalists from the largest news outlets in the country, such as O Globo, Valor Econômico, Folha de São Paulo, and Agência Estado reported on the potential of such a program in improving national saving behavior. A simple back-of-the-envelope calculation on economywide impacts of our intervention complements this positive press coverage. We start by assuming that the 1  percentage point increase in the savings rate is channeled into 2.  Financial education and behavior formation   ◾ 87 domestic investment (note that we ignore compounding over the years).12 Brazil’s current investment stands at 18  percent of GDP, and it supports an economic growth rate of 3.5 percent on average. Dividing the investment share of GDP by the economic growth rate results in an incremental capital-output ratio of 5.14. Hence, a 1  percentage point increase in savings and investment would yield a one-fifth of a percentage point increase in GDP. With a $2 trillion economy, this results in a $4  billion annual increase—a substantial amount. Clearly, this is a simple calculation and does not account for several macro factors. An even larger evaluation and macro-simulation would make this estimation stronger; however, the ballpark improvement in the economy is substantial and definitely outweighs the entire cost associated with the intervention several times over. Another way to assess the impact of our study is to compare it to similar studies in the literature. Unfortunately, the literature that studies financial education in schools is extremely small, and many of the existing studies are fraught with identification concerns as discussed in section 2.2.1. But there are school-based studies outside the realm of financial education that are important for comparison purposes. The type of education interventions in secondary schools that have been tested through random control trials include the provi- sion of school resources, monetary incentives to students, and student tutoring. In general, these studies identify improvements in student learning during the periods studied, although some studies do not find the effects to be statistically significant. Perhaps the most comparable study to ours is one in Brazil that provides financial resources and access to technology to Brazilian high schools—Projeto Jovem de Futuro. Using a randomized control trial methodology, the study finds improvements in test scores in Portuguese by 34 percent and in math by 55 percent. These are substantial effect sizes, but in standard deviation terms, these effects are not very large. Another randomized evaluation of computers for education in Colombia among 97 schools and 5,201 students finds improvements of 0.017 standard deviations in Spanish and 0.008 standard deviations in math (Barrera-Osorio and Linden 2009). Providing monetary incentives to students also tends to improve test scores. Perhaps the most well-known studies in this area are Angrist and Lavy (2009) and Angrist et al. (2002). In the first study, Israeli students are provided cash incentives to pass their graduation exams; while the mean estimates are positive, they are not statistically significant. In the second study, lotteries are used in Columbia to distribute vouchers to partially A caveat here is that not all savings may be channeled into investment, particularly if 12  individuals save at home and not in financial institutions. In addition, an increase in saving represents a loss in consumer spending, which the simple calculation here does not take into account. 88  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES cover the cost of private secondary schooling for 1,600 students who maintain satisfactory academic progress. Three years after the lotteries, winners are about 10 percentage points more likely to have finished eighth grade, and score 0.2 standard deviations higher on achievement tests, although the latter result is only marginally significant. Finally, direct tutoring is beneficial in improving test scores too—a randomized evaluation in Chile finds that providing a three-month program of small group tutoring to fourth graders using college student volun- teers increases reading performance by 0.15–0.2 standard deviations (Cabezas, Paredes, and Quezada 2011). Compared against this literature, our finding of an improvement of 0.2–0.24 standard deviations in financial proficiency lies in the very top end of statistically significant improvements in test scores. 2.9 CONCLUSION This chapter makes an important contribution to the literature by demonstrating that a financial education program targeted to youth can both increase knowl- edge and improve attitudes and behavior. We believe this is due to two factors: (1) the quality and intensity of the program; and (2) the quality, scope, and scale of the study. Brazil embarked on an extraordinary experience with financial education that has attracted national and international attention for its scope, innovation, and results. The ambition is to move policy practice from small targeted interventions to a multi-arm program to improve the intertemporal decisions and economic outcomes of a whole nation. Innovation in design is a central element of this program, with a coalition of financial institutions working with the best educa- tors to develop creative and interactive didactical tools to transform the way young people think about their life choices. The early results are impressive: the approach is effective in improving what high school students know, their current saving and spending behavior, and their attitudes toward investing in their future. Furthermore, educating parents further improves impact on students. The early lessons from this experience are important for Brazil and for the rest of the world. First, increasing knowledge may lower the demands on cogni- tive resources and help reduce anomalies in intertemporal choice. Second, finan- cial education can be understood more widely as the life skills needed for making better intertemporal decisions, being aware of opportunities ahead, and plan- ning to take advantage of them. As such, financial education can strengthen the effectiveness of policies aimed at economic growth and poverty reduction. Third, the partnership between the highly educated and resourceful facets of Brazil’s financial and private sectors with the less resourced education system can play a transformative role in making public schools an exciting place for learning and personal growth, and improving the prospects for equality of opportunity for all. 2.  Financial education and behavior formation   ◾ 89 Fourth, educating parents can strengthen their involvement in their children’s education and generate powerful dynamics within the household. Adult educa- tion is thus an important element in children’s education and poverty reduction. Finally, once brought to scale, this program has the potential to improve national savings rates and, potentially, the rate of economic growth. REFERENCES Administrative Office of the United States Courts. 2012. 2011 Report of Statistics Required by the Bankruptcy Abuse Prevention and Consumer Protection Act of 2005. http:// www.uscourts.gov/uscourts/Statistics/BankruptcyStatistics/BAPCPA/2011/BAPCPA- report.pdf. Angrist, Joshua, Eric Bettinger, Erik Bloom, Elizabeth King, and Michael Kremer. 2002. “Vouchers for Private Schooling in Colombia: Evidence from a Randomized Natural Experiment.” American Economic Review 92 (5): 1535–58. Angrist, Joshua, and Victor Lavy. 2009. “The Effects of High Stakes High School Achievement Awards: Evidence from a Randomized Trial.” American Economic Review 99 (4): 301–31. Barrera-Osorio, Felipe, and Leigh L. Linden. 2009. “The Use and Misuse of Computers in Education: Evidence from a Randomized Experiment in Colombia.”  Policy Research Working Paper Series 4836, World Bank, Washington, DC. Beal, Diana, and Sarath Delpachitra. 2003. “Financial Literacy among Australian University Students.” Economic Papers 22 (1): 65–78. Becchetti, Leonardo, Stefano Caiazza, and Decio Coviello. 2011. “Financial Education and Investment Attitudes in High Schools: Evidence from a Randomized Experiment.” CEIS Research Paper No. 210, Centre for Economic and International Studies, Tor Vergata University, Rome. Becchetti, Leonardo, and Fabio Pisani. 2012. “Financial Education on Secondary School Students: The Randomized Experiment Revisited.” AICCON Working Paper No. 98, Associazione Italiana per la Cultura della Cooperazione e del Non Profit, Forlì, Italy. Benjamin, Daniel, Sebastian A. Brown, and Jesse M. Shapiro. 2012. “Who Is “Behavioral”? Cognitive Ability and Anomalous Preferences.” http://econweb.arts.cornell.edu/ dbenjamin/iq.pdf. Bernheim, B. Douglas, Daniel M. Garrett, and Dean M. Maki. 2001. “Education and Saving: The Long-Term Effects of High School Financial Curriculum Mandates.” Journal of Public Economics 80 (3): 435–65. Berry, James, Dean Karlan, and Menno Pradhan. 2012. “Evaluating the Efficacy of School Based Financial Education Programs with Children in Ghana.” Abdul Latif Jameel Poverty Action Lab. Bruhn, Miriam, and David McKenzie. 2009. “In Pursuit of Balance: Randomization in Practice in Development Field Experiments.” American Economic Journal: Applied Economics 1 (4): 200–232. Cabezas, Victor, Ricardo Paredes, and C. Quezada. 2011. “Peer effect, segregation and school performance in Chile.” Working Paper, Department of Industrial Engineering, PUC. Carlin, Bruce Ian, and David T. Robinson. 2010. “What Does Financial Literacy Training Teach Us?” NBER Working Paper 16271, National Bureau of Economic Research, Cambridge, MA. Cole, Shawn A., Thomas Sampson, and Bilal Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. 90  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Cole, Shawn A., and Gauri K. Shastry. 2009. “Smart Money: The Effect of Education, Cognitive Ability, and Financial Literacy on Financial Market Participation.” Finance Working Paper Series 09-071, Harvard Business School, Cambridge, MA. Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Lührmann, Melanie, Marta Serra-Garcia, and Joachim Winter. 2012. “The Effects of Financial Literacy Training: Evidence from a Field Experiment with German High-School Children.” Discussion Papers in Economics 14101, University of Munich, Munich. Lusardi, Annamaria, and Olivia S. Mitchell. 2007. “Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education.” Business Economics 42 (1): 35–44. —. 2011. “Financial Literacy and Retirement Planning in the United States.” CeRP Working Papers, Center for Research on Pensions and Welfare Policies, Turin, Italy. Lusardi, Annamaria, Olivia S. Mitchell, and Vilsa Curto. 2009. “Financial Literacy Among the Young: Evidence and Implications for Consumer Policy.” NBER Working Paper 15352, National Bureau of Economic Research, Cambridge, MA. —. 2010. “Financial Literacy among the Young.” Journal of Consumer Affairs 44 (2): 358–80. McKenzie, David. 2011. “Beyond Baseline and Follow-up: The Case for More T in Experiments.” Policy Research Working Paper 5639, World Bank, Washington, DC. Mandell, Lewis. 2006. “Financial Literacy: If It’s So Important, Why Isn’t It Improving?” Networks Financial Institute Policy Brief 2006-PB-08. Mandell, Lewis, and L. S. Klein. 2007. “Motivation and Financial Literacy.” Financial Services Review 16 (2): 106–16. Micarello, Hilda, Manuel Palacios Cunha Melo, and Marcelo Baumann Burgos. 2012. “Application of the CAEd Autonomy Scale to Assess the Impact of Financial Education.” CAEd. http://www.pesquisa.caedufjf.net/wp-content/uploads/2012/03/autonomia_ marcelo_burgos.pdf. Schwartzman, Simon. 2010. “Benchmarking Secondary Education in Brazil.” Paper presented at the IDB/OCED/Ministry of Education International Seminar on Best Practices of Secondary Education, May 3–4, Brasilia. Varcoe, Karen P., Allen Martin, Zana Devitto, and Charles Go. 2005. “Using a Financial Education Curriculum for Teens.” Journal of Financial Counseling and Planning 16 (1): 63–71. Walstad, William B., Ken Rebeck, and Richard A. MacDonald. 2010. “The Effects of Financial Education on the Financial Knowledge of High School Students.” Journal of Consumer Affairs 44 (2): 336–57. 2.  Financial education and behavior formation   ◾ 91 ANNEX A: REGRESSION TABLES TABLE 2A.1  Impact on student financial proficiency FINANCIAL PROFICIENCY SCORE FOLLOW-UP 1 FOLLOW-UP 2 (1) (2) Student treatment 3.548*** 3.020*** (0.296) (0.355) R-squared 0.455 0.328 Observations 17,831 18,415 Number of clusters 829 824 Dependent variable mean in control group 56.135 59.003 Dependent variable SD in control group 14.804 14.908 Note: Robust standard errors, clustered at the school level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. SD = standard deviation. Columns 1 and 2 use data from follow-up waves 1 and 2, respectively. The number of students and schools included in the sample fluctuate within a wave because not all students answered every question; and they fluctuate across waves because of student turnover across the school years. The outcome variable in this table is a student financial proficiency score, which aggregates financial knowledge questions included in the surveys. All regressions control for baseline outcomes and include school pair dummies. When baseline outcomes have a missing value, they are replaced by 0 and a dummy variable indicating such missing values is included. TABLE 2A.2  Impact on student saving and spending behavior LIST MONTHLY NEGOTIATE PRICES SAVES MORE PERCENTAGE OF EXPENSES IN A OR PAYMENT THAN ZERO INCOME SAVED BUDGET METHODS (1) (2) (3) (4) (5) (6) (7) (8) Student treatment 0.047*** 0.052*** N/A 1.389*** 0.026*** 0.030*** 0.041*** 0.031*** (0.006) (0.007) (0.316) (0.005) (0.005) (0.005) (0.006) R-squared 0.210 0.106 0.045 0.149 0.093 0.219 0.131 Observations 16288 17320 16695 16358 17475 16175 17343 Number of clusters 825 822 822 825 822 825 822 Dependent variable 0.440 0.404 12.897 0.129 0.139 0.740 0.740 mean in control group Dependent variable SD 18.958 in control group Note: Robust standard errors, clustered at the school level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. SD = standard deviation. Odd- and even-numbered columns use data from follow-up waves 1 and 2, respectively. The number of students and schools included in the sample fluctuate within a wave because not all students answered every question; and they fluctuate across waves because of student turnover across school years. The outcome variables in this table are an indicator variable equal to 1 if the student saves a positive fraction of income; the actual percentage of student monthly income that is saved; an indicator variable equal to 1 if the student makes a list of all monthly expenses; and an indicator variable equal to 1 if the student negotiates either the price or the payment method when making purchases. All regressions control for baseline outcomes and include school pair dummies. When baseline outcomes have a missing value, they are replaced by 0 and a dummy variable indicating such missing values is included. 92  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 2A.3  Impact on student attitudes and future behavior FINANCIAL AUTONOMY INTENTION TO SAVE INDEX INDEX (1) (2) (3) (4) Student treatment 1.703*** 1.774*** 2.471*** 1.914*** (0.234) (0.305) (0.297) (0.324) R-squared 0.449 0.250 0.278 0.168 Observations 14283 16019 14919 16286 Number of clusters 824 822 823 821 Dependent variable mean 49.035 50.544 49.020 51.342 in control group Dependent variable SD in 19.785 20.724 21.157 21.232 control group Note: See table 2A.2. The outcome variables in this table are a student financial autonomy index and an intention to save index. The financial autonomy index aggregates responses to questions that elicit future financial preferences, confidence, and decision making independence. The intention to save index aggregates responses to questions on future hypothetical savings and spending scenarios. TABLE 2A.4  Impact on student participation in household finance STUDENT DISCUSSES FINANCIAL MATTERS WITH STUDENT HELPS ORGANIZE PARENTS HOUSEHOLD BUDGET (1) (2) (3) (4) Student treatment 0.036 *** 0.039 *** 0.048 *** 0.036*** (0.007) (0.006) (0.007) (0.008) R-squared 0.171 0.104 0.194 0.115 Observations 13357 13844 13345 13867 Number of clusters 810 816 809 816 Dependent variable 0.673 0.701 0.487 0.522 mean in control group Note: See table 2A.2. The outcome variables in this table are an indicator variable equal to 1 if a stu- dent discusses financial matters at home, and an indicator variable equal to 1 if a student helps orga- nize the household budget. Both questions are based on responses to the parent questionnaires. 2.  Financial education and behavior formation   ◾ 93 TABLE 2A.5  Trickle-up impact on parent financial knowledge CORRECTLY ANSWERED INTEREST CORRECTLY ANSWERED KNOWS WHAT GOES RATE QUESTION INFLATION QUESTION INTO A BUDGET (1) (2) (3) (4) (5) (6) Student treatment −0.001 0.017 * 0.007 0.014 * N/A 0.063*** (0.008) (0.009) (0.008) (0.008) (0.009) R-squared 0.091 0.071 0.132 0.104 0.080 Observations 13368 12691 13327 12609 12710 Number of clusters 809 816 809 816 816 Dependent variable mean in 0.444 0.443 0.329 0.320 0.678 control group Note: See table 2A.2. The outcome variables in this table are three financial literacy questions in the parent surveys: the interest rate question tests the ability to calculate an interest rate using percentages; the inflation question tests the understanding of how inflation affects future purchasing power; and the budgeting question tests the knowledge of what goes into a budget. TABLE 2A.6  Trickle-up impact on parent saving and spending behavior PERCENTAGE OF LISTS MONTHLY HAS FORMAL SAVINGS INCOME SAVED EXPENSES IN A BUDGET (1) (2) (3) (4) (5) (6) Student treatment 0.003 0.014** N/A 0.667** −0.012 0.021*** (0.006) (0.006) (0.263) (0.007) (0.007) R-squared 0.334 0.206 0.050 0.190 0.116 Observations 13,079 13,533 12,953 13,187 1,566 Number of clusters 810 816 816 810 816 Dependent variable mean in control 0.734 0.762 12.171 0.366 0.369 group Dependent variable 16.521 SD in control group Note: See table 2A.2. The outcome variables in this table are based on responses in the parent survey and include an indicator variable equal to 1 if the parent has formal savings such as a current account, savings account, debit card or checks; the actual fraction of monthly income that is saved; and an indicator variable equal to 1 if the parent makes a list of monthly expenses in a budget. 94  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 2A.7  Impact of parent financial education workshop PARENT STUDENT KNOWS LISTS WHAT HAS % OF MONTHLY % OF GOES INTO FORMAL INCOME EXPENSES IN HAS INCOME A BUDGET SAVINGS SAVED A BUDGET SAVINGS SAVED Attended finan- 0.042 0.018 1.350 0.008 0.024 2.445** cial education (0.026) (0.026) (1.057) (0.030) (0.027) (1.135) workshop R-squared 0.197 0.375 0.202 0.254 0.253 0.170 Observations 1,022 1,059 1,016 1,063 1,273 1,239 Dependent variable mean 0.824 0.705 12.550 0.376 0.474 13.484 in control group Dependent variable SD in     14.838     17.567 control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. SD = standard deviation. The sample includes parents in treatment schools who attended either a financial educa- tion or health education workshop. Since workshop assignment was not revealed in advance, the analysis only includes parents who attended; all parents who were invited but did not attend are excluded. Data in this table are from follow-up survey 2, since the parent workshops occurred between the two follow-up survey waves. Four parent outcomes are presented in this table: a budgeting question that tests the knowledge of a budget; an indicator variable equal to 1 if the parent has formal savings such as a current account, savings account, debit card or checks; the actual fraction of monthly income that is saved; and an indicator variable equal to 1 if the parent makes a list of all monthly expenses. In addition, two student outcomes are presented: an indicator variable equal to 1 if the student has any savings; and the actual fraction of monthly income that is saved. All regressions control for baseline outcomes and include parent work- shop stratification dummies. When baseline outcomes have a missing value, they are replaced by 0 and a dummy variable indicating such missing values is included. 2.  Financial education and behavior formation   ◾ 95 ANNEX B: DETAILS OF THE FINANCIAL EDUCATION CURRICULUM 2B.1.1 General description The material provided to teachers and used in the school financial education program in Brazil includes (1) a student textbook, (2) a student exercise book, (3) a teacher guidebook, and (4) a teacher training DVD. This material was developed by the GAP. This annex describes the content of the student textbook in detail.13 The student textbook is divided into three blocks and covers nine themes. Each theme is taught through case studies/didactic situations, consisting of theo- retical and applied content, activities and self-evaluative questions. The case studies make use of texts, stories, images, and tables to convey the material in an accessible way. They also contain “experiment” sections that are designed to make the material relevant to students’ daily life, and conclude with a short outline of what is expected from the student in terms of learning. The rest of this annex describes each of these blocks and the inclusive themes. 2B.1.2 Block 1 The first block discusses three themes: everyday family life, social life, and personal property. THEME 1: EVERYDAY FAMILY LIFE In everyday family life, students are exposed to common situations where they have to make decisions that affect their family’s financial well-being. This theme is covered in seven sections, which are described below: 1. Agenda and planning —— Record expenses regularly —— Know where you spend money —— Estimate the value of items purchased 2. Calendar —— Make a list of personal and family expenses —— Classify expenses as fixed or variable —— Prepare a monthly budget separating fixed and variable expenses The student exercise books and the teacher guidebooks and DVDs were meant to support 13  the material covered in the student textbooks and did not offer new material of their own. The student exercise book provided assignments to students based on material taught through the textbook. Similarly, the teacher guidebook and DVD provided instructions on teaching and assessment methods for the course material, as well as examples of how to integrate the financial education curriculum into regular school learning. 96  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 3. House repairs —— Compare different repair cost estimates —— Compare interest rates for personal loans —— Decide whether to take a loan or use money saved previously 4. Supermarket —— How to avoid overspending on “temptation” goods —— Distinguish good and bad behaviors when going shopping —— Advance decision making on items to purchase 5. Balancing —— Distinguish and categorize personal and family expenses —— Assess the importance of these different expenses —— Identify spending categories where cuts can be made —— Prepare a 5 percent spending cut plan 6. Unforeseen circumstances —— Understand the value of insurance —— Understand the specific vocabulary of insurance products —— Identify alternative methods of prevention 7. Matching spending to earnings —— Classify income sources as fixed and variable —— Prepare a table with family incomes —— Analyze how family spends and saves money THEME 2: SOCIAL LIFE In social life, students are exposed to situations where they have to make finan- cial decisions about their personal and social lives. This theme is covered in seven sections, discussed below: 1. What a waste —— Analyze personal expenses and identify waste —— Avoid waste —— Identify actions that can lead you to spend more than necessary 2. Let’s get this party started —— Make estimates of the quantity of food and drinks necessary for a party —— Make a budget for a party —— Plan an environmentally friendly party —— Identify pitfalls when making estimates 3. To give in or not to give in to peer pressure…that is the question —— Organize financial information in a way that can be easily explained to others —— Learn and apply concepts such as interest rate, risk, and return to everyday situations 4. Buying on credit —— Identify elements of a credit card bill 2.  Financial education and behavior formation   ◾ 97 —— Identify financial behaviors that lead to credit card debt —— How to use a credit card in a responsible manner 5. Camping —— Identify expenses involved in camping —— Always keep funds for unforeseen events —— Prepare a financial plan to go camping 6. “Viva São Joao!” —— Prepare a plan for a special festivity in terms of a business plan 7. Don’t fall victim to advertising —— Identify financial pitfalls of credit card advertisements —— Analyze various options available for credit THEME 3: PERSONAL PROPERTY Students learn from situations where they have to make personal shopping deci- sions. This theme is covered in seven sections, discussed below: 1. In search of the perfect shoes —— Calculate the difference between the price paid in cash and the one paid with credit —— Decide if it is better to pay in cash or with credit —— Find specific information in the Consumer Defense Code 2. Computer —— Identify the opportunity cost of owning a computer —— Balance wants and needs when choosing a computer —— Compare prices —— Calculate the necessary savings in order to buy a computer 3. Digital camera —— Follow similar steps as with purchasing a computer 4. If by magic… —— Identify the elements of advertising aimed at generating consuming desire —— Identify the conflict between desires and needs —— Be wary of temptation traps and impulsive spending 5. Mobile phone —— Choose a mobile phone that best fit your needs —— Choose a plan that best fits your needs —— Understand your mobile phone bill 6. Consumer protection measures —— Identify cases of abusive practices and consumer rights violations —— When to reach out to the Foundation for Protection  and Consumer Advocacy (PROCON) 7. Changing money —— Convert the value of products priced in foreign currency to local currency 98  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES —— Know how the value of a credit card purchase in foreign currency appears in local currency on the bill 2A.1.3 Block 2 The second block discusses three themes: work, entrepreneurship, and large expenditures. THEME 4: WORK In work, students discuss several aspects of their current and future professional lives. This theme is covered in seven sections, discussed below: 1. What line of work? —— Identify the type of work that attracts you the most —— Decide on the most appealing type of job according to your life ambitions 2. First job —— Prepare a CV —— Identify the skills that are compatible with advertised positions —— Appropriately highlight your professional qualities in a simulated job interview —— Combine your desired job with the type of life you want to have 3. Gross versus net income —— Differentiate gross and net income —— How to explain this concept to others 4. Good times and bad times —— Understand the concept of structural unemployment in a made-up story —— Identify measures to overcome unemployment 5. The incredible case of the 13th salary that disappeared —— Make a budget based on data and estimates —— Make a simulated financial plan in order to achieve a positive balance at the end of the month —— Consider future situations in current monthly planning 6. Lifelines —— Prepare an outline of a retirement plan, harmonizing long-term goals and the means to achieve them 7. Antenor, the wary employee —— Develop a product and message for an information campaign about insurance —— Utilize the vocabulary of insurance as it applies to an information campaign THEME 5: ENTREPRENEURSHIP In entrepreneurship, students learn about practical issues of creating and running a business. This theme is covered in seven sections, discussed below: 2.  Financial education and behavior formation   ◾ 99 1. A great idea —— Differentiate entrepreneurs driven by necessity and by opportunity —— Relate own characteristics with business opportunities —— Identify needs in own community that may generate a business oppor- tunity —— Brainstorm to generate good business ideas 2. What are your talents? —— Distinguish between knowledge, skills, attitudes, and competencies in the context of entrepreneurship —— Evaluate if you possess the necessary knowledge to open a particular business —— Evaluate if you possess the necessary skills to open a particular business
 —— Evaluate if you possess the necessary attitudes to open a particular business 3. Profession: entrepreneur —— Identify the characteristics of an entrepreneur —— Differentiate entrepreneurship from intrapreneurship —— Test if you have the profile of an entrepreneur 4. The soul of a business —— Identify the target audience of a fictitious business —— Create a brand and slogan for a fictitious product or service —— Put together a fictitious marketing plan —— Carry out market research for a fictitious product or service 5. Hands to work —— Identify the resources necessary to open and run a business —— Budget for opening and running a fictitious business —— Determine the knowledge, skills, attitudes, and competencies of the personnel necessary to work in a fictitious business 6. Victory —— Make sales and profit projections for a fictitious business —— Measure the profit of a fictitious business —— Cut costs and expenses related to products or services of a fictitious business 7. Beyond profit —— Distinguish between philanthropy and socio-environmental responsibility —— Make a plan of socio-environmental responsibility for a fictitious business —— Put together in a business plan all the information on entrepreneurship learned in this theme THEME 6: LARGE EXPENDITURES In this theme, students are exposed to situations that involve significant financial outlays. This theme is covered in seven sections, discussed below: 100  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 1. Brick by brick —— Balance the desires and needs of your family when choosing a house to purchase —— Search for information on prices and financing for a house —— Decide how much your family is willing to spend as a function of the household budget —— Plan financially for the down payment and installments of a home mort- gage 2. Surprise —— Create a budget for a party —— Plan a party that suits your financial situation —— Make provisions for unforeseen expenses —— Cut expenses according to your priorities 3. In your corner —— Make investment decisions in a simulated market situation —— Make an initial investment decision, taking into consideration family and personal preferences 4. She speaks about the same thing all day long —— Identify rights and duties that are not being met in a certain situation —— Generate arguments to debate rights and duties of investors 5. Consumption and savings —— Make consumption and savings decisions in a simulated situation 6. Now it’s my turn to help my parents —— Decide between two debt application options, taking interest rates into account —— Explain how to avoid indebtedness —— Come up with options to pay off a debt of R$1,000 (US$583) 7. How much distance separates you from your future —— Estimate fixed and variable expenses in order to study in another city —— Calculate the monthly income necessary to study in another city —— Make a financial plan to study in another city 2A.1.4 Block 3 The third block discusses three themes: public goods, the country’s economy, and the world economy. THEME 7: PUBLIC GOODS In public goods, students address several issues concerning the use and financing of public goods and services. This theme is covered in seven sections, discussed below: 1. Everything has a price 2.  Financial education and behavior formation   ◾ 101 —— It is always the case that someone pays for the public goods you consume for free —— Calculate how much the government spends to sustain a high school class in a public school 2. School budget —— Think about the school and its budget —— Suggest improvements to the school that are feasible 3. School books —— Identify the reasons for the high environmental cost of school books —— Calculate the consumption of paper in school —— Identify actions that can save paper —— Develop and engage in a campaign to save paper 4. Public spaces —— Everyone has the right to access free public spaces —— Maintenance of public spaces is costly and is paid for through taxes —— The individual tax burden can be reduced if all citizens pay their taxes —— Consult the community in order to know which public spaces need to be improved 5. Public services —— A public budget is very similar to a family budget —— The legislature—senators and deputies—decides the public budget —— Link the public duty to pay taxes with the government’s duty to provide public services 6. Corruption —— Corruption affects the lives of everyone because it reduces the money the government can to invest in public services —— Check public accounts through public records 7. Taxation —— Link the public duty to pay taxes with the government’s duty to provide public goods and services —— Understand the purpose of different taxes paid by citizens —— Develop and engage in a campaign to provide incentives for citizens to pay their taxes THEME 8: THE COUNTRY ECONOMY Students are exposed to several aspects of their country’s economy that are rele- vant to their personal lives, including the concept of inflation, the law of supply and demand, the concept of minimum wage, and the basics of the national finan- cial system. This theme is covered in seven sections, discussed below: 1. Culture and sports —— Prepare an outline of a project for a cultural or sporting activity 102  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES —— Align the objectives of a project to the Rouanet Law (law that provides tax incentives to private firms for supporting cultural activities) —— Understand the sections of the laws concerning education that contain financial vocabulary 2. Inflation —— Recognize the problems that inflation can generate when changes in income do not follow the increase in prices —— Make adjustments in the family budget taking inflation into consideration —— Explain the concept of inflation to someone else 3. Supervisors of the national financial system —— Explain the national financial system to someone else 4. Speaking in economic terms —— There exists a direct relationship between the nation’s economic growth and the growth of a family’s personal income —— Families with low income can also organize themselves financially 5. Markets —— Demand and supply simulations 6. Foresight —— Estimate the income and expenses of a retired person —— Prepare a simulated financial plan for a retired person 7. Minimum wage —— Research prices to estimate the total value of the basic needs of a person —— Link the value of the basic needs of a person with the value of the minimum wage THEME 9: THE WORLD ECONOMY Students are exposed to several aspects of the world economy that are relevant to their personal lives, such as the concept of imports and exports, international economic blocks, and measures of a country’s well-being. This theme is covered in seven sections, discussed below: 1. Special issue on money —— Contextualize the role of money in society —— The importance of saving money 2. International cooperation —— Identify the complexities involved in international negotiations —— International economic blocs organize themselves through arrange- ments that are negotiated 3. The game of economic blocs —— Experience, in a game, some of the issues concerning international economic blocs —— Think about simulated strategies of global resolution of conflicts 2.  Financial education and behavior formation   ◾ 103 4. The business of China —— Identify the imported products that you use in your everyday life —— Locate the countries from which the imported products you use in your everyday life come —— Search for data on national and international exports 5. International organizations —— Reflect upon the profile and performance of representatives of a country in an international community —— Develop a funding proposal for an international financial institution 6. The well-being of your country —— Compare the Index of Human Development with the GDP per capita for different countries —— Link the economic performance of a country with its environmental impact 7. Moment of crisis: do I care? —— Establish the relationship between an economic crisis and situations in your personal life —— Identify ways to overcome the impact of an economic crisis for individuals CHAPTER 3 U nderstanding and improving household investment behavior in Brazil A stock market simulator as a learning-by-doing financial literacy tool DENIZ ANGINER, AIDAN COVILLE, VINCENZO DI MARO, MARTIN KANZ, ARIANNA LEGOVINI, CAIO PIZA, AND ASTRID ZWAGER ABSTRACT As equity markets develop rapidly around the world, especially in emerging markets, investing in the stock market is now within reach of a growing segment of the population that did not previously have access to these finan- cial instruments. This opens new opportunities for households to diversify their assets and investments, and at the same time provides liquidity to the market. However, this also introduces potential for consumer risk and capital market volatility if inexperienced investors make uninformed or irrational investment choices. This study explores the effectiveness of an online stock market simulator as a tool to overcome investor biases and improve perfor- mance through a “learning-by-doing” training experience. Using data from about 40,000 simulator participants, we find (1) mixed effects for exposure to the stock market simulator on behavioral biases; (2) a weak relationship between behavioral biases and performance, and training in the simulator and performance; and (3) a strong link between extended participation in the simulator and transition into the actual stock market. The results suggest that learning by doing can motivate investors to participate in the stock market. The authors want to thank Alcides Ferreira, Christiane Barriquelli, Mirella Santolia, and Mariana Vieira (BM&FBOVESPA) for having created the motivation for this research program and for having shared the data without which this study would not have been possible. Also, we thank Richard Hinz and Florentina Mulaj (Human Development Social Protection and Labor) for their encouragement and comments on this study. Financial support from the Russia Financial Literacy and Education Trust Fund, and the World Bank Brazil Country Management Unit are gratefully acknowledged. The list of authors includes all the researchers involved in the broader research program of which the present study is the first output produced. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 105 106  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 3.1 INTRODUCTION AND MOTIVATION In the next 20 years, an estimated 2 billion people will enter the formal financial system (Demirgüç-Kunt, Beck, and Honohan 2008). Access to a broader range of financial instruments offers households an opportunity to diversify their assets. As financial access improves, it becomes increasingly important to understand the mechanisms that drive household investment choices. This is important both to improve consumer protection (ensuring that investors make fully informed decisions that do not expose them to unintended risk) and to fulfill a broader economic development objective of improving savings rates to drive investment and strengthen capital markets. Drawing from the current literature on behav- ioral finance, this study has been designed together with the Brazilian Securi- ties, Commodities, and Futures Exchange (BM&FBOVESPA) to test a range of interventions to understand why people make the decisions they do and identify how to reduce common investor biases observed in the stock market to improve consumer protection and ultimately increase capital market efficiency. In this chapter, we present an analysis of BM&FBOVESPA’s “learning-by-doing” stock market simulator—a teaching tool to familiarize new investors with the stock market; however, this represents one part of a larger evaluation aimed at identi- fying and overcome a broad range of stock market barriers to entry and investor behavioral biases that is outlined in more detail in annex A to this chapter. 3.1.1 New investors create new challenges During the transition to deeper capital markets, two concerns arise. First, despite the potential value in diversifying savings and investment products to spread risks and maximize returns, households often choose not to, exhibiting suboptimal and irrational investment choices. Evidence on savings and investment holdings in the United States shows that, despite the potential equity premium that can be achieved through participation in the stock market, and the predictions from expected utility models indicating that almost everybody should hold stocks described by Haliossis and Bertaut (1995), nearly 20 percent of households at the 80th percentile of the wealth distribution do not invest in public equity (Camp- bell 2006). As wealth increases, understanding the motivation behind household choices to climb the savings and investment ladder from liquid savings accounts to bonds, property, and equity will help tailor interventions aimed at strengthening financial inclusion and ensuring that households maximize the benefits provided through access to the right investment instruments. Conversely, overconfident or ill-informed investors may choose to begin using new financial instruments prematurely and expose their portfolios to unintended risk. It is thus important to make sure that investors on both ends of the spectrum are investing wisely, and ensure that adequate information and support are provided to facilitate these investment decisions. 3.  Understanding and improving household investment behavior in Brazil  ◾ 107 The second concern is that common investor biases may reduce the oppor- tunity for investors to use financial instruments to enhance their own welfare. Evidence shows that overconfident or ill-informed investors with limited financial literacy are much more likely to make costly investment mistakes; they (1) trade too frequently, (2) are not sufficiently diversified, and (3) sell winning stocks too early and hold onto losing stocks for too long (Daniel, Hirshleifer, and Teoh 2002). New investors could reduce their susceptibility to these common investment mistakes by investing amounts in diversified instruments such as index funds. As the range of financial products for potential investors steadily broadens, these opportunities bring with them increased risks and potential pitfalls that have consequences both for household welfare and the stability of financial markets more broadly. However, few new investors know about these financial products or their benefits for risk management. There is thus significant scope for finan- cial education to ensure that the right investment products are chosen and used effectively to allow for calculated risk taking while reducing costly investment mistakes among new investors. For instance, the possibility of buying products automatically diversified (like exchange traded fund [ETFs]) and/or products with characteristics that deter high-frequency trading (e.g., increasing trading fees) may help investors overcome these biases. 3.1.2 Savings and investment in Brazil The study focuses on Brazil, a context in which these issues are particularly relevant. Indeed, over the last two decades, Latin American countries, including Brazil, have been at the forefront of reforms to foster capital market development. Despite the reform effort, capital markets in Latin America remain underdeveloped relative to other regions with similar economic conditions (De la Torre, Gozzi, and Schmukler 2007). Total value of stocks traded as a percentage of gross domestic product (43.2 percent in Brazil) falls short of comparable emerging economies like China (136.6 percent) and South Africa (93.5 percent).1 In particular, Brazil faces the challenge of a culture of high short-term consumption and indebtedness— even Easter eggs can be bought in installments (Economist 2011). Notably, Brazil’s savings rate has fluctuated at around 17 percent of gross domestic product over –50 percent. the last decade, a number that contrasts sharply with China’s 45­ As the Brazilian economy grows, understanding how to introduce a new class of investors to the stock market, while at the same time ensuring that investors are not exposed to unnecessary risks, will become critically important. Under- standing constraints to participation as well as common behavioral investment biases within the market can help inform market participants and policy makers in Brazil and other emerging economies in developing consumer protection policies Figures for capital market development indicators are from the World Bank’s World Devel- 1  opment Indicators. 108  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES that will increase access to new financial tools while mitigating the risks associ- ated with poor investment decisions. 3.1.3 The research agenda Against this backdrop, a large research program has been designed to address a core research question: What interventions are most effective at improving participation in the stock market and reducing biases in investor behavior? In particular, a broad program of impact evaluation studies has been designed to address and disentangle the different constraints and market failures hypothe- sized to be affecting investor participation and behavior, including physical (finan- cial education, financial and nonfinancial entry costs, transaction costs, access) and psychological constraints (such as self-control, time-inconsistent preferences, inertia, menu costs, peer effects). The findings of the study will add to the liter- ature by measuring the effectiveness of new modes of financial education. The study will test multiple hypotheses regarding barriers to effective participation in financial markets within a unified framework, leading to better understanding not only of which investor biases are most prevalent but also how targeted education and product offers can mitigate these biases, as well as illustrating a framework in which impact evaluations can be used to make real-time operational decisions in the private sector. This research program has three main components: (1) testing the effect of an online stock market simulator as a learning-by-doing financial literacy tool to improve knowledge and investor behavior; (2) reducing the barriers to stock market participation offers to randomly selected groups of potential investors participating in the simulator; and (3) providing financial information and details of investment products through a stock broker online platform to individual inves- tors with the aim of reducing investor biases. This chapter presents and discusses the first set of results from this research program. In particular, it focuses on the first component, testing the effect of an online stock market simulator as a learning-by-doing financial literacy tool to improve knowledge and investor behavior.2 BM&FBOVESPA offers people the opportunity to practice trading stocks in an environment that replicates real-time stock data (with a short delay) and simulates the experience of participating in an online “home broker” system that allows real investors to trade directly from their home computer through a stock broker inter- mediary. The simulator, called Simulaçao, includes a wide range of investment products and offers investors the opportunity to build portfolios from current stock options. Participants receive a start-up account with fictional Brazilian reais in order to participate. The motivation for developing this simulator tool was to Please refer to DIME (2012) for the details of the other components of the research 2  program. 3.  Understanding and improving household investment behavior in Brazil  ◾ 109 overcome informational barriers to participation by exposing potential investors to the activities related to investing in the stock market without exposing them to any real risk—and in so doing, educate potential investors through a learning-by- doing experience on how to participate in the stock market. The chapter studies the investment behavior of simulator users using data from Simulaçao in 2011 and 2012. Matching these with data from Brazil's actual stock market for the same period offers a unique opportunity to track both expo- sure and performance in Simulaçao along with transition rates into the actual stock market. In particular, we first analyze basic descriptive investor behavior in the simulator. We then report on the presence and extent of common investor biases, the details of which are explained in the following section. The main objec- tive of the analysis is to assess how the use of the simulator influences trading decisions and simulated portfolio outcomes to see if simulator participation translates into improved investment decisions. Finally, we study how the use of the simulator and behavior within the simulator environment affect entry into the stock market. Our results identify correlations between increased exposure to the simu- lator and behavioral biases, portfolio performance, and transition into the actual stock market. We find that higher exposure to the simulator has mixed effects on users’ behavioral biases. The results indicate that higher exposure to the simulator may have posi- tive effect on user performance (in the simulator), although the evidence is not robust. The relationship between behavioral biases and portfolio performance is mixed, and the results suggest that selection biases might be driving the esti- mated correlations. Finally, high exposure to the simulator as well as a high return in the 2011 simulator is associated with a higher probability of users entering the stock market. Our estimates suggest that the experience with the simulator may have a relevant effect on take-up. The chapter is organized as follows. In the next section, we provide a litera- ture review to motivate the questions and interventions. In particular, we review the available evidence on education constraints to financial market participation, and investment biases among small households with limited financial literacy. We then describe the online stock market simulator Simulaçao. The following section poses the main research questions and describes the methods used to address them. We follow this by describing the simulator and actual stock market data used in this chapter. The last two sections present and discuss the results and draw some policy conclusions based on these findings. 110  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 3.2 WHAT DRIVES INVESTMENT DECISIONS: A REVIEW Measuring household investment is challenging, and the data requirements for accurate measurement are often hard to collect. Many households have compli- cated finances, with multiple accounts at different financial institutions facing different tax treatments and possibly including illiquid investments such as ownership in businesses and real estate in addition to traditional publicly traded investments such as mutual funds, individual stocks, and bonds. Even households that wish to provide data may have some difficulty answering detailed questions accurately. In what is perhaps the most comprehensive review of the financial decisions of households available to date, Campbell (2006) presents several inter- esting facts based on data from the 2001 U.S. Survey of Consumer Finances. Investment choices are strongly correlated with education, age, income, wealth, and gender. The influence of wealth on investment choices is illustrated in figure 3.1 (taken from Campbell 2006). Households initially use their savings for “safe assets” such as savings accounts, and progress to vehicles and real estate; while investment in equity has an approximately linear relationship with wealth percentiles. Particular interest has been in understanding the “participation puzzle” highlighting that, although economic theory suggests that all households should hold at least some portion of stocks in their portfolio, the empirical evidence runs contrary to these predictions. Stock market participation is extremely low FIGURE 3.1  Rate of participation by wealth and asset class Percent 100 Safe assets 80 Real estate Private business 60 Vehicles Public equity 40 20 0 0 20 40 60 80 100 Percentile distribution of total assets Source: Campbell 2006. 3.  Understanding and improving household investment behavior in Brazil  ◾ 111 (less than 50 percent) among households below the 50th percentile of the wealth distribution. But perhaps more strikingly, the evidence indicates that even some of the wealthiest households do not hold stocks (for example, 20 percent of U.S. households at the 80th  percentile of the wealth distribution do not own any stocks). While similarly detailed data are not available for an emerging market, descriptive evidence suggests that these patterns are much more pronounced in less developed financial markets. This lack of financial diversification implies significant welfare losses due to limited financial market participation. Understanding market frictions and deviations from expected utility maximi- zation have come to the foreground as explanations for the observed empirical departures of household investment decisions from the theoretical predictions of how households should allocate their assets. While people respond to financial and nonfinancial incentives and barriers to entry (in line with utility maximization theory), it is also clear from the research in behavioral economics that individuals deviate from the standard (neoclassical) full information choice model by exhib- iting nonstandard preferences, beliefs, and decision making (DellaVigna 2009). At the conceptual level, an important distinction needs to be made between the choices that lead to market entry and participation, and the drivers of asset allocation and investment decisions once a household participates in the financial market. The broader impact evaluation explores the effect of (1) financial and time costs, including fixed entry, variable trading, and per period costs; (2) financial literacy/knowledge; and (3) psychological barriers to stock market participation, while further exploring behavioral biases within the stock market and the effect of these on performance (see annex A). In this chapter, we specifically explore the effectiveness of the stock market simulator as a learning tool to overcome knowledge constraints to stock market participation and mitigate investor behav- ioral biases. 3.3 BARRIERS TO PARTICIPATION IN FINANCIAL MARKETS To understand the role of information and financial education in overcoming barriers to participation in financial markets, our study draws on existing research on barriers to stock market participation and the adoption of less complex finan- cial instruments (such as savings accounts and insurance products). 3.3.1 Financial education Although a number of nonexperimental studies have been conducted in the past, until recently there has been very little rigorous evidence on the causal relation- ship between financial literacy and consumer behavior when it comes to invest- ment choices. Strong correlations between financial literacy and participation in the stock market are well known (Van Rooij, Lusardi, and Alessie 2011); however, experimental evidence is not as compelling—leading to an effort to find new 112  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES methods for delivering financial messages, such as the use of computer games for children to learn how to run a business. The value of practical training as a component of learning has been emphasized since the seminal work of Dewey (1938). Despite this, many programs aimed at promoting financial capabilities do not explicitly incorporate practical training into their program, as this is often seen as an expensive and logistically challenging component of teaching and little rigorous evidence exists as to its relative value. A recent study of financial education in Brazilian schools—conducted through a collaboration with several Brazilian financial institutions and the Development Impact Evaluation's Devel- opment Economics Research Group of the World Bank—is the first of its kind, highlighting how effective a financial literacy program can be when theory and messages are linked directly to practical activities to promote “learning by doing,” finding significant increases in student knowledge and savings rates (Legovini, Bruhn, and Zia 2013). In another example of exploring new, effective ways to present financial information, evidence from a randomized study aimed at using financial informa- tion to deter people from using payday lenders showed strong results when the messages were linked to behavioral cues (Bertrand and Morse 2010). For example, by expressing interest rate repayments as a dollar amount, framing the repay- ment over a longer time horizon, and comparing this to alternative credit options, payday lender use decreased by 5.9 percent and average loan size decreased by US$55, or 23 percent. 3.3.2 Investor biases The literature has identified several biases that affect investor behavior, resulting in suboptimal investment strategies when participating in the stock market. Here we cover only a sample of the most common biases observed. The disposition effect—the tendency of individual investors to hold onto losing stocks and sell winners too quickly to realize gains—has been shown to be an important driver of individual trading decisions. Kahneman and Tversky’s (1979) prospect theory, combined with Thaler’s (1981) “mental accounting’’ frame- work, have been considered the leading explanation for the disposition effect. The main element of this theory is a utility function that is concave (risk averse) in the domain of gains and convex (risk loving) in the domain of losses, both measured relative to some reference point. Mental accounting provides a way in which investors set reference points for the accounts that determine gains and losses. The main idea is that decision makers tend to segregate different types of gambles into separate accounts, and then apply prospect theory to each account, ignoring possible interactions. Investors with a disposition effect have been shown to underperform investors who do not exhibit these biases (Seru, Shumway, and Stoffman 2010). 3.  Understanding and improving household investment behavior in Brazil  ◾ 113 Although financial theory prescribes investors to hold diversified portfolios, many retail investors hold concentrated portfolios with only a handful of stocks (Barber and Odean 2000). Underdiversification has been linked to a lack of sophis- tication (Goetzmann and Kumar 2008) and preference for skewness, but to date the evidence on this relationship is still mixed (see, e.g., Cremers and Petajisto 2009). Prior research suggests that retail investors can substantially reduce their risk (without a reduction in returns) by holding more diversified portfolios (Goetz- mann and Kumar 2008). Retail investors trade actively (without a gain in performance), even though standard finance theory suggests a buy and hold strategy when investors do not have private information. Active trading has been linked to overconfidence, whereby investors overestimate either their abilities or the quality of their private knowledge. Overconfident investors trade excessively, paying high transaction costs without additional gains from active trading (Barber and Odean 2000). Retail investors also exhibit a preference for local stocks. They tend to hold a dispro- portionately large portion of their investments in domestic stocks, stocks head- quartered in geographical locations in which the investor resides, and stocks in the sector in which the investor works. The preference for local stocks has been linked to familiarity bias, whereby investors overweight local stocks because they are more familiar and comfortable with them. While these biases are well known, there has, to date, been no systematic randomized controlled trial designed to test the effectiveness of targeted infor- mation and investment offers aimed at mitigating these biases and lengthening investor investment horizons in the stock market. 3.4 LEARNING BY DOING VERSUS LEARNING BY LOSING The literature that delves into the relationship between financial expertise (experi- ence) and performance in the stock market has identified the three most common mistakes (or behavioral biases) investors make in the stock market: underdiver- sification (more risk for a given return), high turnover (high trading costs), and disposition effect (see Campbell, Ramadorai, and Ranish 2012). It is argued that most of these mistakes can be avoided, and that learning from experience seems to be an effective device to minimize mistakes. Campbell, Ramadorai, and Ranish (2012) use data from the Indian equity market to investigate whether investors’ experience affects behavioral biases and performance. Their main findings are the following: (1) “account performance improves with account age,” (2) holders of older accounts are less prone to incur behavioral biases, and (3) underperfor- mance due to behavior mistakes tends to reduce in the long run. All three findings point to a positive effect of experience on reducing behavior biases and improving performance. 114  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES This chapter follows similar steps as it aims to uncover the relationship between behavior biases and performance using similar indicators. However, unlike all evidence available to date, the objective here is to test whether prac- ticing in a fake stock market (the simulator) helps users avoid those common mistakes and perform better. The idea is to check whether the simulator can be regarded as an effective learning tool by individuals who most likely do not have a strong financial knowledge. In this scenario, individuals can try different things without incurring a real loss. In fact, this could stimulate the simulator users to adopt strategies they would not follow in the real world. Therefore, it must be remembered that the structure of incentives individuals face in the simulator and in the actual stock market are markedly different. Despite the different incentives—and our analysis does not consider a time dimension that could allow one to control for users’ fixed effects—some of our findings are in line with Campbell, Ramadorai, and Ranish (2012), as discussed below. 3.5 THE STOCK MARKET SIMULATOR: SIMULAÇÃO BM&FBOVESPA offers people the opportunity to practice trading stocks in an environment that replicates real-time stock data (with a short delay) and simu- lates the experience of participating in an online “home broker” system that allows real investors to trade directly from their home computers through a stock broker intermediary. Simulaçao includes a wide range of investment products and offers investors the opportunity to build portfolios from current stock options. Participants receive a start-up account with fictional Brazilian reais in order to participate. The motivation for developing this simulator tool was to overcome informational barriers to participation by exposing potential investors to the activ- ities related to investing in the stock market without exposing them to any real risk—and in so doing, to educate potential investors through a learning-by-doing experience in how to participate in the stock market. The simulator is complemented by an array of other financial literacy activ- ities BM&FBOVESPA offers potential investors with the aim of teaching people how to use the stock market to make better use of their money. These activities are organized under an umbrella website called Quer ser socio?3 3.6 RESEARCH QUESTIONS AND METHODS From the literature, we see that, beyond demographics and wealth levels, selec- tion into the stock market may be driven by lack of knowledge/awareness of 3  See http://www.simulacaobmfbovespa.com.br/ and http://www.quersersocio.com.br/. 3.  Understanding and improving household investment behavior in Brazil  ◾ 115 the value and optimal use of financial products linked to equity markets. In addi- tion, investment biases such as overconfidence, underdiversification, and the disposition effect play an important role both in people’s portfolio decisions as well as their investment performance; this may be mitigated through experience. Financial education through experience can potentially reduce information barriers related to stock market participation and protect small investors from making standard investment mistakes that expose them to increased risk. Thus, improved investor behavior has potentially far-reaching consequences by reducing exposure to unintended risk and improving investor risk-adjusted returns, while promoting economic development by strengthening capital markets through the more efficient distribution of resources. Against this backdrop, the study poses the following main research ques- tion: How effective is a learning-by-doing financial education tool at improving knowledge and investor behavior, and reducing barriers to participation in the stock market? In particular, the study tests whether increased simulator expo- sure is associated with improved investment decisions and a higher probability of entering the stock market. The chapter studies the investment behavior of simulator users using 2011 data from Simulaçao. In addition, data from the actual stock market in Brazil is also available for the same period covered by the simulator data and used to define our evaluation strategy. We run three main specifications: Yi = μ + β1Sim + X’iγ + εi (3.1) where Yi refers to the investor’s behavioral bias—disposition effect or diversifica- tion—and Sim corresponds to an individual’s exposure to the simulator. Here we consider three measures of exposure: number of trades in the simulator, number of days traded in the simulator, and length of use of the simulator. The vector of controls Xi includes an individual’s age, gender, and dummy for state of residence; and εi is the error term. The objective of this exercise is to check whether there is a correlation between exposure to the simulator and the common behavioral biases analyzed in this chapter. The hypothesis we will be testing is that individ- uals with a higher exposure to the simulator would be less likely to incur behavior biases. The second model tests how exposure to the simulator and behavioral biases affect the returns and risk of portfolio holdings in the simulator. We use two measures of returns: the adjusted returns and the Sharpe ratio. The model is specified as follow: Ri = μ + β1Sim + β2Yi + X’iγ + εi (3.2) where Ri is a measure of portfolio risk, and Sim, Yi, Xi, and εi are defined as before. The returns are taken as a proxy of how these individuals would have performed in the actual stock market had they behaved in the same way as they did with 116  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES the simulator. Finally, we estimate participation equations to test whether the simulator encouraged people to invest in the actual stock market. To make things simple, we run a simple ordinary least squares regression to estimate a linear probability model. Although the model has some shortcomings, such as the linearity assumption and the fact that the predicted probabilities can go beyond the [0, 1] range, the coefficients of the regression give the marginal effects straightforwardly. Pi = μ + β1Sim + X’iγ + εi (3.3) where Pi is an indicator function that takes the value of 1 if individuals used the simulator in 2011 and are observed in the actual stock market in 2012 (but they were not in the stock market in the previous years 2009–11); and 0 if they used the simulator in 2011, but have not been observed in the stock market in 2012 or previously (2009–11). Although the analysis does not allow us to follow the same individuals over time, we can at least compute variables that inform us of the intensity of use of the simulator. In that sense, all these variables have a time component as they indicate, for instance, how many trades a user made with the simulator, and the number of days a user traded on the simulator. We are therefore able to tell whether those individuals who used the simulator more intensively in 2011 performed better or worse than those who used it less. 3.7 DESCRIPTION OF THE DATA The subsample contains information on each individual simulated transaction during 2011. Consistent with the literature, we aggregate the intraday transac- tions per user and firm. For instance, if in a given day a user traded the stock of a certain company more than once, we computed the net balance of that daily transaction so as to have only one daily observation per user and company. In total, around 517,000 transactions were made by about 40,000 individuals; this amounts to an average of 12.7 transactions per individual. Not all individuals registered for the simulator used it; indeed, about two-thirds of the registered individuals in 2011 ended up using the simulator. The database analyzed in this chapter is the subsample of registered individuals who used the simulator in 2011. Three measures were developed to capture exposure to the simulator: (1) the number of days traded in the simulator; (2) the number of trades made in the simulator; and (3) the length of use of the simulator, given by the difference between the last date the user was active on the simulator and the first date he or she used the simulator. We then compute three behavioral biases and investor characteristics that have been shown empirically to be associated with lower investment perfor- mance. These are the disposition effect, underdiversification, and overconfidence 3.  Understanding and improving household investment behavior in Brazil  ◾ 117 proxied by overtrading. The disposition effect is a behavioral tendency of inves- tors to sell shares whose price has increased, while holding onto shares that have dropped in value—a manifestation of loss aversion. Underdiversification is a tendency of investors to hold nondiversified portfolios, subjecting them to uncompensated idiosyncratic risk (Goetzmann and Kumar 2008). Overtrading (or turnover), a signal of overconfidence, appears when investors trade too much, incurring excessive trading costs without any additional gain in performance (Barber and Odean 2000). Table 3.1 provides descriptive statistics on the subsample and is divided in three blocks. The first block reports statistics for all users, and the second and third blocks present statistics for (1) individuals who used the simulator in 2011 but did not invest in the actual stock market in 2012, and (2) individuals who used the simulator and invested in the actual stock market in 2012. The first three rows show the behavioral biases variables computed with data from the simulator and from the stock market, both in 2011. According to the table, the average user shows a turnover of 17 percent, tends to hold a relatively well-diversified portfolio, and has a disposition effect of 10 percent. The numbers in the second and third blocks are similar, but there is an indication that individ- uals who invested in the actual stock market have a high turnover rate. Since this variable captures overconfidence, this result is not really unexpected: it seems to suggest that those who are overconfident would find it easier to make the transition from the simulator to the actual stock market. As for the diversification variable, because of the way this variable is constructed, the higher the value of this variable, the more concentrated a user’s portfolio. The following four rows summarize exposure to the simulator. The number of trades tells how many times a user set up an order (or transaction) in the system. By taking into account only the executed orders, this variable provides an underestimate of the exposure of the user to the simulator.4 Based on table 3.1, an average user sets up about 17 orders; given that this variable is highly posi- tively skewed, the median number of trades in the simulator is about half of that (8 trades). Figure 3.2 illustrates the distribution of this variable for male and female users, with 86 percent of the users being male. The figure shows that men used the simulator more often than women. This holds for the entire distribution. The distribution of trades for male users exhibits first-order dominance over the female distribution. As with the turnover variable, there is an indication that users who invested in the stock market used the simulator more frequently. The other two variables that summarize user exposure to the simulator are the total number of days a user traded in the simulator, and the length of use of the simulator. The latter is computed as the difference between the first and last We were unable to retrieve reliable information on the number of canceled, expired, 4  received, and rejected orders. 118  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 3.1  Descriptive statistics for simulator users OBSERVA- STANDARD TIONS MEAN DEVIATION MINIMUM MAXIMUM All simulator users Turnover 38,666 0.17 0.19 0.00 1.00 Diversification 39,757 0.58 0.31 0.04 1.00 Disposition 2,350 0.09 0.16 −0.48 0.65 # of orders 40,573 17.08 28.30 1.00 808 # of days 40,573 5.40 7.76 1.00 150.00 Length of use 40,573 35.27 62.44 0.00 360.00 # of stocks 39,757 2.86 2.33 1 39 Age 39,927 27.65 8.17 16.00 70.00 Male 40,573 0.86 0.35 0.00 1.00 Used simulator but not in the actual stock market in 2012 Turnover 35,286 0.17 0.19 0.00 1.00 Diversification 36,289 0.58 0.31 0.04 1.00 Disposition 2,069 0.09 0.16 −0.48 0.65 # of orders 37,044 16.46 26.81 1.00 808 # of days 37,044 5.09 6.94 1.00 150.00 Length of use 37,044 32.58 58.69 0.00 360.00 # of stocks 36,289 2.86 2.31 1 38 Age 36,425 27.27 7.94 16.00 70.00 Male 37,044 0.85 0.36 0.00 1.00 Used simulator and in the actual stock market in 2012 Turnover 3,380 0.22 0.22 0.00 1.00 Diversification 3,468 0.57 0.32 0.05 1.00 Disposition 281 0.08 0.14 −0.33 0.44 # of orders 3,529 23.61 40.16 1.00 589 # of days 3,529 8.70 13.17 1.00 145.00 Length of use 3,529 63.51 88.30 0.00 360.00 # of stocks 3,468 2.86 2.46 1 39 Age 3,502 31.67 9.38 16.00 70.00 Male 3529 0.94 0.24 0.00 1.00 dates a user traded using the simulator. This measure is sometimes taken as a proxy of investor experience.5 On average, a user traded in the simulator about 5 days for an average period (length of use) of 35 days. Since the distribution of both variables is positively Goetzmann and Kumar (2008) used a variable defined very similarly as a proxy for invest- 5  ment experience. 3.  Understanding and improving household investment behavior in Brazil  ◾ 119 FIGURE 3.2  First-order stochastic dominance for the number of trades in the simulator Number 150 100 Male 50 Female 0 0 20 40 60 80 100 Sample proportion Source: BOVESPA-Simulator database 2011. FIGURE 3.3  First-order stochastic dominance for the length of use of the simulator (in days) Length 300 200 Male 100 Female 0 0 20 40 60 80 100 Sample proportion Source: BOVESPA-Simulator database 2011. FIGURE 3.4  First-order stochastic dominance for the number of days traded in the simulator Days 40 30 20 Male 10 Female 0 0 20 40 60 80 100 Sample proportion Source: BOVESPA-Simulator database 2011. 120  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES skewed, the median of the number of days and the length of use are lower than the mean—3 and 7, respectively. Figures 3.3 and 3.4 present the distribution of both variables for male and female users. As before, the distribution for male users exhibits first-order stochastic dominance. These findings suggest that men traded not only more times, but spent more days with the simulator in compar- ison to women. Table 3.1 also reports the total number of stocks held by an investor to capture portfolio diversification (see, e.g., Goetzmann and Kumar 2008). As can be seen, the average number of stocks held by a user is exactly the same in all three blocks. With regard to the demographics, it can be seen that the users who invested in the stock market in 2012 were relatively older and predominantly male. Inter- estingly, only 6 percent of the users who invested in 2012 were female, a propor- tion that is substantially lower than 15 percent—the rate of female participation in the simulator. Finally, to compute returns within the simulator, we use stock market infor- mation obtained from Compustat Global and match them to all stocks traded in the simulator using their corresponding ISIN (International Securities Identifica- tion Number) codes.6 The Compustat Global database provides daily stock price information adjusted for dividends, splits, and repurchases for both active and delisted companies. It accounts for 98 percent of global stock market capitaliza- tion. For the analysis, we use risk-adjusted returns computed as the difference between the stock return and the value-weighted market return. The market return is computed using all stocks trading at a given point in time. In the anal- ysis, we also use the Sharpe ratio—i.e., the average excess returns divided by the standard deviation of average returns over the sample period, where excess returns are computed as portfolio returns of the investor minus the returns on a government bond. Table 3.2 presents descriptive statistics for the adjusted return and Sharpe ratio as well as for the participation dummies. In addition, we compute three dummy variables that inform us whether the simulator user has previous experience in the actual stock market in one of these years. To carry out the analysis, we merged the simulator data set with the actual stock market data from 2009, 2010, 2011, and 2012 as we have the same user identifier in both data sets. The first dummy (exper1) takes the value of 1 if the individual participated in the stock market in only one of the years between 2009 and 2011, and 0 other- wise. The second dummy (exper2) takes the value of 1 if he or she participated in the stock market in two of those years; the third dummy (exper3) takes the value of 1 if the user participated in the stock market in all three years. With these dummy variables, we are able to estimate a participation (take-up) equation to 6  See http://www.isin.org/ (accessed May 2, 2014). 3.  Understanding and improving household investment behavior in Brazil  ◾ 121 TABLE 3.2  Descriptive statistics for participation indicators and return variables OBSERVA- TIONS MEAN SD MINIMUM MAXIMUM Ind09 (1 if in the stock market in 2009) 39,757 0.030 0.172 0 1 Ind10 (1 if in the stock market in 2010) 39,757 0.038 0.191 0 1 Ind11 (1 if in the stock market in 2011) 39,757 0.088 0.284 0 1 Ind12 (1 if in the stock market in 2012) 39,757 0.087 0.282 0 1 Ind (1 if in the stock market in 2012; 0 if 39,757 0.022 0.147 0 1 never participated in the stock market) Sharpe ratio 39,387 −0.078 3.568 −692.907 11.614 Sharpe ratio (if count > 30) 36,322 −0.055 0.070 −0.476 0.459 Risk-adjusted return 39,757 −0.058 0.186 −0.999 6.219 Risk-adjusted return (1% trimmed) 38,963 −0.061 0.144 −0.469 0.556 Note: SD = standard deviation. check whether the simulator is an effective tool in reducing barriers to stock market participation, and to control for previous experience in the stock market in some of the regressions. According to the table, only 3  percent of the simulator users were in the stock market in 2009. This number increases by almost 1 percentage point in 2010 and then jumps to about 9  percent in 2011 and 2012. The dummy variable Ind gives us the proportion of simulator users who participated in the stock market in 2012, as it takes the value of 1 if the individual traded in the stock market in 2012 and 0 if he or she did not do so in 2012 or in any of the previous years under study (2009, 2010, or 2011). Since the simulator data are from 2011, we can use this variable to test whether, after practicing with the simulator, people moved to the actual stock market in the subsequent year. The last three rows of table 3.2 present descriptive statistics for the Sharpe ratio and risk-adjusted return. For the Sharpe ratio, we also report the mean of a truncated distribution, as this measure requires that a stock be held by an indi- vidual a minimum number of days to (in our case, 30). Note that even though the truncation implies a reduction of the sample size by about 9 percent, the standard deviation drops sharply. Figure 3.5 shows the histogram for the truncated distri- bution of the Sharpe ratio. As with the Sharpe ratio, we also report two sets of descriptive statistics for the risk-adjusted return. The truncation makes the mean even more negative and narrows the distribution as measured by the standard deviation. Figure 3.6 illustrates the truncated distribution of this variable. 122  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 3.5  Distribution of the Sharpe ratio measure of simulator users Percent 20 15 10 5 0 −.4 −.2 0 .2 .4 Sharpe ratio Sources: BOVESPA-Simulator database 2011, and COMPUSTAT. Note: Number of nonmissing individual portfolio return days truncated so it is higher than 30. FIGURE 3.6  Distribution of adjusted returns of simulator users Percent 10 8 6 4 2 0 −.5 0 .5 Adjusted returns Sources: BOVESPA-Simulator database 2011, and COMPUSTAT. Note: Distribution of the adjusted returns is 1 percent trimmed due to outliers. 3.  Understanding and improving household investment behavior in Brazil  ◾ 123 3.8 RESULTS The results section is organized in three parts. First we present the results on whether exposure to the simulator is correlated with behavioral biases. We are not able to test whether individuals’ behavior changes over time; however, we test in the cross-section of simulator users whether those more exposed to the simulator are more or less affected by any type of behavioral bias. The second part focuses on how behavioral biases correlate with returns and with the Sharpe ratio. Finally, we provide the results from the estimation of participation to stock market equation. We present two main sets of estimates. In the first set, we have the biases variables on the right-hand side; in the second, we replace the biases by exposure to the simulator variables. We do not keep both vectors of regressors in the same model due to the high correlation between some of these variables. 3.8.1 Does experience in the simulator reduce behavioral biases? The hypothesis we test is whether those users who use the simulator more are more or less likely to make mistakes, such as realize gains too early (disposition effect), underdiversify, or overtrade (turnover). Since there are many potential reasons for selection bias, all results should be interpreted as correlations. In addition, we do not present the results for turnover bias as collinearity between this bias variable and exposure to simulator variables is by construction very high. Table 3.3 presents the results for the disposition effect. The simulator vari- ables used in the specifications are noted in the heading of each column. For instance, in the first column, the simulator corresponds to the number of trades in the simulator; in the third column, it is the number of days. For each different simulator variable, we provide two sets of estimates. In the first column, we have a simple linear model, while in the second column we test for nonlinearities between the exposure to the simulator variables and the behavioral biases of interest. We do that by simply including a second-order polynomial of the simu- lator variable in the regressions. Two main patterns emerge. First, the more a user uses the simulator (as measured by number of trades), the more he or she tends to show a disposi- tion effect. However, this is not confirmed by the other measures of exposure to the simulator. The other interesting insight is that we find some support for a nonlinear relationship between the exposure to the simulator variables and the two measures of behavioral biases.7 This suggests that users make more mistakes Although the coefficient of the squared term is not statistically significant in the last 7  column, it is jointly significant with the linear coefficient at the 1 percent level −F(1 2284) = 9.16. 124  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 3.3  Effect of the usage of the simulator on disposition effect # OF # OF LENGTH OF LENGTH OF TRADES TRADES # OF DAYS # OF DAYS USE USE Simulator 0.00014*** 0.00027*** 0.00013 −0.00022 0.0000034 −0.00026** −3.41 −3.26 −0.82 (−0.52) −0.096 (−2.12) Simulator^2 −0.00000036** 0.0000042 0.00000087** (−2.25) −1.09 −2.35 Length of use Age 0.00029 0.00029 0.00041 0.00042 0.00047 0.00047 −0.78 −0.78 −1.08 −1.1 −1.27 −1.27 Male −0.0062 −0.0071 −0.0049 −0.0047 −0.0045 −0.0037 (−0.59) (−0.68) (−0.47) (−0.44) (−0.43) (−0.35) Dummies for states? Constant 0.032 0.027 0.034 0.038 0.034 0.042 −0.82 −0.68 −0.87 −0.95 −0.86 −1.07 Observations 2,316 2,316 2,316 2,316 2,316 2,316 Adjusted 0.007 0.007 0.003 0.003 0.003 0.005 R-squared Maximum   375   26.2   149.4 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. as they are more exposed to simulator. However, after a certain level of usage, they seem to learn and mistakes start to be less frequent. In other words, after some level of exposure to the simulator, individuals may develop the skills that help them reduce the incidence of behavioral biases. For example, if we compute these threshold levels of exposure based on the estimates in table 3.3, we can infer that users who traded more than 375 times, or have an experience (length of use) higher than 149 days, would be less affected by the disposition bias as they use the simulator more. In fact, except for the measure “length of use,” the thresholds are surpassed for only the top 1  percent of users.8 The correlations suggest that the great majority of users would need more training to minimize the frequency of mistakes due to behavioral biases. We report the results for the diversification effect in table 3.4. Given that a low value for this variable means that the user holds a more diversified port- folio, the story we get from the coefficients of the regressions is one in which users become more diversified the more they use the simulator. However, the relationship between the simulator variable and the diversification effect seems to be a concave one. This suggests that users would settle on some restricted set of stocks (more concentrated portfolios) as they use the simulator more. For example, users who traded more than 236 times, or used the simulator for more 8  For the length of use, the threshold is overcome by about 7 percent of the users. 3.  Understanding and improving household investment behavior in Brazil  ◾ 125 TABLE 3.4  Effect of the usage of the simulator on diversification effect # OF # OF LENGTH OF LENGTH OF TRADES TRADES # OF DAYS # OF DAYS USE USE Simulator −0.0019*** −0.0035*** −0.0058*** −0.012*** −0.00059*** −0.0020*** (−21.7) (−21.4) (−23.7) (−25.3) (−23.7) (−29.6) Simulator^2 0.0000074 *** 0.00012 *** 0.0000061*** −7.59 −11.5 −22.1 Length of use Age −0.0012*** −0.0013*** −0.0011*** −0.0011*** −0.0013*** −0.0013*** (−6.50) (−7.05) (−5.53) (−5.81) (−6.58) (−6.61) Male 0.013*** 0.018*** 0.010** 0.014*** 0.0098** 0.012*** −2.9 −4.15 −2.3 −3.23 −2.19 −2.6 Dummies for states? Constant 0.66*** 0.68*** 0.66*** 0.68*** 0.66*** 0.67*** −48.4 −49.8 −47.8 −49.2 −47.4 −48.6 Observations 39,121 39,121 39,121 39,121 39,121 39,121 Adjusted 0.034 0.049 0.023 0.034 0.016 0.029 R-squared Maximum   236.5   50   163.9 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. than 50 days and for a time range of at least 164 days, would shift from a more diversified portfolio to a more concentrated one. One possible interpretation of this is that users buy several different stocks when they are novices, and then start to focus on a smaller group when they have more experience and familiarity with the stock options. Campbell, Ramadorai, and Ranish (2012) draw on Cremes and Petajisto (2009) to provide a different interpretation. According to them, there is some support for the “idea that skilled investors hold concentrated portfolios.” Although we are unable to control for skills in our regressions, our results could be seen as suggesting that more training in the simulator pays off as individuals learn. The accumulation of learning (or skills) would be translated into a more concentrated portfolio. 3.8.2 How important is simulator exposure and behavioral biases on portfolio performance? This section investigates whether exposure to the simulator variables and the behavioral biases variables affect the (adjusted) returns of users of the simulator. The adjusted returns are computed as the difference between the returns the users make with the simulator and the returns of the actual stock market. We compute investor returns as in Barber and Odean (2000). We first create a time series of day-end portfolio holdings of each stock for each investor. We then 126  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES compute the weights of each stock in the portfolio by dividing the value of a given stock by the total value of all stocks in the investor’s portfolio on that day. To compute the portfolio returns of each investor, we then multiply day-end weights of each stock by that stock’s return over the next day. With this computation, we ignore intraday trades. Once the return for each investor is computed, we subtract from it the daily returns of the market. Figures 3.5 and 3.6 present the distribution of the Sharpe ratio and adjusted return for simulator users, respectively. As expected, the risk-adjusted returns have a slightly negative mean, with most of the distribution around 0. That is, the majority of investors underperform the market but only by a small margin. Simi- larly, the Sharpe ratio is distributed with a mean close to the market Sharpe ratio. When we study the correlation between the behavioral bias measures and the adjusted returns (in tables 3.5 and 3.6), we find that there is an association. In particular, users who are more diversified and who trade more (turnover) show higher adjusted returns. Interestingly, of the few demographic variables we have, it seems that age is negatively associated with returns. In contrast, table  3.5 suggests that there is no clear association between exposure to the simulator and returns. Table 3.6 includes indicators for previous experiences in the stock market, but the results are exactly the same. TABLE 3.5  Effect of the usage of the simulator and behavioral biases on adjusted returns Age −0.00047*** −0.00028 −0.00046*** −0.00047*** −0.00048*** −0.00047*** (−4.06) (−0.55) (−3.93) (−4.05) (−4.16) (−4.05) Male −0.0023 −0.018 −0.0044* −0.0027 −0.0027 −0.0026 (−0.89) (−1.26) (−1.71) (−1.05) (−1.08) (−1.02) Diversification −0.0075** (−2.37) Disposition −0.057 (−1.54) Turnover 0.072*** (16.5) # of trades 0.000061 (1.14) # of days 0.00029 (1.30) Length of use 0.000021 (1.00) Dummies of state? Yes Yes Yes Yes Yes Yes Constant −0.040 *** −0.039 −0.056 *** −0.045 *** −0.045 *** −0.045*** (−4.75) (−1.02) (−6.60) (−5.46) (−5.46) (−5.44) Observations 39,121 2,316 38,047 39,121 39,121 39,121 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 3.  Understanding and improving household investment behavior in Brazil  ◾ 127 TABLE 3.6  Effect of the usage of the simulator and behavioral biases on adjusted returns (include previous experience as controls) 1 year experience 0.0015 0.037 −0.00089 0.0011 0.00053 0.00095 (0.27) (1.00) (−0.15) (0.20) (0.094 (0.17) 2 year experience 0.0042 0.042 0.0023 0.0041 0.0037 0.0037 (0.48) (0.67) (0.26) (0.47) (0.43) (0.43) 3 year experience 0.0048 −0.076 0.0042 0.0046 0.0044 0.0046 (0.35) (−0.90) (0.30) (0.33) (0.32) (0.34) Age −0.00050*** −0.00049 −0.00048*** −0.00049*** −0.00050*** −0.00049*** (−4.21) (−0.87) (−3.93) (−4.17) (−4.24) (−4.15) Male −0.0025 −0.019 −0.0045* −0.0029 −0.0029 −0.0028 (−0.99) (−1.42) (−1.75) (−1.13) (−1.15) (−1.10) Diversification −0.0076** (−2.40) Disposition −0.056 (−1.53) Turnover 0.072*** (16.5) # of trades 0.000060 (1.14) # of days 0.00029 (1.28) Length of use 0.000020 (0.96) Dummies of state? Yes Yes Yes Yes Yes Yes Constant −0.039 *** −0.036 −0.056 *** −0.044 *** −0.044 *** −0.044*** (−4.65) (−0.97) (−6.54) (−5.35) (−5.36) (−5.34) Observations 39,121 2,316 38,047 39,121 39,121 39,121 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. In table 3.7, we run the same specifications with the Sharpe ratio as with the dependent variable. The Sharpe ratio is significantly associated with all the expo- sure measures; in particular, exposure to the simulator leads to higher Sharpe ratios. The results suggest that an increase of 25 trades result in an increase in the Sharpe ratio of 0.03. Given that the market Sharpe ratio since 2005 has been 0.35, what we find is an economically significant result. In terms of behavioral biases, disposition and less diversified portfolios are correlated with a smaller Sharpe ratio, while users who trade frequently seem to be able to achieve a higher risk-adjusted return. As with the adjusted returns, table 3.8 presents the results with the experience dummies as controls. In fact, the coefficient for one of the dummies is positive and highly significant in all regressions. This suggests that simulator users with at least one year of experience in the stock market tended to have higher Sharpe ratios. 128  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 3.7  Effect of the usage of the simulator and behavioral biases on the Sharpe ratio Age −0.00010** 0.00031** −0.00015*** −0.00014*** −0.00017*** −0.00014*** (−2.28) (−2.04) (−3.29) (−3.13) (−3.66) (−3.07) Male −0.0011 0.00093 −0.0016 −0.0015 −0.0017 −0.0013 (−0.98) (−0.22) (−1.47) (−1.33) (−1.56) (−1.19) Diversification 0.020*** (−15.6) Disposition −0.026*** (−3.22) Turnover 0.041*** (−12.2) # of trades 0.000076*** (−6.25) # of days 0.00045*** (−10.2) Length of use 0.000020*** (−4.06) Dummies of state? Yes Yes Yes Yes Yes Yes Constant −0.059 *** −0.057 *** −0.051 *** −0.047 *** −0.047 *** −0.046*** (−16.6) (−4.14) (−14.5) (−13.3) (−13.3) (−13.2) Observations 35,733 2,269 35,732 35,733 35,733 35,733 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. Number of nonmissing individual portfolio return days > 30. 3.8.3 Does the simulator encourage the “right” people to participate in the actual stock market? We use a linear probability model to estimate users’ decision to participate in the actual stock market. As explained in the methods section, the dependent variable is defined as a dummy that takes the value of 1 if the simulator user (in 2011) is in the actual stock market in 2012 and 0 if he or she does not participate in 2012 or in any of the previous years (2009–11). The model includes age, gender (male), and state of residence, exposure to the simulator variables, indicators of returns in the simulator, as well as the behavioral biase variables as predictors. Due to high correlation between some variables on the right-hand side of the model, we opted to run separate regressions. The main results are in tables 3.9 and 3.10, where we present the results of the main specifications on an unrestricted sample of simulator users (table 3.9) and restricted to users with more than five trades (table 3.10). We discuss the results together, as they were consistent across the two samples. There is some evidence of a nudging effect from the simulator. Exposure to the simulator is indeed associated with a higher probability of entering the market. Using the 3.  Understanding and improving household investment behavior in Brazil  ◾ 129 TABLE 3.8  Effect of the usage of the simulator and behavioral biases on adjusted returns (include previous experience as controls) 1 year experience 0.0045*** 0.0075 0.0029** 0.0038*** 0.0027** 0.0038*** (3.26) (1.61) (2.10) (2.76) (1.99) (2.76) 2 years experience 0.0041 0.023** 0.004 0.0043* 0.0037 0.004 (1.60) (1.99) (1.56) (1.67) (1.46) (1.54) 3 years experience 0.0044 −0.019 0.0042 0.0047 0.0043 0.0048 (1.13) (−1.08) (1.08) (1.21) (1.11) (1.23) Age −0.00014 *** 0.00022 −0.00018 *** −0.00018 *** −0.00020 *** −0.00018*** (−3.11) −1.37 (−3.97) (−3.92) (−4.28) (−3.83) Male −0.0014 0.00031 −0.0019* −0.0018 −0.0019* −0.0016 (−1.28) −0.074 (−1.71) (−1.61) (−1.77) (−1.44) Diversification 0.020*** (15.60) Disposition −0.026*** (−3.21) Turnover 0.041*** (12.00) # of trades 0.000074*** (6.07) # of days 0.00043*** (9.84) Length of use 0.000018*** (3.55) Dummies of state? Yes Yes Yes Yes Yes Yes Constant −0.058 *** −0.055 *** −0.050 *** −0.046 *** −0.046 *** −0.045*** (−16.3) (−4.08) (−14.2) (−13.0) (−13.1) (−12.9) Observations 35,733 2,269 35,732 35,733 35,733 35,733 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. Number of nonmissing individual portfolio return days > 30. Includes previous experience as controls. estimates in table 3.9, we find that an increase of 30 trades leads to an increase in the take-up probability by 0.5 percentage points. The magnitude of this coeffi- cient is fairly relevant in this context, given that the average take-up is less than 3 percent. We also find the Sharpe ratio to be associated with higher take-up. This finding suggests positive selection, with investors who perform well on a risk-ad- justed basis to be more likely to participate in the market. There is some evidence of learning and financial sophistication, with investors recognizing and improving upon their investment acumen. At the same time, we find volatility to be asso- ciated with higher participation, suggesting that investors who trade and invest in lottery-type stocks may be drawn to the market for the wrong reasons. More empirical work needs to be done before we can disentangle the different forces at play. In terms of biases, the only clear association is with turnover effect. In other 130  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 3.9  Participation equation: is the simulator an effective nudging tool?— unrestricted sample Age −0.00083** 0.00021** 0.00020** 0.00016* 0.00022** 0.00023** 0.00022** 0.00022** 0.00023** (−2.06) (2.27) (2.24) (1.81) (2.45) (2.52) (2.39) (2.50) (2.52) Male 0.0089 0.015*** 0.015*** 0.015*** 0.016*** 0.016*** 0.016*** 0.016*** 0.016*** (0.92) (9.87) (9.99) (9.81) (10.4) (10.5) (10.3) (10.7) (10.7) # of stocks −0.00015 0.00021 0.00055* 0.00053 0.00059* (−0.41) (0.63) (1.65) (1.59) (1.75) Diversifica- 0.027 −0.0032 tion (1.56) (−1.27) Disposition −0.014 (−0.62) Turnover 0.091*** 0.044*** (4.07) (8.91) # of trades 0.00017*** (4.10) # of days 0.00082*** (5.67) Sharpe ratio 0.000051*** (2.61) Average −0.030 return (−0.24) Return vola- 0.13** tility (2.13) Raw return 0.018 (0.49) Adjusted 0.0026 return (0.64) Dummies of Yes Yes Yes Yes Yes Yes Yes Yes Yes states? 0.016 −0.0083 −0.0040 −0.0050 −0.0047 −0.0049 −0.0072 −0.0035 −0.0034 Constant (0.44) (−1.45) (−0.70) (−0.87) (−0.83) (−0.86) (−1.23) (−0.63) (−0.61) Observa- 2,312 38,047 39,121 39,121 38,756 39,121 38,757 39,121 39,121 tions Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. terms, excess trading—while not necessarily an optimal investment behavior— is correlated with higher participation in the stock market. Finally, in terms of 3.  Understanding and improving household investment behavior in Brazil  ◾ 131 TABLE 3.10  Participation equation: is the simulator an effective nudging tool?— individuals with more than five trades Age −0.00083** 0.00019 0.00019 0.00013 0.00022* 0.00022* 0.00021* 0.00022* 0.00022* (−2.06) (1.56) (1.54) (1.05) (1.80) (1.84) (1.73) (1.84) (1.84) Male 0.0089 0.017*** 0.017*** 0.017*** 0.018*** 0.018*** 0.018*** 0.018*** 0.018*** (0.92) (7.65) (7.69) (7.62) (8.04) (8.01) (7.95) (8.01) (8.01) # of stocks −0.00040 0.000100 −0.000044 −0.000030 0.0000022 (−1.01) (0.25) (−0.11) (−0.078) (0.0055) Diversifica- 0.027 −0.0014 tion (1.56) (−0.36) Disposition −0.014 (−0.62) Turnover 0.091*** 0.044*** (4.07) (7.49) # of trades 0.00015*** (3.29) # of days 0.00077*** (4.76) Sharpe 0.00079* ratio (1.77) Average 0.018 return (0.13) Return vol- 0.17** atility (2.14) Raw return 0.031 (0.61) Adjusted −0.00073 return (−0.14) Dummies Yes Yes Yes Yes Yes Yes Yes Yes Yes of states? Constant 0.016 −0.017** −0.0073 −0.0092 −0.0066 −0.0067 −0.0099 −0.0068 −0.0069 (0.44) (−2.31) (−0.94) (−1.18) (−0.85) (−0.87) (−1.25) (−0.91) (−0.92) Observa- 2,312 22,950 23,656 23,656 23,471 23,656 23,472 23,656 23,656 tions Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. indicators of returns, we find that a better Sharpe ratio and a higher return vola- tility are associated positively with stock market participation. 132  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 3.9 CONCLUSION This study explores the effectiveness of an online stock market simulator as a tool to overcome investor biases and improve performance through a learning-by- doing training experience. Using data from about 40,000 simulator participants, our findings offer several insights. First, exposure to the stock market simulator has some effect on behavioral biases; however, our results are not conclusive regarding the size and magnitude of this effect. It seems that simulator users are more affected by a disposition bias as they use the simulator more often. However, if simulator use is very heavy (which is the case for only 1 percent of the users in our data set), then the disposition bias would be reduced by exposure to the simulator. For diversification, we find that learning from the simulator drives users to hold a more concentrated portfolio. In terms of the relationship between behavioral biases and measures of performance, our results are not conclusive. While there is a clear associa- tion, some biases are correlated with higher adjusted returns, while others are correlated with a decrease in measures of performance. Exposure to the simu- lator does not necessarily affect measures of performance in our data. However, an interesting result is that exposure to the simulator is correlated with a higher Sharpe ratio. When we look at the link between participation in the simulator and tran- sition into the actual stock market, we find quite a strong relationship. We find support for a simulator nudging effect, as exposure to the simulator is asso- ciated with a higher probability of entering the market. In addition, there is evidence of both positive and negative selection into the stock market. Indeed, investors who have a higher Sharpe ratio in the simulator tend to enter more, but investors with higher volatility in the simulator display the same pattern. More empirical work needs to be done to disentangle the different forces at play on this dimension. Overall, our results suggest that learning by doing can be a powerful tool to motivate investors to participate in the actual stock market. At the same time, learning by doing might not be enough to overcome behavioral biases. As such, learning by doing tools might have to be complemented with more specific training to avoid the risk of investors entering the market but not being ready to adopt an optimal investment behavior. REFERENCES Ariely, Dan. 2009. Predictably Irrational: The Hidden Forces That Shape Our Decisions. New York: HarperCollinsPublishers. Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of Economics 121 (2): 635–72. 3.  Understanding and improving household investment behavior in Brazil  ◾ 133 Barber, Brad M., and Terrance Odean. 2000. “Trading Is Hazardous to Your Wealth: The Common Stock Investment Performance of Individual Investors.” Journal of Finance 55 (2): 773–806. Barsky, Robert B., Thomas F. Juster, Miles S. Kimball, and Matthew D. Shapiro. 1997. “Preference Pand Behavioral Heterogeneity: An Experimental Approach in the Health and Retirement Survey.” Quarterly Journal of Economics 112: 537–79. Bertrand, Marianne, Dean Karlan, Sendhil Mullainathan, Eldar Shafir, and Jonathan Zinman. 2010. “What’s Advertising Content Worth? Evidence from a Consumer Credit Marketing Field Experiment.” Quarterly Journal of Economics 125 (1): 263–306. Bertrand, Marianne, and Adair Morse. 2010. “Information Disclosure, Cognitive Biases and Payday Borrowing.” Chicago Booth Research Paper No. 10-01, University of Chicago Booth School of Business, Chicago. Campbell, John Y. 2006. “Household Finance.” Journal of Finance 61: 1553–604. Campbell, John Y., Tarun Ramadorai, and Benjamin Ranish. 2012. “Do Stock Traders Learn from Experience? Evidence from an Emerging Market.” http://www.igidr.ac.in/FSRR/ PDF/EMF2012/Campbell_Ramadorai_Ranish_2012.pdf. Cole, Shawn A., Thomas Sampson, and Bilal Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Cremers, K. J. Martijn, and Antti Petajisto. 2009. “How Active Is Your Fund Manager? A New Measure That Predicts Performance.” Review of Financial Studies 22 (9): 3329–65. Daniel, Kent, David Hirshleifer, and Siew Hong Teoh. 2002. “Investor Psychology in Capital Markets: Evidence and Policy Implications.” Journal of Monetary Economics 49  (1): 139–209. De la Torre, Augusto, Juan Carlos Gozzi, and Sergio L. Schmukler. 2007. “Capital Market Development: Whither Latin America?” Policy Research Working Paper 4156, World Bank, Washington, DC. DellaVigna, Stefano. 2009. “Psychology and Economics: Evidence from the Field.” Journal of Economic Literature 47: 315–72. Demirgüç-Kunt, Asli, Thorsten Beck, and Patrick Honohan. 2008. Finance for All? Policies and Pitfalls in Expanding Access. Washington, DC: World Bank. http://siteresources. worldbank.org/INTFINFORALL/Resources/4099583-1194373512632/FFA_book.pdf. Dewey, J. 1938. Experience and Education. New York: Macmillan. DIME (Development Impact Evaluation). 2012. “Concept Note—From Shop to Stock. Understanding and Improving Household Investment Behavior.” World Bank. Available upon request. Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Economist. 2011. "Consumer Debt in Brazil: Sweet Treats on the Never Never." April 27. http://www.economist.com/node/21256685. Goetzmann, William N., and Alok Kumar. 2008. “Equity Portfolio and Diversification.” Review of Finance 12 (3): 433–63. Haliassos, Michael, and Carol C. Bertaut. 1995. "Why Do So Few Hold Stocks?" Economic Journal 105 (432): 1110–29. Hong, Harrison, Jeffrey D. Kubik, and Jeremy C. Stein. 2004. “Social Interaction and Stock Market Participation.” Journal of Finance 59 (1): 137–63. Kahneman, Daniel, and Amos Tversky. 1979. “Prospect Theory: An Analysis of Decisions Under Risk.” Econometrica 47 (2): 263–91. Legovini, Arianna, Miriam Bruhn, and Bilal Zia. 2013. “Impact Evaluation of Brazil’s School Based Financial Education Program.” Unpublished. 134  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Madrian, Brigitte, and Dennis F. Shea. 2001. “The Power of Suggestion: Inertia in 401(k) Participation and Savings Behavior.” Quarterly Journal of Economics 116 (4): 1149–87. Seru, Amit, Tyler Shumway, and Noah Stoffmann. 2010. “Learning by Trading.” Review of Financial Studies 23 (2): 705–39. Thaler, Richard H. 1981. "Some Empirical Evidence on Dynamic Inconsistency." Economics Letters 8: 201–07. Van Rooij, Maarten, Annamaria Lusardi, and Rob Alessie. 2011. “Financial Literacy and Stock Market Participation.” Journal of Financial Economics 101: 449–72. Vissing-Jørgensen, Annette. 2002. “Towards an Explanation of Household Portfolio Choice Heterogeneity: Nonfinancial Income and Participation Cost Structures.” http://faculty. haas.berkeley.edu/vissing/vissing_nber2_update.pdf. 3.  Understanding and improving household investment behavior in Brazil  ◾ 135 ANNEX 3A: DESCRIPTION OF THE BROADER IMPACT EVALUATION RESEARCH PROGRAM A large research program has been designed to address a core research ques- tion: What interventions are most effective at improving participation in the stock market and reducing biases in investor behavior? A broad program of impact evaluation studies has been designed to address and disentangle the different constraints and market failures hypothesized to be affecting investor participation and behavior, including physical (financial education, financial and nonfinancial entry costs, transaction costs, access) and psychological constraints (such as self-control, time-inconsistent preferences, inertia, menu costs, peer effects). The findings of the study will add to the literature by measuring the effec- tiveness of new modes of financial education. The study will test multiple hypoth- eses regarding barriers to effective participation in financial markets within a unified framework, leading to better understanding of not only which investor biases are most prevalent but also how targeted education and product offers can mitigate these biases, as well as illustrating a framework in which impact evaluations can be used to make real-time operational decisions in the private sector (table 3A.1). This research program has three main components: (1) testing the effect of an online stock market simulator as a learning-by-doing financial literacy tool to improve knowledge and investor behavior; (2) reducing the barriers to stock market participation offers to randomly selected groups of potential investors participating in the simulator; and (3) providing financial information and details of investment products through a stock broker online platform to individual inves- tors with the aim of reducing investor biases. TABLE 3A.1  Barriers to entry and investor trading biases BARRIERS TO ENTRY INVESTOR BIASES Expected change: Interventions reduce participa- Expected change: Financial education " tion barriers " improved transition rates from the increased take-up " reduction in investor bias " simulator to the stock market (increased entry) and reduced exposure to unintended risk " improved improved self-selection into the market portfolio performance (risk-adjusted returns) and longer investment horizons Self-control problems Underdiversification Menu costs (choice avoidance) Disposition effect Inertia/procrastination Overconfidence Lack of information/education Limited attention High start-up costs High ongoing participation costs Ambiguity aversion 136  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Here we discuss in more detail barriers to entry and investor biases. 3A.1 Financial and time costs Direct financial costs as well as the costs associated with the time taken to learn about financial products play an important role in the decision to participate in financial markets. Participation costs can be broken into three categories: (1) fixed entry costs (such as any registration fees, the cost of time required to register, and any training and gathering of information required to make investment decisions); (2) variable trading costs (such as broker transaction fees, bid/ask spreads and custody fees); and (3) per period costs (such as the time required to review and reallocate portfolios, the associated complications in tax filings, and commensurate increase in time required that come with owning stocks). Each of these costs has the potential to price smaller investors out of the market. Viss- ing-Jørgensen (2002) for example shows that a US$50 annual cost associated with the required time and effort to continuously review stock holdings can explain nearly 50  percent of nonparticipation rates in the U.S. stock market. In Brazil, high fixed costs of participation mean that few investment options are targeted to lower-income individuals, thus effectively excluding this group from accessing suitable investment products—resulting in 61 percent of Brazilian stock investors being concentrated in the highest income group.9 Of particular interest in this study is the relationship between participation costs and other factors that may influence investor choices to determine the rela- tive importance of these different barriers to entry. Cole, Sampson, and Zia (2011) find that subsidizing the cost of opening a bank account by US$15 can increase the probability of opening a bank account by 12 percent, whereas providing a finan- cial literacy course for a similar cost (US$17) has little to no effect on opening an account, further enforcing the importance of financial barriers to entry. However, the case for financial education to support stock market participation may be stronger given the more sophisticated context and greater uncertainty related to this investment. 3A.2 Psychological barriers Demographics, financial education, and participation costs, while important, fail to fully explain investor choices. Even when people are fully aware of what the best action to take is, behavioral elements such as procrastination, time-incon- sistent preferences, and self-control influence decisions (Ariely 2009; DellaVigna 2009). For instance, precommitment products have been shown to effectively 9  This figure is based on an investor profile study commissioned by BM&FBOVESPA in 2010. The highest income group is class A, according to the income classification commonly used in Brazil. 3.  Understanding and improving household investment behavior in Brazil  ◾ 137 overcome self-control problems and dramatically increase savings rates by using contractual agreements to specify payment frequencies and amounts to be paid by investors, effectively enforcing time consistency and removing the influence of future temptations (Ashraf, Karlan, and Yin 2006). Behavioral factors are especially important in the decision to participate in more complex financial markets such as the stock market. In contrast to simple investment products, such as savings accounts or insurance, participation in the stock market is associated with significant uncertainty. The participation decision therefore involves not only a simple cost-benefit trade-off, but also individual atti- tudes toward risk and uncertainty, and confidence in one’s ability to pick high-re- turn investments. Risk-averse individuals are less likely to invest in the stock market. In fact, a common approach to measure an individual’s risk aversion is to ask whether he or she invests in the stock market. However, when we consider uncertainty as a possible deterrent to stock market participation, an important distinction arises between general risk aversion and ambiguity aversion. While risk aversion (usually measured by eliciting choices between lotteries with different payoffs and proba- bilities) captures an individual’s general dislike of taking gambles, ambiguity aver- sion measures the aversion to “unknown risks” (Barsky et al. 1997). This is an important concept in understanding the decision to participate in the stock market as opposed to traditional investments: risk-averse individuals may simply dislike participating in markets with uncertainty and would always chose a safe investment over one with variable outcomes. Ambiguity-averse individuals, on the other hand, refrain from investing in the stock market simply because they lack information about the probabilities of gains or losses (i.e., finan- cial education or information about the returns of alternative investments). Even when returns in the stock market may be superior to those of alternative invest- ment, ambiguity-averse investors choose the former when they are ill informed about stock market investments, avoiding the uncertain bet. This has important implications for the ability of financial education to lower barriers to entry. In addition to the extent of information individuals have about financial risks and returns, the way information is provided also plays an important role. For instance, presenting consumers with fewer options has been shown to increase take-up of both loans and retirement savings—a finding that is consistent with the hypothesis that larger investment menus can trigger choice avoidance in consumers (Bertrand et al. 2010). To overcome both choice avoidance and inertia, the use of default options in retirement savings has been shown to increase employee enrollment from 49 percent to 86 percent (Madrian and Shea 2001). Peer effects also play a potentially important role in investment choices. By lowering the barriers to entry for a particular group of people, their peers become significantly more likely to participate in savings products (Duflo and Saez 2003). Similarly, people who socialize or have neighbors who invest in stocks are also more likely to invest themselves (Hong, Kubik, and Stein 2004). 138  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES This study will draw on this body of knowledge on entry barriers by developing interventions directed to address these specific points by, for instance, varying the financial costs of participation, introducing targeted educational campaigns to overcome information constraints, and adapting the way in which this infor- mation is provided (e.g., simplifying the set of investment options offered to over- come choice avoidance or including information about similar investor profiles to leverage the influence of peer effects). The fact that (1) regulatory barriers such as high start-up and participation costs (financial and nonfinancial) create a disincentive to active participation in the market; (2) people may not have the required knowledge to correctly choose the most effective investment strategy (information constraint); and (3) people are aware of the value of diversifying their assets to include stocks but choose not to for a variety of behavioral reasons such as choice avoidance, self-control problems, peer effects or ambiguity aversion. As discussed above, this question fits within a broader research objective. In particular, we will test, through randomized experiments, several incentives and offers/products to reduce investor biases and barriers to stock market participa- tion.10 The research question addressed in this chapter constitutes an important first step in the implementation of the broader research program. 10  See DIME (2012) for a full description of the overall research program. CHAPTER 4 D oes financial education affect savings behavior? Experimental evidence from India MARGHERITA CALDERONE, NATHAN FIALA, FLORENTINA MULAJ, SANTADARSHAN SADHU, AND LEOPOLD SARR ABSTRACT Financial literacy training is growing in popularity across the developing world. Recent research results, though, have failed to find a significant impact from training on savings behavior. We experimentally test the impact of financial literacy training when paired with a branchless banking program that reaches those who do not have easy access to traditional banking. The intervention consisted of a two-day training on a random sample of 3,000 clients served by a branchless banking facility across two adjacent districts in the state of Uttar Pradesh, India. The results reveal that the intervention had a significant impact on savings: savings in the treatment group increased by 29 percent (US$30) within a period of one year. Our results also indicate that attitudes related to financial planning improved after the intervention, but there was no impact on overall financial literacy. These findings suggest that financial education, when paired with branchless banking, can improve savings and behavioral outcomes, even when it does not significantly affect financial knowledge. We thank FINO for implementing this program. For comments, we thank Shawn Cole, William Jack, Toby Linden, Bilal Zia, Sigfried Zottel, and numerous conference and seminar participants. For funding, we are grateful to the Russia Financial Literacy and Education Trust Fund. Finally, Anup Roy provided superb research assistance through the Institute for Financial Management and Research. All opinions in this chapter are those of the authors and do not necessarily represent the views of FINO or the World Bank.   ◾ 139 140  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 4.1 INTRODUCTION In a rising number of developing countries, thanks to innovations in technolo- gy-based branchless banking, formal banking has expanded its outreach to previ- ously unbanked populations, including some of the most vulnerable segments. A recent study shows that between 500 and 800 million of the world’s poor now have access to finance (Deb and Kubzansky 2012). However, there is also evidence indicating that the majority of these individuals are not prepared to interact with the growing complexities of financial products and services. Scholars focusing on both developed (Lusardi and Mitchell 2006) and developing countries (Cole, Sampson, and Zia 2011) have documented low levels of financial literacy, including low knowledge and skills around basic concepts of personal financial manage- ment as well as low understanding of more general banking practices. A growing literature suggests that financial literacy is correlated with house- hold well-being, including participation in savings, credit, investments (Hilgert, Hogarth, and Beverly 2003), and planning for retirement (Lusardi and Mitchell 2007). Therefore, along with the increased focus on financial inclusion, a growing number of countries are developing national strategies for financial education and making more investments in related programs (Grifoni and Messy 2012). The basic premise behind these initiatives is that improved access, complemented with financial knowledge, will lead to responsible financial behavior among consumers. However, in order to provide more concrete policy directions in terms of program design, there is a need to establish causality between financial educa- tion and behavioral outcomes and, to the extent possible, examine the impact of the former on a number of variables along the causal chain. To date, the evidence from field experimental research in analyzing the cause and effect of financial education has provided mixed results in both developed and developing countries. Duflo and Saez (2003) measure the impact of a benefit fair on retirement plan enrollment among employees of a university in the United States, but found small effects on enrollment. In a development context, Cole, Sampson, and Zia (2011) looked at the impact of financial education training among the unbanked in Indonesia, also finding no substantial effect on savings behavior. On the other hand, in Brazil, Bruhn et al. (2013) look at high school financial educa- tion incorporated in the standard curriculum during three academic semesters, and report impact on financial literacy, attitudes, and behavioral change (based on self-reported data). Though not focusing on personal finance, Karlan and Valdivia (2011) examined the impact of business education on female entrepreneurs in Peru and found a significant increase in participants’ engagement in some of the activities included in the training, such as separating money between business and household, reinvesting profits, and maintaining records of sales and expenses. One of the reasons for the lack of consistent findings could be that most poor people are often not familiar with formal savings programs. About 2.5 billion adults—just over half of the world’s adult population—do not use formal financial 4.  Does financial education affect savings behavior?  ◾ 141 services to save or borrow (Chaia et al. 2009). Many of these people participate in other savings options, such as rotating savings and credit associations (ROSCAs), but the majority do not opt for formal savings. This may be due to the lack of knowledge about formal banking’s benefits or to the difficulty of access, since most banks are not near the poor and do not offer services for low-level deposi- tors. Doorstep banking—also called “last mile” banking, as the bank reaches out to those who cannot make it to the banks—can often be found in retail shops, through agents that live in or near the villages, or through mobile banking vehicles, or even mobile phones, such as those being pioneered by M-Pesa and M-Kesho (Demombynes and Thegeya 2012). Doorstep banking makes it easier for people to handle formal savings accounts, though it is a new and still poorly understood idea. There are some advantages to formal banking. Unlike village savings programs, banks offer privacy from family members and other villagers, decreased risk of theft or default, and reliability. When financed by nongovernmental organiza- tions or through government regulation, they can also be low cost or even free of any charges. Dupas and Robinson (2013a) find that giving female micro-enter- prise owners in Kenya access to this type of low-cost savings account increases savings, productive investment, and food expenditures. Sometimes unique savings programs offer the best chance for households to save. Duflo, Kremer, and Robinson (2011) experimented with alternative money storage options by encouraging farmers to spend money at harvest time on fertilizers for the next season, which were then delivered for free. The program was found to increase fertilizer usage. Brune et al. (2011) gave Malawi farmers either normal savings accounts or commitment savings accounts with which the farmers had to specify when money could be withdrawn. The rate of deposits was high for the commitment savings accounts and almost twice that of the normal accounts. Similarly, Ashraf, Karlan, and Yin (2006, 2010) introduced commitment savings accounts in the Philippines for those who already had savings accounts and found increased saving rates. The experimental literature suggests there is reason to be skeptical about financial literacy. We present a randomized field experiment in India to measure the impact of a financial education program, delivered in conjunction with doorstep banking, on savings behavior among low-income households. The methodolog- ical and conceptual contribution of the study is thus to experimentally explore the causal relationship between financial education and savings, and between financial education and financial capabilities, in combination with branchless banking. We find that the financial education intervention had a significant impact on savings. Individuals who received the training saved 29  percent more than the control group. Moreover, we find improvements on attitudes related to finan- cial planning, but we do not find impact on financial knowledge. These findings suggest that financial education can improve behavioral outcomes, even if it does not affect overall financial literacy. 142  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES In common with hypotheses put forward in the literature on financial literacy training, we also look at heterogeneity analysis and discover three interesting findings. First, financial attitudes improve more among less educated individuals. Second, increased interest in financial matters and a shift from informal to formal savings are found among individuals more financially educated at baseline. Third, the intervention was less effective for more impatient individuals. We also discuss some of the mechanisms behind these results and show that the treatment effect is not merely given by a substitution from other forms of savings to the doorstep banking account, but, on the contrary, the program crowded in other formal savings as well. 4.2 EXPERIMENTAL DESIGN 4.2.1 The program The doorstep banking and financial literacy training was conducted in cooperation with FINO Paytech Foundation, a for-profit financial services company based in India and specializing in offering technology-based banking transactions. FINO works with different financial institutions to enable access to financial services for previously excluded segments of society by offering last mile service delivery through a number of portable devices, including biometric smart cards, hand- held devices, and micro-deposit machines with biometric authentication. The model FINO employs to reach out to households in rural areas is based on busi- ness correspondents, also known as bandhus, who are permanently based in the villages where FINO operates and serve as the focal point, or contact person, between the financial institution and community members. This model helps introduce the bank to the poor, who usually are not familiar or comfortable with traditional banking institutions, through a more personal interaction. To date, FINO has trained more than 10,000 bandhus, serving over 58 million customers, and it is adding close to a million clients per month (Forbes India 2012). While doorstep banking has had enormous success in expanding access, as many studies that look at the impact of branchless banking have shown, access in and of itself does not make individuals financially more aware or literate (Thyaga- rajan and Venkatesan 2008, India; Dupas et al. 2012, Kenya). For example, out of the sample of 3,000 individuals randomly drawn from FINO’s administrative data- base who had signed up for FINO bank accounts and were observed over a period of two months, 88 percent were found to have made no transactions, with only 10  percent holding a positive balance.1 While there could be many factors that 1  Source: Administrative data shared by FINO. Such a picture also reflects the fact that bandhus received around Rs 20–25 for signing up each client and so have an incentive to sign up as many clients as possible, not just those with a strong interest in banking. 4.  Does financial education affect savings behavior?  ◾ 143 account for this shortcoming—including lack of financial resources, potential lack of effective access, lack of trust in branchless banking, and behavioral biases— concern exists among many researchers and policy makers that low levels of financial literacy are a major constraint. In the case of FINO, it is important to note that most of the individuals who signed up for bank accounts were provided instructions on how to use the portable cards and on the types of transactions they could make; also, they had ongoing access to the bandhus for any question. However, it is likely that these FINO clients did not have a general understanding about the benefits of responsible financial management or about basic concepts of personal finance. Also, irregular presence of FINO bandhus might result in low transactions by limiting access to transaction points. The financial literacy curriculum was developed in collaboration with FINO. The intervention consisted of a two-day financial education training program, delivered through a video (two to three hours per day) in a classroom setting, followed by interactive discussions on the presentation. It was implemented between May and August 2011 across two adjacent districts in the state of Uttar Pradesh. Table 4.1 illustrates the contents covered by the training, which mostly focused on three topics: the role of formal banking in people’s lives; responsible borrowing, spending, and saving; and cash management. Overall, the training material was standard and based on classic modules used in other financial literacy interventions; the only difference was that the beneficiaries were also TABLE 4.1  The content of the financial education training TRAINING MODULE CONTENTS METHODOLOGY Financial planning ƒƒDiscussion about objective of session Discussion, and budgeting ƒƒHow to keep track of income and expenses pamphlet, storytelling ƒƒCreation of personal budget and its categories ƒƒAllocation of Income among budget categories Saving and ƒƒImportance of regular saving Video, comics, investment ƒƒDifference between savings and investments storytelling, leaflets ƒƒ Importance of savings account and different avenues of saving ƒƒLong-term saving and planning for major future event ƒƒDifferent avenues of investment Borrowing and loan ƒƒConcepts of wise borrowing Video, comics, management ƒƒDifferent avenues of borrowing storytelling, leaflets ƒƒPlanning personal loan management ƒƒPlanning for emergency needs to avoid overindebtedness Mitigating risk and ƒƒMeaning and usefulness of insurance Video, comics, insurance ƒƒDiscussion of different insurance products storytelling, leaflets ƒƒPension planning or target segments Formal financial ƒƒBasic know-how about banking and allied services Video, group services know-how ƒƒNeed for including oneself in formal financial system discussion, leaflets 144  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES told more about how to use the FINO bank card. The content of the training was developed by FINO under the overall guidance of the evaluation team, while the video was designed in collaboration with a company specialized in street plays and movie production. Professional trainers employed by FINO delivered the training, and the bandhu who served the area was present to engage the clients in the workshop discussions. 4.2.2 Hypotheses Our key hypothesis is that financial education can indeed be effective in changing savings behavior among existing FINO account holders from low-income house- holds. More specifically, we are interested in (1) the impact of the financial education training on savings rates; and (2) the impact of the financial education training on financial knowledge, budgeting skills, and attitudes about good money management. The breakdown of the analysis in these two dimensions will allow us to examine in more depth the impact of the intervention along the variables in the causal chain. Furthermore, in order to understand whether there might be alternative and more cost-effective ways to deliver the message that saving is important, we try to investigate which elements are required for the success of the program. For this purpose, we introduced a cross-cutting intervention consisting of a simple 10-minute house visit to stress the importance of saving in formal instruments during the postharvest period (when people have more money), followed up by a monthly five-minute phone conversation to remind people to save. Besides high- lighting the importance of saving and the benefits of formal savings, this post- harvest intervention also included the setting of savings goals and the delivery of information about expected dates of bandhu presence in the area. Hence, we test whether a simple intervention is as effective as the classroom financial literacy training; we also check whether the postharvest reminders can leverage the effect of the training. 4.2.3 The sample The experiment was conducted on a random sample of individuals in villages where FINO operates. Villages were randomly selected to either receive the training or receive no training. Individuals from treatment villages who had FINO bank cards were then randomly selected to be offered the financial education training. The program was rolled out with the clients of 200 bandhus who were working in the villages of the two experiment districts of Varanasi and Azamgarh in Uttar Pradesh. These bandhus were selected from the list of all FINO agents using a distance-based dropping method to prevent contamination between treatment and control groups. From a pair of bandhus who were located in villages very close to each other (less than 1.25 kilometers), one bandhu was randomly dropped to 4.  Does financial education affect savings behavior?  ◾ 145 minimize spillovers; bandhus whose own service areas were far apart (more than 10 kilometers) were also dropped to make data collection and training easier. These 200 bandhus were then randomly assigned into treatment and control groups, and for each bandhu, 25 clients were randomly selected through FINO’s client records updated in January 2011. Using these FINO records, a prebase- line randomization check was undertaken to ensure that the sample was well balanced with respect to available demographic and account activity information. The results of the balance test showed that before the baseline there were indeed no observable differences between treatment and control FINO clients.2 Finally, from the list of 25 clients, a sample of 15 clients per bandhu was drawn for the survey interview.3 So, in total, 3,000 households were selected for the baseline survey, which took place in April 2011; the endline was collected one year later in April 2012. Around November 2011, the sample was further divided, assigning half of the treatment and half of the control groups to receive the postharvest intervention (house visit and reminder phone calls). As a result, four groups were formed: pure control, pure treatment, only postharvest intervention, and treat- ment plus postharvest intervention. 4.3 DATA 4.3.1 Questionnaires The questionnaires were designed to determine clients’ knowledge of financial behaviors and tools, and thus collected detailed information on various indica- tors assumed to play an important role in household financial well-being. They covered the following topics: household demographics, such as the number of family members, age, educational attainment, income, primary and secondary occupation; household ownership of assets; household expenditures, savings, and borrowings; respondent perceptions and behaviors about budgeting, involve- ment in household financial matters, and financial knowledge and attitudes; and respondent time and risk preferences. Additionally, a module on FINO services 2  The variables included in the prebaseline balance test were: percentage female; share of clients in the age groups 18–24, 25–59, and 60 and above; and share of clients who made at least one transaction in the six-month period before February 2011. 3  Buffers of 10 clients per bandhu were kept to ensure that, for each bandhu, the target of 15 clients could be surveyed. The first 15 clients (based on the sorting of randomly assigned client identifications) per bandhu were treated as the priority, and the buffer was only used in the extreme case where, in spite of every effort, the survey team was unable to find the client from the original list. 146  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES was added in the endline survey to assess the experience of the clients with the branchless banking delivered by FINO. 4.3.2 Outcome variables at baseline Table 4.2 presents descriptive information to show how rural households in our sample saved at the time of the baseline survey. To estimate savings, we rely on self-reported data—i.e., the respondents’ recall of the balance amount in each formal account and informal savings tool. In April 2011, mean formal savings were Rs 4,376 (US$51) or about twice the average monthly income, while mean informal savings (mostly home savings or savings in self-help groups) amounted to only Rs 619—showing that, in our sample, formal savings are more prevalent than informal savings. Even though TABLE 4.2  Descriptive statistics on household savings   VALUE OBSERVATIONS Formal savings Formal savings amount (Rs) 4,376 2,926 Has a formal savings account (%) 94 2,926 Amount of formal savings (for those who have at least an account) (Rs) 4,649 2,754 Amount of formal savings (for those who keep a nonzero balance) (Rs) 7,376 1,736 Has a FINO account (%) 87 2,926 Amount in FINO account (for those who have it) (Rs) 569 2,457 Amount in FINO account (for those who keep a nonzero balance) (Rs) 1,984 704 Has a formal savings account other than FINO (%) 57 2,926 Has an account in a nationalized bank (%) 51 2,926 Has an account with post office (%) 6 2,926 Has an account in a private bank (%) 5 2,926 Has an account with a nongovernmental organization (%) 2 2,926 Has an account in a chit fund (%) 1 2,926 Has an account in a nonbanking financial company (%) 1 2,926 Informal savings Informal savings amount (Rs) 619 2,928 Has an informal savings device (%) 28 2,928 Has savings at home (%) 23 2,928 Has savings with a self-help group (%) 2 2,928 Has savings with a neighbor (%) 2 2,928 Has savings with a friend (%) 1 2,928 Has savings with a shopkeeper (%) 1 2,928 Has other informal savings (%) 1 2,928 Note: Baseline values. 4.  Does financial education affect savings behavior?  ◾ 147 technically the entire sample had no-frills savings accounts served by FINO, only 87 percent of households reported having an account through FINO, suggesting that some clients were either not aware they had FINO accounts or did not under- stand what they were signing up for when they opened the account. Noticeably, even though about 94  percent of households reported having a formal savings account at baseline, only 59  percent had a nonzero balance, suggesting that other constraints than access to bank services limited savings amounts. Considering only FINO accounts, the figures are worse: only 24 percent of households appeared to use the account for savings by keeping a nonzero balance.4 Such baseline levels present potential scope for financial education training to help develop better saving behaviors. Almost 60  percent of FINO account holders also had at least one other formal savings account. About 51  percent had an account in a nationalized bank, 6 percent in a post office, 5 percent in a private bank, and 2 percent with a nongovernmental organization (the categories are not mutually exclusive). This picture is similar to the percentages presented by Demirguc-Kunt and Klapper (2012) based on the nationally representative Global Findex data set in India. They showed that in 2011 between 22  and 56  percent of the population (with exact percentages depending on income quintile) had an account at a formal financial institution. The fact that half of the respondents had a national bank account suggests that, while banking may be difficult, considered unimportant, or expen- sive in the areas where FINO operates, people are interested in obtaining formal savings despite the extra costs. This finding also reflects that in 2006 the Reserve Bank of India imposed on all commercial banks the introduction of free no-frills accounts (Thyagarajan and Venkatesan 2008). Thus, in the study area, the extra cost of keeping another formal account consists mostly of the traveling cost of reaching the nearest bank.5 To define our indicator of financial literacy we follow the approach introduced by Cole, Sampson, and Zia (2011) and Carpena et al. (2011). The former presents the first nationally representative measure of financial literacy in a developing world, while the latter identifies “three distinct dimensions of financial knowl- edge”: financial numeracy, basic awareness of financial choices, and attitudes toward financial decisions. Thus, our questionnaire covered different aspects of financial literacy, including budgeting skills, interest in financial matters, basic financial numeracy, financial product awareness, and financial attitudes. Budgeting quality refers to the skills of making a budget, writing it down, eval- uating it as helpful, and being able to stick to it. Interest in financial matters covers 4  The low deposit in FINO accounts might have been aggravated by the absence of bandhus in the areas and/or by trust issues. FINO estimated that in the villages where it chose to operate, a bank branch was at least 5  4–5 kilometers away. 148  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES responses regarding involvement in household financial matters and self-as- sessed understanding of information related to financial products. Table 4.3 pres- ents the summary statistics for these first two measures of financial literacy. In the baseline, budgeting quality appeared to be particularly low: 73 percent of the respondents did not make a budget and, even when they did, they mostly kept it only mentally. A considerable fraction of individuals (24 percent) also reported not being involved at all in financial matters and not being actively interested in finan- cial topics (48 percent). When asked about their understanding of financial prod- ucts, almost half of the respondents stated that, in general, they rarely or never understood financial information, especially on loan and savings products. Even though these percentages might not represent a particularly alarming picture, they still signal a generalized lack of financial understanding and involvement. TABLE 4.3  Descriptive statistics on budgeting quality and interest in financial matters   MEAN VALUE OBSERVATIONS Budgeting quality Makes a budget 0.27 2,922 Writes the budget (if makes it) 0.05 798 Has been helped by the budget (if makes one) 0.04 798 Is able to stick to the budget (if makes one) 0.03 798 Interest in financial matters Is involved in financial matters (dummy) 0.76 2,642 Generally understands loan information (dummy) 0.52 2,630 Generally understands savings information (dummy) 0.58 2,659 Generally understands insurance information (dummy) 0.62 2,726 Actively seeks information about financial topics (dummy) 0.52 2,726 Table 4.4 presents descriptive statistics regarding financial knowledge and compares our results to the findings of Cole, Sampson, and Zia (2011) from rural India and Indonesia and of Carpena et al. (2011) from urban India. The first measure of financial numeracy is based on the study by Cole, Sampson, and Zia (2011), which is in turn very close to the work of Lusardi and Mitchell (2006), who pioneered the research on financial literacy. It includes a question on compound interest, along with one on interest rates versus inflation (see table 4A.1 in the annex to this chapter for details). The main purpose of these questions is to test respondents’ understanding of basic economic concepts (i.e., inflation, interest rate, and compound interest) that are considered indispens- able for making financial decisions. For this reason, we rename our measure of financial numeracy as “understanding of basic economic concepts.” Our indicator appears to be in line with previous estimations: in our sample, the mean share 4.  Does financial education affect savings behavior?  ◾ 149 TABLE 4.4  Descriptive statistics and comparability of our measures of financial knowledge OUR RURAL URBAN SAMPLE INDIAa INDONESIAa INDIAb Understanding of basic economic concepts Compound interest % Correct 70 59 78 % Do not know 15 30 15 Interest rate versus % Correct 71 25 61 inflation % Do not know 11 38 16 Both questions % Correct on average 71 42 70 Observations 2,931 1,496 3,360 Financial awareness Is one crop safer % Correct 31 31 28 than multiple crops? % Do not know 8 6 4 Observations 2,931 1,496 3,360 Knows to include both income and expenses 0.77 0.85 in household budget Knows will get money back if bank closes 0.32 0.70 Knows borrowing money for Diwali is 0.70 0.62 unproductive loan All questions, on average 0.58 0.72 Observations 2,851 221 Financial attitudes Advice to construction worker 0.66 0.81 Advice to friend with bright child 0.77 0.93 Advice to auto driver about loans 0.40 0.92 Advice about buying a TV 0.84 0.95 All question, on average 0.68 0.90 Observations 2,901 221 a. Sample from Cole, Sampson, and Zia 2011. b. Sample from Carpena et al. 2011. of correct answers is 71  percent; it was 70  percent in the Cole, Sampson, and Zia (2011) sample representative of the Indonesian population and 42 percent for their sample of 1,500 poor households in rural Gujarat. Thus, our sample is more comparable with average samples in other developing countries than with Indian subsamples of poorer laborers in subsistence agriculture. Second, to define basic awareness of financial choices, we follow the paper by Carpena et al. (2011); they define this indicator as the “knowledge of funda- mental financial planning concepts, as well as details of financial products, [such as] understanding of deposit insurance or of the purpose of a household budget” (Carpena et al. 2011, 13–14) (see table 4A.1). Since, in this case, the comparison 150  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES is with a subsample from urban India, the means in our sample variables are on average lower than the means showed in Carpena et al. (2011), especially for more complicated concepts such as deposit insurance. Finally, financial attitudes are also measured, as in Carpena et al. (2011), by presenting hypothetical situations to respondents and asking them about the financial products or financial advice they would suggest in the given scenario (see table 4A.1). Some of these questions have an ascending range of correct answers, so they are coded as continuous variables from 0 to 1 with 1 equal to the best financial option and 0.5 weight on the second best option. Again, with an average mean of 0.68, our measure of financial attitudes is on average lower than the one presented by Carpena et al. (2011) based on an urban subsample. 4.3.3 Socioeconomic background of clients and balance test Respondents in our sample are of relatively low socioeconomic status. House- hold heads are mostly males and are on average 45 years old; about 40 percent of them are illiterate. Households are mostly Hindu and have an average of six to seven members, of which four are adults. About 70  percent of the households own land, with income from harvest and livestock contributing about 45 percent of total income. At baseline, the income from primary and secondary occupations was on average only Rs 1,079 (US$20)—slightly above the poverty threshold (the state poverty line for rural Uttar Pradesh in 2010 was fixed at Rs 663). The mean of total household income was Rs 2,028 (US$38), while the mean of total household expenditures was Rs 1,773 (US$33); this suggests that, on average, households did not manage to save much. In fact, about 50 percent of households resorted to loans, with the average loan amount equal to Rs 4,839 (US$90). Table 4.5 presents the results of the balance test relative to all basic house- hold characteristics. The baseline variables seem well balanced, except for the total number of outstanding formal loans. In order to avoid any bias that might arise in estimating treatment effects, we include all unbalanced variables as controls in the empirical analysis.6 Other variables unbalanced at baseline are whether the client has at least secondary 6  education, whether the client had a loan, the number of females in the household, whether the client had a non-FINO bank account, and the level of overall financial literacy. These variables are included as controls in all regressions. 4.  Does financial education affect savings behavior?  ◾ 151 TABLE 4.5  Sample characteristics and balance test CONTROL TREATMENT VARIABLE MEAN MEAN P-VALUE Gender of household head (dummy) 0.71 0.72 0.36 Age of household head 44.68 45.52 0.2 Whether household head is illiterate (dummy) 0.38 0.41 0.32 Whether household head has primary education (dummy) 0.18 0.18 0.88 Whether household head has secondary education (dummy) 0.26 0.26 0.94 Whether household head has higher than secondary education (dummy) 0.11 0.10 0.52 Whether religion is Hindu (dummy) 0.95 0.94 0.77 Whether religion is Muslim (dummy) 0.05 0.06 0.8 Whether belong to general caste (dummy) 0.11 0.13 0.63 Whether belong to schedule caste (dummy) 0.30 0.35 0.16 Whether belong to other backward community (dummy) 0.54 0.49 0.17 Total number of members in the household 6.74 6.96 0.17 Total number of adults (≥ 18 years) 4.03 4.10 0.49 Whether owns land (dummy) 0.71 0.70 0.9 Size of land owned 21.76 22.26 0.59 Household income from primary occupation (Rs) 1,019.24 1,027.3 0.92 Household income from primary and secondary occupations (Rs) 1,094.66 1,062.5 0.85 Household income from harvest, livestock, and other sources (Rs) 821.79 1,074.89 0.54 Total household income (Rs) 1,917.08 2,139.72 0.61 Total household income per capita (Rs) 287.66 311.43 0.66 Amount of household expenditures: consumed at home (Rs) 1,542.38 1,596.49 0.48 Amount of household expenditures: consumed outside home (Rs) 32.83 39.86 0.39 Amount of household expenditures: cigarettes, tobacco, alcohol (Rs) 68.93 58.75 0.12 Total amount of household expenditures in last 14 days (Rs) 1,837.52 1,708.23 0.39 Number of rooms 3.28 3.21 0.59 Score for 1st component of full asset list 0.05 −0.05 0.41 Quality of roof 3.5 3.4 0.12 Total number of outstanding formal loans 0.09 0.11 0.09 Total number of outstanding loans 0.5 0.5 0.96 Total outstanding formal loan amount (Rs) 1,640.9 1,776.64 0.74 Total outstanding loan amount (Rs) 4,623.28 5,053.99 0.61 Index of risk preferences 2.32 2.39 0.23 Discount rate or Index of time preferences 2.84 2.86 0.74 Index of ambiguity preferences 2.11 2.19 0.28 152  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 4.4 IMPACTS ON SAVINGS AND FINANCIAL LITERACY 4.4.1 Summary statistics of the outcome variables The first outcome of interest is whether the program caused any increment in the amount of the following types of savings: FINO savings, formal savings (including FINO, post office, and other commercial bank accounts), non-FINO formal bank savings, savings in other nationalized banks, informal savings, and total savings. All savings amounts are capped at the 99th  percentile in order to eliminate outliers. The second outcome of interest is whether the intervention resulted in any improvement in the different indicators of financial literacy described above. Each indicator is expressed as the average of the answers belonging to the same dimension of financial literacy. Table 4.6 shows the relevant summary statistics and includes all households present in both the baseline and endline surveys. Columns 1–3 give values for the pre-intervention, while columns 4–6 refer to the post-intervention. All standard errors reported are adjusted for clustering at the village level (bandhu service area).7 Column 6 gives a first approximation of the impact of the financial literacy training on savings; the difference between treatment and control endline savings measures is always positive and statistically significant, in spite of the fact that the control group seems to improve along with the treatment group. To mini- mize measurement errors and show only the effect of training, we repeat the comparisons including only pure treatment and pure control (i.e., excluding the beneficiaries of the postharvest intervention). As expected, the difference is even starker in this case: the treatment group increases total savings on average by 154 percent, while the control group experiences an increase of 66 percent. The fact that the control group improves as well is not so surprising if we take into account the fact that Uttar Pradesh is one of India’s states that is growing more rapidly than others. The control group also improves in financial literacy. This could be because people are replying to the same questions again. Looking at the significance levels of column 3 of table 4.6, it is clear that controlling for baseline values is important as not all the financial measures were perfectly balanced at baseline—in spite of the fact that randomization was successful and significant differences appeared only at a rate equal to that given by chance. 7  See table 4A.3 for the nonresponse rates of all the outcome measures in table 4.6. 4.  Does financial education affect savings behavior?  ◾ 153 TABLE 4.6  Pre- and post-intervention differences PRE-INTERVENTION POST-INTERVENTION TREATMENT CONTROL DIFFERENCE TREATMENT CONTROL DIFFERENCE SAVINGS (1) (2) (3) (4) (5) (6) FINO 303.26 324.12 −20.86 180.5 85.37 95.13***     (69)     (26.86) Formal 2,603.96 2,725.83 −121.88 6,505.1 5,267.4 1,237.70*     (341.61)     (665.07) Non-FINO 2,172.39 2,265.39 −93.01 6,263.43 5,147.84 1,115.59* formal     (301.11)     (656.37) Nationalized 1,938.00 2,028.29 −90.29 5,348.78 3,931.25 1,417.53** bank     (302.89)     (589.11) Informal 350.50 333.08 17.42 363.07 375.84 −12.77     (68.17)   (51.43) Total 2,952.45 3,055.17 −102.72 6,868.17 5,643.23 1,224.94*     (353.16)     (675.96) Savings considering only pure treatment and pure control     FINO 364.91 336.53 28.38 152.8 85.80 67.0**     (99.19)     (30.63) Formal 2,579.13 2,971.29 −392.16 7,158.73 5,043.47 2,115.26**     (500.56)     (948.5) Non-FINO 2,089.95 2,461.53 −371.58 6,936.93 4,940.27 1,996.66** formal     (447.8)     (941.3) Nationalized 1,854.82 2,224.42 −369.6 6,165.83 3,823.54 2,342.29*** bank     (458.22)     (865.08) Informal 385.23 288.04 97.19 353.79 372.08 −18.29     (94.42)   (68.73) Total 2,960.63 3,255.58 −294.95 7,512.53 5,415.55 2,096.98**     (517.96)     (961.78) Financial literacy           Budgeting 0.26 0.29 −0.031 0.42 0.4 0.015 quality     (0.03)     (0.04) Interest in finan- 0.41 0.44 −0.03** 0.49 0.48 0.008 cial matters     (0.02)     (0.01) Basic economic 0.8 0.83 −0.034* 0.64 0.63 0.012 understanding     (0.02)     (0.02) Financial 0.55 0.58 −0.027* 0.58 0.58 −0.004 awareness     (0.01)     (0.01) Financial 0.75 0.78 −0.025* 0.60 0.59 0.01 attitudes     (0.01)     (0.01) Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. Columns 1, 2, 4, and 5 show mean values. 154  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 4.4.2 The estimation method For our estimation, we employ an analysis of covariance specification (Bruhn and McKenzie 2009; McKenzie 2011).8 We regress the outcome indicator on the treat- ment status of household h controlling for the baseline value of the indicator: Yh POST = α + βTh + ηYh PRE + δXh PRE + εh POST (4.1) where X represents household control variables unbalanced at the baseline and standard errors are adjusted for clustering at the village/bandhu level. The treat- ment effect is thus estimated by β. As in similar studies, the take-up rate for the financial literacy training was less than full. Defining a client as having attended the program if he or she attended class for at least one hour, class attendance was irregular with only 80 percent of the invited ever attending class. Table 4A.2 presents a list of baseline character- istics that might have influenced attendance status in the treatment group. There are no noticeable differences in savings or financial literacy levels between those who attended and those who were offered the class but did not attend; the only exception being that attendees had more positive financial attitudes. Also, clients who attended seem to be more likely to be females and older. To take into account imperfect compliance, in addition to the standard ordi- nary least squares (OLS) intention-to-treat (ITT) regressions that estimate overall impacts, we also employ instrumental variable (IV) regressions that use the initial assignment (the ITT) as an instrument for actual treatment to assess the treat- ment effect on the treated (ToT). In explaining our results, we focus on the ITT estimates while we present the ToT parameters for comparison. Finally, differential attrition between the treatment and comparison groups could potentially bias our results. To minimize attrition, the survey team under- took a rigorous search for tracking back the baseline sample (including pre-end- line house visits) and, in some cases, used the help of the bandhus to relocate the households. These efforts ensured a low attrition rate: attrition was only 2.8  percent and 2.1  percent, respectively, in the comparison and treatment groups. Furthermore, the baseline characteristics of households that left the sample were similar in the treatment and comparison groups, suggesting that the factors leading to attrition were the same—and, consequently, that attendance and treatment status were unrelated (the results of the estimation regressing attrition on treatment assignment are shown in table 4A.4). Therefore, attrition is unlikely to be a problem in our estimation strategy. As a robustness check, we also replicated the estimation regressing the change in the 8  outcome indicator (post-intervention value in levels minus pre-intervention value in levels) on the treatment status, controlling for the baseline value of the indicator (YPOST − YPRE = α + βT + δXPRE + ηYPRE + ε POST, as in Banerjee et al. 2007). Such a robustness check gave the same results, confirming the validity of our estimates. 4.  Does financial education affect savings behavior?  ◾ 155 4.4.3 Estimates of the average impacts on savings Table 4.7 displays the average impacts of the financial education intervention on increments in the savings amount. These estimates show a positive and substan- tial treatment effect both in the ITT specifications (panel A) and in the ToT (IV) specifications (panel B). Looking at the ITT specification, the program caused an increase in total savings of Rs 1,647 (about US$30) from April 2011 to April 2012, which equals an increase of 29  percent compared to the endline total savings in the control group. Considering the IV specification, the effect of training attendance is even higher and equal to an increase in total savings of almost Rs 2,000 (about US$38), or 35 percent compared to the control group. The effect is robust to the different specifications and even though the magnitude changes slightly depending on the regression, the treatment effect remains positive and significant. Foremost, the increment does not seem to be determined only by a marketing effect on FINO savings because formal non-FINO savings also grew substantially. In fact, still considering the ITT estimates, the program increased FINO savings by Rs 88 and formal savings other than FINO by Rs 1,559. Therefore, given that the program was mostly aimed at increasing savings, it appears to be quite effective. Table 4.8 shows that the postharvest intervention did not have a significant effect on savings. It was successful only in increasing FINO savings by an extra Rs 56 (Rs 122 in total, compared to an increase of Rs 66 for those who were assigned to receive only the training), possibly because the delivery of information TABLE 4.7  Average impacts on savings NON-FINO NATIONAL- FINO FORMAL FORMAL IZED BANKS INFORMAL TOTAL (1) (2) (3) (4) (5) (6) Panel A. ITT estimates Treatment 87.96*** 1,681*** 1,559** 1,392** −13 1,647** (28.51) (630.4) (626.8) (570) (50.9) (640.5)   Panel B. IV estimates  Attendance 104*** 2,000*** 1,855** 1,650** −15.47 1,961*** (33.46) (747.9) (743.4) (672.2) (60.37) (760.1)   Observations 2,666 2,916 2,916 2,661 2,918 2,919 R-squared (OLS estimates) 0.02 0.07 0.07 0.08 0.02 0.07 R-squared (IV estimates) 0.03 0.07 0.06 0.08 0.02 0.07 Mean of endline variable 85.37 5,267.4 5,147.84 3,931.25 375.84 5,643.23 in control group Note: Robust standard errors, clustered at the agent/village level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Controls include the baseline values of the dependent variable and all the vari- ables unbalanced at the baseline: whether the client has at least secondary education, whether the client had a loan, the number of females in the household, whether the client had a non-FINO bank account and the level of overall financial literacy. 156  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 4.8  Average impacts of treatment and postharvest reminders on savings NON-FINO NATIONAL- FINO FORMAL FORMAL IZED BANKS INFORMAL TOTAL (1) (2) (3) (4) (5) (6) Only treatment 65.9** 2,734*** 2,617*** 2,328*** −22.08 2,648*** (31.48) (900.0) (904.1) (831.4) (69.34) (914.0) Treatment and 1,22.1** 1,367* 1,146 863.0 19.0 1,334 postharvest reminders (48.47) (810.5) (812.4) (744.6) (77.52) (828.2) Only postharvest 8.06 930.3 838.5 588.4 21.23 869.4 reminders (25.11) (808.2) (804.6) (707.0) (67.34) (814.3) Observations 2,666 2,916 2,916 2,661 2,918 2,919 R-squared 0.02 0.07 0.07 0.08 0.02 0.07 Note: See table 4.7. Results refer to ITT estimates. about the expected dates of bandhu presence or the actual presence of the bandhus helped improve delivery of the FINO service. The relative ineffectiveness of the reminder intervention suggests that even if the postharvest is a relevant period to boost attention toward savings (Duflo, Kremer, and Robinson 2011) and even if reminders have been found to be effective in previous field experiments in Latin America and Asia (Karlan et al. 2012; Kast, Meier, and Pomeranz 2012), a simple targeted phone call is not as successful as a two-day training class with video-illustrated lessons and interactive discussions to underline the importance of savings and ensure that the contents are internalized. 4.4.4 Estimates of the average impacts on financial literacy Table 4.9 illustrates the average impacts of the financial education intervention on improvements in different aspects of financial literacy. It shows that the only dimension of financial literacy that appears to have been positively affected by the treatment is financial attitudes. The IV estimates show that the intervention increased the financial attitudes indicator of individuals in the treatment group by 3 percent, compared to the endline mean in the control group. This result is somewhat in line with what was previously found in the liter- ature; in particular, it can be related to the conclusions of Carpena et al. (2011), which highlighted positive effects of financial literacy on financial attitudes and basic financial awareness. The exception is that our intervention did not seem to have any effect on financial awareness. However, this may be due to various reasons that might not ensure comparability of the results. First, a problem that has undermined our ability to identify a significant impact is that the survey was answered by the FINO client only when he or she was available; in the remaining cases, it was answered by the other most knowledgeable person in the house- hold. Therefore, we repeat the estimation on the restricted sample of clients who 4.  Does financial education affect savings behavior?  ◾ 157 TABLE 4.9  Average impacts on financial literacy TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY     (1) (2) (3) (4) (5) (6) Panel A. ITT estimates Treatment 0.022 0.013 0.025 0.002 0.016 0.003 (0.04) (0.01) (0.02) (0.01) (0.01) (0.01) Panel B. IV estimates Attendance  0.026 0.016 0.03 0.003 0.019* 0.003 (0.05) (0.02) (0.02) (0.01) (0.01) (0.01) Observations 2,907 2,848 2,739 2,866 2,890 2,883 R-squared (OLS estimates) 0.08 0.01 0.04 0.03 0.02 0.02 R-squared (IV estimates) 0.07 0.01 0.04 0.03 0.03 0.02 Mean of endline variable 0.4 0.48 0.63 0.58 0.59 0.68 in control group Note: See table 4.7. responded to both the baseline and endline surveys because—even though this might be a selected sample—we need to check whether the absence of a signif- icant impact on financial knowledge can be attributed to measurement error. As shown in table 4.10, the new results confirm our previous findings and, as before, the only significant treatment coefficient is the one on financial attitudes. The magnitude of impact is now higher and equal to a 4 percent (ITT) to 5 percent (IV) increase as compared to the endline value in the control group. It could also be because our financial education training was tailored differently than previously evaluated financial education programs, as it was more focused on increasing savings and less oriented toward improving financial knowledge. In fact, even restricting the measure of financial knowledge to include only the questions that must have been well underlined in the training does not modify our results (column 6 of tables 4.9 and 4.10).9 9  We include in this indicator the following questions: question 2 on financial numeracy; questions 1, 2, and 4 on financial awareness; and questions 1 and 4 on financial attitudes. Thus, we exclude the following questions: question 1 on financial numeracy, because it involves numerical skills the training did not cover; question 3 on financial awareness, because it deals with the concept of deposit insurance which was not explicitly included in the training program; and questions 2 and 3 on financial attitudes, because they are framed in a very subjective way. See table 4A.1 for the detailed list of questions. 158  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 4.10  Average impacts on financial literacy for the subsample of clients who answered both the baseline and the endline survey TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY     (1) (2) (3) (4) (5) (6) Panel A. ITT estimates Treatment  0.004 0.019 0.016 −0.005 0.025** 0.0003 (0.05) (0.02) (0.02) (0.01) (0.01) (0.01) Panel B. IV estimates Attendance 0.004 0.02 0.019 −0.005 0.03** 0.0004 (0.05) (0.02) (0.03) (0.01) (0.01) (0.01) Observations 1,584 1,554 1,508 1,565 1,576 1,574 R-squared (OLS estimates) 0.09 0.02 0.04 0.05 0.04 0.04 R-squared (IV estimates) 0.09 0.02 0.04 0.05 0.04 0.04 Note: See table 4.7. Our findings are also similar to the results of Doi, McKenzie, and Zia (2012), who find a positive and significant effect of financial education on financial atti- tudes and financial awareness; however, they use an indicator of awareness that is more relaxed than other measures of financial knowledge and assess it by asking respondents only whether they have heard about different financial products. Interestingly, the authors redefine financial attitudes as “applied financial knowl- edge,” since the questions included in this indicator deal with real-life situation where an individual could use in practice his or her financial understanding—for instance, to suggest an appropriate financial product to someone who is worried about meeting expenses if sick. Finally, table 4.11 replicates the results using standardized indicators to ensure the comparability of the effects. Normalized scores for each dimension of finan- cial literacy are calculated like school test scores, by first summing the results of each question belonging to that dimension and then standardizing by subtracting the mean pre-intervention score of the comparison group and dividing by the standard deviation of the pre-intervention scores of the comparison group as in Banerjee et al. (2007). The ITT estimates show that the intervention increased the average score for financial attitudes of individuals in the treatment group by 0.09 standard deviations (equal to a 4  percent increase). The IV estimates are similar and just slightly higher in magnitude: the effect on financial attitudes for those who attended the training is equal to a 0.11 standard deviation increase or a 5 percent increment. This result is very close to the increment estimated using the mean value of financial attitudes. 4.  Does financial education affect savings behavior?  ◾ 159 TABLE 4.11  Average impacts on financial literacy using the standardized indicators TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY (1) (2) (3) (4) (5) (6) Panel A. ITT estimates Treatment  0.066 0.068 0.052 0.03 0.09** 0.071 (0.13) (0.06) (0.04) (0.04) (0.04) (0.05) Panel B. IV estimates Attendance  0.078 0.081 0.062 0.036 0.107** 0.084 (0.15) (0.08) (0.04) (0.04) (0.05) (0.06) Observations 2,921 2,921 2,921 2,921 2,921 2,921 R-squared (OLS estimates) 0.08 0.01 0.04 0.03 0.02 0.03 R-squared (IV estimates) 0.08 0.01 0.04 0.03 0.02 0.03 Mean of endline variable 1.25 2.4 1.24 2.23 2.26 3.9 in control group Note: See table 4.7. 4.4.5 Heterogeneity To test for heterogeneity in the treatment effect based on observable character- istics, we run the following set of regressions: Yh POST = α + βTh + γTh × TRAITh + ηYh PRE + δXh PRE + εh POST (4.2) where TRAIT is the vector of background characteristics along which theory would predict heterogeneity in the treatment impacts and where X also includes TRAIT among the controls. The effect of the treatment for the subgroup of people with a given trait is given by the sum of the coefficients β and γ, and if γ is significantly different from zero then there is evidence of heterogeneity in the treatment effect for that trait. Since the ITT estimates are more relevant for policy impacts and since the IV estimates are close to the ITT ones, we only show the tables with the ITT estimates for this heterogeneity discussion.10 For comparison purposes, we The ITT estimates are more relevant for policy impacts because people cannot be forced 10  to attend, and policy makers need to know what would be the overall effect of treatment taking into account that not everyone assigned to treatment might comply. In addition, even though in our case attendance is orthogonal to treatment, the sample of those who 160  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES also choose to use the standardized measures of financial literacy instead of the mean values (the tables using the mean values are in annex B). We estimate equation 4.2 for the following different baseline characteristics: client education (at least secondary), the baseline measure of overall financial literacy, client gender, client time preferences, the baseline level of household per capita total expenditures, and an indicator for having a formal savings account other than FINO at baseline.11 In line with analysis conducted by Cole, Sampson, and Zia (2011), who found that a financial education program had a modest effect and positively influenced financial behaviors only for those with limited education and financial literacy, we also test the role of education and baseline financial literacy to check whether our sample offers a similar picture. Tables 4.12 and 4.13 show the heterogeneous treatment impacts on savings increments and on improvements in financial literacy, respectively. For education, there is no heterogeneity in the treatment effect on savings, but there is indeed heterogeneity in the treatment effect on financial attitudes. In particular, more educated people seem to have changed their attitudes less than other clients in the treatment group; this confirms find- ings by Cole, Sampson, and Zia (2011) about greater effects of financial education on the less educated. On the contrary, the heterogeneous impact of baseline financial literacy runs in opposite direction to what would be expected based on Cole, Sampson, and Zia (2011) results. In our sample, those with better baseline financial literacy positively and significantly improved their interest in financial matters by 0.2 standard deviations (0.08 + 0.12); they even slightly increased their basic understanding of basic economic concepts by 0.11 standard deviations (0.06  +  0.05). Also, they significantly decreased their informal savings by Rs 80 (−18 and −62, table 4.12). This finding might indicate that a sufficient prior famil- iarity with financial concepts helps in learning more during a financial education program. Similar to Dupas and Robinson (2013b) on health savings, we also control for heterogeneity in treatment impacts for gender and time preferences. Tables 4.14 and 4.15 show the results of this additional test. There is no heterogeneity for gender, but there is a strong heterogeneous effect for time preferences. Specifi- cally, more impatient individuals (those with higher discount rates) improved their financial attitudes significantly less compared to an average client in the treatment attended is a selected sample with peculiar characteristics different from those of the average population. The tables shown focus only on the treatment effects, but the selected characteristics 11  have significant direct effects, too. In particular, education and baseline financial literacy are significantly and positively correlated with endline savings and financial capability measures, being impatient decreases savings, females have less endline savings in national bank accounts and score worse in financial literacy, and those with higher baseline expen- ditures increased their financial numeracy skills. 4.  Does financial education affect savings behavior?  ◾ 161 TABLE 4.12  Heterogeneity of impacts on savings for client education and baseline financial literacy NON-FINO NATIONAL- FINO FORMAL FORMAL IZED BANKS INFORMAL TOTAL (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for client education (at least secondary) Treatment  90.12*** 1,231** 1,120* 868.4* 9.89 1229** (28.50) (589.7) (586.1) (515.9) (52.64) (601.4) Treatm X Education  −9.17 1,874 1,829 2,220 −95.73 1,747 (55.91) (1,605) (1,597) (1,473) (108.7) (1,639) Observations 2,666 2,916 2,916 2,661 2,918 2,919 R-squared 0.02 0.07 0.07 0.08 0.02 0.07 Panel 2. Heterogeneous impacts for baseline financial literacy Treatment 86.80*** 1,720*** 1,594** 1,439** −18.10 1,683** (29.03) (641.1) (636.4) (583.6) (51.64) (652.3) Treatm X Basel −16.62 499.4 437.3 590.4 −62.31* 428.8 Financ Literacy (25.88) (516.4) (506.2) (454.2) (36.62) (523.1) Observations 2,666 2,916 2,916 2,661 2,918 2,919 R-squared 0.02 0.07 0.07 0.08 0.02 0.07 Note: See table 4.7. TABLE 4.13  Heterogeneity of impacts on financial literacy for client education and baseline financial literacy TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY     (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for client education (at least secondary) Treatment 0.099 0.054 0.062 0.053 0.121*** 0.088 (0.13) (0.07) (0.04) (0.04) (0.05) (0.06) Treatm X Education −0.141 0.058 −0.04 −0.093 −0.13* −0.072 (0.23) (0.1) (0.06) (0.08) (0.07) (0.1) Observations 2,921 2,921 2,921 2,921 2,921 2,921 R-squared 0.08 0.01 0.04 0.03 0.03 0.03 Panel 2. Heterogeneous impacts for baseline financial literacy Treatment 0.064 0.078 0.056 0.027 0.09** 0.07 (0.13) (0.06) (0.04) (0.04) (0.04) (0.05) Treatm X Basel −0.021 0.123** 0.049* −0.044 −0.006 −0.007 Financ Literacy (0.09) (0.05) (0.03) (0.03) (0.03) (0.04) Observations 2,921 2,921 2,921 2,921 2,921 2,921 R-squared 0.08 0.01 0.04 0.03 0.02 0.03 Note: See table 4.7. 162  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 4.14  Heterogeneity of impacts on savings for client gender and time preferences NON-FINO NATIONAL- FINO FORMAL FORMAL IZED BANKS INFORMAL TOTAL (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for client gender (female dummy) Treatment 65.91** 1,703** 1,594* 1,222 −33.12 1,660* (32.59) (835.6) (829.8) (782.6) (63.94) (849.7) Treatm X Female 53.89 −45.34 −76.71 424.1 49.52 −22.85 (34.08) (1,060) (1,061) (950.9) (83.91) (1,076) Observations 2,666 2,916 2,916 2,661 2,918 2,919 R-squared 0.02 0.07 0.07 0.08 0.02 0.07 Panel 2. Heterogeneous impacts for time preferences (discount rate) Treatment  143.6** −510.6 −801.8 −566.5 −109.5 −558.5 (59.84) (1,281) (1,284) (1,195) (102.8) (1,292) Treatm X Discount −18.90 745.6* 800.0** 677.1** 32.92 753.9* Rate  (18.63) (405.4) (400.2) (343.1) (34.73) (412.8) Observations 2,633 2,877 2,877 2,626 2,879 2,879 R-squared 0.02 0.07 0.07 0.08 0.02 0.08 Note: See table 4.7. TABLE 4.15  Heterogeneity of impacts on financial literacy for client gender and time preferences TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY     (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for client gender (female dummy) Treatment  0.073 0.066 0.045 0.006 0.055 0.074 (0.15) (0.07) (0.04) (0.04) (0.05) (0.06) Treatm X Female −0.017 0.008 0.018 0.062 0.088 −0.006 (0.17) (0.08) (0.05) (0.06) (0.06) (0.08) Observations 2,921 2,921 2,921 2,921 2,921 2,921 R-squared 0.08 0.02 0.04 0.04 0.03 0.03 Panel 2. Heterogeneous impacts for time preferences (discount rate) Treatment −0.448** 0.082 0.046 −0.078 −0.04 −0.077 (0.23) (0.13) (0.07) (0.09) (0.09) (0.11) Treatm X Discount 0.179*** −0.004 0.001 0.035 0.045* 0.05 Rate  (0.06) (0.04) (0.02) (0.02) (0.03) (0.03) Observations 2,881 2,881 2,881 2,881 2,881 2,881 R-squared 0.08 0.01 0.04 0.03 0.03 0.03 Note: See table 4.7. 4.  Does financial education affect savings behavior?  ◾ 163 group and even scored worse in budgeting skills. Accordingly, they also saved significantly less than average and their total savings after the training increased only by Rs 195 (Rs 754 − Rs 559, about US$3). This result is in line with the Dupas and Robinson (2013b) findings on the importance of time preference bias in influencing saving behaviors. In fact, the authors showed that a simple safe box significantly helped people in rural Kenya save more through a mental accounting effect; such a basic technology was not useful for people with present-biased preferences who managed to save only when facing social pressure through a savings device with a strong social commitment feature (a health pot at a ROSCA). Finally, we test for heterogeneity in the treatment effects based on baseline expenditures (per capita total expenditures) and on whether the client already had another formal account other than with FINO. Tables 4.16 and 4.17 show that there is almost no heterogeneity in expenditures, while there is a heterogeneous impact for those clients who also had a non-FINO formal savings account. It seems that those who already had a formal non-FINO savings account increased their interest in financial matters by 0.19 standard deviations (0.27 − 0.08) and they even slightly improved their overall financial knowledge (i.e., only including the questions that must have been stressed in the training) by 0.01 standard deviations (0.14 − 0.13). Moreover, after the training, they also saved more, expe- riencing an increment in total savings equal to Rs 2,544 (Rs 2,012 + Rs 532). Thus, these results possibly indicate that the intervention was more effective in influ- encing the behavior of the clients who already had an exposure to formal savings bank accounts, rather than those who were linked with the banking system for the first time through the no-frills savings account served by FINO. 4.4.6 Estimates of the average impacts on household wealth Table 4.18 shows the average impacts of the financial literacy training on consump- tion. Interestingly, clients in the treatment group decreased their unnecessary expenses in cigarettes, tobacco, beetle nuts, and alcohol. This effect is statisti- cally significant, but not economically relevant since the decrease corresponds to only US$0.5. Nonetheless, it is a remarkable finding, and the low magnitude might be due to the short time frame of the evaluation or to general underreporting of tobacco and alcohol expenses. Such results suggest that the intervention was successful in increasing savings at least in part through boosting commitment to save and improving money management. Table 4.19 completes the results on household wealth, illustrating the average impacts of the intervention on loans and assets. As expected, there are no signif- icant effects on total loans and assets sold after the intervention, demonstrating that clients did not increase savings through costly actions—i.e., by borrowing money or disinvesting. On the contrary, it seems that after training, individuals in the treatment group bought more assets. 164  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 4.16  Heterogeneity of impacts on savings for baseline per capita total expenditures and whether client had non-FINO formal savings account NON-FINO NATIONAL- FINO FORMAL FORMAL IZED BANKS INFORMAL TOTAL (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for baseline per capital total expenditures Treatment 84.59*** 1,287* 1,150 1,080* −12.66 1256* (28.49) (696.0) (697.3) (643.5) (50.56) (702.2) Treatm X PC Tot 0.01 1.75 1.8 1.49 −0.03 1.72 Expend (0.03) (1.73) (1.8) (1.48) (0.04) (1.73) Observations 2,656 2,906 2,906 2,653 2,908 2,909 R-squared 0.02 0.08 0.07 0.09 0.02 0.08 Panel 2. Heterogeneous impacts for whether client already had a non-FINO formal savings account Treatment 68.00* 599.4 591.7 989.9 −52.83 532.2 (35.42) (860.8) (854.2) (718.4) (69.13) (875.1) Treatm X Had 36.30 1,949 1,744 746.1 71.86 2,012* Formal Sav Acc (36.32) (1,190) (1,177) (1,062) (81.84) (1,203) Observations 2,666 2,916 2,916 2,661 2,918 2,919 R-squared 0.02 0.07 0.07 0.08 0.02 0.07 Note: See table 4.7. TABLE 4.17  Heterogeneity of impacts on financial literacy for baseline per capita total expenditures and whether client had non-FINO formal savings account TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY     (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for baseline per capital total expenditures Treatment  0.053 0.055 0.055 0.035 0.118*** 0.087 (0.13) (0.07) (0.04) (0.04) (0.04) (0.05) Treatm X PC Tot 0.0001 0.0001 0.00002 0.00002 0.0001*** −0.0001 Expend  (0.0001) (0.0001) (0.00002) (0.0001) (0.00003) (0.0001) Observations 2,911 2,911 2,911 2,911 2,911 2,911 R-squared 0.08 0.01 0.04 0.03 0.03 0.03 Panel 2. Heterogeneous impacts for whether client already had a non-FINO formal savings account Treatment  0.103 −0.083 0.065 0.048 0.075 0.143** (0.15) (0.08) (0.04) (0.05) (0.06) (0.062) Treatm X Had −0.067 0.273*** −0.024 −0.031 0.027 −0.131* Formal Sav Acc (0.17) (0.1) (0.05) (0.06) (0.07) (0.07) Observations 2,921 2,921 2,921 2,921 2,921 2,921 R-squared 0.08 0.01 0.04 0.03 0.02 0.03 Note: See table 4.7. 4.  Does financial education affect savings behavior?  ◾ 165 TABLE 4.18  Average impacts on consumption CIGARETTE, FOOD FOOD CONSUMED TOBACCO, BEETLE TOTAL CONSUMPTION OUTSIDE HOME NUT, ALCOHOL CONSUMPTION   (1) (2) (3) (4) Treatment 56.72 −15.70 −19.28** 21.98 (47.35) (15.56) (9.25) (60.93) Observations 2,850 2,829 2,842 2,885 R-squared 0.07 0.03 0.03 0.07 Note: See table 4.7. Results refer to ITT estimates. TABLE 4.19  Average impacts on loans and assets ASSETS BOUGHT ASSETS SOLD LOANS   (1) (2) (3) Treatment 1,688** −383.4 130.2 (675.6) (317.5) −1,143 Observations 2,921 2,921 2,899 R-squared 0.03 0.02 0.04 Note: See table 4.7. Results refer to ITT estimates. 4.5 DISCUSSION OF THE RESULTS AND CONCLUSION Our key finding is that the financial education program increased total savings on average by 29 percent (as compared to the endline savings of the control group). The success of the intervention in increasing savings is in line with the fast-growing literature on savings in developing countries that has established that even the poorer are not simply “too poor to save.” Collins et al. (2009) document that poor families in Bangladesh, India, and South Africa save resources through a wide variety of informal or semi-formal savings devices. Banerjee and Duflo (2007), looking at household survey data from 13 countries, find that extremely poor households do not use all of their income to afford necessities. Similarly, evidence from bank expansion in developing countries suggests that improving access to the formal financial sector increases savings and income (Aportela 1999; Bruhn and Love 2009; Burgess and Pande 2005). Still, we try to explore various reasons that might have contributed to the increment in savings in our sample. First, this success might be due to the fact that the program was delivered in conjunction with a doorstep banking service, and it might just have worked as a marketing campaign for the FINO account. Indeed, the FINO account offered the major advantage of not only being without fees, but also accessible to clients in their village and thus free of traveling costs. As Schaner (2011) underlines, even if one considers the benefits of formal savings versus home savings (Rb > Rh), 166  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES individuals might still be averse to saving formally when savings are small because bank accounts also have a fixed transaction cost and the difference between Rb and Rh might not be attractive enough to outweigh it. However, when the trans- action cost is reduced, individuals who were already using a bank account will make more deposits and withdrawals, while other individuals who were not using bank accounts will start to use them. Eventually, such increase in account use is likely to lead to higher formal (but not necessarily total) savings levels. In fact, in a parallel and preliminary study about the effect of financial literacy training on use of FINO accounts on the same sample, Sarr, Sadhu, and Fiala (2012) find that the numbers of withdrawals and deposits significantly increased after the inter- vention. However, even offering a formal banking service for free might not be enough if the quality of service is not ensured and trust issues are not addressed. For example, Dupas et al. (2012) emphasize that efforts to expand financial access will effectively achieve financial inclusion only by comprising a communication component that brings awareness of the various financial options available. Accordingly, their randomized experiment suggests that entry costs (including opening fees and the administrative hassle of opening an account) explain only about one-fifth of the low banking rates observed in their sample from rural Kenya. Thus, our financial education program might have been successful just because it worked as a marketing campaign for FINO and helped improve famil- iarity with and perception of quality of the bandhu service. During our endline survey, we asked respondents a full set of questions regarding their satisfac- tion with the FINO account, and we use the responses to investigate how much the marketing of FINO services has contributed toward the program’s impact on savings. Table 4.20 shows how the results change for average impacts on savings increments when we control for ex post quality of service.12 The effect of the financial education program on FINO savings is diminished in magnitude, but nonetheless retains its significance; this suggests that the program indeed had a direct impact on savings, apart from the indirect effect it might have had by increasing familiarity with and perception of quality of the FINO service. Most importantly, the significant and positive effect of the financial education program on average total savings for the treatment group (OLS estimates) and especially for those who attended the training (IV estimates) remains unchanged—and even the magnitude of the coefficient is very close to the one in the estimation that does not control for quality. On the other hand, table  4.20 provides evidence suggesting that the quality of service is an important determinant of the amount of FINO savings. Table 4.21 shows that there is indeed a substantial heterogeneity of impact depending on the frequency with which FINO agents visited the village: We measure quality of service by exploiting the responses of clients to the following 12  question: “How would you rate the overall FINO agent/bandhu service? Very bad, not good, satisfactory, good, or very good?” 4.  Does financial education affect savings behavior?  ◾ 167 TABLE 4.20  Average impacts on saving controlling for quality of FINO services NON-FINO NATIONAL- FINO FORMAL FORMAL IZED BANKS INFORMAL TOTAL (1) (2) (3) (4) (5) (6) Panel A. ITT estimates  Treatment 43.53* 1,632** 1,560** 1,443** −14.98 1,609** (23.17) (672.2) (667.9) (601.8) (51.15) (686.2) Quality of FINO services 81.80*** 186.5 85.97 60.33 −7.083 181.6 (11.83) (262.7) (255.0) (226.6) (23.49) (269.6) Panel B. IV estimates Attendance 52.16* 1,968** 1,880** 1,731** −18.07 1,941** (27.64) (808.3) (802.8) (718.4) (61.46) (825.4) Quality of FINO services 80.63*** 141.9 43.58 25.63 −6.675 137.9 (11.55) (270.0) (262.5) (231.7) (23.55) (277.1) Observations 2,619 2,863 2,863 2,614 2,864 2,865 R-squared (OLS estimates) 0.06 0.07 0.07 0.08 0.02 0.07 R-squared (IV estimates) 0.07 0.07 0.06 0.08 0.02 0.07 Note: See table 4.7. those who were assigned to treatment and were visited by a bandhu in the last three months increased their FINO savings by about Rs 200. (For other types of savings, there was no impact heterogeneity depending on the frequency with which FINO agents visited the village.) The second, and we believe more plausible, explanation for the success of our intervention is that the 29 percent increase in total savings is linked to the 4 percent improvement in financial attitudes. This indicator of financial literacy might represent “applied financial knowledge,” as coined by Doi, McKenzie, and Zia (2012); and it might be a proxy for the important ability of making appro- priate financial decisions in everyday life. Also, it might more generally measure TABLE 4.21  Heterogeneity of impacts on FINO saving for FINO agents’ presence   FINO SAVINGS Treatment −0.28 (13.36) Treatm. X FINO agent visited 199.5*** (74.5) FINO agent visited in the last 3 months  203.5*** (42.67) Observations 2,666 R-squared 0.11 Note: See table 4.7. 168  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES the degree of familiarity with and confidence in the financial system since the questions are mostly about suggesting financially appropriate savings devices over more informal solutions. According to this approach, financial attitudes might be a proxy for trust in the formal financial system, and thus might really be crucial for achieving an effective financial inclusion. For instance, Dupas et al. (2012) maintain that lack of trust is the first reason justifying why people, in rural Western Kenya, did not begin saving in their bank accounts even when an account was offered for free. This hypothesis can explain well why not only FINO savings increased, but also savings in other nationalized bank accounts showed a positive and significant increment. Additionally, FINO bandhus accept only deposits in the basic no-frills account and do not offer term deposits, which are savings products with significantly better returns (approximately 3–4 percent higher). These term deposits are usually offered by other public and private sector banks, which might have attracted the savings of the newly financially literate FINO clients. Another possible reason for the success of the intervention is that the program was so focused on responsible financial behavior (including savings and borrowing) that it directly contributed to boosting attention and commitment toward savings, in addition to its effect on financial attitudes. This hypothesis is consistent with a growing body of literature on savings in developing countries that underlines the power of facilitating the mindset of saving money (Dupas and Robinson 2013b). Finally, we have to note that the endline survey was collected only 9–10 months after the financial education training was provided; thus, the savings increase might be temporary. The welfare effects of a temporary change may still be high, but future work is needed to confirm if the training is effective in changing long-term outcomes. REFERENCES Aportela, Fernando. 1999. “Effects of Financial Access on Savings by Low-Income People.” Banco de México. http://www.lacea.org/meeting2000/FernandoAportela.pdf. Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of Economics 121 (2): 635–72. —. 2010. “Female Empowerment: Impact of a Commitment Savings Product in the Philippines.” World Development 38 (3): 333–44. Banerjee, Abhijit V., and Esther Duflo. 2007. “The Economic Lives of the Poor.” Journal of Economic Perspectives 21 (27): 141–67. Banerjee, A., S. Cole, E. Duflo, and L. Linden. 2007. “Remedying Education: Evidence from Two Randomized Experiments in India.” Quarterly Journal of Economics 122 (3): 1235–64. Bruhn, M., L. de Souza Leão, A. Legovini, R. Marchetti, and B. Zia. 2013. “The Impact of High School Financial Education: Experimental Evidence from Brazil.” Policy Research Working Paper 6723, World Bank, Washington, DC. 4.  Does financial education affect savings behavior?  ◾ 169 Bruhn, Miriam, and Inessa Love. 2009. “The Economic Impact of Banking the Unbanked: Evidence from Mexico.” Policy Research Working Paper 4981, World Bank, Washington, DC. Bruhn, Miriam, and David McKenzie. 2009. “In Pursuit of Balance: Randomization in Practice in Development Field Experiments.” American Economic Journal: Applied Economics 1 (4): 200–232. Brune, Lasse, Xavier Giné, Jessica Goldberg, and Dean Yang. 2011. “Commitments to Save: A Field Experiment in Rural Malawi.” Policy Research Working Paper 5748, World Bank, Washington, DC. Burgess, Robin, and Rohini Pande. 2005. “Do Rural Banks Matter? Evidence from the Indian Social Banking Experiment.” American Economic Review 95 (3): 780–95. Carpena, Fenella, Shawn Cole, Jeremy Shapiro, and Bilal Zia. 2011. ”Unpacking the Causal Chain of Financial Literacy.” Policy Research Working Paper 5798, World Bank, Washington, DC. Chaia, Alberto, Aparna Dalal, Tony Goland, Maria Jose Gonzalez, Jonathan Morduch, and Robert Schiff. 2009. “Half the World Is Unbanked.” Financial Access Initiative Framing Note. http://www.microfinancegateway.org/gm/document-1.9.40671/25.pdf. Cole, Shawn A., Thomas Sampson, and Bilal Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Collins, Daryl, Jonathan Morduch, Stuart Rutherford, and Orlanda Ruthven. 2009. Portfolios of the Poor: How the World’s Poor Live on $2 a Day. Princeton: Princeton University Press. Deb, Anamitra, and Mike Kubzansky. 2012. Bridging the Gap: The Business Case for Financial Capability. Citi Foundation. http://www.citifoundation.com/citi/foundation/ pdf/bridging_the_gap.pdf. Demirguc-Kunt, A., and L. Klapper. 2012. “Measuring Financial Inclusion: The Global Findex.” Policy Research Working Paper 6025, World Bank, Washington, DC. Demombynes, G., and A. Thegeya. 2012. “Kenya’s Mobile Revolution and the Promise of Mobile Savings.” Policy Research Working Paper 5988, World Bank, Washington, DC. Doi, Yoko, David McKenzie, and Bilal Zia. 2012. “Who You Train Matters: Identifying Complementary Effects of Financial Education on Migrant Households.” Policy Research Working Paper 6157, World Bank, Washington, DC. Duflo, E., M. Kremer, and J. Robinson. 2011. “Nudging Farmers to Use Fertilizer: Theory and Experimental Evidence from Kenya.” American Economic Review 101: 2350–90. Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Dupas, Pascaline, Sarah Green, Anthony Keats, and Jonathan Robinson. 2012. “Challenges in Banking the Rural Poor: Evidence from Kenya’s Western Province.” http://www. stanford.edu/~pdupas/Challenges_DupasEtAl2011.pdf. Dupas, Pascaline, and Jonathan Robinson. 2013a. “Savings Constraints and Microenterprise Development: Evidence from a Field Experiment in Kenya.” American Economic Journal: Applied Economics 5 (1): 163–92. —. 2013b. “Why Don’t the Poor Save More? Evidence from Health Savings Experiments.” American Economic Review 103 (4): 1138–71. Forbes India. 2012. “FINO: Taking Banks to India’s Poor.” October 26. http://www. moneycontrol.com/smementor/news/finance-capital/fino-taking-banks-to-indias- poor-766580.html. Grifoni, A., and F. Messy. 2012. “Current Status of National Strategies for Financial Education: A Comparative Analysis and Relevant Practices.” OECD Working Paper 16 on Finance, Insurance and Private Pensions, Organization for Economic Co-operation and Development, Paris. 170  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Hilgert, Marianne A., Jeanne M. Hogarth, and Sondra G. Beverly. 2003. “Household Financial Management: The Connection between Knowledge and Behavior.” Federal Reserve Bulletin 89: 309–22. Karlan, Dean, Margaret McConnell, Sendhil Mullainathan, and Jonathan Zinman. 2012. “Getting to the Top of Mind: How Reminders Increase Saving.” https://www.poverty- action.org/sites/default/files/topmind.pdf. Karlan, D., and M. Valdivia. 2011. “Teaching Entrepreneurship: Impact of Business Training on Microfinance Clients and Institutions.” Review of Economics and Statistics 93 (2): 510–27. Kast, Felipe, Stephan Meier, and Dina Pomeranz. 2012. “Under-Savers Anonymous: Evidence on Self-Help Groups and Peer Pressure as a Savings Commitment Device.” Working Paper 060, Harvard Business School, Cambridge, MA. http://www.hbs.edu/faculty/ Publication%20Files/12-060_4073be1c-88ba-4d5e-9fca-d5275baf3355.pdf. Lusardi, Annamaria, and Olivia S. Mitchell. 2006. “Financial Literacy and Planning: Implications for Retirement Wellbeing.” Working Paper No. 1, Pension Research Council. http://www.dartmouth.edu/~alusardi/Papers/FinancialLiteracy.pdf. —. 2007. “Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education.” Business Economics 42 (1): 35–44. McKenzie, David. 2012. “Beyond Baseline and Follow-up: The Case for More T in Experiments.” Journal of Development Economics 99 (2): 210–21. Sarr, Leopald, Santadarshan Sadhu, and Nathan Fiala. 2012. “Bringing the Bank to the Doorstep: Does Financial Education Influence Savings Behavior among the Poor? Evidence from a Randomized Financial Literacy Program in India.” Working Paper 81036, World Bank, Washington, DC. Schaner, Simone G. 2011. “The Cost of Convenience? Transaction Costs, Bargaining Power and Savings Account Use in Kenya.” http://www.dartmouth.edu/~neudc2012/docs/ paper_75.pdf. Thyagarajan, S., and J. Venkatesan. 2008. “Cost-Benefit and Usage Behavior Analysis of No Frills Accounts: A Study Report on Cuddalore District.” College of Agricultural Banking and Institute for Financial Management and Research/Centre for Micro Finance. http:// www.ifmrlead.org/cmf//wp-content/uploads/attachments/csy/1299/28_NoFrills_ Cuddalore.pdf. 4.  Does financial education affect savings behavior?  ◾ 171 ANNEX A: SUPPORTING MATERIALS TABLE 4A.1  Questions on financial knowledge FINANCIAL KNOWLEDGE INDICATOR QUESTIONS Basic economic 1) “Suppose you need to borrow Rs 5,000. Two people offer you a loan. One loan understanding requires you pay back Rs 6,000 in one month. The second loan also requires you pay back in one month Rs 5,000 plus 15% interest. Which loan would you prefer?” 2) “Imagine that you saved Rs 1,000 in a savings account, and were earning an interest rate of 1% per year. If prices were increasing at a rate of 2% per year, after one year, would you be able to buy more than, less than, or exactly the same amount as today with the money in the account?” Financial 1) “Do you think the following statement is true or false? For a farmer, planting one awareness crop is usually safer than planting multiple crops.” 2) “Shanti is preparing a budget for her household. Which of the following needs to be included in the budget? Income only, Expenses only, Both.” 3) “If you have a savings account in a bank and the bank closes down for some reason, will you get your money back?” 4) “Manoj recently borrowed some money from a local moneylender. He needed this money to buy some clothes for his children for Diwali. Do you think that Manoj’s loan is a productive or an unproductive loan?” 1) “Ramesh does plastering on tall buildings. It is a dangerous job, and he is wor- ried that if he is injured his family’s income will become inadequate to meet their needs. If Ramesh comes to you for advice, what would you suggest among: i. Take up some other (different) work; ii. Purchase health/life/accident insurance; and iii. Increase savings?” 2) “Vimla has a very bright child who is currently in secondary school, but will probably do well in university. She is worried how her family will pay for the child’s education. If Vimla comes to you for advice, what would you suggest among: i. Buy child life insurance policy; ii. Borrow money from a moneylender; iii. Open a sav- Financial attitudes ings account in a bank; iv. Save at home; and v. Discontinue education?” 3) “Naresh currently drives a rented auto rickshaw. He wants to purchase his own auto rickshaw but does not have the money and is considering taking out a loan for the same. If Naresh comes to you for advice, would you suggest for him to take out a loan or not?” 4) “Sajid recently got married. He and his wife are considering buying a TV. They do not have enough savings and will need to take out a loan. Sajid has two options: he can take out a loan from the moneylender and a relative and get a bigger amount of loan to buy a big TV, or he can take a loan only from a relative and buy a smaller TV. What would you advise Sajid and his wife?” 172  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 4A.2  Attendance   ATTENDANCE FINO savings −2.01e−06 (1.13e−05) Formal savings 8.93e−06 (7.68e−06) Non-FINO formal savings −6.25e−06 (4.32e−06) Nationalized bank savings −2.20e−06 (2.62e−06) Total savings −2.82e−06 (6.16e−06) Budgeting quality −0.003 (0.01) Interest in financial matters −3.49e−05 (0.011) Basic economic understanding 0.001 (0.012) Financial awareness −0.015 (0.012) Financial attitudes 0.023* (0.013) Client is female 0.038* (0.022) Client age 0.003*** (0.001) Client education (at least secondary) 0.013 (0.031) Discount rate −0.005 (0.007) Client had a non-FINO formal savings account 0.017 (0.022) Baseline per capital total expenditures −4.70e−05 (4.48e−05) Quality of roof −0.009 (0.009) Observations 1,299 R-squared 0.04 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 4.  Does financial education affect savings behavior?  ◾ 173 TABLE 4A.3  Nonresponse rates for the outcome measures showed in table 4.6 PRE-INTERVENTION POST-INTERVENTION   TREATMENT CONTROL TREATMENT CONTROL   (1) (2) (3) (4) Savings FINO savings 0.03 0.03 0.07 0.05 Formal savings 0.00 0.00 0.00 0.00 Non-FINO formal savings 0.00 0.00 0.00 0.00 Nationalized bank savings 0.07 0.05 0.03 0.04 Informal savings 0.00 0.00 0.00 0.00 Total savings 0.00 0.00 0.00 0.00 Savings considering only pure treatment and pure control FINO savings 0.03 0.03 0.07 0.05 Formal savings 0.00 0.00 0.00 0.00 Non-FINO formal savings 0.00 0.00 0.00 0.00 Nationalized bank savings 0.06 0.05 0.03 0.04 Informal savings 0.00 0.00 0.00 0.00 Total savings 0.00 0.00 0.00 0.00 Financial literacy  Budgeting quality 0.00 0.00 0.00 0.00 Interest in financial matters 0.03 0.02 0.00 0.00 Financial numeracy 0.06 0.05 0.00 0.01 Financial awareness 0.02 0.02 0.00 0.00 Financial attitudes 0.01 0.01 0.00 0.00 TABLE 4A.4  Attrition   ATTRITION Treatment −0.007 (0.007) Observations 3,004 R-squared 0.004 Note: Robust standard errors in parentheses. 174  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 4A.1  Kernel densities a. Pure treatment and control: baseline b. Pure treatment and control: endline Density Density 0.15 0.20 Control Control 0.15 0.10 0.10 0.05 0.05 Treatment Treatment 0 0 0 5 10 15 0 5 10 15 Total savings Total savings c. Postharvest intervention d. Postharvest intervention and pure control: baseline and pure control: endline Density Density 0.15 0.20 Control Control 0.15 0.10 0.10 0.05 0.05 Treatment Treatment 0 0 0 5 10 15 0 5 10 15 Total savings Total savings 4.  Does financial education affect savings behavior?  ◾ 175 ANNEX B: HETEROGENEITY OF IMPACTS ON FINANCIAL LITERACY USING THE MEAN VALUE OF THE INDICATORS TABLE 4B.1  Heterogeneity of impacts on financial literacy for client education and baseline financial literacy TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY     (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for client education (at least secondary) Treatment  0.031 0.01 0.029 0.005 0.02* 0.002 (0.04) (0.01) (0.02) (0.01) (0.01) (0.01) Treatm X Education  −0.037 0.012 −0.014 −0.01 −0.015 0.002 (0.07) (0.02) (0.033) (0.02) (0.02) (0.02) Observations 2,907 2,848 2,739 2,866 2,890 2,883 R-squared 0.08 0.01 0.04 0.03 0.03 0.02 Panel 2. Heterogeneous impacts for baseline financial literacy Treatment  0.022 0.014 0.024 0.002 0.017* 0.003 (0.04) (0.01) (0.018) (0.01) (0.01) (0.01) Treatm X Basel −0.002 0.029*** 0.03** −0.013 0.002 −0.003 Financ Literacy (0.03) (0.01) (0.014) (0.01) (0.01) (0.01) Observations 2,907 2,848 2,739 2,866 2,890 2,883 R-squared 0.08 0.01 0.04 0.03 0.02 0.02 Note: See table 4.7. 176  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 4B.2  Heterogeneity of impacts on financial literacy for client gender and time preferences TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS   QUALITY   (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for client gender (female dummy) Treatment  0.025 0.014 0.018 −0.004 0.012 0.002 (0.05) (0.013) (0.021) (0.01) (0.01) (0.01) Treatm X Female  −0.006 −0.002 0.018 0.017 0.011 0.002 (0.05) (0.02) (0.03) (0.02) (0.02) (0.01) Observations 2,907 2,848 2,739 2,866 2,890 2,883 R-squared 0.08 0.01 0.04 0.03 0.03 0.02 Panel 2. Heterogeneous impacts for time preferences (discount rate) Treatment −0.132* 0.022 0.032 −0.022 −0.027 −0.018 (0.07) (0.03) (0.04) (0.02) (0.02) (0.02)   Treatm X Discount 0.054** −0.003 −0.003 0.008 0.015** 0.007 Rate  (0.02) (0.01) (0.01) (0.01) (0.01) (0.01) Observations 2,867 2,812 2,711 2,830 2,854 2,846 R-squared 0.08 0.01 0.04 0.03 0.03 0.02 Note: See table 4.7. 4.  Does financial education affect savings behavior?  ◾ 177 TABLE 4B.3  Heterogeneity of impacts on financial literacy for baseline per capita total expenditures and whether client had a non-FINO formal savings account TARGETED BY THE BASIC ECONOMIC UNDERSTANDING INTERVENTION KNOWLEDGE AWARENESS INTEREST IN BUDGETING FINANCIAL FINANCIAL FINANCIAL FINANCIAL ATTITUDES MATTERS QUALITY   (1) (2) (3) (4) (5) (6) Panel 1. Heterogeneous impacts for baseline per capital total expenditures Treatment  0.018 0.011 0.026 0.001 0.024** 0.003 (0.04) (0.01) (0.02) (0.01) (0.01) (0.007) Treatm X PC Tot 0.00002 0.00001 −0.0 0.00001 −0.00003*** 0.0 Expend  (0.00004) (0.00001) (0.00001) (0.00001) (0.00001) (0.00001) Observations 2,897 2,838 2,732 2,857 2,880 2,874 R-squared 0.08 0.01 0.04 0.03 0.03 0.02 Panel 2. Heterogeneous impacts for whether client already had a non-FINO formal savings account Treatment  0.03 −0.019 0.032 0.008 0.01 0.014 (0.05) (0.02) (0.02) (0.01) (0.01) (0.01) Treatm X Had −0.014 0.057*** −0.013 −0.011 0.01 −0.02* Formal Sav Acc (0.055) (0.02) (0.03) (0.02) (0.02) (0.01) Observations 2,907 2,848 2,739 2,866 2,890 2,883 R-squared 0.08 0.01 0.04 0.03 0.02 0.02 Note: See table 4.7. CHAPTER 5 C omic FX in Kenya Can cartoons improve the effectiveness of financial education? NADA EISSA, JAMES HABYARIMANA, AND WILLIAM JACK 5.1 BACKGROUND 5.1.1 Rationale and key objectives Kenyan youth face an uncertain and volatile financial landscape, with high un- and underemployment, questionable long-term job prospects, and little or no protection against the vagaries of ill health and injury. On the other hand, young Kenyans have more opportunities to invest in and plan for their futures than perhaps any earlier generation: market liberalization and a stable macroeconomy have facilitated steady growth in recent years, educational and training options abound, and access to financial services has expanded considerably (FinAccess 2011). It is essential to give this population the tools and financial capability needed to grasp these opportunities. This chapter is an evaluation of a financial capability project that seeks not only to inform young Kenyans about financial facts and options, but also to induce them to change the way they think about and make financial deci- sions. In a context in which it is often difficult to see beyond the short term, we focus on encouraging long-term planning as it relates to investment in education, training, and small business creation; and precautionary financial behavior such as saving, asset allocation, and insurance against uncertain events. The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 179 180  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES In addition to fostering a culture of financial sophistication and responsibility, a key objective of the project was to compare two alternative mechanisms for delivering the messages and information that we hope will yield these results. In particular, we compare in-person delivery of a structured course of materials with a series of weekly comic book episodes that personalize, contextualize, and make pertinent to the target population the lessons of the course. 5.1.2 Key activities Reaching and communicating with young people is a challenge, particularly when the material could be construed as dry and technical. We worked with two Kenyan organizations that have sought to help young Kenyans prepare themselves for economic opportunities through innovative means. Well Told Story is a for-profit company that publishes a monthly comic book called Shujaaz, read by upwards of 600,000 young people across the country; as well as producing a radio program broadcast on about 20 radio stations nation- wide featuring the lead character of the comic book. Junior Achievement Kenya (JAK), is a local NGO engaged with about 250 high school–based youth clubs and associations through which a range of educational and behavioral interventions are conducted. The key activities that we jointly engaged in include the following: ◾◾ Development of financial education materials. Using JAK’s existing Company Program materials that have been used over the last four years to develop entrepreneurial knowledge and skills among Kenya’s youth, we developed a curriculum that covers a range of financial skills, both at the business and individual/household levels. ◾◾ Integration of financial education lessons into the Shujaaz storyline. Storylines for the comic exhibiting situations that illustrate the lessons central to the financial education materials described above were devel- oped. These storylines involved the regular Shujaaz characters—both heroes and villains—with whom readers are already familiar, and who serve as role models (or anti-role models, accordingly). ◾◾ Production of radio clips. Additionally, special audio clips were produced that included financial education material that closely relates to that in the comics and the JAK lessons. These segments were integrated with the regular radio shows and delivered by CD to study participants to ensure high rates of exposure. ◾◾ Delivery of JAK programs. JAK has delivered the Company Program in school clubs for four years, as well as a less intensive generic program called It’s My Business. The augmented version of the Company Program, including individual and household-level skills and behavior change mate- rials, was provided over a two-month period to selected clubs, with a focus 5.  Comic FX in Kenya  ◾ 181 on students in the last two years of high school, who typically are between 16 and 20 years of age. ◾◾ Distribution of Shujaaz comics. Shujaaz is currently distributed through a network of M-Pesa agents on a monthly basis.1 However, as part of this project, we arranged for independent delivery of the comic and the CD directly to participating schools. The materials were disseminated to selected JAK clubs to be used as part of the club’s activities, again with a focus on students in the last two years of school. ◾◾ Measurement. We collected data to assess the impact of the financial education material and the relative impact of the different delivery mech- anisms. In particular, we undertook baseline and endline surveys across club participants to measure financial capability; we measured financial decision-making skills using responses to exogenously engineered finan- cial innovations such as a lottery. ◾◾ Analysis and dissemination. We conducted statistical analyses of the data we collected and prepared a report of our findings. We will present our findings at conferences and seminars, and through the Internet, espe- cially in collaboration with the World Bank as requested. 5.2 PROGRAM DESCRIPTION 5.2.1 Area of focus FINANCIAL LITERACY TOPIC The materials we presented through various media focus on increasing the finan- cial sophistication and long-term planning capability of young adults. In addition, we demonstrated the use of financial tools as instruments that facilitate the expansion of real economic activity and income generation, as well as risk mitiga- tion and consumption smoothing. DELIVERY MECHANISMS Financial education materiald were delivered in a classroom lesson environment, through a comic book format, and via prerecorded radio shows delivered by CD. JAK’s Company Program focused on business development and entrepre- neurship, which for this project was augmented as mentioned above with addi- tional information and skills for personal and household financial management. M-Pesa is the mobile banking product offered by Safaricom, Kenya’s largest mobile phone 1  operator. 182  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES The material was delivered to club members either by local business executives, university students/graduates, or teachers from the school who had received the relevant training. Well Told Story publishes a monthly comic book that typically includes three or four themes targeted at youth readers. In the past, these themes have included, for example, prudent business practices, the new Kenyan constitution, and tips on animal husbandry. The lead character in the story is an itinerant disc jockey who communicates with youth over the airwaves. Capitalizing on this, Well Told Story expands the reach of the messages in the comic by producing a complementary radio show featuring this character and others from the storylines; listeners can call or text in feedback and questions. These shows are aired by about 20 radio stations across Kenya. Finally, readers and listeners can be linked through social networking sites such as Facebook, which enable further sharing of experiences. The financial education material described above was included as one of the storylines in the Well Told Story publications that are delivered to clubs at selected schools. Each week, a new financial theme was developed and included, while maintaining a fluid and connected storyline. These entries mix both sugges- tive advice through the evolution of the stories (planning ahead turns out to pay dividends, poor business strategies can lead to failure, etc.), as well as specific instructions on how to perform certain tasks (how to open a bank account, start a business, etc.). TARGET GROUPS Our target population included high school students in the last two years of school, who are typically aged between 16 and 20 years old. We implemented the project in JAK clubs, which operate in 217 high schools across Kenya. Students in forms 1–4 (U.S. equivalent grades 9–12) are eligible to join these clubs, which hold weekly instructional meetings that cover a range of topics and issues. Club membership ranges from 20 to 40 students per class, and clubs can be separated into lower (first and second) and upper (third and fourth) classes for instruction. Descriptive statistics regarding the schools—which include boarding and day schools, boys’, girls’, and coeducational school, etc.—are presented below. A map showing the geographic distribution of schools is also presented (figure  5.1). In Kenya, the average adult has no high school education—mean years of schooling for those over age 25 is 7.0—but children today can expect, on average, to reach within a year of graduating from high school; i.e., the expected number of years of schooling for those now under age 7 is 11.0.2 Thus, while our sample of students is definitely not representative of the entire Kenyan population, we believe it is likely to provide good predictive information for the population of young Kenyans Source: United Nations Development Programme, Kenya Human Development Report, 2  http://hdr.undp.org/en/countries/profiles/KEN (accessed April 20, 2014). 5.  Comic FX in Kenya  ◾ 183 FIGURE 5.1  Location of schools with JAK clubs Source: Google Maps. whose financial and entrepreneurial behaviors we hope to influence. In addition, while club participation is undoubtedly nonrandom, the students who participate in this project could be argued to represent a segment of society for which finan- cial education could be especially productive. The schools in our sample were located across the country, from the capital (Nairobi) to the remote eastern part of the country, and from the coastal areas to Lake Victoria, and thus provide a rich and diverse sample to study. Figure 5.1 illustrates the location of the 217 participating schools. Table 5.1 shows the school distribution by region. The Western and Rift Valley regions had the highest concentrations of schools. The schools exhibit heterogeneity in two potentially important dimensions relating to gender and boarding status, as summarized in table 5.2. Of the schools in our sample, 36  percent are girls-only schools, 29  percent are for boys only, and 34 percent are coeducational. Additionally, 33 percent of schools have day 184  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 5.1  School distribution by region REGION SHARE OF SCHOOLS Western 29% Rift Valley 28% Nairobi 20% Coast 13% Eastern 10% TABLE 5.2  School characteristics SHARE OF SHARE OF GENDER SCHOOLS BOARDING STATUS SCHOOLS Girls only 36% Day students only 33% Boys only 30% Boarding only 56% Coeducational 34% Both 11% students only, 56 percent are boarding schools, and 11 percent have a mix of both day and boarding students. 5.2.2 Specific objectives Our goal was to understand which of the methods of delivering financial educa- tion that we assessed was best able to provide young Kenyans with the tools to grasp the opportunities they face, and to change their proclivity to actually reach out and take hold of those opportunities. Thus, part of the role of financial educa- tion might be to empower individuals to feel that they can, through prudent and creative financial decisions, take steps that will lead them to a better life and—for some of them—to a life out of poverty. Our specific objectives in attempting to achieve this ambitious goal were as follows: ◾◾ To develop and administer two alternative mechanisms for developing financial skills and capabilities ◾◾ To measure indicators of financial knowledge, attitudes, and behavior ◾◾ To use these indicators to assess the relative impacts of each mechanism separately and when used in combination ◾◾ To document the challenges and opportunities for improvement associated with the implementation of each mechanism ◾◾ To understand the experience of consumers (i.e., students in this case), and the potential long-term impacts of the interventions 5.  Comic FX in Kenya  ◾ 185 5.3 EVALUATION METHODOLOGY 5.3.1 Approach We planned to combine a randomized control trial with quasi-experimental quan- titative techniques to address the objectives outlined above. The interventions were randomized across Junior Achievement Clubs, as described below. We described above the population of student clubs supported by JAK at high schools across the country. These clubs were randomly assigned to one of four treatment groups—comic, JAK, placebo, and control—as illustrated in table 5.3. The interventions focused on third form students (11th graders in the U.S. system), of whom there were an average of 21 students per club, so each cell includes about 1,140 target students. TABLE 5.3  Experimental design INTERVENTION Comic and CD JAK Placebo Control (1) (2) (3) (4) Total # schools 60 54 52 51 217 As shown in column 1 of the table, a total of 60 clubs received weekly Shujaaz comics plus a CD with a prerecorded radio show with financial educational mate- rial. As shown in column 2, 54 clubs received the JAK augmented Company Program on a weekly basis. As shown in column 3, 52 clubs received weekly Shujaaz comics that did not contain the financial education material. Finally, as noted in column 4, 51 clubs represented a pure control group, with no financial education delivered either through the Company Program or through Shujaaz. In addition, all four treatment groups received Business Daily newspapers that covered macroeconomic news articles. These newspapers were distributed in order to measure the extent of treatment exposure of the students in each of the treatment groups. 5.3.2 Implementation The project was implemented in 217 schools with JAK clubs in five regions in Kenya (Nairobi, Western, Eastern, Rift Valley, and Coast). Each JAK club has an average of about 21 student members each, and meets once every week on a specific day after regular school hours, designated as “club hours.” These groups formed the unit of randomization of the alternative delivery mechanisms we tested. The implementation time frame for this project was as follows: 186  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES ◾◾ February to mid-March 2012 —— Conducted focus groups with students in different schools in Nairobi in order to get an idea of what contents to include in the comic and JAK materials —— Selected the sample of schools to participate in the study: all schools with JAK clubs were asked to take part in the study; out of 236 schools, 217 consented to take part in the research ◾◾ Mid-March to mid-April 2012 —— Conducted baseline survey in all 217 schools of the JAK Company Program —— Once the baseline survey was completed, randomized the schools to receive the comic, JAK, placebo, or control treatment ◾◾ Mid-April to May 2012 —— Treatment material contents were finalized and printed (JAK materials, Shujaaz comics—both comic and placebo and radio program) —— Business Daily newspapers were developed and printed —— CD players were purchased —— All treatment materials were packaged per school according to the treatment they were supposed to receive —— Bank accounts and mutual fund accounts were opened for the students who chose to put their winnings in either or both accounts ◾◾ June 2012 to mid-July 2012 —— Treatment materials were distributed to all 217 schools according to the randomization —— Each school received the first materials with an orientation on how to implement the materials and when to distribute them —— The weekly treatment materials were delivered to each school there- after for a period of six weeks —— Students received official letters with information regarding their accounts, including their account numbers and how to activate and make deposits into their accounts ◾◾ August 2012 —— ATM cards were distributed to students who opened bank accounts via the various post bank branches ◾◾ September to mid-October 2012 —— Preparation for endline survey of both non-JAK and JAK students; this included preparing the survey instrument, consent forms, bank and mutual forms, etc. —— Training staff on the endline survey and protocols 5.  Comic FX in Kenya  ◾ 187 ◾◾ Mid-October to mid-November 2012 —— Conducted endline survey of both non-JAK and JAK students —— Students were given the opportunity to win money again and open bank and mutual fund accounts ◾◾ Mid-November to December 2012 —— Data cleaning and analysis of endline and baseline data —— Facilitated account opening for the students who won, and arranged for fund transfers into their accounts —— Students were provided with instructions on their accounts and how to access the accounts and automated teller machine (ATM) cards via phone ◾◾ January to February 2013 —— Further data cleaning and analysis and report writing 5.3.3 Program intervention The baseline survey was conducted in 217 of the 236 schools targeted for inclu- sion in the project. We were unable to reach the remaining 19 schools due to school exam schedules or, in some cases, refusal of the school to participate in the research. In schools with clubs with more than 30 students, a random sample of 30 students was chosen to be interviewed. In schools with smaller clubs, all club members who were present were interviewed. As part of the interview, each student was asked to write a short (25 words or less) essay. When all students had been interviewed, a sample of respondents 1,500 (about US$18) each on the was selected to “win” a financial award of K Sh  basis of their writing sample. Although winners were chosen at random, only students who had submitted a writing sample were eligible. The purpose of this design was to make the winners feel that they had earned the money in some sense, so as to more closely approximate a real-world financial event. If there were 20 or more respondents, 10 were randomly chosen to receive the award; if there were 13 or fewer respondents, all students won; if there were 14–19 students, half were chosen as winners. The strategy was adopted so as to avoid a situation in which nearly but not all JAK members won the award, which could easily have created tension within the group. The students who won the award were asked to allocate their winnings across three potential buckets: cash, bank account, and mutual fund account in the Nairobi Stock Exchange. The latter two options would be inaccessible to the student for a three-month period, and were held under the custodianship of Digital Divide Data and JAK. Because schools do not allow students to hold cash on school grounds, if they chose to receive cash, it would be held by the school bursar until the school holidays. 188  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Following the complete analysis of the baseline data, schools were randomly assigned to one of four groups—comic, JAK, placebo, and control—through strat- ification with respect to specific school characteristics, including region, school gender (boys, girls, or mixed), and status (boarding or nonboarding). In our analysis of the baseline data, we found a reasonable degree of financial knowledge (if not capability) among JAK students. Since JAK is an entrepreneur- ship club, we believe that students who are already financially knowledgeable and capable may self-select into the JAK clubs. Therefore, we broadened the sample during the endline survey to include non-JAK students (at the same school in form 3). We printed additional comic books for comic and placebo treatment schools to distribute to non-JAK third form students. Although we do not have baseline data on these students, we will compare those in the treatment and control schools. Of particular interest will be the students in schools that receive the comic treatment (and the placebo), compared with students in the control schools. Non-JAK students in schools that receive the JAK treatment will be of less interest (because they are not directly exposed to the treatment). However, we hope to learn about the spillovers, if any, associated with the JAK treatment. Treatment materials were developed and finalized during the months of April and May. The comic treatment materials referred to as the “Janjaruka” series (which translates to “be smart about your money”) included six editions; the placebo comics included six editions (special editions for this project that had no financial education); there were six editions of the JAK treatment material referred to as “Know Your Money.” In addition, all of these treatment materials, including for the control group, included six editions of the Business Daily newspapers, with selected articles on macroeconomic topics. With the delivery of the treatment materials to the schools, we also delivered official letters to students who had chosen to open a bank or mutual fund account. These letters had full information regarding student account numbers, how to access accounts to deposit money, restrictions on withdrawing money from the account for a period of three months, and contact information for further inquiries. Bank accounts were opened with the Kenya Post Office Savings Bank during the baseline because it offered a specific youth product (a Smata account) meant for youth between 12 and 18 years of age. This account allows more flexibility and autonomy to enable youth to manage their accounts themselves than do other banks in Kenya. Also, they are conveniently located throughout the country and were easily accessible to the students in our study. All bank accounts came with ATM cards that would allow students to withdraw and deposit money into their accounts. In accordance with the research structure, students were informed that they would not be allowed to withdraw money from the accounts until August 1, 2012, but were allowed to deposit money into their accounts prior to that date. In addition, on August 1, all students who opened a bank account received a 10 percent interest on their initial deposit into their account. 5.  Comic FX in Kenya  ◾ 189 Mutual fund accounts, referred to as balanced funds, were opened with Zimele Asset Management Company Ltd. The Zimele Balanced Fund is invested in shares of companies on the Nairobi Stock Exchange and interest instruments such as treasury bills, treasury bonds, and commercial paper, among others. It is ideal for medium- to long-term investment. As with the Kenya Post Office Savings Bank, we decided on this specific organization because it offered flexibility and easy accessibility for the students. Following the treatment intervention, the endline survey was conducted from mid-October till mid-November 2012. The endline survey was to cover both Junior Achievement students (referred to hereafter as JAK students) in the 217 schools covered in the baseline survey; as well as non–Junior Achievement students (referred to hereafter as non-JAK students) in 162 schools. Due to last- minute changes in the Kenyan school calendar, we were able to cover 207 schools of the 217 schools for JAK surveys and 38 schools out of the intended 162 schools for non-JAK surveys. We were unable to cover more schools, especially for the non-JAK surveys, due to two major obstacles with the Kenyan school calendar. First, the final semester of the school year started three weeks late due to a teachers’ strike that resulted in all schools being closed, thus delaying the start of the endline survey by a month and half. Second, because of the already shortened semester, the newly adjusted survey start date conflicted with the Kenyan Secondary Certifica- tion Examination schedule, which started October 15, and with form 3 students’ final semester examination. For this reason, we had to cover more schools within a tight schedule and, in some cases, outside of school hours. Table 5.4 shows the breakdown of JAK and non-JAK schools surveyed, orga- nized into five geographically defined groups. The endline survey differed slightly for each treatment group; it contained additional questions regarding the treatment that a particular school received. These questions were developed in order to test the exposure of the student to the actual treatment materials that were delivered to the schools. In addi- tion, we interviewed the baseline survey winners separately, asking them ques- tions regarding how they spent the money they won in the previous term, who TABLE 5.4  Geographic distribution of schools NUMBER OF BASE- NUMBER OF JAK NUMBER OF NON- REGION LINE SCHOOLS SCHOOLS JAK SCHOOLS Western 54 53 37 Rift Valley 63 58 0 Nairobi 50 47 1 Eastern 23 23 0 Coast 27 26 0 Total 217 207 38 190  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES influenced their spending, and how easy it was for them to access their bank or mutual funds account provided they had chosen to set up an account. We also had a short essay competition in this round of the survey where, as in the baseline sample, students were selected to “win” a financial award 1,500 each on the basis of their writing sample. Although the choice of of K Sh  winners was made randomly, only students who had submitted a writing sample were eligible. Table 5.5 reports indicators of the distributions chosen by the 2,067 base- line JAK winners and the 1,890 JAK winners and 368 non-JAK winners from the endline. In the baseline, about one in five winning students chose to take all their winnings in cash, while 10 percent put all their winnings in a bank account. One in three winners split the total between cash and a bank account, and 26 percent chose all three vehicles. Of all the money allocated in prizes, about 88 percent was split almost equally between cash and bank account (46 percent for cash and 41 percent for bank account), and 12 percent was invested in the mutual fund option. In the endline, about one in five winning JAK and non-JAK students chose to take all their winnings in cash, while 10 percent and 3 percent, respectively, of JAK and non-JAK students put all their winnings in a bank account. Almost none of the JAK or non-JAK students chose to put all their money into a mutual fund. One in four JAK winners and one in five non-JAK winners split the total between cash and a bank account; and 20 percent and 25 percent, respectively, of JAK and non-JAK students chose all three vehicles. Of all the money allocated in prizes, about 93 percent was split almost equally between cash and bank accounts for both JAK and non-JAK students (49 percent and 51 percent for cash, and 44 percent TABLE 5.5  JAK and non-JAK competition winners and winnings % OF % OF % OF ENDLINE BASELINE JAK ENDLINE JAK NON-JAK WINNERS WINNERS WINNERS Winners, total 100.0 100.0 100.0 Allocated 100% to cash 22.5 21.1 18.1 Allocated 100% to bank account 10.0 10.3 2.8 Allocated 100% to mutual fund 1.1 0.2 0.0 Divided money between cash and bank account 30.5 44.8 49.2 Divided money between cash and mutual fund 3.6 1.5 1.7 Divided money between bank account & mutual fund 6.2 3.0 2.8 Divided money between all three options 25.9 19.2 25.4 Winnings, total 100.0 100.0 100.0 Money allocated to cash 46.0 49.1 51.0 Money allocated to bank account 41.5 44.1 40.8 Money allocated to mutual fund 12.5 6.7 8.2 5.  Comic FX in Kenya  ◾ 191 and 41  percent for bank accounts for JAK and non-JAK students, respectively), and 6.7 percent of JAK students and 8.0 percent of non-JAK students invested in the mutual fund option. Following the endline survey, bank accounts and mutual fund accounts were created for the students who won this round of the survey; information about account numbers, access to accounts, and restrictions on withdrawal were communicated accordingly. 5.4 RESULTS We present results on four outcomes: one intermediate outcome on financial literacy and three on behavior (both stated and actual) and aspirations. 5.4.1 Impact of the interventions on financial literacy We measure financial literacy using six questions that are standard in the litera- ture (table 5.6). Questions gauge whether students understand the concepts of interest rates, inflation, stock markets, insurance, and banking. Below we present the results of regressions showing the impact of the JAK and JAK + comic inter- ventions on the likelihood of students responding correctly to these questions. The row corresponding to the constant in table 5.6 shows the fraction of students in the control and placebo arms who answered the questions correctly. As the table shows, there is considerable variation in the financial knowledge of the control group, ranging from 36 percent for the stock market returns question to 86 percent for the interest rates question. Students in the JAK only and JAK + comic arm are more likely to answer this question correctly, but the differences we measure are statistically insignificant. For the inflation question, 64 percent of the control students answered correctly. The same fraction of students in the JAK only arm answered this question correctly; the percentage of JAK + comics TABLE 5.6  Impact of the Interventions on financial literacy STOCK BANKING INTEREST MARKET PORTFOLIO SERVICE RATES INFLATION RETURNS CHOICE INSURANCE FEES JAK class 0.0147 0.00452 0.0339 −0.0491* 0.0170 0.00945 (0.0145) (0.0208) (0.0217) (0.0219) (0.0200) (0.0216) JAK + comics 0.0207 −0.0513** −0.0176 −0.0189 0.0239 0.0328 (0.0134) (0.0198) (0.0200) (0.0205) (0.0187) (0.0203) Constant 0.861*** 0.641*** 0.360*** 0.606*** 0.680*** 0.519*** (0.00872) (0.0122) (0.0126) (0.0127) (0.0118) (0.0127) Observations 3,395 3,347 3,164 3,201 3,360 3,353 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. 192  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES students who answered correctly is 5 percentage points lower. The latter differ- ence is statistically distinguishable from no difference with 95 percent confidence. We find no significant difference for the stock market returns, insurance, and banking service fees questions. Across all study arms, 36  percent, 68  percent, and 52 percent of students answered these questions correctly. For the portfolio choice question, 61 percent of the control students answered correctly. A similar fraction in the JAK + comic arm answered correctly, compared to 56 percent of the JAK only students. This latter difference is statistically significant. Overall, aside from the negative impacts on the inflation question for the JAK  + comics and the portfolio question for the JAK only students, we find no effect of the interventions on other measures of financial literacy. 5.4.2 Impact of the interventions on stated and actual savings We turn to an examination of whether the interventions changed the financial behavior of students. All students were eligible to compete in an essay competi- tion, in which about one-third of them would win about US$18. Before the compe- tition, students were asked how much of those winnings they would be willing to allocate to three categories: cash, bank account, or mutual fund account. Winners of the competition were then asked directly how much they actually wanted to allocate to these three categories. Table 5.7 shows regressions measuring the impact of the interventions on both stated savings for all students and actual savings for competition winners. Columns 1 and 2 show the results for stated savings, while columns 3 and 4 show the results for actual savings. We include an interaction term to measure any differences in behavior depending on students’ stated goals. At baseline, students were asked whether they had any goals. A follow-up question asked the student to indicate what the goal is. We then categorized the goals into long-term goals such as starting a business or going to college versus short-term goals such as purchasing electronics. In column 1, we estimate the negative effects of the JAK only and JAK + comic interventions on stated savings; these differences are not statistically significant. In column 2, we add the interaction with the long-term goals indicator. While the interaction with the JAK + comic indicator is positive, suggesting that the inter- ventions only work for those with long-term goals, the coefficient estimates are small and statistically insignificant. Columns 3 and 4 present the results for actual savings. Once again, we do not find any statistically significant differences in the actual savings of the inter- vention arms relative to the control group. The interaction with the long-term goal indicator suggests that students with such goals actually save less than students with short-term goals. However, the point estimates are again imprecise. Overall, we find no effect of the interventions on stated or actual savings. 5.  Comic FX in Kenya  ◾ 193 TABLE 5.7  Impact of the interventions on stated savings for all students and actual savings for competition winners STATED SAVINGS ACTUAL SAVINGS (1) (2) (3) (4) JAK only −20.35 −19.36 31.66 50.51 (28.35) (31.05) (43.18) (46.46) JAK + comic −10.17 −14.93 −62.20 −51.21 (26.81) (29.24) (43.72) (46.91) Long term 6.697 0.463 20.13 62.78 (30.49) (43.00) (49.95) (64.40) JAK comic* long 26.17 −60.33 (72.85) (124.0) JAK only* long −6.474 −119.3 (76.40) (124.9) Control 963.0*** 964.1*** 786.8*** 779.2*** Observations 1,463 1,463 792 792 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 5.4.3 Impact of the interventions on aspirations Finally, we examine whether the interventions change the goals of the students. We asked the same question about goals at the endline survey. Table 5.8 shows the results of this examination. In particular, we estimate the likelihood of purchasing electronics, starting a business, making money, education, and other goals on the indicators for each of the intervention arms. Students in the intervention arms are less likely to indicate purchasing elec- tronics as a goal. While about 37 percent of control students indicate this as a TABLE 5.8  Impact of the interventions on changing student goals PURCHASING STARTING A MAKING ELECTRONICS BUSINESS MONEY EDUCATION OTHER GOALS (1) (2) (3) (4) (5) Comic −0.0538* 0.0608** −0.0128 −0.000407 0.0476 (0.0243) (0.0232) (0.01000) (0.0121) (0.0249) JAK −0.0333 −0.0205 0.00997 −0.00752 −0.0181 (0.0256) (0.0234) (0.0117) (0.0124) (0.0254) Placebo −0.0553* 0.0392 −0.00316 −0.0108 0.0253 (0.0255) (0.0243) (0.0110) (0.0123) (0.0262) Control 0.374*** 0.269*** 0.0456*** 0.0603*** 0.375*** (0.0186) (0.0170) (0.00800) (0.00913) (0.0186) Observations 2,992 2,992 2,992 2,992 2,992 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. 194  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES goal, only about 32 percent of the students in the comics arm report this as a goal at endline. The comics arm is also more likely to report wanting to start a business than the control group. Just over a quarter of control students report wanting to start a business at endline, compared to more than 32 percent of the students in the comics arm. This difference is statistically significant. However, we find no effects at all for the JAK only intervention. 5.5 POLICY IMPLICATIONS We find little evidence that the interventions implemented here improved the financial literacy of the students. It is important to point out that the absence of an effect is not because of very high levels of financial knowledge. Across the six questions, an average of only 60 percent of students answered the questions correctly. Similarly, we find no effect of the interventions on stated and actual savings behavior. However, we do find impacts of the comics intervention on the likeli- hood that students want to start a business in the future. In addition, students in this arm are less likely to state a goal of purchasing electronics such as a mobile phone/tablet. One potential explanation for these null results is that take-up of the inter- ventions may be low. In results not reported, we find some heterogeneous effects with large positive effects among students with greater exposure to the treat- ments. Future research should seek to improve exposure to the treatments as well as the content of the treatments. REFERENCE FinAccess. 2011. Financial Inclusion in Kenya: Survey Results and Analysis from FinAccess 2009. http://www.fsdkenya.org/finaccess/documents/11-06-27_finaccess_09_results_ analysis.pdf. CHAPTER 6 S ocial networks, financial literacy, and index insurance Evidence from a randomized experiment in Kenya XAVIER GINÉ, DEAN KARLAN, AND MUTHONI NGATIA ABSTRACT We present a randomized field experiment measuring the direct impact and social network spillovers of providing financial literacy and discount vouchers on farmers’ decision to purchase index-based drought insurance. To examine this, we form clusters by grouping together geographically prox- imate households. Clusters were then randomly assigned to receive either a high or low intensity of each of the following treatments: financial literacy materials and discount vouchers for the price of insurance. We find social network spillovers to the provision of financial literacy materials, but no spill- overs to the provision of discount vouchers on farmers’ decision to purchase insurance. Specifically, receiving financial literacy materials when 60 percent or more of their neighbors also receive financial literacy materials increases the likelihood that a farmer will purchase insurance by 4.3  percent (stan- dard error 2.5  percent); while receiving financial literacy materials when 40 percent or less of their neighbors also receive financial literacy materials decreases the likelihood that a farmer will purchase insurance by 2.6 percent We acknowledge financial support from the Russia Financial Literacy and Education Trust Fund and the Agricultural Risk Management Team at the World Bank and Finan- cial Sector Deepening Trust—Kenya. Special thanks to Equity Bank, APA Insurance, ˜ ˜ ˜ Farmer Cooperative Society, Ntima Farmer Cooperative Society and FSD Kenya Ruırı for administrative support. Thanks also to Amrik Heyer, Erin Bryla, Michael Mbaka and Andrea Stoppa for helpful comments and to Pratibha Joshi for valuable research assistance. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions. All errors remain our own.   ◾ 195 196  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES (standard error 2.5 percent). Looking at the discount vouchers, we find significant own effects of receiving discounts on the price of insurance on farmers’ take-up decision, but negligible social network effects. Our results provide suggestive evidence that financial literacy materials are efficacious in encouraging take-up when farmers’ social contacts similarly receive access to such materials. 6.1 INTRODUCTION Agriculture in developing countries is characterized by a high dependence on weather and limited irrigation. Weather shocks are a major source of income fluc- tuation for rural households involved in agricultural activities. Research suggests that households are not fully insured against income shocks. Weather index–based insurance products provide a means by which poor households can hedge against these risks. A significant advantage of the design of index insurance products is that, since payouts are based on measured rainfall, they can be calculated and disbursed quickly and automatically without the need for households to formally file a claim. This in turn reduces transaction costs, which would otherwise tend to drive up the price of the insurance. Fast payouts are also likely to be valued by policy holders in an environment where households are poor and often liquidity constrained. A second advantage is that the product is free of the adverse selection and moral hazard problems that often plague insurance markets. This is because payouts are based only on publicly observed data, rather than private information reported by the person filing the claim. Despite these benefits however, demand for index insurance has been low (Cole et al. 2010; Giné, Townsend, and Vickery 2008; Giné and Yang 2009). One of the reasons for the weak demand may be lack of understanding of the product. If potential buyers misunderstand or underestimate the probability and amount of payouts, there will be little demand and little or no impact on farmer behavior. Conversely, if farmers overestimate the frequency and amount of payouts, they are likely to be disappointed and fail to purchase insurance again in the future. Financial education can help mitigate these problems. Evidence from Malawi (Giné and Yang 2009) and India (Giné, Townsend, and Vickery 2008) suggests that a lack of comprehension of index insurance indeed dampened demand for it. Index insurance, with its complicated triggers and payout schedules, is a particularly difficult product to explain—even more so than standard insurance. There are very few studies that rigorously examine the effects of different means of providing information on insurance products on take-up of insurance (a notable exception is McPeak, Chantarat, and Mude 2010). Social networks could also play a role in individual farmer demand for index insurance. Studies have found that social networks play critical roles in determining a variety of individual and aggregate economic outcomes, ranging from transmitting 6.  Social networks, financial literacy, and index insurance  ◾ 197 information about job networks (Beaman and Magruder 2009; Granovetter 1973, 1995), the basis for trade in noncentralized markets, and the provision of mutual insurance in developing countries (Fafchamps and Lund 2003), to mention but a few. An increasingly popular technique to identify social network effects has been to use randomized experiments (e.g., Angelucci and De Giorgi 2009; Ange- lucci et al. 2010; Duflo and Saez 2003; Godlonton and Thornton 2012; and Miguel and Kremer 2004), wherein a random subset of members of a social network are provided with some intervention to change their behavior. The random variation in the behavior of some members of the social network thus induced is used to instrument the average behavior in the group. Under conditions presented in Imbens and Angrist (1994), an instrumental variables estimator identifies a local average treatment effect that measures the impact of the intervention beyond the targeted group—the peer effect. This chapter presents a randomized field experiment that measures the direct impact and social network spillovers of providing financial literacy and discount vouchers on farmers’ decision to purchase drought insurance. We find suggestive evidence that social networks play an important role in their decision to purchase drought insurance. Receiving financial literacy materials when 60  percent or more of a farmer’s neighbors also receive financial literacy materials increases the likelihood that a farmer will purchase insurance by 4.3 percent (standard error 2.5 percent). In contrast, receiving financial literacy materials when 40 percent or less of a farmer’s neighbors receive them decreases the likelihood that a farmer will purchase insurance by 2.6  percent (standard error 2.5  percent). Discount vouchers, on the hand, have strong individual effects on purchase. Reducing the purchase price of insurance by 10 percent increases take-up by 1.3 percentage points (standard error 0.39 percentage points). There are, however, no spill- overs from the provision of discount vouchers on farmers’ take-up of insurance. Our results imply that financial literacy materials are efficacious in encouraging take-up when farmers’ social contacts similarly receive access to financial literacy materials. The remainder of the chapter is organizes as follows: section 6.2 describes the setting and the experiment, section 6.3 discusses the results, and section 6.4 concludes. 6.2 EXPERIMENTAL DESIGN 6.2.1 Setting ˜ ˜ ˜ and Ntima, two drought- The study was carried out with coffee farmers in Ruırı prone areas in eastern Kenya. The enumeration area covered 14 villages: 12 ˜ ˜ ˜ and 2 villages in Ntima. Interviewers first visited all coffee-growing villages in Ruırı 198  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES households within the enumeration area and collected some basic information about the household: their global positioning system (GPS) coordinates, farm size, number of coffee trees, and contact information. Following the census, households were grouped together into clusters based on geographic proximity. Each household had a block with a 60-meter radius drawn around its GPS marker. All overlapping blocks became part of the same cluster. Households that were isolated were added to the nearest cluster. Enumerators then revisited the households and administered a baseline survey. Table 6.1 presents some summary statistics of our study population. Seventy percent of our study households are headed by men. The average age of the household head is 49, with 7 years of education. We define “somewhat patient” as equal to 1 if the respondent prefers to receive KSh 1,250 (US$15.63) in a month to KSh 1,000 (US$12.50) today. Almost half (47.6 percent) of the respon- dents report understanding what insurance is. Figure 6.1 gives the chronology of the study. We test two interventions: a comic on index insurance and discount vouchers for the purchase of insurance. Randomization of treatment intensity was done at the cluster level and was orthogonal across treatments. After completing the baseline survey, enumerators administered the interventions. Each household was randomly assigned to receive a comic or not, depending on what comic intensity the household’s cluster was assigned. Each household was also given the opportunity to take part in a drawing for a voucher on the price of insurance. Households in high voucher intensity clusters had a 3 in 5 chance of drawing a 50 percent voucher, a 1 in 5 chance of a 25 percent voucher, and a 1 in 5 chance for a 0 percent voucher. Households in low voucher intensity clusters had a 1 in 5 chance of drawing a 50 percent voucher, a 1 in 5 chance of a 25 percent voucher, and a 1 in 5 chance of a 0 percent voucher. Enumerators also completed a base- line survey with survey households. Figure 6.2 illustrates the experiment design. FIGURE 6.1  Study chronology in 2011 End of insurance contract Farmer listing and randomization End of insurance sales Nov 2010 Jan 2011 Feb 2011 Apr 2011 Jun 2011 Jul 2011 Sep 2011 Nov 2011 Dec 2011 Baseline and distribution of comic and vouchers Payment of Followup insurance claims survey 199  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 6.1  Summary statistics by intervention COMIC NO COMIC VOUCHER COMIC INTENSITY CLUSTER P- VALUE MEAN HIGH LOW HIGH LOW HIGH LOW N Male 0.169 0.701 0.027 −0.070 −0.036 0.062 0.104 0.087 904 (0.039) (0.057) (0.059) (0.039) (0.075) (0.086) Age 0.174 48.679 1.886 −2.892 −3.459 0.41 4.47 6.585* 904 (1.714) (2.647) (3.035) (1.792) (3.530) (3.953) Household size 0.033 5.346 0.072 −0.531* −0.332 −0.032 −0.820** 0.17 850 (0.213) (0.292) (0.303) (0.214) (0.405) (0.493) Years of education 0.156 6.682 −0.499 −1.310** −0.751 −0.698* 0.516 1.486 904 (0.396) (0.574) (0.620) (0.394) (0.748) (0.951) Able to read a newspaper 0.296 0.689 −0.043 −0.097* −0.095 −0.019 0.048 0.143 904 (0.041) (0.059) (0.062) (0.041) (0.080) (0.092) Asked for credit in preceding 0.484 0.233 −0.050 −0.082 −0.039 −0.056 0.084 0.106 849 year (0.039) (0.051) (0.056) (0.039) (0.071) (0.087) Agrees with statement: 0.062 0.678 −0.089** −0.041 −0.106* −0.053 0.133* 0.209** 904 Drought is the most important (0.041) (0.057) (0.062) (0.042) (0.080) (0.093) risk my household faces Amount spent on food in last 0.879 1,208.990 5.722 −96.776 −60.187 −61.584 −182.666 −195.997 850 7 days (113.536) (112.324) (139.351) (96.987) (178.270) (209.480) Patient 0.574 0.380 0.004 0.07 0.103 0.024 −0.018 0.089 904 (0.043) (0.061) (0.064) (0.044) (0.083) (0.100) Understand insurance 0.104 0.470 0.006 −0.033 0.142** −0.054 0.035 0.132 849 (0.045) (0.065) (0.065) (0.047) (0.087) (0.103) Total no. of coffee trees 0.683 0.029 −0.062 0.032 −0.006 0.065 0.175 904 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 200  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 6.2  Randomization 904 households 189 clusters No comic Low comic High comic Comic intensity 276 households 307 households 321 households 61 clusters 65 clusters 63 clusters High voucher Low voucher High voucher Low voucher High voucher Low voucher intensity intensity intensity intensity intensity intensity Voucher intensity 138 HHs 138 HHs 163 HHs 144 HHs 164 HHs 157 HHs 31 clusters 30 clusters 33 clusters 32 clusters 31 clusters 32 clusters Comic control households 138 138 115 104 40 38 Comic treatment households 0 0 48 40 124 119 50% disc. voucher treatment 76 45 91 44 105 53 25% disc. voucher treatment 31 40 40 39 34 30 No voucher—marketprice 31 53 32 61 25 74 No comic in high-intensity comic 0 0 0 40 0 Comic in low-intensity comic 0 0 48 40 0 119 No comic in low-intensity comic 0 0 115 104 0 38 6.3 RESULTS This section presents estimates of the causal effect of receiving a comic and a discount voucher on farmers’ decision to purchase insurance. Further, since treatment intensity was assigned randomly at the cluster level, we can estimate the social network effects of being in a high-intensity comic cluster or a high-in- tensity voucher cluster. We estimate: BuyInsij = β1Comicij + β2Comicij × ComicIntensityLevelj + β3 ComicIntensityLevelj + γ1Voucherij + γ2Voucherij × VoucherIntensityLevelj + γ3VoucherIntensityLevelj + ϵij where BuyInsij is an indicator for whether household i in cluster j purchased insur- ance. Comicij is an indicator variable for whether household i in cluster j was 6.  Social networks, financial literacy, and index insurance  ◾ 201 assigned to receive a comic, ComicIntensityLevelj indicates whether cluster j was assigned to be a high comic intensity cluster or a low comic intensity cluster. Voucherij is the discount voucher drawn, VoucherIntensityLevelj indicates whether cluster j was assigned to be a high voucher intensity cluster or a low voucher intensity cluster. Table 6.2 presents the results. Column 1 presents a naïve estimator that ignores spillovers by grouping together treated households in high intensity TABLE 6.2  Impact of comics and vouchers on whether respondents purchased insurance DEPENDENT VARIABLE: BOUGHT INDEX INSURANCE (1) (2) (3) (4) Comic 0.022 (0.019) Voucher 0.131*** (0.039) Low comic intensity −0.014 (0.020) High comic intensity 0.037* (0.023) High voucher intensity 0.029* (0.017) Comic in high comic 0.043* 0.047* intensity (0.025) (0.025) Comic in low comic −0.026 −0.022 intensity (0.025) (0.026) No comic in high comic 0.026 0.032 intensity (0.037) (0.037) No comic in low comic −0.002 −0.002 intensity (0.023) (0.023) Voucher in high voucher 0.149*** 0.148*** intensity (0.046) (0.047) Voucher in low voucher 0.093* 0.092* intensity (0.049) (0.050) Observations 904 904 904 904 R-squared 0.041 0.048 0.071 Mean among controls 0.020 0.036 0 0 Village fixed effects Yes Yes Yes Yes Controls No No No Yes Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Regression (4) includes controls for farmer’s raven score, years of education, gender, ability to read a newspaper, ability to write a letter, score for coffee best practices followed, whether the farmer agrees that drought is the most important risk that he or she faces, a score for whether the farmer is patient, number of coffee trees, whether the farmer understands index insur- ance, whether the farmer can identify 60 millimeters in a drawing. 202  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES clusters with treated households in low-intensity clusters. We find a modestly positive effect of comics on the likelihood of purchasing insurance and a strongly significant positive effect of voucher provision on the likelihood of purchasing insurance. A 10 percent discount in the price of insurance increases the likelihood of purchase by 1.3 percentage points (standard error 0.39 percentage points). In column 2, we compare treated clusters with control clusters; thus, treated and control households in treated clusters are grouped together. We find that households in low comic intensity clusters are less likely to purchase insurance, though this effect is imprecisely estimated. Households in high comic intensity clusters are more likely to purchase insurance by 3.7 percentage points (standard error 2.3 percentage points), and households in high voucher intensity clusters are 0.29 percentage points more likely to purchase insurance for every 10 percent discount on the price of insurance (standard error 0.17 percentage points). Finally, in columns 3 and 4 we interact individual treatment with the intensity assigned to the cluster. Column 3 shows that receiving a comic in a high-inten- sity comic cluster increases the likelihood that a farmer will purchase insurance by 4.3 percentage points (standard error 2.5 percentage points) while receiving a comic in a low-intensity comic cluster decreases the likelihood that a farmer will purchase insurance by 2.6 percentage points (standard error 2.5 percentage points). Drawing a 50 percent discount voucher in a high voucher intensity cluster increases the likelihood a farmer will buy insurance by 14.9 percentage points (standard error 4.6 percentage points), while drawing a 50 percent voucher in a low voucher intensity cluster increases the likelihood that a farmer will purchase insurance by 9.3 percentage points (standard error 4.9 percentage points); these effects are not significantly different from each other. In column 4, we include some controls for the farmer’s raven score, years of education, gender, ability to read a newspaper, ability to write a letter, score for coffee best practices followed, whether the farmer agrees that drought is the most important risk that he or she faces, a score for whether the farmer is patient, number of coffee trees, whether the farmer understands index insurance, and whether the farmer can identify 60 millimeters in a drawing. These controls do not significantly alter the magnitude of our estimates. Our estimates in columns 3 and 4 provide suggestive evidence that finan- cial literacy materials are only efficacious in encouraging take-up when farmers’ social contacts similarly receive access to financial literacy materials. Our results also show that the naïve estimate in column 1 understates the impact of comic provision by conflating the impact of comics in low- versus high-intensity comic clusters, which work in opposite directions. We find similar patterns of impact if we look at spillover effects beyond the geographical cluster (table 6.3). To measure how treatment density affects farmers’ decision to purchase insurance, we rely on exogenous variation in the local density of treated farmers by virtue of the cluster-level randomization. Specif- ically, we compute the number of treated farmers residing in close proximity. We 6.  Social networks, financial literacy, and index insurance  ◾ 203 TABLE 6.3  Ordinary least squares regressions of whether respondents purchased insurance DEPENDENT VARIABLE: BOUGHT INDEX INSURANCE 0–150 M AWAY 0–200 M AWAY 0–300 M AWAY (1) (2) (3) (4) (5) (6) Comic −0.036 −0.031 −0.027 −0.020 −0.032 −0.029 (0.025) (0.025) (0.028) (0.028) (0.032) (0.032) # Comics 0-X m away −0.020* −0.022** −0.010 −0.010 −0.008 −0.009 (0.010) (0.011) (0.009) (0.009) (0.006) (0.006) Comic* # Comics X m 0.047*** 0.045*** 0.025** 0.023* 0.014* 0.013* away (0.017) (0.017) (0.012) (0.012) (0.007) (0.007) Voucher 0.178*** 0.169*** 0.120** 0.110* 0.103 0.096 (0.052) (0.054) (0.058) (0.059) (0.071) (0.072) # 50% vouchers 0-X m 0.007 0.007 0.000 0.000 −0.004 −0.005 away (0.012) (0.012) (0.009) (0.009) (0.007) (0.007) Voucher* # 50% vouchers −0.028 −0.025 0.003 0.006 0.005 0.006 Xm away (0.020) (0.021) (0.016) (0.017) (0.010) (0.011) # Contacts X m away 0.008 0.008 0.004 0.005 0.005 0.005 (0.008) (0.008) (0.006) (0.006) (0.004) (0.004) Observations 904 904 904 904 904 904 R-squared 0.054 0.077 0.05 0.073 0.049 0.073 Mean among controls 0.02 Village fixed effects Yes Yes Yes Yes Yes Yes Controls No Yes No Yes No Yes Note: Robust standard errors, clustered at the village level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Regressions in columns 2, 4, and 6 include controls for farmer’s raven score, years of education, gender, ability to read a newspaper, ability to write a letter, score for coffee best practices followed, whether the farmer agrees that drought is the most important risk that he or she faces, a score for whether the farmer is patient, number of coffee trees, whether the farmer understands index insurance, and whether the farmer can identify 60 millimeters in a drawing. get significant spillovers from farmers living 150 meters away; these diminish as the band is expanded to 200 meters, then 300 meters. If a farmer was assigned a comic, each additional farmer who receives a comic within a 150-meter radius increases the likelihood that a farmer will purchase insurance by 4.7 percentage points (standard error 1.7 percentage points), off a base of 2 percentage points among controls. In contrast, if a farmer was not assigned a comic, each additional farmer who receives a comic within a 150-meter radius decreases the likelihood that a farmer will purchase insurance by 2.0 percentage points (standard error 1.0 percentage points). These estimates are robust to adding farmer-level controls in columns 2, 4, and 6. 204  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 6.4 FOLLOW-UP SURVEY Eight months after the baseline was completed and a few months after payouts were made, we conducted a phone follow-up with respondents. On average, we were able to reach about 56 percent of our sample. This moderate tracking rate was due to the fact that not all respondents had mobile phones and not all respondents were available during the three-week-long follow-up period. Table 6.4 demonstrates that receiving a voucher is mildly correlated with being more likely to be found during the follow-up. TABLE 6.4  Correlates with attrition in follow-up data DEPENDENT VARIABLE: INTERVIEWED IN FOLLOW-UP SURVEY (1) (2) (3) Comic 0.031 (0.035) Voucher 0.168** (0.078) High comic cluster −0.020 (0.042) Low comic cluster 0.054 (0.042) High voucher cluster 0.008 (0.034) Comic in high comic intensity 0.050 (0.045) Comic in low comic intensity −0.018 (0.062) No comic in high comic intensity 0.059 (0.066) No comic in low comic intensity −0.019 (0.046) Voucher in high voucher intensity 0.160* (0.085) Voucher in low voucher intensity 0.181* (0.102) Observations 904 904 904 R-squared 0.021 0.019 0.024 Mean among controls 0.506 0.493 0.417 F-test 1.314 1.088 1.18 p -value of F model 0.186 0.361 0.267 Village fixed effects Yes Yes Yes Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 6.  Social networks, financial literacy, and index insurance  ◾ 205 Table 6.5 examines the impact of the provision of the interventions on farmers’ understanding of index insurance. At the baseline and at the follow-up, we asked farmers the following three questions: Imagine you have bought insur- ance against drought. If it rains less than 70 millimeters by the end of November, you will receive a payout of K Sh 20 for every millimeter of deficient rainfall (that is, each millimeter of rainfall below 70 millimeters). Will you be paid out if (a) it rains 100 millimeters? (b) It rains 60 millimeters (if b = Yes). How much would you receive as a payout? The knowledge score is the sum of these three questions. On average, respondents were able to answer two out of the three questions. Column 1 shows that farmers who received a comic in a low comic intensity TABLE 6.5  Knowledge score on index insurance KNOWLEDGE SCORE ON INDEX INSURANCE AT FOLLOW-UP (1) (2) Comic in high comic intensity −0.021 −0.009 (0.044) (0.045) Comic in low comic intensity −0.137** −0.127* (0.068) (0.069) No comic in high comic intensity 0.041 0.067 (0.063) (0.063) No comic in low comic intensity −0.010 0.002 (0.046) (0.046) Voucher in high voucher intensity 0.041 0.038 (0.086) (0.086) Voucher in low voucher intensity 0.099 0.113 (0.101) (0.100) Baseline knowledge score of index insurance 0.002 (0.032) (0.051) (0.058) Observations 479 479 R-squared 0.078 0.124 Mean of dependent variable among controls 0.686 Village fixed effects Yes Yes Controls No Yes Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. All regressions include village fixed effects. Regressions in column 2 include controls for farmer’s raven score, years of education, gender, ability to read a newspaper, ability to write a letter, score for coffee best practices followed, whether the farmer agrees that drought is the most important risk that he or she faces, a score for whether the farmer is patient, number of coffee trees, and whether the farmer can identify 60 millimeters in a drawing. The knowledge score is the sum of the number of correct answers to the following three questions: Imagine you have bought insurance against drought. If it rains less than 70 millimeters by the end of November, you will receive a payout of K Sh 20 for every millimeter of deficient rainfall (that is, each millimeter of rainfall below 70 millimeters). Will you be paid out if (a) it rains 100 millimeters? (b) It rains 60 millimeters (If b = Yes). How much would you receive as a payout? 206  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES cluster score lower on average. This could be because they were not able to rein- force their knowledge about index insurance, since not many of their neighbors received comics. Table 6.6 presents regressions examining the impact of the interventions on farmers’ attitudes toward insurance. We asked the same set of questions at the baseline and at the follow-up. Column 1 shows that farmers in high comic intensity clusters are more likely to state that “Insurance protects you in times of emergency” by 5.7 percentage points if they received a comic and 7.5 percentage points if they did not. These effects are not, however, significantly different from each other. Farmers in both high- and low-intensity comic clusters are also less likely to state that “Insurance companies try to cheat people even when they have a good claim,” in comparison to farmers in clusters where no one received a comic. Finally, only farmers who received discount vouchers are less likely to TABLE 6.6  Outcomes from follow-up “INSURANCE “I DON’T NEED INSURANCE “INSURANCE COMPANIES TRY TO BECAUSE MY FAMILY, PROTECTS YOU CHEAT PEOPLE EVEN FRIENDS, OR RELATIVES IN TIMES OF WHEN THEY HAVE A PROVIDE COVER WHEN EMERGENCY” GOOD CLAIM” NECESSARY” DEPENDENT VARIABLE (1) (2) (3) Comic in high comic 0.057* −0.223*** −0.092 intensity (0.029) (0.063) (0.061) Comic in low comic −0.009 −0.135 0.012 intensity (0.052) (0.088) (0.088) No comic in high comic 0.075** 0.319*** 0.008 intensity (0.034) (0.086) (0.089) No comic in low comic 0.029 −0.216*** −0.098 intensity (0.033) (0.063) (0.062) Voucher in high voucher −0.022 0.076 −0.309*** intensity (0.059) (0.113) (0.115) Voucher in low voucher 0.038 0.028 −0.126 intensity (0.055) (0.142) (0.140) Baseline attitude −0.023 0.014 0.026 (0.023) (0.056) (0.080) Observations 456 453 458 R-squared 0.06 0.072 0.06 Mean of dependent 0.939 0.441 0.471 variable among controls Village fixed effects Yes Yes Yes Note: Robust standard errors, clustered at the village level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 6.  Social networks, financial literacy, and index insurance  ◾ 207 state that “I don’t need insurance because my family, friends, or relatives provide cover when necessary.” 6.5 CONCLUSION This chapter presents preliminary evidence from a randomized field experiment measuring the direct impact and social network spillovers of providing financial literacy and discount vouchers on farmers’ decision to purchase drought insur- ance. We find social network spillovers to the provision of financial literacy mate- rials but no spillovers to the provision of discount vouchers on farmers’ decision to purchase insurance. We further find that financial materials have spillover effects on farmers’ attitudes toward insurance but limited effects on understanding, as narrowly measured in survey. Our results provide suggestive evidence that financial literacy materials are efficacious in encouraging take-up when farmers’ social contacts similarly receive access to financial literacy materials. REFERENCES Angelucci, Manuela, and Giacomo De Giorgi. 2009. “Indirect Effects of an Aid Program: How Do Cash Transfers Affect Ineligibles’ Consumption?” American Economic Review 99 (1): 486–508. Angelucci, Manuela, Giacomo De Giorgi, Marcos A. Rangel, and Imran Rasul. 2010. “Family Networks and School Enrolment: Evidence from a Randomized Social Experiment.” Journal of Public Economics 94 (3–4): 197–221. Beaman, Lori, and Jeremy Magruder. 2009. “Who Gets the Job Referral? Evidence from a Social Networks Experiment.” http://are.berkeley.edu/~jmagruder/kolkatanetworks. pdf. Cole, Shawn, Xavier Giné, Petia Topalova, Jeremy Tobacman, Robert M. Townsend, and James Vickery. 2010. “Barriers to Household Risk Management: Evidence from India.” Policy Research Working Paper 5504, World Bank, Washington, DC. Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Fafchamps, Marcel, and Susan Lund. 2003. “Risk-Sharing Networks in Rural Philippines.” Journal of Development Economics 71 (2): 261–87. Giné, Xavier, Robert M. Townsend, and James Vickery. 2008. “Patterns of Rainfall Insurance Participation in Rural India.” World Bank Economic Review 22 (3): 539–66. Giné, Xavier, and Dean Yang. 2009. “Insurance, Credit, and Technology Adoption: Field Experimental Evidence from Malawi.” Journal of Development Economics 89 (1): 1–11. Godlonton, Susan, and Rebecca Thornton. 2012. “Peer Effects in Learning HIV Results.” Journal of Development Economics 97 (1): 118–29. Granovetter, Mark. 1973. “The Strength of Weak Ties.” American Journal of Sociology 78 (6): 1360–80. 208  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES —. 1995. Getting a Job: A Study of Contacts and Careers. Chicago: University of Chicago Press. Imbens, Guido W., and Joshua D. Angrist. 1994. “Identification and Estimation of Local Average Treatment Effects.” Econometrica 62 (2): 467–75. McPeak, John, Sommarat Chantarat, and Andrew G. Mude. 2010. “Explaining Index-Based Livestock Insurance to Pastoralists.” Agricultural Finance Review 70 (3): 333–52. Miguel, Edward, and Michael Kremer. 2004. “Worms: Identifying Impacts on Education and Health in the Presence of Treatment Externalities.” Econometrica 72 (1): 159–217. CHAPTER 7 W hy is voluntary financial education so unpopular? Experimental evidence from Mexico MIRIAM BRUHN, GABRIEL LARA IBARRA, AND DAVID MCKENZIE ABSTRACT Take-up of voluntary financial education programs is typically extremely low. We conduct randomized experiments around a large financial literacy course offered in Mexico City to understand the reasons for low take-up and to measure the impact of financial education. We document that the general public displays very little interest in such courses and that participation is low even among individuals who express interest in financial education. We experimentally investigate barriers to take-up, and find no impact of relaxing reputational or logistical constraints and no evidence that time inconsistency is the reason for limited participation. Even relatively sizable monetary incen- tives get less than 40 percent of interested individuals invited to training to actually attend. Using this randomized encouragement design, we measure the impact of the course on financial knowledge and behavior. Attending training results in a 9 percentage point increase in financial knowledge and a 9 percentage point increase in savings outcomes, but no impact on borrowing behavior. Administrative data indicate that the savings impact is relatively short lived. The results suggest people are making optimal choices not to attend financial education courses, and point to the limits of using gener- al-purpose courses to improve financial behavior for the general population. We thank the Russia Financial Literacy and Education Trust Fund for financing this work; Pablo Antón Díaz, Eder González Ramos, and the staff at Innovations for Poverty Action for their assistance in implementing the surveys and impact evaluation; and the staff at our partnering financial institution for their support of this project. All opinions expressed in this work are those of the authors alone and do not necessarily represent those of the World Bank, the Russia Financial Literacy and Education Trust Fund, Inno- vations for Poverty Action, or the partnering financial institution.   ◾ 209 210  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 7.1 INTRODUCTION Access to finance is strongly associated with poverty reduction worldwide (Demirgüç-Kunt, Beck, and Honohan 2008). For Mexico, evidence suggests that many poor and lower-middle-class individuals are credit constrained and that relaxing these constraints has high returns. For example, McKenzie and Wood- ruff (2008) show that small injections of working capital funding have monthly returns of 15 percent or more for microenterprise owners. Bruhn and Love (2009) find that a rapid expansion of bank credit serving low-income individuals led to increases in income and employment. Lack of access to finance has also meant Mexican households have been unable to smooth consumption in the face of income shocks (Attanasio and Székely 2004; McKenzie 2006). These factors illus- trate the great potential for increases in access to savings and credit to benefit low-income individuals. However, as access to financial services expands around the world, there is also a growing concern that many consumers may not have sufficient informa- tion and financial acumen to use these new financial products responsibly. Such concerns are particularly pressing in middle-income countries such as Mexico, which have seen rapid expansions of access to finance in recent years, bringing many low-income individuals into the formal financial system for the first time. The number of credit cards has doubled during the past five years; and in urban areas, an increasing number of credit cardholders are now poorer and less-edu- cated individuals. According to the 2008 Mexican Income and Expenditure Survey (ENIGH), 42 percent of households making a credit card payment had a head with less than high school education, and one in six had only primary education or less. Lower-educated households are more likely to make unsophisticated decisions (Lusardi and Mitchell 2011). These unsophisticated decisions can lead to lower efficiency and competition in the credit card market, and as this market grows, they could also have implications for the stability of the financial system as a whole. The recent global financial crisis has emphasized the importance of overin- debtedness on systemic risk. Low levels of financial literacy were arguably one important factor leading many homeowners in the United States to take out mort- gages exceeding their means.1 During the subprime mortgage crisis, individuals with lower financial literacy were more likely to be delinquent or to default on their mortgage (Gerardi, Goette, and Meier 2010).2 In Mexico, nearly 10 percent of all credit cardholders fell delinquent on their payments in February 2009, double 1  See Willis (2011) for a dissenting opinion. In contrast, individuals with higher financial literacy are more likely to strategically default 2  on underwater mortgages after the crisis (Burke and Mihaly 2012). 7.  Why is voluntary financial education so unpopular?  ◾ 211 the U.S. rate.3 Many of these are low-income individuals who are first-time card- holders. In addition, the financial crisis highlighted the need for building up suffi- cient savings to smooth consumption in the face of adverse shocks. However, savings rates tend to be low in many countries, including Mexico (4.2 percent).4 In response to these concerns, many governments, employers, nonprofit organizations, and even commercial banks have started to provide financial literacy courses with the aim of improving financial education. However, partic- ipation rates for noncompulsory financial education programs are typically extremely low. Willis (2011) sums this up as follows: “Voluntary financial educa- tion is widely available today, yet seldom used.” For example, Brown and Gartner (2007) examine pilot experimental efforts by three credit card providers in the United States to provide online financial literacy training to delinquent and at-risk credit cardholders, as well as to college students with newly received credit cards. Target Financial Services made calls to 80,982 at-risk cardholders, reaching only 6,417 of them, of whom half were invited to use a credit education website: only 684 of these requested a code to log-on, of which only 28 used the code, and only 2 people completed the course. U.S. Bank had only 384 cardholders out of the 42,000 it attempted to reach complete its online program (0.9 percent). Wells Fargo offered college students a 60-minute phone card as an incentive to do the training; it had the highest response rate, with 6.7 percent of college student cardholders offered the treatment logging onto the website, and 6.5 percent completing the training. Thus, despite financial education programs becoming increasingly popular among policy makers and financial providers, they appear to be deeply unpop- ular among customers. This raises two interrelated questions that are important for research and policy. The first is whether there are economic or behavioral constraints that prevent more individuals from participating in such programs. The second question is whether there are any benefits to these marginal individ- uals from doing so, or whether they are rationally choosing not to participate in such training. We investigate these questions through the context of randomized experi- ments conducted in Mexico City. We collaborated with a financial institution to evaluate a free financial literacy course offered on a large scale. We document that there is relatively little interest in such a program among the general population, and then screen recruited subjects on interest in participating. We then randomly 3  Source: “Reigning in the Credit Card Industry,” PRI’s The World, May 11, 2009. 4  This number is based on data from the 2008 Mexican Income and Expenditure Survey. It corresponds to the average household savings rate (total income − total expenditures/total income) using sampling weights and dropping those with savings rates below −100 percent and above 100 percent. The average household savings rate using monetary income and expenditures is 11.4 percent. 212  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES selected half of the interested individuals to be invited to the course. The initial participation rate in the course was low (18 percent), even among the sample of interested individuals. Motivated by different theoretical and logistic reasons why individuals may not attend training, we randomized the treatment group into different subgroups, which received treatments designed to provide evidence on some key barriers to take-up. These treatments included monetary payments for attendance, deferred payments, free-cost transportation to the training location, and a video CD with positive testimonials about the training. We find that the latter two incentives did not increase take-up, but both current and deferred mone- tary payments boosted attendance rates by about 10 percentage points (from 18 to between 27 and 33 percent, depending on the payment size). There was no significant difference in attendance between those who received the payment immediately and those who received a deferred payment, suggesting that time inconsistency or high discount rates are not the main constraint to participation. A follow-up survey conducted six months after the course is then used to measure the impacts of the training on financial knowledge, behaviors, and outcomes. We find attending training results in a 9 percentage point increase in financial knowledge and a 9 percentage point increase in savings outcomes; but there is no impact on credit card behavior, retirement savings, or borrowing. Moreover, administrative data suggest that the savings impact is relatively short lived. Data on credit card balances and repayment rates show no systematic differences across the treatment and control groups related to the course. In addition to limited overall impacts of training, we find relatively little hetero- geneity in impacts. We do not find significantly different effects by gender or whether the individual is a customer of our partner financial institution, but we do find stronger savings impacts for individuals with a bachelor’s degree than those without. Our results therefore suggest that the answer to the question of why volun- tary financial literacy programs are so unpopular is that they do not offer that much in the way of benefits to those who are not currently attending them. That is, it appears individuals may be making rational decisions not to attend such financial literacy training programs. We cannot rule out that such programs offer benefits to the individuals who voluntarily choose to go without being given any additional information or incentives, but our results suggest there are limited gains to trying to encourage more people to attend. There is a growing literature attempting to assess the causal impacts of finan- cial literacy training, which we review in the next section. Much of the existing literature has looked at training provided in a compulsory manner, such as to students (Bernheim, Garrett, and Maki 2001; Cole and Shastry 2009; Lührmann, Serra-Garcia, and Winter 2012; also see Bruhn et al. in chapter 2 of this volume), or otherwise has focused on specialized content given to specific populations such as migrants and their families (Doi, McKenzie, and Zia 2012; Seshan and Yang 2012; also see Gibson, McKenzie, and Zia in chapter 1 of this volume), and farmers 7.  Why is voluntary financial education so unpopular?  ◾ 213 (Cai 2012). We contribute to this literature by focusing on a large-scale voluntary financial literacy program offered to the general population in a large urban city— precisely the type of effort that is being increasingly adopted in a number of coun- tries around the world. As well as evaluating the impact of such a program, we are able to experimentally test the importance of different constraints to participa- tion in such programs, which the existing literature has not done. The remainder of this chapter is structured as follows. Section 7.2 provides background information on financial literacy in Mexico and reviews the existing literature. Section 7.3 describes the financial literacy course. Section 7.4 lays out the experimental design and initial take-up. Section 7.5 then describes our exper- iments to overcome barriers to take-up. Section 7.6 presents estimates of the impact of the course on financial knowledge, behavior, and outcomes. Section 7.7 concludes with policy recommendations. 7.2 BACKGROUND AND LITERATURE REVIEW 7.2.1 Financial literacy in Mexico During the last two decades, Mexicans have experienced significant changes in access to finance, banking products, and exposure to risk. In 1997, the Mexican Pension System switched from a pay-as-you-go system to one based on indi- vidual retirement accounts. Under the new system, all private sector workers have to choose a pension fund administrator to manage their account and invest in portfolios with different levels of risk. A similar pension system was launched for public sector workers in 2007. Banking services and financial product avail- ability have also increased in recent years. While in 2004 around 25 percent of the population had a bank account (Demirgüç-Kunt, Beck, and Honohan 2008), this number had more than doubled by 2009 (Saldívar 2009), reaching 58 percent. The number of credit cards has increased from just below 15 million to almost 28 million during the past five years, so that there are now as many credit cards as households in Mexico.5 As access to financial services expands and many consumers become users of formal financial products for the first time, there is a growing concern that many individuals may not have sufficient information and financial acumen to use these products responsibly. The National Commission for the Protection of Users of Financial Services (CONDUSEF [Comisión Nacional para la Protección y Defensa de los Usuarios de Servicios Financieros]) found that 62 percent of Mexi- cans lack financial education and are not aware of their rights and responsibilities The National Institute of Statistics and Geography (INEGI) reports that there were 28 million 5  households in Mexico in 2010. 214  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES with respect to financial institutions (Fernández 2010). In the 2012 Visa Interna- tional Financial Literacy Barometer, Mexico ranks in the lowest third of a sample of 28 countries on questions related to having a household budget or savings set aside for an emergency.6 Coupled with the evidence of low financial literacy, there is also evidence suggesting that individuals may not be managing their financial products correctly. For example, a recent study by Ponce, Seira, and Zamarripa (2009) analyzing credit card use and payment behavior in Mexico suggests that consumers make financially unsophisticated decisions in the credit card market. Also, Mexican workers tend to choose their pension fund administrator based on features not related to fees or returns (Hastings and Tejeda-Ashton 2008; Lara- Ibarra 2011). The Mexican government has addressed the concern regarding low levels of financial education in several ways. CONDUSEF has held a Financial Education Week annually since 2008. The Ministry of Finance has also launched campaigns to improve financial literacy and financial education awareness. Furthermore, one of the objectives of the 2008–2012 National Development Financing Program was “to develop financial culture and consumer protection, promoting that people realize that they can save, get financing for their productive products, be responsible with their loans and protect against risks” (Saavedra 2012). The government’s efforts have also been matched by the development bank BANSEFI and by the private sector—e.g., commercial bank—initiatives. In 2009, a total of 59 programs throughout the country were promoting financial education among different audiences (Treviño Garza 2011). 7.2.2 Literature review Financial education programs act under the assumption that lack of knowledge may be preventing individuals from making sound financial decisions. This is in part due to growing evidence showing strong associations between finan- cial literacy measures and financial decisions. For example, Lusardi and Tufano (2009) find that individuals who have low measured levels of financial literacy tend to pay minimum balances on credit cards, incur late fees on cards, and use informal sources of credit. Stango and Zinman (2009) show that people who make mistakes in interest and future value calculations tend to borrow more and save less. Lusardi and Mitchell (2009) illustrate that people with low levels of finan- cial literacy think less about retirement, and most of them have not planned for retirement at all. And in Mexico, Hastings and Tejeda-Ashton (2008) conducted a Overall, the barometer ranked Mexico in the top three (Visa 2012). However, this result 6  may be biased by “taste”-related questions such as: “At what age should the government require financial literacy?” “How well do you think teens understand finance?” and “How often do you speak to your kids about money?” 7.  Why is voluntary financial education so unpopular?  ◾ 215 survey that reveals that less literate individuals tend to choose mutual pension funds with higher fees. There are three main problems with moving from these studies to developing policy recommendations. The first is that the results are not necessarily causal. They show a correlation between proxies for financial literacy and outcomes of interest, but these correlations may simply reflect unobserved characteristics of individuals such as their numeracy, ability, self-discipline, parental background, or other such features. Second, it is often conceptually difficult to understand whether individuals are truly undersaving and overborrowing, so even if we see increases in savings it is unclear whether this represents a welfare improvement. Third, even if we knew that individuals were overborrowing and undersaving, and that financial literacy had a causal impact on these outcomes, it is still an open question as to whether and how financial literacy skills can be taught and improved. A growing literature tries to address the first and third of these issues by relying on quasi-experimental or experimental variation in the provision of finan- cial education programs. The studies vary widely in terms of context. In addition, they face the challenge that the concept of financial literacy is often measured and defined differently (see Lewis and Messy 2012; Xu and Zia 2012). Compulsory financial education classes taught in high schools have been subjected to a number of studies. Bernheim, Garrett, and Maki (2001) use exoge- nous variation in high school financial education mandates across U.S. states to show that students exposed to financial education classes save more as adults. However, Cole and Shastry (2009) cast doubt on these findings, showing that they are not robust in controlling for state-fixed effects and examining effects over time. Shorter-term evidence comes from Bruhn et al. (2013), who conducted a randomized experiment providing financial education in Brazilian high schools. They find positive effects on financial knowledge, attitudes, and behaviors; and an increase in savings rates—although the impacts are small in absolute magnitude: a 3 percentage point increase in knowledge, and a 1 percentage point increase in savings. In Germany, Lührmann, Serra-Garcia, and Winter (2012) find teenagers given financial literacy training show increased interest in and knowledge of finan- cial matters, and save more in a hypothetical task, but they do not measure actual savings. Other studies have focused on providing financial education to working adults, recognizing the differences in households’ financial needs and exposure across developed and developing countries. The literature in developed coun- tries tends to study the impact of financial education on planning for retirement or investment portfolio choices. Participation in voluntary seminars for retire- ment savings also tends to be low. Duflo and Saez (2003) show that offering monetary rewards can increase attendance substantially, but that this large increase leads to less than a 1.5 percentage point difference in retirement plan participation. 216  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES In the developing country context, impact evaluations of financial literacy training have studied the unbanked, insurance take-up, and migrants. One of the first papers to examine the impact of financial education in a developing country is Cole, Sampson, and Zia (2009). The authors implemented a field experiment in Indonesia where they offered randomly selected unbanked households a financial literacy course geared toward opening a bank savings account. They find that the financial literacy course had no effect on the likelihood of opening a bank savings account in the full sample, but it had modest effects for uneducated and finan- cially illiterate households. Cai (2012) used a randomized experiment to show that farmers in rural China are more likely to take up crop insurance and become less price sensitive after attending financial education sessions. Gibson, McKenzie, and Zia in chapter 1 of this volume; Doi, McKenzie, and Zia (2012); and Seshan and Yang (2012) analyze how providing information and financial education affects the behavior of migrants and their households. Gibson, McKenzie, and Zia work with migrants in New Zealand and Australia, and find that financial education increases knowledge about remittance transaction costs, but does not lead to changes in the amount of remittances sent or use of the cheapest remittance method. Using a sample of Indonesian migrants, Doi, McKenzie, and Zia (2012) find that impacts on financial knowledge, behavior, and savings are largest when both the migrants and their families receive financial education. The results show that financial education can have large effects when provided at a teachable moment, but that this impact varies with who receives training. GAO (2004) has also recognized the importance of providing education during “teachable moments.” Seshan and Yang (2012) find that Indian migrants in Qatar increase savings after financial literacy training, but only if they had low financial literacy to begin with. A common assumption in most financial literacy interventions is that people are not saving enough. Indeed George Akerlof goes so far as to state in his Nobel address that “it is common wisdom that people save too little” (Akerlof 2001). He notes that in standard economic models, saving is the outcome of utility maxi- mization decisions, so saving too little or too much is not possible. However, behavioral economics offers several reasons for undersaving: present consump- tion is more salient than future consumption, so individuals may procrastinate about saving and have time-inconsistent preferences. He notes that the best evidence of undersaving is the observation that upon retirement, consumption falls substantially. However, the view that people do not save enough for retire- ment has been challenged by Aguiar and Hurst (2005), who show that although expenditure falls dramatically upon retirement, consumption does not, as people increase home production. Nevertheless, evidence from lab experiments (e.g., Brown, Chua, and Camerer 2009) supports two views of undersaving: that people do not know how to save optimally in complex environments; and that even when they do know how to save, they cannot resist short-term temptations to spend. Financial education in principle can reduce undersaving by helping with these two 7.  Why is voluntary financial education so unpopular?  ◾ 217 problems: by giving individuals more information and knowledge about saving strategies, and by providing tools and ideas to help resist temptations to spend. In a recent review, Gale and Levine (2010) find that none of the traditional approaches have generated unambiguous evidence of positive impacts of finan- cial literacy efforts.7 The literature is now moving toward exploring whether inno- vative channels for providing financial education may affect behavior. Ongoing studies in India, Peru, South Africa, and the United States (among others) are testing whether the provision of information via videos, radio, mass media, or video games is effective in improving individuals’ financial decisions. This study builds on and complements this existing literature in several ways. First, we work with a population that is already banked. Thus, we do not study whether financial education makes individuals more likely to use formal finan- cial products, such as savings accounts or credit cards, but rather on whether they use these products to their advantage and responsibly. Second, the financial education course we analyze covers topics that are relevant for a range of users of financial services and is delivered at scale by a financial institution, rather than being a pilot or government program. Third, we examine interest for such training in a general population, and then focus on individuals who state an interest in attending a financial literacy course. We show that even among this population training attendance rates are low, and we examine the barriers to participation experimentally through the use of different incentive treatments. Finally, we obtain administrative data on savings and credit card usage from a financial insti- tution to validate the conclusions drawn from survey data. 7.3 THE FINANCIAL LITERACY TRAINING PRODUCT The financial literacy training course studied in this chapter is a large-scale program offered to the general public. The goal of the program is to convey basic knowledge and tools that individuals may need to manage their personal finances responsibly. It is targeted at adults and is offered free of charge. The program was first launched a few years ago and has trained over 300,000 individuals. In 2011, the course won a regional award for innovativeness in fostering financial education. During the time of our study, the course was being taught at several loca- tions in Mexico City and in a number of other cities in Mexico, as well as through an online platform. Our study focuses on the training locations in Mexico City. Courses are offered on a continuous basis, with one or two sessions per day Monday through Saturday. Each session has capacity for 20 participants, although The authors mention four approaches: employer-provided financial education, state-man- 7  dated financial and consumer education for students, credit and mortgage counseling, and community-based programs providing general financial education. 218  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES typical attendance at the most central location is only four or five people per day on weekdays and more on weekends. Attendance is even lower at the other loca- tions, with participation rates of about one person per day during our pretreat- ment monitoring. Training is administered via individual computer terminals, with an instructor present to show videos and facilitate interactive exercises that are used to strengthen the concepts taught in the material. Participants also receive workbooks that contain the information being presented, as well as exercises to be completed during the course. At the end of the session, participants take a short test and receive a certificate conditional on completing the test success- fully. They also receive a CD to take home. This CD includes the tools used in the exercises performed during the course. The course lasts half a day and consists of modules on saving, retirement, credit cards, and responsible use of credit. The course explains why saving is important and discusses different savings instruments and steps individuals can take to increase the amount they save, such as setting savings goals and keeping a household budget. Saving for retirement and pension funds are also covered. The course then discusses the use of credit cards, associated fees, and how to decipher a credit card statement. Finally, information is provided on good credit card debt management practices, an individual’s credit score and credit history, and steps individuals can take to preserve and improve their credit management. 7.4 EXPERIMENTAL DESIGN, AND THE LOW DEMAND FOR FINANCIAL LITERACY TRAINING To estimate the impact of the training on financial knowledge, behavior, and outcomes, we conducted a randomized experiment. This section describes the design and implementation of the initial experiment, including sample selection, randomization, and take-up. Many programs in finance and private sector development face low take-up (McKenzie 2010), and prior voluntary financial literacy programs have also expe- rienced low take-up issues (Brown and Gartner 2007). While the 300,000 indi- viduals taking part in the financial literacy training program studied here are a substantial number, this is countrywide and over four years. Assuming at most 100,000 were trained in Mexico City in any one year, this represents less than 0.6 percent of Mexico City’s adult population obtaining the training annually.8 Low take-up has severe consequences for statistical power. We therefore tried several approaches to screen individuals for their interest in participating in financial literacy training and thereby measure the impacts of such training on these indi- viduals. This process gives insights into the demand for financial literacy training, 8  The National Population Council (CONAPO) projects that in 2012 there were 17,372,952 individuals 17 years or older living in the metropolitan zone of Mexico City. 7.  Why is voluntary financial education so unpopular?  ◾ 219 and provides training impacts for a policy-relevant group. Since it is difficult to make adult financial education mandatory, a key policy question is whether policy efforts should try and encourage more people who are interested in attending such programs to actually attend them. 7.4.1 Approach 1: low demand for financial literacy demonstrated through a mailing campaign Our first approach to obtain a sample of people interested in financial literacy training was to send a screener survey to clients of our partner financial institu- tion. The institution agreed to partner with us and to provide us with a de-iden- tified list of all its clients in Mexico City who have a savings account and a credit card. By conditioning on these characteristics, we intended to identify individuals for whom debt management and saving advice is likely to be relevant and who have credit card and savings behavior that we could study. We further narrowed down the list of clients to those who lived in a municipality with a financial literacy training location less than 5 kilometers away. From this sample, we randomly selected 40,000 clients to receive a mailing with the screener survey. The randomization was stratified by gender and age. The mailings were sent by our partner institution through its usual provider between January 7 and 12, 2011. The delivery company confirmed delivery of 98.8 percent of the letters. Nondeliveries were due to clients having moved or the address not being found. The mailing contained a letter informing clients of our partnering with their financial institution to help research ways people are managing their savings and credit card debt, and to see if they are interested in participating in a financial education training session. The letter only mentioned the training in general terms and did not refer to the specific program we are studying or the locations where this training was being offered. The mailing also contained a two-page screener survey that clients could mail back to us in a prepaid envelope to indicate their interest in the training. This short screener survey collected information on name, address, phone number, gender, age, education level, occupation, household income and expenditure (in bins), as well as basic usage of savings accounts and credit cards. Clients also had the option of responding to the survey by going to a website or by calling a toll-free number. In order to increase response to the screener survey, half of the letters (20,000) were randomly selected to include an offer for a monetary payment of Mex$75 (about US$5) to the first 200 clients who submitted their answers. The total number of letters sent (40,000) was chosen based on information from our partner financial institution that typically only 2–3 percent of its clients reply to any sort of mail offer it sent to them. This expected response rate would yield 800–1,200 clients who would form the sample for our randomized experiment. However, we received much fewer responses than anticipated—only 42; thus, only 0.1 percent of clients expressed interest in a financial literacy training program. 220  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES We suspect that this low response rate is in part due to the climate of inse- curity that has prevailed in Mexico City during the past few years and that has made people distrustful of unsolicited requests for data. Around the time of our study, fradulent phone and mail extortions had become a threat to the general population. In fact, both the federal government and commercial banks launched a campaign asking people to avoid giving personal information to strangers, espe- cially through phone or mail communications offering money or prizes in return for personal data. Our letters were sent out by our partner financial institution, bearing its logo, but people may still have been suspicious of the request for data.9 7.4.2 Approach 2: low demand for financial literacy demonstrated through an online campaign Our second strategy for obtaining a study sample was to conduct a screener survey through Facebook. We created a Facebook page for financial literacy and launched a Facebook ad that pointed to this page. The Facebook page included the same information as the letters sent in the mail, mentioning the importance of financial literacy. The page invited people to indicate their interest in partici- pating in a (generic) financial literacy course by clicking on a link that redirected them to a page where they could answer our screener survey online. This survey contained the same questions as the one mailed to our partner financial insti- tution’s clients. We did not offer a financial incentive for completing the online survey. The Facebook ad was targeted to individuals residing in Mexico City and ran for two months, from mid-February to mid-April 2011.10 It was displayed about 16 million times. We obtained a total of 1,240 fans of our Facebook page and 119 responses to the online survey. Since this sample was still not large enough for our study, we implemented a third approach to screening people for an interest in a financial literacy course, as described below. 7.4.3 Approach 3: street and branch surveys We conducted screener surveys on the streets of Mexico City and outside branches of our partner institution. Surveyors were placed in busy locations within the city during a period of eight weeks (from April 25 to June 25, 2011) where they tried to interview people passing by. We also placed surveyors outside branches of our 9  All letters included a toll-free number people could call to make i\enquiries. We did not receive any phone calls about the training in response to this mailer, suggesting a low demand for such training. At the time, Facebook had 7,743,220 registered users who reside in Mexico City (approx- 10  imately 87 percent of the population). 7.  Why is voluntary financial education so unpopular?  ◾ 221 partner institution between July 6 and 19, 2011, where they approached exiting customers and people waiting in line. For this approach, interviewers asked people if they would be interested in participating in a financial literacy course, providing the same information as stated in letters sent during the mailing campaign. As in the letters, the name of the course was not disclosed. If the respondents expressed interest in the course,11 interviewers asked them to fill out the screener survey, so that we could contact them later with further information about it. The questionnaire included the same set of questions as the mail and online surveys. We did not offer a monetary incentive for completing the surveys, but people who answered the survey were offered cookies and a pen as a small thank you token. We obtained a total of 6,945 completed questionnaires from the street survey and 2,294 from the branch survey. 7.4.4 Treatment randomization and balance For all individuals who had expressed interest in a financial literacy course either through the mail, online, or via the street or branch screener survey, we conducted phone audits to verify the contact information they had provided. This eliminated about half of the respondents. We further dropped respondents who lived outside the Mexico City metropolitan area or who had participated in a financial literacy program in the past. We also dropped observations that had missing answers to the questions we stratified on in the randomization, as described below. Our final sample includes 8 respondents from the mail survey, 5 from the online survey, 2,490 from the street survey, and 1,000 from the branch survey, for a total sample of 3,503 people. We divided this sample into a control group of 1,752 individuals, and a treatment group of 1,751 individuals, using stratified randomization. The randomization was conducted by the authors by computer. We stratified the randomization by whether we obtained the respondent through the branch versus the mail, online, or street survey; by gender; by having at least a bachelor’s degree or not; and by whether the person was (1) a client of our partner financial institution, (2) a client of another financial institution, or (3) neither. Within the sample of financial institution clients, we further stratified by whether they made a deposit into their savings account during the past month and by whether they have a credit card. For clients with a credit card, we strati- fied by whether they made more than the minimum payment each month during the past six months. For individuals who were not financial institution clients, we stratified by whether they lived closer than 8 kilometers away from a training location or not. Subject recruiters reported that, roughly, only two out of every five people approached 11  expressed an interest in the course and agreed to fill out the survey. 222  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Our original intention with the mail screener survey was to study only indi- viduals who are financial institution clients to learn whether the financial literacy course can improve their financial behavior and outcomes. Through the online, street, and branch surveys, we obtained 1,325 respondents who were not finan- cial institution clients. We decided to keep these in the sample, since the course material could in principle also be relevant for them, and they may start financial institution relationships as a result of taking the course. However, the percentage of nonfinancial institution clients in the treatment group who ended up attending the financial literacy training was very low (18.1 percent), making it difficult to detect any effects on this sample. Since the take-up rates were higher among financial institution clients (28.1 percent), as discussed further below, we decided to conduct our follow-up survey only among financial institution clients; and we dropped the nonfinancial institution clients from the impact analysis, although they are included in our experiments on inducing attendance. Table 7.1 shows baseline variables collected through the screener survey for the sample of financial institution clients. About half of the individuals were clients of our partner financial institution as opposed to clients of another insti- tution. Close to 65 percent made a savings deposit during the past month. About 40 percent had a credit card at baseline, and only half of them made more than the minimum payment in all previous months. In fact, about 20 percent made a late payment on their credit card during the past six months. The demographic variables show that about half of the sample is female, and 40 percent has at least a bachelor’s degree. The average age is 33. As expected given the random assignment, all baseline variables are balanced across the treatment and control groups. 7.4.5 Initial take-up: low attendance for training even among those who say they are interested in attending a financial literacy course Starting on August 1, 2011, each person in the treatment group was contacted by telephone and invited to participate in the financial literacy training program. If the participants confirmed interest, they were offered the opportunity to choose a training location, date, and schedule that best suited them. The phone operator then enrolled the participant based on this information. The operator confirmed the details of the appointment with the participant before ending the call. A second call was made the day prior to the participant’s appointment as a reminder to increase the probability of attendance, and a third and final call was made the day after the appointment to confirm attendance and inquire about the level of satisfaction with the courses. During this third call, if participants responded that they had missed the appointment, they were offered the oppor- tunity to reschedule for a future date, if they said they were still interested in attending. 7.  Why is voluntary financial education so unpopular?  ◾ 223 TABLE 7.1  Confirming randomization using baseline data SAMPLE FULL SAMPLE IN INTERVIEWED AT ADMINISTRATIVE BASELINE FOLLOW-UP DATA SAMPLE TREAT- TREAT- TREAT- MENT MENT MENT CONTROL DIFFER- CONTROL DIFFER- CONTROL DIFFER- MEAN ENCE MEAN ENCE MEAN ENCE Stratification variables Baseline survey conducted in 0.35 −0.0058 0.37 −0.0045 0.62 −0.0005 branch Client of partner financial insti- 0.48 −0.0010 0.48 0.0015     tution (vs. other institution) Made savings deposit during 0.64 0.0012 0.64 0.0011 0.70 0.0534 past month Has credit card 0.41 −0.0039 0.42 0.0161 0.56 −0.0005 Paid more than credit card 0.51 0.0172 0.52 0.0169 0.56 0.0085 minimum in all past 6 monthsa Has bachelor’s degree or 0.40 0.0016 0.41 0.0195 0.45 −0.0386 higher Female 0.47 0.0064 0.50 0.0115 0.48 0.0336 Other baseline variables Age 32.69 0.6308 32.97 0.7372 36.14 0.9789 Occupation is employee 0.51 −0.0171 0.49 0.0031 0.50 −0.0282 Paid credit card late in past 6 0.23 0.0124 0.22 0.0214 0.21 0.0087 monthsa Monthly household income is 0.64 −0.0072 0.63 −0.0111 0.67 0.0367 above Mex$6,500 Monthly household expendi- 0.54 −0.0081 0.54 −0.0141 0.54 −0.0394 ture is above Mex$6,500 Sample size 1,090 1,088 814 772 243 227 Note: Differences for baseline variables not used in the stratification control for randomization strata. Administrative data sample includes clients of our partner financial institution that were found in its database. a. Conditional on having a credit card. During this phase, we signed up 1,049 out of 1,751 individuals (59.9 percent) for the course. About a third of the people who signed up for the course actu- ally attended (312 people). Participants gave a range of reasons for not attending the session they had signed up for, including difficulties attending due to work and family commitments, sickness, and—in some cases—issues with instructors turning up late or their arriving late and being turned away. The overall attendance rate for the 1,751 treatment group individuals who had been screened for interest in attending a financial literacy course was thus only 17.8 percent. This number is low compared to attendance rates for business 224  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES training courses. A study in Mexico found a 65 percent attendance rate for busi- ness training. Attendance rates for business training interventions in other coun- tries range from 39 to 92 percent (McKenzie and Woodruff 2012). 7.5 OVERCOMING BARRIERS TO ATTENDANCE THROUGH EXPERIMENTAL INTERVENTIONS We first examine theoretical reasons why individuals may choose not to partici- pate in financial literacy training, and then describe the experiments designed to test these barriers. 7.5.1 Why might people not want to participate in financial literacy training? Let c be the cost of attending the financial literacy program. While the program itself is free, individuals would incur transportation costs in getting to and from the program, as well as the opportunity cost of lost income (or lost leisure time). Let bt be the benefits the individuals will realize in period t from participating in the course, such as better financial outcomes, and Ebt the certainty-equiva- lent expectation of these benefits. Then theory will predict that an individual will choose to attend a financial literacy course if the expected discounted benefits of the course exceed the costs of attending. i.e., if T ∑t = 0βδt Ebt > c (7.1) Revealed preference would suggest that anyone whom it would benefit to take the financial literacy course would have done so, while anyone who chooses not to take the course is doing so because he or she does not view the benefits as exceeding the costs. This theory then suggests several barriers that may prevent individuals from participating in the financial literacy course, even if it has positive benefits to them (∑tT = 0 bt > 0). A first reason is just that they face costs of attending, so that c is large in magnitude. A second set of reasons concerns the timing of when costs are incurred relative to when benefits are received. In particular, individuals may not participate because the costs are experienced immediately, while the bene- fits may take time to accrue. Individuals with high discount rates (low δ) may find the discounted value of benefits is less than the current costs. Individuals who are present biased (β < 1) can have time-inconsistent preferences; while they would like to have attended financial literacy training in the future, because the benefits occur in the future and attendance occurs today, they keep putting the course off. Third, individuals may not know the benefits of participating, potentially underval- uing them. Even if their expected benefits are accurate, with risk aversion, uncer- tainty as to these benefits will still cause Ebt < bt. Finally, one could also imagine that liquidity constraints prevent individuals from paying the costs today, even if 7.  Why is voluntary financial education so unpopular?  ◾ 225 they see positive net expected benefits. This final explanation seems less relevant in our case where many individuals have credit cards and most have savings. 7.5.2 Experimental interventions to overcome barriers to attendance We designed a second stage of interventions to explore if we could tease out whether high uncertainty about benefits, transport costs, high opportunity costs, impatience, or low appeal for the course could explain individuals’ low attendance and thereby learn what policy efforts could spur attendance. The treatment group was divided further into six different groups—one group that received no further assistance, and the following five booster treatment groups: 1. Offered Mex$1,000 (US$72) for completing the training: participants were given a Walmart gift card of Mex$1,000 if they attended12 2. Offered Mex$500 (US$36) gift card for completing the training 3. Offered Mex$500 gift card that they would receive one month after completing the training 4. Offered a free taxi ride to and from the course location 5. Provided a video CD containing positive testimonials from people who had attended the course Treatments 1 and 2 enable us to examine whether individuals are more likely to attend the course as the benefits of attending the course increase. This helps get at the issue of whether individuals are making rational decisions by responding to changes in the net benefits, as well as to measure how the demand for training varies with these benefits. The comparison of Treatments 2 and 3 enables us to see whether high discount rates or present bias is a reason for lack of attendance—people might think the course has benefits, but because these benefits occur in the future and attendance occurs today, they keep putting the course off. If this was the issue, we would expect much greater response to Treatment 2 than to Treatment 3. Treatment 4 lowers the costs of attending the training, which enables examination of whether one important component of c is the constraint. Treatment 5 aims to reduce informational constraints that may prevent people from attending the course if they are not sure it will be helpful, thereby attempting to reduce the difference between Ebt and bt. The incentive treatments were assigned through stratified randomization using the entire treatment group. We stratified by whether the respondent was screened through the branch versus mail, online, or street survey; by income 12  For comparison, at follow-up, median monthly income in our sample was about Mex$9,000 (US$650). 226  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES group; by whether the person was a client of our partner financial institution, a client of another financial institution, or neither; and by attendance status thus far. The attendance status categories were attended, was scheduled but did not attend, was reached but did not want to be scheduled, and cannot be reached. Individuals who had already attended the course did not receive the incentives, but they were included in the randomization to allow us to report post-incentive treatment take-up rates by incentive type for the complete treatment group. 7.5.3 Impacts of experiments to increase attendance Figure 7.1 shows the overall percentages attending in each incentive treatment group (after the initial treatment offer and the booster treatments). Some partic- ipants were unreachable by phone, so we show results as a percentage of all participants assigned to a treatment, and of all who could be reached to actually offer them the treatment. Table 7.2 provides regression estimates of the impacts of the various treatment arms on attendance. As a result of the booster interventions, we succeeded in getting an addi- tional 114 individuals to attend the course, giving a total of 426 attendees out of 1,751 treatment group individuals (24.3 percent). Offering a monetary incentive of US$36 increased the take-up rate from about 18 percent to 27 percent of those assigned to treatment, while the US$72 incentive increased take-up further to 33 percent. While the difference between the two monetary amounts is not statis- tically significant, they both suggest that individuals are rationally responding to FIGURE 7.1  Financial literacy training take-up rates by incentive group Percent 40 39 Full treatment group Only individuals who could be reached 33 32 32 30 27 27 25 23 21 21 19 20 18 10 0 No extra incentive US$72 now US$36 now US$36 later Free transportation Testimonials Note: Full treatment group includes 1,751 individuals. The subsample who could be reached to offer them the treatment includes 1,485 individuals. 7.  Why is voluntary financial education so unpopular?  ◾ 227 TABLE 7.2  Impact of incentive treatments on take-up in treatment group DEPENDENT VARIABLE: ATTENDED COURSE (1) (2) US$72 now dummy  0.1480*** 0.1758*** (0.0352) (0.0401) US$36 now dummy  0.0931*** 0.1123*** (0.0343) (0.0393) US$36 later dummy  0.0918*** 0.1110*** (0.0343) (0.0392) Free transportation dummy  0.0330 0.0407 (0.0326) (0.0376) Testimonials dummy  0.0142 0.0153 (0.0320) (0.0367) F-test p -value: US$72 now = US$36 now 0.1435 0.1357 F-test p -value: US$36 now = US$36 later 0.9725 0.9761 Control group mean of outcome variable 0.1815 0.2129 Observations 1,751 1,457 Sample Full treatment Only individuals group who could be reached Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Ordinary least squares regressions of a dummy for having attended the course on a set of dummies indicating to which incentive group the individual was randomly assigned (the omitted category is the control group, i.e., received no extra incentives). Regressions control for ran- domization strata dummies. higher benefits of training by being more likely to attend. The treatment impact is exactly the same when US$36 is offered immediately at the completion of training, or one month after training. This suggests that high discounts or present bias are not the main barriers to participation in training. In contrast to the monetary incentives, we find that transportation assistance and the testimonials did not significantly increase attendance. One reason for the lack of impact of transportation assistance was security concerns in Mexico City, with people reluctant to take a taxi cab that came to their home. The lack of impact of the testimonials could reflect people updating their beliefs about the efficacy of the training very little after the receipt of this treatment, or that lack of information about the benefits and quality of the training was not the main reason for nonparticipation. Finally, note that even when participants were offered US$72 to attend, it is still the case that the majority of individuals who had initially expressed interest in attending financial literacy training do not attend.13 This dissociation of intentions and follow-through has been documented in studies of 13  retirement seminars (Madrian and Shea 2001). 228  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 7.5.4 Which individuals attend training? If we break down the take-up rate by individuals who are clients of a financial insti- tution and those who are not, we find that the attendance rate was 28.1 percent among clients versus 18.1 percent among nonclients. Since we do not expect to be able to detect treatment effects with a take-up rate of 18.1 percent and since our randomization was stratified by being a financial institution client, we decided to collect follow-up data only on the 2,178 individuals who are clients of a financial institution. From this point on, all tables and analysis in this chapter cover only these individuals. Using data from the screener survey, we analyze which individual background characteristics are correlated with training attendance in the treatment group. Table 7.3 displays this analysis, using two different measures of take-up: (1) a dummy vari- able for whether the individual attended the course prior to our booster incentive interventions, and (2) a dummy variable for whether the individual attended either before or after we offered the incentives.14 The bottom row of table 7.3 shows that the take-up rate increased from 20.8 percent to 28.5 percent after the incentive intervention (in our sample of financial institution clients). Two characteristics are strongly and positively correlated with attendance: education and age. Individuals who have a bachelor’s degree or higher are up to 14 percentage points more likely to have attended the training. The take-up rate among this group was 38 percent after the incentive intervention. Older individuals are also more likely to have partic- ipated in the training. In addition, we find weak evidence that females and indi- viduals who are clients of our partner financial institution as opposed to another institution are slightly more likely to have attended the training. 7.6 IMPACTS OF FINANCIAL LITERACY TRAINING In this section, we first discuss our data and estimation methodology, and then turn to impacts of the financial literacy training on financial knowledge, behavior, and outcomes. 7.6.1 Follow-up survey We conducted a follow-up survey between February and July 2012 to measure post-training financial knowledge, behavior, and outcomes. We kept the question- naire relatively short (about 15 minutes) to encourage participation. The questions We do not include income among the variables displayed in table 7.3, since this variable 14  has many missing values and including it would thus reduce the sample size. When we add income to the analysis, it is not correlated with take-up, and the coefficients on the other variables remain similar to the ones shown in table 7.3. 7.  Why is voluntary financial education so unpopular?  ◾ 229 TABLE 7.3  Determinants of program take-up in treatment group DEPENDENT VARIABLE ATTENDED COURSE WITH- ATTENDED COURSE WITH OUT INCENTIVES OR WITHOUT INCENTIVES (1) (2) Baseline survey conducted in branch 0.0437 0.0250 (0.0298) (0.0330) Client of partner financial institution 0.0232 0.0540* (versus other institution) (0.0271) (0.0300) Made savings deposit during past month 0.0284 0.0359 (0.0258) (0.0285) Has credit card −0.0450 −0.0296 (0.0355) (0.0393) Paid more than credit card minimum in 0.0570 0.0309 all past 6 months (0.0392) (0.0435) Has bachelor’s degree or higher 0.1189*** 0.1420*** (0.0254) (0.0282) Female  0.0445* 0.0405 (0.0245) (0.0272) Age 0.0035*** 0.0039*** (0.0011) (0.0012) Occupation is employee  −0.0127 −0.0137 (0.0243) (0.0270) Paid credit card late in past 6 months  0.0524 0.0503 (0.0454) (0.0467) Observations 1,081 1,081 Mean of outcome variable 0.2077 0.2849 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. focused on concepts and behaviors taught in the course. We discuss specific questions and outcome variables below. For logistic reasons, we first attempted to conduct the follow-up survey over the phone. If the person did not respond to the survey during the first attempt, we offered them a Mex$500 (US$36) Walmart gift card for completing the survey during the second attempt. If we were still not able to interview the person over the phone, a surveyor visited the individual’s house to conduct a face-to-face interview. If the participant was not at home, the surveyor delivered a letter with information about our study and instructions for how to contact us to participate in the survey and to receive the Walmart gift card. Surveyors made two more attempts (three attempts in total) to conduct a face-to-face interview if a respon- dent was not at home. We were able to interview 72.8 percent of our sample during the follow-up survey. The attrition rate was slightly higher in the treatment group (29 percent) 230  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES than in the control group (25.3 percent). The last two columns of table 7.1 show baseline characteristics for the sample of individuals interviewed at follow-up. The characteristics are very similar to the full sample, and we do not find any statistically significant differences between control and treatment group means in the follow-up sample. This shows the difference in attrition rates between the two groups is not leading to imbalance on observable characteristics. 7.6.2 Administrative data We obtained administrative data on savings account balances and credit card outcomes from our partner institution. For confidentiality reasons, our partner cannot disclose individual-level data, but it offered to generate summary statis- tics at the treatment and control group level. This was going to be straightforward with the sample we screened through the mail survey, since the list of individ- uals in this sample came from our partner financial institution, meaning that it could easily find these individuals in its records. The low response rate to this survey implied, however, that our sample now consists almost entirely of individ- uals who were screened through the street or branch surveys. About half of the current sample (1,034 individuals) reported being clients of our partner financial institution, and 470 of these individuals could be found in the institution’s records based on name, address, and phone number.15 The last two columns of table 7.1 show our baseline statistics for this sample of 470 clients. Overall, this group is similar to the full sample. Also, the characteristics of the treatment and control group clients who were matched with administrative data were not statistically different at baseline. 7.6.3 Estimating treatment impacts We estimate the impact of the financial literacy training with the following inten- tion-to-treat equation yi,s,m = α + βTrainingInvitei,s,m + ∑γsds+ ∑δmdm + εi,s,m (7.2) where yi,s,m is a follow-up survey measure of the financial knowledge, behavior, or outcome of individual i, in randomization strata s, who was surveyed in month m. The variable TrainingInvitei,s,m indicates whether an individual was invited to the course or not and is thus equal to 1 for the treatment group and equal to 0 for the control group. We control for randomization strata dummies ds, as well as month of follow-up interview dummies dm. The main coefficient of interest is β, which represents the treatment effect of being invited to a financial literacy course. The coefficient β in equation 7.2 is also equal to the difference in means of the For ethical reasons and given the security concerns in Mexico City, we did not ask for 15  date of birth or national ID numbers in the survey. 7.  Why is voluntary financial education so unpopular?  ◾ 231 outcome variable yi,s,m across the treatment and control groups, conditional on strata and interview month dummies. In addition, we estimate the following local average treatment effect regres- sion: yi,s,m = α + βAttendedTrainingi,s,m + ∑γsds+ ∑δmdm + εi,s,m (7.3) where AttendedTrainingi,s,m is equal to 1 for treatment group individuals who attended a course and 0 otherwise. We instrument this variable with our indi- cator variable for whether an individual was invited to the training or not (TrainingInvitei,s,m). The coefficient of β here represents the local average treat- ment effect: that is, the impact of the financial literacy course on the individuals who attended a course as a result of being invited to it but who would not have otherwise attended. 7.6.4 Impact on financial knowledge We measure financial knowledge through eight follow-up survey questions based on the material that was taught in the course. For each question, we create a dummy variable indicating whether the respondent gave the correct answer to the question. These eight questions are then aggregated into a financial knowl- edge index that is the average of the eight dummy variables, and thus represents the fraction of questions the respondent answered correctly. Finally, we asked individuals to rate their own level of financial education on a scale from 1 to 5 (1 = excellent, 2 = good, 3 = satisfactory, 4 = unsatisfactory, 5 = no knowledge of the topic). Based on the responses to this question, we coded a dummy variable indi- cating whether the self-assessed level of financial literacy is satisfactory or higher. Table 7.4 lists the eight dummy variables used to measure financial literacy, as well as the knowledge index and self-assessed knowledge level. The table shows the control group mean for each variable and the intention-to-treat and local average treatment effect estimates that correspond to the coefficients β in equations 7.2 and 7.3. The results in table 7.4 show that the course had a positive and statistically significant impact on financial knowledge. The knowledge index increased from 0.31 in the control group to 0.34 in the treatment group due to the training, meaning that control group individuals answer 31 percent of the eight questions correctly, while treatment group individuals answer 34 percent correctly. The size of this impact is relatively small, but the effect is larger for some individual questions. For example, only 13 percent of control group individuals know that bank deposits are insured up to 400,000 UDIs,16 but the training increased this number to 20 percent in the treatment group. Similarly, 24 percent of the control group knows what total 16  A unidad de inversion (UDI) is an inflation-adjusting currency unit. 232  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 7.4  Impact on financial knowledge ITT LATE CONTROL TREATMENT TREATMENT SAMPLE SIZE MEAN DIFFERENCE DIFFERENCE Panel A: Knowledge index and components Knowledge index (average of 8 1,586 0.31 0.0307*** 0.0871*** components below)  (0.0094) (0.0261) (1) Knows what UDI isa  1,580 0.10 0.0044 0.0125 (0.0150) (0.0426) (2) Knows deposit insurance 1,578 0.13 0.0732*** 0.2073*** exists up to 400,000 UDIs (0.0188) (0.0533) (3) Knows what a credit report is 1,575 0.39 0.0518** 0.1464** (0.0245) (0.0684) (4) Knows credit card cycle is 1,568 0.46 0.0190 0.0535 30 days (0.0251) (0.0701) (5) Knows they have 20 days to 1,568 0.12 0.0143 0.0402 pay credit card without interest (0.0168) (0.0472) (6) Knows what total annual 1,557 0.24 0.0369* 0.1036* cost isb (0.0216) (0.0602) (7) Knows what an AFORE 1,559 0.72 0.0499** 0.1409** (pension fund) is (0.0219) (0.0613) (8) Knows retirement age is 65 1,551 0.29 0.0253 0.0714 (0.0234) (0.0659) Panel B: Self-assessed financial literacy Says their financial knowledge is 1,550 0.58  0.0537** 0.1500** satisfactory or higher (0.0248) (0.0683) Note: Robust standard errors in parentheses. *, **, and *** indicate statistically different from control mean at the 10%, 5%, and 1%, respectively, after controlling for randomization strata and month of follow-up interview dummies. ITT = intention to treat; LATE = local average treatment effect. a. A unidad de inversion (UDI) is an inflation-adjusting currency unit. b. Total annual cost of credit, including all interest rates and fees. annual cost of credit is; this number increased to 28 percent in the treatment group due to the training. The training also increased the self-assessed level of financial literacy, with 63 percent of the treatment group saying their knowledge is satisfac- tory or higher, compared to 58 percent in the control group. When interpreting the magnitude of the effects, it is important to keep in mind that only about 30 percent of the treatment group attended the financial literacy course. The local average treatment effect estimates in the last column of table 7.4 take into account this relatively low take-up rate. They show much larger impacts on individuals who actually went to the training as a result of being invited through the intervention. The knowledge index increased by 8.7 percentage points due to the course for these individuals, and the probability of answering some of the specific knowledge questions correctly increased by up to 20 percentage points. 7.  Why is voluntary financial education so unpopular?  ◾ 233 7.6.5 Impact on savings behavior and outcomes The financial literacy course emphasized specific behaviors that may help individ- uals save more money, including checking financial institution transactions regu- larly and keeping track of expenses, making a budget and setting a savings goal, and identifying necessary and unnecessary expenses to reduce overall expendi- tures. The follow-up survey asked whether individuals engage in these behaviors, and we code five dummy variables indicating that they do, as listed in panel A of table 7.5. We aggregate these behaviors into a savings behavior index by taking an average of the five dummies, giving the fraction of the five behaviors in which individuals engage. The control group means in panel A of table 7.5 show that many individuals check their financial institution transactions regularly (69 percent), keep track of expenses (79 percent), and make a budget (77 percent), even without having taken the financial literacy course. We do not detect a significant treatment effect of the training on these behaviors, perhaps because these behaviors were already quite common in the control group. The remaining two savings behaviors we study are not as frequently used in the control group: 57 percent of individuals have a savings goal, and 59 percent have cut expenses in the past three months. We find that the course increased the percentage of individuals who cut expenses during the past three months to 63 percent, but this effect is only statistically significant at the 10 percent level. In addition to studying savings behavior, we also examine whether the training had an impact on savings outcomes. Panel B of table 7.5 lists three measures of personal savings: a dummy variable indicating whether the indi- vidual has any type of savings/money set aside, a dummy variable for whether the individual reports saving at least some fraction of his or her income during the past six months, and a dummy variable indicating whether the respondent said he or she saves more each month than a year ago. We construct a savings outcome index by taking the average of these three dummy variables. All three savings outcomes are higher in the treatment group than in the control group at follow-up, although the differences are not statistically significant for the individual variables. We do, however, find a positive and significant impact of the course on the savings outcome index. Control group individuals reach on average 65 percent of all three savings outcomes, increasing to 68 percent in the treatment group. Recall that we provided monetary incentives of Mex$500 or Mex$1,000 (US$36 or US$72) to some randomly chosen treatment group individuals in order to encourage them to attend the financial literacy course. These amounts are equivalent to 5.5 and 11 percent of median monthly income in our sample. To check whether the positive effect on savings outcomes is driven by the monetary payments instead of by the course itself, we add a dummy variable to estima- tion questions (1) and (2), indicating whether the individual received an incentive 234  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 7.5  Impact on savings behavior and outcomes ITT LATE CONTROL TREATMENT TREATMENT SAMPLE SIZE MEAN DIFFERENCE DIFFERENCE Panel A: Savings behavior index and components Savings behavior index (average 1,586 0.68 0.0133 0.0376 of 5 components below) (0.0124) (0.0348) (1) Checks financial institution 1,586 0.69 −0.0235 −0.0666 transactions regularly (0.0226) (0.0644) (2) Keeps track of expenses 1,586 0.79 0.0072 0.0204 (0.0206) (0.0582) (3) Makes a budget 1,585 0.77 0.0264 0.0748 (0.0211) (0.0596) (4) Has a savings goal 1,582 0.57 0.0130 0.0367 (0.0250) (0.0705) (5) Cut expenses in past 3 1,584 0.59 0.0428* 0.1212* months (0.0247) (0.0698) Panel B: Savings outcomes index and components Savings outcome index (average 1,586 0.65 0.0335** 0.0948** of 3 components below) (0.0147) (0.0414) Has any type of savingsa 1,586 0.80 0.0288 0.0814 (0.0200) (0.0566) Saved more than zero during 1,413 0.83 0.0293 0.0800 past 6 months (0.0192) (0.0524) Saves more each month than a 1,547 0.36 0.0408 0.1151* year ago (0.0250) (0.0697) Panel C: Savings outcomes controlling for monetary incentives Savings outcome index (average 1,586 0.65 0.0268* 0.0902* of 3 components below) (0.0151) (0.0506) Has any type of savingsa  1,586 0.80 0.0187 0.0629 (0.0205) (0.0693) Saved more than zero during 1,413 0.83 0.0253 0.0823 past 6 months  (0.0198) (0.0642) Saves more each month than a 1,547 0.36 0.0347 0.1161 year ago  (0.0255) (0.0845) Note: Robust standard errors in parentheses. *, **, and *** indicate statistically different from control mean at the 10%, 5%, and 1%, respectively, after controlling for randomization strata and month of follow-up interview dummies. Regressions in panel C additionally include a dummy for whether the individual received a monetary incentive payment for participa- tion in the financial literacy course. ITT = intention to treat; LATE = local average treatment effect. a. Includes savings account, caja de ahorro, tanda, and other nonretirement savings. payment for course participation. Panel C of table 7.5 replicates the results in panel B, controlling for the monetary incentive dummy variable. The estimated impact of the course on savings outcomes is slightly smaller, but remains statis- tically significant. 7.  Why is voluntary financial education so unpopular?  ◾ 235 The course included material on retirement savings and pension funds. Table 7.6 shows the impact of the course on retirement savings behavior and outcomes. In the control group, 57 percent of individuals have a pension fund; the course did not change this percentage in the treatment group. Pension funds are typically provided through an individual’s employer, implying that we did not necessary expect to find an impact of the course on this outcome since acquiring a pension fund may involve switching jobs. The other variables reported in this table are based on follow-up survey questions that were only answered by individuals who have a pension fund. In order to account for possible selection bias, we fill the variables in with 0 for individuals who do not have a pension fund, so that they are defined for the complete sample, except for individuals who did not know whether they have a pension fund or not. In terms of retirement savings behaviors, we measure whether (1) individ- uals choose a pension fund administrator based on fees or returns, as opposed to using the default option, relying on friends’ advice etc.; (2) they check their pension fund statement regularly; and (3) they have calculated how much money they will need upon retirement. We construct an index of retirement savings behavior that is the average of the three behavior dummy variables. As the retire- ment savings outcome, we examine whether individuals report saving money for TABLE 7.6  Impact on retirement savings behavior ITT LATE CONTROL TREATMENT TREATMENT SAMPLE SIZE MEAN DIFFERENCE DIFFERENCE Has a pension fund 1,471 0.57 −0.0031 −0.0088 (0.0263) (0.0739) Panel A: Retirement savings behavior index and components  Retirement savings behavior index 1,471 0.20 0.0114 0.0320 (average of 3 components below) (0.0144) (0.0404) Pension fund administrator choice 1,471 0.16 0.0058 0.0162 based on fees or returns (0.0195) (0.0547) Checks pension fund statement 1,467 0.34 0.0340 0.0958 (0.0250) (0.0704) Has calculated how much money 1,465 0.11 −0.0026 −0.0074 will need upon retirement (0.0158) (0.0447) Panel B: Retirement savings outcomes Is saving money for retirement 1,470 0.17 0.0122 0.0343 (0.0199) (0.0559) Note: Robust standard errors in parentheses. *, **, and *** indicate statistically different from control mean at the 10%, 5%, and 1%, respectively, after controlling for randomization strata and month of follow-up interview dummies. Variables in panels A and B are based on questions that were only answered by individuals who have a pension fund. To account for potential selection bias, we fill these variables in with 0 for individuals who do not have a pension fund. ITT = intention to treat; LATE = local average treatment effect. 236  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES retirement (on top of their pension fund).17 We do not find an impact of the training on any of the retirement savings behaviors or outcomes listed in table 7.6. 7.6.6 Impact on borrowing behavior and outcomes The course also discussed responsible use of credit cards and healthy borrowing behavior. Table 7.7 shows that 48 percent of the control group had a credit card at follow-up. This percentage was not statistically different in the treat- ment group. For individuals with a credit card, we define several measures of credit card behavior and outcomes, as listed in panels A and B of table 7.7. The measures are based on questions that were only posed to individuals with a credit card and refer to their most frequently used card. The variables are filled in with 0 for individuals who do not have a credit card, to account for possible selection bias. We examine six credit card behaviors. Three are dummy variables indicating whether the person knows the credit limit, knows the credit card interest rate, and checks the credit card statement every month. The other three variables are coded as the fraction of the past six months, where the individual paid the balance in full, made only the minimum payment, and received cash through the credit card. We construct a credit card behavior index by first converting all vari- ables to z-scores using the control group mean and standard deviation and then taking the average of these z-scores. The two behaviors that are last on the list are coded with a negative sign in the average, since they represent undesirable credit card behavior. We study three credit card outcomes: (1) whether the issuer blocked the credit card during the past six months, (2) the fraction of the past six months when the individual was charged late payment fees, and (3) the fraction of the past six months where the individual was charged overdraft fees. The incidence of all three events is quite low in the sample, ranging from 1 to 4 percent. We also construct an index of these three outcomes, using the average of the z-scores. The results in table 7.7 show no impact of the course on credit card behavior or outcomes. Table 7.8 examines borrowing behavior more broadly. We look at three loan behaviors, coded as dummy variables that indicate whether the individual applied for a loan from any source during the past six months, went to a pawn shop for credit during the past six months, and stopped servicing outstanding debt during the past six months. We aggregate these behaviors into a loan behavior index by taking the average of the three dummy variables. In the control group, 23 percent We intended to also ask questions on having calculated how much money is needed 17  upon retirement and whether the individual is saving money for retirement in the complete sample, independent of whether the person has a pension fund. However, by mistake, these questions were only posed to individuals who reported having a pension fund. 7.  Why is voluntary financial education so unpopular?  ◾ 237 TABLE 7.7  Impact on credit card behavior and outcomes ITT LATE CONTROL TREATMENT TREATMENT SAMPLE SIZE MEAN DIFFERENCE DIFFERENCE Currently has at least one credit 1,560 0.48 −0.0287 −0.0811 card (0.0210) (0.0596) Panel A: Credit card behavior index and components Credit card behavior index (average 1,556 0.00 −0.0233 −0.0660 of z-scores of 6 components below)  (0.0229) (0.0652) (1) Knows credit limit  1,550 0.45 −0.0266 −0.0756 (0.0209) (0.0594) (2) Knows interest rate  1,519 0.20 −0.0017 −0.0050 (0.0193) (0.0552) (3) Checks statement every month  1,543 0.41 −0.0278 −0.0786 (0.0209) (0.0594) (4) Fraction of past 6 months where 1,536 0.20 −0.0180 −0.0505 paid balance in full  (0.0170) (0.0479) (5) Fraction of past 6 months where 1,539 0.12 0.0022 0.0062 made only the minimum paymenta  (0.0138) (0.0389) (6) Fraction of past 6 months where 1,540 0.07 −0.0094 −0.0267 got cash through credit carda (0.0098) (0.0281) Panel B: Credit card outcomes index and components Credit card outcomes index (average 1,554 0.00 0.0434 0.1228 of z-scores of 3 components below)  (0.0416) (0.1184) (1) Issuer blocked credit card during 1,547 0.04 0.0009 0.0026 past 6 months  (0.0096) (0.0270) (2) Fraction of past 6 months where 1,546 0.03 0.0102 0.0289 late payment fees were charged (0.0064) (0.0183) (3) Fraction of past 6 months where 1,545 0.01 0.0026 0.0075 overdraft fees were charged (0.0034) (0.0098) Note: Robust standard errors in parentheses. *, **, and *** indicate statistically different from control mean at the 10%, 5%, and 1%, respectively, after controlling for randomization strata and month of follow-up interview dummies. All variables in panels A and B refer to the most frequently used credit card. a. Included in credit card behavior index with negative sign. of individuals applied for a loan during the past six months, 10 percent went to a pawn shop, and 13 percent stopped servicing outstanding debt. These percent- ages are not statistically different in the treatment group. We also find no effect of the course on the index of loan behavior. Panel B of table 7.8 shows the impact of the course on loan outcomes. One-third of control group individuals have a loan from any source, and their total outstanding debt represents about 15 percent of annual income. These numbers are statistically identical in the treatment group, implying that the training had no effect on loan outcomes. 238  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 7.8  Impact on loan behavior and outcomes ITT LATE CONTROL TREATMENT TREATMENT SAMPLE SIZE MEAN DIFFERENCE DIFFERENCE Panel A: Loan behavior index and components Loan behavior index (average of 3 com- 1,570 0.15 0.0075 0.0212 ponents below)      (0.0118) (0.0334) (1) Applied for a loan from any source 1,564 0.23 −0.0074 −0.0208 during past 6 months      (0.0210) (0.0594) (2) Went to a pawn shop to get credit 1,568 0.10 0.0054 0.0152 during past 6 months      (0.0152) (0.0429) (3) Stopped servicing outstanding debt 1,434 0.13 0.0205 0.0591 during past 6 months      (0.0180) (0.0523) Panel B: Loan outcomes index and components Loan outcomes index (average of 1,560 −0.01 −0.0132 −0.0372 z-scores of 2 components below)      (0.0427) (0.1206) (1) Currently has a loan (from any 1,555 0.33 −0.0058 −0.0165 source)     (0.0234) (0.0660) (2) Total outstanding debt as 1,209 15.38 −0.6563 −1.7753 percentage of annual income      (1.1899) (3.2220) Note: Robust standard errors in parentheses. *, **, and *** indicate statistically different from control mean at the 10%, 5%, and 1%, respectively, after controlling for randomization strata and month of follow-up interview dummies. ITT = intention to treat; LATE = local average treatment effect. 7.6.7 Heterogeneous treatment effects We now ask whether the impact of the course was different for different groups of people. The take-up regressions in table 7.3 show that individuals with a bach- elor’s degree, as well as females and clients of our partner financial institution, were more likely to attend the training compared to other people in the treatment group. The treatment effects may be larger for these groups since more of them were exposed to the course material. It could also be the case that individuals who know that they will benefit more from the course are more likely to attend in the first place. The treatment randomization was stratified by gender, being a client of our partner institution, and having a bachelor’s degree. We can thus examine the effect of the training on these groups of people separately by adding an interac- tion term between the treatment group dummy (TrainingInvitei,s,m ) and a dummy variable indicating whether the individual was female, a client of our partner insti- tution, or had a bachelor’s degree, to equation 7.2. Table 7.9 reports the heterogeneous treatment effect regression, using all indexes from tables 7.4 through 7.8 as the outcome variables. We do not study the index components here to minimize multiple hypothesis testing. The results in panel A show that the course had a similar impact on clients of our partner 7.  Why is voluntary financial education so unpopular?  ◾ 239 TABLE 7.9  Heterogeneous treatment effects (intention to treat) DEPENDENT VARIABLE: INDEX OF CREDIT CARD CREDIT CARD KNOWLEDGE RETIREMENT OUTCOMES OUTCOMES OUTCOMES BEHAVIOR BEHAVIOR BEHAVIOR BEHAVIOR SAVINGS SAVINGS SAVINGS LOAN LOAN Panel A: Type of financial institution Treatment group 0.0315** 0.0095 0.0343* 0.0099 −0.0550* 0.0157 0.0065 −0.0066 dummy (0.0130) (0.0174) (0.0199) (0.0194) (0.0294) (0.0559) (0.0161) (0.0548) Treatment*Client of −0.0016 0.0078 −0.0018 0.0030 0.0658 0.0572 0.0021 −0.0135 partner institution  (0.0188) (0.0246) (0.0295) (0.0287) (0.0460) (0.0836) (0.0237) (0.0859) F-test p-value: 0.028 0.321 0.135 0.543 0.761 0.240 0.622 0.761 Treatment + Inter- action = 0 Panel B: Gender Treatment group 0.0182 0.0057 0.0546*** −0.0035 −0.0345 0.0587 0.0144 0.0414 dummy  (0.0134) (0.0169) (0.0205) (0.0215) (0.0331) (0.0564) (0.0171) (0.0626) Treatment*Female  0.0250 0.0150 −0.0421 0.0297 0.0223 −0.0304 −0.0137 −0.1086 (0.0188) (0.0246) (0.0295) (0.0287) (0.0459) (0.0842) (0.0237) (0.0858) F-test p-value: 0.001 0.249 0.554 0.172 0.702 0.648 0.967 0.249 Treatment + Inter- action = 0 Panel C: Education           Treatment group 0.0288** −0.0067 0.0148 −0.0139 −0.0174 0.0585 0.0017 −0.0421 dummy (0.0116) (0.0164) (0.0192) (0.0179) (0.0284) (0.0575) (0.0161) (0.0546) Treatment*Has 0.0046 0.0486* 0.0455 0.0613** −0.0144 −0.0369 0.0141 0.0702 bachelor or higher  (0.0196) (0.0248) (0.0300) (0.0297) (0.0477) (0.0819) (0.0237) (0.0877) F-test p-value: 0.034 0.025 0.009 0.046 0.406 0.713 0.363 0.681 Treatment + Inter- action = 0 Observations 1,586 1,586 1,586 1,471 1,556 1,554 1,570 1,560 Note: Robust standard errors in parentheses. , , and * ** indicate statistically different from control mean at the 10%, 5%, *** and 1%, respectively, after controlling for randomization strata and month of follow-up interview dummies. financial institution and on clients of other institutions. The positive effect on financial knowledge and on savings outcomes is not statistically different across these two groups. Panel B similarly shows no statistically significant differences in treatment effects by gender. The only indication of heterogeneity in impacts is found in panel C, which shows that individuals with a bachelor’s degree are more likely to improve savings behavior and retirement savings behavior than individuals without this education level. Since the take-up rates for training were also higher for the more educated, this could just reflect the fact that this group was more likely to attend training. 240  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES We therefore estimate local average treatment effect impacts by education level to see whether the effect of actually receiving training varies with education status. We find that receiving training had a larger effect on individuals with a bachelor’s degree for both the savings behavior and retirement savings behavior index (although the heterogeneous treatment effect on savings behavior is only statistically significant at the 10.4 percent level). 7.6.8 Comparison of treatment and control group outcomes with administrative data Our partner financial institution provided monthly data on savings and credit card outcomes at the treatment and control group level for the 470 clients found in their database, from December 2010 through May 2011. These data cover several pre-intervention months, since we started inviting clients to the financial literacy course in August 2011. It also overlaps with the follow-up survey, which was conducted between February and July 2012. Figures 7.2 to 7.4 plot the administrative data for the treatment and control groups over time. Figure 7.2 shows that the median savings account balance in the treatment and control groups followed a more or less parallel trend before our intervention started in August 2011. In October 2011, the savings balance starts rising in the treatment group, going from about Mex$900 in September 2011 to Mex$1,600 in December 2011, while the savings balance in the control group stays relatively constant. After January 2011, the savings balance in the treatment group starts falling again and goes back to its original level in April 2012. Although the aggregate data do not allow us to conduct a test of statistical FIGURE 7.2  Median savings account balance Pesos 3,500 3,000 Treatment group 2,500 2,000 1,500 1,000 Control group 500 0 Dec Feb Apr Jun Aug Oct Dec Feb Apr 2010 2011 2011 2011 2011 2011 2011 2012 2012 Note: Administrative data. Sample includes 470 individuals who are clients of our partner financial institution. 7.  Why is voluntary financial education so unpopular?  ◾ 241 significance in the difference of medians, the observed pattern is consistent with the course having a positive effect on savings account balances. However, this effect appears to be temporary. Figures 7.3 and 7.4 display median credit card balances and average percentage of credit card debt paid off each month, respectively. The treatment and control group values follow a similar pattern throughout the period, indicating no impact of the course on credit card balances or debt paid off each month. FIGURE 7.3  Median credit card balance Pesos 9,000 8,000 7,000 Treatment group 6,000 5,000 Control group 4,000 3,000 2,000 1,000 0 Dec Feb Apr Jun Aug Oct Dec Feb Apr 2010 2011 2011 2011 2011 2011 2011 2012 2012 Note: Administrative data. Sample includes 470 individuals who are clients of our partner financial institution. FIGURE 7.4  Average percentage of credit card debt paid off each month Percent 35 30 Control group 25 20 15 10 Treatment group 5 0 Dec Feb Apr Jun Aug Oct Dec Feb Apr 2010 2011 2011 2011 2011 2011 2011 2012 2012 Note: Administrative data. Sample includes 470 individuals who are clients of our partner financial institution. 242  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Overall, the patterns in the administrative data are consistent with the follow-up survey results. The course seems to have had no effect on credit card behavior or outcomes, but it led to a small increase in savings. The administra- tive data suggest that this increase was temporary and dissipated within a few months. 7.7 CONCLUSIONS AND POLICY IMPLICATIONS Despite the popularity of financial literacy workshops among policy makers and financial institutions, voluntary participation rates are typically very low for these programs. Working with a half-day-long training course in Mexico City that aims to teach participants the importance of savings and responsible debt and credit card use, we use randomized experiments to investigate the lack of demand. We find very little interest in participating in financial education among samples of financial institution clients and Facebook users. Screening our sample to focus on those individuals who express interest in attending training, we find the majority of them also do not attend. Experiments to increase take-up suggest that this low participation rate is not mainly due to high discount rates, time inconsistency, or lack of information, but rather appears to be due to indi- viduals thinking that the benefits of such training are not high enough. Mone- tary incentives that increase the benefits lead to more attendance. A follow-up survey conducted about six months after the course enabled us to measure just what these benefits are. We find participating in the course leads to increases in financial knowledge and savings, but this increase in savings appears to dissi- pate quickly, and there is no evidence that the training changed credit card usage or borrowing behavior. The lack of interest in training therefore appears to be a rational choice, since users see relatively little benefit from it. One natural response to the modest impacts of training measured here is to note that the training is only a few hours long, and to thus argue that much longer and intensive training sessions are needed to really affect financial behavior. However, our study shows that most of the general population has very little interest in attending even a short financial literacy course, and that it is very difficult to even get individuals who state they are interested in partici- pating to show up and attend. Monetary incentives did help boost attendance, but even an immediate payment of US$72 motivated less than 40 percent of people who had expressed interest in a financial literacy program to attend. Longer and more intensive programs would appear likely to have even more trouble attracting participants. Moreover, Kim, Garman, and Sorhaindo (2003) provide an example suggesting that even 18 months of adult credit counseling has not been effective. We caution that our study measures local treatment effects, showing the effect of training for people who are induced to attend as a result of our 7.  Why is voluntary financial education so unpopular?  ◾ 243 interventions, but who would not attend it without the information, logistical support, and incentives we provided. It seems plausible that the impacts for these individuals who have self-selected into not attending a program that is available to anyone who wants to participate will have less benefit from such training than people who choose to attend of their own accord. However, as our study notes, less than 2 percent of the eligible population voluntarily chooses to take part in training. Even if the training has large benefits for them, the results of our study suggest that the benefits of encouraging more people to participate in such training are likely to be slight. This concurs with recent skepticism about the value of such training (Economist 2013; Willis 2011). Finally, we note two issues with the increasing focus of financial institu- tions and policy makers on extending financial education to the masses through general multipurpose financial literacy courses. First, the classroom setting for the provision of financial education may not be appealing to the general public. We find higher take-up rates among the population of bachelor graduates who may be more comfortable in such a setting, or may be better at digesting informa- tion provided through this channel. Alternative methods to educate the general public should be tested. For example, Berg and Zia in chapter 11 of this volume provide evidence of an increase in financial knowledge and reductions in certain types of borrowing following financial education taught through a soap opera in South Africa. Second, the variety of topics covered in the course conflicts some- what with the idea that some concepts are best taught at teachable moments. Workshops about pension savings provided to workers at the time of making their pension savings allocations, or to college students or young graduates when they receive their first credit card, may be more effective in influencing these outcomes. In contrast to credit and retirement savings, there is less of an obvious teachable moment for education about savings. Indeed, since individuals make spending and savings decisions on a very regular basis, it could be argued that general financial literacy courses are likely to work better for this topic since individuals have the ability to quickly translate the concepts learned into prac- tice. This might help explain why we find impacts (albeit transitory ones) only on savings outcomes in our study, and be consistent with the savings impacts of financial education programs taught in schools. REFERENCES Aguiar, Mark, and Erik Hurst. 2005. “Consumption versus Expenditure.” Journal of Political Economy 113 (5): 919–48. Akerlof, George. 2001. “Behavioral Macroeconomics and Macroeconomic Behavior.” Nobel Prize Lecture. http://www.nobelprize.org/nobel_prizes/economics/laureates/2001/ akerlof-lecture.pdf. Attanasio, Orazio, and Miguel Székely. 2004. “Wage Shocks and Consumption Variability in Mexico during the 1990s.” Journal of Development Economics 73 (1): 1–25. 244  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Bernheim, B. Douglas, Daniel M. Garrett, and Dean M. Maki. 2001. “Education and Saving: The Long-Term Effects of High School Financial Curriculum Mandates.” Journal of Public Economics 80 (3): 435–65. Brown, Alexander, Zhikang Chua, and Colin Camerer. 2009. “Learning and Visceral Temptation in Dynamic Savings Experiments.” Quarterly Journal of Economics 124 (1): 197–231. Brown, Amy, and Kimberly Gartner. 2007. “Early Intervention and Credit Cardholders: Results of Efforts to Provide Online Financial Education to New-to-Credit and at-Risk Consumers.” Center for Financial Services Innovation, Chicago. http://cfsinnovation. com/system/files/imported/managed_documents/earlyintervention.pdf. Bruhn, Miriam, and Inessa Love. 2009. “The Economic Impact of Banking the Unbanked: Evidence from Mexico.” Policy Research Working Paper 4981, World Bank, Washington, DC. Burke, Jeremy, and Kata Mihaly. 2012. “Financial Literacy, Social Perception and Strategic Default.” RAND Working Paper WR-937, Santa Monica. Cai, Jing. 2012. “Social Networks and the Decision to Insure: Evidence from Randomized Experiments in China.” University of California, Berkely. http://gspp.berkeley.edu/ assets/uploads/research/pdf/JingCai_Revised_paper.pdf. Cole, Shawn, Thomas Sampson, and Bilal Zia. 2009. “Financial Literacy, Financial Decisions, and the Demand for Financial Services: Evidence from India and Indonesia.” Working Paper 09-117, Harvard Business School, Cambridge, MA. Cole, Shawn A., and Gauri K. Shastry. 2009. “Smart Money: The Effect of Education, Cognitive Ability, and Financial Literacy on Financial Market Participation.” Finance Working Paper Series 09-071, Harvard Business School, Cambridge, MA. Demirgüç-Kunt, Asli, Thorsten Beck, and Patrick Honohan. 2008. Finance for All? Policies and Pitfalls in Expanding Access. Washington, DC: World Bank. http://siteresources. worldbank.org/INTFINFORALL/Resources/4099583-1194373512632/FFA_book.pdf. Doi, Yoko, David McKenzie, and Bilal Zia. 2012. “Who You Train Matters: Identifying Complementary Effects of Financial Education on Migrant Households.” Policy Research Working Paper 6157, World Bank, Washington, DC. Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Economist. 2013. “Teacher, Leave Them Kids Alone.” February 16. http://www.economist. com/news/finance -and- economic s/21571883-financial- education-has-had- disappointing-results-past-teacher-leave-them. Fernández, Diana. 2010. “¿Cuál es tu nivel de cultura financiera?” CNNExpansión October 5. http://www.cnnexpansion.com/mi-dinero/2010/10/05/condusef-tras-la-educacion- financiera. Gale, William G., and Ruth Levine. 2010. “Financial Literacy: What Works? How Could it be More Effective?” Brookings Institution, Washington, DC. http://www.brookings.edu/ research/papers/2010/10/financial-literacy-gale-levine. Gerardi, Kristopher, Lorenz Goette, and Stephan Meier. 2010. “Financial Literacy and Subprime Mortgage Delinquency: Evidence from a Survey Matched to Administrative Data.” Working Paper 2010-10, Federal Reserve Bank of Atlanta, Atlanta. http://www. frbatlanta.org/documents/pubs/wp/wp1010.pdf. GAO (Government Accountability Office). 2004. “The Federal Government’s Role in improving Financial Literacy: Highlights of a GAO Forum.” http://www.gao.gov/ assets/210/202486.pdf. 7.  Why is voluntary financial education so unpopular?  ◾ 245 Hastings, Justine, and Lydia Tejeda-Ashton. 2008. “Financial Literacy, Information, and Demand Elasticity: Survey and Experimental Evidence from Mexico.” NBER Working Paper 14538, National Bureau of Economic Research, Cambridge, MA. Kim, Jinhee, Thomas Garman, and Benoit Sorhaindo. 2003. “Relationships among Credit Counseling Clients’ Financial Well-Being, Financial Behaviors, Financial Stressor Events, and Health.” Financial Counseling and Planning 14 (2): 75–87. Lara-Ibarra, Gabriel. 2011. “The Effects of Framing on Retirement Management Decisions: Evidence form a Field Study in Mexico.” Unpublished. Lewis, Sue, and Flore-Anne Messy. 2012. “Financial Education, Savings and investments: An Overview.” OECD Working Papers on Finance, Insurance and Private Pensions, No. 22, Organisation for Economic Co-operation and Development, Paris. Lührmann, Melanie, Marta Serra-Garcia, and Joachim Winter. 2012. “The Effects of Financial Literacy Training: Evidence from a Field Experiment with German High-School Children.” Discussion Papers in Economics 14101, University of Munich, Munich. Lusardi, Annamaria, and Olivia S. Mitchell. 2009. “How Ordinary Consumers Make Complex Economic Decisions: Financial Literacy and Retirement Readiness.” NBER Working Paper 15350, National Bureau of Economic Research, Cambridge, MA. ­—. “Financial Literacy and Retirement Planning in the United States.” CeRP Working Papers, Center for Research on Pensions and Welfare Policies, Turin, Italy. Lusardi, Annamaria, and Peter Tufano. 2009. “Debt Literacy, Financial Experience and Overindebtedness.” Dartmouth College Working Paper. http://www.dartmouth. edu/~alusardi/Papers/Lusardi_Tufano.pdf. Madrian, Brigitte, and Dennis Shea. 2001. “Preaching to the Converted and Converting Those Taught: Financial Education in the Workplace.” Working Paper, University of Chicago, Chicago. McKenzie, David. 2006. “The Consumer Response to the Mexican Peso Crisis.” Economic Development and Cultural Change 55 (1): 139–72. —. 2010. “Impact Assessments in Finance and Private Sector Development: What Have We Learned and What Should We Learn?” World Bank Research Observer 25 (2): 209–33. McKenzie, David, and Christopher Woodruff. 2008. “Experimental Evidence on Returns to Capital and Access to Finance in Mexico.” World Bank Economic Review 22 (3): 457–82. —. 2012. “What Are We Learning from Business Training and Entrepreneurship Evaluations around the Developing World?” Policy Research Working Paper 6202, World Bank, Washington, DC. Ponce, Alejandro, Enrique Seira, and Guillermo Zamarripa. 2009. “Do Consumers Borrow on Their Cheapest Credit Card? Evidence from Mexico.” http://papers.ssrn.com/sol3/ papers.cfm?abstract_id=1364722. Saavedra, Ana Luisa. 2012. “Developing National Strategies for Financial Education: The Mexican Case.” Paper presented at the Colombia–Organisation for Economic Co-operation and Development–World Bank Conference on Financial Education, “Progress of Global Policies and Practices and Latin American Experiences,” Cartagena, October 31–November 1. Saldívar, Germán. 2009. “Mexican National Strategy on Financial Education.” Presentation. http://www.oecd.org/finance/financialeducation/44265373.pdf. Seshan, Ganesh, and Dean Yang. 2012. “Transnational Household Finance: A Field Experiment on the Cross-Border Impacts of Financial Education for Migrant Workers.” http://www.afd.fr/webdav/shared/PORTAILS/RECHERCHE/evenements/Migrations- developpement-2012/seshan-yang-2012-transnational-household-finance.pdf. 246  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Stango, V., and J. Zinman. 2009. “Exponential Growth Bias and Household Finance.” Journal of Finance 64 (6): 2807–49. Treviño Garza, Luis. 2011. “Financial Capabilities Measurement.” Presentation at the AFI-BNM Financial Inclusion Policymakers Forum, Kuala Lumpur, April 5–6. http:// www.bnm.gov.my/microsites/fipf2011/day2_session_5/06_FIPF%20Session%205_ Luis%20Trevino.pdf. Visa. 2012. “Visa International Financial Literacy Barometer 2012.” http://www. practicalmoneyskills.com/summit2012/decks/bodnar.pdf. Willis, Lauren. 2011. “The Financial Education Fallacy.” American Economic Review Papers and Proceedings 101 (3): 429–34. Xu, Lisa, and Bilal Zia. 2012. “Financial Literacy around the World: An Overview of the Evidence with Practical Suggestions for the Way Forward.” Policy Research Working Paper 6107, World Bank, Washington, DC. CHAPTER 8 F inancial (dis-)information Evidence from an audit study in Mexico XAVIER GINÉ, CRISTINA MARTÍNEZ CUELLAR, AND RAFAEL KEENAN MAZER ABSTRACT We conduct an audit study in peri-urban Mexico to understand the quality of information and products offered to potential low-income customers. Trained auditors visited multiple financial institutions seeking credit and savings products. Consistent with Gabaix and Laibson (2006), staff provide little information about avoidable fees, especially to auditors trained to reveal little experience with the market. In addition, clients are almost never offered the cheapest product, most likely because staff are incentivized to offer more expensive products that are thus more profitable to the institu- tion. This suggests that disclosure and transparency policies may be ineffec- tive if they undermine the commercial interest of financial institutions. 8.1 INTRODUCTION Many financial decisions are made infrequently and without immediate feedback that can be used to improve decision making (Thaler and Sunstein 2008). In addition, these decisions may involve unfamiliar concepts, espe- cially to individuals with limited financial capabilities (Lusardi and Mitchell We thank the National Commission for the Protection of Users of Financial Services (CONDUSEF [Comisión Nacional para la Protección y Defensa de los Usuarios de Servi- cios Financieros]), and in particular Jesus Chavez Ugalde and Maria Fernanda Saldivar Cortes for their constant encouragement and support. We also thank the team at Ipsos for data collection. We are grateful to Santosh Anagol, David Medine, Claudia Ruiz, and Bilal Zia for helpful comments and to the Russia Financial Literacy and Education Trust Fund for funding. The views expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 247 248  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 2006). In a series of focus groups carried out as part of this research with low- to middle-income consumers in the Mexico City greater metropolitan area, many savers were not aware of various key terms of their accounts; those who were had only become aware of them when these had already negatively affected their account. Similarly, most borrowers did not understand key product terms of their consumer loans, such as total annual cost, and were unaware of possible fees for early and late payment of installments. This suggests, as evidenced in a burgeoning literature, that financial consumers do not necessarily choose the most cost-effective product or the one most suitable to their needs (see, e.g., Agarwal et al. 2013; Campbell et al. 2011; Choi, Laibson, and Madrian 2011; DellaVigna 2009; Duarte and Hastings 2011; Gross and Souleles 2002; and Hastings, Hortaçsu, and Syverson 2012 for reviews). Agarwal et al. (2013) find that if feedback is fast enough, then financial learning can occur, albeit the hard way: credit card fees paid by new cardholders fall during the first three years after account opening because, after paying a fee, consumers adjusted their behavior to avoid paying it in the future. More generally, mistakes in financial decision making would be less of a concern if market forces led to a set of relatively cheap financial products. However, this is rarely the case. Table 8.1 reports summary statistics for all credit and savings products offered by the 26 institutions that were part of the study targeted to low-income households. The total annual cost of a Mex$10,000 loan (US$748) ranges from 3.48 percent to 58.53 percent. The total annual yield for savings products with an initial deposit of Mex$5,000 (US$374) exhibits a similar pattern. The yield ranges from −31.80 percent to 6.17 percent for an investment account and from −100 percent to 5.12 percent for a checking account. This lack of transparency and high-cost environment are the result of a market failure created by the information asymmetry between less informed customers and better informed financial institutions and the misalignment of their incen- tives. Many governments around the world have tried to reduce the information asymmetry by introducing legislation to improve disclosure and transparency; in some cases, they have mandated that low-cost savings products be offered in the marketplace or have imposed usury laws capping the interest rate that can be charged on credit products. For example, in 2009 Mexico enacted the Law for Transparency and Regulation of Financial Services, which requires that consumers be presented with key financial terms for both savings and credit products, and standardizes calculation of certain key product terms and disclosure formats. In addition, as part of the Law on Credit Institutions of 2007 the Bank of Mexico requires that all deposit-taking institutions offer a basic account without account opening, deposit, withdrawal, balance inquiry, or debit card fees.1 Table 8.1 also reports the summary statistics for mandated basic savings accounts offered 1  by the institutions in the study. As expected, the mean in annualized yield is higher and 8.  Financial (dis-)information  ◾ 249 It remains unclear if these efforts have been successful in improving the financial outcomes of the population (Lowenstein, Sunstein, and Golman 2013). Indeed, the reason consumers may remain misinformed and prone to making mistakes is because financial institutions shroud prices, even in competitive envi- ronments, in order to maximize profits (Ausubel 1999; Gabaix and Laibson 2006). Firms will therefore adjust their behavior to undermine transparency initiatives (Duarte and Hastings 2011) and may not put much effort in marketing mandated products. This discussion brings to the fore two interrelated questions that are the focus of this chapter. First, what is the quality of information provided by financial institutions to low-income customers when choosing among credit and savings products? Second, do financial institutions offer the product that best meets the customer needs, in particular as it relates to cost and intended usage? To answer these questions, we implemented an audit study in peri-urban areas near Mexico City. Auditors visited the branches of financial institutions seeking to acquire a loan or a savings product. Since the goal was to capture all the information given to the auditor until the product was contracted (or the auditor was rejected), several follow-up visits were required, especially for credit auditors. Credit auditors therefore had to be local residents in each study area, since follow-up visits at the house were common. For these reasons, instead of using professional auditors, we recruited and trained credit auditors from the street. The scripts used by auditors were designed in collaboration with Mexico’s National Commission for the Protection of Users of Financial Services (CONDUSEF [Comisión Nacional para la Protección y Defensa de los Usuarios de Servicios Financieros]) and differed along four dimensions. First, we introduced variation specific to the product sought. Savings auditors expressed a preference for either a checking account or an investment account where funds would be deposited for a minimum duration of one year. Credit auditors requested a loan amount of either 20 percent or 70 percent of their household’s annual income, thus creating exogenous variation in the level of household indebtedness. Second, we varied the financial sophistication of the auditor made salient by the language used and the level of engagement during the visits. Third, we varied the level of competition among sophisticated auditors by stating that a competing institution had offered them better terms. Finally, we created variation in the dress code used during the visit. Each auditor was given a randomized list of branches to visit and was randomly assigned to a script. The model of Gabaix and Laibson (2006) provides a useful framework to understand the results from the audits. In the presence of enough uninformed variation is lower than that of checking accounts, and they do not vary by whether avoidable fees are included or not, since most of the basic accounts do not charge avoidable fees. 250  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES customers, their model predicts that firms in equilibrium will fail to advertise certain fees and commissions that uninformed customers will end up paying and informed customers will try to avoid. Thus, advertising will be deceptive rather than informative; in fact, firms will have no incentive to de-bias customers.2 Consistent with these predictions, we find that the staff provided enough information to allow auditors to apply for the loan or to open the savings account, but that very little information about avoidable fees and commissions was provided to neophyte auditors. In addition, the printed materials given to auditors of all profile types contained too little information to be useful for comparisons across products. Relatedly, while savings auditors were offered products that matched their preference for maturity, they were never offered the cheapest product that the institution could offer. The mandated basic account was only offered in 2 of the 54 visits in which the auditor expressed a preference for a checking account. By contrast, when faced with credit auditors requesting large amounts, finan- cial institutions demonstrated ability to assess borrowers’ capacity to pay both by being more likely to reject their loan application and by reducing the amount granted if the application was finally approved. Thus, firms seem to behave rationally by following their own self-interest, which may explain why the recent policy initiatives around product transparency and basic savings accounts have met with little success. The fact that finan- cial institutions will resist transparency initiatives and mandated products with low-cost designs is not new; in this sense, the chapter contributes to the recent literature that uses audit studies to assess financial advice for financial invest- ment (Mullainathan, Nöth, and Schoar 2010) and life insurance (Anagol, Cole, and Sarkar 2012). The remainder of the chapter is organized as follows. Section 8.2 describes the financial market for low-income households in Mexico. Section 8.3 reviews the predictions of Gabaix and Laibson (2006). Section 8.4 describes the experimental design, while section 8.5 describes the data and empirical strategy. Section 8.6 reports the results, and section 8.7 concludes. We note that if consumers were rational, then shrouding would actually hurt the finan- 2  cial institutions, as consumers faced with impartial disclosure of fees and commissions would assume they are high and purchase from banks that disclose more information. Shrouding behavior by banks would thus unravel, forcing banks to disclose all the infor- mation (Grossman 1981; Jovanovic 1982; Milgrom 1981). There is evidence suggesting that consumers tend to ignore not only information that is not disclosed (see Nisbett and Ross 1980 for an early review), but also disclosed information that is not made salient (Chetty, Looney, and Kroft 2009; Finkelstein 2009). 8.  Financial (dis-)information  ◾ 251 8.2 CONTEXT Financial markets in Mexico have recently been transformed by the appearance of new providers such as the consumer goods chains Elektra, Walmart, and Coppel, and new intermediation channels such as agent and mobile banking. This increase in the supply of financial services has coincided with a rapid expan- sion of household debt to gross domestic product from 8.7 percent in 2000 to 14.3 percent in 2011, with a recent study estimating debt service at 35 percent of household income (López Bolaños 2012). Among the components of household debt, consumer debt has been a key driver of this increase, rising from 7.7 percent in 1994 as a share of banking sector debt to 23 percent in 2011.3 Consumer credit in 2011 represented 40 percent of all bank earnings in interest, commissions, and annual fees. Much of this growth in consumer credit comes from bringing low-income individuals into the formal financial system for the first time. These individuals tend to have lower levels of education and may thus be worse equipped when deciding among financial products (Lusardi and Mitchell 2011). According to the 2008 Mexican Income and Expenditure Survey (ENIGH), 42 percent of households making a credit card payment had a head of household with less than high school education, and one in six had only primary educa- tion or less. The irregular (and sometimes unpredictable) income flows of many low-income individuals suggest that more flexible credit and savings products will be more suitable than others with stiff penalties for missed or partial install- ment payments or high minimum balance fees. In addition, products offered to low-income households tend to be relatively more expensive because they entail either smaller loan amounts or smaller savings balances that are more expensive to administer.4 For all these reasons, this study focuses on consumer credit and savings products targeted to low-income households. Savings and credit products in Mexico are offered by a range of financial institutions, including the new players that have a particular focus on low- and middle-income Mexicans. These include banks established by large retail chains, such as Banco Walmart, Banco Azteca (part of the Elektra Group, affiliated with the Elektra retail chain), Banco Coppel, and Banco Ahorro Famsa; as well as credit providers that have recently obtained banking licenses, such as Banco 3  According to a series of focus groups we conducted at the time of the study, low-income households mainly use credit to either finance an emergency or a durable asset. Using CONDUSEF’s online calculator, we find a total annual cost for a 12-month loan 4  of Mex$100,000 ranges from 19.3 percent to 154.4 percent; the total annual cost for a 12-month loan of Mex$10,000 ranges from 22.6 percent to 190.1 percent. Source: http://e- portalif.condusef.gob.mx/condusef_personalnomina/captura.php (accessed October 16, 2013). 252  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Compartamos. Besides these banks, the commercial banking sector also includes traditional banks such as HSBC, Banorte, Bancomer, and Banamex. Cooperative societies (SCAPs [sociedades cooperativas de ahorro y prés- tamo] or cajas de ahorro) are nonprofit, cooperative-based institutions that offer both credit and savings services. Popular financial societies (SOFIPOs [sociedades financieras populares]) also offer both credit and savings products, typically serving low-income consumer segments; unlike SCAPs, they are not considered nonprofit institutions. Multipurpose societies (SOFOMEs [sociedades financieras de objeto múltiple]) only offer credit products, as they are not permitted to collect deposits.5 SOFOMEs were nevertheless included in the savings audit to verify that they were indeed not accepting deposits.6 Institutions that offer savings products include banks, SCAPs, and SOFIPOs. All of these institutions offer both checking and fixed-term accounts with an initial deposit as low as Mex$1. Despite the diversity in institutions offering financial services, there are two related indications that the market is not competitive: lack of product information made available to the consumer, and high cost of credit and savings products. Focus groups conducted with low-income clients recruited from the study areas around the time of the audit study revealed that terms advertised in marketing materials and communicated by sales representatives did not accu- rately reflect the actual costs of the credit products they selected, making it hard to assess the true cost of a product before signing the contract. In addition, few of the participants understood the concepts of total annual cost and total annual earnings, and therefore did not use these terms in their decision to purchase credit or savings products. Consumers tended to focus instead on the periodic payment amounts (such as weekly or monthly installments), but these could lead to more expensive choices if the bulk of fees and commissions were not reflected in the installment amount. Participants also focused on the peso-value cost of the product, which was easier to understand than when the cost was expressed in percentages. Similarly, participants cited several cases where they had lost a significant fraction of their savings balance due to fees assessed on their accounts that were not known at the time when the savings account was opened. The fees respon- sible for the reductions in savings balances were mostly avoidable fees such as 5  SOFOMEs are divided between regulated and nonregulated SOFOMEs; the latter do not have to report financial information to the National Banking Commission, but must comply with consumer protection regulations set forth by CONDUSEF. While SOFOMEs accounted for only 1.3 percent of all complaints registered by consumers 6  with CONDUSEF in 2011, the number of cases leading to a fine being assessed were nearly twice as many for SOFOMEs as for any other provider type, despite their being a relatively small share of the market and only offering credit products (CONDUSEF 2012). 8.  Financial (dis-)information  ◾ 253 the monthly maintenance fee, minimum balance requirement fee, activity fee (charged when a certain number of withdrawals per month is exceeded), and inactivity fee (for not depositing additional funds over a certain period of time). From a policy perspective, mistakes in financial decision making due to limited information would be less of a concern if all financial products offered were rela- tively cheap. However, despite the recent entry of new players in the Mexican market, the cost of financial products in Mexico is relatively high compared to other countries, and it varies drastically depending on whether the fees just described are incurred. According to the Banco de México (2013), competition in consumer credit markets remains low, especially when compared to mortgages or business loans. Table 8.1 reports summary statistics for all credit and savings products that could have been offered by all 26 financial institutions in the audit study to clients between June and August 2012, the period in which the study was implemented. For consumer credit loans, table 8.1 reports the total annual cost of a Mex$10,000 (US$780) loan.7 For savings products, table 8.1 reports the total annual earnings from a Mex$5,000 (US$390) deposit in a term account and in a checking account. The table presents two different costs (or returns) depending on whether avoidable fees are incurred. The reason for presenting costs and returns this way is that information (or lack thereof) on avoidable fees is relevant, since behavior can be adjusted to avoid paying them. For example, a well-informed borrower could have borrowed from elsewhere to meet the installment payment obligation and avoid a high late payment fee; while a well-informed saver could have timed the deposits and withdrawals from the account so as to prevent the balance from falling below the minimum, thus avoiding the minimum balance fee. Thus, when the total cost (and return) includes only fees that must be incurred to contract the product (i.e., unavoidable fees), the cost refers essen- tially to the total annual cost and total annual earnings (equivalent to the U.S. annual percentage rate or annual percentage yield). However, since avoidable fees are only incurred if the client engages in certain behavior, the reported cost and returns inclusive of avoidable fees are computed using a hypothetical usage profile. For credit, we assume that the client misses one payment; for fixed term, we assume that the client withdraws the money before maturity (at six months); and for the checking account, we assume that the client maintains an average balance below the minimum for two months per year. According to table 8.1, the total cost of credit products offered by the 26 institutions that were included in the audit study was 24 percent without avoidable fees, and 26 percent if avoidable fees are also taken into account. Annualized yields for savings products exhibit a 7  Since auditors asked for either 20 percent or 70 percent of household income, the average loan size was Mex$18,388, and the median was Mex$13,500. 254  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 8.1  Cost and return of financial products N MEAN S.D. MIN P10 P25 P50 P75 P90 MAX Credit product Total annualized cost with unavoidable fees only 46 24% 15% 3.48% 9.20% 10.39% 21.33% 36.37% 45.37% 58.53% Total annualized cost inclusive of avoidable fees 46 26% 23% 3.48% 9.20% 10.39% 22.43% 36.37% 46.12% 138.04% Number of products per institution 46 2.88 3.12 1 1 1 2 3 13 13 Investment account Total annualized yield with unavoidable fees only 35 1.01% 10.00% −31.80% 0.38% 0.59% 2.66% 4.39% 5.12% 6.17% Total annualized yield inclusive of avoidable fees 35 0.39% 5.71% −63.60% 0.00% 0.21% 1.15% 2.10% 2.52% 3.04% Number of products per institution 35 3.63 2.52 1 2 2 2 4 8 8 Checking account Total annualized yield with unavoidable fees only 77 −7.36% 20.33% −100.00% −22.20% −6.24% 0.00% 0.25% 1.57% 5.12% Total annualized yield inclusive of avoidable fees 77 −9.10% 19.93% −100.00% −22.20% −8.40% −3.60% 0.00% 1.51% 5.12% Number of products per institution 77 8.90 5.92 1 2 4 7 12 18 18 Basic account Total annualized yield with unavoidable fees only 13 −0.77% 3.39% −12.00% 0.00% 0.00% 0.00% 0.00% 0.40% 1.36% Total annualized yield inclusive of avoidable fees 13 −0.86% 3.39% −12.00% −1.20% 0.00% 0.00% 0.00% 0.40% 1.36% Number of products per institution 13 1.15 0.38 1 1 1 1 1 2 2 Note: S.D. = standard deviation. An observation is a credit or savings product offered by 1 of the 26 institutions in the study. Unavoidable fees include the premium for compulsory unemploy- ment or life insurance, administrative commissions, credit bureau check and membership fees for credit products, and the opening fee and management fee for savings products; avoidable fees include a late installment payment for credit products, early withdrawal for investment accounts, and two months of balances below the minimum for checking and basic accounts. Information on product terms was collected by the National Commission for the Protection of Users of Financial Services (CONDUSEF) between June and August 2012, the period in which the audit visits were made. 8.  Financial (dis-)information  ◾ 255 similar pattern. Investment accounts yield on average 1.01 percent without avoid- able fees, and 0.39  percent if avoidable fees are included. Checking accounts have lower (negative) annualized yields of −7.36 percent without avoidable fees, or −9.10  percent on average if avoidable fees are taken into account. Consid- ering only government-mandated basic checking accounts, the average yield was −0.77 percent without any avoidable fees, and −0.86 percent with the avoidable fees. Table 8.1 suggests that small differences in behavior can have large impacts on the cost or return of a product, especially for credit and checking accounts; thus, accurate information on overall costs—in particular, avoidable fees—can save customers sizable amounts. The cost comparisons also demonstrate how, for most small-balance savers, the government-mandated account is cheaper than most checking accounts, and actually offers average returns that are only Mex$89 (approximately US$6.93) lower over a 12-month period without avoidable fees, and Mex$62 (approximately US$4.83) lower over a 6-month period with the avoidable fees than the average return of the investment accounts in the sample. The Gabaix and Laibson (2006) model to which we now turn can account for the variation in costs and yields reported in the data. It therefore provides a useful starting point to study the quality of information about financial products offered to prospective clients by the staff of financial institutions. 8.3 THEORY Recasting the example in Gabaix and Laibson (2006), imagine that a bank can offer a 2 percent deposit rate on a savings account so long as it can also charge a fee whenever the average monthly balance falls below a certain minimum, to break even.8 If the fee is not assessed, the institution can only offer a 1 percent deposit rate. Suppose that there are two types of customers, naïve and sophisti- cated. Naïve customers are not informed about the minimum balance fee (or do not understand it) and thus decide which account to open based on the highest deposit interest rate offered. In contrast, sophisticated customers know about the fee. Assume further that if customers do not take action, the timing of their deposits and withdrawals is such that the average balance will fall below the minimum, thus resulting in the minimum balance fee being assessed by the bank. Customers can, however, exert some effort to change the timing so that the balance never falls below the minimum. In this setup, banks will market accounts with a 2 percent deposit rate, failing to mention the minimum balance fee, to attract naïve customers. Given the 8  Note that this fee will only be assessed to the subset of customers whose average balance falls below the minimum. We assume that the bank knows this fraction of customers in the population and that, given the deposit rate offered, it can calibrate the fee to break even. 256  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES assumptions made, all naïve customers will end up paying the minimum balance fee, unaware of it. Sophisticated customers will also be attracted to the 2 percent deposit rate, but will never pay the minimum balance fee as they will take action to avoid it. Banks will therefore make enough money from naïve customers to cover the losses from offering the account to sophisticated consumers: naïve customers will cross-subsidize the sophisticated ones. Note that if a bank decided to price the savings account more transparently, offering savings accounts at 1 percent without the minimum balance fee, no customer would be attracted because it offers a lower interest rate. All customers would still demand the 2  percent — naïve customers failing to realize that they will end up paying savings account­ the minimum balance fee, and sophisticated ones realizing that they are better off earning 2 percent and avoiding the minimum balance fee altogether.9 The equi- librium is thus one in which financial products offered have hidden fees that are only taken into account by sophisticated consumers. More formally, the following testable implications can be derived: 1. Banks will not engage in informative marketing. Marketing campaigns, if any, will advertise the deposit rate but not the minimum balance fee for savings. Similarly, marketing campaigns for credit products will not adver- tise the total cost of the credit nor the detailed fees and charges. In other words, campaigns will be mostly persuasive. 2. Both naïve and sophisticated customers are aware of all basic require- ments, unavoidable fees, and other terms to open an account. For example, all customers know the interest rate earned on deposit products or the interest rate charged on credit products. 3. Naïve customers are unaware of avoidable fees and commissions such as the minimum balance fee on deposit accounts or the late payment fee on credit products. 4. Naïve and sophisticated customers are offered the same product, but naïve customers will earn less (if any) interest on savings (or pay more interest for credit) as they also pay fees and commissions. In section 8.6, we test these predictions more formally. Alternatively, naïve customers could understand the fee structure but be overly optimistic 9  about the chances of their savings balances falling below the minimum. In either case, they would prefer the 2 percent account over the 1 percent account offer. 8.  Financial (dis-)information  ◾ 257 8.4 EXPERIMENTAL DESIGN The audit study was conducted in collaboration with CONDUSEF, the government authority established in 1999 responsible for financial consumer protection and financial education programs. Four towns near the greater metropolitan region of Mexico City were selected with populations ranging from 30,000 to 1.6 million: Amecameca, Cuernavaca, Ecatepec, and Los Reyes de la Paz. These towns were selected for their high penetration of financial institutions targeting low- to middle-income consumers, and for their proportion of low- to middle-income population.10 From the town center, a radius of 1.5 kilometers was drawn, and every financial institution inside the circle was included in the study. We found a total of 26 distinct institutions, of which 13 were banks, 2 were SCAPs, 4 were SOFIPOs, and 7 were SOFOMEs. For the purposes of the analysis, we group the banks as either commercial banks (Scotiabank, HSBC, Banorte, Banamex, etc.) or “low-income” banks that target low-income households (Banco Ahorro Famsa, Bancoppel, Compartamos Banco, Banco Walmart, and Banco Azteca). Of the low-income banks, all but Banco Compartamos had their branches located inside the department stores of which the banks were a subsidiary. Credit auditors requested a six-month loan for a household expenditure, preferably with monthly installments. The expenditures included house repairs, medical expenses, and children’s school supplies, among others. Savings auditors were assigned a quantity of Mex$5,000 (US$378) that they wanted to deposit. Since we wanted to capture all the information that staff provided to prospec- tive clients up until the signing of the credit contract or opening of the savings account, a visit by the auditor was deemed completed when either the institution refused to open the account or grant the loan or when the auditor was asked to sign the contract. While most savings auditors needed only one visit to gather all the information and be offered the chance to open the account, credit audi- tors needed up to four follow-up visits (with an average of 1.87 visits), some at their residence, before they were offered the opportunity to sign the credit contract. Because credit auditors had to reside locally, we could not use profes- sional auditors. Instead, we recruited 18 auditors from low-income households (4 or 5 auditors in each town) who were trained by a full-time professional from a survey firm. There were 7 men and 11 women, with ages ranging from 21 to 53. Education levels were average for the population: 13 had secondary educa- tion, and 5 had undergraduate education. In terms of income- generation activi- ties, 7 were self-employed and 11 had a salaried job. Credit auditors made 215 visits, and a total 115 auditor-branch pairs since multiple visits were needed to Although there are considerable disparities in access by region and between rural and 10  urban populations, the towns selected are similar to other peri-urban towns in Mexico. 258  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES complete the loan application process. Savings auditors were professionals from the same survey firm, but they all also came from low-income households. Each auditor visited roughly seven different institutions in each town and collectively carried out a total of 112 visits. Since SOFOMEs cannot take deposits by law, we included them in the list of institutions to assess whether they complied with the law.11 Credit and savings visits took place between the months of March and June 2012. Given that a total of 26 different financial institutions were visited, a branch in a given site was always visited by more than one auditor. Each visit was recorded with a hidden digital recorder; after each visit, the auditor was required to complete a questionnaire that was validated with the audio recording. Immediately after recruitment, auditors were randomized into savings and credit scripts. To minimize confusion among credit auditors, a given auditor was assigned and trained on the same script that would be implemented in all of his or her visits. Savings auditors were professionals and thus alternated between different scripts. Auditors memorized the script(s) and used a uniform language when asking about the products. During the training, the auditors and manager were never told about the purpose of the study nor the specific hypotheses we wanted to test. The scripts varied along three dimensions: financial sophistication or literacy, the degree of perceived competition, and the dress code used in the visit. ◾◾ Financial literacy. Auditors were assigned to be either experienced or neophyte. Experienced auditors were trained to mention at the outset of the visit that they were shopping around for the best product and that they had reviewed product offers from other institutions. During the visit, they had to demonstrate knowledge of certain terms; if the staff did not voluntarily provide certain product information, they had to ask for it. For savings products, they had to ask for the minimum balance required to open the account, interest rate, total annual earnings, fees and commis- sions, and whether there was a linked debit card. For credit products, the experienced auditor had to ask about the total loan amount, total amount received, the term of the loan, the frequency and amount of each install- ment, other costs and fees, the interest rate, total annual cost, and the time lag for approval and disbursement. Neophyte auditors did not mention that they were shopping around nor that they had received other product offers. In sum, experienced shoppers explicitly asked for avoidable fees and commissions if the staff did not provide information voluntarily, while neophyte auditors did not. The reason for this treatment was to assess 11  The law regulating SOFOMEs is the Ley General de Organizaciones y Actividades Auxi- liares de Crédito, National Banking Commission of Mexico. 8.  Financial (dis-)information  ◾ 259 the amount of information provided by the staff either voluntarily or when prompted. ◾◾ Competition. Among experienced auditors, half mentioned that they were offered a lower interest rate in another institution, while the other half mentioned that they were offered a higher interest rate while shopping around at other institutions. The goal of this treatment was to assess the degree of flexibility of the staff in offering product terms that could match the competition. ◾◾ Dress code. Auditors were also instructed to alternate the way they dressed for each branch. In half of the branches that each shopper visited, he or she would dress formally (the attire for a special event or church); for the rest, he or she would dress informally. Since credit auditors could have follow-up visits to the same branch, they were instructed to dress formally or informally for all the visits to that particular branch.12 In addition, scripts differed in a product-based variation related to the savings needs or level of overindebtedness: ◾◾ Specific savings needs. Savings auditors wanted to make a deposit for either one full year, indicating a preference for an investment product; or to make frequent transactions, indicating a preference for a checking account. ◾◾ Overindebtedness. Credit auditors were assigned a loan amount resulting in a total household level of indebtedness of either 20 percent (low indebt- edness) or 70 percent (high indebtedness). This treatment amounted to loan sizes that ranged from Mex$1,000 to Mex$55,000. Auditors assigned to the high indebtedness script asked on average for Mex$10,350 more (US$774), an amount that is statistically significant at the 1 percent level. The reason for this treatment was to assess whether credit officers reduced the loan amount approved or whether they fudged the income numbers in the application to approve the loan (Berg, Puri, and Rocholl 2013). The combination of these variations resulted in six different scripts for credit and savings auditors.13 Although most shoppers followed this instruction, there are 13 cases in credit where the 12  shoppers alternated their dressing profile within the multiple visits to the same institution. For these cases, in the analysis we used the dressing profile treatment followed in the first visit. The six scripts used for credit auditors are: experienced, high competition, and high 13  indebtedness; experienced, high competition, and low indebtedness; experienced, low competition, and high indebtedness; experienced, low competition, and low indebtedness; neophyte and high indebtedness; and neophyte and low indebtedness. The six scripts used 260  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 8.5 DATA AND EMPIRICAL STRATEGY We use data from three different sources: ◾◾ Questionnaire. The questionnaire filled out by auditors after each visit covers various aspects of the visit, including the length of the visit, the products offered and their features, and the quality of service provided. ◾◾ Printed materials. The printed materials the staff provided to the auditor consist of either leaflets about the product or, less frequently, personal- ized amortization tables when credit was requested. For credit, out of the 54 printed materials given to the auditors, 49 were advertisements and generic leaflets; only 5 were personalized amortization tables. ◾◾ CONDUSEF data set. This includes the terms of all the savings and credit products offered by the institutions at the time of the study. Many of the terms were not available online and had to be requested directly from the institution. The data set only includes credit products with loan sizes in the range requested by the auditors and savings products that can be contracted with an initial deposit of Mex$5,000. These data were used to generate table 8.1 and to match the products that were actually offered to the auditors to compare their cost against the rest of products offered by the institution. We use data from the questionnaire, printed materials, and CONDUSEF to construct variables related to the amount of information provided by the institu- tions. We divide the information into the following categories: legal requirements to open the account or take out the loan, terms of the account, and fees and commissions—most of which are avoidable, as they are only assessed depending on usage. The items of information the staff could provide for each category vary by product. For savings products, legal requirement items include the need to show an official identification number and an address certificate. The terms of account items include whether the client needs to be a member of the institution to open an account, the minimum balance required to open an account, maturity of the account, whether a debit card is provided, the interest rate offered and whether it is fixed or variable, total annual earnings, deposit insurance tax, and tax on earnings. for savings auditors are similar, replacing high indebtedness with preference for invest- ment account and low indebtedness with preference for checking account. As mentioned, each credit auditor was assigned one script, while savings auditors were assigned as either experienced or neophyte, and to have a preference for either an investment account or a checking account; if they were assigned the experienced profile, they alternated between high and low competition scripts. 8.  Financial (dis-)information  ◾ 261 Fees and commissions include fee if the average balance falls below the minimum balance, inactivity fee, early withdrawal fee for investment accounts, withdrawal fee, balance inquiry fee, and debit card reposition fees. In sum, the legal requirements and the terms of the account include requirements and unavoidable costs to open the account, while fees and commissions are addi- tional costs that can be avoided through usage behavior. For credit products, the legal requirements additionally include an income certificate. The terms of the account include the need to be a member/client of the institution, save before acquiring a loan, have a checking account in the institution, amount lent, amount disbursed, total amount due, amount of each installment, value added tax, interest rate and whether it is fixed or variable, total annual cost, maturity, frequency of payments, and number of installments. The fees and commissions include premium for unemployment and/or life insurance, administrative commissions, cost of the credit bureau check, membership fee, early payment fee, late payment fee, and fee for payment of installments in other branches or institutions. Similar to the savings categories, the legal requirements and terms of the account include requirements and unavoidable costs to obtain the loan, while fees and commissions are additional costs that can mostly be avoided through usage behavior. There are therefore up to 16 items of information for savings products and 23 items for credit products we measured that could potentially be provided to auditors by sales staff. Clearly, not all are equally relevant or important. For this reason, we construct two sets of measures for each product offered. The first simply involves counting the items of information in each category. For example, if the staff mentioned that a savings auditor needed to show an official identifica- tion number and an address certificate, the variable would take the value 2 (for the possible two items). Because certain items of information may be irrelevant if they are not required or do not apply, we also construct a set of variables that reflect the percentage of items of information that are present in the product offered by the staff. The numerator in this variable is the number of items mentioned (among those that are present in the product), and the denominator is the number of items either required or that apply. For example, if a product has a withdrawal fee, then the item will be included in the denominator for the fees and commission category and will appear in the numerator if the staff mentioned the fee.14 In addition to the information provided, we assess the cost of the product offered and compare it to similar products offered by the institution and the market. As with the costs reported in table 8.1, we compute the total costs and earnings with unavoidable fees only, using the formulas for total annual cost and In this example, if there was no withdrawal fee, it would not appear in the numerator or 14  denominator, although the staff could have mentioned that such a fee was not assessed. 262  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES total annual earnings, and the total costs including the most commonly assessed avoidable fees. In particular, we assume that credit clients will miss one payment and will thus be assessed a late payment fee and that savings clients will either withdraw early for those contracting an investment account or will maintain a balance for two months below the minimum.15 Because the assignment of scripts to auditors was random, we can run the following ordinary least squares regres- sion for savings visits: yij = β1HLi * HIi + β2HLi * LIi + β3 INVi + β4 Fi + Lj + εij (8.1) where yij is the outcome of interest for auditor i visiting institution j, HLi is an indicator for high literacy, HIi is an indicator for high interest rate (high compe- tition), and LIi is an indicator for low interest rate (low competition). In addition, INVi is a dummy that takes the value of 1 if the auditor has a preference for an investment account or fixed-term deposit (0 if savings/checking account), and the dummy Fi takes the value of 1 if the auditor was dressed formally during the visit. The variable εij is a mean zero error term. The regression specification also includes dummies for each type of financial institution (Lj). Robust standard errors are included.16 For credit auditor-branch pairs, a similar specification is used. Note that if the pair involved more than one visit, the outcomes are aggregated up to the pair level. We include a dummy that takes the value of 1 if the auditor asked for a loan representing 70 percent of household income (0 for a loan representing 20 percent of household income). In this specification, we include a vector of auditor char- acteristics, because given the relatively small number of credit auditors, there are some imbalances in the characteristics of auditors assigned to the different scripts. As the table in the annex to this chapter shows, there is balance in the assignment of the financial literacy profile (experienced and neophyte) with high and low competition ( p-value of an F-test that all auditor characteristics included are jointly zero is .42 and .97, respectively), but not in the indebtedness profile ( p-value of the F-test that all variables are jointly zero is 0.01). Overindebted audi- tors tend to be older women with a regular salary and to be more likely to have an outstanding loan. As a result, we include as controls the variables that cause the imbalance—namely, gender, age, whether the auditor is self-employed, and whether he or she is currently a borrower (Bruhn and McKenzie 2009). We there- fore run the following specification: 15  These usage profiles come from discussions with CONDUSEF and focus groups about the average behavior of low-income consumers with personal credit products and savings accounts. Notice that other auditor individual characteristics were not included as controls 16  because, with only four auditors, some of these characteristics were near collinear with the treatment assigned. 8.  Financial (dis-)information  ◾ 263 yij = β1HLi * HIi + β2HLi * LIi + β3 HDi + β4 Fi + Lj + Xi + εij (8.2) In this specification, the indicator for high interest rate HIi (low interest rate LIi) denotes low (high) competition because a credit at lower cost is more attrac- tive. As before, the regression also includes dummies for each type of financial institution (Lj). 8.6 EMPIRICAL RESULTS This section presents the results. Columns 1 and 2 of table 8.2a report, respec- tively, the average waiting time and the interview time or face-to-face time of savings auditors with the staff. For auditors dressed informally with low experi- ence and a preference for a checking account (dubbed “control” auditors because all treatment dummies take value zero), each visit lasted on average 18.5 minutes, including 10.2 minutes of face-to-face interaction with the staff. Auditors visiting commercial banks had to wait on average 18.5 minutes, while the wait time in the other institutions was approximately 7 minutes. Commercial banks and SOFIPOs had the shortest face-to-face time (approximately 11.5 minutes), while low-income banks and SCAPs had longer interview times (14.5 and 25 minutes, respectively). Experienced auditors reported significantly longer interview times relative to neophyte auditors by 6.3 and 8.3 minutes, depending on the interest rate quoted from other institutions. This difference is, however, not significant, suggesting that the extent of competition does not affect how long staff spends with the auditor. We also do not find differences in interview time between audi- tors seeking an investment account compared to those interested in a checking account. Column 3 reports the number of different savings products offered by the staff. Commercial banks and SOFIPOs offered the least number of products. Similarly, auditors seeking an investment account are offered fewer products, simply because the institutions have less distinct investment products. According to the CONDUSEF data set, of the 125 savings products, only 35 percent are investment products. Experienced auditors are not provided with more choices. We note that the number of observations drops from 112 to 107 because all the auditors who were assigned to visit SOFOMEs were turned away empty handed, suggesting that SOFOMEs do comply with the law. Interestingly, while in 92 percent of the visits, the auditor was offered a product that aligned with his or her preferences (column 4), in only 2 of the 54 visits where auditors sought a checking account did the staff offer the mandated basic account. In sum, expe- rienced auditors spent more time with the staff, but were not offered more or cheaper products. Table 8.2b reports similar results to those of table 8.2a for credit audits. Since an observation is now an auditor-branch pair that may involve more than one visit, columns 1 and 2 report the sum of each visit’s length in minutes. For control auditors, now defined as informally dressed, with low experience and low 264  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 8.2A  Interview time, number of products offered, and product alignment with needs for savings products PRODUCT TOTAL OFFERED OFFERED WAIT TIME INTERVIEW MORE THAN ALIGNED WITH (MINS) TIME (MINS) ONE PRODUCT NEEDS (1) (2) (3) (4) High experience X high interest 1.920 6.309*** 0.109 (0.110) (3.862) (1.999) (0.089) (0.075) High experience X low interest −2.055 8.306*** 0.125 −0.014 (2.205) (2.327) (0.102) (0.059) High Indebtedness (1 = yes) −0.441 1.073 −0.466*** −0.027 (2.419) (1.603) (0.080) (0.056) Formal dress (1 = yes) 0.126 −0.689 −0.234*** −0.05 (1.906) (1.484) (0.087) (0.051) Observations 112 110 107 107 R-squared 0.473 0.755 0.707 0.923 Mean of dependent variable 8.3 10.2 1.00 0.92 (among controls) p -value of F-test: high experience 0.501 0.000 0.334 0.340 p -value of t-test: high interest– 0.322 0.482 0.882 0.231 low interest Mean Commercial banks 18.5 11.7 0.47 0.82 Low-income banks 7.5 14.5 0.61 0.94 SCAP 7.0 25.0 1.00 1.00 SOFIPO 6.6 11.3 0.26 1.00 p -value of t-test Commercial banks = low-income 0.001 0.213 0.206 0.099 banks Commercial banks = SCAP 0.196 0.000 0.027 0.320 Commercial banks = SOFIPO 0.017 0.842 0.144 0.053 Low-income banks = SCAP 0.867 0.044 0.087 0.578 Low-income banks = SOFIPO 0.597 0.254 0.009 0.277 SCAP = SOFIPO 0.895 0.000 0.002 — Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SCAP, and SOFIPO). Control auditors were assigned the script with low experience (neophyte), a preference for a checking account, and were dressed informally. 8.  Financial (dis-)information  ◾ 265 indebtedness, the average wait time was around 6 minutes and the total time in face-to-face interactions throughout all visits was 33.1 minutes. Similar to the savings audits, experienced auditors had more face-to-face time with the staff, especially for those who mentioned a higher interest rate offer from another insti- tution.17 Using nonexperimental variation, we find that shoppers with an under- graduate degree have about 10 more minutes of face-to-face interaction with the staff. Again, commercial banks had more wait time and less interview time compared to the rest of the institutions. Column 3 reports the number of visits in each auditor-branch pair. Control auditors had on average 2.2 visits, 0.6 of which were at home or at the place of business. None of the treatment conditions affect the number of visits, although auditors in commercial banks had significantly less visits. Auditors with a busi- ness, however, had on average 0.46 more visits—perhaps to verify that the business did exist and that the income reported could be generated from its reve- nues. In column 4, we report the probability that the loan application was rejected by the lender. On average, 70 percent of the loan applications were rejected. Interestingly, auditors with the high indebtedness profile—that is, auditors who requested amounts that brought household indebtedness up to 70 percent of household income—were 47 percent more likely to be rejected. This is particu- larly true for the commercial banks compared to the rest of the institutions. Using nonexperimental variation, we find that female and educated auditors as well as those with a loan were less likely to be rejected, while auditors with a business were more likely to be rejected. In addition, among auditors who were offered a loan amount, those with the high indebtedness profile were approved amounts significantly smaller than those requested (column 5).18 Commercial banks did provide, on average, larger loan sizes compared to other institutions. Thus, it seems banks are willing to give larger loans, but may have more stringent approval processes. In sum, lenders seem to adjust their approval decisions and loan amounts rationally in response to the risk profile of prospective clients to mitigate their exposure. In tables 8.3a and 8.3b, we explore whether the information provided to audi- tors conforms to the predictions of the Gabaix and Laibson (2006) model reviewed in section 8.2. It is not clear a priori whether the extent of competition should play a role in the length 17  of the interview because, when faced with an auditor who has an alternative offer with a higher interest rate, say, the staff could spend more (or less) time convincing the auditor about the better product offered in the institution. The number of observations drops from 115 to 88 because 27 loan applications were 18  rejected without an amount being offered. Among the 88 loan offers, 54 were ultimately rejected. Results from the regression in column 5 using the 34 loans that were approved are similar to those reported, except that experienced auditors with the low interest rate (high competition) script were offered larger loans. 266  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 8.2B  Interview time, number of visits, and approval rate for credit products TOTAL DIFF. BET. WAIT TIME INTERVIEW TOTAL # OF DENIED AMTS (MINS) TIME (MINS) VISITS CREDIT OFFERED & (1) (2) (3) (4) REQUESTED High experience X high −2.189 18.862** −0.099 0.124 −2,124.7 interest (3.090) (7.534) (0.183) (0.135) (2,262.8) High experience X low −0.415 3.652 −0.085 0.121 −2,168.8 interest (3.433) (5.550) (0.165) (0.140) (2,133.1) High Indebtedness (1 = yes) −0.404 12.209 0.369 0.467* −6,858.3* (4.511) (9.902) (0.340) (0.241) (3,535.5) Formal dress (1 = yes) 3.486* 6.005 −0.057 −0.100 1,139.2 (1.870) (3.897) (0.118) (0.089) (1,404.4) Observations 115 115 115 115 88 R-squared 0.45 0.749 0.923 0.749 0.391 Mean of dependent variable 6.0 33.1 2.2 0.70 −944.0 (among controls) p -value of F-test: high expe- 0.754 0.046 0.764 0.539 0.532 rience p -value of t-test: high 0.596 0.084 0.954 0.988 0.984 interest– low interest Mean Commercial banks 11.9 16.2 1.2 0.84 1,500.0 Low-income banks 8.9 30.9 2.0 0.59 −2,545.0 SCAP 3.6 40.4 2.0 0.65 −2,900.0 SOFIPO 6.1 32.4 2.1 0.65 −3,368.0 p -value of t-test Commercial banks = low-in- 0.367 0.004 0.000 0.056 0.090 come banks Commercial banks = SCAP 0.008 0.001 0.000 0.134 0.163 Commercial banks = SOFIPO 0.103 0.006 0.000 0.179 0.120 Low-income banks = SCAP 0.031 0.092 1.000 0.603 0.829 Low-income banks = SOFIPO 0.341 0.784 0.827 0.660 0.646 SCAP = SOFIPO 0.312 0.266 0.811 0.992 0.828 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SOFOME, and SOFIPO) and the following auditor characteristics: gender, age, whether the auditor is self-employed and whether he or she is currently a borrower. Control auditors were assigned the script with low experience (neophyte), a preference for a loan size of 20% of household income, and were dressed informally. 8.  Financial (dis-)information  ◾ 267 Column 1 of table 8.3a reports the number of printed materials (generic leaf- lets or personalized information) that auditors were given by staff. The average number of materials among control savings auditors was less than one. In fact, only about 67 percent of auditors were given any material, and among those who were, the mode was one leaflet. Perhaps more interesting, column 2 confirms hypothesis 1 in section 8.2, which states that financial institutions do not engage in informative marketing. Indeed, the printed materials from the institutions do not provide factual and detailed information to help customers choose among different financial products. An example of one such printed material for a credit product is shown in figure 8.1. Column 2 of table 8.3a reports that printed mate- rials contained one item of information (out of 17 items in total as described above). While experienced auditors received materials with an additional two items, especially if they reported an alternative inferior interest rate offer, the average level of information that printed materials provide is very low and certainly insufficient to make informed decisions about the suitability and cost of financial products. Columns 1 and 2 of table 8.3b echo the results of table 8.3a for credit audits. In only 30 percent of the credit audits (22 out of 88) did auditors receive any printed material. While experienced auditors received 2 or 3.4 more items of information (depending on the interest rate mentioned) than neophyte auditors who only received 1.5 items, in neither case is the information received sufficient for making informed decisions. We note that by law all printed materials must FIGURE 8.1  Sample marketing materials for consumer credit collected during auditor visits 268  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 8.3A  Printed and oral information for savings products PRINTED INFORMATION ORAL INFORMATION TOTAL # OF PRINTED TOTAL TOTAL LEGAL TERMS OF FEES AND MATERIALS INFORMATION INFORMATION REQUIREMENTS ACCOUNT COMMISSIONS (1) (2) (3) (4) (5) (6) High experience X high interest 0.121 2.450*** 5.936*** 0.013 1.546*** 4.377*** (0.322) (0.466) (0.464) (0.140) (0.340) (0.195) High experience X low interest −0.077 2.236*** 6.074*** −0.004 1.701*** 4.377*** (0.286) (0.440) (0.416) (0.118) (0.293) (0.223) High Indebtedness (1 = yes) 0.365 0.197 −0.567* 0.053 0.108 −0.728*** (0.242) (0.343) (0.336) (0.106) (0.246) (0.158) Formal dress (1 = yes) −0.359 −0.740** −0.871** 0.036 −0.816*** −0.09 (0.250) (0.371) (0.342) (0.116) (0.262) (0.163) Observations 107 107 107 107 107 107 R-squared 0.460 0.944 0.972 0.892 0.948 0.950 Mean of dependent variable (among controls) 0.8 5.8 7.1 1.5 4.6 1.0 p -value of F-test: high experience 0.871 0.000 0.000 0.992 0.000 0.000 p -value of t-test: high interest–low interest 0.601 0.719 0.804 0.903 0.700 0.999 Mean of dependent variable by institution Commercial banks 0.9 7.1 9.4 1.4 5.2 2.8 Low-income banks 1.3 6.6 9.1 1.5 5.0 2.6 SCAP 0.5 6.2 9.4 1.4 5.2 2.8 SOFIPO 1.1 7.1 9.4 2.0 4.8 2.6 p -value of t-test Commercial banks = low-income banks 0.230 0.262 0.686 0.601 0.626 0.682 Commercial banks = SCAP 0.387 0.398 0.994 0.955 0.993 0.984 Commercial banks = SOFIPO 0.752 0.942 0.968 0.000 0.427 0.730 Low-income banks = SCAP 0.171 0.717 0.853 0.857 0.822 0.867 Low-income banks = SOFIPO 0.543 0.344 0.783 0.000 0.649 0.960 SCAP = SOFIPO 0.415 0.484 0.988 0.020 0.669 0.876 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Legal requirements in column 4 include whether the officer requested an official identification and an address certificate. In column 5, terms of account include need to be member client to open account, minimum balance required to open account, maturity, debit card, interest rate and whether it is fixed or variable, total annual earnings, deposit insurance tax, and tax on earnings. In column 6, fees and commissions include opening fee, management fee, fee if minimum balance not maintained, inactivity fee, early withdrawal fee (for investment accounts), withdrawal fee, balance inquiry fee, and debit card reposition fee. Quan- tity of information aggregates legal requirements, terms of account, and fees. All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SCAP, and SOFIPO). Control auditors were assigned the script with low experience (neophyte), a preference for a checking account, and were dressed informally. 269  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 8.3B  Printed and oral information for credit products PRINTED INFORMATION ORAL INFORMATION TOTAL # OF PRINTED TOTAL TOTAL LEGAL TERMS OF FEES AND MATERIALS INFORMATION INFORMATION REQUIREMENTS ACCOUNT COMMISSIONS (1) (2) (3) (4) (5) (6) High experience X high interest −0.276 3.368* 3.178*** 0.3 0.186 2.692*** (0.322) (1.692) (1.002) (0.255) (0.654) (0.570) High experience X low interest 0.364 2.060* 4.135*** 0.558*** 1.131** 2.447*** (0.270) (1.114) (0.853) (0.191) (0.516) (0.523) High Indebtedness (1 = yes) 0.356 1.471 −0.593 −0.58 −1.384 1.371* (0.410) (0.998) (1.462) (0.364) (0.996) (0.801) Formal dress (1 = yes) −0.197 −0.32 −0.841 −0.069 −0.588 −0.184 (0.181) (0.714) (0.614) (0.124) (0.354) (0.379) Observations 88 22 88 88 88 88 R-squared 0.567 0.817 0.882 0.911 0.807 0.696 Mean of dependent variable (among controls) 0.9 1.5 4.7 1.8 2.7 0.2 p -value of F-test: high experience 0.184 0.157 0.000 0.009 0.084 0.000 p -value of t-test: high interest– low interest 0.078 0.352 0.371 0.422 0.160 0.689 Mean of dependent variable by institution Commercial banks 0.8 1.33 3.5 1.0 1.8 0.7 Low-income banks 0.7 1.20 5.8 1.8 2.5 1.5 SCAP 0.8 1.50 6.5 1.8 3.3 1.4 SOFIPO 0.5 1.50 6.2 1.6 2.7 1.8 p -value of t-test Commercial banks = low-income banks 0.755 0.855 0.067 0.006 0.306 0.342 Commercial banks = SOFOME 1.000 0.833 0.061 0.010 0.137 0.353 Commercial banks = SOFIPO 0.194 0.846 0.101 0.135 0.313 0.157 Low-income banks = SOFOME 0.640 0.622 0.395 0.865 0.100 0.917 Low-income banks = SOFIPO 0.203 0.638 0.673 0.208 0.627 0.526 SOFOME = SOFIPO 0.161 1.000 0.766 0.278 0.378 0.462 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Legal requirements in column 4 include whether the officer requested offi- cial identification and address certificate. Terms of account in column 5 include total payment, amount of each installment, interest rate, total annual cost, maturity, frequency of payment, and number of installments. Fees and commissions in column 6 include unemployment insurance, life insurance, administrative commissions, credit bureau check, membership fee, anticipated payment fee, late payment fee, and payment in other institutions. Total information adds the items in each category (legal requirements, terms of account, and fees and commissions). All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SOFOME, and SOFIPO) and the following auditor characteristics: gender, age, whether the auditor is self-employed, and whether he or she is currently a borrower. Control auditors were assigned the script with low experience (neophyte), a preference for a loan size of 20 percent of household income, and were dressed informally. 270  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES contain the total annual earnings or total annual cost, as the examples in figure 8.1 show. The formulas for computing the total annual earnings and total annual cost, however, do not take into account avoidable fees and commissions—which can be relevant when choosing the cheapest product. In addition, the total annual cost and total annual earnings are never made salient, as they are typically buried in a paragraph with other disclosures printed in small type. Columns 3–6 in tables 8.3a and 8.3b test hypotheses 2 and 3 of section 8.2, stating that all auditors should be equally informed about the requirements and unavoidable costs in opening a savings account or contracting a loan and that only experienced auditors would be aware of avoidable costs. Column 3 reports all information items in one aggregate variable called “Total information,” while columns 4–6 separate the information provided by the staff into the categories discussed in the previous section: legal requirements (column 4), terms of the account (column 5), and fees and commissions (column 6). According to column 4 in table 8.3a, staff provided oral information to control auditors on 7.1 of 22 items (1.5 legal requirements, 4.6 term items, and 1 fee item). Experienced auditors were provided on average six more items of informa- tion (an increase of 80 percent in number of items). Inspecting columns 4–6, we can see that the bulk of the increase comes from fees and commissions, which are all avoidable. Indeed, there are no differences among auditors on the legal requirements needed to open an account. Control auditors are provided 1.5 of 2 items of information on average. There are also few differences in the informa- tion provided to experienced and neophyte auditors. Neophyte (control) auditors receive 4.6 items of information, while experienced auditors receive 1.6 more items of information—an increase of only 13.5 percent (column 5). In contrast, column 6 reports that neophyte auditors only receive information about one fee and commission, while experienced auditors receive information on 4.4 fees and commissions—more than a fourfold increase. Table 8.3a thus confirms hypoth- eses 2 and 3 because all auditors receive roughly the same information needed to open an account, but only experienced auditors learn about the avoidable fees and commissions when they ask about them. We note that auditors seeking investment products were provided less information on fees simply because these products tend to have fewer fees. Also noteworthy is the fact that all insti- tutions tend to report the same amount of information. Table 8.3b echoes the results of table 8.3a. Experienced auditors are given 3.2 and 4.1 more items of information depending on the interest rate quoted from a competing institution. This increase in information does not concern legal requirements or terms of the account but rather fees and commissions. Indeed, control auditors are given information on 0.2 fees and commissions; experienced auditors are given information on 2.7 and 2.5 fees and commissions, depending on the interest rate quoted. Again, all auditors are provided enough informa- tion to be able to apply for the loan, but only experienced auditors are provided information about (mostly avoidable) fees and commissions. We also note that 8.  Financial (dis-)information  ◾ 271 commercial banks tend to give less information to all auditors, probably because staff is more likely to reject applications and after the rejection no further infor- mation was provided. Columns 1–4 of tables 8.4a and 8.4b complement columns 3–6 of tables 8.3a and 8.3b in that they report the percentage of items in each category that are relevant for the products offered. Put differently, if, for example, the savings account offered has only one legal requirement (instead of the maximum two), four relevant terms instead of the maximum eight, and only fees fees instead of the maximum six, the percentage of items of information is constructed with the number of items mentioned by the staff among those in the product. Continuing with the example, if the staff mentions the legal requirement needed to open the savings account, then column 2 would record 100 percent; if three out of the four fees are mentioned, then column 4 would report 75 percent. In computing these percentages, information provided by the staff about, say, fees that are not assessed or requirements that are not imposed is ignored. Columns 2 and 3 of table 8.4a shows that control auditors are provided most of the information needed to open an account (70 percent of the legal requirements and 50 percent of the relevant terms of the account), but are barely provided with information about avoidable fees and commissions (only 11 percent). In addition, experienced auditors are provided a threefold increase in information about fees and commissions (column 4). These two findings again confirm hypotheses 2 and 3. Columns 1–4 of table 8.4b are again consistent with the findings in table 8.4a. Control credit auditors appear to be more informed about the fewer fees and commissions of credit products than their savings counterparts, but again expe- rienced credit auditors are more informed about fees and commissions. In fact, they now learn about all of them. Columns 5 and 6 of table 8.4a report the total yield after one year that would accrue in the savings account under the assumptions of “no usage” (column 5) and “usage” (column 6). No usage refers to no deposits and withdrawals being made for one year. In this case, without inactivity fees, the formula to calcu- late the total yield coincides with the total annual earnings formula. In column 6, usage refers to the situation where the average balance in the account is below the minimum allowed for two months. Columns 5 and 6 can be used to test hypothesis 4 in section 8.2, namely that all customers will be offered similar products. Indeed, control auditors are offered accounts with a small negative total yield, and experienced auditors are given roughly the same accounts. As expected, auditors with a preference for investment accounts are offered higher yields, so auditors are offered different products based on their preference but not based on their financial literacy. When usage fees are taken into account, it appears that experienced auditors (espe- cially those who quote a low interest rate) are offered higher-yielding accounts. Columns 7 and 8 report the difference in total yields between the cheapest savings product that meets the needs of the customer that the institution could 272  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 8.4A  Quality of information and return on savings DIFF. IN TOTAL INFOR- RE- COM- TOTAL YIELD YIELD MA- QUIRE- AC- MIS- NO US- NO TION MENTS COUNT SIONS AGE USAGEa USAGE USAGEa (1) (2) (3) (4) (5) (6) (7) (8) High experience X high 0.191*** 0.09 0.081 0.362*** 0.001 −0.004 0.001 0.005 interest (0.026) (0.087) (0.064) (0.023) (0.006) (0.005) (0.008) (0.007) High experience X low 0.178*** −0.035 0.129** 0.339*** 0.009 0.014** −0.011 −0.016** interest (0.019) (0.069) (0.060) (0.019) (0.007) (0.007) (0.008) (0.008) High Indebtedness (1 = −0.015 0.026 0.019 −0.053*** 0.044*** 0.027*** −0.009 −0.017*** yes) (0.017) (0.073) (0.054) (0.018) (0.005) (0.005) (0.006) (0.006) Formal dress (1 = yes) −0.032* 0.034 −0.101* 0.006 −0.002 (0.009) 0.002 0.006 (0.018) (0.072) (0.052) (0.018) (0.008) (0.007) (0.008) (0.008) Observations 80 80 80 80 80 80 80 80 R-squared 0.972 0.893 0.907 0.938 0.627 0.553 0.331 0.525 Mean of dependent vari- 0.33 0.73 0.51 0.11 −0.007 −0.011 0.027 0.030 able (among controls) p -value of F-test: high 0.000 0.381 0.065 0.000 0.452 0.036 0.334 0.046 experience p -value of t-test: high 0.642 0.167 0.571 0.369 0.305 0.010 0.195 0.015 interest–low interest Mean Commercial banks 0.41 0.74 0.63 0.24 −0.011 −0.028 0.034 0.046 Low-income banks 0.39 0.73 0.57 0.22 0.028 0.013 0.008 0.006 SCAP 0.44 0.67 0.62 0.29 0.029 0.029 0.000 0.000 SOFIPO 0.43 1.00 0.70 0.22 0.037 0.018 0.001 0.000 p -value of t-test Commercial banks = 0.516 0.871 0.232 0.688 0.000 0.000 0.003 0.000 low-income banks Commercial banks = 0.701 0.723 0.913 0.648 0.178 0.046 0.286 0.144 SCAP Commercial banks = 0.673 0.010 0.404 0.752 0.005 0.004 0.059 0.010 SOFIPO Low-income banks = 0.519 0.731 0.690 0.531 0.942 0.071 0.310 0.326 SCAP Low-income banks = 0.372 0.003 0.091 0.950 0.287 0.277 0.098 0.095 SOFIPO SCAP = SOFIPO 0.923 0.064 0.610 0.637 0.309 0.021 0.319 0.319 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The percentages in columns 1–4 are computed with the same information in each category in table 8.3a in the numerator and the number of requirements, terms, and fees reported to CONDUSEF. All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SCAP, and SOFIPO). Control auditors were assigned the script with low experience (neophyte), a preference for a checking account, and were dressed informally. a. For investment products, usage includes fee and penalties incurred in an early withdrawal. For checking accounts, usage includes fees and commissions paid for keeping a balance below the minimum balance allowed (if it exists) for two months. 8.  Financial (dis-)information  ◾ 273 TABLE 8.4B  Quality of information and cost of credit TOTAL COST TOTAL LEGAL FEES AND ON-TIME INFORMA- REQUIRE- TERMS OF COMMIS- REPAY- ONE LATE TION MENTS ACCOUNT SIONS MENT PAYMENTa (1) (2) (3) (4) (5) (6) High experience X high 0.268*** 0.150 −0.048 1.154*** −0.284 −0.142 interest (0.083) (0.150) (0.061) (0.352) (0.189) (0.318) High experience X low 0.308*** 0.259** 0.073 1.229*** 0.109 0.009 interest (0.065) (0.115) (0.049) (0.310) (0.199) (0.390) High Indebtedness (1 −0.108 −0.262 −0.106 0.167 −0.583 −0.538 = yes) (0.124) (0.241) (0.096) (0.563) (0.429) (0.755) Formal dress (1 = yes) −0.004 0.065 −0.027 −0.431* −0.009 −0.092 (0.043) (0.077) (0.030) (0.220) (0.145) (0.234) Observations 56 53 53 56 56 56 R-squared 0.905 0.921 0.84 0.808 0.89 0.777 Mean of dependent vari- 0.27 0.80 0.20 0.40 1.20 1.43 able (among controls) p -value of F-test: high 0.000 0.028 0.107 0.000 0.280 0.895 experience p -value of t-test: high 0.654 0.609 0.049 0.845 0.151 0.797 interest–low interest Mean Commercial banks 0.28 0.33 0.20 0.88 0.30 0.32 Low-income banks 0.33 0.92 0.16 1.21 0.49 0.65 SOFOME 0.42 0.80 0.25 0.79 1.74 1.86 SOFIPO 0.42 0.82 0.16 1.03 1.02 1.39 p -value of t-test Commercial banks = 0.627 0.007 0.605 0.592 0.000 0.000 low-income banks Commercial banks = 0.326 0.057 0.593 0.887 0.001 0.006 SOFOME Commercial banks = 0.177 0.036 0.570 0.698 0.006 0.096 SOFIPO Low-income banks = 0.253 0.251 0.066 0.288 0.000 0.000 SOFOME Low-income banks = 0.154 0.333 0.915 0.588 0.000 0.011 SOFIPO SOFOME = SOFIPO 0.983 0.831 0.061 0.464 0.002 0.223 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The percentages in columns 1–4 are computed with the same information in each category in table 8.3a in the numerator and the number of requirements, terms, and fees reported to CONDUSEF. All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SOFOME, and SOFIPO) and the following auditor characteristics: gender, age, whether the auditor is self-employed, and whether he or she is currently a borrower. Con- trol auditors were assigned the script with low experience (neophyte), a preference for a loan size of 20% of household income, and were dressed informally. a. Late payment is calculated assuming one late monthly payment in the first month after acquiring the loan. 274  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES have offered and the product that was actually offered. We see that auditors are offered more expensive products, especially those offered checking accounts. Staff offered the basic account in only 2 of the 54 visits where a checking account was requested. Table 8.4b again echoes the findings of table 8.4a. All auditors are offered similar products, and yet there are large differences in the total cost of credit once usage costs are included. The last section of the questionnaire included a set of subjective questions about the auditor’s perceptions of the service provided by the staff. Table 8.5 reports these perceptions for the nonprofessional credit auditors.19 In about 30 percent of the visits, the control auditors thought that the official was kind and that the information provided was clear. Interestingly, none of the control audi- tors would acquire the product. Experienced auditors provide higher ratings of the officer’s performance and would acquire the credit in 19.4 and 28.6 percent, depending on the interest rate quoted. Auditors quoting a lower interest rate and thus signaling that they had a better offer were more likely to acquire the product, although they were not necessarily offered a better product. In addition, auditors also perceived that the staff were kinder and were more likely to be willing to acquire the product when they dressed formally. 8.7 CONCLUSIONS This study provides evidence of the quality and quantity of information that financial institutions provide to potential customers on their savings and credit products, as well as the ways in which mismatching of product features and low- income consumers’ financial activity profiles can lead these consumers to pay higher costs for the financial services they use in their daily lives. Consistent with the predictions of Gabaix and Laibson (2006), this study finds that consumers who behave in ways that indicate they have little experience with formal financial products are provided enough information to acquire the product but very little information about its costs, especially those that relate to avoidable fees and commissions. In contrast, experienced auditors who were instructed to ask specific questions about the product if the official did not disclose this infor- mation voluntarily end up better informed about avoidable fees and commissions. The first key finding then is that voluntary provision of basic product informa- tion is exceedingly low. Of course, one should not disclose all terms and conditions of the products because too much information may also be ineffective, but at least aggregate terms similar to the total annual cost and total annual earnings should Results for the professional savings auditors are available but are less relevant since the 19  number of auditors is small and the fact that they are professional suggests that they may be less influenced by the staff. 8.  Financial (dis-)information  ◾ 275 TABLE 8.5  Credit auditor’s assessment of staff AUDITOR WOULD OFFICER WAS CLARITY OF ACQUIRE KIND INFORMATION PRODUCT (1) (2) (3) High experience X high interest −0.066 0.053 0.194** (0.126) (0.120) (0.093) High experience X low interest 0.296** 0.232* 0.286** (0.126) (0.133) (0.118) High Indebtedness (1 = yes) −0.385 −0.051 −0.115 (0.243) (0.254) (0.190) Formal dress (1 = yes) 0.158* 0.061 0.139** (0.084) (0.083) (0.060) Observations 115 115 115 R-squared 0.487 0.491 0.405 Mean of dependent variable (among controls) 0.30 0.30 0.00 p -value of F-test: high experience 0.021 0.210 0.043 p -value of t-test: high interest–low interest 0.010 0.190 0.351 Mean Commercial banks 0.32 0.37 0.21 Low-income banks 0.39 0.36 0.15 SOFOME 0.22 0.24 0.19 SOFIPO 0.35 0.35 0.25 p -value of t-test Commercial banks = low-income banks 0.616 0.945 0.599 Commercial banks = SOFOME 0.424 0.335 0.852 Commercial banks = SOFIPO 0.826 0.908 0.777 Low-income banks = SOFOME 0.113 0.278 0.687 Low-income banks = SOFIPO 0.799 0.947 0.378 SOFOME = SOFIPO 0.282 0.401 0.599 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. All regressions include a dummy for each type of financial institution (commercial bank, low-income bank, SOFOME, and SOFIPO) and the following auditor characteristics: gender, age, whether the auditor is self-employed, and whether he or she is currently a borrower. Control auditors were assigned the script with low experience (neophyte), a preference for a loan size of 20% of household income, and were dressed informally. be adequately disclosed. Despite the fact that the total annual cost and total annual earnings should by law have been disclosed as they enhance compara- bility across similar products, in every visit the staff failed to explain their meaning correctly to experienced auditors who had to ask about them according to their script. The aggregate key terms to be disclosed should, however, include avoidable fees, unlike the total annual cost and total annual earnings which do not reflect them, because they can significantly affect the total yield and cost of products. 276  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES The second key finding is that staff responded to incentives by limiting the offer of basic savings accounts and by rejecting loan applications more frequently from overly indebted auditors or by reducing the loan amounts approved. The analysis of the average returns and fee structures for savings products in table 8.1 suggests that traditional checking accounts have a much stronger business case for providers than the mandated basic accounts, which limit potential revenue that can be earned from savings accounts through both avoidable and unavoidable fees. The findings point to an evident misalignment of incentives between officials and potential customers, both in terms of suitability of products to consumers’ needs, and in willingness and ability of sales staff to disclose key product infor- mation. They also demonstrate how disclosure and transparency policies may be difficult to implement successfully because financial institutions have strong incentives to undo them. Lowenstein, Sunstein, and Golman (2013) argue that disclosure policy implemented successfully may have little effect on consumers but large effects on financial institutions. This is, however, not the case in Mexico, where financial disclosure policy can be improved. For example, since experienced auditors received more product information because they were required to ask certain questions, basic guidance could be provided to consumers on key questions to ask when looking for a credit or savings product. In addition, information could be provided more transparently as tested in a follow-up to this study. This study covers standard consumer credit and savings products that may be expensive to offer, especially to low-income populations. A promising avenue is the take-up of low-cost mobile or online savings accounts, or access to financial products as a by-product of government-to-person payment programs, as is the case with the Oportunidades social payment program in Mexico. REFERENCES Agarwal, Sumit, John C. Driscoll, Xavier Gabaix, and David I. Laibson. 2013. “Learning in the Credit Card Market.” http://dx.doi.org/10.2139/ssrn.1091623. Anagol, Santosh, Shawn Allen Cole, and Shayak Sarkar. 2012. “Understanding the Advice of Commissions-Motivated Agents: Theory and Evidence from the Indian Life Insurance Market.” Working Paper No. 12-055, Harvard Business School, Cambridge, MA. Ausubel, Lawrence M. 1999. “Adverse Selection in the Credit Card Market.” http://www. ausubel.com/creditcard-papers/adverse.pdf. Banco de México. 2013. “Reporte sobre las condiciones de competencia en el mercado de emisión de tarjetas de crédito.” Mexico City. Berg, Tobias, Manju Puri, and Jörg Rocholl. 2013. “Loan Officer Incentives and the Limits of Hard Information.” http://dx.doi.org/10.2139/ssrn.2022972. Bruhn, Miriam, and David McKenzie. 2009. “In Pursuit of Balance: Randomization in Practice in Development Field Experiments.” American Economic Journal: Applied Economics 1 (4): 200–232. 8.  Financial (dis-)information  ◾ 277 Campbell, John Y., Howell E. Jackson, Brigitte C. Madrian, and Peter Tufano. 2011. “Consumer Financial Protection.” Journal of Economic Perspectives 25 (1): 91–114. Chetty, Raj, Adam Looney, and Kory Kroft. 2009. “Salience and Taxation: Theory and Evidence.” American Economic Review 99 (4):1145–77. Choi, James J., David Laibson, and Brigitte C. Madrian. 2011. “$100 Bills on the Sidewalk: Suboptimal Investment in 401(k) Plans.” Review of Economics and Statistics 93 (3): 748–63. CONDUSEF (National Commission for the Protection of Users of Financial Services). 2012. 2011 Anuario Estadístico. Mexico City. Duarte, F., and J. S. Hastings. 2011. “Fettered Consumers and Sophisticated Firms: Evidence from Mexico’s Privatized Social Security Market.” Brown University. Unpublished. Finkelstein A. 2009. “E-ZTAX: Tax Salience and Tax Rates.” Quarterly Journal of Economics 124 (3): 969–1010. Gabaix, Xavier, and David Laibson. 2006. “Shrouded Attributes, Consumer Myopia, and Information Suppression in Competitive Markets.” Quarterly Journal of Economics 121 (2): 505–40. Gross, David B., and Nicholas S. Souleles. 2002. “Do Liquidity Constraints and Interest Rates Matter for Consumer Behavior? Evidence from Credit Card Data.” Quarterly Journal of Economics 117 (1): 149–85. Grossman S. 1981. “The Informational Role of Warranties and Private Disclosure sbout Product Quality.” Journal of Law and Economics 24 (3): 461–83. Hastings, J. S., A. Hortaçsu and C. Syverson. 2012. “Advertising and Competition in Privatized Social Security: The Case of Mexico.” Brown University. Unpublished. Jovanovic, B. 1982. “Truthful Disclosure of Information.” Bell Journal of Economics 13 (1): 36–44. López Bolaños, Alejandro. 2012. “El Endeudamiento de los Hogares en México.” Boletín Momento Económico. Universidad Autónoma de México. Lowenstein, George, Cass R. Sunstein, and Russell Golman. 2013. “Disclosure: Psychology Changes Everything.” RPP-2013-20, Mossavar-Rahmani Center for Business and Government, Harvard Kennedy School, Cambridge, MA. Lusardi, Annamaria, and Olivia S. Mitchell. 2006. “Financial Literacy and Planning: Implications for Retirement Wellbeing.” Working Paper No. 1, Pension Research Council. http://www.dartmouth.edu/~alusardi/Papers/FinancialLiteracy.pdf. —. 2011. “Financial Literacy Around the World: An Overview.” CeRP Working Papers, Center for Research on Pensions and Welfare Policies, Turin, Italy. Milgrom, Paul R. 1981. “Good News and Bad News: Representation Theorems and Applications.” Bell Journal 12 (2): 380–91. Mullainathan, Sendhil, Markus Nöth, and Antoinette Schoar. 2010. “The Market for Financial Advice: An Audit Study.” http://dev3.cepr.org/meets/wkcn/5/5571/papers/ N%C3%B6thFinal.pdf. Nisbett, Richard E., and Lee Ross. 1980. Human Inference: Strategies and Shortcomings of Social Judgment. Englewood Cliffs, NJ: Prentice-Hall. Thaler, Richard, and Cass Sunstein. 2008. Nudge: Improving Decisions about Health, Wealth, and Happiness. New Haven: Yale University Press.World Bank. 2005. Broadening Access to Financial Services Among The Urban Population: Mexico City’s Unbanked. Washington, D.C. 278  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES ANNEX: ORTHOGONALITY CHECKS EXPERIENCED X EXPERIENCED X HIGH HIGH COMPETITION LOW COMPETITION INDEBTEDNESS (1) (2) (3) Gender 0.248 −0.193 0.477** (0.326) (0.375) (0.213) Age −0.02 0.003 0.021* (0.015) (0.017) (0.010) Education (1 = has licenciatura) −0.227 −0.079 0.051 (0.296) (0.340) (0.193) Household montly salary 0.000 0.000 0.000 (0.000) (0.000) (0.000) Owns business (1 = yes) 0.364 −0.068 −0.643*** (0.269) (0.309) (0.175) Has loan (1 = yes) 0.177 0.01 0.626*** (0.274) (0.315) (0.179) Reported in bureau (1 = yes) −0.112 0.051 0.066 (0.276) (0.317) (0.180) Constant 0.741 0.599 −0.715 (0.710) (0.815) (0.463) N 18 18 18 R-squared 0.439 0.141 0.809 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The dependent variable takes value one if the auditor was assigned an experienced and high competition script in column 1, experienced and low competition script in column 2, and a loan corresponding to 70 percent of household income in column 3. CHAPTER 9 L earning by doing? Using savings lotteries to promote financial inclusion in Nigeria MARTIN KANZ, ALEX KAUFMAN, KELLY SHUE, AND BENEDIKT WAHLER ABSTRACT Prize-linked savings products—such as savings lotteries—are a popular class of financial products widely used around the world. Prize-linked prod- ucts combine an otherwise standard financial product with the possibility of winning additional returns. Prize-linked savings have broad popular appeal because they use lottery elements familiar to large segments of the popula- tion, use simple savings products that do not require a high level of financial literacy, and have no downside risk for the client. We evaluate the impact of a large-scale savings lottery in Nigeria. The I-Save I-Win program was a nationwide savings incentive campaign run by one of the country’s largest banks. The program incentivized customers to maintain a prespecified balance in their savings account for a period of 90 days. Bank customers who met this target qualified for cash and noncash prizes. We find that the publicity generated by the lottery increased individual savings balances by 3 percentage points, and the usage of additional financial products by up to 5 percentage points, within a week of lottery events. However, we find no evidence that the incentive program led to persistent changes in savings, wider use of financial products, or intensified banking relationships that persisted after explicit incentives were removed. In addition, new media promotional campaigns, which encouraged customers to maintain minimum savings account balances to become eligible for the lottery, did not signifi- cantly affect individual banking choices. The patterns and effect sizes we The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 279 280  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES document are consistent with results from the literature on financial education, and underscore the challenge of designing incentive products and programs that can generate persistent changes in financial behaviors. 9.1 INTRODUCTION Low-income households in developing countries are often confronted with the dual challenge of highly volatile incomes and an inability to insure such income risk through access to insurance or formal credit. Without access to insurance markets, the accumulation of precautionary savings is often the only viable risk-management strategy available to poor households. Mobilizing precautionary savings has, however, proven difficult. In many contexts, access to basic banking services is limited; even where such services are available, there are often signifi- cant barriers preventing borrowers with limited financial literacy from harnessing their full potential. In this chapter, we explore the effectiveness of prize-linked savings prod- ucts as a device to encourage savings and build financial capabilities through “learning by doing.” Prize-linked savings products combine an otherwise standard financial product, such as a savings account, with a lottery component. In this way, customers are incentivized to adopt the saving product through its lottery component, which, in addition to regular interest payments, promises the pros- pect of a stochastic return in the form of one-time awards or annuities. Some early examples of prize-linked savings products include “prize bonds” issued by the French court to close the financing gap left by the profligate spending of Louis XIV. In recent times, lottery-linked savings products have often been used by banks in many emerging markets to extend their client base to previously unbanked populations. While policies to encourage savings have often relied on mandating saving, prize-linked products may provide an appealing alternative for building financial capabilities among households with low financial literacy. Many households in this target population have never had the experience of explicitly thinking through and committing to a savings strategy that will insure against future risk. Prize- linked savings products are attractive in this context, because they are easy to run through regular savings accounts, significantly less complex than dedicated commitment savings products, and have no downside risk for the client. More- over, the lottery component of the product is often an extremely effective factor in raising the appeal of such savings programs: for instance, 82  percent of all South Africans play the lottery at least once a week (Bar and Collins 2006), but only 53 percent have a formal savings account, and at 20 percent of gross domestic product (GDP), South Africa’s domestic savings rate is one of the lowest in the world (FANews 2011). In Nigeria, only 30 percent of all adults have an account at a formal financial institution, while lotteries are widespread. 9.  Learning by doing?  ◾ 281 Savings lotteries share many features of a basic commitment savings product. Such products, which have been introduced by many banks across the globe, are designed to build household savings through automatic deductions or mandatory contributions to a savings plan—e.g., a 401(k) retirement savings account. The available evidence suggests that commitment savings products can do much to boost savings among low-income households. Ashraf, Karlan, and Yin (2006), for example, report that individuals who were offered a commitment savings product in a field experiment in the Philippines were able to increase their savings by up to 81 percent relative to households in a control group without access to the product. In practice, however, using commitment savings products to build financial capabilities has been challenging. Dedicated commitment savings products are relatively costly to introduce, to market, and to administer. Take-up of these prod- ucts has therefore often been suboptimally low and unevenly distributed, with more financially literate households being disproportionately likely to demand these products. By contrast, prize-linked saving products, such as the savings lottery we study in this chapter, are usually based on an existing financial product, such as a savings account, and create direct incentives to commit to a savings plan by offering stochastic returns. Although prize-linked products have been shown to be effective at reaching low-income nonsavers, an important question is whether they can generate persistent changes in saving behavior. While some incentive savings programs, such as prize bonds, have been offered consistently and over an extended period of time, many savings lotteries offered by private lenders have been limited in time, and it is unclear whether this relatively short exposure to the incentive program is sufficient to build financial capabilities. We investigate the impact of the I-Save I-Win (ISIW) promotion, a nationwide financial capability program launched by a leading Nigerian bank (InterContinental Bank [ICB]) in the spring of 2011 with the aim of mobilizing precautionary savings through mass market savings accounts. The ISIW promotion was a heavily publi- cized lottery incentive scheme designed to educate consumers about savings through first-hand sustained use of individual savings accounts. All consumers who opened new accounts or maintained existing savings accounts above partic- ular balance thresholds were eligible for the lottery. Over 600 lottery prizes, ranging from cash payouts to television sets and cars, were awarded in stages over the promotional period. The promotion delivered a simple educational message: lottery awards are the immediate draw, but every saver is a winner. From a financial capability perspective, the promotion resembled a simple commitment savings product (see, e.g., Brune et al. 2011; and Ashraf, Karlan, and Yin 2006) that incentivized the establishment of new savings accounts and the more active use of accounts among existing account holders. Rather than relying on explicit financial education content, the program sought to change financial behaviors directly by providing incentives for “learning by doing” through the active use of formal financial services. 282  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES In this way, banking relationships and learning by doing induced by ISIW may change financial behavior even after the promotion is over, leading to greater financial capabilities, as measured by precautionary savings and the use of finan- cial products beyond savings accounts. In the empirical analysis, we examine in detail which of these goals the program achieved by exploring the impact of the incentive program on savings balances, the intensity of banking relationships, and changes in the use of financial services among program participants. While many past efforts to increase financial capabilities have offered educa- tion hoping it would improve later financial behaviors, ISIW focused on changing financial behaviors directly. Was this approach effective, and did it have lasting effects? Direct experience with savings, whether the experience is gained because of a desire to self-insure or because of the promise of a prize, may serve to familiarize customers with formal financial services and lower the barriers to future participation. It may also demonstrate to customers that they can manage to maintain levels of savings previously thought impossible. In this way, savings induced by the ISIW promotion may lead to learning by doing, changing financial behavior even after the promotion is over (as measured by long-term level changes in savings balances and adoption of other financial products). Analysis of the ISIW promotion will be of interest to both banks that wish to draw poor, unbanked clients into the formal banking system; regulators who wish to encourage precautionary savings among vulnerable households; and researchers exploring methods other than direct education that may be used to increase financial literacy. Nigeria’s emerging market for consumer finance is a particularly useful testing ground for programs like ISIW that combine private sector initiatives with “nudges” intended to build financial capabilities. Like other countries in the region, Nigeria does not have a significant history of state banking and lacks an extensive network of state banks through which government-led financial capa- bility programs could be delivered. This increases the role of the private sector in bringing individuals with limited financial literacy into the fold of the formal banking system. ISIW is an example of a simple, commercially viable program that has successfully aligned private and social incentives to achieve this goal. Several important details of implementation allow us to plausibly estimate the impact of the ISIW program. First, ISIW was heavily promoted over a seven-month period through a variety of innovative media channels. The media push included celebrity endorsements from popular politicians and athletes. For example, Sullivan Chime, the governor of Nigeria’s Enugu State, publicly endorsed ISIW via a televised ceremony. The popular Nigerian football star Segun Odegbami also endorsed ISIW and encouraged personal savings as a way of transforming the lives of participants (WorldStage 2011). In addition, the ISIW lottery was promoted on television and via social media portals such as Facebook pages and YouTube videos. Finally, the stag- gered public drawing of lottery prizes across geographical regions itself functioned as a form of promotion for the ongoing portion of the ISIW lottery. 9.  Learning by doing?  ◾ 283 By exploiting the staggered introduction of various components of the ISIW promotional campaign over time, we can measure the marginal increase in savings applications and banking activity following each component of the savings campaign, including lottery drawings and promotional events. We will control for time trends using data from a second major Nigerian bank, Access Bank. By comparing banking behavior across these two banks, before and after the introduction of each ISIW media promotion, we can estimate the differenc- es-in-differences impact of each type of media push on short- and medium-term consumer banking behavior. In addition, the presence of a lottery winner in the local area likely attracted participants through their social networks. Because the location of winners was randomly determined by the lottery, this additional source of variation allows us to estimate both the short- and long-term impacts of participation in the ISIW program. We find some evidence that the publicity generated by the lottery had a short-run impact on financial behaviors and increased individual savings balances by 2–3 percentage points, and the usage of additional financial products by up to 5 percentage points, within a week of lottery events. However, we find no evidence that the incentive program led to persistent changes in savings, intensi- fied banking relationships, or the use of financial products. In addition, new media promotional campaigns, which encouraged customers to maintain minimum savings account balances to become eligible for the lottery, did not significantly affect individual banking choices. These patterns are consistent with results from the literature on financial education and underscore the challenge of designing education and incentive programs that can generate persistent changes in finan- cial behaviors. 9.2 BACKGROUND 9.2.1 Savings and financial literacy A number of recent papers investigate household savings in developing countries and find both low rates of savings and low access to formal savings channels. Banerjee and Duflo (2007) summarize survey data from 13 countries from 1998 to 2005 and find that few people with expenditures below US$2 per day manage to maintain formal savings accounts. Instead, households—if they save at all—use alternative methods, documented in Rutherford (2000). Such alternative savings channels may include self-help groups, popular in India; rotating savings and credit associations (ROSCAs), popular in Africa and Latin America (see also Besley, Coate, and Loury 1993); as well as moneylenders and other informal means. Compounding the problem of low savings rates is the fact that individuals and small businesses in developing countries tend to borrow at high rates. As pointed out by Banerjee and Duflo (2008), only a few small businesses with high rates 284  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES of return persist—since they are forced to borrow at extremely high rates—and many do not maintain cash buffers that would otherwise help them avoid costly borrowing. Another branch of the savings literature argues that demand for savings accounts is high, but overall savings rates are low because of lack of access to formal savings technologies. Prina (2011) describes an experiment that offered savings accounts to households in the slums of Nepal. The take-up rate was high, and the average household saved 7 percent of its weekly income. In a similar vein, Collins et al. (2009) argue that many poor people wish to use better savings tech- nologies—more reliable, more convenient, cheaper—than they currently use. In fact, people often pay fees to save money—i.e., they are willing to accept nega- tive interest in exchange for access to mechanisms that allow them to save. While the literature finds that many people who already save are willing to change overt to formal savings channels, it has been more challenging to increase savings participation overall. Results from the literature exploring the effect of financial literacy and education programs on money management decisions are, for example, extremely mixed. Carpena et al. (2011) study the effects of a financial literacy training consisting of five weekly video screenings, and find that the program affects participants’ numeracy but not their awareness of financial products. Other studies have found that financial literacy training tends to have only a small impact on actual behaviors—especially when educational content is delivered at a point where the target audience does not have the opportunity to translate this information into action. Relatedly, studies of targeted financial literacy interventions generally find newly acquired financial knowledge to have a relatively short half-life and therefore very limited effects on longer-run behavior. Xu and Zia (2012) provide an overview of other targeted financial literacy interven- tions; they find that lectures do not significantly affect financial behavior, while interactive lessons can generate stronger effects. In contrast to this line of research, we explore the effect of a program that directly incentivized individuals to save and to try out a new savings technology. The objective of this program was twofold. First, by encouraging people to save for a few months by offering lottery rewards, the hope is that program participants will learn from the experience and continue to save even after explicit incentives are removed (learning by doing). Second, among individuals who already have an account, the program creates incentives for more frequent interactions with the formal banking system. This may build financial capability by improving familiarity with formal financial transactions and a wider range of financial products. In part, our project is motivated by findings by Cole, Sampson, and Zia (2011), who compare the effect of financial literacy programs with small financial induce- ments to open a bank account. The ISIW program is similar in its underlying objective—individuals who open or maintain an account and receive additional expected returns through the opportunity to win lottery prizes, conditional on achieving a prespecified savings target. 9.  Learning by doing?  ◾ 285 Our focus on learning by doing in the use of financial services is also motivated by Lusardi, Keller, and Keller (2009), who find that learning about the process and difficulties of saving from others’ stories is highly effective at motivating savings behavior. To the extent that holding a savings account (and having to maintain a given balance level for several months) may provide an even more salient experi- ence, learning by doing could also be helpful in creating a long-term savings habit. 9.2.2 Time inconsistency and commitment savings products The ability to engage in welfare-enhancing financial behaviors, such as saving, may be constrained by other factors beyond financial literacy and the availability of appropriate savings instruments. Much of the available evidence suggests that a wide range of behavioral and psychological factors such as time inconsistency may prevent people from saving as much as they would like. Against this background, a number of studies explore whether savings prod- ucts can be designed to help people commit to save. Much of this research is motivated by the hyperbolic discounting model (Laibson 1997; O’Donoghue and Rabin 1999; Thaler 1990). Hyperbolic discounting describes the case in which an agent sharply discounts all future consumption relative to today. This is a popular model that explains low rates of precautionary savings and is consistent with observed behaviors. Even individuals who understand that saving would be welfare enhancing and would like to commit to a savings plan may be tempted to place a greater weight on current consumption at the end of the month when their paycheck arrives and they have the opportunity to trade future versus current consumption. This is the commitment problem captured by the hyper- bolic discounting model: an individual expects that in the future, he or she will commit money to savings, but when the time comes, he or she no longer finds it worthwhile. If the agent can anticipate this tendency, he or she will demand commitment savings devices that prevent reneging on the savings goal. In response to the recognition that time-inconsistency problems may pose a significant barrier to a household’s ability to save, banks and microfinance insti- tutions have introduced so-called commitment savings products (Ashraf et al. 2003). A typical commitment savings product restricts or penalizes withdrawals before a specified date or penalizes insufficient deposits. A study by Ashraf, Karlan, and Yin (2006) shows that commitment savings products can significantly affect behavior. The authors conduct a randomized trial of a commitment savings product offered by a bank in the Philippines. The study demonstrates that there is significant demand for products offering individuals a way to save—28 percent of bank customers took up the product and used it to save. More interestingly, the authors show that, consistent with the theory, more impatient individuals (i.e., those with the highest elicited discount rates) were also the most likely to take up the product. 286  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Mullainathan and Shafir (2009) discuss some additional behavioral issues that may generate low savings rates among low-income households. Their analysis focuses mainly on U.S. households, although the lessons apply more generally. They argue that mental accounting explains why people who save using cash will have a greater tendency to spend their savings on impulse purchases than people who keep their savings in formal savings accounts (Thaler 1999). Lack of reminders and limited attention may be another factor. For example, credit card companies give timely reminders to pay, while savings accounts do not regularly prompt the holder to contribute available money toward a financial goal. Overall, the behavioral economics view of savings behavior suggests that observed savings behavior may run counter to the long-run preferences of the agents themselves. This motivates efforts to create products and incentive programs that try to build financial capability by inducing people to save more, including promotions and future efforts that this project hopes to inform. While the main aim of the ISIW promotion was to raise savings balances and to promote savings through learning by doing, it shares some characteristics usually associated with commitment savings schemes. By opening an account and main- taining a certain account balance, the agent receives an anticipated reward (the expected lottery payout). If the agent were to withdraw from the savings account prior to the end of the promotion period, he or she then suffers a loss of the expected lottery rewards. This creates an incentive mechanism similar to that of a standard commitment savings product, where the prospect of expected returns combined with penalties for not maintaining the savings target creates an incen- tive to commit to a savings plan. When it comes to reaching low-income households which face particularly severe constraints in committing to a savings plan, prize-linked savings products may be especially effective. Traditional commitment savings products suffer from the limitation that they are relatively complex to market and administer. They may also be prone to problems of selection. Because it requires some level of financial sophistication to be aware of one’s own commitment problems and sign on to financial products that can help mitigate these problems, it is likely that there is a disproportionate selection of financially literate households in the take-up of these products. Prize-linked products generally do not suffer these selection problems. On the contrary, by using lottery elements with wide popular appeal, such products are especially likely to be successful among hard-to-reach segments of the population that would not otherwise sign up for a savings plan. Using a survey among potential clients of prize-linked commitment savings products in the United States, Tufano, Maynard, and De Neve (2008) show that demand for this class of product is greatest among low-income individuals who do not have regular savings habits and have little actual savings, but regularly play the lottery. 9.  Learning by doing?  ◾ 287 9.2.3 The history of prize-linked savings products Prize-linked savings products have been used by governments and financial insti- tutions for at least three centuries. These entities learned early on that combining savings with lottery features was an effective means of raising capital. In cases where savings lotteries have been run by the state, they have often been combined with the issuance of specialized bonds, so that in addition to the potential upside of the lottery, they often provided consumers with an incentivized introduction to new savings products. The earliest use of lotteries as a device to encourage savings dates back to the “Million Lottery” or “Million Adventure,” which was used to finance King William of Orange’s efforts in the Nine Years War (Murphy 2005). Launched in 1694, the Million Adventure was designed to appeal to a wide range of inves- tors, as it promised large prizes to fortunate winners for their £10 ticket with no downside risk. For every 100,000 tickets sold, the lottery offered 2,500 generous prizes, the highest being an annuity of £1,000 for 16 years; the lowest an annuity of £10. Moreover, unlike other lotteries, the non-prize-winning, or blank, tickets also entitled the holder to a modest but reasonable return on investment of £1 per year until 1710. As a result, the tickets were swiftly sold, and soon after the draw, a secondary market for the drawn tickets, including the blanks, formed— which, after all, now had the features of a government bond. Savings ideas soon became popular elsewhere. In France, Louis XV intro- duced the first “borrowing lottery,” in part to make up for excessive spending under the reign of Louis XIV (Pfiffelmann 2009). In a borrowing lottery, bonds issued by the state paid bondholders through a draw. Each bondholder was given a ticket to participate in the draw. The winners received funds that were significantly higher than what they had originally invested, while losers received interest compensation in the form of a rente viagères (a form of perpetuity that ends upon the death of the beneficiary). A few decades later, in 1777, Jacques Necker, then French comptroller-general, resorted to a borrowing lottery to raise capital after France had declared war on England the previous year. He arranged for issuance of “the most attractive bonds possible”—20,000 titles at a face value of 12,000 livres each. The drawing comprised two draws without replacement; the first offered 3,000 lots of annuities and the second, 15,000 perpetuities and 2,000 batches of rente viagères. Prize-linked savings products were launched successfully by several govern- ments throughout the 20th century. In 1918, Sweden needed to find new sources of funding to finance postwar development and began issuing bonds annually or biannually. The coupons of these bonds were determined by a lottery drawing. The lottery became so popular Sweden’s Debt Department faced stiff opposition when it tried to discontinue the program in 1931. In the United Kingdom, the government of Harold Macmillan introduced so-called “premium bonds” with the budget of 1956, and Ireland saw the 288  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES introduction of “prize bonds” in the same year (Lobe and Hölzel 2007; Tufano 2008). Accessible in small denominations, both systems paid the interest into a prize fund out of which a electronic random number generator would select the winning bond numbers on a regular (monthly for the United Kingdom, weekly in Ireland) basis. With about one in three Britons holding premium bonds, this form of savings has gained widespread appeal. Banks have noticed the wide appeal of government-sponsored savings lotteries, and have launched a range of similar products. First introduced by a local cooperative bank in Germany in 1950, a program known as GewinnSparen (WinSaving) is now offered by many cooperative and savings banks across the country.1 As an incentive for the client to commit to a regular savings amount (deducted as a direct debit from the client’s current account), part of this amount is effectively invested as a ticket for a lottery which typically has a 55  percent payout quota and is run by an nonprofit association independent of the bank.2 Prize-linked savings products have been successfully used by public and private entities around the world, particularly in Latin America, Southeast Asia, Scandinavia, and the United Kingdom. More recently, lottery-linked products have been introduced in the United States, where legislation still limits the use of prize- linked savings products for commercial purposes. 9.2.4 Savings lotteries in Sub-Saharan Africa Savings lottery promotions are relatively common in a number of Sub-Saharan African economies. However, only a few banks have rolled out a single promo- tion across the breadth of the countries in which they operate. One of the first instances of a launch of a savings lottery across different markets occurred in September 2012, when Ecobank launched its Win Big with Ecobank promotion in several African markets including the Gambia, Malawi, Tanzania, and Zambia. Banks in many African markets have been open to the introduction of savings promotions with lottery features, as these are often used as a short-term instru- ment for new client acquisition and rapid deposit mobilization. However, legal barriers to lottery-linked savings products persist in several countries. In South Africa, for example, First National Bank’s very successful Million-a-Month Account (MaMA) program was banned by a court in 2008, and other South African banks have been prevented by law from running similar programs. In this subsection, we review the MaMa program and selected examples of lottery promotions from Kenya and Nigeria. 1  Wiesbadener Volksbank (2010). While this product has become a standard component in the cooperative and public 2  savings bank space, its impact on the bank’s balances appear limited, judging by an annual savings mobilization of €18 million in the case of some 50 cooperative banks in northern Germany. 9.  Learning by doing?  ◾ 289 SOUTH AFRICA: FIRST NATIONAL BANK MILLION-A-MONTH ACCOUNT In 2005, First National Bank, one of South Africa’s leading banks, launched MaMA, a lottery-linked savings program. The idea was to draw on the popularity of traditional lotteries among South Africans to encourage savings behavior in the country (see also Cole et al. 2008). For example, while nearly 80 percent of South Africans spend on average R 80 on the lottery at least once a week, South Africa’s domestic savings rate is one of the lowest in the world at about 20 percent of its GDP (FANews 2011), and its household savings rate is below 1 percent. At zero cost, the MaMA program encouraged South Africans to open a 32-day savings account with a minimum balance of R 100 (about US$12) (account holders were not permitted to withdraw their savings without giving a 90-day notice to the bank); they were offered a nominal low interest rate of 0.25 percent. 1,000 In return, savers were automatically offered a chance to win between R  1 million (US$115,000) for every R  (US$115) and R  100 on deposit in a monthly draw. MaMA became very popular, as First National Bank opened nearly 1 million savings accounts with an average balance of R 1,000 and mobilized over R 1 billion (US$115 million). The impact of this program on financial inclusion was consider- able: approximately 20 percent of those signing up for an account were reported as previously unbanked (De Waal 2012). The program was introduced around the same time that South Africa enacted new national lotteries legislation that provided legal clearance for the bank to launch the product. First National Bank obtained additional formal approvals from Uthingo (the national lottery operator) and the National Lotteries Board, which viewed the program as a promotional tool rather than a form of lottery and raised no objection to it. However, taken by surprise by MaMA’s rapid success, six months after the launch, the National Lotteries Board ordered the bank to stop the program and later obtained an injunction from the High Court of South Africa against First National Bank. A protracted legal battle ensued that resulted in the Supreme Court of Appeal ruling in favor of the National Lotteries Board. After an initial wave of withdrawals by participants of the MaMA program, the majority of client savings remained with the bank after the demise of the program. KENYA POST OFFICE SAVINGS BANK’S PREMIUM BONDS As part of its mission to “encourage and facilitate savings” in Kenya, the Kenya Post Office Savings Bank, a state entity, introduced the “premium bond” in 1978. The premium bond was a savings product akin to a lottery savings scheme whereby holders, mostly low-income individuals or families, could buy premium bonds to enhance their savings and stand a chance to win monthly cash prizes (Kamewe and Radcliffe 1999; Wright 1999). These premium bonds were offered 10 (US$0.14) and K Sh  in denominations of K Sh  20 (US$0.28). Holders could not make any withdrawals on their interest-free deposits within the first three months after the initial deposit. By 1998, these deposits constituted about 0.62 percent 290  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES of the Kenya Post Office Savings Bank’s total deposits (or US$0.56 million out of US$89.6 million). Interest in the premium bonds was particularly focused in rural areas where, until the early 2000s, hardly any alternatives to the Kenya Post Office Savings Bank existed. Plans to introduce larger denominations of K Sh 500 (US$7.14), K  1,000 (US$14.28), and K  Sh  10,000 (US$142.80) and increase the Sh  prizes were abandoned, though, when the bank ceased to offer the premium bonds in October 2004 in the face of increasing competition and restructuring. Likely one of the first lottery savings programs in Sub-Saharan Africa, the Kenya Post Office Savings Bank’s premium bonds did not face the regulatory hurdles that, for example, First National Bank’s MaMA did. Several Kenyan and international banks such as Barclays and Ecobank are now turning to savings promotions to incentivize Kenyans to save and increase their deposit base. In 2012, Barclays Bank Kenya and Ecobank Group launched savings promotions with features similar to the ones described below for Nigeria. NIGERIA In 2010, several Nigerian banks including Union Bank, Oceanic Bank, and Spring- Bank (renamed Enterprise Bank after the takeover by the country’s “bad bank” AMCON) began to deploy lottery-based savings promotions to increase their deposit base and expand their customer base among unbanked and under- banked segments of the population. Each of these savings promotions has been structured uniquely, but share several commonalities. The prizes on offer are most often cash, vehicles (cars, scooters, or motorbikes), household appliances, personal electronics, or some combination of the four. Promotions are held for a limited period of time, ranging in duration from three months to one year. Draws are usually held monthly and on a rotating basis (in different geographic zones). This was also the case for the ISIW promotion, which we analyze in detail below. There are some nontrivial regulatory hurdles for financial institutions wishing to launch a prize-linked savings product in Nigeria. To launch a lottery promo- tion, a bank must first register with the National Lottery Regulatory Commission, apply for a permit and pay permit processing fees of between N = 2 million and = 10 million (US$63,400), depending on the length of the promotion. Each subse- N quent promotion requires a permit of its own. Among other stipulations, the commission demands that the location of each draw be open to members of the public and that its officials be present. Additionally, banks are required to register their promotions with the Consumer Protection Council of Nigeria and pay the requisite registration fees. We describe below a number of savings promotions that have been successful completed in Nigeria. Diamond Bank’s SavingsXtra promotion. Initially launched in June 2008, Diamond Bank Plc.’s annual SavingsXtra promotion was created with the stated aim of engendering a culture of saving among the Nigerian populace and to reward customers for their loyalty. For the bank, the mobilization of low-cost deposits as 9.  Learning by doing?  ◾ 291 a means of funding and driving client growth as Diamond Bank expanded its retail footprint were central. However, by the same token, this implied an extension of the bank’s customer base to segments of the population with previously limited access to formal banking. Since its inception, the bank has rolled out several phases of the scheme: Season 2 (September 2009–August 2010), Season  3 (September 2010–August 2011), Season 4 (September 2011–August 2012), and Season 5 (September 2012–August 2013). As of mid-2011, 30,000 new savings accounts were opened each month through SavingsXtra (ABN Digital 2011). To be eligible to participate in the monthly and weekly draws, existing customers must simply maintain a minimum balance of N = 5,000 (US$31.70) in their Diamond SavingsXtra account prior to the draw. New customers must open an account and maintain a minimum balance of the same amount. All account holders with the required balance are equally likely to win prizes, but for every additional = 5,000 held in the account, customers receive more lottery tickets—thereby N increasing their chances of winning. Prizes include cash gifts of varying amounts = 500,000, N = 250,000, N (N = 2 million), household appliances, and a unique offer enti- = 100,000 (US$634) monthly for tled Salary4Life where a lucky customer receives N a period of 20 years. In addition to these prizes, SavingsXtra account holders also receive life insurance coverage of N= 50,000 and automatic eligibility to apply for a SavingsXtra Visa credit card (with a credit line capped at 75 percent of their cash balance). While the savings account permits withdrawals, it still offers 2 percent annusl interest (daily accrual) independent of the account balance. Given these favorable conditions, the multiple repetition indicates the bank’s interest in this product, as it appears to have driven up new accounts strongly with more than 75 percent of new accounts remaining active after the promotion.3 First Bank of Nigeria’s Save & Excel promotion. FirstBank of Nigeria’s Save & Excel promotion (February–December 2012) also stated a broader goal of promoting financial inclusion by compelling unbanked/underbanked groups such as farmers, women, and senior citizens to test out different savings products. The top prizes in the quarterly draws were all-expense-paid trips to the London Olym- pics and Peugeot 307 cars. Cash prizes of varying amounts and household appli- ances such as home theater sets and chest freezers were offered in the monthly draws. Thresholds were put in place for participation in draws and for eligibility to win the Olympics trip. To participate in the monthly draws, customers need to have saved and maintained a balance of N = 20,000 (US$127) in the period leading up to the draw—i.e., four times the minimum amount at Diamond Bank. To be eligible for the quarterly draw, customers need to have maintained a minimum balance = 20,000 = 60,000 for three months. Alternately, they need to have saved N of N incrementally every month for three consecutive months. Finally, to qualify for 3  Source: conversations with Diamond Bank executives. 292  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES the Olympics trip, customers need to have saved N= 30,000 every month for four = 120,000 for four months. In the consecutive months or maintained a balance of N wake of previous savings promotions, First Bank aims to look at continuing this product category in the future, as it credits a growth of up to 40 percent in its client numbers to these initiatives.4 First City Monument Bank’s Millionaire Savings promotion. First City Monu- ment Bank’s Millionaire Savings promotion, which ran from June 2011 to April 2012, was also developed to foster a savings culture and reward bank customers for their loyalty. The top prize was a N= 1 million cash gift; consolation prizes = 10,000. To qualify to participate in the = 100,000 (US$634) and N included gifts of N consolation prize draws, customers were required to have or to open a First City = 10,000 for Monument Bank Millionaire Savings certificate and save a balance of N 30 days or more prior to the draw. Every subsequent unit held for 30 days or more qualified a saver for the millionaire top prize draws. The promotion was unique in that it was carried out on a rolling basis; nonwinning certificates were rolled over every month to subsequent draws. In generating awareness of the scheme, management emphasized that the certificate could be liquidated for its full value at any point and that customers could exit from the scheme at their convenience. The bank’s experience with the product, though, was less enthusiastic compared to initial sales projections. The nature of the certificate—vis-à-vis a more standard account—might have played a role in diminishing the product’s appeal, as well as the lower reward compared to peers and limited advertising expenditures. The fees levied by the National Lottery Regulatory Commission further challenged the overall profitability of the product for the bank; it is currently undergoing revision.5 While savings promotions have become a standard feature in the product offerings of Nigerian banks, it is also evident that banks continue to test product features and marketing interventions to guarantee the economic success of these products as well as ensure they do not just generate a one-off effect of clients signing up only to relapse into dormancy as soon as a lottery expires. In this respect, it appears that the incentives of banks are highly congruent with a finan- cial inclusion agenda, aiming to draw new customers into formal banking relation- ships and ensuring that accounts do not just get opened but are also actively used. 9.3 THE I-SAVE I-WIN SAVINGS PROMOTION We study a savings promotion program designed to incentivize the opening of new savings accounts and raise savings balances among existing account holders. The 4  Source: conversations with executives of First Bank. 5  Source: conversations with executives of FCMB. 9.  Learning by doing?  ◾ 293 program we evaluate is the ISIW savings promotion, a commercial savings incen- tive program offered by ICB, one of Nigeria’s leading retail banks with a network of 362 branches across the country and a client base of more than 2 million active savings account holders in all states in Nigeria. The ISIW savings promotion was announced in March 2011 and ran from April to November 2011. The bank’s motivation for offering the program was twofold: first, to raise savings balances among existing clients, and second, to extend the appeal of basic savings accounts to Nigeria’s large unbanked population—many of whom reside in remote areas of the country and have limited prior financial literacy exposure to formal financial services. The remainder of this section describes the design of the program. The first subsection reviews the program rules and eligibility criteria, the second describes the advertising campaign that accompanied the program, and the third provides details about lottery winners and the timing of regional and national prize drawings. 9.3.1 Program rules and eligibility Program rules were kept deliberately simple to achieve widespread participation. Bank clients qualified for participation in one of the six regional draws of the = 50,000 (US$300) for a minimum period savings lottery by maintaining a balance of N of 90 days.6 For the drawings that were held more than 90 days after the contest was announced, the requirement was amended so that a customer needed at = 50,000 between least 90 days—not necessarily consecutive—with a balance of N the contest announcement and the prize drawing. Savings account holders qual- = 100,000 (US$600) for a ified for the national draw by maintaining a balance of N minimum period of 90 days. The savings lottery was open to both existing bank clients as well as clients who opened a new savings account between April 4 and October 31, 2011. The program applied only to dedicated savings accounts. Balances held in current, checking, and other types of accounts did not qualify bank customers for participation in the lottery. 9.3.2 Social media campaign and celebrity endorsements The lottery was advertised through print media and direct marketing as well as radio and television commercials. The marketing campaign accompanying the program targeted both existing and potential new bank clients and had three main components (figure 9.1). First, to mobilize savings among existing clients, the bank launched a direct marketing campaign utilizing relationship banking interactions as well as direct marketing by mail and new media to generate awareness about the program. This direct marketing campaign had several components. Branch managers, sales 6  As a basis for comparison, Nigeria’s per capita GDP in 2011 was US$1,502 (World Bank). 294  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 9.1  Social media campaign and celebrity endorsements 9.  Learning by doing?  ◾ 295 representatives, and relationship managers provided information materials to customers at the branch level. In addition, information about the savings promo- tion was included in the bank’s regular mailings to its customers. Finally, the bank undertook several waves of direct marketing using mobile phone text messages. Targeted SMS messages were sent to more than 100,000 bank customers in three waves. The messages contained a list of regional and national prizes and informa- tion on how to sign up for the savings lottery. Second, the bank used television and radio advertisements aired in English and multiple regional languages across all states of Nigeria. Television and radio commercials were designed specifically to reach out to marginal clients in remote and underbanked areas of the country where existing banking relationships and direct marketing via mobile phones and social media are less effective. Televi- sion and radio advertisements were used throughout the promotion, initially to announce the program and generate awareness about the benefits of increased savings. At later stages in the program, radio and television commercials were also used to publicize the prize drawing events and announce the winners of the regional and local prize drawings. In addition to traditional television and radio commercials, the bank launched a comprehensive social media campaign, which included a dedicated Facebook page, a YouTube video stream, and a Twitter channel. The social media campaign was used to publicize prize drawings and included a basic financial education message: “By saving more, everybody wins.” Finally, to extend the appeal of the savings promotion beyond the bank’s traditional customer base, the bank secured a number of high-profile celebrity endorsements. For example, in April 2011, international soccer star Segun Odeg- bami endorsed ISIW, in a move that was widely publicized in local and national media. Over the course of the promotion, ISIW received additional endorsements from popular Nigerian media and sports personalities. 9.3.3 Prize drawings The bank held prize drawings for each of its 10 regions between July and November 2011, followed by a final drawing for the grand national prize on November 3, 2011. To be eligible for the regional draw, a customer needed to maintain a balance of = 50,000 for 90 days prior to the contest; for the national draw, the threshold N = 10,000 = 100,000. In each regional draw, 20 customers won a cash prize of N was N = = (US$60), 10 each won cash prizes of N20,000 (US$120) and N50,000 (US$300), 10 won 32” LCD TVs, and one won a Toyota Corolla. The prize in the national drawing = 15 million (US$95,000). To was an apartment in Lagos, or its cash equivalent of N guarantee the integrity of the lottery, regional and national prize drawings were overseen by the accounting firm PriceWaterhouseCoopers. Regional prize drawings took place in six phases between July and October 2011. In each phase, winners were drawn from all qualifying account holders in one or more of the bank’s 10 operational regions. Each region encompasses 296  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES between 5 and 10 states of Nigeria and between 10 and 30 of the bank’s 362 branches. The first regional prize draw occurred on July 21, 2011, and covered the capital territory of Abuja and the north central region of Nigeria; subsequent regional prize drawings followed on August 8 for the southeast region, August 25 for the south region, September 14 for the northwest region, and October 21 for the southwest region. The final regional prize drawings for northeast Nigeria, the Lagos mainland, and Lagos island took place concurrently with the national draw on November 3, 2011. The prize drawings were televised, and prize winners were announced in regional and national print media, including Nigeria’s leading daily newspapers, the Guardian and This Day, following each prize drawing. Table 9.1 summarizes the timing of prize drawings; figure 9.2 shows the sequence of prize drawings by state. 9.4 DATA AND DESCRIPTIVE STATISTICS We analyze the impact of the ISIW savings promotion using detailed microdata on daily account balances for more than 4 million account holders at all of the bank’s 362 branches for the period between October 1, 2010, and August 30, 2012. We merge this data set with information on (1) demographic data on each account holder; (2) data on the complete set of financial products used by each client before and after the program; and (3) data on the timing of lottery draws, prizes won, and the geographical location of lottery winners. We additionally use data on the media campaign that accompanied the savings lottery. These include records on the timing of television commercials and newspaper and radio promo- tions. We were able to collect data for all clients at ICB, as well as for a control sample of 620,035 customers of a second leading Nigerian bank—Access Bank— which did not have a comparable promotion during the time period covered by our data. The remainder of this section describes our data sources and provides descriptive statistics on customer demographics, savings balances and transac- tions, and summary statistics on the lottery drawings to motivate the subsequent econometric analysis. 9.4.1 Savings balances and transactions The main data set consists of detailed transaction-level data of 2,958,714 savings account customers. This includes all 2,338,679 savings account holders at the bank that ran the ISIW promotion, as well as a random sample of 620,035 account holders at the control bank. The data series consists of daily data between October 1, 2010, and August 30, 2012, for a total of 1,979,379,666 daily account observations. 9.  Learning by doing?  ◾ 297 TABLE 9.1  Prize drawings REGION DATE DRAW NUMBER Abuja July 21, 2011 1 North central July 21, 2011 1 Southeast August 8, 2011 2 South August 25, 2011 3 Northwest September 14, 2011 4 Southwest October 21, 2011 5 Northeast November 3, 2011 6 Lagos mainland November 3, 2011 6 Lagos island November 3, 2011 6 National Draw November 3, 2011 FIGURE 9.2  Sequence of prize drawings by state 4 4 4 6 4 4 4 6 6 4 6 1 1 6 1 1 1 5 6 5 5 1 1 5 5 6 3 2 2 2 3 3 2 2 3 3 3 These data allow us to compute daily balances and changes for every account over any subperiod. Accounts at the two banks were combined following their merger in March 2012. For the purpose of this chapter, we consider only balances and transactions up to February 2012. Customers with large balances might have 298  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES large balance changes as well, and thus heavily influence both the average and total figures; it is likely that the impact of the lottery is particularly pronounced among customers with balances close to the eligibility threshold of N = 50,000 and = 150,000. In the statistical analysis, we restrict our sample accordingly. Second, N power calculations indicate that our minimum detectable effect size is very small, on the order of a few naira, even with relatively low statistical power. This allows us to further focus the statistical analysis and consider the effect of the lottery on very specific subpopulations in the sample. It is worth noting an important caveat about the account data underlying our analysis. Because of inconsistencies in the data on account balances, we currently restrict our attention to data on balance changes within a time window of ±7 or ±14 days around ISIW events, for which we can be sure that our results are not biased by measurement error. This places some limitations on our anal- ysis. First, we have only an approximate measure of whether a bank customer is above or below the N = 50,000 cutoff at the date the lottery is introduced. Hence we currently cannot exploit the sharp discontinuity in eligibility for the program that = 50,000 in our analysis. Second, one would expect the response to the occurs at N lottery incentive to be heterogeneous, with take-up and balance changes being larger for households with smaller initial savings balances. Since our measure of initial savings balances is imprecise, we cannot currently explore this heteroge- neity in the response to the program. Finally, the account data also enable us to calculate measures of the type and frequency of banking transactions, which we use as a proxy for the intensity of a customer’s banking relationship. 9.4.2 Customer profile: demographics and geography The data set contains the demographic details and locations for all individuals who opened a savings account before September 2012. Table 9.2 lists the vari- ables available for each individual in the data set. Demographic data and location for each individual can be matched to account and transaction data through an anonymized identifier code. An important consideration for our analysis is whether the population of customers at the control bank is sufficiently similar to serve as a useful compar- ison group vis-à-vis the population of borrowers were eligible to participate in ISIW. To verify that this is the case, this section presents a number of summary statistics by bank. We first consider the geographic distribution of the two banks’ population of savings accounts customers. Both banks are large nationwide banks that operate a large network of branches (362 and 325 branches, respectively) in all states and territories of Nigeria. Figure 9.3 plots the branch network of the two banks and shows that the locations served by them are very similar, with the control bank having a slightly more extensive network in central Nigeria. The same pattern 9.  Learning by doing?  ◾ 299 TABLE 9.2  Geographical distribution of customers, by bank and state ICB ACCESS Abia 2.16 1.54 Adamawa 0.31 0.48 Akwaibom 2.78 1.24 Anambra 4.72 2.02 Bauchi 0.53 0.71 Bayelsa 0.82 0.76 Benue 2.67 1.07 Borno 0.66 0.42 Crossriver 1.16 1.1 Delta 2.06 2.27 Ebonyi 0.5 0.68 Edo 4.06 3.96 Ekiti 1.12 1.08 Enugu 4.34 3.23 Federal Capital Territory 4.76 6.3 Gombe 0.64 0.83 Imo 2.56 1.56 Jigawa 0.65 0.11 Kaduna 2.67 1.69 Kano 3.36 0.98 Katsina 0.83 0.37 Kebbi 0.46 0.54 Kogi 1.21 0.8 Kwara 1.32 1.32 Lagos 25.34 45.16 Nasarawa 1.68 0.8 Niger 2.4 1.19 Ogun 5.45 1.37 Ondo 2.42 1.26 Osun 1.67 0.97 Oyo 5.72 3.39 Plateau 1.37 1.29 Rivers 5.51 7.81 Sokoto 0.65 0.81 Taraba 0.21 0.43 Yobe 0.39 0 Zamfara 0.82 0.48 Note: This table summarizes the distribution of savings accounts across Nigerian states for both ICB and Access Bank. 300  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES holds for the percentage of accounts in each geographical region (figure 9.4). The general pattern is similar for the two banks in that the vast majority of accounts is concentrated in the same urban centers (Lagos, Abuja, Kano, Ibadan). Overall, the control bank has a higher fraction of accounts in Lagos and some of the southern Nigerian states, while the treatment bank’s branch network is somewhat more evenly distributed around the country. We next consider the demographic characteristics of bank customers of the treatment and control banks. Table 9.3 shows the breakdown of the two banks’ customers by gender, marital status, and age. The clientele of both banks is about two-thirds male, with the customers of the control bank being slightly younger and, on average, less likely to be married. Overall, the demographic profile of the two banks is quite similar, indicating that differential selection into the client pool of the two banks is unlikely to pose a threat to our empirical approach. 9.4.3 Prize drawings and lottery winners We match account data with information on prize drawings and lottery winners to explore the effectiveness of the savings promotion as a tool to attract new FIGURE 9.3  Branch network of treatment and control banks, January 2011 Sokoto Water body Sokoto Water body Katsina Yobe Katsina Yobe Jigawa Jigawa Zamfara Zamfara Kebbi Kano Borno Kebbi Kano Borno Bauchi Bauchi Kaduna Gombe Kaduna Gombe Niger Niger Adamawa Adamawa Plateau Plateau Kwara Kwara Federal Capi Federal Capi Nassarawa Nassarawa Oyo Oyo Taraba Taraba Ekiti Kogi Ekiti Kogi Osun Osun Benue Benue Ogun Ondo Ogun Ondo Lagos Edo Enugu Lagos Edo Enugu Anambra Ebonyi Anambra Ebonyi Cross River Cross River Delta Delta Imo Abia Imo Abia Rivers Akwa Ibom Rivers Akwa Ibom Bayelsa Bayelsa Note: ICB: N = 362; Access: N = 323. 9.  Learning by doing?  ◾ 301 FIGURE 9.4  Geographical distribution of bank customers by state a. ICB Sokoto Water body Katsina Yobe Jigawa Zamfara Kebbi Kano Borno Bauchi Kaduna Gombe Niger Adamawa Plateau Kwara Abuja Nassarawa Oyo Taraba Ekiti Kogi Osun Benue Ogun Ondo Lagos Edo Enugu Anambra Ebonyi Cross River % of Accounts Delta (2.77,25.48] Imo Abia (1.63,2.77] (.83,1.63] [.38,.83] Bayelsa Rivers Akwa Ibom No data b. Access Sokoto Water body Katsina Yobe Jigawa Zamfara Kebbi Kano Borno Bauchi Kaduna Gombe Niger Adamawa Plateau Kwara Abuja Nassarawa Oyo Taraba Ekiti Kogi Osun Benue Ogun Ondo Lagos Edo Enugu Anambra Ebonyi Cross River % of Accounts Delta (1.43,49.62] Imo Abia (1.06,1.43] (.63,1.06] [0,.63] Bayelsa Rivers Akwa Ibom No data 302  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 9.3  Basic demographic statistics for ICB and Access Bank customers (%) ICB ACCESS MALE FEMALE TOTAL MALE FEMALE TOTAL Marital status Married 41.5 23.2 64.6 39.2 22.2 61.4 Single 22.1 13.2 35.4 25.6 13.0 38.6 Total 63.6 36.4 100.0 64.9 35.1 100.0 Age 0–17 0.06 3.84 18–24 17.46 22.68 25–34 48.38 40.17 35–44 20.38 18.35 45–54 9.05 9.03 55–64 3.41 4.28 65+ 1.26 1.65 Median 30 30 Mean 33.3 32.2 Observations 2,338,679 620,035 Note: In this and future tables, we limit the sample of customers who are likely to be affected by the ISIW promotion. We exclude accounts with balance changes that exceed N = 3 million (about US$20,000) in a single transaction. We also exclude accounts that are inactive over the period October 2010–Feb- ruary 2012. customers to the bank and change behavior among existing bank clients. By matching contest winner data with demographic and geographic data, we will also be examine to examine the local network effects of the lottery. That is, we will be able to estimate the impact of having a winner at a given branch on the account balances and new account openings in that community. Lottery drawings for the ISIW promotion occurred in six phases between July and November 2011. Prize drawings were conducted separately among qualifying bank customers in each of the bank’s six regions, beginning with the Abuja and north central tegions of Nigeria on July 21, 2011, and concluding with the Lagos and northeast Nigeria regions on November 3, 2011. The drawing for the grand national prize took place concurrently with the last regional drawing on November 3, 2011. The data on prize winners contains the name, address, and account informa- tion for all 611 individuals who won prizes in the regional and national rounds of the ISIW lottery. 9.  Learning by doing?  ◾ 303 9.5 METHODOLOGY AND RESULTS 9.5.1 Empirical strategy The introduction of the ISIW promotion and each type of new media campaign (e.g., celebrity endorsement, television commercial, YouTube video, intermediate lottery drawing) was staggered over time from April to October 2011 as shown in figure 9.5. To evaluate the effect of each type of media campaign on borrower behavior, we focus on a narrow time window around the introduction of each type of media launch and perform the following event study: ∆Yi = β 0 + εi (9.1) where ∆Yi is the change in the outcome of interest in the narrow time window around the event, e.g., account balances for customer i, and β 0 is the coefficient on the constant term capturing the effect of the promotion. εi is the error term, which we allow to be clustered at the bank branch level. We include a single observation for each account. Note that, because the outcome is specified in terms of changes in account characteristics, the regression implicitly controls for fixed individual differences affecting the level of account activity. In supple- mentary tests, we also allow for more flexible nonlinear time trends as well as corrections for seasonality in account activity. To account for the possibility that there are other changes affecting savings behavior that happen to coincide with the intervention, we employ a differ- ence-in-difference approach to account for such changes. To do this, we add data from accounts at Access Bank (the control bank) and estimate the following: ∆Yi = β 0 + β1IiICB + β2 Xi + εi (9.2) in which IiICB is a dummy equal to 1 if the customer banks with ICB and equal to 0 if the customer banks with Access Bank. We again focus on a narrow time window around the introduction of each type of media launch, with a single observation for each account. β 0 measures the average change in account activity among Access Bank customers. β1 measures the marginal difference in the average change in account activity for ICB customers relative to Access Bank customers. We also include controls for demographic characteristics that might affect the change in account activity around the event. Note again that because the outcome is spec- ified in terms of changes, the regression implicitly controls for fixed individual differences affecting the level of account activity. As before, we also test specifi- cations allowing for more flexible nonlinear time trends as well as corrections for seasonality in account activity. For the evaluation of the lottery component of ISIW we use the following base specification: 304  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 9.5  Timeline of the ISIW savings promotion April 4: 2011 ISIW is launched April 5: Nigerian football star Segun Odegbami endorses ISIW April 13: YouTube video campaign launched May 6: Facebook page and online message boards launched July 1: Toyota Nigeria endorses ISIW, gives 10 Corollas as prizes July 21: Winners drawn for Abuja and north central regions August 8: Governor of Enugu State endorses ISIW August 8: Winners drawn for southeast region August 25: Winners drawn for south region September 14: Winners drawn for northwest region October 21: Winners drawn for southwest region November 3: Winners drawn for Lagos region November 3: Winners drawn for northeast region November 3: Winners drawn for grand national prize Yirt = β 0 + β1Irt postlottery + β3 Xi + R + εit (9.3) where R is a set of regional dummies, and rt I postlottery is an indicator equal to 1 if the observation corresponds to a region that already had a lottery. The sample is at the account by day level and is restricted to observations corresponding to ICB (the treatment bank). β1 measures the effect of the lottery drawing on savings account activity. Standard errors are clustered at the regional level. 9.  Learning by doing?  ◾ 305 To further explore the effect of lotteries on savings behavior, we exploit branch-level variation in whether customers at the bank won a disproportionate number of lottery prizes or won particularly large prizes, such as cars. We hypoth- esize that other customers banking at branches with major lottery winners may become more aware of the lottery and of the possible benefits associated with banking activities. We use the following specification: Yibt = β 0 + β1Ibt postlottery + β2W  ◻ bt + β3Ibt postlottery ◻ W  bt + B + β4 Xi+ εit (9.4) where B is a set of branch dummies, Ibt postlottery is an indicator for the time period after the lottery drawing, and W  bt is a measure of whether the branch won a ◻ disproportionate number of lottery prizes or was the home of a car lottery winner. Our strategy for evaluating components of ISIW (other than the lottery component) relies on the assumption that Access Bank customers serve as an appropriate control for ICB customers. We believe this is a reasonable assumption for several reasons. Both Access and ICB are among Nigeria’s largest national banks and serve a broad clientele. Like ICB, Access Bank previously ran a savings campaign, although it occurred outside the time frame of our study and had fewer new media elements. This suggests that Access and ICB executives had similar approaches to running their businesses, and merely the timing of their implemen- tation differed. However, as noted previously, we must also confront several important concerns when using Access Bank as a control. First, Access launched a sign-up drive for new accounts shortly after the launch of ISIW, as evidenced by a signif- icant increase in new account openings for Access Bank starting in July 2011. Therefore, we cannot use Access Bank as a control group to examine the effect of ISIW on new account openings. Instead, we explore the effect of the savings promotion on the behavior of customers with accounts opened prior to the launch of the ISIW promotion. Second, ISIW was launched during a turbulent period in which both ICB and Access Bank saw large declines in balances and new account openings just prior to the initiation of the ISIW promotion. Given the general volatility in account activity, one may be concerned that ICB and Access Bank may have reacted in different unobserved ways that directly affected savings behavior. Therefore, in the next section, we present the results using Access Bank as a control, but refrain from drawing strong conclusions. Instead, we focus on the lottery-related results, which use regional and branch-level variation in the timing and outcomes of the lottery drawings. These lottery-related results do not require the use of Access Bank as a control group and are less subject to endog- eneity concerns. 9.5.2 Short- and medium-term impact on savings balances We begin our analysis by examining how the ISIW promotion affected savings balance activity. We find evidence that the initiation of the savings promotion led 306  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES to a small increase in the savings balances of existing account holders. However, this effect was short lived. Moreover, subsequent media campaigns, such as celebrity endorsements and the release of YouTube promotional videos, did not significantly affect savings behavior. Figure 9.6 shows the mean savings account balances at ICB and Access Bank around the initiation of the ISIW promotion in April 2011. To focus on the set of customers most likely to be affected by the promotion, we exclude “high-roller” accounts, defined as accounts with balance changes exceeding N= 3 million (about US$20,000) in a single transaction in our sample period. We also exclude accounts that are inactive over a large window surrounding the launch of the ISIW promo- tion (from October 2010 to February 2012). Comparing the trends for ICB and Access Bank, we do not observe obvious large differential changes in savings balances across the two banks. In unreported results, we also look at mean and median account balances around the initiation of ISIW and around the introduction of subsequent new media campaigns and find similar results. In all cases, new media campaigns did not lead to economically meaningful changes in savings activity. In table 9.4, we explore changes in the savings account balances following the launch of the ISIW promotion in April 2011 using the single-difference and differences-in-differ- ences specifications described in the previous section. The sample is restricted to accounts active prior to April 2011. Outcomes are the change in the month-end account balance relative to March 2011 (the month prior to the ISIW promotion). For example, for event window [−1,2], the outcome is the month-end balance in June 2011 minus the month-end balance in March 2011. Columns 1, 3, and 5 report a single-difference specification in which observations are limited to accounts at ICB. Columns 2, 4, and 6 report a differences-in-differences specification FIGURE 9.6  Mean account balance, by month Naira 26,000 Access Bank 24,000 22,000 2,0000 ICB 18,000 Feb 1, 2011 Mar 1, 2011 Apr 1, 2011 May 1, 2011 Jun 1, 2011 Jul 1, 2011 9.  Learning by doing?  ◾ 307 TABLE 9.4  Pre-post and differences-in-differences estimation, net balance changes 1 2 3 4 5 6 Constant 528.3 *** 720.1 *** 642.4 *** (27.2) (34.6) (42.7) ICB dummy 171.2 284.5** −32.0 (130.2) (136.9) (185.3) State FE N Y N Y N Y Demographic N Y N Y N Y controls Event window [−1,0] [−1,0] [−1,1] [−1,1] [−1,2] [−1,2] (months) Observations 14,701 158,969 14,701 15,896 14,701 15,896 03 9 03 99 03 99 Note: Standard errors, at the bank branch level, in parentheses. Columns 1, 3, and 5 report a single difference specification in which observations are limited to accounts at ICB. Columns 2, 4, and 6 report a differences-in-differences specification which includes observations at both ICB and Access Bank. We do not report the coefficient for the constant term in Columns 2, 4, and 6 because they rep- resented the omitted category after including state and profession fixed effects. which includes observations at both ICB and Access Bank. Demographic controls include gender, marital status, a quadratic in age, and dummies for the employ- ment profession of the primary account holder. We do not report the coefficient for the constant term in columns 2, 4, and 6 because they represent the omitted category after including state and profession fixed effects. We find that account balances at ICB increased in the event window around = 500 (approximately US$3.20). However, at the launch of the ISIW by more than N approximately 2.5 percent of the mean balance, this change is relatively small in economic magnitude. Further, Access account balances also grew in this period, and the marginal difference between ICB and Access is not always statistically significant. In particular, columns 4 and 6 show that ISIW may have had a small significant impact on ICB balances in the month after the ISIW launch, but this effect quickly disappears if the event window is extended to two months. In additional results, not reported here, we conduct similar regression anal- ysis to explore the effect of new media promotions on savings account balances. We do not find significant effects. Finally, we find that adjustments for seasonality do not substantively change our results. In figure 9.7, we test whether the launch of the ISIW encouraged bank clients = 50,000 over a minimum nonconsecu- to maintain a minimum savings balance of N tive period of 90 days to be eligible for the lottery drawings. We limit the sample to saving accounts that were created prior to the launch of ISIW and plot the = 50,000. We find that fraction of all active accounts with balances exceeding N both ICB and Access Bank experienced steady declines in the fraction of accounts = 50,000 in the months preceding the launch of the ISIW with balances exceeding N 308  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 9.7  Fraction of accounts qualifying for lottery, by month Fraction .16 Access Bank .15 .14 ICB .13 .12 .11 Mar 1, 2011 Apr 1, 2011 May 1, 2011 Jun 1, 2011 Jul 1, 2011 promotion. In the months following ISIW, the fraction of accounts qualifying for the lottery remains steady at both banks. This suggests that ISIW did not signifi- cantly encourage ICB clients to maintain minimum account balances. Instead, we believe there were general time trends leading to broad declines in account balances in early 2011, which subsequently abated starting in April 2011. 9.5.3 Impact on use of other financial products Next, we explore changes in the use of other bank products (excluding the main account holder’s primary savings account) following the launch of the ISIW promotion in April 2011. In table 9.5, we again restrict the sample to accounts active prior to April 2011. Outcomes are the change in the rate of product initia- tion in May 2011 relative to March 2011 (the month prior to the ISIW promotion). Each column reports a differences-in-differences specification as described in the previous section. We find that ISIW led to small significant increases in the usage of other products at ICB relative to the same trends at Access Bank. In the short run, ISIW led to a 5 percentage point increase in the use of other bank prod- ucts. Gains in the usage of credit cards and nonsavings products are smaller in magnitude, indicating that customers primarily used other similar products—e.g., a secondary savings account or checking account. This evidence is suggestive that customers may have opened additional accounts to become “doubly” eligible for the savings lottery. In unreported results, we also test whether the staggered release of new media promotions affected the usage of other bank products and find insignif- icant effects. Finally, we explore the long-run effect of ISIW on other product usage and find insignificant effects. 9.  Learning by doing?  ◾ 309 TABLE 9.5  Differences-in-differences estimation, product usage after ISIW launch CHANGE IN PRODUCT USAGE NON-CREDIT ANY PRODUCT CREDIT CARDS CARDS   (1) (2) (3) ICB dummy 0.0521*** 0.0001** 0.0028*** (0.0027) (0.0001) (0.0005) State FE Y Y Y Demographic controls Y Y Y Event window (months) [−1,1] [−1,1] [−1,1] Observations 1,549,686 1,549,686 1,549,686 Note: Standard errors, at the bank branch level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The sample is restricted to accounts active prior to April 2011. Outcomes are the change in the rate of product initiation in May 2011 relative to March 2011 (the month prior to the ISIW promotion). Each column reports a differences-in-differences specification as described in the previous table. 9.5.4 Impact of lottery drawings on savings balances Table 9.6 explores changes in savings account balances and activity following the ISIW lottery drawings. Panel A exploits regional variation in when lottery draw- ings were held. Panel B exploits branch-level variation in whether the branch was the home of winners of the largest prize, the car. Panel C exploits branch-level variation in the percentage of account holders at the branch who won any lottery prize. Observations in panels B and C are limited to the sixth round of lottery drawings (this round was the largest regional drawing, including all branches in Lagos and northeastern Nigeria). We limit the sample to this round of drawings because it is currently the only round for which we can link all lottery winners to home branches. The outcome in column 1 is the daily change in the savings account balance (account balance at day end minus the account balance in the previous day). The outcome in column 2 is an indicator for whether the daily change in savings account balance is positive. The outcome in column 3 is an indicator for whether there was any account activity on each day. We explore these indicator variables because we currently are unable to link lottery winners to individual savings accounts. Therefore, we cannot exclude lottery winners from the specifications in column 1. Lottery winners may directly experience large increases in savings balances, which can affect our estimates. To ensure that this does not bias our results, we focus on columns 2 and 3, which use indicator variables as the outcome variables. In these specifications, a small handful of lottery winners with poten- tially large changes in balances are unlikely to significantly alter the results. All event windows measure calendar days relative to the day of the lottery drawing (day 0). All other variables are as defined in previous tables. 310  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 9.6  Treatment effects, lottery-induced variation DAILY POSITIVE DAILY CHANGE IN NET BALANCE CHANGE DAILY ACCOUNT BALANCES DUMMY ACTIVITY DUMMY   (1) (2) (3) Panel A: Regional variation in the timing of lottery drawings After lottery 15.752 −0.001 0.015 (20.642) (0.003) (0.009) Region FE Y Y Y Demographic controls Y Y Y Event window (days) [−7,7] [−7,7] [−7,7] Observations 25,364,445 25,364,445 25,364,445 Panel B: Branch variation in car lottery winners After lottery X car 4.069 0.010*** 0.011*** (16.837) (0.003) (0.004) Car 21.943** −0.008** −0.009** (10.889) (0.003) (0.004) After lottery −1.244 −0.009** −0.005 (6.520) (0.003) (0.003) Demographic controls Y Y Event window (days) [−7,7] [−7,7] [−7,7] Observations 7,046,625 7,046,625 7,046,625 Panel C: Branch variation in % of accounts winning lottery prizes After lottery X winner pct 1.853** 0.001* (0.834) (0.000) Winner pct −0.785 −0.001** (0.521) (0.000) After lottery −9.543 −0.007 (7.674) (0.004) Demographic controls Y Event window (days) [−7,7] [−7,7] Observations 7,046,625 7,046,625 Note: Standard errors, at the bank branch level, in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Observations in panels B and C are limited to region 6 (the largest region, including Lagos), because of data limitations regarding the identity of lottery winners. The outcome in column 1 is the daily change in the savings account balance (account balance at day end minus the account balance in the previous day). The outcome in column 2 is an indicator for whether the daily change in savings account balance is positive. The outcome in column 3 is an indicator for whether there was any account activity on each day. Event windows measure calendar days relative to the day of the lottery drawing (day 0). All other variables are as defined in previous tables. In panel A, we find that the staggered timing of lottery drawings across regions did not significantly affect savings balances. Nonetheless, the magnitude of the coefficient in column 1 is suggestive. In each of the seven days following the regional lottery draw, savings accounts experienced a N= 16 increase relative to 9.  Learning by doing?  ◾ 311 = 100 the seven days preceding the lottery draw, totaling an average increase of N across all accounts active prior to the lottery drawings. In panels B and C, we find that focusing on local branch-level effects leads to more precisely estimated effects. Customers at branches with a high percentage of lottery winners or at branches with at least one winner of a large prize increased savings balances relatively more following the lotteries. In panel B, we find that banking with a branch associated with a winner of a car leads to significant increases in account activity and in positive changes in the account balances. Relative to the mean proportion of days with a net positive change in account balances of 0.026, the lottery leads to an increase of 0.10 (an increase of 28 percent). Relative to the mean proportion of days with any account activity of 0.067, the lottery leads to an increase of 0.11 (a 16 percent increase). However, we do not find significant effects with regard to mean daily changes in net balances, although this measure may be more subject to outlier bias. In panel C, we find similar effects of the lottery on those banking with branches associated with a high proportion of lottery winners. In general, 71.7 percent of branches had a least one lottery winner, averaging to 6.4 winners per 10,000 customers. We find that a 1 percentage point increase in the propor- tion of branch customers who win the lottery leads to a 3.8 percent increase the likelihood of positive changes in account balances and a 1.5 percent increase in the proportion of days with any account activity. Mean daily changes in net balances also increase. However, we focus on the results in columns 2 and 3, which use indicator variables as the outcome variables, because they are less likely to be influenced by balance changes directly related to lottery winnings. In unreported results, we extend the event windows to 14 days around the lottery drawings and find similar effects. However, medium- and long-run effects several months after the lottery are not significant. 9.6 CONCLUSION The literature on financial education documents that it has often proven difficult to design education programs that can lead to significant changes in financial behavior. This chapter has examined the impact of a large-scale savings lottery as a potential alternative. Instead of teaching welfare-enhancing financial behav- iors, ISIW incentivized savings. In this way, the program aimed to change financial behaviors directly, by “turning a dollar saved into a lottery ticket” that would give the account holder the opportunity to earn additional returns in the form of cash and noncash prizes. We find that the program affected savings balances in the short run, but did not have a significant long-run impact on financial behavior. When comparing bank customers exposed to ISIW with the clients of another bank that did not offer a similar promotion, we find that the initiation of the program led to a temporary 312  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES increase in the savings balances of existing account holders. At approximately 2.5 percent of the mean account balance, this change is small in economic terms, but within the range of estimates found in the literature on financial education (see, e.g., Bruhn et al. 2013; Xu and Zia 2012). The effect of ISIW on savings balances is, however, short lived, and disappears when we extend the time window of our analysis more than one month beyond the introduction of the program. The intro- duction of ISIW also led to a 5 percent increase in the total number of financial products used by customers at the treatment versus the control bank. This result should, however, be interpreted with caution, as it may reflect existing customers opening additional accounts to become “doubly” eligible for the program. Finally, using variation generated by the lottery, we analyze how effective the highly publicized lottery drawings are as a means to incentivize savings. We find that the lottery drawing itself did not lead to significant changes in balances of accounts exposed to the drawing versus branches in regions that did not have a drawing at that date. We do find that the demonstration effect of having a lottery winner at one’s branch led to significant increases in net savings balances at the branch within a seven-day window after the event. Taken together, the results presented in this chapter suggest that savings incentive programs, although a potentially promising channel to build financial capability, appear to suffer from many of the same limitations documented in the literature on financial education: their impact on behavior is relatively short lived and quite small relative to the income and assets of the target population. Despite the fact that prize-linked products have a number of features that could make them an attractive tool to serve as a stepping stone into formal banking relationships, it appears difficult to realize this potential even in widely publicized, large-scale incentive programs such as the one studied here. The fact that we document significant short-run responses to the introduction of the program and the publicity associated with the lottery drawings is sugges- tive of a missed opportunity. Both the bank and potential participants would benefit from a program that could provide a credible long-term incentive to save and establish deeper formal banking relationships. One possible way to do this would be to offer savings accounts with an incentive feature as a regular product, rather than a one-off promotion. Exploring how to best build financial capabilities by integrating such incentive features into financial products is beyond the scope of this chapter, but an exciting topic for future research. REFERENCES ABN Digital. 2011. “Diamond Bank Disburses NGN1bn to SavingsXtra Customers.” August 30. http://www.abndigital.com/page/news/nigeria-market-news/1003497-Diamond- Bank-disburses-NGN1bn-to-SavingsXtra-customers. 9.  Learning by doing?  ◾ 313 Ashraf, Nava, Dean Karlan, Nathalie Gons, and Wesley Yin. 2003. “A Review of Commitment Savings Products in Developing Countries.” Financial Access Initiative. http://www. financialaccess.org/sites/default/files/publications/a-review-of-commitment-savings- products-in-developing-countries.pdf. Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of Economics 121 (2): 635–72. Banerjee, Abhijit V., and Esther Duflo. 2007. “The Economic Lives of the Poor.” Journal of Economic Perspectives 21 (27): 141–67. —. 2008. “What Is Middle Class about the Middle Classes around the World?” Journal of Economic Perspectives 22 (2): 3–28. Bar, Graham, and Peter Collins. 2006. “The National Prevalence Study 2006: Gambling and Problem Gambling in South Africa.” htttp://www.responsiblegambling.co.za. Besley, Timothy, Stephen Coate, and Glenn Loury. 1993. “The Economics of Rotating Savings and Credit Associations.” American Economic Review 83 (4): 792–810. Bruhn, Miriam, Arianna Legovini, and Bilal Zia. 2013. “Financial Literacy for High School Students and Their Parents: Evidence from Brazil.” Unpublished. Brune, Lasse, Xavier Giné, Jessica Goldberg, and Dean Yang. 2011. “Commitments to Save: A Field Experiment in Rural Malawi.” Policy Research Working Paper 5748, World Bank, Washington, DC. Carpena, Fenella, Shawn Cole, Jeremy Shapiro, and Bilal Zia. 2011. ”Unpacking the Causal Chain of Financial Literacy.” Policy Research Working Paper 5798, World Bank, Washington, DC. Cole, Shawn A., Thomas Sampson, and Bilal Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Cole, Shawn A., Peter Tufano, Daniel Schneider, and Daryl Collins. 2008. “First National Bank’s Golden Opportunity.” Case 208-072, Harvard Business School, Cambridge, MA. Collins, Daryl, Jonathan Morduch, Stuart Rutherford, and Orlanda Ruthven. 2009. Portfolios of the Poor: How the World’s Poor Live on $2 a Day. Princeton: Princeton University Press. De Waal, Mandy. 2012. “National Savings ‘Lottery’—An Idea Whose Time Has Come. Again.” Daily Maverick April 24. http://www.dailymaverick.co.za/article/2012-04-24-national- savings-lottery-an-idea-whose-time-has-come-again/#.U2K1K8e59q4. FANews. 2011. “South Africa’s Savings Rate Lags behind Other Emerging Markets.” July 15. http://www.fanews.co.za/article/life-insurance/9/investing-saving/1084/south-africa- s-savings-rate-lags-behind-other-emerging-markets/10307. Kamewe, H. T., and I. A. Radcliffe. 1999. “Reviving Postal Savings Banks in East Africa.” Laibson, David. 1997. “Golden Eggs and Hyperbolic Discounting.” Quarterly Journal of Economics 112 (2): 443–78. Lobe, Sebastian, and Alexander Hölzl. 2007. “Why Are British Premium Bonds So Successful? The Effect of Saving With a Thrill.” University of Regensburg Working Paper, Regensburg, Germany. Lusardi, Annamaria, Punam Anand Keller, and Adam M. Keller. 2009. “New Ways to Make People Save: A Social Marketing Approach.” NBER Working Paper 14715, National Bureau of Economic Research, Cambridge, MA. Mullainathan, S., and E. Shafir. 2009. “Savings Policy and Decision-Making in Low-Income Households.” In Insufficient Funds: Savings, Assets, Credit, and Banking among Low-Income Households, edited by R.  M. Blank and M.  S. Barr, 121–45. New York: Russell Sage Foundation Publications. 314  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Murphy, Anne L. 2005. “Lotteries in the 1690s: Investment or Gamble?” Financial History Review 12 (2): 227–46. O’Donoghue, Ted, and Matthew Rabin. 1999. “Doing It Now or Later.” American Economic Review 88 (1): 103–24. Pfiffelmann, Marie. 2009. “Savings and Lottery: An Historical Approach (Le mariage efficace de l’épargne et du jeu: une approche historique).” Working paper, Strasbourg University, Strasbourg. Prina, Silvia. 2011. “Do Simple Savings Accounts Help the Poor to Save? Evidence from a Field Experiment in Nepal.” http://www.econ.yale.edu/conference/neudc11/papers/ paper_317.pdf. Rutherford, Stuart. 2000. The Poor and Their Money. New Delhi: Oxford University Press. Thaler, Richard H. 1990. “Anomalies: Saving, Fungibility, and Mental Accounts.” Journal of Economic Perspectives 4 (1): 193–205. —. 1999. “Mental Accounting Matters.” Journal of Behavioral Decision Making 12: 183–206. Tufano, Peter. 2008. “Saving whilst Gambling: An Empirical Analysis of UK Premium Bonds.” American Economic Review 98 (2): 321–26. Tufano, Peter, Nick Maynard, and Jan-Emmanuel De Neve. 2008. “Consumer Demand for Prize-Linked Savings: A Preliminary Analysis.” Working Paper 08-061, Harvard Business School, Cambridge, MA. WorldStage. 2011. “Odegbami Endorses Intercontinental Bank’s I Save I Win Promo.” April 8. http://worldstagegroup.com/worldstagenew/index.php?active=page&id=2557 &pgcat=news&start=745. Wright, Graham. 1999. “A Critical Review of Savings Services in Africa and Elsewhere.” Centre for Microfinance. Xu, Lisa, and Bilal Zia. 2012. “Financial Literacy around the world: An Overview of the Evidence with Practical Suggestions for the Way Forward.” Policy Research Working Paper 6107, World Bank, Washington, DC. CHAPTER 10 N igeria’s Nollywood nudge An entertaining approach to saving AIDAN COVILLE, VINCENZO DI MARO, SIEGFRIED ZOTTEL, AND FELIPE ALEXANDER DUNSCH ABSTRACT Can edutainment be an effective tool to strengthen financial inclusion? In collaboration with a local nongovernmental organization (Credit Awareness) and a microfinance bank (Accion), we explore the short- and medium-term savings decisions of a group of microentrepreneurs in Lagos, Nigeria, by inviting business owners to one of four randomly allocated events: a movie screening of The Story of Gold —a Nollywood (the Nigerian version of Holly- wood) film encouraging entrepreneurs to save responsibly; an event where business owners are shown a “placebo” screening of a movie with no finan- cial education content and offered on-the-spot microsavings accounts through Accion; a combined event, screening The Story of Gold and offering on-the-spot accounts; and a screening of the placebo film only as our control group. We find that entrepreneurs watching The Story of Gold were 5 percentage points more likely to open a savings account on the spot than those in placebo screenings, and this effect was mostly driven by male busi- ness owners. In contrast, less than 1 percent of entrepreneurs who were not offered an on-the-spot opportunity signed up for a savings account after the screening. In the longer run, only moderate changes in attitudes and percep- tions were found, while savings and borrowing behavior was unchanged four We are thankful for the financial support and guidance of the Russia Financial Literacy and Education Trust Fund. We would also like to thank Marcus Holmlund for his comments on previous drafts, as well as Billy Jack and Florentina Mulaj for technical inputs on the evaluation design. Finally, we are grateful to Segun Adebamiji, Edwin Daniels, Nneka Eneli, and Ladi Smith for implementation support. All views expressed in this chapter should be considered those of the authors alone, and do not neces- sarily represent those of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 315 316  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES months after the screening. This suggests that, while influencing short-term deci- sions is possible, longer-run behavior is far less malleable through one-off events. This chapter contributes to the literature by directly testing the importance of linking emotional stimulus to financial messages in order to influence short-term savings decisions and identifying the important interaction between emotional stimulus and the opportunity to act on this stimulus. 10.1 BACKGROUND Traditional rational agent economic models rely on the assumption that people make decisions based on a rational and deliberate consideration of all costs and benefits associated with the action, conditional on available knowledge. However, low-income individuals regularly make seemingly suboptimal financial decisions, and there are strong correlations between financial knowledge, sound financial decisions, and the use of financial products (e.g., Hilgert, Hogarth, and Beverly 2003). This has led to a growing body of literature exploring the importance of providing financial education and training to individuals and entrepreneurs to effectively improve knowledge, leading to improved financial capabilities and decisions. Despite strong correlations (e.g., Lusardi 2007), rigorous causal impact evaluations of financial literacy training programs have shown mixed results, often with little to no effect on actual behavior (e.g., Cole, Sampson, and Zia 2011) or with positive impacts only through resource-intensive interventions (see, e.g., Bruhn, Ibarra, and McKenzie in chapter 7 of this volume). These limited effects could be explained by (1) only small increases in actual knowledge, or (2) the fact that people do not fully apply this knowledge when making financial decisions such as when and how much to save. Evidence from psychology and behavioral economics highlights the fact that people act within “bounded rationality,” often relying on heuristics to simplify their choices. Kahneman (2003) presents a frame- work that differentiates between two states that drive human decision making: intuition and reasoning. Decisions based on intuition are “fast, automatic, effort- less, and often emotionally charged,” whereas reasoning is “slower, effortful, and deliberately controlled” (Kahneman 2003, 1451). He argues that most decisions are based on intuition, where reasoning acts as a safeguard, rather than moti- vator, of many behaviors. This insight has important potential implications on how best to influence financial behavior. Even when people are fully aware of the most appropriate action to take, cognitive biases and heuristics may prevent this knowledge from translating into action. Thus, the traditional causal framework linking improved financial knowledge to changes in awareness, perceptions, atti- tudes, and behavior may underestimate important psychological barriers to finan- cial inclusion that weaken the suggested causal chain. Acknowledgment that we base many decisions on heuristics rather than full information helps to explain why, for instance, “rule-of-thumb” approaches to financial education can be more 10.  Nigeria’s Nollywood nudge  ◾ 317 effective at changing behavior than teaching more detailed accounting principles (Drexler, Fischer, and Schoar 2012). This evaluation explores the effectiveness of mass and social media in deliv- ering financial messages in order to induce behavior change beneficial to recipi- ents. Specifically, building on the evidence that emotions and heuristics are likely to influence decisions, this study explores the effectiveness of using a Nollywood movie, The Story of Gold, to relay a simple message of safe saving and responsible borrowing through an emotionally charged storyline to a group of 2,938 micro- entrepreneurs in Lagos, Nigeria. By intertwining the main message of respon- sible financial behavior into an accessible, entertaining, and relatable story about twin sisters trying to succeed in business, the movie appeals to emotion, without providing specific information related to common measures of financial literacy such as understanding interest rates and inflation. The underlying assumption is that a movie loses its entertainment value when people start explaining how to calculate risk-adjusted returns to investments. The Story of Gold is a one-off event aiming to influence transient emotions and lower transaction costs. However, responsible saving is a long-term commit- ment requiring continued and deliberate effort. The objective of the study was to identify whether this one-off event could spur action (in our case, opening a microsavings account) and serve as a catalyst to build financial capabilities through direct and continued exposure to financial institutions and products. The theory of behavioral consistency—where actions based on transient emotions have been identified to influence later decisions derived from people’s desire to be consistent with previous actions—justifies the possible effectiveness of this “foot-in-the-door” hypothesis, but there is limited evidence on how this might influence savings behavior in particular.1 Hence, shedding some light on whether and how interventions that work through affecting perception and emotions in the short term can produce change in behavior and commitment in the longer term is an important empirical topic. The study uses a 2x2 randomized factorial design to exogenously vary (1) exposure to The Story of Gold and (2) access to financial products by offering free on-the-spot microsavings accounts through a microfinance bank (MFB) at selected screening events. Through this framework, we are able to test the rela- tive effectiveness of (1) using “edutainment” (i.e., education through entertain- ment) to motivate action, (2) reducing access constraints to financial products, and (3) the interaction of these two. We find that entrepreneurs in all three treatment arms increase self-reported trust in MFBs, but the treatment arms including The Story of Gold had a larger 1  More generally, this can be related to the path-dependence principle in economics and sociology (Pierson 2000). 318  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES effect on male self-reported trust. The combination of the movie with the pres- ence of an MFB to help facilitate the opening of a savings account (at the time of the screening) was substantially more effective in motivating business owners to open an account than the presence of an MFB combined with a placebo screening—and this was most effective for influencing male decisions, increasing savings account sign-up rates from 1 percent to 11 percent. Four months after the event, we find limited or no sustained impacts on perceptions of MFBs and intention to borrow and save, and no effect on the likelihood of having a savings account (we find that many of the business owners who opened an account at the screening already had a savings account, resulting in this null effect). This suggests that, even with relatively low-budget productions, it is possible to use entertainment to motivate action in the short term, but long-term behavior is less malleable.2 Furthermore, having a direct opportunity to act in the moment may significantly increase the impact of edutainment activities that influence transient emotions. Care needs to be taken when developing the choice architec- ture designed to nudge people toward more “optimal” financial decisions, as this may induce unexpected behavior leading to further suboptimal outcomes. The rest of the chapter is structured as follows: in section 10.2, we explain our rationale to test edutainment—in contrast to more standard financial education programs—as a means to change savings behavior. In section 10.3, we describe the interventions; sections 10.4 and 10.5 provide an overview of the identifica- tion strategy, sampling, baseline balance, and attrition. Section 10.6 presents the econometric framework for analysis. Section 10.7 presents results, with robust- ness checks included in section 10.8. We provide a discussion and conclude in section 10.9. 10.2 A NUDGE FOR BETTER SAVINGS OUTCOMES? This section explains the reasoning behind this chapter’s approach to test enter- tainment media to nudge savings behavior. It first presents the state of poor financial literacy and access to finance in Nigeria. We then show that traditional financial education programs have mostly failed to deliver results to ameliorate this condition. We next argue that psychological biases might partly cause this inefficient savings behavior, and that they cannot be overcome by learning about the right way to do things alone. We show how to make existing biases work in favor of sound financial decision making, “work[ing] around human nature to help people save as they aspire to” (Karlan, Ratan, and Zinman 2013).3 We then present 2  This could indicate that commitment savings accounts might be necessary to solidify longer-term behavior. 3  See, e.g., Sunstein and Thaler (2003) for a discussion of libertarian paternalism. 10.  Nigeria’s Nollywood nudge  ◾ 319 how edutainment has previously been used to aim at these biases to transform behavior. Lastly, we briefly describe Nollywood and its potential to serve as a vehicle to spread messages broadly. 10.2.1 Financial literacy and access to finance in Nigeria Although improvements have been registered in the last three years, 46 percent of the Nigerian population remains financially excluded, with no access to formal or informal financial services.4 This compares unfavorably to countries such as Kenya and Botswana (33  percent); in South Africa, only one-quarter of the population is financially excluded. Only 25  percent of Nigerian’s population has a formal savings account, excluding 66 million adults. The use of MFB accounts is even less widespread, with only 4.6  percent of the adult population having a savings account with an MFB. This lack of access is not derived from a lack of interest or demand. According to recent survey results, almost 75 percent of the unbanked population in Nigeria report that they would like to have a bank account, and over 80  percent of the population receives financial advice from family and friends. In theory, saving helps individuals and businesses by enabling consumption smoothing for volatile incomes, serving as insurance for the poor, growing investments, and allowing better access to microfinance (e.g., Deaton 1989; Karlan, Ratan, and Zinman 2013). However, “…very few people possess the extensive financial knowledge conducive to making and executing complex plans” (Lusardi and Mitchell 2013). But knowledge and acting on knowledge are two different concepts, and individuals often make poor financial decisions— even when better options are readily available (Pathak, Holmes, and Zimmerman 2011; Willis 2011), and even when they express the desire to act differently (Thaler and Benartzi 2004). Building financial capacity in Nigeria represents a big step in helping consumers acquire the skills and knowledge to be capable, confident, and self-reliant when making financial decisions. Evidence on the best way to build this capacity is, however, lacking. It is within this context that the World Bank has worked closely with the Central Bank of Nigeria to develop and implement the World Bank–funded Micro, Small, and Medium Enterprises project to test innova- tive consumer education programs such as the one evaluated here.5 4  Results presented here are based on a recent nationally representative survey of 20,000 consumers conducted by EFiNA in 2010, http://www.efina.org.ng/our-work/research/ access-to-financial-services-in-nigeria-survey/ (accessed April 23, 2014). The project financed the production of the film, but Credit Awareness was responsible for 5  both overseeing this production and the subsequent roll-out. 320  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 10.2.2 Financial education and business training programs In order to improve financial decision making, a common strategy is to offer finan- cial or business training. Evidence on the impact of these programs is mixed. While financial literacy is correlated with household well-being (Mulaj and Jack 2012) and less financial decision-making errors (Lusardi and Tufano 2009; Stango and Zinman 2009) research results do not fully support a causal chain leading from financial education to higher financial literacy and subsequently improved behavior (Duflo and Saez 2003; Willis 2011).6 Financial literacy may therefore be a secondary or even tertiary determinant of individual financial behavior (Cole and Fernando 2008). Intensity, exposure, quality, and training content also vary widely (Drexler, Fischer, and Schoar 2012). Willis (2011) argues that effective financial education would need to be “extensive, intensive, frequent, mandatory, and provided at the point of decision making, in a one-on-one setting, with the content personalized for each consumer.” Also, participation levels for voluntary financial education programs are “extremely low,” even for very short courses (see chapter 7). This presents a concern regarding the power of the analysis; but more broadly, not attending the courses might be an expression of economically optimal behavior by the potential recipient, reflecting the poor perceived efficacy of these programs.7 The poor results of traditional education programs made us think about alternative interventions such as making use of existing behavioral biases to change detrimental behavior. 10.2.3 Bounded rationality A large body of literature from the fields of psychology and behavioral economics attempts to shed light on the fact that individuals often make irrational decisions or “mistakes” (being limited by “bounded rationality”), even when they know better. To present a framework of this bounded rationality, Kahneman (2003) introduces the “architecture of cognition,” distinguishing two models of thinking and deciding, broadly (and metaphorically) summarized as intuition—System 1— and reasoning—System 2: The operations of System 1 are fast, automatic, effortless, associative, and often emotionally charged; they are also governed by habit, and therefore difficult to 6  An increase in knowledge does not necessarily change attitudes and habits, also among more educated populations (Thaler and Benartzi 2004). In their literature review, McKenzie and Woodruff (2012) come to the conclusion that many 7  impact evaluations of training programs are inconclusive due to technical shortcomings such as heterogeneity in length, content, and types of firms participating. Many studies are underpowered, with hurried follow-up surveys (within one year of the training) covering small sample sizes, making it difficult (or impossible) to detect long-term effects. They also suffer from attrition and measurement problems of relevant business indicators. 10.  Nigeria’s Nollywood nudge  ◾ 321 control or modify. The operations of System 2 are slower, serial, effortful, and deliberately controlled; they are also relatively flexible and potentially rule-gov- erned… (Kahneman 2003, 1451–52). The two systems can provide crucial insights on how to influence financial decision making. If System 1 mainly drives financial behavior (intuition), models aiming to affect behavior through System 2 (reasoning) such as information campaigns or business training, assuming a “rational agent of economic theory” (Kahneman 2003), might prove to be ineffective (which is supported by some evidence; see, e.g., Cole, Sampson, and Zia 2009).8 10.2.4 Accessing System 1 References (such as expectations, emotional and motivational arousal, and other phenomena) can increase the accessibility of thoughts that are important for decision making (Andrade and Ariely 2009). Loewenstein and Lerner (2003) argue that even small “primers” can influence behavior, even when this “priming” is unnoticeable by the stimulated individual.9 In the field of marketing, Bertrand et al. (2010), for example, find that “persuasive” advertising can play a significant role in decision making, even if the content of the advertising is not directly related to the product being sold. There are different kinds of references applicable to our setting, as discussed below. THE “AFFECT HEURISTIC” People tend to base decisions that are being taken now on past decisions (uncon- sciously), shortcutting the thought-intensive System 2 process of deliberately evaluating the pros and cons of the respective decision at hand. They also base decisions on whether they like something, rather than carefully evaluating bene- fits and disadvantages (Slovic et al. 2007), answering a difficult question (what are the pros and cons?) by answering the easier question instead (how do I feel about it?)—a cognitive shortcut, where intuition (which resembles perception) acts as a substitute for reasoning (Kahneman 2003). Advertising professionals often make use of these phenomena by focusing on conveying a good feeling about their product to their audience rather than stressing the beneficial effects of a purchase. 8  Kahneman, e.g., argues that the assumption that deciders evaluate outcomes by the utility of final asset positions is “easily” proven to be wrong. Willis (2011): “Decisions can be strongly affected by even transitory emotions related to 9  nothing more than the weather.” 322  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES BEHAVIORAL CONSISTENCY Another important heuristic is the tendency to behave consistently with previous decision making (Cialdini, Trost, and Newsom 1995). Although the incidental effect of emotions might be short lived, the influence of mild incidental emotions can last longer than the emotional experience itself (Andrade and Ariely 2009). Goldberg, Lerner, and Tetlock (1999), for example, illustrate the effects of an anger-inducing film on subsequent—unrelated—actions. Decisions based on a short-lived inci- dental emotion can develop the foundation for future choices and hence outlive the original cause (the emotion) for the behavior (Andrade and Ariely 2009). Retrospec- tively, people tend to identify their past choice as an expression of their past prefer- ence (Schwarz and Clore 1983); in reality, thoughts and actions are rather intuitive most of the time (as argued in Kahneman 2003). In this manner, initial emotions serve as an “anchor” for later decisions (Tversky and Kahneman 1974), reinforcing behavioral consistency.10 Similarly, hypothetical commitment carries over to real decisions if they are presented later (Ariely, Loewenstein, and Prelec 2003).11 KNOWLEDGE AND TRUST An initial reference or action can have longer-lasting effects by fostering cooper- ative behavior based on knowledge and trust in the institution generated through repeated interaction (Mailath and Samuelson 2006). Once the initial burden of interacting in a new environment is overcome, subsequent interactions might become easier, as benefits become more salient. Following this rationale, expo- sure to media that induce emotions can trigger an initial action, providing a “foot- in-the-door,” which may influence later actions (Freedman and Fraser 1966). EMOTIONS AND DECISIONS: GENDER DIFFERENTIALS A sizable body of research looks into the question of whether emotions show differential gender effects on risk preferences, social preferences, and compet- itive preferences. Harshman and Paivio (1987) review evidence on studies showing that women experience emotions more strongly than men. Women are often more risk averse (Croson and Gneezy 2009; Sunden and Surette 1998) and tend to save more conservatively than men (Hinz, McCarthy, and Turner 1997).12 However, Finucane et al. (2000) find gender differences only for whites (“white The so-called “sunk cost fallacy” or the “endowment effect” are related concepts. People 10  have a hard time to correct previous actions by realizing financial losses, consequentially making things worse (Arkes and Blumer 1985; Thaler 1981). 11  Other relevant studies on past decisions affecting the present include Ottati and Isbell (1996) and Pocheptsova and Novemsky (2010). Inability to determine who makes the financial decisions in a household is a potential 12  problem for the validity of these results. 10.  Nigeria’s Nollywood nudge  ◾ 323 male effect”), which hints at cultural biases causing gender differences. Brought together, the literature suggests that gender differentials tend to be context (and culture) specific with few clear and unambiguous traits across population groups and activities. 10.2.5 Edutainment and behavior change Drawing from the above-mentioned studies and findings, the question arises as to whether (1) commercial entertainment media could be used to combine infor- mation (education) delivery with (2) behavioral treatment arms, such as nudges, varying choice architecture, and/or emotional stimulation. Could combining the two perhaps help improve literacy levels and, at the same time, overcome some of the psychological barriers that stimulate bad behavior? While commer- cial media have for a long time been associated with effective changes in social behavior (both positive and negative), they have rarely been used in the field of finance. In other sectors, such as health and education, these tools have been used with success for a long time. For instance, as Brazil’s Rede Globo network grew through the 1970s and 1980s, women also began having fewer children, experiencing the same decrease in fertility as with two extra years of education (La Ferrara, Chong, and Duryea 2012). While using mass media to transmit educational messages is not a novel approach, using edutainment to improve financial capabilities is less explored. The telenovela Nuestro Barrio is a prominent example from the United States aimed at Hispanic immigrants, where research found that it successfully conveyed the importance of formal bank accounts to the largely underbanked community (Spader et al. 2009). Most recently, a World Bank–supported study evaluated the impact of a South African soap opera with financial messages (Scandal!). The study made use of an encouragement design to compare outcomes between a randomly selected group that watched Scandal! and another group that watched a “placebo” show without financial education content. Watching Scandal! resulted in higher financial knowledge scores, increased borrowing from formal sources, and decreased the likelihood of entering into hire purchase agreements (see chapter 11). Edutainment, as an alternative to more formal classroom learning, has the potential to be distributed more widely at lower marginal costs and may appeal to a broader base, reaching out to people who may not otherwise be interested in the topic. By creating emotional connections to the characters and the storyline, the process is believed to help internalize and operationalize the learning. Since this is a relatively new approach in the field of finance, there is a need for rigorous evaluation of these programs to assess the extent to which entertainment media are indeed effective in changing individuals’ financial behavior. In particular, one question is about the role of edutainment through a one-off event (as is the case for The Story of Gold ) as opposed to continued exposure to the message (as in 324  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES the case of the soap operas mentioned above) that could make the emotional connections much stronger. 10.2.6 Nollywood Movies from the Nigerian film industry penetrate almost all households in Nigeria—and across much of Africa—making them the ideal platform to deliver edutainment content. Although producing relatively low-budget films, Nollywood is now the second largest movie industry in the world in terms of production, only trailing India’s Bollywood, with an output of about 200 films every month. The industry is also the second largest employer in Nigeria, after the govern- ment. Films are largely made for home consumption rather than for the bigger cinema screenings. The stories told put fundamental human emotions and strong narratives front and center: love, hate, envy, upward mobility, urban culture, and witchcraft. Due to their ubiquity, movies have the potential to reach large audi- ences with ease, surpassing traditional ways of conveying messages. Even poli- ticians have understood the potential of these movies, posing with their stars at rallies and events. President Goodluck Jonathan recently announced support for aN= 3 billion facility to support the Nigerian movie industry (Vanguard 2013). With financial and political backing, together with large demand, Nollywood provides a unique opportunity to disseminate knowledge and build a culture of responsible financial decision making, reaching out to otherwise marginalized communities. 10.2.7 Application Under the assumption that System 1 is a driver of many financial decisions and accessibility and “narrow framing” (Kahneman and Lovallo 1993) and references are indeed important, The Story of Gold was developed to place more weight on intuition than reasoning to influence decision making.13 The movie seeks to address System 1 in order to encourage behavior change by promoting the take-up and use of savings accounts in the short term and encourage sustained use by building experience (offering a foot in the door) and promoting longer-term behavioral consistency with the original action. Thus, while the Nollywood movie could possibly also augment knowledge and aware- ness that in turn leads to better reasoning, the main intention of using the movie is to target business owners’ intuitive behavior by influencing emotions, making relevant thoughts more accessible—especially when coupled with the immediate availability of signing up for savings accounts after the screening (reduction of transaction costs). Kahneman (2003) stresses the point that preferences of System 1 are shaped by emotions 13  of the moment and need not be internally coherent or reasonable. The preferences of Systems 1 and 2 therefore do not have to be consistent. 10.  Nigeria’s Nollywood nudge  ◾ 325 10.3 DESCRIPTION OF INTERVENTION The Story of Gold is a feature-length Nollywood movie produced and distributed by Credit Awareness, a local nongovernmental organization promoting “safe savings and responsible borrowing.”14 It tells the story of identical twin sisters in Nigeria. Although identical in appearance, the decisions they make when faced with different financial choices affect their lives as well as those around them and ultimately lead them down different paths, one making sound financial deci- sions and succeeding in business and the other falling into a debt trap. The movie aims to impress upon low-income individuals with limited formal education the importance of saving with a formal financial institution and borrowing respon- sibly. Focusing on this simple message and highlighting the repercussions of poor financial decisions, The Story of Gold focuses on the heuristic and emotional elements of human decisions to promote a stronger savings culture, facilitated by Credit Awareness. A partner MFB, Accion, participated in selected screening events and briefly presented its main savings and borrowing products after the show.15 It then provided all the necessary paperwork for participants to open a “Brighta Purse” business savings account on the spot if they were interested in doing so. The microsavings account is geared toward microentrepreneurs as an entry savings and transaction account, requiring no initiation fees (although a minimum balance of N= 500 is needed—one-third of average daily profits from our sample of entrepreneurs). Interest in this savings account is then a function of the amount of savings held. If entrepreneurs expressed an interest in opening an account but did not have the opening balance on hand, they could sign up their names and contact details and follow up with Accion at a later date to confirm the account opening. In this case, the combined intervention aimed at simultaneously encouraging people to save through the movie’s message while reducing access barriers almost to zero with the presence of the MFB at the screening events. The hypothesis was that the movie would serve to inform, and also motivate business owners to act and open a new savings account. The motivational effect of the movie was expected to wear off soon after the screening; giving business owners the opportunity to act in the moment may increase the potential for this short- term motivation to translate into action. By overcoming these barriers to formal financial participation, the study could then explore whether this engagement resulted in longer-term interactions, leading to improved use of financial products over time. While Credit Awareness plans to roll out the screening events across the country, the evaluation focused on a series of early pilot screenings to test the 14  http://www.creditawarenessnigeria.com/home.php (accessed April 23, 2014). 15  http://www.Accion.org/our-impact/nigeria (accessed April 23, 2014). 326  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES modality and learn before scale-up. The pilot screenings were conducted at local community halls in the Ikotun region of Lagos—home to a sprawling street market. The typical screening event would be held in a hall, with local traders invited to attend. The event lasted approximately three hours, starting with a brief introduc- tion, the screening of the movie, and an open discussion after the event to reflect on the story’s core messages. This would be followed by engagement with the MFB. For the purpose of the evaluation, two extra elements were included to the standard Credit Awareness model: (1) to ensure compliance with the assignment strategy, each participant received a personalized invitation with a photograph to confirm his or her identity; and (2) to improve participation rates, a lottery was held at the end of the event in which participants could win spot prizes. 10.4 SAMPLING AND IDENTIFICATION STRATEGY Two community halls large enough to hold 200 people were identified in the Ikotun area of Lagos. A radius of 2 kilometers was used to set the boundaries to ensure that all participants could easily access the halls without needing to use public transport. A census of the area was then taken in July 2012, together with a short baseline listing questionnaire used to stratify the sample as to whether they had a savings account, whether they kept financial records, and if their store was in the main (official) market area or in the surrounding Lagos streets. In total, 2,938 microentrepreneurs were recorded with geopositioning and photographs to confirm their identity in follow-up interactions and verify intervention compli- ance (see figure 10.1 for an example of the invitation created from this infor- mation to verify identity at the event). The criterion used for selection into the sample was being the owner/operator of a business operating within the study area. These businesses were then randomized into one of five groups: (1) pure control, (2) placebo screening, (3) The Story of Gold screening (Movie), (4) placebo screening plus presence of MFB (MFB), and (5) The Story of Gold screening plus presence of MFB (Movie/MFB). The pure control group was not invited to attend any screening. The other four groups were invited to attend one of eight screenings (two per group). Invi- tations were delivered one week before the screening, and two screenings took place every Thursday during September 2012 for four weeks. Invitations to each screening were identical, and events were held at the same time each week (8 –11 a.m.), chosen because the cleaning of the market took place at this time, ensuring low opportunity costs to participation since businesses were not allowed to trade during this time. This uniformity of invitations and event dates was used to mini- mize the possibility of differential take-up across screening events. In placebo screenings, people were shown a Nollywood movie that had no financial messages associated with it, but were given a brief talk after the event about the importance of hygiene in markets to provide quality products 10.  Nigeria’s Nollywood nudge  ◾ 327 FIGURE 10.1  Sample invitation and services. This was done explicitly to control for the event effect of having received a personalized invitation and participation in a big screening event possibly confounding results, and also to create a comparable group of compliers in both treatment and control groups to simplify the analysis. The standard Credit Awareness program (screening The Story of Gold and interacting with an MFB) was split in order to differentiate the impact of the movie from the increased access of financial products coming from the MFB’s presence. As such, a 2x2 factorial design was implemented for the treatment arms in order to detect the differential impact of each component and the interaction effect relative to the placebo screening. In total, 1,261 people (60 percent of those invited) attended the movie screen- ings; a short questionnaire was administered at the end of the event to measure perceptions and attitudes about savings, borrowing, and MFBs. Administrative records were kept at the MFB and Movie/MFB events to record the people who (1) engaged with Accion to open an account at a later stage and (2) actually opened an account at the event. Four months later, in February 2013, a follow-up survey was conducted on all baseline respondents to collect longer-term data on attitudes, intentions, and behaviors with respect to saving and borrowing activities to assess the longevity of any impacts identified at the screenings. 328  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 10.5 OUTCOME MEASURES, BASELINE BALANCE, AND ATTRITION 10.5.1 Outcome measures The main outcome measures are aligned with the essential messages of the Nolly- wood movie. They can be divided into four categories that capture (1) percep- tions of MFBs, (2) perceptions of women, (3) intentions to save or borrow, and (4) savings and borrowing behavior. Regarding the perceptions of MFBs, the survey asked the microentrepreneurs if they agree or disagree with statements such as, “I would trust an MFB to keep my money safe,” “MFBs treat people with respect,” and “If I apply to an MFB for a loan, my application will be accepted.” Since the movie focused on female entre- preneurs as the main protagonists, we also explore self-reported perceptions of female business competence and access to financial opportunities. Questions designed to explore perceptions of women as business owners or financial deci- sion makers ask respondents if they agree or disagree with statements such as “Women can run businesses just as well as men,” “Women make better financial decisions than men,” and “It is easier for men to receive loans than for women.” The intention to save or borrow questions capture whether respondents agree with statements such as “I plan to apply for a loan in the next six months” or “I will save some money next month.” Self-reported savings and borrowing behavior is captured through responses to questions such as “I saved money last month,” the amount of total savings relative to the monthly income earned, savings kept at MFBs, savings at commercial banks, outstanding loans from commercial banks, MFBs, suppliers, moneylenders, or family/friends. Actual savings behavior is measured through administrative records of those who engaged with represen- tatives of Accion to open an account and those who actually opened an account at the screening event. Neither financial knowledge nor basic numeracy skills were specifically addressed in the movie’s storyline. Nevertheless, the survey also included six quiz- like questions with true and false choices to assess respondents’ understanding of basic financial concepts as well as their numeracy skills. The underlying motivation for including these questions is that economic models of savings and investment choice consider both as indispensable for good financial decision making (Lusardi and Mitchell 2013). In particular, respondents were required to do simple division, perform basic calculations related to interest rates, identify the better bargain among two different savings and loan products, and demonstrate their under- standing of how inflation affects their savings. Lastly, one question aimed to eval- uate respondents’ know-how in successfully interacting with financial institutions (awareness of required documentation for being able to open an account). 10.  Nigeria’s Nollywood nudge  ◾ 329 Since single questions provide a rather incomplete picture of respondents’ levels of financial knowledge, an arithmetic financial knowledge score ranging from 0 to 6 was calculated by summing up the correct answers to these six ques- tions. To reflect the level of difficulty associated with each question, an alternative financial knowledge score has been developed, which weights every question with the inverse of the proportion of respondents who were able to provide a correct answer. Larger weights are given to questions that fewer people answered correctly. 10.5.2 Baseline balance Table 10.1 reports summary statistics for the entire sample, as well as for each of the five assignment groups for all exogenous variables including information from the baseline listing and time-invariant variables measured at follow-up. Results are thus reported on balance for business owners who were included in both the baseline and follow-up surveys (n = 2,357). The microentrepreneurs comprising the total sample are on average 38 years old, predominantly female (71 percent), married (84 percent), Christian (64 percent), and able to speak English (70 percent); they completed high school as their highest level of education (50 percent), and live in households with an average size of 4.5 individuals. They are experienced in running a business (on average around 11 years of experience), and more than half of the sample (57 percent) already holds a savings account. Given that treatment was randomly assigned, the five assignment groups are expected to have similar characteristics. Columns 4, 6, 8, and 10 in table 10.1 show the mean baseline characteristics of all microentrepreneurs surveyed at baseline by treatment group (including the pure control). Columns 5, 7, 9, and 11 report the p-values of the t-test for equality of each of these mean baseline characteristics against those in the placebo control group. No characteristics are significantly different from the placebo control group at the 5 percent level for the three treatments, except for the proportion of Igbo business owners in the Movie/ MFB group. The expectation of balance on observable baseline characteristics also holds across treatment groups, which supports our claim that the random- ization worked well. We see for the pure control group, however, that 3 of the 26 characteristics are significantly different at a 5  percent level (we would expect significant difference in 1 of every 20 measures by chance). Of particularly concern is an imbalance in having a savings account (56  percent in the placebo control group; 63 percent in the pure control group). This is likely to have been driven by differential nonresponse at follow-up, where we find higher nonresponse rates in the pure control group. We also explore balance across treatment groups for male and female business owners separately (tables 10A.1, 10A.2, 10A.3, and 10A.4 in the annex to this chapter) and find similar results. 330  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.1  Baseline balance TOTAL SAMPLE CONTROL MOVIE MFB MOVIE/MFB PURE CONTROL N MEAN MEAN MEAN P-VALUE MEAN P-VALUE MEAN P-VALUE MEAN P-VALUE VARIABLE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) Personal characteristics Age of respondent 2,314 37.76 37.90 37.52 0.553 37.89 0.996 37.31 0.339 38.44 0.427 Gender (male) 2,358 0.29 0.26 0.30 0.173 0.30 0.220 0.29 0.371 0.31 0.138 Married 2,357 0.84 0.85 0.82 0.211 0.86 0.557 0.82 0.206 0.86 0.845 Widowed 2,357 0.02 0.02 0.03 0.094* 0.01 0.284 0.02 0.984 0.03 0.264 Single 2,357 0.14 0.13 0.15 0.551 0.13 0.795 0.16 0.190 0.12 0.494 Muslim 2,356 0.36 0.35 0.40 0.136 0.35 0.793 0.36 0.717 0.33 0.421 Christian 2,356 0.64 0.64 0.60 0.154 0.65 0.958 0.63 0.621 0.67 0.387 Can speak English 2,346 0.70 0.70 0.67 0.321 0.72 0.450 0.71 0.636 0.73 0.382 Igbo 2,356 0.20 0.17 0.17 0.925 0.21 0.141 0.22 0.104 0.24 0.012** Yoruba 2,356 0.75 0.78 0.78 0.873 0.75 0.219 0.72 0.035** 0.71 0.025* Other ethnicitiy 2,356 0.05 0.05 0.05 0.635 0.04 0.777 0.06 0.242 0.04 0.839 Education No completed school 2,356 0.07 0.06 0.07 0.421 0.08 0.180 0.08 0.297 0.08 0.347 Primary school 2,356 0.22 0.24 0.24 0.968 0.21 0.164 0.21 0.209 0.19 0.067* High school diploma 2,356 0.50 0.49 0.48 0.749 0.50 0.754 0.51 0.527 0.53 0.329 Diploma 2,356 0.10 0.11 0.10 0.512 0.11 0.825 0.09 0.276 0.11 0.945 Graduate school 2,356 0.10 0.09 0.10 0.866 0.10 0.626 0.11 0.425 0.09 0.916 Household characteristics Household size 2,343 4.53 4.58 4.57 0.902 4.43 0.168 4.48 0.395 4.61 0.825 Number of children below 12 2,311 1.33 1.38 1.29 0.230 1.30 0.311 1.25 0.080* 1.44 0.524 Number of dependents 2,322 2.44 2.45 2.39 0.671 2.41 0.769 2.41 0.747 2.57 0.385 # of dependents outside household 2,213 1.55 1.50 1.53 0.843 1.53 0.827 1.54 0.784 1.66 0.330 Business characteristics Months in operation 2,310 97.40 98.69 97.58 0.847 96.98 0.771 101.02 0.698 91.03 0.218 Has a savings account 2,350 0.57 0.56 0.57 0.732 0.54 0.624 0.57 0.753 0.63 0.035** Keeps written financial records 2,340 0.37 0.36 0.35 0.684 0.37 0.708 0.38 0.619 0.40 0.315 Operating inside main market 2,324 0.25 0.24 0.26 0.500 0.24 0.985 0.26 0.535 0.27 0.287 Number of employees 2,352 1.44 1.57 1.46 0.345 1.40 0.169 1.39 0.161 1.36 0.168 Business experience in years 2,350 10.75 10.84 10.77 0.892 10.78 0.907 10.48 0.497 10.97 0.834 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 10.  Nigeria’s Nollywood nudge  ◾ 331 Table 10.2 reports the mean characteristics of those who were assigned to a screening event (column 1), which excludes individuals in the pure control group; and details observable differences of those who attended (column 2) with those who did not (column 3). As indicated in column 4, the selection into screenings is strongly correlated with more educated microentrepreneurs, who are more likely to speak English, enjoy higher access to financial products, and keep financial TABLE 10.2  Selection into screenings TOTAL PARTICIPATED DID NOT PARTICIPATE N MEAN N MEAN N MEAN P-VALUE VARIABLE (1) (2) (3) (4) (5) (6) (7) Personal characteristics Age of respondent 1,946 37.63 1,242 38.26 704 36.52 0.000*** Gender (male) 1,984 0.29 1,260 0.28 724 0.30 0.368 Married 1,983 0.84 1,259 0.85 724 0.82 0.054* Widowed 1,983 0.02 1,259 0.02 724 0.01 0.031** Single 1,983 0.14 1,259 0.13 724 0.17 0.004*** Muslim 1,983 0.36 1,260 0.35 723 0.39 0.112 Christian 1,983 0.63 1,260 0.64 723 0.61 0.111 Can speak English 1,974 0.70 1,255 0.72 719 0.66 0.005*** Igbo 1,982 0.19 1,260 0.20 722 0.18 0.149 Yoruba 1,982 0.75 1,260 0.75 722 0.75 0.965 Other ethnicitiy 1,982 0.05 1,260 0.04 722 0.07 0.012** Education No completed school 1,983 0.07 1,260 0.06 723 0.10 0.006*** Primary school 1,983 0.22 1,260 0.22 723 0.24 0.386 High school diploma 1,983 0.50 1,260 0.50 723 0.49 0.843 Diploma 1,983 0.10 1,260 0.11 723 0.09 0.137 Graduate school 1,983 0.10 1,260 0.11 723 0.09 0.101 Household characteristics Household size 1,972 4.51 1,251 4.52 721 4.51 0.873 Number of children below 12 1,948 1.30 1,234 1.31 714 1.29 0.761 Number of dependents 1,954 2.41 1,241 2.47 713 2.31 0.090* # of dependents outside HH 1,862 1.53 1,179 1.52 683 1.54 0.882 Business characteristics Months in operation 1,947 98.59 1235 98.76 712 98.30 0.917 Has a savings account 1,977 0.56 1260 0.59 717 0.52 0.002*** Keeps written fin. records 1,968 0.37 1254 0.39 714 0.32 0.002*** Op. inside main market 1,979 0.25 1260 0.28 719 0.20 0.000*** Number of employees 1,980 1.45 1259 1.45 721 1.45 0.987 Business experience in years 1,977 10.70 1256 10.88 721 10.40 0.218 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 332  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES records for their business. This selection process may be explained by the way the screening events were framed: business owners were told that they were invited to a business development event, and the invitation was in English (see figure  10.1). Since a major aim of edutainment is to reach out to the “bottom of the pyramid,” future edutainment activities may want to consider framing the event less as business development and more as entertainment, as well as promoting and designing it in such a way that language is not perceived as a barrier to attendance. Overall participation rates are reasonably high (60 percent) when compared to other financial literacy programs, but it is clear that nonpartic- ipants present a target group that potentially has the most marginal added value to participation but is at the same time the most difficult group to entice into these types of events. Although there is strong evidence of self-selection into screening events, table 10.3 shows that the drivers of this selection across screening events appear to be the same. For those who participated, we see balance across observable characteristics—which is in line with the fact that all screening events were marketed in the same way with the same characteristics. This balance of selec- tion across events supports the possibility of comparing attendees against each other, rather than needing to rely on the intention to treat estimates. 10.5.3 Attrition The attrition rate in this study is 21.1 percent, which is relatively high compared to other household surveys (e.g., EFInA 2010 had an attrition rate of 6 percent), but within reason when compared to enterprise surveys. Intensive efforts were made to reach all respondents who were listed at the baseline, but around 12 percent could not be contacted again, some refused to be reinterviewed (2.9  percent), and a very few (0.3 percent) were unable to participate (e.g., for health-related reasons). This attrition rate also includes former microentrepreneurs (5.7 percent), who may not be considered as being eligible anymore, because they shut down their business between the baseline listing and the endline survey. If former microbusiness owners are not taken into account, the attrition rate is reduced to 16.3 percent. There is some evidence for selective attrition for the pure control group, but good balance between the placebo and three treatment arms. Attrition is largest in the pure control group (25.5 percent) when compared to the control and treatment groups (20.2  percent). Table 10.4 suggests a random pattern of attrition for the three treatment arms when compared to the placebo control group, but a large and significant differential attrition in the pure control group. This differential attrition is reinforced by the balance results from table 10.1, and may result from the fact that pure control business owners were only contacted at baseline and follow-up. In contrast, all other groups had another intermediate contact to receive the screening invitation, making them (1) more aware of the activities, and (2) easier to track. Given the significantly lower response rate in the 333  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.3  Balance across screening participants TOTAL CONTROL MOVIE MFB MOVIE/MFB N MEAN N MEAN N MEAN P-VALUE N MEAN P-VALUE N MEAN P-VALUE VARIABLE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) (12) (13) Personal characteristics Age of respondent 1,243 38.27 309 38.13 327 38.46 0.79 287 37.92 0.81 307 38.52 0.60 Gender (male) 1,261 0.28 313 0.25 333 0.26 0.78 292 0.30 0.21 310 0.30 0.19 Married 1,260 0.85 312 0.84 333 0.84 0.87 292 0.87 0.41 310 0.85 0.67 Widowed 1,260 0.02 312 0.02 333 0.04 0.15 292 0.01 0.24 310 0.02 0.56 Single 1,260 0.13 312 0.13 333 0.12 0.55 292 0.12 0.68 310 0.13 0.83 Muslim 1,261 0.35 313 0.34 333 0.40 0.080* 292 0.35 0.79 310 0.32 0.71 Christian 1,261 0.64 313 0.66 333 0.60 0.096* 292 0.65 0.72 310 0.67 0.78 Can speak English 1,256 0.72 311 0.71 331 0.70 0.76 292 0.71 0.97 309 0.77 0.13 Igbo 1,261 0.20 313 0.19 333 0.19 0.94 292 0.23 0.30 310 0.22 0.40 Yoruba 1,261 0.75 313 0.78 333 0.77 0.88 292 0.74 0.34 310 0.73 0.15 Other ethnicitiy 1,261 0.04 313 0.04 333 0.04 0.62 292 0.03 0.95 310 0.06 0.18 Education No completed school 1,261 0.06 313 0.05 333 0.08 0.26 292 0.05 0.98 310 0.06 0.85 Primary school 1,261 0.22 313 0.24 333 0.23 0.96 292 0.22 0.61 310 0.19 0.15 High school diploma 1,261 0.50 313 0.50 333 0.47 0.57 292 0.51 0.77 310 0.52 0.57 Diploma 1,261 0.11 313 0.11 333 0.11 0.94 292 0.12 0.76 310 0.11 0.92 Graduate school 1,261 0.11 313 0.11 333 0.10 0.74 292 0.10 0.91 310 0.13 0.44 Household characteristics Household size 1,252 4.52 311 4.49 332 4.63 0.40 289 4.37 0.36 307 4.54 0.72 Number of children below 12 1,235 1.31 307 1.39 331 1.31 0.34 285 1.27 0.19 299 1.26 0.20 Number of dependents 1,242 2.47 306 2.44 331 2.51 0.69 287 2.40 0.85 305 2.51 0.67 # of dependents outside HH 1,180 1.52 297 1.50 308 1.47 0.94 274 1.65 0.43 288 1.51 0.93 Business characteristics Months in operation 1,420 97.23 350 96.94 369 101.37 0.47 334 96.54 0.95 352 95.03 0.76 Has a savings account 1,448 0.59 356 0.58 378 0.61 0.30 343 0.58 0.80 356 0.60 0.48 Keeps written fin. records 1,442 0.39 355 0.40 377 0.36 0.36 341 0.40 0.89 354 0.42 0.48 Op. inside main market 1,448 0.27 356 0.27 378 0.26 0.80 343 0.26 0.86 356 0.28 0.73 Number of employees 1,448 1.52 356 1.58 378 1.56 0.89 343 1.47 0.54 355 1.48 0.53 Business experience in years 1,257 10.89 312 10.95 330 11.02 0.96 292 10.45 0.45 310 11.03 0.92 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 334  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.4  Attrition in endline survey DEPENDENT VARIABLE INTERVIEWED IN ENDLINE SURVEY   (1) Movie −0.014 (0.02) MFB −0.032 (0.02) Movie/MFB −0.021 (0.02) Pure control −0.069** (0.02) Observations 2,437 R-squared 0 p -value of F model 0.6 Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. pure control group, we subsequently analyze treatment effects by comparing the placebo screening group with the different treatment arms. When data are analyzed by simply excluding respondents with missing values for any relevant outcome measures—item nonresponse (INR)—this could again cause biased results if missingness is systematically related to a respondent’s potential outcomes. Table 10.5 presents INR rates for main outcome measures across different treatment and control groups. For instance, for the question of basic understanding of inflation, it can be seen that 100 percent of the surveyed microentrepreneurs are asked this question (column 1) and that 2.37 percent of those who are asked do not give a response (column 2). Overall, the data in table  10.5 indicates that INR for main outcome measures is not a critical issue (most of the time, INR rates are less than 5 percent) and nondifferential across treatment and control groups. Interestingly, INR is the lowest for measures of intentions, saving, and borrowing behavior, whereas highest INR rates (between 10 and 20 percent) can be observed for questions related to perceptions about MFBs—possibly reflecting cases where business owners have not interacted with MFBs and therefore have not been able to form an opinion. Table 10.5 also reveals a striking increase in INR for the questions of perceptions about MFBs at the endline survey relative to the data that were collected shortly after the screening. This increase does not interact with a particular treatment status and may be due to different modes of interviews and the design of the questionnaires. While the short survey conducted right after the screenings was self-adminis- tered by attendees, the endline survey was conducted face to face. To avoid unit nonresponse and potential measurement errors, the self-administered question- naire was designed to be as simple as possible and only asked dichotomous (yes or no) type of questions with no explicit “don’t know” or “refusal” choices. This 10.  Nigeria’s Nollywood nudge  ◾ 335 TABLE 10.5  Item nonresponse across screening participants (%) TOTAL SAMPLE CONTROL MOVIE MFB MOVIE/MFB HAVE HAVE HAVE HAVE HAVE VARIABLE ITEM INR ITEM INR ITEM INR ITEM INR ITEM INR Knowledge Simple division 100 7.21 100 5.75 100 8.62 100 7.11 100 6.63 Inflation 100 2.37 100 2.18 100 2.40 100 1.46 100 2.21 Necessary documentation 100 3.77 100 3.57 100 3.21 100 3.97 100 3.61 Better savings product 100 1.74 100 2.18 100 1.60 100 1.67 100 2.01 Interest rate 100 4.07 100 4.37 100 5.21 100 3.56 100 3.61 Better loan product 100 2.67 100 3.37 100 2.40 100 2.30 100 3.61 Perceptions MFB will accept loan ap. (S) 52 0.00 62 0.00 65 0.00 59 0.00 59 0.00 MFB will accept loan ap. (E) 100 19.34 100 19.05 100 19.24 100 20.50 100 18.27 Taking a loan is too risky (S) 52 0.00 61 0.00 66 0.00 60 0.00 60 0.00 Taking a loan is too risky (E) 100 4.03 100 2.98 100 3.41 100 3.41 100 4.62 Trust in MFBs (S) 52 0.00 61 0.00 66 0.00 59 0.00 61 0.00 Trust in MFBs (E) 100 9.88 100 8.53 100 10.62 100 12.13 100 8.63 MFBs treat people w/ respect (S) 50 0.68 59 0.00 63 0.00 56 2.60 60 0.33 MFBs treat people w/ respect (E) 100 20.23 100 19.44 100 19.44 100 21.34 100 19.88 Perceptions about women Women can run businesses as 100 0.81 100 0.60 100 0.20 100 0.63 100 1.41 well as men Easier for men to receive loans 100 9.88 100 9.52 100 9.62 100 9.62 100 9.04 than for women Women make better financial 100 2.50 100 2.38 100 2.00 100 2.72 100 2.61 decisions than men Intentions Plan to apply for loan in next 6 52 0.16 62 0.00 67 0.00 59 0.71 61 0.00 months (S) Plan to apply for loan in next 6 100 4.66 100 3.17 100 5.21 100 4.18 100 4.62 months (E) Will save money next month (S) 52 0.00 62 0.00 66 0.00 59 0.00 61 0.00 Will save money next month (E) 100 4.24 100 3.77 100 4.21 100 4.81 100 3.82 Savings behavior Opened acc. day of screening 1 0.00 0 0.00 0 0.00 1 0.00 5 0.00 Follow-up with Accion 1 0.00 0 0.00 0 0.00 7 0.00 0 0.00 Plan to follow up with Accion 5 0.00 5 0.00 6 0.00 6 0.00 7 0.00 Saved money last month 100 0.47 100 0.79 100 0.40 100 0.00 100 0.60 Savings relative to income 100 8.57 100 9.13 100 8.22 100 8.79 100 8.84 Supplier credit 100 0.25 100 0.00 100 0.20 100 0.21 100 0.40 Loan from family/friends 100 0.25 100 0.00 100 0.00 100 0.00 100 0.60 Note: (S) = screening; (E) = endline. 336  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES means that direct comparison over time (e.g., through a difference-in-difference approach) would present challenges; however, similar response patterns across treatment groups support the idea that responses are at least internally consis- tent. Given the rather low INR rates for most outcome measures and the fact that they are indistinguishable across control and treatment groups, we take no specific measures to address this type of missingness. Nevertheless, we do account for missing data on covariates. In the regression analysis, coefficients of predictors of interest are adjusted using a procedure advocated by Cohen et al. (2002), whereby measures with missing values are replaced by zero and a dummy variable indicating such missing values is included. The logic behind this approach is that the dummy variables adjust the parameters for theoretically rele- vant predictors by removing variance that can be attributed to missing data that is lurking in the dependent variable (McKnight, McKnight, and Figueredo 2007). This also avoids losses in sample size during regression analysis in cases where observations would otherwise be dropped due to missing covariate responses. 10.6 MODEL SPECIFICATIONS In this study, we effectively have three treatment arms: MOVIE, MFB, and MOVIE/ MFB. Given that the intervention assignment was randomly allocated, we can measure the causal impact of these interventions through a simple linear regres- sion that identifies the average treatment effect using the intention-to-treat esti- mator: Yi = α + ∑3 γ T + Xi+ εi (10.1) j = 1 j ij where Yi is the outcome interest for participant i, and Tij is the treatment status for person i with regard to treatment j. Treatment j = {1,2,3}, for each of the three treatment groups. Xi is a vector of exogenous control variables collected at base- line or time-invariant variables collected in the endline survey.16 We run the same regression without controls and find point estimates to be unchanged in the anal- ysis, consistent with the balanced nature of the selected control variables, and as such we report the adjusted results in the chapter. Since we are particularly interested in gender differentials, our second spec- ification explores the impact heterogeneity by gender: 3 Yi = α + βGi + ∑j = 1(γj + Gδj)Tij + Xi+ εi (10.2) 16  The control variables included in the analysis are: business owner age, marital status, ethnicity, ability to speak English, education level, household size, religion, business expe- rience, number of employees at baseline, whether they had a savings account or kept financial records at baselines, and whether they operated in the main market area or in the outskirts (geographically defined through global positioning system [GPS]). 10.  Nigeria’s Nollywood nudge  ◾ 337 Here Gi = 1 if male, 0 if female. The regression results presented in the tables generated from the analysis include the effect of treatment j on females (γj), the additional impact for males (δj) and the overall gender differential Gj. Each table of results presents results from equation 10.1 first, followed by gender-disaggre- gated results from equation 10.2. In section 10.3, we see that overall selection into the movie screening is such that those who attended the events were slightly different from those who did not. However, we find that this selection pattern is the same across all screening events (based on balance of observable characteristics) and, importantly, there are no differential selection patterns between the three treatment arms and the placebo screening. In this case, we run a restricted analysis on those business owners who actually attended the event. Relying on the balance across an exten- sive set of baseline variables and the manner in which the events were imple- mented (randomized invitations at the individual level), we reasonably expect this comparison to provide an unbiased estimate of the average treatment effect on the treated—the impact for those who actually attended the event, using equa- tions 10.1 and 10.2 with the restricted sample of 1,261 participants. We acknowledge that, if there are large positive spillovers, this may result in a downward bias of the estimate of impact. As such, the survey included control clusters that were created through geographic discontinuities, where a self-con- tained cluster meant that all businesses within the cluster were at least 20 meters away from the next closest business outside of the cluster.17 This sampling method creates a “pure” control group less exposed to treatment neighbors, thus exogenously varying the level of intensity of treatment in any particular area of the market, theoretically allowing us to explore spillovers. We see, however, in the pure control group that we experience differential attrition resulting in an imbalance based on baseline observable variables. As such, we exclude this group from analysis in this chapter. In the following section, we present results using equation 10.1 with the restricted sample of business owners who actually attended a screening, using the placebo group as our control comparison. 10.7 RESULTS 10.7.1 Exposure Administrative records were kept on who participated in the screenings, using the personalized invitations to verify details and treatment status, which was a requirement for entry into the movie screening. The screenings were secured and We use the rule of 20 meters for businesses outside of the main market area. Density is 17  too high for businesses inside the main market area, in which case we use a 5-meter rule. 338  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES private with complete control over the entrance and exit of the events. Although participation rates averaged around 60 percent, contamination was very low as a result of this process. Table 10.6 highlights this fact, where less than 1 percent of invited guests went to a different screening than the one to which they had been assigned—strengthening the justification of using equations 10.1 and 10.2 with our restricted sample to measure the average treatment effect on the treated. In the follow-up survey, we asked for self-reported exposure, partly to confirm attendance, but also to understand whether people could remember the main activities and messages from the events; this is presented as a summary in table 10.7. While people have no problem recalling the screening, they express some confusion about the details of the event. We find that 95 percent of people recall receiving an invitation and 96 percent of the people who were recorded through administrative records as attending the event confirmed that they had attended. When asked specifically about whether they saw The Story of Gold, 90 percent TABLE 10.6  Compliance table (%) TREATMENT DID NOT ATTENDED THE FOLLOWING SCREENING ASSIGNMENT ATTEND PLACEBO MOVIE MFB MOVIE/MFB Pure control 99.0 0.0 0.2 0.4 0.4 Placebo control 41.0 57.9 1.0 0.2 0.0 Movie 38.0 0.2 61.5 0.3 0.0 MFB 42.6 0.3 0.5 56.6 0.0 Movie/MFB 41.1 0.0 0.2 0.5 58.3 in the Movie group and 93 percent in the Movie/MFB group acknowledged that they had done so; 77 percent and 82 percent, respectively, could recall the main message of the movie without prompting. However, placebo screening and MFB groups also reported having seen the movie, although at significantly lower levels (59 percent and 58 percent, respectively). Since the movie was tightly controlled and not released to the public, this suggests a potential confusion between The Story of Gold and the placebo movie screening—possibly confounded by the fact that neighboring businesses may have seen and mentioned something about the movie. Recall of Accion presence was much lower. We find significant increases in recall for MFB and Movie/MFB compared to Movie and control as expected, but the proportions are still low. Only 16 percent of MFB attendees and 17 percent of Movie/MFB attendees recalled Accion’s presence at the event. We also asked a falsification question to assess the level at which respondents may have been adjusting their answers to respond positively to the interview. We find that only 1 percent of people responded positively to a question asking whether a certain MFB (Jaiz Bank), which is only based in Abuja, had visited them (an impossibility); 10.  Nigeria’s Nollywood nudge  ◾ 339 TABLE 10.7  Self-reported exposure to interventions REMEMBERED REMEMBERED ATTENDING CORRECTLY REMEMBERED SEEING MOVIE EVENT WHERE IDENTIFIED RECEIVING ATTENDED CALLED THE ACCION MESSAGE OF INVITATION EVENT STORY OF GOLD PRESENTED MOVIE (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Movie only 0.01 −0.00 0.04 −0.01 0.23*** 0.30*** 0.02 0.02 0.26*** 0.36*** (0.014) (0.011) (0.029) (0.016) (0.031) (0.031) (0.019) (0.027) (0.030) (0.035) MFB −0.00 −0.01 0.01 −0.00 −0.01 −0.01 0.04** 0.06** −0.02 −0.04 (0.014) (0.011) (0.030) (0.016) (0.031) (0.033) (0.019) (0.028) (0.030) (0.036) Movie/MFB −0.00 −0.00 0.00 0.01 0.21*** 0.33*** 0.04** 0.07** 0.26*** 0.41*** (0.014) (0.011) (0.029) (0.016) (0.031) (0.032) (0.019) (0.028) (0.030) (0.036) Observations 1,976 1,259 1,975 1,259 1,974 1,258 1,974 1,259 1,979 1,261 R-squared 0.00 0.00 0.00 0.00 0.05 0.14 0.00 0.01 0.08 0.18 Controls No No No No No No No No No No Restricted sample No Yes No Yes No Yes No Yes No Yes Control mean 0.948 0.984 0.673 0.958 0.404 0.593 0.0734 0.102 0.286 0.419 Note: Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. this is similar across treatment arms, suggesting that positive response bias does not seem to be a problem in our case. Since the interventions were monitored carefully and Accion was indeed present at these events, this contrast between Accion and The Story of Gold recall highlights the differential salience of each of the interventions. 10.7.2 Financial literacy The quiz questions test basic numeracy and financial concepts. Since the movie screening aimed to influence emotions and perceptions rather than formal financial literacy, we expected these indicators to show balance across groups, which they do. Aggregating the questions into a single index, we find two things (table  10.8): (1) scores are very similar across all groups; and (2) the aggregate scores are relatively high, with the weighted and arithmetic scores yielding similar results—perhaps reflecting a lack of variation and cognitive separating ability of the set of questions. However, when exploring the covariates associated with these financial literacy scores, we find strong relationships between the overall score and (1) whether business owners had a savings account at baseline and (2) whether they had any schooling, supporting the assertion that the indexes are informative in distinguishing between financial literacy levels, and the similarities in scores across groups reflect balance induced by the randomization. 340  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.8  Financial literacy indexes FINANCIAL LITERACY SCORE ARITHMETIC WEIGHTED (1) (2) Treatment Movie −0.11 −0.14 (0.075) (0.112) MFB 0.04 0.10 (0.078) (0.115) Movie/MFB −0.05 −0.04 (0.077) (0.114) Gender-disaggregated interaction effects (female base) Movie −0.11 −0.12 (0.088) (0.130) MFB 0.10 0.14 (0.092) (0.136) Movie/MFB −0.09 −0.10 (0.091) (0.134) Gender-disaggregated interaction effects (male interaction) Male 0.11 0.18 (0.131) (0.193) Male*Movie −0.03 −0.07 (0.172) (0.254) Male*MFB −0.18 −0.14 (0.176) (0.261) Male*(Movie/MFB) 0.12 0.19 (0.173) (0.257) δ1 + γ1 ≠ 0 0.36 0.38 p -values for F-tests δ 2 + γ2 ≠ 0 0.57 0.98 δ 3 + γ3 ≠ 0 0.82 0.65 Observations 1,261 1,254 R-squared 0.14 0.12 Controls Yes Yes Restricted model Yes Yes Control mean 5.262 7.556 Note: Standard errors in parentheses. 10.  Nigeria’s Nollywood nudge  ◾ 341 10.7.3 Perceptions We find increases in self-reported trust and perceptions of MFBs directly after the screening events; however, when asked the same questions in the follow-up survey, many of the initial differences reduce or disappear.18 While males are influenced most strongly by the movie stimulus in the short run, differentials in self-reported trust only sustain for females in the longer run. Table 10.9 presents the results from the screening and endline surveys. While the movie on its own has some impact on whether people report that they would trust an MFB to keep their money when they were asked this question at the screening, the presence of Accion seems to have a much larger effect than the movie, and there is no additivity of the interventions (although both are significant and positive). In the second follow-up survey, we see that the differential between control and treat- ment group trust declines; however, it is the movie treatment arms that sustain results, where the impact on MFB reduces to insignificance. This sustained impact is almost entirely driven by females, even though males were most affected by the movie in the short run. A supporting question identifying positive perceptions of MFBs (“MFBs treat people with respect”) shows similar results, with larger impacts for males in the short run, followed by some limited but sustained differ- ences for females in the longer run—even when male differentials disappear. This significant impact is only found in the combined Movie/MFB arm. We also explore perceptions of ease in obtaining a loan and riskiness of doing so. Both the Movie and MFB treatments have a significant positive effect on busi- ness owners’ perception of how likely it is that they may receive a loan if they applied for one in the short run (this falls away completely in the longer run), but none of the interventions have any impact on beliefs of the risk in taking out a loan. We also explore perceptions of female business owners (table 10.10). 10.7.4 Intentions We tested business owners’ intentions about their saving and borrowing plans, once again through the screening questionnaire and in the follow-up, with results presented in table 10.11. Here there is mixed evidence, with some impact on borrowing intentions, but no changes on what are already very high intentions to save. Intention to save is almost universal—90 percent of respondents at the screening and 95  percent in the follow-up indicated that they planned to save Direct comparison between the two follow-up surveys should be handled carefully. 18  Although the questions asked were identical, the response method varied across data collection activities. In the immediate follow-up, the question responses were yes/no, and the questionnaire was self-administered. In the four-month follow-up survey, the question- naire was administered by an interviewer and response options were strongly agree, agree, disagree, strongly disagree. 342  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.9  Perceptions of microfinance banks MFB LOAN APPLICATION TAKING A LOAN IS TOO I WOULD TRUST AN MFB TO WILL BE ACCEPTED RISKY FOR ME KEEP MY MONEY MFBS TREAT PEOPLE WITH RESPECT SCREENING ENDLINE SCREENING ENDLINE SCREENING ENDLINE: S ENDLINE: A+S SCREENING ENDLINE: S ENDLINE: A+S (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Treatment Movie 0.06** 0.04 −0.01 0.01 0.15*** 0.08** 0.01 −0.05 0.03 0.01 (0.026) (0.033) (0.038) (0.039) (0.034) (0.038) (0.033) (0.031) (0.038) (0.034) MFB 0.10*** −0.01 −0.02 0.01 0.26*** 0.05 0.01 0.01 0.06 0.02 (0.027) (0.034) (0.039) (0.041) (0.035) (0.039) (0.034) (0.032) (0.040) (0.035) Movie/MFB 0.08*** 0.05 −0.02 0.01 0.27*** 0.08** 0.06* 0.10*** 0.10** 0.06* (0.027) (0.034) (0.039) (0.040) (0.034) (0.039) (0.033) (0.031) (0.039) (0.035) Gender-disaggregated interaction effects (female base) Movie 0.04 0.00 −0.01 −0.01 0.08** 0.06 0.01 −0.08** −0.00 −0.01 (0.031) (0.038) (0.044) (0.046) (0.039) (0.044) (0.038) (0.036) (0.044) (0.040) MFB 0.10*** −0.03 −0.03 −0.01 0.25*** 0.07 0.03 −0.01 0.07 0.01 (0.032) (0.040) (0.046) (0.048) (0.041) (0.046) (0.040) (0.038) (0.047) (0.042) Movie/MFB 0.08** 0.05 −0.02 −0.02 0.22*** 0.12*** 0.05 0.07* 0.13*** 0.05 (0.032) (0.040) (0.045) (0.047) (0.040) (0.046) (0.039) (0.037) (0.046) (0.041) Gender-disaggregated interaction effects (male interaction) Male −0.03 −0.02 −0.08 −0.10 −0.17*** 0.01 −0.00 −0.09 0.03 −0.02 (0.046) (0.057) (0.066) (0.068) (0.058) (0.066) (0.057) (0.054) (0.066) (0.059) Male*Movie 0.06 0.13* −0.01 0.10 0.28*** 0.04 0.03 0.14* 0.11 0.08 (0.061) (0.075) (0.087) (0.090) (0.077) (0.087) (0.075) (0.071) (0.087) (0.078) Male*MFB −0.02 0.07 0.03 0.08 0.04 −0.06 −0.06 0.08 −0.01 0.03 (0.062) (0.077) (0.089) (0.092) (0.078) (0.089) (0.076) (0.074) (0.089) (0.080) Male*(Movie/MFB) 0.01 0.00 −0.02 0.11 0.19** −0.14* 0.02 0.11 −0.09 0.03 (0.061) (0.076) (0.087) (0.090) (0.077) (0.087) (0.075) (0.071) (0.088) (0.079) δ1 + γ1 ≠ 0 0.05 0.04 0.81 0.21 0 0.18 0.55 0.36 0.17 0.28 p -values δ 2 + γ2 ≠ 0 0.11 0.56 0.98 0.14 0 0.93 0.63 0.27 0.48 0.61 for F-tests δ 3 + γ3 ≠ 0 0.1 0.42 0.58 0.25 0 0.77 0.23 0 0.66 0.21 Observations 1,215 1,261 1,223 1,261 1,226 1,261 1,261 1,174 1,261 1,261 R-squared 0.04 0.05 0.05 0.04 0.11 0.05 0.05 0.06 0.05 0.05 Controls Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Restricted model Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Control mean 0.820 0.754 0.356 0.495 0.586 0.581 0.757 0.808 0.559 0.722 Note: A = agree; S = strongly agree. Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 10.  Nigeria’s Nollywood nudge  ◾ 343 TABLE 10.10  Perception of female financial performance at endline WOMEN CAN RUN IT IS EASIER FOR MEN WOMEN MAKE BETTER BUSINESSES JUST AS TO RECEIVE LOANS FINANCIAL DECISIONS WELL AS MEN THAN WOMEN THAN MEN (1) (2) (3) Treatment Movie −0.00 0.07* 0.05 (0.020) (0.038) (0.030) MFB 0.00 0.07* 0.04 (0.020) (0.039) (0.031) Movie/MFB 0.00 0.07* 0.06* (0.020) (0.039) (0.031) Gender-disaggregated interaction effects (female base) Movie −0.01 0.04 −0.02 (0.023) (0.044) (0.035) MFB 0.01 0.05 0.01 (0.024) (0.046) (0.037) Movie/MFB 0.00 0.03 0.01 (0.024) (0.046) (0.036) Gender-disaggregated interaction effects (male interaction) Male −0.13*** 0.09 −0.48*** (0.034) (0.066) (0.052) Male*Movie 0.04 0.13 0.25*** (0.045) (0.087) (0.069) Male*MFB −0.04 0.06 0.15** (0.046) (0.088) (0.071) Male*(Movie/MFB) −0.01 0.16* 0.19*** (0.045) (0.087) (0.070) δ1 + γ1 ≠ 0 0.55 0.03 0 p -values for δ 2 + γ2 ≠ 0 0.54 0.13 0.01 F-tests δ 3 + γ3 ≠ 0 0.88 0.01 0 Observations 1,261 1,261 1,261 R-squared 0.09 0.08 0.19 Controls Yes Yes Yes Restricted model Yes Yes Yes Control mean 0.936 0.342 0.751 Note: Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. some money in the following month. When we compare this to actual saving in the past month (65  percent in the endline survey—table 10.12), it is clear that there is a disconnect between intentions and behavior, with many more business owners planning to save but not necessarily following through with these plans, reinforcing the possibility that various frictions may be reducing people’s ability to 344  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.11  Intentions I PLAN TO APPLY FOR A LOAN IN I WILL SAVE SOME MONEY NEXT THE NEXT 6 MONTHS MONTH SCREENING ENDLINE SCREENING ENDLINE (1) (2) (3) (4) Treatment Movie 0.05 −0.02 0.03 0.02 (0.039) (0.039) (0.023) (0.017) MFB 0.08* −0.06 −0.04* −0.01 (0.041) (0.040) (0.024) (0.018) Movie/MFB 0.10** 0.00 0.02 −0.03 (0.040) (0.040) (0.024) (0.018) Gender-disaggregated interaction effects (female base) Movie 0.05 −0.03 0.04 0.02 (0.045) (0.046) (0.027) (0.020) MFB 0.06 −0.06 −0.04 −0.01 (0.048) (0.048) (0.029) (0.021) Movie/MFB 0.09* 0.01 0.01 −0.03 (0.047) (0.047) (0.028) (0.021) Gender-disaggregated interaction effects (male interaction) Male −0.01 0.15** −0.00 0.03 (0.068) (0.068) (0.041) (0.030) Male*Movie 0.03 0.03 −0.03 −0.01 (0.089) (0.090) (0.053) (0.040) Male*MFB 0.06 −0.02 −0.03 0.02 (0.092) (0.091) (0.055) (0.040) Male*(Movie/MFB) 0.04 −0.02 0.03 0.02 (0.090) (0.090) (0.054) (0.040) δ1 + γ1 ≠ 0 0.34 0.92 0.76 0.73 p -values δ 2 + γ2 ≠ 0 0.12 0.31 0.16 0.84 for F-tests δ 3 + γ3 ≠ 0 0.09 0.87 0.36 0.81 Observations 1,233 1,259 1,232 1,259 R-squared 0.04 0.05 0.04 0.07 Controls Yes Yes Yes Yes Restricted model Yes Yes Yes Yes Control mean 0.547 0.530 0.902 0.949 Note: Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. translate intention into action. The reason for this disconnect could be manifold— hyperbolic discounting, lack of disposable funds, overconfidence, limited access to financial products—and we cannot necessarily disentangle all of these factors. However, we do see that the interventions provided have little influence on what are already very strong self-reported intentions to save, suggesting that this is not likely the channel through which any behavior change occurs. 10.  Nigeria’s Nollywood nudge  ◾ 345 10.7.5 Savings behavior At screening events with MFBs present, business owners were able to discuss savings opportunities with the MFB and sign up for a savings account on the spot if they were interested. Participants had two options when expressing interest in opening an account with the MFB: (1) business owners would meet with the MFB and sign up for a follow-up visit to open an account, or (2) business owners would sign up for an account on the spot. Table 10.13 reports on the data collected at the two types of screening events (MFB, Movie/MFB) showing that people were more likely to express interest in opening an account by visiting the MFB stand directly after the event in the MFB group (13 percent versus 8 percent). However, differ- entiating this visit into each of the two options available (signing up on the spot or agreeing to a follow-up visit to sign up for an account), we find substantial differ- ences. The majority of people in the MFB group who visited the MFB stand opted for a follow-up visit rather than signing up on the spot. However, the Movie/MFB combination event was substantially more effective at incentivizing on-the-spot savings account sign-ups at the event; this effect was strongest for male partici- pants. The Movie/MFB combination event motivated 7 percent of participants to open an account on the spot (compared to 2 percent in the MFB group). This effect was substantially different between male and female participants (5 percent of females and 11 percent of males). The overall difference is statistically significant, but the gender-disaggregated differences are only significant for males. Although the MFB event was moderately successful in encouraging people to visit the MFB stand and agree to a follow-up visit (11 percent), on further inspec- tion we find that none of the people in this category actually followed up after the event (table  10.12). In fact, the only people who followed up with an MFB after the screening came from the Movie group, where the MFB had not been present. Although it is a small fraction (2 percent for both males and females), this is the only group with a statistically significant increase. The results provide the following insights: (1) reducing access barriers to virtually zero (MFB condition) increases engagement with the MFB and reported interest in opening an account, but has only a modest effect on actual sign-up rates; (2) even without having an immediate call to action (the ability to open an account on the spot) The Story of Gold has some (although very limited) impact on short-term behavior, inducing 2 percent of participants to follow up with an MFB afterwards (Movie condition); but (3) combining the reduced access constraint with the movie designed to promote savings (Movie/MFB) provides the strongest incentive to open a savings account, mostly driven by male participant choices. The evaluation design helps to deconstruct some of the potential barriers to demand for a savings account and identifies that an educational event attached to an emotional stimulus can be an effective tool to increase take-up, but only when combined with an interven- tion that allows for immediate action. However, this tells us little about savings behavior after the event. 346  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10.12  Savings behavior FOLLOWED UP CURRENTLY HAS CURRENTLY HAS WITH AN MFB ANY FORM OF SAVED SOME SAVINGS OF AFTER THE FORMAL SAVINGS MONEY LAST ≤ 1 MONTH OF EVENT ACCOUNT MONTH INCOME (1) (2) (3) (4) Treatment Movie 0.02*** −0.01 0.02 0.01 (0.006) (0.029) (0.037) (0.039) MFB 0.00 −0.04 0.01 0.07* (0.006) (0.030) (0.038) (0.040) Movie/MFB 0.00 −0.04 −0.04 0.02 (0.006) (0.030) (0.038) (0.040) Gender-disaggregated interaction effects (female interaction) Movie 0.02*** 0.02 0.05 0.03 (0.007) (0.034) (0.043) (0.045) MFB 0.00 −0.05 0.02 0.05 (0.008) (0.035) (0.045) −0.047 Movie/MFB 0.00 −0.04 0.01 0.03 (0.008) (0.035) (0.044) (0.047) Gender-disaggregated interaction effects (male interaction) Male 0.00 0.07 0.03 0.05 (0.011) (0.050) (0.064) (0.067) Male*Movie −0.02* −0.09 −0.10 −0.05 (0.014) (0.066) (0.084) (0.089) Male*MFB −0.01 0.02 −0.05 0.05 (0.015) (0.068) (0.086) (0.091) Male*(Movie/MFB) −0.01 −0.01 −0.17** −0.02 (0.014) (0.067) (0.085) (0.089) δ1 + γ1 ≠ 0 0.92 0.18 0.43 0.73 p -values δ 2 + γ2 ≠ 0 0.82 0.64 0.76 0.17 for F-tests δ 3 + γ3 ≠ 0 0.78 0.38 0.03 0.93 Observations 1,261 1,261 1,256 1,261 R-squared 0.03 0.34 0.08 0.05 Controls Yes Yes Yes Yes Restricted model Yes Yes Yes Yes Control mean 0 0.738 0.650 0.415 Note: Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. Despite the strong impacts observed, important concerns arise from the follow-up findings. First, we find that 67 percent of all participants who opened a savings account at the event reported having a savings account at baseline (significantly higher than the average for our sample). While there may be rational reasons to hold multiple accounts (or to change accounts), the finding reinforces 10.  Nigeria’s Nollywood nudge  ◾ 347 TABLE 10.13  Savings account sign-up rates EXPRESSED DID NOT OPEN INTEREST IN SIGNING ACCOUNT AT OPENED ACCOUNT UP FOR SAVINGS SCREENING BUT ON DAY OF ACCOUNT PLANS TO FOLLOW UP SCREENING (1) (2) (3) Treatment Movie/MFB −0.05* −0.09*** 0.05*** (0.024) (0.019) (0.017) Gender-disaggregated interaction effects (female base) Movie/MFB −0.07** −0.10*** 0.03 (0.029) (0.022) (0.020) Gender-disaggregated interaction effects (male interaction) Male −0.04 −0.02 −0.02 (0.040) (0.030) (0.027) Male*(Movie/MFB) 0.09* 0.02 0.07** (0.054) (0.041) (0.037) p -values for F-tests: δ1 + γ1 ≠0 0.73 0.02 0 Observations 607 607 607 R-squared 0.08 0.09 0.10 Controls Yes Yes Yes Restricted model Yes Yes Yes Control mean 0.128 0.108 0.0203 Note: Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. the fact that the intervention may be inducing action only in a subpopulation that has lower marginal gains in doing so when compared to the unbanked target population. The second related concern is that in the follow-up, we find no distin- guishable difference in whether respondents have a savings account—which is not surprising given that the majority of those induced to open an account already had one prior to the screenings. Of greater concern, however, we find that males in the Movie/MFB group report having been less likely to save some money in the month prior to the follow-up survey and show no differences in savings amounts relative to their income. While it is not clear what may be driving this result, it is possible that the event—although successfully motivating business owners to act in the moment and put money in a new savings account—only served to displace future savings, with no net gain. 10.7.6 Borrowing behavior For borrowing behavior, we rely only on self-reported responses in the follow-up survey. The movie message centered on responsible borrowing, highlighting the problems with relying on moneylenders, and we reflect on this through two particular indicators: (1) borrowing rate in last 4 months and (2) the source of 348  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES borrowing. In particular, we were interested in identifying whether business owners used formal or informal sources for financing. We find first that borrowing rates are substantial—about half of all business owners reported taking out a loan in the past four months, and half of those who took a loan did so from an informal source. The interventions have no effect on borrowing rates (although there is a reduction in all treatment groups, it is not significant). Similarly we find little to no evidence in changes in the form of lending (table 10.14), although TABLE 10.14  Borrowing behavior TAKEN OUT A LOAN IN LAST LOAN WAS FROM 4 MONTHS INFORMAL SOURCE (1) (2) Treatment Movie −0.06 −0.02 (0.039) (0.070) MFB −0.07* 0.07 (0.040) (0.070) Movie/MFB −0.06 −0.08 (0.040) (0.069) Gender-disaggregated interaction effects (female interaction) Movie −0.06 −0.07 (0.045) (0.081) MFB −0.06 0.05 (0.047) (0.081) Movie/MFB −0.05 −0.14* (0.047) (0.081) Gender-disaggregated interaction effects (male interaction) Male 0.01 −0.11 (0.067) (0.121) Male*Movie 0.01 0.19 (0.089) (0.166) Male*MFB −0.03 0.11 (0.091) (0.161) Male*(Movie/MFB) −0.01 0.21 (0.089) (0.159) δ1 + γ1 ≠ 0 0.5 0.47 p -values δ 2 + γ2 ≠ 0 0.25 0.27 for F-tests δ 3 + γ3 ≠ 0 0.36 0.61 Observations 1,261 410 R-squared 0.06 0.11 Controls Yes Yes Restricted model Yes Yes Control mean 0.508 0.470 Note: Standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 10.  Nigeria’s Nollywood nudge  ◾ 349 females in the Movie/MFB group reduce informal lending by 14 percentage points, which is borderline significant. Interestingly, there seems to be more congruency between intentions to borrow and actual borrowing than for savings intentions and behavior. While 54 percent of people mentioned that they were planning to take out a loan in the next six months immediately after the screening, we find four months later that 51 percent of people did so. This contrasts sharply with the intended savings (90 percent) and actual savings rates (60 percent); this seems to confirm that, in terms of savings behavior, there are several additional barriers at play besides those that the interventions address directly. 10.8 ROBUSTNESS CHECKS Our results show a significant effect of Movie/MFB on motivating business owners to open a savings account, but little to no evidence of longer-term impact on a broad range of savings and borrowing perceptions and behavior. A null effect could be a result of (1) limited power, driven by sample sizes too small to detect true impacts; (2) spillovers improving outcomes for the control group; or (3) selec- tion bias resulting from the control group participants having different participa- tion decisions than our treatment groups. Power is of particular concern when we measure heterogeneous impacts by gender, given that only 28 percent of our sample is male. We run each of the regressions reported in this chapter for the entire sample (without differentiating by gender) and continue to find mostly null to low effects on our outcomes of interest in the four-month follow-up.19 Here our sample is substantial, and power is less of a concern. However, in most cases, the point estimate of the effects is so small that the interpretation of the results would not change even in cases we were to have enough sample power to estimate these small changes. The study was originally designed to account directly for potential spill- overs, given that all participants came from the same market area and interac- tion between participants was expected. The pure control group was generated using cluster randomization to address this; however, as mentioned previously, we are unable to use this group due to selective attrition and cannot rule out potential spillovers. Given that we see the strongest effects of the intervention in the immediate term, and given the nature of the program (increasing short-term motivation rather than focusing on financial content), it seems somewhat unlikely that second-hand information passed from treatment to control business owners is likely to be a serious concern. As expected, we do find cases where significant results in the gender-disaggregated 19  analysis become nonsignificant in the pooled specifications, particularly when male and female effect coefficients have opposite signs. 350  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Our restricted regression analysis used throughout the chapter effectively reports on the average treatment effect on the treated, without reference to the intention-to-treat results—which limits the scope of interpretation to effects on those who were actually convinced to attend the event. We run intention-to-treat regressions, including all business owners invited to the screening events on outcomes who were recorded at the endline, but do not report these results here. Unsurprisingly (see discussion above on why we can rely on the treatment effect on the treated in this context), the null effects remain; and our outcomes where impacts were found mostly remain significant, albeit with lower point estimates for impact.20 Finally, reflection on the savings account take-up rates on which we find significant impacts is required. Why is it that males react most strongly to the screening event in the short run? This could reflect the fact that male emotions are affected more than females, inducing action; but may equally reflect the possi- bility that females have added constraints beyond motivation that affect take-up, such as low liquidity or limited autonomy in financial decision making. The liter- ature has found that females often make decisions jointly with their spouse or other counterpart, when compared to male business owners. However, we find that business autonomy is balanced across gender in our sample, with 92 percent of males and females reporting that they make business decisions on their own. We do find, however, that business revenues and profits across gender differ significantly, with males having nearly twice the yearly profits of females. Selec- tion equations regressing profits and revenue with the likelihood of opening an account show no relationship. Furthermore, we find that intermediate outcomes such as increased self-reported trust in MFBs are substantially stronger for males than females. This suggests that, rather than females facing added constraints that the screening event does not overcome, the events have a differential effect on perceptions by gender that seems to be driving the differential take-up of savings accounts at the event. 10.9 DISCUSSION AND CONCLUSION The primary role of the evaluation was to explore the use of a new medium to transmit financial messages, focusing on the use of heuristics and emotions to spur action in the short run with the intention of getting business owners a foot in the door to use financial products more regularly, learning and building expe- rience thereafter. The second objective was to identify how access to financial products and motivation interact to induce action, and whether choice architec- ture can be effectively utilized to promote welfare-enhancing financial decisions. 20  Informal lending is no longer significant. 10.  Nigeria’s Nollywood nudge  ◾ 351 The results from the evaluation are mixed, and warrant further discussion on three issues of importance for policy dialogue: (1) the ability of edutainment to reach out to the targeted population, (2) the role of choice architecture on influ- encing short-term decisions, and (3) ensuring sustained behavior change. Recent evidence has highlighted the challenges to encouraging people to attend voluntary financial literacy workshops and other training programs (see chapter 7). Low take-up rates are common, and this is especially true for interven- tions targeting business owners. Business owners may be making a rational deci- sion to avoid the training because of low perceived benefits. Using edutainment to transmit financial messages is a new approach that has the potential benefit of being more inclusive, lowering barriers to participation. Response rates in this study of approximately 60 percent reflect the fact that, even though these events are able to reach out to the majority of potential participants, this is far from universal and more effort is needed to find ways to market these events to have more mass appeal. In particular, the least educated people with lowest access to financial products were the ones who selected out of the screening events, highlighting the difficulty of reaching out to this subpopulation. The study identifies a strong interaction between offering a stimulus (the movie) together with a direct outlet (the presence of the MFB) for acting on this motivation. This result is not surprising, and replicates what is well known among marketers in a development setting. However, applying choice architecture to a development setting requires careful attention to the potential unexpected outcomes that may result. In our case, the one-off screening was effective at encouraging people to open new accounts; but on closer inspection, nearly two-thirds of these people already had savings accounts, possibly limiting the potential marginal impact of the work. This highlights the importance of testing potential interventions at a pilot level, measuring and understanding the determi- nants of take-up before scaling up. While the intervention was able to influence decisions in the short run, people make financial decisions on a daily basis, and more sustained behavior change is critical in the context of saving. Our limited longer-term impacts emphasize this point. The ability to spur people to action through the use of edutainment may have more development impact for activities that are beneficial as one-off actions, particularly given the intervention’s relatively low cost and simple logistics. Exam- ples of where these types of interventions could work in other development areas include, for instance, encouraging people to test themselves at mobile clinics for HIV/AIDS or taking vaccinations, where one-time actions of groups of people at once can have important private and public benefits. This approach could also be tailored to more sustained financial behavior change if coupled with commitment savings accounts—where decisions taken in the moment have a more binding effect in the longer run (Ashraf, Karlan, and Yin 2006). However, take-up of finan- cial instruments tells us little about how this increased exposure may strengthen financial capabilities—responsible use of these instruments and financial decision 352  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES making more generally. The literature has traditionally explored the direction for strengthening financial capabilities as going from education to better financial decision making and increased use of financial products. There is less under- standing of how a learning-by-doing approach—focusing on providing access to financial instruments and exploring how this translates into experiential learning and ultimately improved decision making. While we have seen that nudges can be developed to help overcome the access constraint, it is still unclear as to whether this can be effectively translated into strengthened financial capabilities in the longer run. REFERENCES Andrade, Eduardo B., and Dan Ariely. 2009. “The Enduring Impact of Transient Emotions on Decision Making.” Organizational Behavior and Human Decision Processess 109 (May): 1–8. Ariely, Dan, George Loewenstein, and Drazen Prelec. 2003. “Coherent Arbitrariness”: Stable Demand Curves without Stable Preferences.” Quarterly Journal of Economics 118 (1): 73–106. Arkes, Hal R., and Catherine Blumer. 1985. “The Psychology of Sunk Cost.” Organizational Behavior and Human Decision Processes, 35 (1): 124–40. Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of Economics 121 (2): 635–72. Bertrand, Marianne, Dean Karlan, Sendhil Mullainathan, Eldar Shafir, and Jonathan Zinman. 2010. “What’s Advertising Content Worth? Evidence from a Consumer Credit Marketing Field Experiment.” Quarterly Journal of Economics 125 (1): 263–306. Cialdini, Robert B., Melanie R. Trost, and Jason T. Newsom. 1995. “Preference for Consistency: The Development of a Valid Measure and the Discovery of Surprising Behavioral Implications.” Journal of Personality and Social Psychology 69 (2): 318–28. Cohen, J., Patricia Cohen, Stephen G. West, and Leona S. Aiken. 2002. Applied Multiple Regression/Correlation Analysis for the Behavioral Sciences, 3rd edition. Mahwah, NJ: Lawrence Erlbaum Associates. Cole, Shawn, and Nilesh Fernando. 2008. “Assessing the Importance of Financial Literacy.” ADB Finance for the Poor 9 (3). Cole, Shawn, Thomas Sampson, and Bilal Zia. 2009. “Financial Literacy, Financial Decisions, and the Demand for Financial Services: Evidence from India and Indonesia.” Working Paper 09-117, Harvard Business School, Cambridge, MA. —. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Croson, Rachel, and Uri Gneezy. 2009. “Gender Differences in Preferences.” Journal of Economic Literature 47 (2): 448–74. Deaton, Angus S. 1989. “Saving and Liquidity Constraints.” NBER Working Paper 3196, National Bureau of Economic Research, Cambridge, MA. Drexler, Alejandro, Greg Fischer, and Antoinette Schoar. 2012. “Keeping It Simple: Financial Literacy and Rules of Thumb.” http://www.mit.edu/~aschoar/KIS-DFS-March2013.pdf. 10.  Nigeria’s Nollywood nudge  ◾ 353 Duflo, Esther, and Emmanuel Saez. 2003. “The Role of Information and Social Interactions in Retirement Plan Decisions: Evidence from a Randomized Experiment.” Quarterly Journal of Economics 118 (3): 815–42. Finucane, M. L., P. Slovic, C. K. Mertz, J. Flynn, and T. A Satterfield. 2000. “Gender, Race, and Perceived Risk: The ‘White Male’ Effect. Health, Risk and Society 2 (2): 159–72. Freedman, Jonathan L., and Scott C. Fraser. 1966. “Compliance without Pressure: The Foot- in-the-Door Technique.” Journal ol Personality and Social Psychology 4 (2): 195–202. Goldberg, J. H., J. S. Lerner, and P. E. Tetlock. 1999. “Rage and Reason: The Psychology of the Intuitive Prosecutor.” European Journal of Social Psychology 29 (56), 781–95. Harshman, Richard A., and Allan Paivio. 1987. “Paradoxical” Sex Differences in Self-Reported Imagery.” Canadian Journal of Psychology 41 (3): 287–302. Hilgert, Marianne A., Jeanne M. Hogarth, and Sondra G. Beverly. 2003. “Household Financial Management: The Connection between Knowledge and Behavior.” Federal Reserve Bulletin 89: 309–22. Hinz, R.  D. McCarthy, and J.  P., D.  A. Turner. 1997. “Are Women Conservative Investors? Gender Differences in Participant-Directed Pension Investments.” In Positioning Pensions for the Twenty-First Century, edited by M. S. Gordon, O. S. Mitchell, and M. M. Twinney, 91–103. Philadelphia: University of Pennsylvania Press. Kahneman, Daniel. 2003. “Maps of Bounded Rationality: Psychology for Behavioral Economics.” American Economic Review 93 (5): 1449–75. Kahneman, Daniel, and Dan Lovallo. 1993. “Timid Choices and Bold Forecasts: A Cognitive Perspective on Risk Taking.” Management Science 39 (1): 17–31. Karlan, Dean, Aishwarya Lakshm Ratan, and Jonathan Zinman. 2013. “Savings by and for the Poor: A Research Review and Agenda.” Economic Growth Center Discussion Paper No. 1027, Yale University, New Haven. La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. 2012. “Soap Operas and Fertility: Evidence from Brazil.” American Economic Journal: Applied Economics 4 (4): 1–31. Loewenstein, George, and Jennifer Lerner. 2003. “The Role of Affect in Decision Making.” In Handbook of Affective Science, edited by R. Davidson, H. Goldsmith, and K. Scherer, 619–42. Oxford, UK: Oxford University Press. Lusardi, Annamaria. 2007. “Household Saving Behavior: The Role of Financial Literacy, Information, and Financial Education Programs.” CFS Working Paper No. 250. Center for Financial Studies, Goethe University, Frankfurt. Lusardi, Annamaria, and Olivia S. Mitchell. 2013. “The Economic Importance of Financial Literacy: Theory and Evidence.” NBER Working Paper 18952, National Bureau of Economic Research, Cambridge, MA. Lusardi, Annamaria, and Peter Tufano. 2009. “Debt Literacy, Financial Experiences, and Overindebtedness.” NBER Working Paper 14808, National Bureau of Economic Research, Cambridge, MA. Mailath, George J., and Larry Samuelson. 2006. Repeated Games and Reputations: Long-Run Relationships. Oxford, UK: Oxford University Press. McKenzie, David, and Christopher Woodruff. 2012. “What Are We Learning from Business Training and Entrepreneurship Evaluations around the Developing World?” Policy Research Working Paper 6202, World Bank, Washington, DC. McKnight, P. E., K. M. McKnight, and A. J. Figueredo. 2007. Missing Data: A Gentle Introduction. New York: Guilford Press. Mulaj, Florentina, and William Jack. 2012. “Evaluating the Efficacy of Mass Media and Social Marketing Campaigns in Changing Consumer Financial Behavior.” Social Protection and Labor Discussion Paper 1220, World Bank, Washington, DC. 354  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Ottati, Victor C., and Linda M. Isbell. 1996. “Effects of Mood during Exposure to Target Information on Subsequently Reported Judgments: An On-Line Model of Misattribution and Correction.” Journal of Personality and Social Psychology 71: 39–53. Pathak, Payal, Jamie Holmes, and Jamie Zimmerman. 2011. “Accelerating Financial Capability among Youth: Nudging New Thinking.” New America Foundation. h t t p : // w w w. n e w a m e r i c a . n e t / s i t e s / n e w a m e r i c a . n e t / f i l e s / p o l i c y d o c s / AcceleratingFinancialCapabilityamongYouth.pdf. Pierson, Paul. 2000. “Increasing Returns, Path Dependence, and the Study of Politics.” American Political Science Review 94 (2): 251–67. Pocheptsova, Anastasiya, and Nathan Novemsky. 2010. “When Do Incidental Mood Effects Last? Lay Beliefs Versus Actual Effects.” Journal of Consumer Research 36 (6), 992–1001. Schwarz, Norbert, and Gerald L. Clore. 1983. “Mood, Misattribution, and Judgments of Well- Being: Informative and Directive Functions of Affective States.” Journal of Personality and Social Psychology 45 (3): 513–23. Slovic, Paul, Melissa L. Finucane, Ellen Peters, and Donald MacGregor. 2007. “The Affect Heuristic.” European Journal of Operational Research 177 (3): 1333–52. Spader, Jonathan, Janneke Ratcliffe, Jorge Montoya, and Peter Skillern. 2009. “The Bold and the Bankable: How the Nuestro Barrio Telenovela Reaches Latino Immigrants with Financial Education.” Journal of Consumer Affairs 43 (1): 56–79. Stango, V., and J. Zinman. 2009. “Exponential Growth Bias and Household Finance.” Journal of Finance 64 (6): 2807–49. Sunden, A.  E., and B. J. Surette. 1998. “Gender Differences in the Allocation of Assets in Retirement Savings Plans.” American Economic Review 88 (2): 207–11. Sunstein, Cass R., and Richard H. Thaler. 2003. “Libertarian Paternalism Is Not an Oxymoron.” University of Chicago Law Review 70 (4): 1159–202. Thaler, Richard. 1981. “Some Empirical Evidence on Dynamic Inconsistency.” Economics Letters 8 (3) 201–07. Thaler, Richard H., and Shlomo Benartzi. 2004. “Save More TomorrowTM: Using Behavioral Economics to Increase Employee Saving.” Journal of Political Economy 112 (S1): S165–87. Tversky, Amos, and Daniel Kahneman. 1974. “Heuristics and Biases: Judgment under Uncertainty.” Science 185 (4157): 1124–30. Vanguard. 2013. “Beyond Project Nollywood.” March 14. http://www.vanguardngr. com/2013/03/beyond-project-nollywood/. Willis, Lauren. 2011. “The Financial Education Fallacy.” American Economic Review Papers and Proceedings 101 (3): 429–34. 10.  Nigeria’s Nollywood nudge  ◾ 355 ANNEX:  DETAIL TABLES 356  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10A.1  Descriptive statistics (female) TOTAL SAMPLE CONTROL MOVIE MFB MOVIE/MFB PURE CONTROL MEANS BY GENDER VARIABLE N MEAN MEAN MEAN P-VALUE MEAN P-VALUE MEAN P-VALUE MEAN P-VALUE MALE FEMALE Personal characteristics Age of respondent 1,642 38.16 38.59 38.38 0.783 38.13 0.547 37.34 0.081* 38.48 0.900 36.79 38.16 Married 1,674 0.89 0.89 0.88 0.682 0.90 0.676 0.87 0.428 0.90 0.647 0.72 0.89 Widowed 1,674 0.03 0.02 0.04 0.103 0.01 0.321 0.02 0.929 0.03 0.465 0.00 0.03 Single 1,674 0.08 0.09 0.07 0.563 0.09 0.998 0.10 0.410 0.07 0.343 0.27 0.08 Muslim 1,674 0.34 0.34 0.38 0.341 0.32 0.470 0.37 0.484 0.29 0.192 0.40 0.34 Christian 1,674 0.66 0.66 0.62 0.341 0.68 0.524 0.63 0.436 0.71 0.192 0.59 0.66 Can speak English 1,667 0.68 0.66 0.64 0.448 0.68 0.561 0.69 0.468 0.72 0.169 0.77 0.68 Igbo 1,674 0.17 0.15 0.15 0.971 0.19 0.130 0.18 0.271 0.20 0.050* 0.28 0.17 Yoruba 1,674 0.78 0.81 0.80 0.746 0.76 0.086* 0.77 0.178 0.75 0.058* 0.67 0.78 Other ethnicitiy 1,674 0.05 0.04 0.05 0.588 0.05 0.512 0.05 0.503 0.05 0.842 0.05 0.05 Education No completed school 1,673 0.08 0.08 0.08 0.689 0.09 0.415 0.08 0.856 0.07 0.939 0.06 0.08 Primary school 1,673 0.23 0.25 0.25 0.995 0.20 0.091* 0.22 0.341 0.20 0.107 0.20 0.23 High school diploma 1,673 0.48 0.46 0.45 0.936 0.50 0.253 0.48 0.607 0.51 0.178 0.56 0.48 Diploma 1,673 0.11 0.12 0.09 0.366 0.11 0.819 0.10 0.625 0.13 0.543 0.09 0.11 Graduate school 1,673 0.10 0.10 0.11 0.762 0.10 0.956 0.12 0.486 0.08 0.441 0.09 0.10 Household characteristics Household size 1,665 4.63 4.73 4.71 0.861 4.51 0.060* 4.53 0.103 4.68 0.703 4.29 4.63 # of children below 12 1,644 1.35 1.38 1.38 0.970 1.33 0.591 1.26 0.193 1.41 0.743 1.27 1.35 # of dependents 1,647 2.28 2.32 2.24 0.560 2.25 0.616 2.27 0.716 2.32 0.993 2.82 2.28 # of dependents outside HH 1,572 1.41 1.23 1.38 0.291 1.58 0.024 1.33 0.477 1.58 0.030** 1.88 1.41 Business characteristics Months in operation 1,632 96.50 95.77 94.33 0.828 99.27 0.616 101.12 0.447 90.33 0.452 99.58 96.50 Has a savings account 1,668 0.54 0.51 0.55 0.307 0.51 0.913 0.54 0.438 0.64 0.001*** 0.64 0.54 Keeps written fin. records 1,662 0.37 0.36 0.36 0.902 0.35 0.968 0.38 0.458 0.39 0.363 0.38 0.37 Op. inside main market 1,648 0.30 0.27 0.32 0.147 0.28 0.679 0.29 0.470 0.33 0.082* 0.14 0.30 Number of employees 1,672 1.27 1.38 1.31 0.551 1.29 0.482 1.28 0.435 1.02 0.004*** 1.86 1.27 Business exp. in years 1,667 10.49 10.89 10.86 0.956 10.51 0.553 10.08 0.190 10.04 0.219 11.37 10.49 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 357  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10A.2  Descriptive statistics (male) TOTAL SAMPLE CONTROL MOVIE MFB MOVIE/MFB PURE CONTROL MEANS BY GENDER VARIABLE N MEAN MEAN MEAN P-VALUE MEAN P-VALUE MEAN P-VALUE MEAN P-VALUE MALE FEMALE Personal characteristics Age of respondent 672 36.79 35.97 35.53 0.715 37.34 0.231 37.23 0.296 38.35 0.054* 36.79 38.16 Married 683 0.72 0.73 0.68 0.315 0.77 0.446 0.69 0.438 0.75 0.817 0.72 0.89 Widowed 683 0.00 0.00 0.01 0.349 0.00 0.00 0.02 0.129 0.00 0.03 Single 683 0.27 0.27 0.31 0.376 0.23 0.446 0.31 0.438 0.23 0.585 0.27 0.08 Muslim 682 0.40 0.38 0.45 0.250 0.41 0.617 0.35 0.648 0.40 0.734 0.40 0.34 Christian 682 0.59 0.61 0.55 0.307 0.57 0.469 0.63 0.732 0.60 0.828 0.59 0.66 Can speak English 679 0.77 0.80 0.74 0.318 0.81 0.810 0.77 0.600 0.75 0.375 0.77 0.68 Igbo 682 0.28 0.26 0.23 0.638 0.27 0.824 0.31 0.297 0.33 0.210 0.28 0.17 Yoruba 682 0.67 0.69 0.71 0.662 0.72 0.628 0.60 0.129 0.63 0.367 0.67 0.78 Other ethnicitiy 682 0.05 0.05 0.05 0.991 0.01 0.072* 0.08 0.315 0.03 0.490 0.05 0.05 Education No completed school 683 0.06 0.02 0.05 0.187 0.06 0.101 0.08 0.041** 0.09 0.024** 0.06 0.08 Primary school 683 0.20 0.21 0.21 0.980 0.22 0.901 0.17 0.435 0.17 0.451 0.20 0.23 High school diploma 683 0.56 0.59 0.55 0.456 0.51 0.165 0.60 0.860 0.56 0.587 0.56 0.48 Diploma 683 0.09 0.10 0.11 0.822 0.10 0.998 0.06 0.186 0.07 0.418 0.09 0.11 Graduate school 683 0.09 0.08 0.07 0.939 0.11 0.299 0.09 0.652 0.11 0.316 0.09 0.10 Household characteristics Household size 678 4.29 4.15 4.23 0.713 4.26 0.620 4.38 0.369 4.45 0.242 4.29 4.63 Number of children below 12 667 1.27 1.39 1.09 0.040** 1.24 0.331 1.20 0.238 1.50 0.557 1.27 1.35 Number of dependents 675 2.82 2.79 2.75 0.870 2.78 0.961 2.74 0.849 3.12 0.249 2.82 2.28 # of dependents outside HH 641 1.88 2.25 1.86 0.259 1.41 0.015** 2.06 0.598 1.82 0.258 1.88 1.41 Business characteristics Months in operation 678 99.58 106.76 105.08 0.885 91.69 0.175 100.78 0.607 92.53 0.245 99.58 96.50 Has a savings account 682 0.64 0.70 0.62 0.176 0.61 0.144 0.64 0.347 0.62 0.189 0.64 0.54 Keeps written fin. records 678 0.38 0.38 0.32 0.342 0.42 0.506 0.36 0.796 0.40 0.693 0.38 0.37 Operating inside main market 676 0.14 0.16 0.12 0.344 0.14 0.674 0.17 0.845 0.14 0.643 0.14 0.30 Number of employees 680 1.86 2.10 1.80 0.309 1.65 0.135 1.67 0.162 2.15 0.901 1.86 1.27 Business experience in years 683 11.37 10.70 10.57 0.883 11.42 0.481 11.48 0.448 13.04 0.033** 11.37 10.49 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 358  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 10A.3  Balance across screening participants (female) TOTAL ATTENDED TREATMENT MISSED TREATMENT VARIABLE N MEAN N MEAN N MEAN P-VALUE Personal characteristics Age of respondent 1,386 38.10 893 38.73 493 36.96 0.001*** Gender (male) 1,416 0.00 908 0.00 508 0.00 Married 1,415 0.89 907 0.89 508 0.88 0.79 Widowed 1,415 0.02 907 0.03 508 0.01 0.047** Single 1,415 0.09 907 0.08 508 0.10 0.16 Muslim 1,416 0.35 908 0.34 508 0.37 0.19 Christian 1,416 0.65 908 0.66 508 0.63 0.22 English 1,410 0.67 903 0.69 507 0.64 0.044** Igbo 1,415 0.16 908 0.17 507 0.15 0.36 Yoruba 1,415 0.78 908 0.78 507 0.79 0.86 Other ethnicitiy 1,415 0.05 908 0.05 507 0.06 0.23 Education No completed school 1,415 0.08 908 0.07 507 0.10 0.034** Primary school 1,415 0.23 908 0.22 507 0.25 0.33 High school diploma 1,415 0.47 908 0.47 507 0.47 0.91 Diploma 1,415 0.11 908 0.12 507 0.09 0.069* Graduate school 1,415 0.11 908 0.11 507 0.09 0.33 Household characteristics Household size 1,409 4.62 903 4.57 506 4.71 0.14 Number of children below 12 1,394 1.34 892 1.32 502 1.37 0.40 Number of dependents 1,393 2.27 895 2.28 498 2.24 0.72 # of dependents outside HH 1,329 1.38 850 1.36 479 1.41 0.73 Business characteristics Months in operation 1,384 97.60 885 96.58 499 99.42 0.58 Has a savings account 1,410 0.53 908 0.55 502 0.48 0.013** Keeps written fin. records 1,404 0.36 902 0.38 502 0.33 0.033** Op. inside main market 1,412 0.29 908 0.32 504 0.24 0.001*** Number of employees 1,414 1.32 907 1.32 507 1.32 0.99 Business experience in years 1,409 10.57 904 10.84 505 10.09 0.11 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 10.  Nigeria’s Nollywood nudge  ◾ 359 TABLE 10A.4  Balance across screening participants (male) TOTAL ATTENDED TREATMENT MISSED TREATMENT VARIABLE N MEAN N MEAN N MEAN P-VALUE Personal characteristics Age of respondent 560 36.48 349 37.08 211 35.49 0.071* Gender (male) 568 1.00 352 1.00 216 1.00 Married 568 0.72 352 0.75 216 0.66 0.019** Widowed 568 0.00 352 0.00 216 0.00 0.434 Single 568 0.28 352 0.24 216 0.34 0.015** Muslim 567 0.40 352 0.38 215 0.42 0.408 Christian 567 0.59 352 0.61 215 0.57 0.342 English 564 0.78 352 0.81 212 0.73 0.021** Igbo 567 0.27 352 0.29 215 0.24 0.172 Yoruba 567 0.68 352 0.68 215 0.68 0.998 Other ethnicitiy 567 0.05 352 0.03 215 0.08 0.005*** Education No completed school 568 0.05 352 0.04 216 0.08 0.047 Primary school 568 0.21 352 0.20 216 0.21 0.914 High school diploma 568 0.56 352 0.57 216 0.55 0.737 Diploma 568 0.09 352 0.09 216 0.09 0.855 Graduate school 568 0.09 352 0.10 216 0.06 0.127 Household characteristics Household size 563 4.25 348 4.39 215 4.03 0.036** Number of children below 12 554 1.22 342 1.30 212 1.11 0.086* Number of dependents 561 2.76 346 2.95 215 2.47 0.007*** # of dependents outside HH 533 1.89 329 1.92 204 1.84 0.740 Business characteristics Months in operation 563 101.02 350 104.27 213 95.68 0.282 Has a savings account 567 0.64 352 0.67 215 0.59 0.046** Keeps written fin. records 564 0.37 352 0.41 212 0.31 0.023** Op. inside main market 567 0.15 352 0.17 215 0.11 0.038** Number of employees 566 1.80 352 1.81 214 1.78 0.891 Business experience in years 568 11.03 352 10.98 216 11.12 0.851 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. CHAPTER 11 H arnessing emotional connections to improve financial decisions Using savings lotteries to promote financial inclusion in South Africa GUNHILD BERG AND BILAL ZIA ABSTRACT This chapter exploits the emotional connections and viewer attentiveness of mainstream media to evaluate the economic impact of financial education messages on debt management delivered through a popular television soap opera in South Africa. The study uses a symmetric encouragement design to compare outcomes of individuals who were randomly assigned to watch a soap opera with financial messages, Scandal!, to those of individuals who were invited to watch a similar soap opera without financial messages, Muvhango. Both shows overlapped in evening primetime and had similar past viewership profiles. The financial storyline spanned two months and featured one of the leading characters of the show borrowing excessively We would like to thank our direct project counterparts, Ochre Media, the production company of Scandal!, in particular Stan Joseph and Lebogang Ramafoko who coor- dinated the development of the storyline; the National Debt Mediation Association for supporting the project, in particular Magauta Mphalele; and TNS South Africa, in particular Bobby Berkowitz, who offered keen intellectual insights into the survey design. We are also grateful for comments and input received from Claus Astrup, Michael Fuchs, Richard Hinz, Ruth Kagia, Elaine Kempson, Lawrence Kincaid, Thomas Losse-Mueller, Matthias Lundberg, David McKenzie, Margaret Miller, Florentina Mulaj, Leopold Sarr, Robert Walker, and the South Africa Advisory Panel consisting of repre- sentatives from the National Treasury, the Financial Services Board, the National Credit Regulator, and the Credit Ombudsman. Special thanks to Christian Salas Pauliac and Julia Thelen for exceptional research assistance. We gratefully acknowledge funding received from the Russia Financial Literacy and Education Trust Fund. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 361 362  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES and irresponsibly through hire purchase, gambling, ending up in financial distress, and eventually seeking help to find her way out. Two intermediate and one final follow-up surveys were conducted as part of the study. The analysis finds individ- uals assigned to watch Scandal! had significantly higher financial knowledge of the issues highlighted in the soap opera storyline, in particular messages deliv- ered by the leading character. On behavior, Scandal! viewers were almost twice as likely to borrow from formal sources, less likely to engage in gambling, and less prone to enter hire purchase agreements. Messages promoting a national debt mediation helpline delivered by an external character did not sustain traction beyond immediate interest. Three qualitative focus groups highlight the impor- tance of emotional connections with the leading character in motivating behavior change. 11.1 INTRODUCTION Financial education is important, yet there is a considerable knowledge gap in determining how best to deliver it. The literature on careful evaluations of finan- cial literacy is small but growing, and has moved away from classroom-based interventions, such as Cole, Sampson, and Zia (2011), to more innovative delivery mechanisms, such as videos and DVDs (Carpena et al. 2011). Yet the scope and reach of even the best produced DVDs is limited on the supply side, and attracting viewership can be significantly challenging on the demand side. Entertainment media—television and radio—offer not only broader outreach, since nearly every household nowadays has a TV, but also a captive audience. Furthermore, as emotional connections are established between a show and its audience, the program provides a potentially powerful platform for communi- cating messages and influencing behavior. There is considerable anecdotal and some survey evidence on the impact of mass media on behavior, especially in the areas of health and education. For instance, as Brazil’s Rede Globo television network grew through the 1970s and 1980s, women began having fewer children, and experienced the same decrease in fertility as with two extra years of education (La Ferrara, Chong, and Duryea 2008). Similarly, one of the most successful public health campaigns in the Arab Republic of Egypt—the central component of which was a soap opera—was highly successful in improving the use of oral rehydration therapy and reducing infant mortality rates by 70  percent (Abdulla 2004). In other studies, the introduction of television has been linked to decreased domestic violence and fertility rates in India (Jensen and Oster 2009), and lower adolescent drug use and increased contraception adoption in Brazil (Verner and Cardoso 2007). The emotional connections channel is particularly central to media campaigns. Advertisers, for example, consistently and successfully use influence on emotions to affect consumer choices (Bagwell 2007; Russo, Carlson, and Meloy 2006; Stigler 1987). In the financial context, Bertrand et al. (2010) and Agarwal and Ambrose 11.  Harnessing emotional connections to improve financial decisions  ◾ 363 (2007) show that persuasive advertising and direct mail solicitation can influence consumer loan contract choices. Perhaps the most direct anecdotal evidence on the power of emotional connections with television personalities comes from the United States, where Kennedy et al. (2004) find that the greatest number of calls to the Centers for Disease Control and Prevention’s AIDS hotline in one year came not as a result of a million-dollar public awareness campaign or through news and documentary broadcasts, but rather in response to a beloved soap opera char- acter dealing with his HIV-positive diagnosis.1 This chapter contributes to the literature by carefully analyzing the impact of financial education messages delivered through a popular South African television soap opera, Scandal! In partnership with the production company of Scandal!, the soap storyline features one of the main characters borrowing excessively through hire purchase, gambling, falling into a debt trap, and eventually seeking help to find her way out and manage her debt responsibly. The storyline plays out over a two-month period, and the study evaluates the effectiveness of these messages through two intermediate surveys and one final follow-up survey conducted four months later. Further, three qualitative focus groups provide insight into the mechanisms behind measured impacts. The methodological challenge associated with such an evaluation is nontrivial. Critically, it is impossible to externally restrict viewership since the soap opera is nationally televised with no variation in TV signal strength, at least in urban areas where this study focuses. Even an encouragement design methodology is limited in scope as we can only encourage a set of individuals to view Scandal!, but in no way can we prevent others (particularly those in the control group) from doing the same. To overcome this constraint, we employ a symmetric encouragement design in this study. Specifically, while a randomly selected treatment group is encour- aged through financial incentives to view Scandal!, another set of individuals—the control group—is similarly encouraged through identical financial incentives to view a comparable soap opera (Muvhango) that overlaps with Scandal! in tele- vision primetime. Hence, we are able to create an artificial separation between treatment and control groups that enables us to identify the treatment effects of financial literacy delivered through Scandal! Our results are quite encouraging. First, we confirm that the overwhelming majority of individuals (96  percent) assigned to watch the two soap operas in fact watched them, with very little overlap—less than 12 percent of Muvhango viewers claim to have also watched Scandal! during the study period, and a third of these viewers watched several other shows too. Moreover, such crossovers will only underestimate treatment effects. 1  See Chong and Ferrara (2009) and Kenny (2009) for similar findings. 364  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES We measure and study outcomes along the causal chain of impact from financial knowledge to financial behavior, as per Carpena et al. (2011). The anal- ysis shows significant improvements in content-specific financial knowledge—a 4.5  percentage point increase—and no differences in knowledge of financial concepts that were external to the soap storyline. Central to the theme of the soap was highlighting the high interest charges and hidden costs of borrowing through hire purchase. Our analysis on finan- cial behaviors finds a significant and substantial increase in the likelihood of borrowing through formal channels—22 percent compared to 13 percent in the control group—and a greater likelihood to borrow for productive purposes such as investments in household durable goods and vehicles. We also directly measure hire purchase use and find a significant reduction in the likelihood of entering into such contracts—15 percent compared to 19 percent in the control group. Further, we find a 5 percentage point lower propensity to gamble among the treatment group, yet another theme highlighted in the show. These are promising results and point to successful dissemination of themes depicted in the soap storyline. One aspect of the storyline that did not sustain traction beyond immediate effects was a public call to action for the South African National Debt Mediation Association (NDMA). The soap featured an NDMA counselor who helps the main character build a financial plan to enable her to save enough to repay her debts. The actual NDMA toll-free helpline also appeared at the end of several shows. The findings on this public call to action are quite stark. While we see a huge uptake in the number of calls received within days of the NDMA being featured in the soap, the longer-term analysis shows no recall of the NDMA or even memory that a debt counselor appeared in the soap storyline. These quantitative results hint at the mechanism of our measured impacts. Specifically, the behavior change in our study tracks well with the storyline that featured the main soap character, who herself depicted poor financial behavior before transforming her habits. This character was part of the show prior to the two-month financial storyline and remained part of the show afterwards. The nonresult on knowledge of the NDMA could at least partly be explained by the fact that the NDMA messages were delivered not by the main character, but by a different character who was external to the main soap cast and who appeared in two or three episodes as an NDMA counselor and then disappeared. Hence, the audience did not get the opportunity to establish or maintain a connection or following with this character. In order to further assess the merits of this emotional connections channel, we conducted three qualitative focus group discussions with study participants three months after the final quantitative survey. In these meetings, the focus group leaders were unable to elicit recall of NDMA from the participants; only a few women seemed to remember, and they associated their memory more with the fact that the NDMA character was female rather than her representation of the NDMA. In contrast, all participants readily recalled the main soap character 11.  Harnessing emotional connections to improve financial decisions  ◾ 365 and aspects of the financial storyline delivered by her. Upon further inquiry, they attributed better recall to the fact that the character was a mainstay in the soap opera, and was popular and likable. Overall, our findings suggest an important role for entertainment media as an accessible and important tool for policy makers to deliver carefully designed educational messages that resonate with the audience and can potentially influence financial knowledge and behavior. Further, our findings suggest that emotional connections and familiarity with media personalities certainly play a role in motivating knowledge and behavior change among viewers, and that harnessing such potential can be an important channel for achieving develop- ment impact. This chapter is organized as follows. Section 11.2 provides a brief litera- ture review on financial literacy and mass media. Section 11.3 focuses on finan- cial education in South Africa and explains the study context and soap opera storyline. Section 11.4 outlines the research methodology and study timeline, and section 11.5 presents summary statistics. Section 11.6 provides the main analysis and discussion of results, and section 11.7 concludes. 11.2 BACKGROUND AND LITERATURE 11.2.1 General financial education evaluations Making informed financial decisions and evaluating complex financial choices remains a challenge for large parts of the population across the world (Lusardi and Mitchell 2007; Lusardi, Mitchell, and Curto 2010). While financial literacy training has been heavily promoted as a solution, evidence of its effectiveness is limited. A recent study finds muted evidence that financial education influences the demand for savings accounts in Indonesia (Cole, Sampson, and Zia 2011). A more comprehensive video-based financial education program in India is asso- ciated with large and statistically significant improvements on knowledge and awareness of financial products and services, and some moderate improvements in savings rates (Carpena et al. 2011). To date, debt literacy research has been limited to survey-based studies. Lusardi and Tufano (2009) find that debt literacy levels are generally low in the United States, especially among women; and that some groups perceive their knowledge to be higher than it actually is. More work has been published on the effects of mortgage counseling,2 however, such facilities are rarely available in developing countries. See for example, Agarwal et al. 2010; Elliehausen, Lundquist, and Staten 2007; Hartarska 2  and Gonzalez-Vega 2005; and Hirad and Zorn 2001. 366  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 11.2.2 Mass media for edutainment Mass media outlets, including television and radio, have been used successfully to deliver educational messages in developing and developed countries (Kenny 2009). For example, Verner and Cardoso (2007) find that access to TV lowers the probability of teen pregnancy and risk-taking in Brazilian favelas. Similarly, Jensen and Oster (2009) analyze the power of TV to shift cultural roles of women by studying the effects of satellite or cable service installation in Indian villages. Access to TV increases women’s autonomy and female school enrollment, while fertility, tolerance of domestic violence, and the preference for sons appear to decrease. Entertainment education through soap operas is quite popular in Spanish-lan- guage telenovelas. The Peruvian series Simplemente Maria became a success in educating and inspiring people to take actions to improve their own lives (Singhal, Rogers, and Brown 1993). Similarly, the U.S. soap opera The Bold and the Beautiful successfully reached at-risk minority women when a storyline was accompanied by display of a free hotline number. Further, after an HIV-positive character was diagnosed—a subplot that lasted only two shows—the number of calls to the U.S. Centers for Disease Control and Prevention’s AIDS hotline spiked substantially. Examples of education through entertainment media come from other parts of the world as well. The radio soap opera Apwe Plezi was broadcast between February 1996 and September 1998 in St. Lucia, with themes including family planning and sexually transmitted disease prevention. Vaughan, Regis, and St. Catherine (2000) carried out personal interviews (753 pretest and 1,238 posttest) and find that although only small changes were observed, the program had a positive effect on knowledge and change of attitudes with a minimal change in behavior. Similarly, Paluck and Green (2009) evaluate the effectiveness of a radio program on conflict resolution and intergroup tolerance in Rwanda using a group-randomized research design. For one year, the treatment group was exposed to the reconciliation program, while the control group listened to a program about health and HIV. The authors observe increased dissent from popular opinion and also increased collective action among the treatment group. In Tanzania, the soap opera Twende Na Wakati was broadcast between 1993 and 1997, with messages promoting adoption of family planning methods and HIV/ AIDS prevention. Rogers et al. (1999) and Rogers and Vaughan (2000) evaluate the effectiveness of this program and find that the soap opera had strong behavioral effects on family planning. Yet another example is from Brazil, where La Ferrara, Chong, and Duryea (2008) study the effects of novelas on fertility rates in the 1970s and 1980s. The authors find that as a result of the expansion of the Rede Globo network (which enjoyed a monopoly position as the supplier of novelas until the 1990s), family size was reduced, women began naming their children after soap opera characters, 11.  Harnessing emotional connections to improve financial decisions  ◾ 367 and separation and divorce figures went up in the areas that were exposed to the network. The results can partly be explained by the strong influence of television in Brazil. In Kazakhstan, Seeking Happiness, a six-episode soap opera funded by the U.S. Agency for International Development, was used to explain capitalism—especially privatization and rule of law—to the former Soviet country. According to Mandel (1998), the project failed since the local production company was unfit to produce a social marketing soap opera, the budget was too small to produce the series well, and the U.S. Agency for International Development decision makers lacked experience in rural Kazakhstan. Another soap opera, Crossroads, funded by the British government, took a different approach to social engineering in Kazakhstan; it ended up becoming one of the most popular Kazakhstani television programs in 1998. The most important element for success was ensuring high quality at all production levels, especially a good storyline and convincing actors. Finally, Brown and Cody (1991) explore the effects of a pro-social soap opera, Hum Log, in promoting the status of women in India. The authors find that although the soap was popular, it did not significantly influence society’s percep- tion of women. Overall, the use of mass media for spreading social messages is quite common both in developed and developing countries, with mixed evidence on effectiveness. 11.2.3 Mass media and financial education The use of mass media, and specifically soap operas, to communicate financial capability-related messages is a relatively new approach to financial education, and the few existing financial literacy mass media programs have not yet been rigorously tested for their effectiveness. The telenovela Nuestro Barrio in the United States was one of the earliest soap operas to incorporate financial literacy messages. It was aimed at educating Hispanic immigrants about financial matters and promoting the use of bank accounts. It was primarily directed toward households outside the formal finan- cial system with the goal of making these products and services attractive. Similarly, Makatuno Junction, a Kenyan soap opera that is still in produc- tion, includes financial literacy themes. Since 2005, the series has reached large audiences in Kenya (over 7 million viewers in 2009), Uganda (3 million viewers in 2009), and Tanzania (2 million viewers). Among the many storylines featured in the soap, the series has encouraged viewers to open bank accounts, to save, and to budget. The soap has also featured other social messages, such as malaria prevention, contraception, and schooling. Yet, to date, no careful evaluation of these messages has been conducted. Apart from television, the online world has stepped into the fray of finan- cial education. The most prominent example is the U.S. nonprofit, Doorway to 368  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Dreams (D2D), which offers online games that teach players important lessons on managing finances. Tufano, Flacke, and Maynard (2010) analyze the impact of these video games on low-income and minority adults and find that the games improved financial knowledge and confidence, although the sample size and scope of this study were fairly limited. Yet another entertainment outlet, theater, has been used to spread financial messages. The VISA Financial Literacy Road Show is a traveling theatrical perfor- mance that provides educational messages on financial matters in unbanked communities in Sub-Saharan Africa. However, the program has never been eval- uated. 11.2.4 Evidence on emotional connections through media The evidence on using emotional connections to connect with target audiences is varied. Several factors and external impulses have been shown to affect consumer decisions. Numerous factors—from the messenger, to the setting, to the method of delivery and phrasing of the message—matter when communi- cating information. The clearest example of persuasive messaging comes from advertising. Advertisers regularly use influence of emotions to affect and complement consumer preferences (Bagwell 2007; Stigler 1987). Russo, Carlson, and Meloy (2006) further argue that persuasive advertising can lead decision makers to make inferior choices. Bertrand et al. (2010) are among the first to systematically study how persuasive advertising influences consumer choice. The authors carry out a large-scale direct mail field experiment to study the effects of advertising content on real decisions in South Africa, and find that factors such as framing and phrasing—such as including a picture of an attractive female accounts manager—influences loan demand. Similarly, Agarwal and Ambrose (2007) study the impact of direct mail advertising on financial decisions and find that adver- tising campaigns have a persuasive effect on loan contract choices. Outside of the financial realm, Evans (1963) finds that demographic similar- ities between the client and the salesperson influence retail demand. Similarly, an experiment involving door-to-door charitable fundraising finds that attractive female solicitors secure significantly more donations (Landry et al. 2006). Further evidence on the influence of emotions through media comes from Anschutz et al. (2008), who find that body size of television actors affects personal body dissatis- faction and actual food intake. 11.  Harnessing emotional connections to improve financial decisions  ◾ 369 11.3 FINANCIAL EDUCATION IN SOUTH AFRICA 11.3.1 The South African context Household debt has been a large and growing problem for South Africa over the last decade. The ratio of household debt to disposable income was 76.3 percent in the second quarter of 2012, up from 50 percent in 2002 (South African Reserve Bank 2012). In June 2012, the National Credit Regulator reported that, of the 19.6 million consumer borrowers in its records, 9.22 million (47 percent) had impaired records (NCR 2012). Of the consumers with impaired records, 19.5 percent were in arrears for three months or more, 13.3 percent had adverse listings, and 14.2 percent had judgments and administration orders issued against them. What is particularly troubling is that the percentage of consumers with impaired records has been above 45 percent since December 2009. Financial woes in South Africa are not limited to indebtedness. The savings rate of households is, at 1.7 percent in the second quarter of 2012, very low (South African Reserve Bank 2012). In fact, according to Finmark Trust (2012), a large percentage of the South African population (67 percent) does not save at all, even though the majority of adults believe that saving is important. In addition, of those who save, only 22 percent save through formal channels, either with banks or other formal financial institutions. The Financial Services Board, the statutory regulator of the nonbanking finan- cial sector in South Africa, undertook a nationally representative study on finan- cial literacy levels in South Africa in 2011.3 The survey focused on four key areas of financial literacy: day-to-day money management, financial planning, choosing appropriate products, and financial knowledge and understanding. One of the central findings from this survey was that financial literacy is, to a large extent, dependent on educational attainment and income levels (Roberts and Struwig 2011). For example, the presence of a household budget—indicative of a higher awareness of financial management—is strongly related to income levels, with 79 percent of households in the top income quintile having household budgets, compared to 36 percent in the lowest income quintile. Income differences are also prevalent in the sources of information that influ- ence individuals’ decisions regarding financial products. While individuals in the richest quintile are far more likely to consult the Internet, independent financial advisers, or friends in the financial services industry, the poor are more likely to rely on public resources such as newspapers, radio, and television. In general, 3  A total of 3,112 South Africans (individuals aged 16 and above living in South Africa) partic- ipated in the survey. The study was undertaken in collaboration with the International Network on Financial Education of the Organisation for Economic Co-operation and Devel- opment. 370  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES the importance of public media for financial decisions is high across all income groups, with 33 percent of respondents indicating that TV and radio programs or advertisements were the source of information that most influenced their deci- sions about financial products; this shows the potential media has to influence financial decisions and behavior in South Africa. Against this backdrop, financial education through popular media is poten- tially a powerful tool to stimulate improvements in financial knowledge and behavior. This chapter exploits this potential to study the effectiveness of finan- cial messages delivered through a popular soap opera. 11.3.2 Scandal! South Africa has a well-developed media industry with popular television shows reaching millions of viewers. Further, educational messages in health and educa- tion have previously been incorporated in soap operas such as Soul City. The target soap opera for this study is Scandal!, which has been running for eight years (more than 1,500 episodes) and produces four weekly episodes. The show is broadcast on eTV, the second most popular station in South Africa,4 and is especially popular among low-income South Africans. The storyline for this study was developed by the production company of Scandal!, Ochre Media, together with the NDMA and a team of financial capability and social marketing experts. Messages in the storyline were tested through focus groups to ensure that they were correctly understood by the target group, that the plot appeared realistic, and that the target group could identify with the characters and their problems. All suggested changes and recommendations by the focus group members were included in the final storyline. The storylines focused on debt aspects of financial capability, including sound financial management (e.g., using credit wisely and borrowing from formal sources), avoiding debt traps (e.g., impulse buying and living beyond one’s means), and getting out of debt (e.g., practical steps for seeking help). The financial educa- tion storyline stretched over a period of two consecutive months, a time that was deemed necessary for the viewers to emotionally connect with the characters and their problems. Specifically, between February 13 and March 27, 2012, 26 episodes with financial literacy messages were aired. The objective of the storyline was to positively affect the behavior of medium- to low-income South Africans regarding financial literacy and management by increasing their knowledge and understanding of personal financial management. The story highlighted five major themes: ◾◾ Getting into debt—caused by financial mismanagement, e.g., impulse buying and/or living beyond one’s means 4  For more information, see http://www.etv.co.za/scandal (accessed April 22, 2014). 11.  Harnessing emotional connections to improve financial decisions  ◾ 371 ◾◾ The effects of financial mismanagement and debt on family and home life ◾◾ Acknowledgment that a problem exists with managing finances that has led to debt, which in turn has led to other problems ◾◾ Getting out of debt—practical steps for seeking help, e.g., debt manage- ment, assessment tools, and debt recovery ◾◾ Sound financial management, e.g., using credit wisely, borrowing respon- sibly, feeling confident. In addition, the NDMA was featured prominently in the storyline, and the main character and her husband seek advice from an NDMA employee who helps them find their way back to financial responsibility. At the end of each episode in which the NDMA was highlighted, the actual NDMA toll-free number was displayed on screen with a public announcement offering NDMA’s free financial advisory service. A detailed synopsis of the show is included in the annex to this chapter. Further, a short summary video clip of the storyline is available on YouTube.5 11.4 RESEARCH DESIGN AND TIMELINE 11.4.1 Research design Evaluating the impact of a nationally televised soap opera is quite challenging. First, it is difficult to separate the effect of the soap opera’s message from other messages on similar issues that individuals and families may be receiving from alternative sources. Second, certain types of individuals may self-select into watching a particular soap opera, and hence any subsequent behavior change is confounded by these selection issues. Third, it is difficult to identify the cause of any subsequent behavior change as the message itself because the effect of the message is confounded by the mode of delivery. Specifically, since on-screen actors typically have a mass following, many of those watching may simply follow what their favorite actors advocate regardless of the content of the message. To overcome these concerns, the project used a random encouragement design methodology with the following setup. First, a listing exercise of the study population was conducted during which basic socioeconomic and TV viewer- ship information was gathered from all households. The evaluation, and hence the listing exercise, was concentrated in the area of urban Gauteng (with 2,500 respondents from Johannesburg and 500 respondents living in Pretoria). The listing was done for 3,000 people through physical visits (face to face). It was ensured that all individuals listed were responsible for their household’s spending 5  http://www.youtube.com/watch?v=ys5eSxTetF4 (accessed April 22, 2014). 372  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES decisions, watched television soap operas, were 20 years of age or older, and belonged to medium- to low-income groups (i.e., classified as 5 to 7 on the Living Standards Measure scale).6 To avoid individuals in the sample communicating with each other about the soap operas they watch (and thus potentially influ- encing each other), only one interview per home/family was conducted, only one interview per street was carried out, and two streets were mechanically skipped before visiting another household. Next, the study sample of 1,031 individuals was chosen and randomized into treatment and control groups. The original study design called for stratification by gender, income level, education, and the likelihood of watching soap operas to allow for measuring heterogeneous impacts; however, due to an undetected coding error, the randomization occurred unconditionally rather than within strata. Nevertheless, as the summary statistics in table 11.1 show, the sample is balanced across treatment and control on these variables. Since the soap operas were nationally televised, we could not control or restrict viewership. In order to encourage individuals to abide by our assignments and to create a separation between treatment and control groups, a financial incentive was provided for individuals in both groups to watch their respective assigned show. Initial telephone calls were made to explain these details to the study sample. They were told that we would call them every three to four weeks during the next three months and ask them questions about the soap storyline, and if they answered three out of four questions correctly they would win R 60 (US$7) each time. This amount was agreed upon after consulting with the survey and production firm and was deemed high enough to encourage participation. An auxiliary advantage of providing financial incentives to both treatment and control groups was that any confounding effects of the incentive itself on subsequent financial attitude or behavior change (i.e., Hawthorne effects) would be netted out.7 During these calls, in addition to questions about the storyline, the surveyors asked questions related to financial knowledge, attitudes, and behavior. Since the scope of questions asked over the telephone was limited, a face-to-face follow-up survey was conducted four months after conclusion of the storyline with more comprehensive knowledge and behavior questions. In addition, administrative daily call data were obtained from the NDMA so we could track incoming calls 6  The South African Audience Research Foundation’s Living Standards Measure is the most widely used marketing research tool in Southern Africa. It divides the population into 10 groups, with 10 being the highest and 1 the lowest. The scale cuts across race, gender, age, or any other variable used to categorize people. Instead, it groups people according to their living standards using criteria such as degree of urbanization and ownership of cars and major appliances. This design is similar to an experimental evaluation of radio programming to change social 7  norms and behavior regarding dispute resolution in Rwanda (Paluck and Green 2009). 11.  Harnessing emotional connections to improve financial decisions  ◾ 373 TABLE 11.1  Summary statistics and tests of randomization FULL SAMPLE TREATMENT GROUP CONTROL GROUP   N MEAN N MEAN N MEAN P-VALUE Respondent is female 1,031 0.70 553 0.69 478 0.71 0.577 Respondent has at least 1,031 0.70 553 0.69 478 0.71 0.394 secondary schooling Respondent has above 1,031 0.53 553 0.53 478 0.53 0.914 median financial literacy Respondent’s household 1,031 0.44 553 0.46 478 0.43 0.357 belongs to low LSM group Respondent has a job or paid 1,031 0.49 553 0.49 478 0.48 0.689 work Respondent has not borrowed 1,031 0.58 553 0.58 478 0.58 0.986 money in prior 6 months Respondent has watched 1,031 0.47 553 0.48 478 0.47 0.830 Scandal! in the past Note: This table presents demographic summary statistics for respondents at baseline by treatment status; p -values for equality of means tests across treatment and control groups are presented in the last column. LSM = Living Standards Measure. to the financial advisory call center following the public call to action. Finally, in order to be able to understand mechanisms of impact, three qualitative focus groups were conducted after the final follow-up survey. An important aspect of the above research design is the comparability of viewers across Scandal! and Muvhango. Clearly, if the demographic profiles varied substantially, we could have significant differences in take-up among our random sample of treatment and control groups. Hence, prior to the study, we researched all available alternative soap operas to Scandal! and found Muvhango to be very similar in content and in viewership profile.8 Both series had approximately 2.5  million adult viewers who came predominantly from mid- to low-income groups, and viewership was distributed similarly across race with most viewers being nonwhite and primarily Sotho-speaking. 11.4.2 Study timeline Figure 11.1 illustrates the study timeline. The study spanned a period of one year, from December 2011 to December 2012. The initial listing exercise described in the previous subsection was conducted in early December 2011, after which our study sample was selected. An initial computer-assisted telephone interview was Muvhango has run for the last three years on SABC2. It is set in a Venda community 8  both in rural Venda and in urban Soweto. For more information, see http://www.tvsa.co.za/ showinfo.asp?showid=540 (accessed April 22, 2014). 374  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 11.1  Study timeline Initial CATI Face-to-face 1st CATI follow-up interview Focus group Listing 2nd CATI follow-up discussions phase Dec Jan Feb March April May June July Aug Sept Oct Nov Soap opera aired Note: CATI = computer-assisted telephone interview. conducted in February 2012, prior to the airing of the storyline. The soap opera storyline then began airing from mid-February to end March. The first follow-up computer-assisted telephone interview was conducted in early March, the midpoint of the storyline; the second was in early April, at the conclusion of the storyline. The final face-to-face follow-up survey was conducted four months later, in late July 2012. The results from these surveys were then analyzed and issues highlighted for the qualitative study. Three focus group sessions were conducted in mid-November 2012 to conclude the study. 11.5 SUMMARY STATISTICS, RANDOMIZATION, AND TAKE-UP Table 11.1 presents baseline summary statistics that were gathered as part of the listing exercise. The table first presents means and standard deviations for the full sample, then separately for treatment and control groups, and finally pres- ents p-values for a test of differences in means between treatment and control groups. None of the differences are significant, which confirms that the random- ization was successful. Overall, the sample consists of 70  percent females, which is unsurprising since soap opera viewership is predominantly female. At the same time, however, we strived to include males in the sample; the surveyors made the extra effort to find males who watched soap operas during the listing phase. Since our sample is entirely urban, the level of formal education was relatively high, with 70 percent of the sample having at least graduated from secondary school. A little more than 50 percent scored above the median in the financial literacy test, which highlights 11.  Harnessing emotional connections to improve financial decisions  ◾ 375 that the mean and median are almost the same—thus, a nice bell-shaped distri- bution. A little less than half the sample (43–46 percent) belongs to the low Living Standards Measure group, thus providing variation in income levels. Half the sample also has a full-time job or other paid work. Nearly 60 percent of the sample had not borrowed money in the past six months at the time of the baseline survey; finally, about 50 percent of the sample had watched Scandal! in the past. Table 11.2 presents attrition results and notes two important points. First, the general level of attrition is very low—in the final follow-up survey from which we obtain knowledge and behavioral outcomes for our analysis, we were able to inter- view more than 93 percent of the baseline sample. (The response rates on the first two telephonic surveys were relatively lower, however: 85 percent and 76 percent, respectively.) Second, there is no significant difference between response rates of treatment and control groups. Some of the final interviews could not be conducted face to face, and we investigate whether the likelihood of face-to-face interview is correlated with treatment status; reassuringly, it is not (column 2). Nevertheless, we control for mode of interview in all subsequent analysis. TABLE 11.2  Sample attrition PRESENT IN FINAL FACE-TO-FACE INTERVIEW FOLLOW-UP IN FINAL FOLLOW-UP (1) (2) Invited to watch Scandal! 0.002 0.003 (0.016) (0.023) R-squared 0.000 0.000 Observations 1,031 1,024 Dependent variable mean in control group 0.933 0.839 Note: Robust standard errors in parentheses. This table presents attrition analysis for our study in column 1, and the effect that treatment might have had on the survey mode in column 2. All variables of interest used in the subsequent analysis come from the final follow-up survey. Note that the alternative to a final face-to-face interview was a telephonic interview. The total sample in both columns is all respondents present at baseline; column 2 has only 1,024 observations because of missing information on survey mode for 7 observations. Finally, the take-up within both treatment and control groups was very high. Out of the sample of individuals assigned to either group, 96  percent reported watching their assigned soap during the study period, and there was no varia- tion across treatment and control groups. Furthermore, table 11.3 investigates whether there was substantial crossover, or treatment contamination, with the control group also watching the treated show, Scandal!. In either follow-up 1 or 2, we find that less than 12 percent of Muvhango viewers claim to have also watched Scandal! in the last four weeks, and a third of these viewers watched several other shows too. The tabulations show that 2 percent of respondents watched Scandal! and another show, about 1  percent watched two other shows, and 0.5 percent watched three or more other shows. A large proportion—38 percent 376  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 11.3  Sample crossovers (%) SCANDAL! SCANDAL! SCANDAL! SCANDAL! ONLY + 1 OTHER + 2 OTHER + 3 OTHER + 4 OTHER OTHER SCANDAL! SHOW SHOWS SHOWS SHOWS SHOWS (1) (2) (3) (4) (5) (6) Follow-up 1 7.4 2.2 1.5 0.5 0.2 37.7 Follow-up 2 7.2 1.9 0.3 0.3 0.0 28.5 Note: This table tabulates responses to the following survey question asked in follow-up 1 and follow-up 2: “Apart from Muvhango, which other shows did you watch in the last four weeks?” The question was only asked of the sample in the control group—i.e., those individuals who were assigned to watch the soap opera Muvhango. The purpose was to assess whether individuals assigned to Muvhango also watched Scandal! during our study period. in follow-up 1 and 29 percent in follow-up 2 watched other shows entirely and did not watch Scandal!. While these are reassuring findings, we note that such crossovers will only underestimate treatment effects in our subsequent analysis. 11.6 RESULTS AND DISCUSSION This section presents the main analysis on financial knowledge and behavior. Since our research methodology uses random assignment, we compare the cross-sectional averages between the treatment and control groups in a regres- sion framework, and conclude that any difference between the two groups is due to viewing the soap opera’s financial messages in Scandal! Additionally, we control for mode of interview and use robust standard errors. Note that while the two intermediate computer-assisted telephone inter- view surveys provide insight into the immediate effects of our intervention, the outcomes measured in these surveys are not directly comparable with the outcomes in the final face-to-face survey since a richer set of questions were asked in the latter. Hence, our analysis focuses primarily on outcome measures obtained from the final follow-up survey, though we briefly discuss intermediate results as well. Finally, the discussion below also explains our qualitative results from the three focus groups. Note that the focus groups were held after we analyzed the quantitative data, and were designed and directed at understanding mechanisms and heterogeneity of some of our quantitative findings. 11.6.1 Financial knowledge The first outcome of interest is financial knowledge. While our intermediate computer-assisted telephone interview surveys did ask questions on financial knowledge, we could only ask very basic questions and do not detect any signifi- cant differences between treatment and control groups in these surveys. 11.  Harnessing emotional connections to improve financial decisions  ◾ 377 However, in the final face-to-face survey, we were able to ask more compre- hensive questions, and further were able to distinguish between general financial literacy and content-specific financial literacy. General financial literacy refers to questions that test financial concepts that were not at the forefront of the soap opera storyline, such as separating business and personal expenses, proba- bilities in lotteries (although gambling was highlighted, less attention was paid to probabilities), and writing down budgets. In contrast, content-specific financial literacy refers to questions that focused on knowledge of hire purchase agree- ments, which was the central contract through which the main soap character overspent and ended up with unaffordable high monthly payments. Table 11.4 presents our findings. Column 1 shows no significant differences between treatment and control groups on general financial literacy, but signifi- cantly higher scores on content specific financial literacy in column 2—compared to 42.5 percent in the control group, 47 percent of respondents in the treatment group answered these questions correctly, a 4.5 percentage point increase. Comparing across subgroups of gender, ex ante financial literacy, schooling, and ex ante income status, we do not see significant differences on these literacy measures. TABLE 11.4  Financial knowledge SCORE ON GENERAL SCORE ON CONTENT-SPECIFIC FINANCIAL LITERACY TEST FINANCIAL LITERACY TEST (1) (2) Invited to watch −0.010 0.045* Scandal! (0.019) (0.024) R-squared 0.005 0.014 Observations 963 963 Dependent variable 0.564 0.425 mean in control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The sample in both columns is composed of respondents present in the final survey who answered the respective questions. The outcome variable in column 1 is the average of three variables indicating correct answers to: (a) does not combine personal and business budgets, (b) writes down budget when spending excessively, and (c) thinks chances of winning the lottery are small. The outcome variable in column 2 is the score on knowledge about hire purchase products which averages correct answers to: (a) hire purchase is a convenient way to buy things when you do not have money, (b) hire purchase is helpful because it does not take very long, and (c) it is a good idea to take out longer hire purchase agreements. All regressions control for interview mode. 11.6.2 Debt management Next we turn to behavior. Clearly the most important financial behavior high- lighted in the soap storyline was borrowing and debt management. The first outcome we study is borrowing propensity—whether there was an increase or decrease in borrowing due to the storyline. Ex ante, we did not expect any partic- ular effect on this margin since the storyline did not discourage borrowing per se, 378  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES but rather only discouraged irresponsible borrowing. And the results confirm our prior assumption, as we do not observe any significant difference in borrowing likelihood between the treatment and control groups, as shown in table 11.5. TABLE 11.5  Debt management BORROWED LARGEST MISSED LOAN MONEY IN PAST 6 BORROWING IS FROM TOTAL LENGTH PAYMENTS IN MONTHS A FORMAL BANK OF LOAN PAST 6 MONTHS (1) (2) (3) (4) Invited to watch −0.001 0.088** 4.470* 0.035 Scandal! (0.030) (0.043) (2.588) (0.043) R-squared 0.003 0.014 0.010 0.015 Observations 963 315 269 304 Dependent variable mean 0.327 0.130 12.136 0.819 in control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The sample in column 1 is composed of respondents present in the final survey who answered the question, and the samples in columns 2–4 are composed of respondents present in the final survey who both borrwed money in the past six months and answered the respective question. All regressions control for interview mode. While borrowing is similar across the two groups, we identify a very strong treatment effect in the composition of borrowing. Column 2 of table 11.5 shows that, compared to 13 percent of respondents in the control group, 22 percent of treated respondents reported borrowing primarily from a formal bank. This is a very large effect size, equivalent to a 69 percent increase from the control group. Further, this result is directly relatable to the storyline, as the main soap character falls into financial distress precisely because she underestimates the financial charges associated with borrowing outside of formal financial institutions—in her case, hire purchase borrowing. Further, we detect significant differences in the length of loan, with the treat- ment group obtaining loans 4.5 months longer in term than the control group (column 3 of table 11.5). This difference in loan terms complements the finding of a shift toward formal borrowing as nonformal borrowing is typically of shorter duration. Nevertheless, we do not detect any differences in missed payments between the treatment and control groups. Finally, we examine reasons for borrowing in table 11.6 and find that the treatment group is more likely to borrow for investment purposes, which includes loans for vehicles, home improvement, or business development—all productive loans. Some interesting findings do emerge on heterogeneous margins. First, the effect of borrowing formally is driven entirely by men, while the coefficient for women is positive but not significant. Similarly, those with high ex ante financial 11.  Harnessing emotional connections to improve financial decisions  ◾ 379 TABLE 11.6  Reasons for borrowing TO COVER TO PAY OTHER FOR UNEXPECTED FINANCIAL FOR CONSUMPTION EXPENSES OBLIGATIONS INVESTMENT (1) (2) (3) (4) Invited to watch Scandal! −0.012 −0.019 0.026 0.075* (0.045) (0.054) (0.056) (0.041) R-squared 0.013 0.008 0.001 0.013 Observations 304 304 304 304 Dependent variable mean 0.196 0.312 0.355 0.116 in control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The sample in all columns is composed of respondents present in the final survey who declare having borrowed in the past six months. All regressions control for interview mode. literacy and high education also have larger (and significant) effect sizes. The latter results make sense, since more educated people would be more likely to be eligible and approach formal financial institutions for loans. The gender difference is interesting, and the qualitative analysis below sheds more light on this finding. Finally, though not statistically significant because of small sample size and large standard errors, we do see point estimates higher for women in borrowing for unexpected events, and higher for men in borrowing for consumption purposes. On the qualitative side, women generally came across as being incredibly reluctant to borrow at all, and only preferred to borrow as a last resort. This helps explain the muted impact on formal borrowing and borrowing only for unex- pected emergencies. One female response was, “If you do loans you will end up paying a lot of money back, they are very costly. It is better to budget and save for what you want.” Another female reacted to information from a colleague about her multiple loans, “Three accounts? That is damaging!” Men, on the other hand, were more willing to use borrowing, often to pay for consumption items and consumer electronics. What is particularly interesting is that men themselves recognized that women tend to be more responsible in their borrowing decisions. One male offered his insight, “Women are good with money most of the time. When they need something and they are short [R] 500, you should know that she really needs [R] 1,500 but she already has the [R] 1,000.” Another male concurred, “Women don’t like borrowing money, they have a conscience.” Perhaps as a consequence of these differences in borrowing preferences, women in our focus group tended to have more credibility among friends and family, and reported less resistance on the rare occasions they had to borrow in the past. One female summed up, “Men are so scared [to borrow from friends and family]. If my brother does not have money, he will phone me over and over again and then I must make a plan for him.” 380  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Finally, despite thinking differently about reasons to borrow, both male and female respondents agreed that formal borrowing was key to successful finan- cial management. An older female provided her opinion, “Sometimes people are negligent in using money, as human beings we have to manage every cent we get, we have to. If you are negligent you will end up with loan sharks and in trouble.” The men concurred with this opinion, with one male responding, “When I started working, I bought a TV and the salesman advised me to buy it over a six-month credit. And because hire purchase is expensive, you pay for years, and when you don’t pay, they come and take your stuff.” 11.6.3 Hire purchase and gambling The discussion of qualitative interviews above leads to our next set of results—use of hire purchase and gambling. The severe consequences of both these financial decisions were highlighted centrally in the soap storyline, with the main char- acter purchasing items on hire purchase and gambling and ending up in significant financial distress. Our quantitative results confirm the effectiveness of the soap opera’s message. As shown in table 11.7, respondents in the treatment groups were signifi- cantly less likely to have used hire purchase in the past six months (19 percent in the control group versus 15  percent in the treatment) and to have gambled in the past six months (31 percent control versus 26 percent treatment). These effect sizes are also large, representing an increase of 27 percentage points and 19 percentage points, respectively. When we examine heterogeneous effects, we see that these effects are stron- gest for respondents with low initial financial literacy and low formal schooling, both of which are consistent with having strong treatment effects given the low base level of financial knowledge and schooling. TABLE 11.7  Hire purchase and gambling SOMEONE IN HOUSEHOLD HAS USED SOMEONE IN HOUSEHOLD HAS HIRE PURCHASE IN PAST 6 MONTHS GAMBLED MONEY IN PAST 6 MONTHS (1) (2) Invited to watch −0.043* −0.052* Scandal! (0.024) (0.029) R-squared 0.004 0.003 Observations 963 963 Dependent variable mean in 0.188 0.307 control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The sample in all columns is composed of respondents present in the final survey who answered the respective questions. All regressions control for interview mode. 11.  Harnessing emotional connections to improve financial decisions  ◾ 381 Turning to the qualitative study, we find a strong disdain for gambling among both men and women in our focus groups. One woman expressed her frustration, “People are crying, it is addictive. There is a lady I work with, when the money hits her account and she is working the midnight shift with me, when we finish in the morning she doesn’t go home, she goes to Gold Reef City [a casino].” The men agreed, with one relating a personal experience with gambling, “I regretted that day, my whole salary went down the drain.” 11.6.4 Savings and well-being Next we investigate whether there were any treatment impacts on other finan- cial outcomes, such as savings, budgeting, and feeling of financial well-being (table  11.8). We should note that we did not expect any significant treatment effects on these margins ex ante since the soap opera storyline did not focus on these issues. However, there may have been indirect impacts on these measures as a result of improved financial behavior elsewhere. Our results do not find signif- icant differences except in feeling confident about solving financial problems. TABLE 11.8  Savings and well-being SAVED MONEY DOES NOT SPEND FEELS STRESSED FEELS CONFIDENT IN PAST 6 MORE THAN AND WORRIED SOLVING FINANCIAL MONTHS EARNINGS ABOUT FINANCES PROBLEMS (1) (2) (3) (4) Invited to watch −0.034 0.048 −0.041 0.053* Scandal! (0.031) (0.031) (0.029) (0.032) R-squared 0.002 0.005 0.003 0.003 N 963 963 963 963 Dependent variable mean 0.657 0.354 0.749 0.574 in control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The sample in all columns is composed of respondents present in the final survey who answered the respective questions. All regressions control for interview mode. The nonresult on savings is, in fact, encouraging, as it provides some evidence against survey bias—that individuals were not simply providing answers that pleased the surveyors—else we would have seen a positive treatment effect here also. 11.6.5 Seeking financial advice Finally, we turn to an important feature of the soap opera—a public call to action through the NDMA. As explained in previous sections, the NDMA was highlighted prominently in the storyline and its toll-free number was broadcast at the end 382  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES of each episode that featured an NDMA representative helping the main soap character. We examine quantitative outcomes in two separate ways. First, we obtained daily call volume data from the NDMA call centers and observe a huge spike in incoming calls immediately following the episode where the NDMA was intro- duced into the soap storyline. Figure 11.2 plots the daily call volume, with the orange vertical line representing the date of the soap feature. The call volume immediately jumped from an average of 120 calls per day to over 500, a more than 300 percent increase. Unfortunately, we cannot classify and match callers to our sample and hence can only present a before and after picture. However, our second intermediate survey, conducted just a short while later, included a question on consumer awareness of financial advice outlets. The regression analysis confirms a signif- icant increase in awareness of formal avenues for financial advice (such as the NDMA). Specifically, as shown in column 1 of table 11.9, compared to an average of 69 percent in the control group, nearly 80 percent of the respondents in the treatment group stated that they would seek financial advice from a formal source such as the NDMA if it was needed. Surprisingly, however, this effect dissipated in the long run. Columns 2–4 of table 11.9 show that, while the average proportion of respondents in the treat- ment and control groups who would seek formal financial advice remained rela- tively high (about 60  percent), there was no discernable statistical difference FIGURE 11.2  NDMA daily call frequency Total calls 500 400 300 200 100 0 01 Jan 2012 01 Feb 2012 01 Mar 2012 01 Apr 2012 01 May 2012 Note: This figure plots the daily call frequency by customers calling into the NDMA toll-free helpline in South Africa. The vertical orange line indicates when the helpline number was displayed in the televi- sion soap opera, Scandal! 11.  Harnessing emotional connections to improve financial decisions  ◾ 383 TABLE 11.9  Seeking financial advice IMMEDIATE IMPACT LONG-TERM IMPACT WILL SEEK WILL SEEK WILL SEEK ADVICE NOWHERE TO GO ADVICE FROM ADVICE FROM FROM SEMIFORMAL FOR FINANCIAL FORMAL SOURCE FORMAL SOURCE SOURCE ADVICE (1) (2) (3) (4) Invited to watch 0.102*** 0.019 −0.008 0.001 Scandal! (0.032) (0.031) (0.018) (0.026) R-squared 0.014  0.012 0.001 0.003 Observations 770 963 963 963 Dependent variable mean 0.689 0.601 0.087 0.202 in control group Note: Robust standard errors in parentheses. *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respec- tively. The sample in column 1 is composed of respondents present in the second follow-up round. The samples in columns 2–4 are composed of respondents present in the final survey who answered the respective questions. All regressions control for interview mode. between the two groups. Also, the treatment group was no less likely to believe that there was nowhere to go for financial advice. The qualitative analysis provides key insight into these differences in short- and long-term effects. In the focus groups, very few respondents recognized the NDMA logo, and no link was made between the Scandal! storyline and the NDMA until there was more probing into the related soap characters. Men in particular needed more prompting to recall the show, while women could remember more but primarily based on the fact that the debt advisor on the show was female rather than her affiliation with the NDMA. However, neither women nor men could accurately recall the role of the NDMA advisor in the storyline. One leading explanation for this lack of recall is that the NDMA advisor only appeared in the show for two or three episodes, was an external character, and never reappeared after her short stint. Hence, the audience did not get the oppor- tunity to establish or maintain a connection or following with this character. In contrast, all participants readily recalled the main soap character and aspects of the financial storyline delivered by her. Further, they attributed better recall to the fact that the main character was a mainstay in the soap opera, and was popular and likable. 11.7 CONCLUSION This chapter analyzes the economic impact of financial literacy messages on debt management delivered through a primetime television soap opera in South Africa. Using a symmetric randomized encouragement design methodology, the results show significant improvements in content-specific financial knowledge, strong 384  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES affinity toward borrowing formally, moving away from hire purchase deals, and gambling less—all messages that were conveyed in the soap storyline. A comple- mentary qualitative analysis confirms these findings and further highlights some key gender differences in the way men and women think about borrowing, with women generally using debt as a last resort while men are more willing to borrow for purchases. Interestingly, while a televised public call to action toward seeking financial advice through the NDMA led to a significant upsurge in calls immediately after the messages were shown on TV, this effect dissipates over time. The qualita- tive discussions suggest that lack of emotional connection with an external char- acter acting as the NDMA agent was a leading cause for the erosion of memory regarding the NDMA. Conversely, respondents easily and fondly recalled other aspects of the show delivered by the main soap character, who was part of the show prior to the financial literacy storyline and remained part of it afterwards. These findings suggest that emotional connections and familiarity with media personalities certainly play a role in motivating knowledge and behavior change among viewers, and that erosion of memory is an important concern when incorporating public service messages in mass media productions. Hence, incorporating messages through mainstay actors, delivering complementary interventions, or scheduling regular reappearances of external characters could potentially lead to greater retention and continuation of impact. Overall, unlike most other financial literacy evaluations, our analysis shows significant and favorable impacts on financial knowledge and behavior, which highlights the importance of delivery mechanisms in financial education. Indeed, entertainment media has the power to capture the attention of individuals unlike any other medium, and thereby provides policy makers with an effective and accessible vehicle to deliver carefully designed educational messages. REFERENCES Abdulla, Rasha A. 2004. “Entertainment-Education in the Middle East: Lessons from the Egyptian Oral Rehydration Therapy Campaign.” In Entertainment-Education and Social Change, edited by A. Singhal, M. Cody, E. Rogers, and M. Sabido, 301–20. Mahwah, NJ: Lawrence Erlbaum Associates. Agarwal, Sumit, and Brent Ambrose. 2007. “Does It Pay to Read Your Junk Mail? Evidence of the Effect of Advertising on Financial Decisions.” Working Paper WP-08-09, Federal Reserve Bank of Chicago, Chicago. Agarwal, Sumit, Gene Amromin, Itzhak Ben-David, Souphala Chomsisengphet, and Douglas D. Evanoff. 2010. “Learning to Cope: Voluntary Financial Education and Loan Performance during a Housing Crisis.” American Economic Review 100 (2): 495–50. Anschutz, Doeschka, Rutger Engels, Eni Becker, and Tatjana van Strien. 2008. “The Bold and the Beautiful. Influence of Body Size of Televised Media Models on Body Dissatisfaction and Actual Food Intake.” Appetite 51 (3): 530–37. 11.  Harnessing emotional connections to improve financial decisions  ◾ 385 Bagwell, Kyle. 2007. “The Economic Analysis of Advertising.” In Handbook of Industrial Organization, vol. 3, edited by Mark Armstrong and Rob Porter, 1701–1844. Amsterdam: Elsevier. Bertrand, Marianne, Dean Karlan, Sendhil Mullainathan, Eldar Shafir, and Jonathan Zinman. 2010. “What’s Advertising Content Worth? Evidence from a Consumer Credit Marketing Field Experiment.” Quarterly Journal of Economics 125 (1): 263–306. Brown, William J., and Michael Cody. 1991. “Effects of a Prosocial Television Soap Opera in Promoting Women’s Status.” Human Communication Research 18 (1): 114–44. Carpena, Fenella, Shawn Cole, Jeremy Shapiro, and Bilal Zia. 2011. ”Unpacking the Causal Chain of Financial Literacy.” Policy Research Working Paper 5798, World Bank, Washington, DC. Chong, Alberto, and Eliana La Ferrara. 2009. “Television and Divorce: Evidence from Brazilian Novelas.” Working Paper 651, Inter-American Development Bank, Washington, DC. Cole, Shawn A., Thomas Sampson, and Bilal Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Elliehausen, Gregory, E. Christopher Lundquist, and Michael E. Staten. 2007. “The Impact of Credit Counseling on Subsequent Borrower Behavior.” Journal of Consumer Affairs 41 (1): 1–28. Evans, Franklin. 1963. “Selling as a Dyadic Relationship.” American Behavioral Scientist 6: 76–79. Finmark Trust. 2012. FinScope South Africa 2012: Survey Highlights. http://www.finmark. org.za/wp-content/uploads/pubs/FinScope_SA_Booklet_2012.pdf. Hartarska, Valentina, and Claudio Gonzalez-Vega. 2005. “Credit Counseling and Mortgage Termination by Low-Income Households.” Journal of Real Estate Finance and Economics 30 (3): 227–43. Hirad, Abdighani, and Peter M. Zorn. 2001. “A Little Knowledge Is a Good Thing: Empirical Evidence of the Effectiveness of Pre-Purchase Homeownership Counseling.” http:// www.jchs.harvard.edu/sites/jchs.harvard.edu/files/liho01-4.pdf. Jensen, Robert, and Emily Oster. 2009. “The Power of TV: Cable Television and Women’s Status in India.“ Quarterly Journal of Economics 124 (3): 1057–94. Kenny, Charles. 2009. “Revolution in a Box.” Foreign Policy 175: 68–75. Kennedy, May, Ann O’Leary, Vicky Beck, Katrina Pollard, and Penny Simpson. 2004. “Increases in Calls to the CDC National STD and AIDS Hotline Following AIDS-Related Episodes in a Soap Opera.” Journal of Communication 54 (2): 287–301. La Ferrara, Eliana, Alberto Chong, and Suzanne Duryea. 2008. “Soap Operas and Fertility: Evidence from Brazil.” Working Paper 663, Inter-American Development Bank, Washington, DC. http://www.iadb.org/res/publications/pubfiles/pubWP-633.pdf. Landry, Craig, Andreas Lange, John List, Michael Price, and Nicholas Rupp. 2006. “Towards an Understanding of the Economics of Charity: Evidence from a Field Experiment.” Quarterly Journal of Economics 121: 747–82. Lusardi, Annamaria, and Olivia S. Mitchell. 2007. “Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education.” Business Economics 42 (1): 35–44. Lusardi, Annamaria, Olivia S. Mitchell, and Vilsa Curto. 2010. “Financial Literacy among the Young.” Journal of Consumer Affairs 44 (2): 358–80. Lusardi, Annamaria, and Peter Tufano. 2009. “Debt Literacy, Financial Experiences, and Overindebtedness.” NBER Working Paper 14808, National Bureau of Economic Research, Cambridge, MA. Mandel, Ruth. 1998. “Structural Adjustment and Soap Opera: A Case Study of a Development Project in Central Asia.” Central Asian Survey 17 (4): 629–38. 386  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES NCR (National Credit Regulator). 2012. Credit Bureau Monitor, June. http://www.ncr.org.za/ press_release/CBM.pdf. Paluck, Elizabeth, and Donald Green. 2009. “Deference, Dissent, and Dispute Resolution: An Experimental Intervention Using Mass Media to Change Norms and Behavior in Rwanda.” American Political Science Review 103 (4). Roberts, Benjamin, and Jarè Struwig. 2011. “Financial Literacy in South Africa: Results of an OECD/INFE Pilot Study.” Financial Services Board. https://www.fsb.co.za/ Departments/consumerEducation/Documents/FSB%20FINANCIAL%20LITERACY%20 REPORT%2009062011%20prt1.pdf. Rogers, Everett, Peter Vaughan, Ramadhan M.  A. Swalehe, Nagesh Rao, Peer Svenkerud, and Suruchi Sood. 1999. “Effects of an Entertainment-Education Radio Soap Opera on Family Planning Behaviour in Tanzania.” Studies in Family Planning 30 (3): 193–211. Rogers, Everett, and Peter Vaughan. 2000. ”A Staged Model of Communication Effects: Evidence from an Entertainment-Education Radio Soap Opera in Tanzania.” Journal of Health Communication 5: 203–27. Russo, Edward, Kurt Carlson, and Margaret Meloy. 2006. “Choosing an Inferior Alternative.” Psychological Science 17 (10): 899–904. Singhal, Arvind, Everett M. Rogers, and William J. Brown. 1993. “Harnessing the Potential of Entertainment-Education Telenovelas.” International Communication Gazette 51 (1): 1–18. South African Reserve Bank. 2012. Quarterly Bulletin 265 (September). Stigler, George. 1987. The Theory of Price, 4th ed. New York: MacMillan. Tufano, Peter, Timothy Flacke, and Nicholas W. Maynard. 2010. “Better Financial Decision Making among Low-Income and Minority Groups.” Working Paper WR-795-SSA, Financial Literacy Center. http://www.rand.org/content/dam/rand/pubs/working_ papers/2010/RAND_WR795.pdf. Vaughan, Peter, Alleyne Regis, and Edwin St. Catherine. 2000. “Effects of an Entertainment- Education Radio Soap Opera on Family Planning and HIV Prevention in St. Lucia.” International Family Planning Perspectives 26 (4): 148–57. Verner, Dorte, and Ana Cardoso. 2007. “Youth Risk Taking Behavior in Brazil: Drug Use and Teenage Pregnancies.” IZA Development Paper Series No. 3030, Institute for the Study of Labor, Bonn. http://ftp.iza.org/dp3030.pdf. 11.  Harnessing emotional connections to improve financial decisions  ◾ 387 ANNEX: SCANDAL! STORYLINE: MALETSATSI’S DEBT Lead cast: Maletsatsi, Eddie Support cast: Daniel, Palesa, Gloria (Maletsatsi’s neighbor), Society Members (Maletsatsi’s new neighbors), Constance, NDMA Official, Financial Advisor Universal truth: If you can’t afford it, don’t buy it. Central question: Will Maletsatsi take the necessary steps to get out of debt and successfully manage her finances in the future? Premise: Maletsatsi, through good deeds and wanting to create a beautiful home for her husband and family, finds herself caught in a spiraling vicious cycle of debt. The objective of the storyline is to positively affect targeted behaviors around financial literacy and management by increasing the knowledge and understanding of low-income South Africans regarding personal finan- cial management. The storyline will improve financial capability attitudes and behaviors related to debt management and help-seeking through the NDMA. The story will make it clear that anyone can fall prey to debt and that it can occur through actions that have the best and frequently unselfish intentions. Contact details for the NDMA will appear on screen at the end of every episode in which the story plays out. Synopsis ◾◾ Maletsatsi has started to socialize with the women in her new neighbor- hood of Orlando, Soweto. She’s becoming comfortable enough to start mobilizing. ◾◾ A neighbor can’t afford to bury a family member. Maletsatsi plays an active role in collecting donations from the neighborhood to help with the burial, but the money is not enough. ◾◾ Gloria (Maletsatsi’s new neighbor) tries to talk Maletsatsi into accompa- nying her to a casino. Maletsatsi is initially completely opposed to the idea but, when she realizes that she may be able to win money to put into the burial fund, she decides to try it, although she refuses to go to the casino herself. She gives money to Gloria and tells her what number to play on her behalf at the roulette table. 388  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES ◾◾ Maletsatsi’s number is lucky for her, and she wins some money which she puts into the burial fund. But it is still not enough. ◾◾ Encouraged by her initial success, she gets Gloria to play for her again… and again. Sometimes she wins, sometimes she loses, but on the whole, she stays ahead. Of course, she doesn’t tell Eddie what she is doing and convinces herself that, because she isn’t actually going to the casino herself, what she is doing isn’t wrong—particularly as any money she wins will go to a good cause. ◾◾ Maletsatsi doesn’t manage to raise quite enough money for the funeral. Daniel Nyathi comes to the rescue and magnanimously foots the rest of the bill, much to Eddie’s annoyance. ◾◾ The experience with the burial fund leaves Maletsatsi surprised that the neighborhood doesn’t have a society (stokvel ). ◾◾ Maletsatsi sees an opportunity to start a society with the neighbors. She runs her idea by Eddie, who thinks it is a good idea. ◾◾ Maletsatsi pitches her idea to Gloria. Gloria tells Maletsatsi that she’s tried to get the women to start a society, but there has been a lack of commitment. ◾◾ Maletsatsi speaks to some of the neighbors about her vision. The people to whom she speaks are keen to get involved in the society. ◾◾ Maletsatsi calls a meeting, and the interested neighbors respond well. Maletsatsi pitches her idea for the society, and most of them love it. They admit that they tried this before but it didn’t work due to a lack of commit- ment. (What becomes apparent is that they lacked a strong leader to inspire and keep them motivated, which Maletsatsi is good at.) ◾◾ Gloria tells Maletsatsi that some of the women gossip about her behind her back. They say that her lounge, dining room, and kitchen (which visitors see) are well furnished, but the bedrooms (which most visitors don’t see) don’t match up. (Gloria checked out the house after using the bathroom during one of her visits and started the gossip herself. Gloria is jealous that Maletsatsi seems to be succeeding with starting a society while she, Gloria, failed to get the initiative off the ground.) Maletsatsi pretends to laugh this off. ◾◾ In actual fact, Maletsatsi is troubled by what she heard from Gloria. She is worried that people will think that she doesn’t care for her family properly; that she only makes the public areas of her house nice and neglects the areas where her family sleeps. Eddie advises her not to pay attention to town gossip, but Maletsatsi is unappeased. Her commitment to being a good wife, mother, and grandmother has been brought into question. ◾◾ To make herself feel better (and to prove to herself and her neighbors that she cares for her family), Maletsatsi buys new curtains for all the 11.  Harnessing emotional connections to improve financial decisions  ◾ 389 bedrooms—and, while she’s at it, new lounge and passage carpets and a new tea set with which to impress the society ladies. ◾◾ Eddie is surprised to see the new things. Maletsatsi sheepishly admits that she wanted her family’s bedrooms—and the rest of the house—to be nice for her family. Eddie is pleased that the house is looking so good. (Note: Some younger males in the focus groups felt that Eddie is too under- standing and supportive of Maletsatsi, while the older males felt that Eddie is doing the right thing. We have elected to go with Eddie being supportive. This also avoids the clichéd situation of the spendthrift wife who squan- ders money on luxuries without her husband’s knowledge. By making Eddie pleased with Maletsatsi’s purchases, he encourages her to continue making a nice home—which makes him partly responsible for the negative financial situation in which they find themselves.) ◾◾ The society kicks off. The women give it a name, and Maletsatsi suggests that they draft a constitution by which they will all abide. The society agrees to meet fortnightly. (Most of the first meetings will take place at the Khumalo home.) ◾◾ The first contributions come in. It is agreed that Maletsatsi will be treasurer of the society and will take care of the money while they are in the process of opening a bank account. ◾◾ Constance comes to see Maletsatsi. She and Benjamin are having some renovations done to their house in a few weeks’ time and she wants to know if they can stay in the Khumalos’ back rooms for a week or so. Maletsatsi and Eddie agree that Palesa’s in-laws can come and stay. ◾◾ Maletsatsi tells Gloria about Constance and Benjamin coming to stay in the back rooms and asks whether she has a spare bed or even some mattresses that she can borrow. Gloria is horrified. She knows Constance and doesn’t think that Maletsatsi is doing a good thing by allowing her daughter’s in-laws to sleep on borrowed, makeshift furniture. Maletsatsi really needs to furnish the back rooms properly or she will lose face. ◾◾ Maletsatsi tells Eddie that she would like to furnish the back rooms for Constance and Benjamin’s stay. Eddie agrees, but says that Maletsatsi must only buy the necessities—a bed, sheets and blankets. The rest can wait. ◾◾ Maletsatsi comes back with brochures from the furniture shops. She spends time looking at them and is particularly taken with a bedroom suite made of expensive wood. ◾◾ Maletsatsi purchases the bedroom suite but, when it arrives, a small table and chairs for the back rooms, plus a new refrigerator and flat screen TV for the house, arrive along with it. Eddie is concerned. Maletsatsi explains that when she went to look at furniture, the salesperson checked her credit status and found that she could afford to open an account. Eddie asks how 390  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES much the stuff came to. Maletsatsi says R 30,000. She has opted to pay in installments over 24 months. ◾◾ Eddie goes through the credit agreement and breaks down the various costs involved, including the interest. It turns out that by the end of that period, she will have paid almost double that amount, due to the interest and other costs. (The salesperson sweet-talked her into this and told her that there was no deposit needed and no payments due for the first two months.) ◾◾ Eddie thinks Maletsatsi should send the things back. Maletsatsi pleads with him. Their old refrigerator from the flat is too small for the space in the new kitchen, the TV will be great for Eddie to watch sports on, and their family will have lovely back rooms to stay in. And anyway, she will be paying for it. Eddie asks with what money. ◾◾ Maletsatsi tells a little white lie. She has been saving some of her salary every month (it’s actually her gambling winnings), and Daniel is giving her a small increase that she can put toward the monthly installments. Also, she has started the society and, when it is her month to get the money, she will be able to pay off the full amount. ◾◾ Eddie softens. He likes the things, and Maletsatsi’s arguments make good sense. He is happy that his new house is looking so good. But he tells her there is a condition attached. She must pay the debt off quicker to avoid paying so much interest. Maletsatsi undertakes to go to the furniture store and shorten the repayment period. ◾◾ Maletsatsi does this, but the monthly installments will go up substantially and the two-month grace period falls away. She will have to start paying immediately. She tells Eddie that she has dealt with his conditions, but in reality she hasn’t. The new monthly installments are more than she can afford, and she doesn’t have enough money for the first installment which is now due. ◾◾ Gloria is impressed with Maletsatsi’s purchases and that she has “managed her husband so well.” Abel (Gloria’s husband) would never allow her (Gloria) to get away with this. Maletsatsi starts to wonder if Gloria’s advice is what she really needs in her life. But she tells Constance that the back rooms have been tastefully and comfortably furnished in preparation for their stay. Constance is pleased and touched that Maletsatsi has gone to so much trouble for them. ◾◾ Constance arrives to stay (Benjamin has had to go away on Evershine busi- ness) and makes herself comfortable in the back rooms. ◾◾ Maletsatsi still has the cash from the society, as the bank account hasn’t been opened. She borrows some of it (quite a lot, actually) to make up the shortfall on her furniture payment, telling herself that she will put it back soon. 11.  Harnessing emotional connections to improve financial decisions  ◾ 391 ◾◾ As per the constitution of the society, if a member is in dire need of money before it is their turn, the society can vote to help them out. One of the member’s husbands loses his job and, even though it is the first month of existence, the society votes to donate the money to the member. Trapped, Maletsatsi says she will get the money, which she claims she has locked in the community center safe until the bank account has been opened. ◾◾ Maletsatsi doesn’t know where to turn. But then she remembers the community center. Daniel handed the running of the center’s money over to her when he went away some months ago and decided to leave her in charge on his return. Maletsatsi draws money from the community center’s account to pay back the society’s money. She draws a little extra to help her with next month’s installment for the furniture. ◾◾ Maletsatsi pays the society money over to the needy woman. ◾◾ Daniel asks Maletsatsi for the community center’s books, etc. He wants to take a look at them. Maletsatsi stalls, buying time by telling Daniel that she has taken everything home. Daniel is fine with that, and says Maletsatsi can bring the stuff in at her leisure. ◾◾ Panicked, Maletsatsi can think of only one way of getting the money back. She gives Gloria some more money to gamble with. She loses. ◾◾ Desperate to be able to pay back the community center money, Maletsatsi goes to the automated teller machine (ATM) to see what she has left in her account. The ATM offers her the chance of a loan. ◾◾ She takes a deep breath and requests the loan, which is granted. ◾◾ Maletsatsi pays the money she took from the community center back into the account and hands the books over to Daniel. ◾◾ Maletsatsi manages to cover the remainder of the month’s expenses with the rest of her loan, but when the new month starts, she realizes that she is in debt to the bank, has to pay the installment on the furniture, and has lost all of her savings to gambling. ◾◾ In an unguarded moment, Maletsatsi lets slip to Constance that she has been gambling and borrowed money from the stokvel to help with her debts. Judgmental and horrified, Constance is expressing her shock when Gloria walks in and overhears everything. She immediately insists that Maletsatsi has to confess to the society ladies. Maletsatsi begs Gloria to give her some time. ◾◾ Daniel notices something odd in the community center’s books. An amount was withdrawn and then repaid a few days later. He questions Maletsatsi about this. Maletsatsi breaks down and confesses her financial woes to Daniel. Daniel is taken aback. This is the last thing he expected of Maletsatsi, of all people. The irony of the situation (the woman who saved 392  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES him from sin actually stole money from him) tickles Daniel, but he cares too much for Maletsatsi to allow her to see his amusement. ◾◾ Palesa arrives back from her POWA tour unexpectedly and comes to the community center to surprise her mother. She is surprised to find Maletsatsi weeping in Daniel’s arms. ◾◾ Maletsatsi is forced to come clean to her daughter about what she has done. Palesa wants her to tell Eddie straight away (she is horrified that Maletsatsi has confided in Daniel but not her husband), but Maletsatsi refuses. There has to be a way of fixing this before Eddie finds out. She is also terrified of the society ladies hearing that she has landed herself in financial trouble—she is the treasurer of the society! ◾◾ Palesa advises her mother to go back to the furniture store and get them to make a plan to allow her to pay it off at a lower installment rate. Maletsatsi does so, and the store agrees to extend the loan period again. But the interest she will pay will go up substantially. ◾◾ Not wanting to worry Palesa, Maletsatsi tells her that she has made a plan with the store and everything is going to be all right. Palesa still wants her mother to talk to Eddie, but Maletsatsi insists that she is handling the problem and that she is fine. ◾◾ But she isn’t. Terrified that she will never be able to pay back the loan and the interest on the furniture installments, Maletsatsi starts to ignore other bills, leaving them unopened—and unpaid. ◾◾ At the next society meeting, the ladies all pay in their contributions. Maletsatsi is about to take the money, but “supportive” Gloria says she’ll take some of the responsibility off Maletsatsi’s shoulders of having to take care of the cash until the bank account is opened. She’ll take care of the money this time. For a moment, Maletsatsi fears that Gloria is going to expose her, but she doesn’t. ◾◾ Maletsatsi becomes moody. She makes mistakes at work and is irritable with her family at home. ◾◾ Eddie gets a letter saying that a household expense hasn’t been paid. He questions Maletsatsi, who almost faints but covers and tells him she must have forgotten to pay it. She will sort it out today. Eddie thinks nothing of it. ◾◾ Maletsatsi breaks down at work again. Daniel walks in on her crying, sits her down, and asks her how much she owes, including the outstanding amount on the furniture. She tells him. It is a lot of money to her, but to Daniel (with his recent R2m windfall and the property business taking off) it is not much. He offers to pay off her debts for her. She doesn’t even have to pay him back. She must consider it his gift to her. 11.  Harnessing emotional connections to improve financial decisions  ◾ 393 ◾◾ Daniel also makes Maletsatsi realizes that she will have to come clean with Eddie. (Daniel is, of course, thrilled to have the opportunity to be able to help Maletsatsi out and to rub Eddie’s nose in it.) Maletsatsi tells Palesa that she is going to confess to her husband. ◾◾ Maletsatsi and Palesa arrive home, and Palesa tells Eddie that Maletsatsi has something to talk to him about. She leaves them alone. ◾◾ Slowly, painfully, and with many tears, Maletsatsi tells Eddie what has happened. He is shocked, but allows her to complete her story without interrupting too much. Only when she gets to the part where Daniel has offered to pay off her debt does Eddie erupt. There is no way that his wife is touching that man’s money! Maletsatsi says she understands Eddie’s concerns, but what other way out is there? Eddie says he doesn’t know. He needs some time to think about it, but he will come up with another plan. He leaves a distraught Maletsatsi alone. ◾◾ The next day Eddie is calm and practical. He tells Maletsatsi that he was very angry with her, but that is now over. What’s done is done and they have to move forward to fix things. He spent a large part of the night researching debt, debt counseling, and finding out about organizations that can help. ◾◾ Firstly, they are going to approach the NDMA to get them to help negotiate a reasonable repayment option with the furniture store for some of the items. (Note: It is clear that most people have never heard of the NDMA. So, at this point, Eddie will explain in simple and clear terms what the NDMA is and what it does. This will be expanded on in later scenes. These scenes will be written with input and final approval from the NDMA.) ◾◾ Secondly, Eddie continues, they are going to return some of the items to the store. Thirdly, they are going to go to her bank and get them to work out a way for her to pay off her loan. And they are going to take a financial management course to learn how to manage their finances properly. All of this they will do together. But there are two things that Maletsatsi is going to have to do on her own. One, she is going to make sure that she never gambles again. And two, she is going to come clean with her society and step down as treasurer until such time as she and they feel that she is financially responsible enough to occupy such a position. ◾◾ Maletsatsi is horrified at the thought of having to tell her neighbors what she has done but, as a Christian, she agrees that she has to do it. Eddie concurs. ◾◾ Eddie softens. He tells Maletsatsi that he only wants her to do all these things for her own good. He knows that she only got into debt because she wants the best for her family and because she wants to help others. He reminds her that she has done this before (the leather couch debacle). She is a good woman. But even good people make mistakes. There is no shame 394  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES in that. The only shame is if lessons are not learned from those mistakes. And this time Maletsatsi really has to take responsibility and learn her lesson. As an alcoholic, he knows that better than anyone. Maletsatsi agrees. Eddie hugs her and tells her he loves her. ◾◾ Maletsatsi thanks Daniel for his kind offer, but tells him that she and Eddie will deal with the debt on their own. Daniel is not surprised. Eddie Khumalo would never take anything from him. But he could still give Maletsatsi some money without Eddie knowing. Maletsatsi holds up her hand. She is never going to hide anything from Eddie again. Ever. Daniel is disappointed but understands. He wishes he could find a woman who is as loyal as Maletsatsi. ◾◾ We see a scene where the NDMA successfully negotiates a reasonable repayment option for Maletsatsi with the furniture store. This scene will show that approaching the NDMA is nonthreatening and easy to do. It will also make it clear that the NDMA does not charge for its services (one of the misconceptions that was picked up in the focus groups). Maletsatsi and Eddie will have to develop an income and expenditure statement (these will be carefully explained) to determine the amount available to offer to the credit provider. They also need to plan how they will prevent previous mistakes and understand the consequences of not sticking to the new restructured agreement. They decide to do this for the refrigerator and TV and to return the furniture that they bought for the back rooms. (Note: this scene will be written with input and approval from the NDMA.) ◾◾ The furniture from the back rooms is returned to the store. Maletsatsi is forced to borrow a camping bed from Gloria, who is horrified that Maletsatsi is doing this. With great difficulty, Maletsatsi has to swallow her pride and explain to Constance why the furniture that she bought specially for Constance’s stay will no longer be there. As usual, Constance can’t help being a bit judgmental, but Eddie, in full-on support of Maletsatsi, tells Constance that if she isn’t happy with what they are able to provide for her, she is welcome to check into a hotel. Constance backs down. ◾◾ We see Maletsatsi and Eddie at a financial management course, learning the rules of sound financial management and planning. (Note: During this scene, terms that the focus groups had difficulty understanding— converting to cheaper debt, downgrading, and productive borrowing—will be simply and clearly defined and explained.) ◾◾ Maletsatsi plucks up the courage to admit to the society ladies that she took money from them. They are all shocked, but some are more supportive and understanding than others. The overall feeling, though, is that Maletsatsi will have to step down as treasurer. She will have to work hard to regain 11.  Harnessing emotional connections to improve financial decisions  ◾ 395 the members’ trust. Maletsatsi accepts this, but the shame is hard for her to bear. ◾◾ Maletsatsi organizes a talk at the community center where a financial planning expert comes and gives simple yet sound advice about finan- cial management. The whole Khumalo family attends, as well as people from the community. They are again told about downgrading, converting to cheaper debt, budgeting, and saving (all of which terms will be clearly defined and explained again). Borrowing can be a potential risk, but produc- tive borrowing and saving can help consumers take control of their debt. The expert also explains the role of the NDMA and the differences between that organization and the National Credit Regulator, which is tasked with education, research, and registration of credit providers and debt coun- selors; complaints related to debt counselor behavior; and ensuring the enforcement of the Credit Act. The difference in the roles of the NDMA and debt counselors is also explained. (Note: Most people have heard of the National Credit Regulator but not of the NDMA, so we are taking the oppor- tunity to define and differentiate clearly the roles and functions of the two organizations. This scene will be written with input and approval from the NDMA.) Maletsatsi shares her experience with the assembled group, explaining that she is going to take control of her situation and implement some of these measures in her own life. ◾◾ Maletsatsi uses the money that she is no longer paying off on the furniture to service her high-interest debt (her loan and other purchases), thereby lowering the interest she will be paying over the long term. ◾◾ Maletsatsi gets Daniel to help her structure a simple savings plan. She is going to put a part of her salary into a special bank account from which she cannot draw and which will give her good interest on her savings. CHAPTER 12 T he impact of financial education on financial service use Evidence from a financial diaries study in Uganda GUY STUART ABSTRACT This chapter presents findings on a study of a financial education program in the Luweero district of Uganda, about 150 kilometers north of the capital, Kampala. The chapter focuses on how financial diaries generate data on the financial behavior of low-income individuals that cannot be readily gener- ated from more traditional survey research or focus group discussions. In combination with in-depth interviews, it provides considerable detail on the knowledge, skills, and attitudes and behaviors of low-income people that can inform the debate about how best to change knowledge, skills, and atti- tudes and behavior through financial education and other interventions. In addition, the chapter shows how researchers can track change in indicators of financial capability over time using diaries and in-depth interview data. 12.1 INTRODUCTION This chapter demonstrates the circumstances under which financial diaries can be used to evaluate the impact of a financial education program. Such diaries generate data on the financial behavior of low-income individuals that cannot be readily generated from more traditional survey research or focus group discussions. The chapter addresses the following questions: The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 397 398  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 1. What can diaries, in combination with in-depth interviews, tell us about the financial capabilities of low-income people that we might not know otherwise? 2. How can change in indicators of financial capability be tracked through diaries, in combination with in-depth interviews, over time? 3. Under what circumstances is it appropriate to use financial diaries to eval- uate the impact of a financial education program? Answers to these questions have important implications for both research and practice. The study exploits the fact that the diaries produce detailed behav- ioral data—they track the transactions respondents perform, the relationship between the respondent and the other party in the transaction, and where the transaction was performed. As we will elaborate in the following section, one theory of change in the financial capabilities literature is that education changes knowledge, skills, and attitudes (KSAs), which in turn result in a change in behavior, out of which improvements in economic well-being emerge. We use the interview data to identify the sample’s KSAs and changes in KSAs, and relate this informa- tion to their economic behavior and changes in behavior. In this way, we were able to examine the relationship between KSAs and behavior. In addition, though the diaries data from this project only concern low-income rural Ugandans in the Luweero district (about 150 kilometers north of Kampala), they will add to the growing evidence from other diaries studies regarding the economic behavior of low-income individuals. As noted above, such evidence is critical for understanding how education, through changes in KSAs, might affect behavior. But it is also critical for understanding how alternative strategies aimed at directly affecting behavior might work within low-income populations. Advances in behavioral economics suggest that financial capabilities might be enhanced through behavioral prompts (“nudges”) that bypass changes in KSAs (Holzmann 2010). Detailed behavioral data from the diaries can be used to inform the work of researchers who are interested in designing and evaluating nudges aimed at changing the behavior of low-income individuals. For example, evidence from other diaries studies (Collins et al. 2009; Stuart and Cohen 2011; Stuart, Ferguson, and Cohen 2011) suggests that low-income individuals handle a lot of cash and have a wide variety of cash flows, depending on their livelihoods and other factors. A behavioral approach to improving finan- cial capabilities might segment populations by cash flow structure, and deliver nudges timed to coincide with the flow of money through the hands of the people in targeted segments to remind them to save their money. A researcher inter- ested in testing such a strategy might use a traditional survey to gather informa- tion on how much cash a person handles each week and what pattern their cash flow follows, to segment the population and apply the different nudges. Informa- tion from the diaries should be useful in informing survey questions designed to capture information on cash flow. 12.  The impact of financial education on financial service use  ◾ 399 In addition to these substantive contributions to the understanding of indi- viduals’ economic behavior, the chapter evaluates the role financial diaries, in combination with qualitative interviews, can play in the evaluation of the impact of financial education programs. Though this project’s final sample composition makes it difficult to attribute changes in KSAs and behavior to the Habitat for Humanity Uganda (HFHU) intervention, it nevertheless provides important guid- ance to researchers wishing to use the diaries for an impact evaluation.1 It docu- ments the data-gathering process and provides examples of how diaries can be used to measure changes in behavior. The framework for the analysis is informed by a number of different sources: the literature on financial literacy and financial capabilities in developing coun- tries, data from participatory exercises conducted in Uganda in which partici- pants defined financial capabilities in their own terms, and findings from previous financial diaries regarding the behavior of low-income individuals. The remainder of this chapter is organized as follows: section 12.2 provides an overview of the project, describing the intervention, conceptual framework, research design, research sites and sample characteristics, and research meth- odology. Section 12.3 presents the findings of the data analysis in full, and section 12.4 presents a conclusion that summarizes the major points covered in the chapter. A more detailed version of this chapter, including an analysis of respon- dents’ cash flow management and planning for the future, can be found in Stuart (2012). 12.2 PROJECT OVERVIEW 12.2.1 Habitat for Humanity intervention HFHU’s eight-week financial education program adapted a curriculum developed by Microfinance Opportunities which includes modules on savings, budgeting, bank services, debt management, and financial negotiations. The program is designed to give participants the basic knowledge and tools they need to manage their money more wisely and to encourage positive changes in their financial behavior. By the end of the training, participants are expected to have gained 1  Our sampling yielded a self-selected group of individuals almost all of whom attended an HFHU financial education program, and who are overwhelmingly members of either a savings group or a savings and credit cooperative organization (SACCO). It has also yielded a comparison group, randomly selected from similar communities to those from which the treatment group was drawn, who show little use of SACCOs and savings groups, and some use of banks. More information on the original purpose of the project and the sampling problems we faced can be found in Stuart (2012), annex 1. 400  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES financial knowledge and skills leading them to positive changes in financial atti- tudes and behavior, as described in the financial education model of change. The learning objectives for each session were used in formulating outcomes indicators for assessing the impact of the financial education program. We obtained infor- mation on the schedule, rosters, and attendance of the sessions from HFHU in order to cross-check participants’ self-reported events data and to help correlate session attendance with outcome measures. HFHU conducted seven training sessions from July through September 2011 for each of four groups of financial education clients in two locations: Ssenyomo village and Kiwayirembe village. In the former, HFHU delivered financial education training to the Basoko Kwavuula Group; in the latter village, it delivered financial education training to the Bakuseka Magende Group, Suubi Farmers’ Group, and Ssenyomo F.F. Association Group. Two sessions were required to cover each of the first two chapters in the curriculum on budgeting and saving. The remaining three sessions covered financial negotiation, debt management, and bank services (table 12.1). Though HFHU reported a 95 percent attendance rate for participants in the diaries study, self-reported attendance suggests much lower rates of participation, from a low of 23 percent at the second budgeting session to a high of 70 percent at the second savings session. More information on the HFHU curriculum and attendance rates can be found in Stuart (2012), annex 3. TABLE 12.1  Financial education training syllabus SESSION TOPICS COVERED 1. Budgeting How to make a budget; benefits of a budget 2. Budgeting Setting financial goals 3. Saving Purpose for saving; savings services 4. Saving Benefits of saving 5. Loan management & negotiation Preparing for negotiation and benefits of negotiation 6. Debt management Reasons for taking a loan; loan terms and conditions 7. Bank services Overview of different service providers; HFHU services In total, HFHU recorded 110 individuals attending some or all of its classes, including 41 in the diaries study. All but one pair of diaries respondents were from different households, meaning that 40 diaries households were exposed to financial education. 12.2.2 Indicators of financial capability Stuart (2012) identifies three major elements of financial capability: day-to-day cash flow management, planning ahead, and financial service use. Each of these 12.  The impact of financial education on financial service use  ◾ 401 three has a cognitive (KSAs and self-confidence), behavioral, and environmental dimension. For the purposes of this chapter, our focus is on financial service use and indicators of change in use. We are especially interested in cognitive and behavioral indicators, because the prevailing theory of change within the financial education literature is that financial education has a cognitive impact that then translates into behavioral change (Stuart 2012). Table 12.2 outlines the elements and associated indicators used to describe what it means to be financially capable. TABLE 12.2  Financial capability concepts and indicators ELEMENT INDICATOR How people save Frequency # of savings transactions per week Size Amount per week; average size per transaction Where/with whom (this will yield information on # of savings transactions by entity types of individuals and organizations people use) Return Interest rate paid to saver by entity How people manage debt, to others and from others Frequency of loans # of loans per week Size of loans Average size of loans Where/from/to whom (this will yield information on # of loans by entity types of individuals and organizations people use) Interest rate on loans Interest paid by entity How people make financial service choices Interest earned/paid Articulated in in-depth interviews Do they know how to calculate this? Able to answer question asking them to calculate simple interest rate yield Convenience Proximate Articulated in in-depth interviews; average distance traveled by financial service users Flexible access—hours and number of transactions Articulated in in-depth interviews; size and allowed per period frequency of transactions per week per entity Something else Articulated in in-depth interviews As we will explain more fully later in this chapter, the financial diaries produce transaction data that can be used to measure the indicators associated with each element, except in cases where it is indicated that the source is in-depth inter- views. Furthermore, the interviews provide explanations for some of the behav- iors we observe in the transaction data generated by the diaries and are invaluable in helping construct the right measures of indicators of financial capability. 402  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 12.2.3 Research design The original research design of this project was a quasi-experimental, differ- ence-in-difference evaluation of the impact of HFHU’s financial education program. It was set up to compare changes from the pre-intervention period to the post-intervention period in the KSAs and behavior of a treatment group, in comparison to pre-/post-intervention changes in the KSAs and behavior of a similar comparison group in a different community that had not been exposed to the education program. The research design of the Uganda Financial Education Impact Project divides the study into two phases in order to capture both the short-term and long-term impacts of the financial education intervention. Specifically, the design calls for data to be collected immediately after the intervention ends and again eight months after the fact. This approach allows for there to be an overlap in the months in which short-term pre-intervention data and longer-term post-interven- tion data are collected, in turn ensuring that our comparison of respondent KSAs along these time horizons is based on data collected during the same season of the year. Phase 1 spanned a 28-week period between May and November 2011. During this phase, we planned to collect two sets of in-depth interview data before and immediately after the intervention in addition to a series of diaries data that ran from about 10 weeks before the intervention, through the eight-week interven- tion itself, and then about 10 weeks after it. The study design called for a three- month hiatus before Phase 2 diaries collection picked up again on March 12, 2012. During Phase 2, we planned to conduct a third round of in-depth interviews and collect another 14 weeks of diaries data. The project’s design was quasi-experimental in that the treatment group was not randomly selected, but recruited using HFHU’s normal process of responding to community demand for its education program. We anticipated that this would introduce selection bias into the recruitment of the treatment group and intended to mitigate the effect of this on our results by both recruiting a comparison group ex ante with similar observable characteristics and through post hoc controls on variables revealed by the diaries data. The most important method for recruiting a comparison group with similar characteristics ex ante was the identification of communities from which to draw the comparison sample that were similar to the communities in which the intervention took place. As we will discuss in more detail below, the two samples we generated using the sampling strategy described above were similar in many ways. But, as noted, there was a crucial and overwhelming difference in the pattern of their financial service use: almost all respondents in the treatment group were members of a savings group or savings and credit cooperative organization (SACCO) and used these forums frequently; very few respondents in the comparison community transacted with such entities, and those who did did so infrequently. It is likely that 12.  The impact of financial education on financial service use  ◾ 403 this is a product of the way that HFHU recruited the participants into its educa- tion program, but there may also be a fundamental difference in the financial landscape of each community. Our landscape study of the communities suggests that there were active SACCOs near each community and that each community was home to savings groups, although there may have been more groups in the treatment communities. As a result, we cannot attribute the differences in the behavior of the two groups to differences in the groups and organizations active in their communities. Rather, the data suggest that the treatment group in the sample was a group of self-selected savings group and SACCO members. Though there was some variation in the level of participation in savings groups and SACCOs within the treatment group, and there was some savings group and SACCO activity in the communities we identified to serve as a compar- ison group, the variance was not sufficient to allow us to control for the inherent bias in the data post hoc. As a result, we have reoriented the project to answer the questions raised in the introduction to this chapter: 1. What can diaries, in combination with in-depth interviews, tell us about the financial capabilities of low-income people that we might not know other- wise? 2. How can change in indicators of financial capability be tracked through diaries, in combination with in-depth interviews, over time? 3. Under what circumstances is it appropriate to use financial diaries to eval- uate the impact of a financial education program? Though the original research design was set up to allow us to conduct a difference-in-difference analysis, it still provides a useful framework for answering the questions above. With respect to the first question, we have two samples of significant interest: one a self-selected group of individuals actively participating in savings groups and SACCOs in low-income, rural communities in Uganda; and another a group of randomly selected individuals from similar communities. We have 1,456 weeks of data for the first group and over 1,919 weeks of data for the second group. Following McKenzie (2011), even with intra-class correlations due to the clustered nature of the data resulting in larger standard errors, the size of the data set from each sample allows for quantitative analysis of the indicators of financial capability discussed elsewhere in this chapter. With respect to the second question, the data set will be, again, of sufficient size to allow for quantitative comparisons of the behavior of the respondents at different points in the study. Furthermore, the continuity of the diaries data, especially in the first phase, will allow us to track the behavioral pathways followed by individual respondents or groups of respondents using a case analysis approach to the data. Finally, with respect to the third question, our analysis of the data can proceed on an “as if” basis (as if the two groups were not so different) that can demonstrate how one might use the diaries to conduct a difference-in-difference impact evaluation. 404  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 12.2.4 Research sites and sample characteristics At the time of the start of the project, HFHU’s Luweero office had a number of requests from community groups in the area to offer its financial education program in their communities. HFHU selected two communities, Ssenyomo and Kiwamirembe, in the Kikyusa subdistrict of the Luweero district, based on the criteria provided by Microfinance Opportunities. Those criteria were that the communities be ◾◾ a rural village; ◾◾ away from any main tarmac road; and ◾◾ about a U Sh 2,000 (US$2 PPP) boda boda, or motorcycle taxi, ride from a trading center.2 The two communities selected are about 20 kilometers east of the Kampala-Gulu highway by road. The nearest trading center is Kiziba. Kiwa- mirembe is about 1 kilometer from the trading center, while Ssennyomo is about 4 kilometers from it. In the case of the former, the boda boda ride costs less than U Sh 2,000; in the case of the latter, it costs about U Sh 2,500, with rates increasing in the rainy season. Our field workers also note that the price of a boda boda ride is not fixed and can vary depending on weather conditions and the bargaining power of the customer. With respect to the comparison communities, HFHU assisted Microfinance Opportunities and its field research team in identifying the geographical areas in which residents had neither requested nor received financial education but were similar in other ways to those in the treatment area. It is from these areas that Microfinance Opportunities and its field team identified the two compar- ison communities. The communities we selected were Kibowa and Kimwanyi, which are about 16 and 17 kilometers west of the Kampala-Gulu Highway by road, and thus over 35 kilometers by road from the treatment communities. The nearest trading center is Nakaseke, which is about 3 kilometers from Kibowa and 4 kilometers from Kimwanyi—although, due to the geography of Kibowa, which stretches along a dirt road heading into Nakaseke, the nearest respondent to Nakaseke is only 1.7 kilometers from the trading center. At the end of Phase 1 of the project, our financial diaries sample repre- sented 103 respondents living in 88 households. Forty-seven of these respon- dents (41  households) live in treatment communities and 56 (47 households) in comparison communities, with a fairly even split between men and women across communities. All currency data in this chapter are recorded in Ugandan shillings. There are about 2  U Sh 1,000 to US$1 in purchasing power parity (PPP) terms. 12.  The impact of financial education on financial service use  ◾ 405 Analysis of Phase 2 transaction data for the 13 weeks between March and June 2012 revealed that a total of 91 respondents living in 83 households rejoined the project for the second phase of data collection, with 12 respondents having dropped out. The data show that only 5 of the 103 original respondents declined to participate in the study further because of a lack of interest. The other seven who dropped out did so because they relocated, died, or suffered some other calamity. The Phase 2 respondent retention analysis in Stuart (2012), annex 4, contains a detailed report on the changes in the sample from Phase 1 to Phase 2. As a result, there were 39 respondents in the treatment communities and 52 in the comparison communities, with a fairly even split between men and women. There are 41 households in the treatment communities and 47 households in the comparison communities. The average household size is just over six people, with little difference across three of the four communities, but a markedly smaller average household size in Senyomo, which is a treatment community. In sum, the treatment and comparison samples are drawn from communities that are located in geographically similar areas. The data suggest that the occu- pations of members of those communities are roughly similar, and though the sample of respondents itself shows some differences in income, these are not statistically significant at the 5 percent level. The diaries data confirm that almost all respondents in both samples are farmers, but also suggest that there are some differences between the two groups with respect to the share of respondents who earn income from farming, wage work, and/or running a business. A more detailed description of the research sites and sample characteristics can be found in Stuart (2012), annex 4. 12.2.5 Methodology We used a variety of methods to generate the data on which this chapter is based. These were as follows. ◾◾ Financial diaries: Weekly interviews with each respondent in which the enumerators recorded all economic transactions performed by the respondent in the preceding week, including all formal and informal finan- cial transactions; the enumerators also asked respondents to report any unusual events that occurred during the week. ◾◾ In-depth interviews: These were conducted three times during the study period—near the start, at the end of the first phase, and during the second phase. The questions in the interview focused on respondents’ under- standing of money management and financial concepts. ◾◾ Case study interviews. ◾◾ Financial landscape study. More information on these methodologies can be found in Stuart (2012), annex 5. 406  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 12.3 FINDINGS: IMPACT ON FINANCIAL SERVICE USE The conceptual framework identified three major areas of financial capability: managing cash flows, planning ahead, and financial service use. It also noted that there are three dimensions to these areas: cognitive, behavioral, and envi- ronmental. The analysis that follows focuses on financial service use, and, where available, reports evidence regarding changes in cognition and behavior. It should be noted that for the sake of this analysis, we have split the sample temporally into pre- and post-intervention periods, with the cutoff being the 22nd week of the study, when the HFHU financial education was completed. The choice of the cutoff was difficult to make, given the fact that the HFHU program ran for eight weeks and covered a variety of topics. It is possible that during the training program behavior changed temporarily, and so it would not be appropriate to use the end-of-program date as a cutoff, but a review of the data suggests that that was not the case. The longer baseline period gained from using the end-of-pro- gram date as the cutoff gives us more precision in our measures of baseline behavior, because we have more data. In addition, to increase the precision of the post-intervention data, we combined the data from the end of Phase 1, after the intervention occurred, with the data from Phase 2. 12.3.1 Overview The respondents in the study used a wide variety of financial services. The most common services used were savings accounts held with savings groups, SACCOs, and banks. Respondents used loans from those sources as well, but much less frequently. On top of these organizational sources of services, respondents also saved money at home, transferred cash to/from a spouse, and gave and received cash gifts and loans to/from family members and friends. Respondents also gave and received in-kind loans and loan repayments to/from other individuals. Overall, respondents engaged in some sort of financial transaction with an individual other than their spouse once every two and a half weeks, averaging about U Sh 17,642 per transaction (tables 12.3 and 12.4). Their financial interac- tions with banks, SACCOs, savings groups, or some other type of organization were more frequent, over once every two weeks, and the amounts involved were much larger. The average transaction involved about U Sh 33,000. The frequency of transactions did not vary considerably by gender, but the amounts involved did. The average amount of transactions men conducted with other individuals was about 1.5 times the average amount of women’s transactions with other individ- uals. In the case of transactions with organizations, the men’s average amount was almost twice that of the women’s. As noted above, there was a considerable difference between the treatment and comparison groups in terms of their financial transactions. Respondents in the treatment group conducted a transaction with an organization, mostly a 12.  The impact of financial education on financial service use  ◾ 407 TABLE 12.3  Number of financial transactions per week WOMEN MEN TREAT- COMPARI- TREAT- COMPARI- TOTAL MENT SON TOTAL MENT SON TOTAL ALL Individuals 0.62 0.30 0.45 0.52 0.33 0.41 0.43 (0.10) (0.07) (0.07) (0.12) (0.05) (0.06) (0.05) All organizations 1.18 0.16 0.62 1.13 0.14 0.55 0.59 (0.13) (0.06) (0.09) (0.15) (0.06) (0.10) (0.08) Bank 0.01 0.02 0.02 0 0.08 0.05 0.03 (0.01) (0.02) (0.01) — (0.03) (0.02) (0.01) SACCO 0.86 0.01 0.39 0.77 0.03 0.33 0.36 (0.18) (0.01) (0.08) (0.21) (0.02) (0.09) (0.08) Savings group 0.26 0.11 0.17 0.35 0.02 0.16 0.17 (0.10) (0.04) (0.06) (0.10) (0.01) (0.04) (0.04) Other 0.06 0.02 0.03 0.01 0.02 0.02 0.02 (0.03) (0.01) (0.02) (0.00) (0.01) (0.01) (0.01) Total with others 1.80 0.47 1.07 1.65 0.48 0.96 1.01 (0.19) (0.09) (0.16) (0.19) (0.08) (0.10) (0.08) Missing 0.00 0.00 0.00 0.00 0.00 0.00 0.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) Note: Robust standard errors in parentheses. TABLE 12.4  Mean amount of financial transactions (U Sh) WOMEN MEN TREAT- COMPARI- TREAT- COMPARI- TOTAL MENT SON TOTAL MENT SON TOTAL ALL Individuals 15,241 13,810 14,704 15,894 26,766 21,085 17,642 (2,624) (3,452) (1,830) (2,421) (5,145) (2,515) (1,928) All 16,820 67,945 24,090 25,993 140,773 43,730 32,966 organizations (3,182) (24,927) (4,411) (9,451) (79,527) (13,172) (6,037) Bank 157,688 278,710 244,132 228,884 228,884 233,576 — (68,029) (105,041) (52,161) (94,780) (134,162) (69,258) SACCO 14,071 39,360 14,440 32,050 9,000 30,981 21,718 (2,583) (25,422) (2,207) (21,044) (1,885) (19,618) (6,413) Savings 16,675 24,717 19,435 11,018 11,333 11,040 15,627 group (7,556) (6,869) (5,671) (1,461) (13,158) (1,097) (2,666) Other 34,111 73,367 44,092 110,160 57,900 68,352 51,312 (8,950) (24,954) (10,755) (31,883) (33,258) (29,717) (12,654) Total with 13,557 9,560 12,028 15,134 18,029 16,229 13,941 others (2,578) (2,824) (2,011) (3,269) (5,428) (2,937) (1,633) Missing 182,000 10,000 124,667 120,800 72,500 88,600 106,633 (134,124) (4,708) (87,172) — (42,754) (30,816) (44,772) Note: Robust standard errors in parentheses. 408  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES savings group or a SACCO, more than once a week; they conducted transac- tions with individuals other than their spouse more than once every two weeks. In contrast, respondents in the comparison group conducted a transaction with another individual, other than their spouse, once every three weeks; and with some sort of organization, mostly banks, savings groups, or SACCOs, once every seven to eight weeks (table 12.3). Furthermore, the size of transactions conducted with financial institutions differed between treatment and comparison groups. The treatment groups engaged in smaller transactions with smaller variance than the comparison group, members of which conducted large and infrequent trans- actions, mostly with banks (table 12.4). The data summarized above are for all financial transactions with other indi- viduals and organizations. In the rest of this section, we document the use of services in terms of number of transactions and the amounts involved by type of transaction. For savings and cash gifts, we will build a model to identify factors that prompted respondents to deposit and withdraw savings or to give or receive cash gifts, and then evaluate whether behavior changed from the period before the financial education intervention to the period after it by comparing changes in the treatment and comparison groups’ use of savings over the study period. We do not perform the same analysis of loan activity, as the number of loans given or received was too small. 12.3.2 Savings The HFHU financial education curriculum included two sessions on savings, as described earlier in the chapter. The interview data illustrate the savings knowl- edge and attitudes of participants in the study, identifies barriers to savings, and can help to explain the motivations and context of participant savings behavior. Consistent with the transactions data reported in tables 12.3 and 12.4 and below, the interview data revealed that treatment and comparison participants saved in different locations. Treatment community participants were very active in what they referred to as “groups.” A small number of treatment participants also reported saving at home. In contrast, very few comparison participants reported participating in groups, and many more saved at home or in banks. Some inter- viewees saved in multiple locations, particularly in the comparison sample. No drastic changes were found in where comparison group participants saved over the last two rounds of interviews. However, there were two note- worthy changes among treatment participants. First, more treatment participants mentioned saving in groups and saving at home during the final two rounds. This finding is linked to the second change, which concerned the number of locations where each participant saved. During the first round of interviews, the number of locations where treatment group participants reported saving averaged just over one location apiece. During the second and third interviews, the number of locations where treatment group participants reported saving rose to nearly two 12.  The impact of financial education on financial service use  ◾ 409 locations apiece, with small increases coming in SACCOs, livestock or farming, property, and unspecified investments. Another important finding from the first round of interviews concerned saving location preferences; this can help to assess participant knowledge of formal, semi formal, and informal savings options, which are addressed in the HFHU curriculum. Initially, more people in the comparison sample felt that banks were the best savings location, while treatment sample participants were split between whether banks or groups were preferable. It was not uncommon for people in the comparison group to say that they would like to save in banks for security reasons, but some complained that banks were inaccessible due to distance. This suggested that comparison sample participants are aware of banks and their benefits. Treatment sample participants expressed this opinion less frequently, but it is not clear whether this was due to lower awareness or to their satisfaction with their savings groups. Very few participants in either sample felt that saving at home was the best method, and the only advantage they cited for home savings was the convenience of immediate access. The preferred savings locations of treatment participants changed from the first round to the final rounds of the research; this was not true of the comparison group, despite slight fluctuations. Treatment participants increasingly favored saving in groups during the second and third rounds of the research, while fewer favored saving in banks. At the same time, while a small number of treatment participants during the first round felt that the best way to save was with friends or family, or in livestock or farming, the numbers decreased during the second round and none felt that either way was best by the third round. Participants who felt that banks were the best savings location felt that way mainly because banks offer security (table 12.5). No participants mentioned earning interest as a benefit of using banks. By contrast, earning profits or interest from savings groups was a key motivation participants cited for keeping money with groups. Two other important benefits that participants cited for using groups—not cited when discussing banks—were convenience in accessing their money and the ability to borrow funds. The number of treatment participants citing interest and convenience as benefits of savings groups doubled from the first round to the last, suggesting an added awareness of the financial benefits that groups offer. Treatment participants increasingly favored using groups, but because very few thought banks were the best savings location by the end of the study, their specific level of knowledge of banks was not measurable. One interpretation of this could be that knowledge of these relative benefits may still not be enough to convince participants to open and use bank accounts because of the associated transaction costs. Treatment participants, as will be shown, were also less likely at the end of the study to think of money with savings groups as an investment; this may be a result of exposure to information in the HFHU curriculum about savings options, causing them to alter the way they think about the money they keep with their groups. 410  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 12.5  Quotes about the benefits of different savings locations TOPIC QUOTE AND PARTICIPANT CODE Distance of banks “Banks are the best [savings option], but the only problem is that they are far from us. You have to go up to Wobulenzi or Luwero to access one.” —II 305 (Round 1, comparison) “It is very safe, besides that you take a long time to withdraw your money because it is not near you.” —II 904 (Round 3, comparison) Security of banks “[Banks] have a lot of security unlike in [savings] groups and at home. You are assured that your money is safe.” —II 111 (Round 1, treatment) “You do not use your money at all times because it is not near you and thieves cannot access your money like when you keep it all at home. I believe if you kept your money in the bank you also need to keep a little at home because of the abrupt problems that may come up.” —II 903 (Round 3, comparison) Profits from “When you keep money in a group, you get profits. We are able to get profits savings groups because some people borrow from the same group at an interest rate.” —II 101 (Round 1, treatment) “Groups give you the opportunity to borrow when you do not have money and they earn you some profit when time to share comes; that profit comes from the interest that people pay on loans.” —II 115 (Round 3, treatment) Convenience of “Saving groups [are best] because they are more accessible and you can easily get savings groups money from there in case you need it.” —II 407 (Round 1, treatment) “Saving groups [are best] because [your] money is always available and no condi- tions are attached to the money you borrow.” —II 107 (Round 3, treatment) The data from the financial diaries show that the bulk of the financial activity of the respondents in the study was savings related—either with a bank, SACCO, or savings group (henceforth referred to collectively as “financial institution”) or at home. Respondents made a deposit or withdrawal from a financial institution about once every two weeks and from their home savings about once a week (tables 12.6 and 12.7). The treatment and comparison groups had very different levels of transaction activity with financial institutions, both in terms of frequency and average amounts per transaction. The treatment group interacted far more frequently with financial institutions than did the comparison group (once a week versus once every eight weeks), and the amounts involved each time were smaller (U Sh 21,000 versus U Sh 102,000). This difference was driven by the treatment group’s extensive use of savings groups and SACCOs. In contrast, the home savings activity of each group was roughly similar in terms of frequency and amount per transaction. Respondents in both study groups deposited or withdrew money into/from their home savings just under once a week, and the average amount of the transaction was U Sh 46,000 in the case of the comparison group, and U Sh 41,000 in the case of the treatment group (tables 12.6 and 12.7). Differences across gender varied depending on the group and the type of transaction. Men in the comparison community deposited or withdrew money 12.  The impact of financial education on financial service use  ◾ 411 TABLE 12.6  Number of savings transactions per week WOMEN MEN TREAT- COMPAR- TREAT- COMPAR- TOTAL MENT ISON TOTAL MENT ISON TOTAL ALL With others Individuals 0.00 0.00 0.00 0.00 0.00 0.00 0.00 (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) (0.00) All organizations 1.02 0.13 0.53 1.01 0.13 0.49 0.51 (0.12) (0.05) (0.09) (0.11) (0.05) (0.07) (0.06) Bank 0.01 0.02 0.01 0 0.06 0.04 0.03 (0.01) (0.01) (0.01) — (0.03) (0.02) (0.01) SACCO 0.79 0.01 0.36 0.69 0.03 0.30 0.33 (0.17) (0.00) (0.09) (0.18) (0.02) (0.07) (0.06) Savings group 0.21 0.10 0.15 0.32 0.02 0.14 0.14 (0.08) (0.04) (0.04) (0.10) (0.01) (0.05) (0.03) Other 0.01 0.00 0.01 0 0.00 0.00 0.01 (0.01) (0.00) (0.01) — (0.00) (0.00) (0.00) Total with others 1.03 0.13 0.53 1.01 0.12 0.48 0.51 (0.13) (0.05) (0.08) (0.10) (0.05) (0.10) (0.06) 0.66 0.73 0.70 0.76 0.87 0.83 0.76 Home savings (0.06) (0.04) (0.03) (0.06) (0.03) (0.04) (0.02) Note: Robust standard errors in parentheses. TABLE 12.7  Mean amount of savings transactions WOMEN MEN TREAT- COMPAR- TREAT- COMPAR- TOTAL MENT ISON TOTAL MENT ISON TOTAL ALL With others Individuals 2,200 5,200 3,400 16,000 4,667 9,200 6,300 (1,152) (2,890) (1,368) (9,978) (2,732) (6,242) (3,222) All organizations 11,917 65,486 19,209 22,447 152,179 42,192 29,783 (1,963) (25,942) (4,685) (8,417) (76,171) 15,661 (8,406) Bank 157,688 304,364 257,428 — 228,884 228,884 236,993 (72,995) (100,329) (83,960) (108,010) (126,856) (84,742) SACCO 11,060 39,360 11,507 29,262 9,000 28,213 18,731 (1,841) (21,613) (1,587) (18,245) (1768) (15030) (6,140) Savings group 7,190 18,793 11,376 7,996 11,333 8,253 9,881 (1,411) (5,911) (3,071) (1,017) (12,913) (1,049) (1,326) Other 24,545 55,000 32,667 — 48,333 48,333 35,278 (7,877) (24,940) (8,056) (16,876) (16,185) (7,661) Total with others 12,291 60,217 18,698 22,428 133,997 38,285 27,656 (1,942) (23,762) (4,755) (10,289) (87,888) (12,131) (6,935) 30,419 28,491 29,313 51,533 60,319 56,973 43,802 Home savings (2,496) (7,192) (5,291) (7,023) (6,605) (5,769) (3,024) Note: Robust standard errors in parentheses. 412  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES into/from home savings slightly more frequently than the women in that group— the difference was statistically significant but substantively not great. Within the treatment group, the differences were not statistically significant. In both treat- ment and comparison communities, the amounts men deposited into or withdrew from their home savings were far larger than the amounts women transacted. The average amount women transacted was about U Sh 30,000, while the average amount for men ranged from U Sh 51,500 (treatment) to U Sh 60,000 (compar- ison). In the case of financial institution transactions, there was little difference in the frequency of transactions conducted by men and women. The variance of the transaction amounts was high in most cases, so interpreting the differ- ences across groups is difficult. But the data do suggest that, on average, the size of men’s and women’s transactions with savings groups were roughly the same (U Sh 7,996 and U Sh 7,190 respectively). The case study data revealed the logic behind the highly consistent pattern of deposits made by many respondents in savings groups and SACCOs. They explained that the logic of interest earnings combined with an annual share-out period motivated them to do so. That is, they understood that the more they saved in these forums, the more they would earn in interest. Having their savings “frozen” until share-out appealed to them because it helped avoid temptations to spend (box 2.1). Additionally, many respondents described having joined savings groups and SACCOs in order to gain access to loans, either for unexpected emer- gency expenses or for business investments. CHANGES IN SAVINGS BEHAVIOR The diaries data allow us to look at the coincidence of transactions in time, and also to calculate the overall balance of income and expenditures within a given week. As a result, we can model savings behavior, and then see whether there was any change in behavior between the period before and during the financial education program and the period after it. The models below include variables BOX 12.1  A case study in group savings discipline Isaac saved with two different SACCOs, Muyizi Kasubwa SACCO and Suluma Farmers Group, making consistent weekly deposits of an average of U Sh 6,988. Isaac explained that he saved with these SACCOs in order to be prepared for emergencies and for weeks in which his income was low. Isaac remarked that, even though there are no penalties for missing a weekly deposit, it is important for him to save each week in order to increase his earnings at share-out. This commitment is reflected in the fact that he continued to make deposits every week in Phase 2, despite the decrease in income he experienced. 12.  The impact of financial education on financial service use  ◾ 413 likely to be correlated with savings deposits into formal and semiformal accounts and home savings. We have presented models that look at both the amount of the deposits as the dependent variable and the incidence of deposits (1 if the deposit happened; 0 if it did not). They include the following: ◾◾ Post: 1 if week fell after the end of the whole financial education program (weeks 22+), 0 otherwise ◾◾ Balance_iht: the amount by which the respondent’s net income during the week exceeded his or her household expenses (business expenses are taken into account through the calculation of net income) ◾◾ Deficit_iht: 1 if balance_iht is less than zero and 0 otherwise ◾◾ Harverst1: 1 if week fell during the harvest period (weeks 13–21), 0 other- wise ◾◾ Gender1: 0 if the respondent is a woman, and 1 if the respondent is a man ◾◾ Medbill: the amount spent on medical bills during the week ◾◾ Medbill1: 1 if a medical bill was paid, 0 otherwise ◾◾ Schoolfees: the amount spent on school fees during the week ◾◾ Schoolfees1: 1 if school fees were paid, 0 otherwise ◾◾ Deposits: amount deposited in a formal or semiformal account ◾◾ Deposits1: 1 if deposit was made, 0 otherwise ◾◾ Home_Dep: amount deposited in home savings ◾◾ Home_Dep1: 1 if home deposit was made, 0 otherwise ◾◾ Tot_fin_in: the amount of inflows from financial sources, including cash gifts or loan repayments from other individuals, withdrawals from a savings account, a loan from an individual or financial institution Medical bills are included in the model in order to capture the impact of a large, unexpected expenditures on the savings behavior of the respondents; while school fees capture the impact of large but anticipated expenditures on this behavior. If we run the model for each group, the results suggest that the treatment group changed its savings behavior from the period before it finished its finan- cial education program to the period afterwards, while the comparison group’s behavior did not change. This is the case in terms of both the amount deposited and the incidence of deposits for home savings; it is also the case in terms of the amounts deposited in formal and semiformal accounts. This result from the trans- action data is consistent with the findings from the in-depth interviews, which also found that treatment group respondents reported an increase in savings. The results also suggest that, other than the post dummy variable, the independent variables were correlated with the two groups’ deposit behavior in similar ways. 414  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES HOME DEPOSITS Controlling for other factors that might have affected how much people deposited in their home savings, we find that individuals in the treatment group deposited more in their home savings in the post-intervention period than they did in the pre-intervention period (table 12.8). The coefficient on the post dummy variable is strongly significant, suggesting that there was an increase in average deposits of about U Sh 9,600 between the pre- and post-intervention periods after controlling for other factors. In contrast, the coefficient on the post dummy variable in the model of comparison group behavior was in the opposite direction, suggesting a decrease in home savings deposits, though this was not statistically significant at the 5 percent level. The variables correlated with home deposits were similar for both the treat- ment and comparison groups. In both cases, men deposited more than women; deposits in formal and semiformal accounts were negatively correlated with home deposits; home deposits went up during the harvest period; home deposits went down when the respondents spent more than they earned; and medical bills had no impact on home savings deposits (box 2.2). The one exception was TABLE 12.8  Model of home savings deposit behavior OBSERVED BOOTSTRAP NORMAL BASED VARIABLE COEFFICIENT STANDARD ERROR Z P > |Z| 95% CONFIDENCE INTERVAL Comparison group (replications based on 52 clusters in respondent ID) Post −6,429.57 3,386.78 −1.90 0.06 −13,067.53 208.39 Gender1 11,143.47 4,115.49 2.71 0.01 3,077.27 19,209.68 Medbill 0.01 0.18 0.07 0.95 −0.34 0.37 Schoolfees 0.24 0.15 1.59 0.11 −0.06 0.54 Deposits −0.55 0.16 −3.42 0.00 −0.87 −0.24 Harvest1 7,234.46 3,723.48 1.94 0.05 −63.42 14,532.35 Balance_iht 0.56 0.11 5.12 0.00 0.34 0.77 Tot_fin_in 0.55 0.10 5.39 0.00 0.35 0.75 _cons 12,301.75 4,988.39 2.47 0.01 2,524.68 22,078.82 Treatment group (replications based on 39 clusters in respondent ID) Post 9,591.84 2,132.24 4.50 0.00 5,412.72 13,770.95 Gender1 9,567.55 3,024.48 3.16 0.00 3,639.68 15,495.42 Medbill 0.07 0.14 0.48 0.63 −0.21 0.35 Schoolfees 0.29 0.11 2.66 0.01 −0.08 0.50 Deposits −0.35 0.10 −3.45 0.00 −0.55 −0.15 Harvest1 3,803.55 18,87.63 2.01 0.04 103.86 75,03.24 Balance_iht 0.35 0.11 3.24 0.00 0.14 0.56 Tot_fin_in 0.45 0.08 5.51 0.00 0.29 0.60 _cons 1,114.30 1,882.26 0.59 0.55 −2,574.86 4,803.46 12.  The impact of financial education on financial service use  ◾ 415 BOX 12.2  A case study in increased savings Harriet independently reported that after attending an HFHU training, she and her husband realized they had been overspending and started working to decrease their expenditures and save more, which in turn allowed them to invest in a plot of land and building material. She shared having cut down on unnecessary expenditures such as monthly trips to Kampala to visit family and giving her children money for snacks when she had already paid for school lunches. the payment of school fees. The impact of paying these fees on home savings behavior was positive and statistically significant in the case of the comparison group; there was no correlation between fees and home savings deposits for the treatment group. It is unclear why this was the case, especially given that the school fees paid by comparison group respondents were almost twice as high, on average, as those paid by treatment group respondents. This suggests that, overall, the drivers of savings behavior in the treatment and comparison groups were similar, and that the major difference between them was how much they saved in the pre- and post-intervention periods. Figure 12.1 partially depicts what happened, showing the amount deposited in home savings as a share of net income for each group, during weeks when the respondents reported that they did not receive any financial inflows. The data suggest that there was a savings increase as a share of income in the immediate FIGURE 12.1  Home deposit amounts as a share of income by week and group, when financial inflows are zero Percent 70 Comparison Treatment Comparison average Treatment average 60 50 40 30 20 10 0 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 46 48 50 52 54 56 58 Study week 416  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES post-intervention period for those in the treatment group, and that this increase remained, at a lesser level, six months later. In contrast, the comparison group decreased its home savings deposits in the post-intervention period. FORMAL AND SEMIFORMAL SAVINGS ACCOUNT DEPOSITS The data also suggest that respondents in the treatment group increased the amount they deposited in savings accounts each week after the financial educa- tion program was over. In contrast, the comparison group did not. Controlling for other factors that might have affected how much people deposited in their savings accounts, we find that the difference in savings behavior of the treatment group between the pre- and post-intervention periods holds true (table 12.9). The coefficient on the post dummy variable is strongly significant, suggesting that there was an increase in average deposits of about U Sh 6,200 between the pre- and post-intervention periods after controlling for other factors. In contrast, the coefficient on the post dummy variable in the model of comparison group behavior was not statistically significant. The match between the statistical TABLE 12.9  Model of semiformal and formal deposit behavior OBSERVED BOOTSTRAP NORMAL BASED VARIABLE COEFFICIENT STANDARD ERROR Z P > |Z| 95% CONFIDENCE INTERVAL Comparison group (replications based on 52 clusters in respondent ID) Post 789.4402 1,215.14 0.65 0.52 −1,592.192 3,171.07 Gender1 1,071.887 3,537.52 0.30 0.76 −5,861.518 8,005.29 Medbill 0.02 0.07 0.22 0.82 −0.12 0.15 Schoolfees 0.08 0.05 1.79 0.07 −0.01 0.18 Deposits −0.19 0.10 −1.91 0.06 −0.39 0.00 Harvest1 553.3675 845.00 0.65 0.51 −1,102.807 2,209.54 Balance_iht 0.19 0.09 2.05 0.04 0.01 0.37 Tot_fin_in 0.19 0.09 2.12 0.03 0.01 0.36 _cons 4,032.797 1,882.78 2.14 0.03 342.6219 7,722.97 Treatment group (replications based on 39 clusters in respondent ID) Post 6,242.98 2,691.11 2.32 0.02 968.51 11,517.45 Gender1 5,433.21 3,472.29 1.56 0.12 −1,372.35 12,238.77 Medbill 0.29 0.23 1.24 0.22 −0.17 0.74 Schoolfees 0.54 0.29 1.88 0.06 −0.02 1.11 Deposits −0.57 0.28 −2.03 0.04 −1.11 −0.02 Harvest1 8,136.16 5,196.98 1.57 0.12 −2,049.73 18,322.05 Balance_iht 0.62 0.29 2.11 0.04 0.04 1.19 Tot_fin_in 0.56 0.27 2.10 0.04 0.04 1.08 _cons 568.18 3,137.54 0.18 0.86 −5,581.28 6,717.65 12.  The impact of financial education on financial service use  ◾ 417 significance and direction of the coefficients was less similar in the treatment and comparison groups, suggesting that drivers of deposit behavior across the two groups were different (table 12.9). This is not surprising given the fact that, as we have observed before, the respondents in the two groups used formal and informal accounts very differently. Figure 12.2 partially depicts what happened, showing the amount deposited in semiformal and formal savings accounts as a share of income for each group, during weeks when the respondents reported that they did not receive any finan- cial inflows. In contrast to the home deposit data, the figure suggests that there was an initial small increase in deposits in the period right after the end of the financial education program, and that there was a greater increase six months later. Furthermore, the figure shows how the savings deposits of the comparison group also increased—but not by much. FIGURE 12.2  Formal and semiformal deposit amounts as a share of income by week and group, when financial inflows were zero Percent 30 Comparison Treatment Comparison average Treatment average 25 20 15 10 5 0 5 7 9 11 13 15 17 19 21 23 25 27 29 31 33 46 48 50 52 54 56 58 Study week Note: Data depicted exclude eight deposits in amounts greater that U Sh 500,000, which are outliers. 12.3.3 Cash gifts As in the other studies Microfinance Opportunities has conducted in East Africa, the transactions data from Uganda suggest that the respondents frequently exchanged cash gifts. On average, respondents received a gift about once every six weeks and gave one about every seven weeks—excluding transfers between husbands and wives (table 12.10). The average amount given was U Sh 2,800, and the average amount received was U Sh 992 (table 12.11). There was considerable 418  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 12.10  Cash gifts received and given: transactions per week OBSERVED BOOTSTRAP NORMAL-BASED GENDER AND GROUP MEAN STANDARD ERROR 95% CONFIDENCE INTERVAL Gifts received Women, comparison 0.15 0.04 0.07 0.23 Women, treatment 0.12 0.02 0.07 0.16 Men, comparison 0.24 0.06 0.13 0.36 Men, treatment 0.09 0.03 0.02 0.15 Gifts given Women, comparison 0.09 0.02 0.04 0.14 Women, treatment 0.12 0.02 0.07 0.16 Men, comparison 0.18 0.04 0.11 0.26 Men, treatment 0.18 0.04 0.11 0.25 Note: Replications based on 91 clusters in respondent ID). TABLE 12.11  Cash gifts received and given: amount per transaction OBSERVED BOOTSTRAP NORMAL-BASED GENDER AND GROUP MEAN STANDARD ERROR 95% CONFIDENCE INTERVAL Gifts received Women, comparison 2,514 800 946 4,082 Women, treatment 2,578 575 1,451 3,705 Men, comparison 4,883 1,298 2,338 7,427 Men, treatment 1,141 577 9 2,272 Gifts given Women, comparison 755 227 310 1,200 Women, treatment 1,357 383 605 2,108 Men, comparison 770 252 275 1,264 Men, treatment 1,072 270 544 1,600 Note: Replications based on 91 clusters in respondent ID. difference in the frequency with which men and women gave and received cash gifts. Men gave gifts more frequently than women, and the latter received gifts more than did the former. Looking across the treatment and comparison groups, the two groups were very similar in terms of gifts given, but the men in the comparison group received gifts far more frequently than the men in the treat- ment group. There were no discernible changes in cash gift activity between the pre- and post-intervention periods, except that the men in the comparison group expe- rienced a dramatic decline in the number of cash gifts they received over that period. It is unclear what drove this change in behavior. 12.  The impact of financial education on financial service use  ◾ 419 12.3.4 Debt The respondents in the diaries study received a limited number of loans from financial institutions and individuals during the study period—155 in total. Seven- ty-one of these were from financial institutions, and 68 were from savings groups or SACCOs. There were no loans from banks reported. Not surprisingly, most of the loans from savings groups and SACCOs went to treatment group respon- dents. In addition, treatment respondents received the bulk of the loans from individuals—63 of 83. One interpretation of this latter finding is that the treatment group’s social networks were much richer than those of the comparison group. Not only were they part of active savings groups, but they were also able to call on members of their community for loans. In the qualitative interviews, it was clear that respondents thought of asking someone else for a loan as imposing a burden on them and was something that should not be done casually. There was no change in the rate at which respondents used loans across the pre- and post-intervention time periods, but the limited number of loans that the respondents received makes the detection of any change very difficult due to the lack of precision in the data. 12.3.5 Financial service use: summing up ◾◾ Transaction data from the diaries suggest that by far the most common savings mechanism among the respondents was saving at home. ◾◾ Transaction data from the diaries suggest that respondents in the treat- ment group increased the amount they saved at home and in their savings groups and SACCOs from the pre-intervention period to the post-interven- tion period. The same was not the case for the comparison group. —— In-depth interview data support the finding on increased savings: respondents in the treatment group reported saving more in the second and third rounds of interviews, both of which occurred after the inter- vention. ◾◾ Transaction data suggest that respondents borrowed little during the period of the study: they received 155 loans, of which 71 were from a savings group, either a SACCO or a nongovernmental organization. There were no bank loans reported. The rest of the loans were from other individuals. 12.4 CONCLUSION This chapter began with three questions: 1. What can diaries, in combination with in-depth interviews, tell us about the financial capabilities of low-income people that we might not know other- wise? 420  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 2. How can change in indicators of financial capability be tracked through diaries, in combination with in-depth interviews, over time? 3. Under what circumstances is it appropriate to use financial diaries to eval- uate the impact of a financial education program? In answer to the first question, the chapter draws on a rich set of qualitative and quantitative data to describe the lives of low-income Ugandans. The data were able to provide considerable insight into how the respondents manage their cash flow and risk, and use financial services. The data demonstrated that low-in- come Ugandans form multiple financial relationships, have highly varied incomes, and are prone to exogenous shocks. Anyone hoping to implement a financial education program must keep these complexities in mind and find ways to enable low-income people to deal with the realities of their lives as they live them. In answer to the second question, the combination of in-depth interviews, diaries, and case studies identified situations where changes in KSAs resulted in changed behavior and situations where, despite changes in the former, there was no discernible change in the latter. As such, it provided insight into the issue of the connection between changes in KSAs and changes in behavior. In addition, the chapter took a step away from traditional quantitative anal- yses that rely on cross-sectional data and examine behavior in terms of respon- dent attributes (such as gender or average income). Instead, the chapter used the diaries to analyze behavior in one aspect of a person’s life—say saving money in a particular week—with behavior in another aspect of their lives—say income earned in a particular week or whether they made a medical expenditure in that week. In essence, the analysis in this chapter sought to explain, or at least correlate, behaviors with behaviors, rather than attributes with behaviors. The chapter used the same approach in tracking changes in behavior—looking at whether behavior changed in one aspect, controlling for changes in behavior in other aspects of people’s lives. To the extent that financial education is about behavior change, the analytical approach described above demonstrates the power of diaries data for under- standing the impact of financial education. Diaries data allow a researcher to track changes in behaviors of interest, such as saving or spending on unneces- sary items. This chapter showed how this could be done while controlling for changes in behavior in other, related aspects of respondents’ lives. The analysis was fairly simple, and was limited by the size of the sample; there is more that can be done to unpack the causal chain linking financial education to behavior change, by looking at, for example, intermediate behavioral changes. Despite these limitations, we believe this chapter has demonstrated how one might go about this, by operationalizing measures of financial capability using the diaries data and analyzing whether those data demonstrated any behavioral changes. Finally, despite the sampling problems this project encountered, the findings of the chapter suggest that the financial education program in Uganda—however 12.  The impact of financial education on financial service use  ◾ 421 flawed in its implementation, and however biased in its recruitment processes— had some effect on the financial behavior of the participants. In particular, the transactions and Phase 2 data suggest that the respondents in the treatment group changed their savings behavior. In particular, the finding that the treatment group experienced an increase in the amount they saved at home—an activity common to both groups—was most striking. The findings also showed the potential disjuncture between a change in KSAs on the one hand and behavior on the other hand. The example presented in this study was the finding in the qualitative data that the treatment group felt it had improved its control over money management, but there was no change in incidence of weeks when group members spent more than they earned. REFERENCES Collins, Daryl, Jonathan Morduch, Stuart Rutherford, and Orlanda Ruthven. 2009. Portfolios of the Poor: How the World’s Poor Live on $2 a Day. Princeton: Princeton University Press. Holzmann, Robert. 2010. “Bringing Financial Literacy and Education to Low and Middle Income Countries: The Need to Review, Adjust, and Extend Current Wisdom.” Paper prepared for the Conference on Financial Literacy: Implications for Retirement Security and the Financial Market Place, Wharton School of the University of Pennsylvania, Philadelphia, April 29–30. Stuart, Guy. 2012. Uganda Financial Education Impact Project: Final Report. World Bank. http://responsiblefinance.worldbank.org/~/media/GIAWB/FL/Documents/Publications/ MFO-Uganda-Evaluation-Report.pdf. Stuart, Guy, and Monique Cohen. 2011. Cash In, Cash Out Kenya: The Role of M-PESA in the Lives of Low-Income People. Washington, DC: Financial Services Assessment Project. http://www.gsma.com/mobilefordevelopment/wp-content/uploads/2012/03/ cashincashoutkenya.pdf. Stuart, Guy, Michael Ferguson, and Monique Cohen. 2011. Cash In, Cash Out: Financial Transactions and Access to Finance in Malawi. Washington, DC: Financial Services Assessment Project. https://opportunity.org/content/News/Publications/Knowledge%20 Exchange/Cash%20In-Cash%20Out-Financial%20transactions%20and%20access%20 to%20finance%20in%20Malawi.pdf. CHAPTER 13 Increasing the impact of conditional cash transfers through financial literacy Evaluation pilot in the Dominican Republic XAVIER GINÉ, DEAN KARLAN, AND GREG FISCHER 13.1 INTRODUCTION This chapter gives an overview of the status of the evaluation pilot in the Dominican Republic, including the design, sample, and evaluation method- ology, and preliminary results from the baseline. While this project began in February 2012, there was significant turnover in the Dominican government in August 2012, which delayed the process of gaining government approval for the research design, accessing sample data, and identifying and hiring government-approved surveyors. The baseline survey was completed in June 2013, and the financial education material has been finalized. The financial literacy implementation is under way and should be completed in June 2014. 13.2 BACKGROUND Progresando con Solidaridad is a conditional cash transfer (CCT) program in the Dominican Republic which began in 2004 and focuses on promoting investments in human capital among low-income households. Beneficiaries receive monthly transfers to a debit card that can be used to purchase basic goods in qualifying stores (colmados). The program currently serves approxi- mately 25 percent of the national population. The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 423 424  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Poverty levels are assessed by a countrywide census that uses 17 indica- tors, including living conditions, employment status of household members, and education level for household members. Households defined as poor Type I (extreme poverty) and poor Type II (moderate poverty) are eligible to participate in Progresando con Solidaridad. Evaluations of CCT programs such as Oportunidades in Mexico and Fami- lias en Acción in Colombia have shown that such programs are successful in increasing usage of health care and education services (Attanasio et al. 2005; Baez and Camacho 2011; Gertler and Boyce 2003). However, even with improved health and education outcomes, it may remain difficult for CCT beneficiaries to manage their household finances, find stable employment, or start profitable businesses. All of these problems affect beneficiaries’ ability to graduate from the CCT program and achieve a sufficient level of economic stability on their own. The program does not have a graduation strategy, so the government is inter- ested in ways to improve the income-generating opportunities of current benefi- ciaries. In this context, it is piloting a series of projects to improve financial literacy and access to credit and savings. Previous research conducted by the Dominican government has shown that Progresando con Solidaridad beneficiaries tend to have low levels of financial literacy and little access to financial services (GCPS 2011). This financial literacy pilot program aims to improve the financial knowl- edge of Progresando con Solidaridad beneficiaries, thus increasing their ability to better use household resources and increase access to formal financial services. Innovation for Poverty Action plans to evaluate the impact of a package of financial training services on beneficiaries’ economic well-being. The package includes training in financial literacy and household financial management, credit and savings as products offered by the formal financial sector to address finan- cial needs, job skills training including how to find and keep a job, and entrepre- neurship/business administration training. The evaluation will also measure the impact of increased access to financial products, particularly savings. 13.2.1 Theory of change In the short term, we expect the program to have an impact on beneficiaries’ knowledge in terms of personal and household finances, take-up of formal sector financial services, job skills, and small business administration. In the long term, the change in knowledge should in turn affect behavior, specifically the following: ◾◾ Improved management of household finances, by identifying planning, and keeping track of recurrent income and expenses, and better control in terms of spending ◾◾ Increased savings, both in kind and monetary, resulting from more consci- entious spending habits as well as awareness of the importance and utility of saving 13.  Increasing the impact of conditional cash transfers through financial literacy   ◾ 425 ◾◾ Improved usage of credit resulting from more informed choices of products and avoiding overindebtedness ◾◾ Increased usage of formal sector (regulated) savings and credit products ◾◾ Improved ability to search for and retain formal employment, resulting in a better overall employment situation ◾◾ Improved management of small businesses, resulting in higher income levels; similarly, we expect to see a higher incidence of new businesses opened 13.2.2 Hypotheses Based on the theory of change, the impact evaluation will attempt to test the following hypotheses. 1. Financial literacy: Beneficiaries who participate in the financial literacy training will improve their knowledge of household financial management, including budgeting techniques, the importance of saving, and knowledge of credit and savings options available to them. They are also expected to change their behavior in accordance with these principles, leading us to witness more frequent use of budgets and increased use of available savings mechanisms. 2. Business training: We expect beneficiaries who already have their own businesses to improve their knowledge of best practices in business finan- cial management, including budgeting, recording income and sales, paying themselves a salary, and keeping business and personal accounts separate. This change in knowledge should also lead to changes in related behaviors. We also expect some beneficiaries who do not already have a business to open one. 3. Job skills training: Beneficiaries who are not currently employed might lack “soft” job skills, such as self-esteem, life planning, time management, and decision making, that would help them find secure employment. We hypothesize that providing this training will increase the likelihood that beneficiaries will become employed and retain employment. 4. Different training methods: The impact of financial training might vary depending on who provides it. Financial literacy training can be done by professional trainers or community peers. While we expect professional trainers to be more effective in providing the training and increasing bene- ficiaries’ knowledge of financial management, being trained by peers might make beneficiaries more likely to change their behavior and implement the teaching tools they have learned. This is because peer trainers are part of the community, and beneficiaries should be able to use them as a resource during and after training. Furthermore, should the government choose to 426  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES scale up the program, it would be much more cost-effective if community peers can facilitate training as well as professionals. 5. Recordkeeping materials: Beneficiaries may be more likely to budget and keep records if they are given a tool to help them perform this activity. To this end, we can provide a notebook to keep records for those who received entrepreneurship training. We hypothesize that providing this notebook will increase the likelihood that beneficiaries will engage in recordkeeping and improve management of their finances. 13.3 DESIGN 13.3.1 Randomized controlled trials This evaluation uses a randomized controlled trial design to evaluate the impact of the training program. In a randomized controlled trial, one group of benefi- ciaries is randomly selected to participate in the training (the treatment group), while another is randomly selected not to receive the training at the time of the evaluation (the control group).1 Randomly assigning beneficiaries to treatment or control allows us to ensure that any differences observed between the two groups after the training program is over can be directly attributed to the training. If all beneficiaries or a nonrandom group of beneficiaries were invited to participate, those who chose to participate might systematically differ from those who did not (for instance, they might be more motivated or already have higher levels of financial management knowl- edge). Comparing these two self-selected groups would not allow us to identify the causal impact of the training program, because people with enough moti- vation to participate in the training program might have had different outcomes from the nonparticipating group due to their different characteristics. Random assignment ensures treatment and control groups are the same on average. Thus, the only difference between them can be attributed to the program. TREATMENT ASSIGNMENT Progresando con Solidaridad beneficiaries are assigned to a group (núcleo) of 45–60 members. Núcleos are groups of beneficiaries who attend meetings every three months on a variety of subjects such as health and education, and are assigned a community liaison, an enlace, that is responsible for communicating with and organizing the núcleo. The study will include 240 núcleos in the greater If the training proves successful, the government may wish to offer it to the control group 1  after the study period ends and scale up nationally to all CCT beneficiaries, so that they can also benefit from it. 13.  Increasing the impact of conditional cash transfers through financial literacy   ◾ 427 Santo Domingo area in the Dominican Republic with a total of 3,840 individuals. Núcleos will be selected from administrative data and randomly assigned to either the treatment or control group. The treatment group will include 120 núcleos, and the control will have 120. The original design based on 240 núcleos had to be rethought midway through the baseline survey, since beneficiaries were undergoing reassignment to new núcleos. The people originally selected were split up into multiple new núcleos in their area. This happened throughout our whole sample area and was mostly due to administrative restructuring, given staff changes and absorption of new beneficiaries. This made the núcleo structure irrelevant, since more than half the sample had already been surveyed but were spread into different, new núcleos. Since the beneficiaries still lived in the same places but were reorganized into new administrative groups, project leaders elected to change the treatment and control selection; instead of basing selection on groups, treatment and control would be randomly assigned on an individual basis. Since this restructuring was just an administrative change, the original design and plan for treatment implementation remains intact. Approximately 240 groups will be synthetically created based on geographic clusters and randomly assigned to treatment or control. That is, beneficiaries who live nearest each other will form a cluster, or training group, simply for implementation purposes. All members of the treatment group will receive financial literacy training intended to improve household financial management skills. Within this group, however, the treatment design is broken up into two layers of subtreatments, which will be assigned on an individual basis and grouped based on the nearest community training center. An additional treatment level will also be assigned at the individual level (budget notebooks). CLUSTER-LEVEL INTERVENTIONS Layer 1: Professional versus peer trainers. Of the approximately 120 clusters in the treatment group, 60 will receive financial literacy training from professional trainers, and 60 from peer trainers (figure 13.1). This will allow us to test whether being trained by peers is more or less effective than being trained by professional trainers. Layer 2: Business training versus job skills training. Half of the clusters in both the professional and peer training groups will receive an additional training session on financial management for businesses. The other half of the clusters in each group will receive additional training on job skills and the job search process. This allows us to test whether receiving small business administration training leads to improved business outcomes and whether receiving soft job skills training increases employment of beneficiaries. We can also compare the economic outcomes of beneficiaries who received each type of training. 428  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 13.1  Training delivery Total sample 240 clusters Financial literacy Control 120 clusters 120 clusters Peers Professional 60 clusters 60 clusters Business Job skills Business Job skills 30 clusters 30 clusters 30 clusters 30 clusters INDIVIDUAL-LEVEL INTERVENTIONS The feasibility and sample size of the interventions that will be randomized at the individual level will be determined at a later time. Furthermore, the control group may include individuals who receive the credit only and a pure control. Layer 3: Budgeting notebooks. Within each cluster in the treatment group, a random subset of beneficiaries will be selected to receive notebooks that can be used to keep household or business budgets. Of the sample of 3,840 individuals across all clusters, a subset to be determined will receive notebooks. This will allow us to test whether receiving a notebook to aid with the implementation of financial literacy training increases the impact of the training. 13.3.2 Implementation and monitoring The Social Cabinet of the government of the Dominican Republic and Innovation for Poverty Action contracted a local nogovernmental organization, the Domin- ican Institute for Integral Development, to develop the educational material in conjunction with Junior Achievement Dominicana. Since the Dominican Institute for Integral Development has experience working with vulnerable populations in the country, it was best suited to implement the training as well, and will be contracted for that phase. Implementation will be carried out through collabora- tion with operations of Progresando con Solidaridad, which will need to help coor- dinate the community liaisons to invite beneficiaries to the training. The Social Cabinet and Innovation for Poverty Action will monitor training for quality assur- ance and support logistics. 13.  Increasing the impact of conditional cash transfers through financial literacy   ◾ 429 The Dominican Institute for Integral Development was contracted for the material development, after which the finished product was sent for review by Freedom from Hunger, a reputable international organization experienced in developing financial education material. Freedom from Hunger was contracted to review and finalize the material, as well as train the Dominican Institute for Integral Development’s trainers to ensure quality training of all facilitators, both community leaders and professionals. The financial education material will consist of five modules: 1. Entrepreneurship: making a business plan, small business administration, and keeping a business budget 2. Job skills: steps needed to find and keep a job, including how to write a curriculum vitae and have a successful interview 3. Credit: tools and information for choosing between credit products and learning about formal lending institutions, as well as planning for capacity to repay loans 4. Savings: tools and information for choosing between savings products and formal and informal savings options, as well as making a savings plan 5. Budgeting: understanding the usefulness of a household budget and steps for making one 13.4 SAMPLE The sample was randomly selected from among four municipalities that fall within two regions of Santo Domingo. These municipalities are Distrito Nacional, in the Distrito Nacional region; and Santo Domingo Norte, Los Alcarrizos, and La Victoria, which are in the Santo Domingo region. All of these are within the urbanized area of greater Santo Domingo, although some parts are peri-urban and in some households there may have been more rural than urban characteristics. Prior to randomly selecting núcleos, the selection was filtered for those núcleos with greater than 25 people to conservatively account for beneficiaries who may have moved and to include substitute invitees. It was also filtered for households categorized as Type II (moderate) poverty. The evaluation targets beneficiaries in moderate rather than extreme poverty because of the graduation aspect of the hypothesized impact. Financial literacy training may give these beneficiaries an increase in income, along with access to credit, savings, and improved budgeting that will improve their financial situation enough for them to no longer rely on the program. The age for individuals was also capped at 55, under the assumption that beneficiaries much older than that may not benefit as much from financial literacy training. This is based on the assumption that those 55 and older will most likely retire sooner, and their financial literacy training would not be as useful since they are less likely to begin a new business or look for a new job. 430  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Of 1,763 núcleos that were in the sample area before selection, 240 were selected at random to be surveyed, as well as 16 people within those núcleos, plus 4 substitutes. Throughout the course of the baseline, three more substitutes were added in each núcleo to address the issue of occasional poor attendance at survey meetings. Additionally, 28 núcleos were replaced because of administra- tive constraints in working with those groups (lack of staff). During the project’s planning phase, power calculations were used to estimate the minimum sample size necessary to detect the desired effect on the key indica- tors. Since treatment levels are implemented at the group level, the impact (effect size) will be measured at the group level. Power calculations were done using núcleos as groups, and are still valid even though we are no longer using actual núcleos as clusters, since the new clusters are from within the same communities. Recent evidence on the impact of financial literacy training, business admin- istration, and job skills on knowledge and behavior reveals considerable variation of the effect size of these programs.2 We conducted power calculations to iden- tify the minimum sample size that would allow us to detect changes in the key outcomes of interest for the financial education intervention. With an estimated effect size of 0.15 and an intra-class correlation of 0.05, a sample of 240 núcleos with 15 beneficiaries per núcleo was selected in order to reach the desired power of 0.8 for the three treatment levels.3 In other words, we will detect an expected 15 percent increase in the number of beneficiaries displaying stronger financial management skills as a result of this intervention. 13.5 MEASUREMENT 13.5.1 Baseline survey Baseline surveying is conducted at scheduled meetings with selected núcleos. The meetings are organized in coordination with existing Progresando con 2  Carpeno et al. (2011) found the treatment group 5 percent more likely to understand the concept of a family budget, 17 percent more likely to know the minimum requirements for opening a savings account, and 20 percent more likely to understand what a nonperforming loan is. Fischer, Drexler, and Schoar (2010) find that the treatment group is 11 percent more likely to keep personal and business accounts separate, 11  percent more likely to keep accounting records, 6 percent more likely to calculate income, and 6 percent more likely to have savings—although the latter is only marginally significant. 3  The estimated effect size is based on evidence from similar experimental studies and the minimum effect size necessary (as determined by the principal investigators) to justify the scaling up of the program, taking into account the cost of the intervention. The interclass correlation was estimated by educational level data (chosen as one of the determining vari- ables) from 2004 Progresando con Solidaridad data. Eighty percent power is the conven- tional power level in econometric studies. 13.  Increasing the impact of conditional cash transfers through financial literacy   ◾ 431 Solidaridad field staff, from regional coordinator to field supervisor to community liaisons (enlaces), who are the direct organizers of beneficiaries. The enlace is responsible for communicating with heads of households to inform them about the survey meeting and, when implementation of the training begins, the financial literacy training sessions. The baseline survey is divided into the following sections of topics: ◾◾ Poverty-level information, including the education level of every member of the household, household expenses (as a proxy for income), and the income-generating activities of each member of the household ◾◾ Budgeting behavior ◾◾ Work experience, including strategies for job seeking ◾◾ Business information for those who have small businesses (including sales estimates, recordkeeping habits, business assets, etc.) ◾◾ Credit and savings (including how much and how often beneficiaries save, the amount and prevalence of loans, as well as questions to gauge knowl- edge of formal credit and savings options) ◾◾ Financial knowledge (knowledge of budgets, recordkeeping, prior financial training, as well as basic arithmetic and risk aversion) ◾◾ Dependency questions, both in terms of decision making and reliance on others for income ◾◾ Personality questions to gauge self-confidence and entrepreneurial spirit Only Progresando con Solidaridad beneficiaries (who are the heads of their households), or their spouses/partners were interviewed in the baseline; this is because beneficiaries, as the heads of their household, are the primary breadwin- ners and thus the best situated person in the household to apply the skills learned in financial literacy training. Children of beneficiaries or other household members are not necessarily in a position to take full advantage of financial literacy training, nor do they know enough about the income-generating activities of the head of household to provide the data necessary. 13.5.2 Endline survey The endline survey will measure the impacts of the training program one year after the program’s implementation. The survey will include everyone in the treatment and control groups, including those who were invited to training but did not attend, to capture the intend-to-treat effect. This was added to the design since it became apparent that the take-up rate for training may be lower than anticipated. Many beneficiaries did not attend baseline survey meetings; despite inviting 7 substitutes, in some cases fewer than the targeted 16 attended. Since beneficiaries move often, the endline survey will require surveyors to leave the survey area and conduct some surveys at beneficiaries’ homes when they do not attend the meetings. 432  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 13.5.3 Survey staff A total of 40 individuals were hired to implement the survey: 32 surveyors, 4 supervisors, and 4 auditors. These were divided into four teams, comprising one supervisor, eight surveyors, and one auditor. Auditors work separately. Each team member was selected from a database of previous government surveyors. After interviewing potential candidates, about 48 were invited to surveyor training; of those, the 40 most qualified were selected according to experience and perfor- mance during training. The supervisors were trained and selected first in order to assist in training the surveyor candidates. Both training sessions included pilot sessions of the survey at núcleos that were not part of the sample area. The survey team received training on the randomized controlled trial methodology and survey implementa- tion, and was carefully instructed in the protocols of the questionnaire designed for this study. The procedure for the endline survey will be the same. 13.5.4 Computer-assisted interviewing Computer-assisted interviewing is preferable to paper-based surveying in that there is a quicker turnover of data, since the data can be immediately exported electronically and do not require data entry, and there is a smaller risk of error due to programmable validity and logic checks in most computer-assisted inter- viewing software. We decided to use computer-assisted interviewing for this project for the following reasons: (1) the government offered to lend 40 laptops for this purpose, free of charge; and (2) it eliminated the need for data entry staff and time. The platform used for the baseline was Surveybe, chosen because of its ease in programming and because of its compatibility with the borrowed equip- ment. For the endline survey, these laptops may not be available to us, so we are considering using different survey software and purchasing tablets, since they are more mobile. 13.6 DATA QUALITY CONTROL 13.6.1 Confidentiality All data that can be used to identify a participant such as name, address, and phone number will be kept confidential and will only be accessed by key project staff. A unique ID code was assigned to each respondent, which will be used for data analysis, dropping all other personally identifiable information. Furthermore, before each interview, the surveyor reads a consent script, which is printed on the consent forms that all respondents must sign. This script informs respondents that data are anonymous and in no way affect their status as a beneficiary. 13.  Increasing the impact of conditional cash transfers through financial literacy   ◾ 433 13.6.2 Survey monitoring To ensure data quality, surveyors in each team are monitored by their super- visor as they collect data. This is easily done since survey teams collect data from an entire núcleo as meetings are held in a single site. Innovation for Poverty Action project staff also conduct unannounced monitoring visits to survey teams. In addition, field audits (also called back checks) are conducted with a subset of around 15 percent of beneficiaries, who are randomly selected by a surveyor, to ensure we audit all surveyors. In a field audit, the respondents are visited by an auditor, who asks them a few questions from the questionnaire. This provides us with a second set of responses to check the quality of the surveyors and the reliability of the data. If the field audit reveals discrepancies, the surveyor may be asked to reinterview the respondent. There have been no cases in which the respondent was not the same or the data were fabricated, although there are some higher than normal discrep- ancy rates on some variables. For example, there is a high discrepancy rate for answers that are estimates of amounts spent or earned. The question “How much do you estimate you spend weekly on food for the household?” has an 80 percent discrepancy rate, both small and large differences. In practice, many beneficiaries have trouble estimating this expense in weekly terms, so their estimates differ when asked the same question a few days later in the back check survey. There are discrepancies that are due to surveyor error, and those, when detected, are discussed with surveyors individually or as a group if the error is common. The question “How many economic activities do you have?” had a discrepancy rate of 23  percent. This is problematic, because in some of those cases the respondent reported that he or she had his or her own business in the back check, and in the original survey reported that he or she did not. This is most likely because respondents did not understand the script the surveyor reads to describe what we mean by economic activity, and the surveyor did not clarify. Innovation for Poverty Action project staff address these issues as they are gleaned from back check data throughout the survey. In an effort to reduce error rates and encourage higher-quality data collection for the endline survey, we will introduce an incentive system whereby surveyors and survey teams with the fewest errors are rewarded with a bonus in compen- sation and those with the highest errors are penalized. Surveyors will also receive incentives for improved performance. 13.7 BASELINE RESULTS The baseline survey was conducted between February and June 2013. A total of 3,280 beneficiaries were surveyed. The following summary statistics from the baseline data include a sample from the 237 questions on the survey, chosen 434  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES because they reflect key outcome variables we expect to be affected by the finan- cial literacy pilot program. 13.7.1 Income The amount spent weekly on food we are using as a proxy for income, since many beneficiaries are unemployed or have informal small businesses and may not be able to accurately express how much they earn or receive as income. The average spent on food weekly is RD$2,695 (about US$65). Seven  percent of the sample could not estimate how much they spent on food weekly. 13.7.2 Jobs and businesses Many people supplement their main income with small businesses, which can be anything from selling food from the home to obtaining small amounts of merchandise and selling them to friends and neighbors from time to time. In light of this, there is a question on the survey that explains “economic activities” and asks the number of economic activities each person in the household engages in. An economic activity is defined as anything that generates income, whether formal or informal. Fifty-one percent of those surveyed do not participate in any economic activity, whether it is a formal job or self-employment. Of the other 49 percent that participate in at least one economic activity, 64 percent have sala- ried employment and the remaining 34 percent are self-employed small or micro- business owners or collaborators. Only 3 percent of business owners claimed to have formal (legally registered) businesses. Of those who have their own busi- ness, average sales in a normal week are RD$4,710, about US$115. 13.7.3 Budgeting In response to the question “Do you regularly (at least once a month) keep track of your expenses?,” 57 percent of responded that they do not. While we want to gather accurate information on the budgeting practices of the beneficiaries, many do not know or misunderstand what a budget is. They were asked a question that breaks down budgeting: “Do you ever think about the money you expect to get and money you expect to spend and then make a plan for what you will do with your money?” The data show that 63 percent of respondents do not budget. 13.7.4 Credit and savings An important component of the financial literacy training is the informed and conscientious take-up of savings and credit products. The majority of those surveyed (72 percent) do not hold any loans. Among those who currently hold a loan, 82 percent accessed formal sector loans. Finally, 22 percent stated that they have had problems paying back their loans. The average amount of the largest 13.  Increasing the impact of conditional cash transfers through financial literacy   ◾ 435 loan, either formal or informal, is RD$28,972 (US$706). Of those who already have a loan, 57 percent responded that they plan to take out another one. As for savings, 27 percent affirm that they save, whether with the formal or informal sector (12 percent have savings accounts at formal financial institutions). Furthermore, 22 percent save regularly every month. Mean monthly savings per participant are approximately RD$435 (US$11) but increase to RD$1,989 (US$49) when excluding participants who do not save on a monthly basis. 13.8 RESULTS AND POLICY RELEVANCY Results should be finalized in early or mid-2015, with a paper published by the end of that year. The data will be made publicly available on the American Economic Association Registry to ensure transparency and reproducibility of results. Should the results be positive, the Dominican Republic government is interested in scaling up the program—especially if peer facilitators are proven as effective as professional facilitators, since the program will be more cost-ef- fective that way. Although the results from this study are immediately relevant to the govern- ment of the Dominican Republic, they will also be of great interest to other countries implementing similar CCT or unconditional cash transfer programs (especially given the recent surge in discussion around these programs). As noted previously, there is little evidence on how to graduate beneficiaries from CCT programs. The characteristics of our population include low levels of financial literacy, poor knowledge of and access to formal financial services, high unem- ployment and underemployment, and high rates of informality among business owners. These characteristics are very similar to many people in middle- and low-income countries and can also be found in vulnerable segments of the popu- lation in high-income countries. This is the first impact evaluation of financial literacy and employment or business training of CCT beneficiaries. Governments that are thinking about graduation strategies from CCT or similar programs—or, more broadly, financial inclusion and better employment options and income for poor and vulnerable populations—will benefit from the unique data and results of this evaluation. 13.9 TIMELINE The project began mid-2012 and will be a 3.5-year project, including the time necessary for data analysis and publication after completion of fieldwork. The baseline survey took place from January to June 2013, and the financial literacy and labor market training will take place from February to June 2014. The timeline in figure 13.2 shows all completed and planned activities for this project. 436  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 13.2  Project timeline 2012 Activity Jan Feb Mar Apr May Jun Jul Aug Sept Oct Nov Dec Present research plan Develop survey instrument Define and select sample Planning for baseline survey 2013 Selection and training Baseline data collection Baseline data cleaning Development of training Assignment to treatment and control Planning for implementation 2014 Financial literacy training implementation Planning for endline survey 2013 Endline data collection Endline data cleaning Final reporting REFERENCES Attanasio, Orazio, Luis Carlos Gómez, Patricia Heredia, and Marcos Vera-Hernandez. 2005. “The Short-Term Impact of a Conditional Cash Subsidy on Child Health and Nutrition in Colombia.” Institute for Fiscal Studies. http://www.ifs.org.uk/edepo/rs_fam03.pdf. Baez, Javier E., and Adriana Camacho. 2011. “Assessing the Long-Term Effects of Conditional Cash Transfers on Human Capital: Evidence from Colombia.” Policy Research Working Paper 5681, World Bank, Washington, DC. http://elibrary.worldbank.org/doi/ pdf/10.1596/1813-9450-5681. Carpena, Fenella, Shawn Cole, Jeremy Shapiro, and Bilal Zia. 2011. ”Unpacking the Causal Chain of Financial Literacy.” Policy Research Working Paper 5798, World Bank, Washington, DC. Fischer, Greg, Alejandro Drexler, and Antoinette Schoar. 2010. “Keeping It Simple: Financial Literacy and Rules of Thumb.” Paper presented at the Centre for Economic Policy Research Development Economics Workshop, Barcelona, October 8–9. GCPS (Social Cabinet of the Government of the Dominican Republic). 2011. “Reporte Final— Estudio Cuantitativo de Mercado: Diagnostico Sobre el Nivel de Bancarización y Capacidades Micro-Financieras de los Beneficiarios Programa Solidaridad.” Gertler, Paul J., and Simone Boyce. 2003. “An Experiment in Incentive-Based Welfare: The Impact of PROGESA on Health in Mexico.” Royal Economic Society. http://ideas.repec. org/s/ecj/ac2003.html. CHAPTER 14 D irect deposit and commitments to save in Malawi XAVIER GINÉ 14.1 INTRODUCTION This project investigates innovative ways to address low levels of formal savings in developing countries. The study explores to what extent psycho- logical mechanisms can be leveraged to increase formal savings. In partic- ular, we study to what extent “mental accounting,” status quo bias, and aspirations can be targeted to help individuals match desired savings and expenditure patterns with actual behavior. The study is motivated by market failures and behavioral phenomena that limit the ability of poor households to save and invest in productive activities. Incomplete credit markets in developing countries force many to rely heavily on accumulating personal savings to finance investments. However, behav- ioral phenomena, such as self-control problems or limited aspirations, may lead to suboptimal savings and underinvestment relative to desired levels. The study has three parts, each targeting a different psychological concept. In the first part, respondents within 10 kilometers of a banking branch are offered a subsidized ordinary savings account. In addition, some respondents are encouraged to “label” the account for a specific purpose. Labeling is a form of mental accounting that can render funds assigned to one category (e.g., school fees) less fungible for expenditures in another cate- gory (e.g., unplanned expenditures on luxury items). The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 437 438  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES The second part of the study explores the effectiveness of financial literacy education and its potential to increase people’s aspirations by going through a detailed savings plan and showing respondents videos of successful savers from a similar background. This strategy may be effective if aspirations of the poor and their belief in their own capacity for change are depressed, creating a feedback of decreased effort and decreased expectations. The third part of the study tests the power of status quo bias by comparing the expenditure and savings behavior of two groups that receive a randomly allo- cated grant either in the form of cash or directly into their bank accounts. If a bias toward the status quo leads people to disproportionately follow the default, then depositing income into accounts by default may facilitate savings relative to receiving cash. The third intervention will also test whether time between the announcement of winning money and the receipt of the money affects respon- dents’ expenditure and savings decisions. Within the “cash” group and “direct deposit” group, respondents randomly receive the money immediately, one day, or eight days after prize announcement. Last year, the project conducted a baseline survey and simultaneous account offers, after having completed a household census as a sampling frame. The households were revisited for financial literacy sessions and sessions to increase aspirations. In the current stage of the project, a subset of the same households will receive cash grants and will be interviewed repeatedly to follow expenditure and savings patterns. In the final state, a follow-up survey with the entire sample will measure effects of the interventions on formal and informal savings, aspira- tions, income, and expenditures. 14.2 MOTIVATION Savings enable people to smooth consumption over time, to buy durable consump- tion goods, and to make use of profitable business and investment opportunities. Returns to capital in particular are high in developing countries (de Mel, McKenzie and Woodruff 2008; Duflo, Kremer, and Robinson 2008; Fafchamps et al. 2011), and savings are important to generate the required capital in developing coun- tries due to the lack of well-functioning local credit markets. Access to formal savings accounts in particular has the potential to increase savings, and recent empirical work suggests substantial benefits from formal savings accounts for overall welfare. Aportela (1999) uses data from an expansion of branches set up in post offices at the end of 1993. He finds that the expansion resulted in an average increase in savings rate of 3 to 5 percentage points, with higher effects (up to 7 percentage points) for low-income individuals compared to other low-income households located in towns without the expansion. Burgess and Pande (2005) find that a policy-driven expansion of rural banking reduced poverty in India, and provide suggestive evidence that deposit mobilization and 14.  Direct deposit and commitments to save in Malawi  ◾ 439 credit access were intermediating channels. Despite positive social effects, the program was discontinued in 2001 due to losses from defaults. Bruhn and Love (2009) examine the opening of bank branches in consumer durable stores in Mexico in 2002 and find an increase in the number of informal business owners by 7.6 percent, in total employment by 1.4 percent, and in average income by about 7 percent. The effects are concentrated among low0income households and in municipalities with lower preexisting bank penetration. Dupas and Robinson (2013a) offer ordinary savings accounts to Kenyan urban entrepreneurs, finding positive impacts on investment and income for women. Prina (2011) finds that random assignment of basic savings account access to households in Nepal leads to increases in financial assets and human capital investments. In this project, we focus on two sets of barriers to savings: self-control and low levels of aspiration. The former is discussed first. The latter follows further below. Common reasons for people to save less than they would like to are psycho- logical factors such as impatience (a strong preference for the present over the future) and issues of self-control (competing preferences that dictate different actions at different times). There is evidence from both developed and developing countries that self-aware individuals seek to limit their options in anticipation of future self-control problems. Ashraf, Karlan, and Yin (2006) investigate demand for and impacts of a commitment savings device in the Philippines and find that demand for such commitment devices is concentrated among women exhibiting present-biased time preferences. Duflo, Kremer, and Robinson (2011) find that offering a small, time-limited discount on fertilizer immediately after the harvest has an effect on fertilizer use comparable to that of much larger discounts offered later, around planting time. Giné et al. (2012) find that Malawian tobacco farmers with present-biased preferences are more likely to revise a plan about how to use future income, even when that plan is made under commitment. In this project, we test two interventions by addressing the psychological barriers described above—the first encourages mental accounting; the second targets status quo bias. 14.2.1 Mental accounting through account labeling Mental accounting, as introduced by Thaler (1990), describes the phenomenon that money is not always fungible across expenditure categories. Savings that are mentally assigned to a specific savings goal or set of expenditures, such as inputs for farming, may become less available for another set of expenditures, such as unplanned luxury purchases. Formal savings accounts, among other things, may help people address the above-described psychological phenomena of impa- tience and self-control by helping them perform mental accounting through phys- ically—not just mentally—separating funds. Recent empirical work provides evidence that mental accounting can help people increase savings. Dupas and Robinson (2013b) provide a number of simple 440  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES informal savings technologies, such as a simple metal lockbox, to members of existing savings groups in Ghana. The unifying feature shared by all the various savings technologies is that, while funds were technically fungible, they were designated as “savings for health.” Take-up was initially high and remained high after 33 months, and overall average savings for health increased substantially. A specific earmarking condition in which the type of earmarking was not chosen by participants was effective for emergency savings but ineffective for preventative health savings, suggesting that the type of labeling can matter. In this study, participants are randomly separated into two groups. Both groups receive account opening assistance and an offer for one fully subsidized account. In addition, members of one group are encouraged to label their newly opened accounts for a specific, individually chosen savings goal. The labeling is physically reinforced by writing the savings goal on the automated teller machine (ATM) card that comes with the new account. 14.2.2 The power of default options A robust finding of behavioral economics that uses insights from psychology is the disproportional impact a status quo has on decision making (Kahneman, Knetsch, and Thaler 1991; Samuelson and Zeckhauser 1988). Given a set of choices, people are more likely to choose a particular option if that option is framed as the default or status quo. Since people are relatively more averse to losses, the perceived loss of moving away from the status quo tends to outweigh potential benefits. The idea that defaults can matter in the context of savings has been shown by Thaler and Benartzi (2004) with the Save More Tomorrow program that offered employees the opportunity to precommit parts of future salary increases to retire- ment savings. Take-up of offers was high and remained high over time, and savings rates increased substantially for those who were offered the precommitment option. Status quo effects can be expected to be more important for the poor since the effects of bad decisions have relatively larger repercussions; at the same time, the poor are more likely to be affected by status quo effects, since limited attention is relatively scarcer given a larger number of concurrent urgent items that need to be handled (Mullainathan and Shafir 2009). In this study, some participants receive grants in the form of cash, while others receive grants directly deposited into their individual savings account. If status quo bias matters for savings in this context, then those who receive grants as direct deposits should have higher formal (and overall) savings, since cash will not be designated as savings and would be more likely to be spent. 14.2.3 Increasing aspirations A determinant in the decision making of the poor that recently is receiving more attention by economists in both theoretical (Dalton, Ghosal, and Mani 2011) and empirical work (Bernard, Dercon, and Taffesse 2011) is aspirations. Given external 14.  Direct deposit and commitments to save in Malawi  ◾ 441 circumstances, individuals need to believe that their actions can make a differ- ence in their lives and that change toward the better is possible. However, since aspirations and beliefs are more broadly at least in part determined by individ- uals’ local environment, initial conditions can matter with regard to how much effort is exerted. Low achievement in the immediate environment can trigger a negative feedback loop: lower effort—in expectation of low rewards—can lead to low individual achievement, thereby perpetuating low expectations about future cost-benefit ratios. If aspirations can be changed by external intervention, this could lead individuals to try harder and achieve outcomes that are in fact possible in a given environment, creating positive reinforcement. This study tests a cognitive intervention that targets respondent aspirations by showing individual respondents short inspirational videos in order to set a higher reference point for respondents to aim at and to increase their belief in achieving those goals. A control group is shown a neutral video instead. Recent pioneering empirical work by Bernard et al. (2012) uses a similar design in which a documentary was shown in a group setting in rural Ethiopia. The researchers find suggestive evidence of positive effects on education-related aspirations and on demand for credit that are still visible six months after the intervention. 14.3 CONTEXT OF THE STUDY The study takes place in rural and peri-urban areas in Malawi. The household sample will be chosen from the village population in a 10-kilometer radius surrounding the trading center of Chitakale in Mulanje, a prominent trading center in the Mulanje District in the southern region of Malawi. With two commercial banks, a major microfinance institution, a microcredit institution, and a chain grocery store, Chita- kale is a commercial hub frequented by residents of the local villages. The market in Chitakale runs every day, experiencing particularly high traffic on Fridays, when the largest number of informal vendors come and set up their sale items on tarps on the ground. Chitakale is located close to a number of large commercial tea estates, which form a major source of economic activity in the region. Chitakale is within 75 kilometers of Blantyre, Malawi’s commercial capital. The study partners with a national commercial bank, NBS, to implement the account intervention. NBS was originally established as the New Building Society in 1964 and transitioned to a full-service bank in 2004. The bank currently oper- ates 13 branches and 21 agencies throughout the country. The Mulanje NBS branch is located in the center of Chitakale’s market area. 14.4 HOUSEHOLD LISTING For the purpose of identifying a representative sample of the villages surrounding the Chitakale trading center, a household listing exercise was conducted. The 442  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES study sample includes households from five group villages. Each group village is represented by a group village headman that oversees the village chiefs. The total sample area consists of 14 villages ranging in size from about 50 house- holds at the low end to about 800 households at the other extreme. Villages are further divided into subvillages. A team of trained research assistants listed each household in these 14 villages, asking an adult member of the household for basic demographic and income information. The listing survey ran from the end of January to the end of February 2013, including a clean-up period during which a small team revisited each village to capture any households missed during the first visit. The listing exercise recorded a total of 6,089 household interviews within an approximate radius of 6–7 kilo- meters outside of Chitakale. Fewer than 200 village households (3 percent of the population listed) were unavailable for interview. Households in the urban centers were excluded, so the sample represents a peri-urban and rural population. Following are sample averages for key variables of the household listing. ◾◾ Household head demographics: Among interviewed households, 69 percent reported having a male head of household. The average age for all household heads is 42 years. Fifteen percent reported no formal educa- tion. Twenty-one  percent had completed some schooling between stan- dards one and four; 37 percent had completed some schooling between standards five and eight (end of primary school). Twenty-seven  percent of household heads had completed schooling beyond standard eight, meaning secondary or tertiary education. The survey did not ask respon- dents to specify a level beyond standard eight. ◾◾ Household members: Respondents reported an average of 4.4 household members. Household members are people who have regularly slept and taken meals in the home for the majority of the last three months, including nonrelatives and servants. Children under age 18 who are boarding at school but whose parent or guardian is a household member are also considered members. ◾◾ Income sources: Interview respondents were asked to list the most important sources of income (cash or in kind) from the last year, naming up to five sources. Forty-three percent reported self-employment or a small business as an important source; 33 percent reported piece labor for an income source; 23  percent sell pigeon peas for income, and 20  percent work an unskilled job for a private sector company. Other important sources of income mentioned were remittances (11 percent), selling maize (8 percent), working a skilled or professional private sector job (7 percent), selling cassava (5 percent), and working with an organization or association (5 percent). Respondents were asked to list the months in the last year in which each source of income was received. Seventy-nine percent reported having at least one source of income for each month of the last year. 14.  Direct deposit and commitments to save in Malawi  ◾ 443 ◾◾ Household assets: Based on interviewer observation, 60  percent of households interviewed had some iron roofing. Forty-nine  percent of households had some glass windows, and 27 percent of homes had at least one room with cement flooring. Interviewers read out a list of common assets, and respondents were asked to self-report all items that any house- hold members owns. Forty-six  percent of households own at least one bicycle; 53 percent own a radio; 55 percent own a table; 49 percent own a bed. Forty-four  percent own a mobile phone, although some reported not owning a SIM (subscriber identity module) card to be able to use it. Fifty-six percent of homes own one or more axes, and 28 percent possess an iron (for clothing). Regarding livestock, the survey inquired about chickens (55 percent ownership) and goats (17 percent). Households own an average of 1.2 acres of land, including the land on which the house sits. ◾◾ Household financing: The survey included questions about banking and informal savings use. In 20 percent of households, at least one member has a bank account. On average, households with accounts used the bank 17.5 times in the past year. Account use was defined as a withdrawal, deposit, transfer, direct deposit, or checking of the balance. Forty percent of house- holds had participated in an informal savings group in the past year. 14.5 STUDY DESIGN 14.5.1 Original study design The original study design included an evaluation of the introduction of direct deposits for a large agricultural firm that was going to switch workers from paying in cash to paying directly into individual bank accounts. The partner firm was cooperating with a microlending institution that had set up a small branch and housings for several ATMs. The research team had reached an agreement with the firm and the bank to randomize the roll-out of direct deposit for a pilot division of workers who had already been given accounts, and extend the randomization of the roll-out at the appropriate level for the remainder for the approximately 6,000 workers. The bank had already opened several thou- sand accounts for smallholder farmers in the area that were transacting with our partner firm. Roll-out of direct deposit for the same farmers was under way and has been fully completed in the meantime. Despite the seemingly fortuitous timing of the study, the established pres- ence of the bank, the introduction of direct deposit for smallholder farmers, and assurances as well as expectations by all parties involved about the imminent start, the roll-out was stalled and has not yet begun. The main reason for this is the bank’s lack of capacity (management, information technology system) to 444  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES take over smallholder loans from the firm, which constitute an integral part of the bank’s business proposition. Given the revealed capacity constraints of the bank and the possibility that it would eventually have to abandon the project on economic grounds, the research team decided to pursue the research with a different, geographically close population with independent access to banks, and to work with a different bank as a partner. 14.5.2 Overview of project stages The study sample of 872 households was drawn randomly from the households included in the listing exercise. The study—lasting approximately one year—will be conducted in five stages: (1) baseline and account offers, (2) financial literacy training and immediate follow-up survey, (3) short-term follow-up survey, (4) direct deposits and expenditure surveys, and (5) endline survey with optional account closures. Each stage consists of one or more individual household visits. STAGE 1: BASELINE SURVEY AND ACCOUNT OFFERS Each of the 872 households in the sample were asked to complete a baseline survey; they were offered a special promotional account from a national commer- cial bank that has a local branch in Chitakale. The baseline survey took approximately 90 minutes to complete. Ques- tion topics include demographics, expenditures, assets, sharing within social network, savings, savings aspirations, formal banking experience, financial knowledge, and agricultural activities. The account offer lasted approximately 30 minutes. Households received one of two offers: (1) an offer for a promotional savings account, or (2) an offer for a promotional savings account with a savings goal label. The respondents were offered the accounts at home, according to their assigned group, immediately after completing the baseline survey. The account script explained the benefits of savings accounts and the specific benefits of the promotional account. The research assistant walked the respondent through filling out the necessary account opening forms, organized the other documents required to bring to the bank, and helped the respondent formulate a plan for how and when to go to the bank to complete the account opening process. Respondents received a bank account voucher with their name, appointment time, and bank hours. The voucher was only valid for two weeks after the home visit to encourage participants to open an account soon without procrastination. STAGE 2: FINANCIAL LITERACY TRAINING AND IMMEDIATE FOLLOW-UP SURVEY Each of the 872 households in the study were randomly selected for one of three financial literacy trainings (one group receiving no training). A research assistant visited the house to complete an individual training session with the household 14.  Direct deposit and commitments to save in Malawi  ◾ 445 head; immediately following the training, the research assistant administered a short survey measuring financial literacy and savings aspirations. The training lasted approximately 0–100 minutes, depending on the treat- ment group assigned. The “info only” treatment watched a video about budgeting monthly income and choosing the right savings method for their needs. This group then completed personal worksheet exercises that demonstrate how to create a monthly budget and monthly savings plan. The “info plus aspirations” group completed the same training as the “info only” group; in addition, they watched a video about community members who were able to save and improve their households. This group also completed a worksheet with questions that asked participants to envision their life in the future when they have reached their savings goal. The “control” group did not receive any training. Following the training, the participants completed a 20-minute survey including topics such as financial knowledge and savings aspirations. Modules were similar to those administered at baseline. STAGE 3: SHORT-TERM FOLLOW-UP SURVEY The short-term follow-up survey was administered two weeks after the financial literacy training and immediate follow-up survey. This survey took approximately 30 minutes to complete, and was similar to the immediate follow-up survey with question topics including financial knowledge and savings aspirations. STAGE 4: DIRECT DEPOSIT AND EXPENDITURE SURVEYS Enumerators will visit a subset of 600 study participants, all of whom have savings accounts with our partner bank. The household head will be asked to complete an expenditure survey; after the survey, some respondents will be offered a voucher to receive a monetary prize. The respondent must travel to the bank in two days to present the voucher and determine the prize. A research assistant will return one and two weeks later, after the prize has been announced, and administer the same expenditure survey. The three rounds of surveys and prizes will occur over two months. Six hundred households with savings accounts will be chosen from the greater study sample to win a monetary prize of approximately US$60. Money will be delivered in one of two ways, with the method to be chosen at random: (1) cash, to be picked up from a location in Chitakale; or (2) directly deposited into the winner’s savings account. The timing of the disbursement will also be randomized: (1) immediately following prize announcement, (2) one day after prize announcement, or (3) eight days after prize announcement. The expenditure survey will be administered before the prize is announced and again one and two weeks after. The research assistant will ask the participant to detail all household expenditures from the last week. The survey is set to last approximately 45 minutes. 446  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES STAGE 5: ENDLINE SURVEY WITH OPTIONAL ACCOUNT CLOSURE The endline survey will be administered approximately one year after the base- line survey. Survey administration will take approximately 60 minutes; the topics are similar to those at baseline, including demographics, expenditures, assets, sharing within social network, savings, savings aspirations, formal banking expe- rience, financial knowledge, and agricultural activities. Those respondents who opened a promotional bank account at the begin- ning of the study will be reminded that the promotional aspects of the account will expire at the end of the month. They will be informed of the fees that will take effect next month, and will be given assistance in closing the account if they wish to do so. The discussion will last approximately 25 minutes, depending on the action the participant decides to take. 14.6 INTERVENTION DETAILS 14.6.1 Savings accounts All study participants were offered a subsidized savings account with the specifications listed below. In addition to the account offer, half of the sample (436 households) were randomly chosen to receive the labeling treatment. In the labeling treatment, the study participant was encouraged to label the subsidized account with a personal savings goal. The participant could choose any goal label, with no restrictions except word length. The label appears on the participant’s ATM card as a sticker with wording such as “saving for house” or “saving for school fees.” The ATM card label is intended as a regular reminder to the partic- ipant of his or her savings goal, potentially reducing the likelihood that he or she will use the account funds for other expenditures. Labeling could increase savings and the ability of participants to reach their savings goal. The accounts offered have the following characteristics: ◾◾ Control group: regular savings account —— Minimum balance —— No initial deposit —— No monthly fee for the first 12 months after account opening —— No withdrawal fee for the first 12 months after account opening —— Some interest earned on balance —— ATM card included ◾◾ Treatment group: labeled savings account —— All the features mentioned above for regular savings account —— ATM card has a label that mentions the owner’s chosen savings goal (a sticker saying e.g., “saving for house” or “saving for fertilizer”) 14.  Direct deposit and commitments to save in Malawi  ◾ 447 14.6.2 Account opening During the household visit in which the baseline survey was administered, the research assistant also offered the subsidized accounts. The research assis- tant read from a script that explained the special offer and how to complete the account opening process. The research assistant helped the respondent fill out the bank account opening forms, if he or she wished. The research assistant also worked with the respondent to identify how and when to travel to the bank to complete the account opening process, and made an appointment for the spec- ified time. Two research assistants were stationed at the local bank branch to assist study participants with opening their savings accounts. If participants brought all the necessary documentation, they were able to open an account that day. Clients received their account number immediately and an ATM card during a subsequent bank visit two to four weeks later. Study participants could choose to open the account as a joint account. At the time of account opening, both account owners had to be present. The joint account owners then decided whether one or both needed to be present to with- draw money. 14.6.3 Financial literacy training Study participants were randomly assigned to one of three financial literacy training groups, listed below. They received a one-on-one training session in their household with a research assistant. TREATMENT 1: INFORMATIONAL TRAINING The study participants watched a video explaining financial concepts, such as common banking fees, interest rates, budgeting, and choosing the best savings method for one’s needs. At the conclusion of the session, the research assis- tant and participant reviewed a sample savings goal plan and completed a corre- sponding worksheet. This informational training was meant to increase participant knowledge of financial concepts, savings methods, and budgeting. The materials for the training were based on a locally developed financial literacy training from a microfinance bank. The training was adapted for time length and research purposes; the resulting material was piloted several times and adjusted to improve comprehension. TREATMENT 2: INFORMATIONAL AND ASPIRATIONAL TRAINING In addition to the informational training described above, the aspirational training included a second video that featured five local residents—Fred, Bornface, Janet, Mac, and Dickson—who have worked hard to earn income and have used the profits to improve their lives. The stories are true, describing real hardships, 448  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES how the farmers or businesspeople overcame them, and the positive impact of increased wealth on their livelihoods. The second activity to increase aspira- tions was an additional worksheet that asked participants to imagine their main savings goal in detail, how it would improve their lives, and how it could increase their social standing in the family and wider community. This training was meant to increase participants’ aspirations for the future. CONTROL GROUP: NO TRAINING This group watched a video about the benefits of family planning. They did not receive informational or aspirational financial literacy training. 14.6.4 Direct deposit grants Among the study households with a savings accounts (promotional or existing), 600 will be randomly chosen to participate in a study of direct deposit. All 600 households will be visited by a research assistant, who will present a voucher to earn a cash prize. The participant will need to present the voucher to a research assistant in Chitakale two days later. The participant will choose a token from a bag to deter- mine the prize (or treatment group): ◾◾ Treatment Group 1: cash prize immediately. The research assistants will inform 80 households that they have won MK 25,000 (approximately US$60) and will receive the cash immediately. ◾◾ Treatment Group 2: direct deposit immediately. The research assis- tants will inform 80 households that they have won MK 25,000 and will receive the money immediately into their savings account. ◾◾ Treatment Group 3: cash prize + 1 day. The research assistants will inform 80 households that they have won MK 25,000 and will need to return tomorrow to pick up the cash. ◾◾ Treatment Group 4: direct deposit + 1 day. The research assistants will inform 80 households that they have won MK 25,000 and will receive the money tomorrow directly into their savings account. ◾◾ Treatment Group 5: cash prize + 8 days. The research assistants will inform 80 households that they have won MK 25,000 and will need to return in eight days to pick up the cash. ◾◾ Treatment Group 6: direct deposit + 8 days. The research assistants will inform 80 households that they have won MK 25,000 and will receive the money directly into their savings account in eight days. ◾◾ Control group. The research assistants will inform 120 households that they have won MK 1,000 (approximately US$3) and will receive the cash immediately. 14.  Direct deposit and commitments to save in Malawi  ◾ 449 14.7 ADDITIONAL DATA ANALYSIS OF HOUSEHOLD LISTING: CHARACTERISTICS OF BANK ACCOUNT HOLDERS Among the households surveyed as part of the village listings, 80  percent do not hold a bank account. Among the bank account holders, we categorize “infre- quent bank account users” as those who use their bank account less than once a month over the last 12 months (roughly half of all account users); we catego- rize the others as “frequent bank account users.” Comparing basic demographic and income characteristics across bank account usage status shows that bank account holders were more often male. Thirty-four percent of those who do not hold an account were women, compared to 20 percent of account holders. Bank account ownership and frequency of use do not vary substantially with the age of the respondent: infrequent bank account users had the lowest average age (39.4), slightly younger than frequent account users (40.2 years) and those with no account (42.2 years). Bank account holders are more educated than those without an account. Those who did not hold an account had, on average, 5 years of school; in comparison infrequent and frequent account users averaged 7.6 and 8.1 years of school, respectively. Participation in a rotating savings and credit association (ROSCA) over the last year was most pronounced among frequent bank account users (60 percent); participation was slightly less among infrequent account users (58 percent), and much less among those who do not hold accounts (36 percent). The types of household income source vary by bank account ownership. Those who did not hold bank accounts were more likely to mention income through piece labor work (37 percent), unskilled work in the private sector (23 percent), and remittances (12 percent). Account holders—both frequent and infrequent— were more likely to mention income through crop sales (28 and 27  percent, respectively), self-employment (57  percent of infrequent users), skilled work in the private sector (13 and 18  percent, respectively), and work with nongovern- mental organizations or associations (7 and 8 percent, respectively). REFERENCES Aportela, Fernando. 1999. “Effects of Financial Access on Savings by Low-Income People.” Banco de México. http://www.lacea.org/meeting2000/FernandoAportela.pdf. Ashraf, Nava, Dean Karlan, and Wesley Yin. 2006. “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines.” Quarterly Journal of Economics 121 (2): 635–72. Bernard, Tanguy, Stefan Dercon, and Alemayehu Seyoum Taffesse. 2011. “Beyond Fatalism: An Empirical Exploration of Self-Efficacy and Aspirations Failure in Ethiopia.” IFPRI Discussion Paper 01101. International Food Policy Research Institute. http://www.ifpri. org/sites/default/files/publications/ifpridp01101.pdf. Bernard, Tanguy, Stefan Dercon, Kate Orkin, and Alemayehu Seyoum Taffesse. 2012. “Aspirations and Well-Being Outcomes in Ethiopia: Evidence from a Randomized 450  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Experiment.” Presentation at the CSAE Conference 2012, Economic Development in Africa, Center for the Study of African Economies, Oxford, March 18–20. Bruhn, Miriam, and Inessa Love. 2009. “The Economic Impact of Banking the Unbanked: Evidence from Mexico.” Policy Research Working Paper 4981, World Bank, Washington, DC. Burgess, Robin, and Rohini Pande. 2005. “Do Rural Banks Matter? Evidence from the Indian Social Banking Experiment.” American Economic Review 95 (3): 780–95. Dalton, Patricio S., Sayantan Ghosal, and Anandi Mani. 2011. “Poverty and Aspirations Failure.” CentER Discussion Paper Series No. 2011-124. http://papers.ssrn.com/sol3/ papers.cfm?abstract_id=1966402. de Mel, Suresh, David McKenzie, and Christopher Woodruff. 2008. “Returns to Capital in Microenterprises: Evidence from a Field Experiment.” Quarterly Journal of Economics, 123 (4): 1329–72. Duflo, Esther, Michael Kremer, and Jonathan Robinson. 2008. “How High Are Rates of Return to Fertilizer? Evidence from Field Experiments in Kenya.” American Economic Review Papers and Proceedings 98 (2): 482–88. Dupas, Pascaline, and Jonathan Robinson. 2013a. “Savings Constraints and Microenterprise Development: Evidence from a Field Experiment in Kenya.” American Economic Journal: Applied Economics 5 (1): 163–92. —. 2013b. “Why Don’t the Poor Save More? Evidence from Health Savings Experiments.” American Economic Review 103 (4): 1138–71. Fafchamps, Marcel, David McKenzie, Simon R. Quinn, and Christopher Woodruff. 2011. “When Is Capital Enough to Get Female Microenterprises Growing? Evidence from a Randomized Experiment in Ghana.” NBER Working Paper 17207, National Bureau of Economic Research, Cambridge, MA. Giné, Xavier, Jessica Goldberg, Dan Silverman, and Dean Yang. 2012. “Revising Commitments: Field Evidence on the Adjustment of Prior Choices.” NBER Working Paper 18065, National Bureau of Economic Research, Cambridge, MA. Kahneman, Daniel, Jack L. Knetsch, and Richard H. Thaler. 1991. “Anomalies: The Endowment Effect, Loss Aversion, and Status Quo Bias.” Journal of Economic Perspectives 5 (1): 193–206. Mullainathan, S., and E. Shafir. 2009. “Savings Policy and Decision-Making in Low-Income Households.” In Insufficient Funds: Savings, Assets, Credit, and Banking among Low-Income Households, edited by R.  M. Blank and M.  S. Barr, 121–45. New York: Russell Sage Foundation Publications. Prina, Silvia. 2011. “Do Basic Savings Accounts Help the Poor to Save? Evidence from a Field Experiment in Nepal.” http://www.aeaweb.org/aea/2012conference/program/ retrieve.php?pdfid=555. Samuelson, William, and Richard Zeckhauser. 1988. “Status Quo Bias in Decision Making.” Journal of Risk and Uncertainty 1 (1): 7–59. Thaler, Richard H. 1990. “Anomalies: Saving, Fungibility, and Mental Accounts.” Journal of Economic Perspectives 4 (1): 193–205. Thaler, Richard H., and Shlomo Benartzi. 2004. “Save More TomorrowTM: Using Behavioral Economics to Increase Employee Saving.” Journal of Political Economy 112 (S1): S165–87. CHAPTER 15 E valution of Old Mutual’s On the Money program Financial literacy in South Africa SHAWN COLE, BILAL ZIA, MARTIN ABEL, LUCAS CROWLEY, CHRISTIAN SALAS PAULIAC, AND VERONICA POSTAL 15.1 PROJECT BACKGROUND AND SCOPE The last few decades have seen a tremendous expansion of the availability of financial services to millions of individuals previously deemed “unbankable.” Microfinance and microinsurance institutions have grown dramatically and have now entered mainstream finance. Innovations in financial products and service delivery methods now make it possible to offer affordable products even to very poor or remote individuals. These developments have success- fully provided millions of households with new opportunities to save, borrow, and manage liquidity and risk. While this expansion in financial access has the potential to significantly increase the welfare of consumers in developing economies, several challenges lie ahead. The new client base is largely comprised of poorer, less educated individ- uals than the traditional consumers of financial services. This poses the ques- tion of whether consumers will be able to make complex financial decisions or if low financial literacy will limit the spread of financial products to the less educated segments of the population. Moreover, even among the fraction of the population already taking part in the financial system, behaviors that significantly depart from those predicted by standard economic theory are prevalent. These departures raise the concern that disadvantaged individuals may fail to realize the full benefits generated by financial access because they are ill prepared to make complex financial decisions. The authors thank the Russia Financial Literacy and Education Trust Fund for financing this work. All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 451 452  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES There has been a growing interest among practitioners in financial educa- tion courses to mitigate these issues. These programs have the objective of illustrating the costs and benefits of savings, credit, and insurance, as well as familiarizing individuals with financial products. Financial education is expected to help individuals become more comfortable with financial institutions and help them amend behaviors detrimental to their financial security. The underlying assumption is that suboptimal behaviors—such as insufficient savings, exces- sive amounts of debt, use of expensive sources of credit when cheaper sources are available, and insufficient adoption of insurance—are due to consumers’ lack of financial knowledge. The provision of financial education would correct this market failure, allowing mutually beneficial transactions between consumers and financial institutions to occur. To this end, several governments, firms, and international organizations around the world have started funding financial education programs. These proj- ects aim to improve people’s financial skills, with the ultimate goal of reducing inequality by giving individuals the opportunity to benefit from financial access. For example, in 2004, the government of South Africa adopted the Financial Sector Charter, which requires financial institutions to contribute 0.2 percent of after-tax profits to financial education. Old Mutual—an international investment, savings, insurance, and banking group operating in South Africa since 1845— dedicated R 8.69 million to its financial education program in 2011 alone, training almost 36,000 individuals. This considerable outlay had the objective of improving people’s financial skills, helping them make sounder financial decisions, and encouraging saving. However, there is little substantive evidence on the effective- ness of these financial education programs. While low levels of financial literacy seem to be correlated with worse financial outcomes, it is unclear if improve- ments in financial knowledge will result in better financial choices. This study adds to the growing body of evidence analyzing the relation- ship between levels of financial literacy and poor financial choices and robust evidence on the effects of financial education on financial knowledge, behavior, and outcomes. The evaluation sheds light on the channels through which financial education is assimilated and internalized. The results of this study will help Old Mutual improve the effectiveness of its financial education program and provide valuable insights for financial education programs elsewhere. 15.1.1 Purpose and scope of the evaluation Financial access is widely recognized as being crucial to development. Thanks to innovations in products and technology, financial access has dramatically increased over the last few decades. Now more than ever, people in developing countries have the possibility of participating in the formal and semiformal finan- cial system. For instance, microfinance and microinsurance products are popular among millions of lower-income households. Despite this success, many questions 15.  Evalution of Old Mutual’s On the Money program  ◾ 453 remain unanswered with regard to individuals and their behavior vis-à-vis the financial system. While the benefits of savings, borrowing, and insurance products seem unde- niable, a large share of the world population does not take advantage of these opportunities. Chaia et al. (2009) report that 5.5 billion people—over half of the world’s adult population—does not save or borrow through formal or semiformal financial institutions. While part of this phenomenon is largely explained by a lack of access, take-up is relatively low even when formal financial institutions are available. Karlan, Morduch, and Mullainathan (2010) note that financial participa- tion is especially low in Sub-Saharan Africa, averaging at about 20 percent of the population. While supply factors partly explain low take-up of financial services, they point out that demand-side factors should not be dismissed as important explanations. Low demand for beneficial financial products and services is only one of the many divergences from the behaviors predicted by the neoclassical economic model of life-cycle consumption and savings. There is vast evidence attesting to the fact that individuals undersave, borrow excessively, or borrow at an excessive interest rate. Although no single explanation can account for such large depar- tures from the predictions of rational economic models, several authors have imputed part of the responsibility for these occurrences to imperfect information (Lusardi 2008) and to behavioral biases such as hyperbolic discounting (Laibson 1997), lack of trust (Guiso, Sapienza, and Zingales 2008), and lack of understanding of compound interest rates (Lusardi and Tufano 2009). Financial education has often been viewed as a way to solve the problems created by behavioral biases and imperfect information. Low levels of financial literacy are correlated with poor financial outcomes (Cole, Sampson, and Zia 2011), and several studies find that the vast majority of the population struggles with basic financial concepts in developed countries and developing countries alike (Lusardi and Mitchell 2007; Lusardi and Tufano 2009; Cole, Sampson, and Zia 2011). In an attempt to mitigate the problems arising from this disconnect, there has been growing attention over the last few years to programs aimed at promoting financial education. Financial literacy has been shown to be closely associated with portfolio diversification (Guiso and Jappelli 2008), retirement planning (Lusardi and Mitchell 2009), the use of lower-cost sources of credit (Lusardi and Tufano 2009), and stock market participation (Van Rooij, Lusardi, and Alessie 2007). Similarly, financially illiterate households have a tendency to underestimate the cost of compound interest rates (Stango and Zinman 2009) and fail to refinance mortgages at more favorable conditions (Campbell 2006). Financial literacy might also have more far-reaching implications for financial inclusion, since several studies find that vulnerable groups such as minorities, women, youth, and less educated indi- viduals fare much worse with respect to financial knowledge than the average individual (Lusardi and Mitchell 2008; Lusardi, Mitchell, and Curto 2010). Despite 454  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES these findings, it is unclear whether increasing financial knowledge is the most effective means of improving financial outcomes. Practitioners have viewed financial education as a way to increase savings, asset accumulation, and stock market participation, and to improve credit management. It is clear that low financial literacy prevents individuals from making optimal decisions concerning consumption, asset allocation, and debt management, but it is yet to be established if an intervention aimed at increasing financial education would be successful in positively affecting these variables. For instance, Cole, Sampson, and Zia (2011) conducted a field experiment in Indo- nesia, finding that despite the correlation between financial literacy and finan- cial access, an exogenously assigned financial literacy program has only modest effects on financial participation. Caution should be used when interpreting the results, however, and varying programs and teaching methods may result in significantly different outcomes. For example, Fischer, Drexler, and Schoar (2010) obtained positive results on financial management when offering people a rules-of-thumb-based education rather than a more comprehensive approach, while Karlan and Valdivia (2011) found that short entrepreneurship training sessions over an extended period of time were successful at improving business practices in Peru. Thus, while financial education may appear to be an important remedy for the problems of low financial participation and suboptimal financial behavior, reality presents a much more varied picture. Despite the positive correlation between financial literacy and financial outcome, the causal relationship between the two is unclear. This chapter aims to create a better understanding of whether improving financial literacy through an educational program would constitute a viable conduit to improve financial outcomes. 15.1.2 Main research questions Existing evidence on the effect of financial literacy is generally inconclusive, largely because existing studies fail to establish a causal link or suffer from limited data. To address these shortcomings, our study measures the causal effects of finan- cial education on financial behaviors and comprehensively examines the scope and pathways of these effects. To establish causality, this study incorporates random assignment of burial society support plan (BSSP) groups and borrowing groups of Women’s Development Businesses (WDB) organizations to receive Old Mutual’s On the Money financial literacy training. This procedure ensures that those who receive training are statistically indistinguishable from those who do not, and that any observable difference between the groups (in terms of saving, borrowing, etc.) after the intervention was implemented can be attributed solely to the training program. Both the treatment and the control groups participated in baseline and endline surveys so as to detail the scope and pathways by which financial 15.  Evalution of Old Mutual’s On the Money program  ◾ 455 education affects financial behaviors. The survey questionnaires included a module on household demographic characteristics, saving and borrowing, finan- cial knowledge, attitudes toward the financial system, numeracy, and well-being. Study participants were divided randomly into treatment and control groups. The treatment group received eight hours of financial education training provided by Old Mutual. The course aimed at promoting sensible financial behaviors, such as higher savings, reducing the debt burden, and financial planning. The intervention was specifically designed to improve subjects’ financial knowledge and skills, targeting common misconceptions about money manage- ment and financial institutions, and aimed to improve their ability to understand financial concepts. Individuals could, for example, improve the way they manage their income, debt, and household expenditures, potentially becoming more able to cope with negative income shocks. The study results may be used to improve the current curriculum of the financial education program to increase its effec- tiveness. This study aims to examine a range of important channels and mechanisms through which financial education could affect outcomes. Beyond evaluating the effectiveness of one specific financial education program, we also hope to provide some deeper insights into how financial education might work more generally both in South Africa and elsewhere in the world. The goal of the study is to examine a range of important channels through which financial education could work. In particular, prior to the start of the study, we formulated the following hypotheses: 1. Financial knowledge: Does financial education increase the participants’ financial knowledge and understanding of financial concepts? 2. Financial planning: Does financial education increase household finan- cial planning (e.g., drawing up budgets, making a will, forecasting future expenses)? 3. Financial outcomes: Does more financial knowledge or better financial planning result in improved financial outcomes? 4. Savings: Does financial education increase the participants’ savings (e.g., savings amounts, frequency of deposits)? 5. Financial choices: Does financial education affect the types of financial products participants choose (e.g., informal borrowing, fixed deposits, loans, insurance plans)? 6. Heterogeneous effects: Does the impact of financial education vary with initial levels of financial literacy, education, or other demographic charac- teristics? 7. Well-being: Does financial education affect the mental well-being of the participants (e.g., stress levels, aspirations, confidence)? 456  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 15.1.3 Old Mutual’s On the Money program Half of the sample was randomly selected to participate in a financial educa- tion course four to six months after completing the baseline survey. The finan- cial education training course replicated Old Mutual’s On the Money curriculum, and was taught by facilitators selected and trained by Old Mutual. Instructors used materials developed to modify the audience’s financial behavior, encourage responsible financial decision making, and improve people’s money manage- ment skills. The On the Money course was divided into five modules, each to be completed in 1¼–1½ hours, over a single day of training. Various didactic activi- ties complemented the curriculum, including individual and group exercises. The instruction was supplemented with video clips and PowerPoint slides. In addition, a workbook was used during interactive group activities and discussions, with attendants practicing creating budgets together and discussing their financial plans for the future. The course started by exposing some misconceptions about money manage- ment. In particular, the focus was on conveying the fact that even low-income individuals can attain financial security if they develop good money management habits and relinquish careless financial behaviors. The facilitator used two exam- ples to reinforce this concept, the ostrich and the peacock. The ostrich buries its head under the sand to hide from the truth and avoid facing problematic situa- tions, just as some individuals attempt to ignore their financial worries in the hope that they will disappear. The peacock uses its extravagant plumage to impress potential mates, just as individuals spend excessive amounts of money to assert their status within their social group. The facilitator then explained how aban- doning behaviors like those of the ostrich and the peacock will enable individuals to enjoy their money while working toward financial security. Individuals were then asked to envision a scenario in which they had attained financial security and sufficient savings for a comfortable retirement and to focus on how satisfying reaching such a state would be. The facilitator explained that if they start embracing the secrets of the “Big Five of Africa,” that scenario could become a reality. Each of the five secrets comprises one of the modules and aims at conveying a core idea and two or three related ideas pertaining to behaviors that would improve individuals’ financial management. The “Secret of the Lion” is saving—in particular, prespending saving, in order to make it easier to set aside a suitable portion of income whenever a pay check is cashed. The “Secret of the Leopard” is setting a clear and realistic goal to motivate and reward good financial management. The “Secret of the Elephant” is to draft comprehensive budgets, listing sources of income and expenses. The “Secret of the Rhino” is to reduce the existing amount of debt insofar as possible and to avoid becoming unnecessarily indebted. Lastly, the “Secret of the Buffalo” is using insurance as a protection from life risks and increasing assets through careful investing. 15.  Evalution of Old Mutual’s On the Money program  ◾ 457 This curriculum hoped to generate positive attitudes toward saving and budgeting, to encourage individuals to rely less on debt, and to present financial institutions as a medium to achieve these goals. The treatment was expected to improve individuals’ financial knowledge, attitudes, and perceptions as measured by the endline survey and to affect behavior along the lines described above. 15.2 METHODOLOGY Examining the scope and pathways by which financial education affects financial behaviors requires that we obtain a rich data set on trained participants and the control group. To that end, we obtained outcome measures through a baseline and an endline questionnaire, which allow us to understand how financial literacy evolves and the ways individuals change their financial behavior in response to training. All individuals participating in the study completed a baseline survey. For the next four to five months, individuals randomly assigned to the treatment group took part in the Old Mutual financial education module. Five to six months afterwards, the entire sample completed the endline survey (see figure 15A.1 in the annex to this chapter for a more detailed timeline of the study). To estimate the true causal impact of financial education, we need to estab- lish the correct counterfactual—i.e., what our trained participants would have done had they not received the training. Studies that simply compare individuals who receive financial education to those who do not are susceptible to selec- tion bias, meaning that people who choose to take financial literacy courses may differ from those who choose not to take such courses. Our evaluation meth- odology—the randomized controlled trial—eliminates that bias. By randomizing assignment to treatment or control group, we ensure the offer to attend training is not correlated with any potentially confounding factors such as level of educa- tion, income, or motivation.1 The randomization thus allows us to interpret differences between treatment and control groups in the endline as the causal effects of financial education. The endline questionnaire allows us to discern these effects by collecting information on a number of variables, including financial knowledge, attitudes, and outcomes. This information lets us evaluate the possible pathways through which financial education can affect behavior. 15.2.1 Intention to treat analysis The study’s setup allows for an intention to treat analysis. This means that the results presented in the findings section depend on the initial random assignment 1  This introductory information was drawn from the baseline report. 458  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES of individuals to a treatment and control group and not on the actual level of participation of individuals in the financial education module. In effect, individuals in the treatment group were merely offered the opportunity to participate in Old Mutual’s financial education program, and were not required to do so. Whether individuals completed the assigned treatment or not is disregarded in an intention to treat analysis, to avoid the risk that underlying factors correlated with the treatment’s take-up might bias results. This is because some unobserv- able individual characteristics—personality traits such as conscientiousness, for example—might affect both the program’s take-up and financial outcomes such as savings. In measuring the effect of financial education only among individ- uals who attended the training without accounting for conscientiousness or any other unobservable characteristic affecting both participation and outcomes, we would in effect attribute the effect of these characteristics to financial education, creating a spurious estimate for the program’s effect. Measuring the differences between the two groups as assigned in the randomization process ensures that the effect detected is truly attributable to financial education. However, it is important to keep in mind that these differ- ences are to be attributed to the option of freely attending a financial education training course, rather than participation in the program itself. 15.2.2 Sample selection The sample was selected among BSSP and WDB borrowing groups because these groups resembled the target population of Old Mutual’s financial education program. The selected sample includes 43 BSSP and 36 WDB groups located across 13 geographic areas in the provinces of Eastern Cape and KwaZulu-Natal. Clusters of BSSP and/or WDB groups were located in Flagstaff, Bizana, Mount Frere, Matatiele, Queenstown, Mthatha, Butterworth, East London, Port Elizabeth, Underberg, Empangeni, Mtubatuba, and Port Shepstone (see figure 15A.2 for a map indicating these areas). Table 15A.1 presents a comprehensive list of the groups surveyed in the baseline; table 15A.2 presents a list of the groups surveyed in the endline. These groups are of varying sizes, ranging from those with 10 or fewer members to those consisting of more than 50 individuals. All members of the WDB groups are women, and men are a minority within the burial societies in the sample. Overall, the sample includes 1,291 individuals: 589 received the On the Money training, and 661 were in the control group. In the final sample, 673 individuals belonged to a WDB group, while 585 individuals were members of a burial society. 15.2.3 Data collection methods The financial education training module was implemented by Old Mutual facilita- tors, who were selected and trained by Old Mutual. Surveys were conducted by 10 surveyors from a professional survey firm, Reform Development Consulting, 15.  Evalution of Old Mutual’s On the Money program  ◾ 459 based in Durban. The surveyors were hired and trained by the Abdul Latif Jameel Poverty Action Lab (J-PAL Africa). Baseline data collection began July 4, 2011, and ended November 10, 2011; the endline data collection began May 31, 2012, and ended November 20, 2012. The baseline data collection was conducted in a single wave, with surveyors visiting each location and moving on to the next one once collection was completed; the endline was articulated in two distinct waves to reduce attrition. Surveyors visited all locations in the first wave, between May 31 and September 3, 2012. During the second wave, from October 24 to November 20, 2012, surveyors traced indi- viduals who could not be contacted during the first wave to ensure continuity between the baseline and the endline data. The baseline and endline survey questionnaires included questions on the following:2 ◾◾ General household and personal characteristics ◾◾ Financial literacy and behavior ◾◾ General and mathematical literacy ◾◾ Attitudes toward banks, hire purchase, and gambling ◾◾ Past experience with hire purchases ◾◾ Gambling behaviors ◾◾ Risk aversion ◾◾ Savings, debt, and insurance behavior ◾◾ Past, current, and future expenditure behavior ◾◾ Mental distress and well-being ◾◾ Future goals and aspirations. The information thus collected in the baseline was used to ensure that the treatment and control groups displayed similar characteristics and to assess how the initial levels of certain variables (e.g., income, education, financial literacy, or mathematical abilities) can lead to heterogeneous impacts of the financial educa- tion training. In order to improve statistical power, data from the baseline survey were used to create group-level pairs among the BSSP and WDB groups, respec- tively. Within each pair, treatment was randomly assigned to one group with the other serving as control. The data collected in the baseline and endline surveys were recorded on smart phones and automatically transferred to a server using a software called Pendragon VI, from which data were imported for collating and cleaning. 2  See annex B for full survey. 460  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 15.2.4 Randomization The assignment of individuals to the treatment group that would receive financial education or to the control group was determined through group-level random- ization. Random assignment occurred on a rolling basis: once all groups in a predefined geographical area were surveyed, the assignment of those groups occurred, before the surveyor began data collection in the next region. Assign- ment was done on the basis of pair-wise matching using the Mahalanobis distance algorithm based on eight variables: ◾◾ Number of households ◾◾ Median household income ◾◾ Median household size ◾◾ Median number of dependents ◾◾ Whether the median household saved in any way ◾◾ Median number of savings accounts ◾◾ Median baseline math score ◾◾ Median baseline financial literacy score. Once the groups were thus defined, one group was randomly assigned to the treatment group, and the other to the control group. When the number of sampling units inside a stratum was odd, a “pair” containing three sampling units had to be generated. Within this type of “pair,” sampling units were randomly ordered, the first sampling unit being assigned to the control group, the second to the treatment group, and the third assigned randomly to one of the two groups. This rolling-basis process was necessary because of timeline constraints, and it resulted in a de facto stratification by area and organization type (BSSP or WDB), while the matching mechanism ensured that the samples were balanced. 15.2.5 Sample description Of the 1,258 individuals in the final sample, 1,117 are women and 130 are men. Within BSSPs, 78.6  percent of the members surveyed are female compared to more than 99 percent in the WDB groups. The median respondent is a 51-year-old woman who has seven years of schooling, earns R 1,050 a month, and receives at least one government grant of R 260 per month. While the collection of survey data at baseline served the primary purpose of ensuring that the treatment and control groups were comparable after the randomization, we can also use this information to draw a comparison between our sample and the rest of the South African population. For instance, it is possible to compare our data set with that of the National Income Dynamics Study (NIDS) collected by the School of Economics at the University of Cape Town. The NIDS data set is a comprehensive panel of surveys administered to 28,000 individuals in 7,300 households in South Africa, collecting detailed information on demographic 15.  Evalution of Old Mutual’s On the Money program  ◾ 461 characteristics, income and employment, health, education, and well-being. We use the results from Wave 2 of the panel, collected in 2010–11, as a point of refer- ence for our sample. There are some small but significant differences between the samples from WDB and BSSP groups with respect to the NIDS sample. Members of WDB groups are on average younger than BSSP members—respectively, 48 and 55 years old, on average. The average NIDS respondent to the “Adult (15+) Questionnaire” is significantly younger: 33 years old. While the native language of the majority of our sample is IsiXhosa (64  percent of WDB respondents, and 71  percent of BSSP respondents), IsiXhosa is the second-most common language in the NIDS survey, spoken by less than 16  percent of the respondents. The most widely spoken language in the NIDS sample is IsiZulu, spoken by almost 33 percent of individuals. The remaining 36 percent of WDB members also speak IsiZulu; among BSSP members, 20 percent speak IsiZulu and 9.4 percent speak Sesotho. It is thus clear that because of the geographic concentration of our sample in the Eastern Cape and KwaZulu-Natal provinces, certain populations were oversampled with respect to the national average. While 17.3 percent of BSSP and 12.2 percent of WDB respondents have not been educated, the two groups have on average completed about 7.54 years of education if they were enrolled in school; this puts them slightly below the national average, with almost 14 percent of individuals having no schooling, and an average of 9.17 years of education among individuals who did receive some education. Only 12.3 percent of BSSP and 6.2 percent of WDB respondents have completed grade 12, obtaining a matric degree; in contrast, more than 23 percent of the national sample did so. Individuals in our sample report a lower under- standing of English than the national average; about 29  percent of WDB and 35 percent of BSSP respondents said they understood some English. In the NIDS sample, almost 63 percent of individuals reported they read English well or fairly well, and almost 62 percent said they wrote English well or fairly well. Our sample is also somewhat poorer than the national average, reporting a median income of R 1,050 among BSSP members and R 1,260 among WDB members. Individuals in the NIDS survey report a median gross income of R 1,850 and R 1,600 net of taxes and deductions. Only a minority of the national sample reports receiving government grants; while in the sample under investigation this is fairly commonplace, with almost 85 percent of respondents receiving at least one government grant. Individuals in our sample are also somewhat less likely to be salaried employees and more likely to be self-employed than the national average. Slightly more than 10 percent of WDB members and less than 5 percent of BSSP members reported being in salaried employment; the figure reported in the NIDS survey is over 24 percent. While a little over 10 percent of NIDS individ- uals reported being self-employed (or homemakers), this figure is at 20 percent for BSSP members and at over 75 percent for WDB members. 462  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 15.2.6 Attendance Randomization was carried out at the group level, meaning that all individuals belonging to the selected BSSP or WDB group were sent an invitation to partici- pate in the financial education training. Before the training, Old Mutual facilitators sent an open invitation to the entire group, and individuals were given the choice of whether or not they wanted to participate. While group members external to our sample were allowed to participate in the financial education module as well, because of the geographical distance between surveyed groups, it is extremely unlikely that any members of the control group participated in the training courses. Treatment spillover is thus not a major concern, and the control group can be thought of as an appropriate counterfactual for the treatment group. 15.2.7 Balance checks Because of the randomization in the treatment assignment, demographic charac- teristics as well as average responses to baseline survey questions are expected to be balanced across treatment and control groups. Furthermore, the random- ization was performed within the subsamples of BSSP and WDB groups; there- fore, even within these subsamples, observations are expected to be balanced. Table 15A.3 in the annex to this chapter reports the relevant statistics on demo- graphic characteristics and on the set of outcome variables later employed in the analysis of the effects of financial education training. The responses to these variables were collected in the baseline survey—i.e., before the financial educa- tion intervention—and the table reports a randomization balance test in the full sample as well as in the subsamples of BSSP and WDB groups. All variables, with the exception of the respondents’ risk aversion, are balanced both in the full sample and within each subsample. Risk aversion is measured by a question asking the respondent to choose between a gamble and a safe choice of equal expected pay. There is no reason to believe that this variable should be unbalanced, and we attribute this solely to chance. In a relatively small sample of about 400 respondents, it is not out of the ordinary to occasionally have a slight imbalance for a specific characteristic. While statistically significant, the differ- ence in the fraction of risk-averse respondents in the full sample is very small in absolute terms—83 percent in the control group compared to 87 percent in the treatment group—and we do not expect this discrepancy to affect the endline results. Furthermore, we conducted the randomization ourselves, by computer. 15.3 FINDINGS To estimate the effect of financial education on the groups that were offered to take part in Old Mutual’s On the Money program, we use regression analysis to estimate the difference between the financial outcomes of the treatment 15.  Evalution of Old Mutual’s On the Money program  ◾ 463 and control groups. We use the ordinary least squares method to estimate how financial outcomes depend on treatment assignment. The estimated coefficient reports the treatment effect—i.e., the average difference in outcome between the treatment and control groups. The control mean reports the control group average for a particular outcome. The mean outcome for the treatment group can be calculated by adding the value of the coefficient to the control mean. A statistically significant coefficient means that we can, with high confi- dence, reject the hypothesis that the difference between the treatment and control groups is due solely to chance. The level of confidence is indicated by asterisks in the following tables: one asterisk next to the coefficient means that there is a 90 percent likelihood that the observed difference is the causal effect of treatment assignment, as opposed to a difference due to chance; two asterisks are reported when this likelihood is higher than 95 percent; and three asterisks indicate that this likelihood is beyond 99 percent. The tables report these values for the entire sample as well as separately for the subsamples corresponding to the BSSP and WDB groups. 15.3.1 The effect of financial education on financial awareness and attitudes Table 15.1 reports the effect of financial education on financial awareness. The first five variables are indexes created to summarize various responses on each topic. The last variable reports the response to survey question F21. Responses were coded using dummy variables, which took the value of 1 if the individual was aware of the discussed topic, and the value of 0 if she was not. These values were then averaged to create a financial awareness index. See table 15A.13 for the breakdown of each index into the original variables. TABLE 15.1  Impact on financial awareness FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN Awareness of savings 848 0.023 0.825 436 0.063*** 0.805 412 −0.032 0.845 accounts Awareness of loans and 848 0.009 0.596 436 0.038 0.595 412 −0.043 0.596 interest rates Awareness of insurance 852 −0.032 0.504 437 −0.009 0.49 415 −0.063 0.517 Awareness of purchasing 848 0.031 0.583 436 −0.001 0.606 412 0.066*** 0.561 on credit Awareness of budgeting 852 0.061*** 0.471 437 0.098*** 0.451 415 0.013 0.491 F21: Awareness of sepa- 848 −0.017 0.696 436 0.047 0.699 412 −0.103** 0.693 rating business and house- hold accounts Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 464  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Financial education seems to have a significant effect on budgeting both in the full sample and within the BSSP sample. It should be noted that this is the topic with the lowest initial level of awareness in the sample, and thus the topic for which the largest potential gains from financial education could be accrued. Within the BSSP sample, there is also an effect on the awareness of savings accounts, while WDB groups increased their awareness of purchasing on credit. Within this sample there is also a negative effect on the awareness of sepa- rating business and household accounts. Results are more mixed for financial attitudes and perceptions. The BSSP sample seems to have considerably improved its attitudes and perceptions vis-à-vis financial institutions and products, feeling less nervous going into a bank, attaching more importance to understanding the contractual terms of hire purchase agreements before undertaking them, and appreciating the benefit of lowering interest payments through saving (table 15.2). However, BSSP members seem to display a lower recognition of the importance of saving early to ensure their children’s education. In contrast, WDB group members display a slight decline in financial atti- tudes. They seem doubtful that they can improve their finances and seem to fail to appreciate the cost of borrowing versus saving. Results reported in table 15A.5 show that these differences between the BSSP and WDB groups are not driven by education levels. TABLE 15.2  Impact on financial attitudes and perceptions FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN G4: Never too late to 848 −0.016 0.82 436 0.026 0.819 412 −0.067* 0.82 improve finances G1: Saving early for child’s 847 −0.059* 0.556 436 −0.141*** 0.604 411 0.029 0.509 education B18: Does not feel nervous 876 0.079** 0.69 440 0.095** 0.694 436 0.058 0.686 going into a bank B17: People at the bank are 874 0.036 0.816 439 0.031 0.833 435 0.041 0.799 there to help I3: Purchasing accident 849 −0.008 0.519 435 0.037 0.516 414 −0.063 0.521 insurance in dangerous job E3: Understanding details 848 0.061** 0.718 436 0.094** 0.735 412 0.016 0.701 of hire purchase contracts is important E23: Conservative 847 0.028 0.652 435 0.119** 0.629 412 −0.082** 0.674 spending to avoid large interest payments Overall financial attitudes 881 −0.009 0.674 442 0.039** 0.682 439 −0.064** 0.666 and perceptions Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 15.  Evalution of Old Mutual’s On the Money program  ◾ 465 15.3.2 The effect of financial education on behavior Financial education seems to have significant effects on savings. It significantly raises self-reported saving, both in terms of the number of people who report saving and in terms of the average amount of savings, as shown in table  15.3. Within the treatment group, not only do more BSSP and WDB respondents report saving money in any way, but both the average monthly savings and the expected future monthly savings increase. BSSP members are less likely to withdraw money as soon as it is deposited; the opposite is true for WDB individuals. Our analysis in table 15.4 also shows an overall decrease in the average number of individuals who applied for a loan in the last six months. Within the BSSP sample, there is also evidence of a decrease in the use of hire purchase agreements as a source of consumer credit. This result does not hold for WDB groups, which instead seem to rely more heavily on hire purchase agreements. Again, this divergence in results does not seem to be driven by education level (see table 15A.7). TABLE 15.3  Impact on savings behavior FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN B1: Has bank account 879 0.023 0.554 442 0.001 0.615 437 0.028 0.496 B4: Has Mzansi account 459 −0.021 0.342 247 −0.002 0.265 212 −0.065 0.427 B16: Withdraws money as 873 −0.008 0.455 438 −0.085* 0.546 435 0.061* 0.37 soon as it is deposited C1: Saves money in any way 851 0.129*** 0.733 436 0.126*** 0.767 415 0.123*** 0.701 C12: IST (monthly savings) 822 0.686*** 4.323 416 0.821*** 4.466 406 0.401** 4.193 C16: IST (expected future 851 0.463*** 5.701 436 0.548** 5.393 415 0.327 5.992 monthly savings) Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. IST = inverse sine transformation. TABLE 15.4  Impact on borrowing behavior FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN D1: Applied for a loan in past 851 −0.027** 0.051 436 −0.002 0.032 415 −0.055*** 0.068 6 months E7-E8: Bought an Item on hire 846 0.000 0.069 435 −0.044* 0.077 411 0.058** 0.062 purchase in past 6 months E9: Number of hire purchase 846 0.015 0.049 435 −0.042** 0.061 411 0.089*** 0.037 contracts in past 6 months Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 466  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Financial education seems to play a role in reducing gambling and decreasing risk aversion (table  15.5). The treatment effects vary significantly between the two groups: in the WDB sample, there is a slight decrease in gambling and an increase in the self-reported preference for saving versus gambling; these vari- ables have a virtually null treatment effect in the BSSP sample. Within the BSSP sample, there is instead a decrease in risk aversion. Lastly, financial education does not seem to have an effect on financial plan- ning in the full sample (table 15.6). There is, however, a negative effect on retire- ment plans within the WDB sample. TABLE 15.5  Impact on gambling bahavior FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN L7: Has gambled in past 6 848 −0.013 0.11 433 0.007 0.113 415 −0.05** 0.106 months L9: Prefers saving money 847 −0.013* 0.572 436 −0.02 0.61 411 0.202*** 0.535 versus gambling Risk averse 844 −0.038* 0.908 434 −0.087** 0.919 410 0.005 0.897 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. TABLE 15.6  Impact on financial planning FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN F9: Has written a budget 848 0.004 0.155 436 0.013 0.177 412 −0.004 0.134 F7: Has thought about what 847 0.007 0.656 435 0.052 0.669 412 −0.066 0.644 to do in financial emergency F15: Has made financial 844 −0.032 0.122 435 0.018 0.129 411 −0.057*** 0.115 plans for retirement Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 15.3.3 The effect of financial education on reported well- being Financial education does not seem to have an effect on the well-being of the entire sample (table 15.7). Individuals in the BSSP sample report feeling less over- whelmed, while results are mixed within WDB groups. WDB individuals feel some- what more confident and less stressed, but also more overwhelmed. These effects are null if the sample is disaggregated along educational lines (table 15A.10). 15.  Evalution of Old Mutual’s On the Money program  ◾ 467 TABLE 15.7  Impact on well-being FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN M2: Feels confident 848 −0.019 0.625 435 −0.076 0.605 413 −0.072* 0.645 M1: Feels stressed 847 −0.033 0.379 435 0.016 0.383 412 −0.12*** 0.375 M3: Feels overwhelmed 844 −0.002 0.471 435 −0.11** 0.472 413 0.145*** 0.469 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 15.3.4 The effect of financial education on numeracy Financial education does not seem to have any effect on numeracy skills (table 15.8), which is not surprising, since the training program did not focus on either mathematical skills or problem solving. TABLE 15.8  Impact on numeracy skills FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN Basic math skills 749 0.014 0.51 373 0.01 0.512 376 0.016 0.508 Basic numeric problem-solving 850 0.003 0.25 436 0.017 0.252 414 −0.023 0.247 skills Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. The breakdown in table 15A.14 shows two isolated outcomes with statisti- cally significant treatment effects; this allows for no relevant conclusions to be drawn. REFERENCES Campbell, John Y. 2006. “Household Finance.” Journal of Finance 61: 1553–604. Chaia, Alberto, Aparna Dalal, Tony Goland, Maria Jose Gonzalez, Jonathan Morduch, and Robert Schiff. 2009. “Half the World Is Unbanked.” Financial Access Initiative Framing Note. http://www.microfinancegateway.org/gm/document-1.9.40671/25.pdf. Cole, Shawn A., Thomas Sampson, and Bilal Zia. 2011. “Prices or Knowledge? What Drives Demand for Financial Services in Emerging Markets?” Journal of Finance 66 (6): 1933–67. Fischer, Greg, Alejandro Drexler, and Antoinette Schoar. 2010. “Keeping It Simple: Financial Literacy and Rules of Thumb.” Paper presented at the Centre for Economic Policy Research Development Economics Workshop, Barcelona, October 8–9. Guiso, Luigi, and Tullio Jappelli. 2008. “Financial Literacy and Portfolio Diversification.” CSEF Working Paper 212, Centre for Studies in Economics and Finance, University of Naples, Italy. 468  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES Guiso, Luigi, Paola Sapienza, and Luigi Zingales. 2008. “Trusting the Stock Market.” Journal of Finance 63 (6): 2557–600. Karlan, Dean, Jonathan Morduch, and Sendhil Mullainathan. 2010. “Take-Up: Why Microfinance Take-Up Rates Are Low and Why It Matters.” Financial Access Initiative Research Framing Note. Karlan, D., and M. Valdivia. 2011. “Teaching Entrepreneurship: Impact of Business Training on Microfinance Clients and Institutions.” Review of Economics and Statistics 93 (2): 510–27. Laibson, David. 1997. “Golden Eggs and Hyperbolic Discounting.” Quarterly Journal of Economics 112 (2): 443–78. Lusardi, Annamaria. 2008. “Household Saving Behavior: The Role of Financial Literacy, Information, and Financial Education Programs.” NBER Working Paper 13824, National Bureau of Economic Research, Cambridge, MA. Lusardi, Annamaria, and Olivia S. Mitchell. 2007. “Financial Literacy and Retirement Preparedness: Evidence and Implications for Financial Education.” Business Economics 42 (1): 35–44. —. 2008. “Planning and Financial Literacy: How Do Women Fare?” NBER Working Paper 13750, National Bureau of Economic Research, Cambridge, MA. —. 2009. “How Ordinary Consumers Make Complex Economic Decisions: Financial Literacy and Retirement Readiness.” NBER Working Paper 15350, National Bureau of Economic Research, Cambridge, MA. Lusardi, Annamaria, Olivia S. Mitchell, and Vilsa Curto. 2010. “Financial Literacy among the Young.” Journal of Consumer Affairs 44 (2): 358–80. Lusardi, Annamaria and Peter Tufano. 2009. “Debt Literacy, Financial Experiences, and Overindebtedness.” NBER Working Paper 14808, National Bureau of Economic Research, Cambridge, MA. Stango, V., and J. Zinman. 2009. “Exponential Growth Bias and Household Finance.” Journal of Finance 64 (6): 2807–49. Van Rooij, Maarten, Annamaria Lusardi, and Rob Alessie. 2007. “Financial Literacy and Stock Market Participation.” NBER Working Paper 13565, National Bureau of Economic Research, Cambridge, MA. 15.  Evalution of Old Mutual’s On the Money program  ◾ 469 ANNEX A: SUPPORTING MATERIALS FIGURE 15A.1  Study timeline Baseline survey July 4–November 10, 2011: 1,117 individuals are surveyed by Reform Development Consulting Randomization 590 individuals are assigned to the treatment group and 660 individuals are assigned to the control group Old Mutual financial education training September–December 2011: Old Mutual delivers the 8-hour financial education training to the BSSP and WDB samples assigned to the treatment group Endline survey First Wave: Second Wave: May 31–September 3, 2012 October 24–November 20, 2012 470  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES FIGURE 15A.2  Study map Source: Google Maps 2011. Note: Yellow flags = BSSP clusters; red flags: WDB clusters. 15.  Evalution of Old Mutual’s On the Money program  ◾ 471 TABLE 15A.1  Baseline survey records LOCATION DAY DATE SURVEYS GROUP 1 GROUP 2 Monday 04-Jul-11 33 216044 216014 Tuesday 05-Jul-11 42   Wednesday 06-Jul-11 42 216044 216014 Thursday 07-Jul-11 37 216027 216012 Flagstaff Monday 11-Jul-11 33 216012   Tuesday 12-Jul-11 46 216049 216039 Wednesday 13-Jul-11 44 216022 216028 Thursday 14-Jul-11 27 216036 216020 Friday 15-Jul-11 24 Laphumilanga   Monday 18-Jul 14 Mtshokotshi   Tuesday 19-Jul 23 Chulunca   Mthatha Wednesday 20-Jul 38 Phakamani Siyaya Thursday 21-Jul 17 Ma Africa   Friday 22-Jul 19 Khulasizwe   Tuesday 26-Jul 38 Mzwana Simanye Wednesday 27-Jul 40 Galisa Masiphekisane Butterworth Thursday 28-Jul 12 Somana   Friday 29-Jul 32 Qayiso Phila Tuesday 02-Aug 4 Nobunto   Queenstown Wednesday 03-Aug 9 Rainbow Group   Thursday 04-Aug 9 Khanyisa   Monday 08-Aug 36 214038   Mt Frere Wednesday 10-Aug 39 214028 214035 Thursday 11-Aug 38 214050 214053 Tuesday 16-Aug 36 Masizakhe Mangweni Wednesday 17-Aug 20 Dubu Masibambane East London Thursday 18-Aug 10 Masihlume   Friday 19-Aug 25 Tildin Lathitha Monday 22-Aug 10 Ndileka   Port Elizabeth Thursday 01-Sep 3 Speelman   Queenstown Friday 02-Sep 13 Amangelengele   Tuesday 06-Sep 12 Sikhulisele   Mt Frere Wednesday 07-Sep 14 Nontsizi   Thursday 08-Sep 6 Mehlomakhulu   Matatiele Friday 09-Sep 7 Tshwaranang   Monday 12-Sep 22 213035   Wednesday 14-Sep 12 210024   Bizana Thursday 15-Sep 11 213013   Friday 16-Sep 12 Masizakhe Mzamomuhle Wednesday 21-Sep  20 Nhliziyonye   Port Shepstone Thursday 22-Sep  19 Ndlozana   Monday 10-Oct-11 15 217025 217029 Tuesday 11-Oct-11 28 217035 217011 Underberg Wednesday 12-Oct-11 25 217014 217010 Thursday 13-Oct-11 27 217023   (continued) 472  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 15A.1  Baseline survey records (continued) LOCATION DAY DATE SURVEYS GROUP 1 GROUP 2 Friday 14-Oct-11 53 Zubagcine   Wednesday 19-Oct-11 11 Hamba Masiphekisani Matatiele Thursday 20-Oct-11 8   Friday 21-Oct-11 6     Sunday 23-Oct-11 20 Ubumbano   Monday 24-Oct-11 15   Wednesday 26-Oct-11 29 Mzingazi Halalisani Thursday 27-Oct-11 18 208044   Empangeni/PMB Wednesday 02-Nov-11 27 203007   Thursday 03-Nov-11 23 203064   Saturday 05-Nov-11 10 Thembalethu   Sunday 06-Nov-11 12 Christian Charity   Monday 07-Nov-11 35 201054 201063 Tuesday 08-Nov-11 31 201030   Mtubatuba Wednesday 09-Nov-11 16 201065   Thursday 10-Nov-11 18 201013   15.  Evalution of Old Mutual’s On the Money program  ◾ 473 TABLE 15A.2  Endline survey records LOCATION DAY DATE SURVEYS GROUP 1 GROUP 2 Friday 22-Jun-12 11 Laphumilanga   Saturday 23-Jun-12 1 Laphumilanga   Friday 29-Jun-12 1 Laphumilanga   Monday 16-Jul-12 8 216048   Tuesday 17-Jul-12 9 216028   Wednesday 18-Jul-12 8 216044 216048 Thursday 19-Jul-12 15 216039   Friday 20-Jul-12 12 216044 216008 Saturday 21-Jul-12 6 216017 216013, 216021 Thursday 26-Jul-12 6 216039   Monday 27-Aug-12 22 216012   Tuesday 28-Aug-12 11 216028 216044, 216039, 216012 Flagstaff Wednesday 29-Aug-12 6 216013 216017, 216039 Friday 31-Aug-12 1 216048   Wednesday 24-Oct-12 7 216012   Thursday 25-Oct-12 6 216028 216014 Friday 26-Oct-12 2 216022   Saturday 27-Oct-12 5 216036   Sunday 28-Oct-12 2 216026   Tuesday 30-Oct-12 3 216028 216014 Wednesday 31-Oct-12 10 216027 216049 Thursday 1-Nov-12 7 216012 216024, 216026 Friday 2-Nov-12 2 216022   Saturday 17-Nov-12 2 216012 216036 Wednesday 29-Aug-12 8 Mtshokotshi   Thursday 30-Aug-12 14 Siyaya   Friday 31-Aug-12 14 Chulunca   Mthatha Saturday 1-Sep-12 12 Phakamani   Monday 3-Sep-12 15 Mtshokotshi Chulunca, Ma Africa Monday 12-Nov-12 11 Nxanga   Thursday 28-Jun-12 8 Somana   Friday 29-Jun-12 8 Phila   Tuesday 3-Jul-12 9 Galisa   Wednesday 4-Jul-12 14 Simanye   Thursday 5-Jul-12 24 Masiphekisani (BTW)   Butterworth Friday 6-Jul-12 7 Mzwana   Saturday 7-Jul-12 15 Qayiso   Thursday 15-Nov-12 2 Somana Phila Thursday 15-Nov-12 3 Mzwana Friday 16-Nov-12 7 Mzwana Phila, Masiphekisani (BTW) Saturday 16-Jun-12 5 Rainbow Group   Thursday 15-Nov-12 2 Khanyisa   Queenstown  Friday 15-Jun-12 8 Masiphekisani (MtFrere)   Sunday 17-Jun-12 6 Masiphekisani (MtFrere)   (continued) 474  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 15A.2  Endline survey records (continued) LOCATION DAY DATE SURVEYS GROUP 1 GROUP 2 Tuesday 19-Jun-12 3 214038   Wednesday 20-Jun-12 10 214035   Thursday 21-Jun-12 8 214038   Friday 22-Jun-12 13 214053   Saturday 23-Jun-12 8 214050   Monday 25-Jun-12 9 214028   Tuesday 26-Jun-12 11 214038 214028 Mt Frere Wednesday 27-Jun-12 10 214053 214028 Monday 9-Jul-12 5 214050   Tuesday 17-Jul-12 4 214028   Wednesday 31-Oct-12 1 214035   Saturday 10-Nov-12 5 214038   Friday 9-Nov-12 6 214035 214028 Monday 12-Nov-12 1 214050   Monday 4-Jun-12 6 Dubu   Tuesday 5-Jun-12 6 Masihlume   Wednesday 6-Jun-12 9 Lathitha   Thursday 7-Jun-12 12 Mangweni   East London Friday 8-Jun-12 2 Ndileka   Saturday 9-Jun-12 2 Dubu   Monday 11-Jun-12 9 Masizakhe (KWT)   Tuesday 12-Jun-12 8 Masibambane   Wednesday 13-Jun-12 3 Masizakhe (KWT)   Thursday 14-Jun-12 1 Amangelengele   Queenstown  Friday 15-Jun-12 4 Amangelengele     Saturday 16-Jun-12 1 Amangelengele   Wednesday 14-Nov-12 1 Amangelengele   Monday 18-Jun-12 11 Sikhulisele   Mt Frere Tuesday 19-Jun-12 11 Nontsizi   Tuesday 12-Jun-12 3 Mehlomakhulu   Wednesday 13-Jun-12 5 Tshwaranang   Matatiele Thursday 8-Nov-12 1 Mehlomakhulu   Friday 9-Nov-12 2 Mehlomakhulu   Wednesday 20-Jun-12 14 213035 Mzamomuhle Thursday 21-Jun-12 9 Mzamomuhle Masizakhe (Biz) Saturday 23-Jun-12 8 Mzamomuhle Masizakhe (Biz) Monday 25-Jun-12 10 213013   Bizana Friday 29-Jun-12 1 213013   Saturday 21-Jul-12 1 213013   Friday 2-Nov-12 7 213035   Saturday 3-Nov-12 5 213035 210024 Saturday 17-Nov-12 1 213035   (continued) 15.  Evalution of Old Mutual’s On the Money program  ◾ 475 TABLE 15A.2  Endline survey records (continued) LOCATION DAY DATE SURVEYS GROUP 1 GROUP 2 Tuesday 26-Jun-12 10 Nhliziyonye   Port Wednesday 27-Jun-12 9 Ndlozana   Shepstone Thursday 28-Jun-12 7 Nhliziyonye   Friday 29-Jun-12 6 Ndlozana   Thursday 31-May-12 11 217010   Monday 4-Jun-12 21 217023   Tuesday 5-Jun-12 10 217011 217010 Wednesday 6-Jun-12 6 217029   Underberg Thursday 7-Jun-12 16 217034   Friday 8-Jun-12 3 217014   Saturday 9-Jun-12 6 217014 217023, 217011, 217034 Monday 11-Jun-12 10 217025   Thursday 14-Jun-12 15 Hamba   Friday 26-Oct-12 2 Zubagcine Hamba Saturday 27-Oct-12 1 Hamba   Matatiele Tuesday 6-Nov-12 7 Zubagcine   Wednesday 7-Nov-12 14 Zubagcine   Thursday 8-Nov-12 4 Zubagcine   Tuesday 3-Jul-12 9 Mzingazi   Wednesday 4-Jul-12 9 Halalisani   Thursday 5-Jul-12 9 Mzingazi   Friday 6-Jul-12 10 Thembalethu   Empangeni/ Wednesday 22-Aug-12 11 203007   PMB Tuesday 28-Aug-12 7 Christian Charity   Wednesday 24-Oct-12 16 203007 208044 Monday 19-Nov-12 5 Ubumbano   Tuesday 20-Nov-12 7 Ubumbano   Friday 24-Aug-12 9 201054   Saturday 25-Aug-12 11 201065   Mtubatuba Sunday 26-Aug-12 7 201054 201065 Monday 27-Aug-12 25 201030   476  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 15A.3  Balance checks FULL SAMPLE BURIAL SOCIETY SAMPLE WDB SAMPLE ALL CONTROL TREATMENT P- CONTROL TREATMENT P- CONTROL TREATMENT P-   N MEAN N MEAN N MEAN VALUE N MEAN N MEAN VALUE N MEAN N MEAN VALUE Respondent characteristics X9: Is female 1,224 0.90 647 0.89 577 0.90 0.760 251 0.77 189 0.79 0.665 262 1.00 172 0.99 0.377 X15: Is married 1,207 0.48 635 0.49 572 0.47 0.456 244 0.53 186 0.46 0.275 259 0.49 172 0.47 0.650 X14: Has no schooling 1,229 0.15 649 0.16 580 0.14 0.461 249 0.21 190 0.18 0.562 266 0.14 174 0.13 0.845 X14: Has 12 or more years of education 1,229 0.09 649 0.10 580 0.09 0.746 249 0.13 190 0.12 0.818 266 0.05 174 0.06 0.583 I1-I2: Is risk averse 1,079 0.85 573 0.83 506 0.87 0.080* 221 0.82 173 0.93 0.002*** 231 0.81 146 0.86 0.323 Household characteristics A2-A9: Bottom 25% of income distribution 1,235 0.20 653 0.21 582 0.19 0.599 253 0.27 190 0.25 0.665 266 0.16 174 0.16 0.896 X16: Number of people in household 1,228 6.46 648 6.41 580 6.52 0.682 249 5.80 190 6.04 0.468 266 6.69 174 7.14 0.389 X17: Number of dependents 1,228 4.77 648 4.71 580 4.83 0.621 249 4.18 190 4.41 0.459 266 5.05 174 5.52 0.401 Baseline outcomes K1-K4, K6: Basic math skills 1,206 0.40 638 0.40 568 0.39 0.803 245 0.41 188 0.45 0.453 264 0.39 169 0.35 0.301 B18: Not nervous when going to a bank 1,218 0.58 642 0.59 576 0.56 0.311 245 0.58 187 0.57 0.880 265 0.62 174 0.62 0.997 E3: Understanding details of HP contracts is imp. 1,218 0.71 646 0.71 572 0.71 0.964 247 0.75 188 0.80 0.203 265 0.66 171 0.66 0.931 B1: Has bank account 1,221 0.55 643 0.58 578 0.52 0.262 244 0.62 188 0.62 0.982 266 0.52 174 0.48 0.609 B4: Has Mzansi account 402 0.31 248 0.31 154 0.31 0.982 111 0.24 62 0.26 0.816 95 0.41 36 0.36 0.622 B16: Withdraws as soon as she deposits 1,217 0.47 642 0.47 575 0.46 0.700 245 0.48 187 0.49 0.873 265 0.50 174 0.43 0.219 C1: Saves money in any way 1,222 0.51 647 0.50 575 0.52 0.631 247 0.53 187 0.51 0.792 266 0.46 173 0.50 0.547 D1: Applied for a loan in past 6 months 1,210 0.25 641 0.28 569 0.23 0.449 244 0.18 186 0.10 0.145 264 0.37 171 0.29 0.442 E9: Number of HP contracts in past 6 months 1,210 0.06 642 0.05 568 0.07 0.356 244 0.04 187 0.03 0.666 266 0.06 170 0.09 0.440 F9: Has written a budget 1,210 0.24 641 0.25 569 0.23 0.551 246 0.24 189 0.24 0.942 265 0.23 170 0.19 0.493 F7: Has thought what to do in fin. emergency 1,206 0.42 638 0.44 568 0.41 0.384 245 0.50 188 0.50 0.972 262 0.41 169 0.39 0.726 F15: Made financial plans for retirement 1,211 0.12 640 0.12 571 0.11 0.648 246 0.11 188 0.12 0.826 264 0.11 170 0.11 0.844 M1: Feels stressed 1,200 0.54 637 0.54 563 0.53 0.774 243 0.47 187 0.50 0.535 265 0.56 168 0.58 0.708 M2: Feels confident 1,198 0.30 637 0.30 561 0.31 0.583 243 0.29 187 0.31 0.718 265 0.29 167 0.31 0.708 M3: Feels overwhelmed 1,198 0.43 636 0.44 562 0.43 0.591 243 0.38 187 0.42 0.492 264 0.49 168 0.43 0.272 15.  Evalution of Old Mutual’s On the Money program  ◾ 477 TABLE 15A.4  Heterogeneous effects by education level: financial awareness FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN Awareness of savings 848 0.023 0.825 402 0.053* 0.775 442 0.006 0.873 accounts Awareness of loans and 848 0.009 0.596 402 −0.006 0.541 442 0.013 0.648 interest rates Awareness of insurance 852 −0.032 0.504 405 −0.012 0.48 443 −0.045 0.525 Awareness of purchasing 848 0.031 0.583 402 0.084* 0.569 442 0.009 0.6 on credit Awareness of budgeting 852 0.061*** 0.471 405 0.091*** 0.41 443 0.048 0.53 F21: Awareness of sepa- 848 −0.017 0.696 402 −0.068 0.663 442 0 0.723 rating business and house- hold accounts Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. TABLE 15A.5  Heterogeneous effects by education level: financial attitudes and perceptions FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN G4: Never too late to 848 −0.016 0.82 402 −0.013 0.789 442 −0.019 0.85 improve finances G1: Saving early for child’s 847 −0.059* 0.556 401 −0.056 0.548 442 −0.065 0.567 education B18: Does not feel nervous 876 0.079** 0.69 419 0.063 0.592 453 0.094*** 0.782 going into a bank B17: People at the bank are 874 0.036 0.816 419 0.061 0.776 451 0.027 0.851 there to help I3: Purchasing accident 849 −0.008 0.519 402 0.116** 0.411 443 −0.069 0.621 insurance in dangerous job E3: Understanding details of 848 0.061** 0.718 402 0.112** 0.703 442 0.019 0.731 hire purchase contracts is important E23: Conservative spending 847 0.028 0.652 402 0.058 0.642 441 0.001 0.66 to avoid large interest pay- ments Overall financial attitudes 881 −0.009 0.674 423 0.011 0.629 454 −0.023 0.717 and perceptions Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 478  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 15A.6  Heterogeneous effects by education level: savings behavior FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN B1: Has bank account 879 0.023 0.554 422 0.027 0.42 453 0.063 0.676 B4: Has Mzansi account 459 −0.021 0.342 159 −0.22 *** 0.42 297 0.048 0.308 B16: Withdraws money as 873 −0.008 0.455 420 −0.029 0.454 449 0.005 0.46 soon as it is deposited C1: Saves money in any way 851 0.129*** 0.733 405 0.112 ** 0.715 442 0.124*** 0.75 C12: IST (monthly savings) 822 0.686*** 4.323 387 0.765*** 4.059 432 0.624*** 4.571 C16: IST (expected future 851 0.463*** 5.701 405 0.34 5.415 442 0.412* 5.985 monthly savings) Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. IST = inverse sine transformation. TABLE 15A.7  Heterogeneous effects by education level: borrowing behavior FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN D1: Applied for a loan in 851 −0.027** 0.051 404 −0.022 0.056 443 −0.017 0.046 past 6 months E7-E8: Bought an Item on 846 0.000 0.069 400 0.038 0.057 442 −0.018 0.081 hire purchase in past 6 months E9: Number of hire purchase 846 0.015 0.049 400 0.073** 0.043 442 −0.016 0.055 contracts in past 6 months Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. TABLE 15A.8  Heterogeneous effects by education level: gambling behavior FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN L7: Has gambled in past 6 848 −0.013 0.11 404 0.001 0.088 440 −0.014 0.132 months L9: Prefers saving money 847 0.077* 0.572 402 0.143*** 0.557 441 0.024 0.587 versus gambling Risk averse 844 −0.038* 0.908 400 −0.042 0.894 440 −0.051** 0.919 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 15.  Evalution of Old Mutual’s On the Money program  ◾ 479 TABLE 15A.9  Heterogeneous effects by education level: financial planning FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN F9: Has written a budget 848 0.004 0.155 402 0.06* 0.106 442 −0.017 0.204 F7: Has thought about what 847 0.007 0.656 401 −0.003 0.58 442 0 0.727 to do in financial emergency F15: Has made financial 846 −0.032 0.122 400 0.05 0.078 442 −0.069** 0.162 plans for retirement Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. TABLE 15A.10  Heterogeneous effects by education level: well-being FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN M2: Feels confident 848 −0.019 0.625 404 −0.061 0.69 440 −0.009 0.566 M1: Feels stressed 847 −0.033 0.379 404 −0.034 0.331 439 −0.022 0.424 M3: Feels overwhelmed 848 −0.002 0.471 404 −0.076 0.52 440 0.017 0.426 Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. TABLE 15A.11  Heterogeneous effects by education level: numeracy skills FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN K1: Addition 639 0.051 0.69 226 0.046 0.489 410 0.018 0.805 K2: Multiplication 744 0.028 0.826 311 0.04 0.692 429 0.003 0.925 K3: Division 375 −0.066 0.621 70 0.103 0.381 303 −0.072 0.672 K6: Fractions 482 0.04* 0.91 135 0.008 0.936 345 0.06 ** 0.899 K4: Percentages 288 0.015 0.643 49 −0.237 0.448 239 0.028 0.683 Basic math skills 749 0.014 0.51 316 0.007 0.318 429 0.014 0.653 K5: Percent calculation 364 0.031 0.797 77 0.208 0.609 284 −0.034 0.85 K7: Interest calculation 345 0.032 0.363 72 0.17* 0.227 270 0.026 0.405 K8: Compounding calculation 130 −0.012 0.038 17 0 0 113 −0.013 0.044 I1: Insurance calculation 850 0.01 0.494 403 0.025 0.417 443 −0.016 0.563 Basic numeric problem- 850 0.003 0.25 403 0.021 0.143 443 −0.022 0.35 solving skills Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 480  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 15A.12  Heterogeneous effects by education level: financial awareness FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN H1: Bank account opening 848 0.03 0.873 402 0.061 0.789 442 0.009 0.95 requirement G8: Security of savings in 848 0.013 0.939 402 0.034 0.939 442 −0.001 0.938 banks H4: Interest rates on savings 845 0.02 0.669 399 0.056 0.605 442 0.011 0.731 Awareness of savings 848 0.023 0.825 402 0.053* 0.775 442 0.006 0.873 accounts H5: Interest rate on loans 833 −0.049 0.743 396 −0.033 0.689 434 −0.065 0.791 E25: Loan terms and 846 0.063 0.69 401 −0.015 0.673 441 0.084 * 0.708 interest payments E24: Installment payments 848 0.001 0.743 402 0.003 0.752 442 −0.007 0.735 H2: Stop order 848 0.01 0.225 402 0.008 0.065 442 0.032 0.377 Awareness of loans and 848 0.009 0.596 402 −0.006 0.541 442 0.013 0.648 interest rates I2: Accident insurance 852 −0.004 0.455 405 −0.032 0.454 443 0.005 0.456 I4: Insurance premiums 851 −0.062 0.554 404 0.007 0.508 443 −0.096 0.594 Awareness of insurance 852 −0.032 0.504 405 −0.012 0.48 443 −0.045 0.525 G7: Credit versus cash price 843 0.036 0.743 398 0.123*** 0.712 441 −0.002 0.776 differences G5: Opportunity cost of 848 0.023 0.429 402 0.041 0.435 442 0.017 0.427 spending Awareness of purchasing 848 0.031 0.583 402 0.084* 0.569 442 0.009 0.6 on credit H3: Composition of a 844 0.067** 0.488 400 0.113** 0.424 440 0.048 0.548 budget J3: Importance of tracking 852 0.009 0.628 405 0.032 0.574 443 −0.012 0.678 expenditure G2: Options of tracking 846 0.102** 0.307 401 0.123*** 0.241 441 0.107** 0.371 income and expenditure Awareness of budgeting 852 0.061*** 0.471 405 0.091*** 0.41 443 0.048 0.53 F21: Awareness of separating 848 −0.017 0.696 402 −0.068 0.663 442 0 0.723 business and household accounts Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 15.  Evalution of Old Mutual’s On the Money program  ◾ 481 TABLE 15A.13  Heterogeneous effects by survey group: financial awareness FULL SAMPLE BSSP SAMPLE WDB SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN H1: Bank account opening 848 0.03 0.873 436 0.041 0.859 412 0.001 0.885 requirement G8: Security of savings in 848 0.013 0.939 436 0.05** 0.928 412 −0.031* 0.95 banks H4: Interest rates on savings 845 0.02 0.669 434 0.09** 0.632 411 −0.071 0.704 Awareness of savings 848 0.023 0.825 436 0.063*** 0.805 412 −0.032 0.845 accounts H5: Interest rate on loans 833 −0.049 0.743 431 0.011 0.735 402 −0.139*** 0.751 E25: Loan terms and 846 0.063 0.69 434 0.093 0.673 412 0.012 0.705 interest payments E24: Installment payments 848 0.001 0.743 436 0.045 0.735 412 −0.071 0.751 H2: Stop order 848 0.01 0.225 436 −0.004 0.253 412 0.008 0.199 Awareness of loans and 848 0.009 0.596 436 0.038 0.595 412 −0.043 0.596 interest rates I2: Accident insurance 852 −0.004 0.455 437 −0.009 0.424 415 0.01 0.485 I4: Insurance premiums 851 −0.062 0.554 437 −0.008 0.556 414 −0.138* 0.551 Awareness of insurance 852 −0.032 0.504 437 −0.009 0.49 415 −0.063 0.517 G7: Credit versus cash price 843 0.036 0.743 433 0.088** 0.748 410 −0.039 0.738 differences G5: Opportunity cost of 848 0.023 0.429 436 −0.1 0.474 412 0.174* 0.387 spending Awareness of purchasing 848 0.031 0.583 436 −0.001 0.606 412 0.066* 0.561 on credit H3: Composition of a 844 0.067** 0.488 434 0.07* 0.488 410 0.058 0.488 budget J3: Importance of tracking 852 0.009 0.628 437 0.101** 0.572 415 −0.104* 0.682 expenditure G2: Options of tracking 846 0.102** 0.307 436 0.12* 0.297 410 0.077* 0.317 income and expenditure Awareness of budgeting 852 0.061*** 0.471 437 0.098*** 0.451 415 0.013 0.491 F21: Awareness of separating 848 −0.017 0.696 436 0.047 0.699 412 −0.103 ** 0.693 business and household accounts Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 482  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES TABLE 15A.14  Heterogeneous effects by survey group: numeracy skills FULL SAMPLE LOW EDUCATION SAMPLE HIGH EDUCATION SAMPLE CONTROL CONTROL CONTROL OUTCOME MEASURE N COEFF MEAN N COEFF MEAN N COEFF MEAN K1: Addition 639 0.051 0.69 326 0.097* 0.661 313 −0.028 0.716 K2: Multiplication 744 0.028 0.826 370 0.023 0.827 374 0.037 0.826 K3: Division 375 −0.066 0.621 199 −0.083 0.638 176 −0.03 0.605 K6: Fractions 482 0.04* 0.91 243 0.059 0.865 239 0.021 0.952 K4: Percentages 288 0.015 0.643 149 −0.071 0.707 139 0.151 0.584 Basic math skills 749 0.014 0.51 373 0.01 0.512 376 0.016 0.508 K5: Percent calculation 364 0.031 0.797 192 −0.028 0.798 172 0.108 0.797 K7: Interest calculation 345 0.032 0.363 188 −0.01 0.406 157 0.085 0.325 K8: Compounding calculation 130 −0.012 0.038 76 −0.019 0.073 54 0 0 I1: Insurance calculation 850 0.01 0.494 436 0.058 0.498 414 −0.067 0.49 Basic numeric problem- 850 0.003 0.25 436 0.017 0.252 414 −0.023 0.247 solving skills Note: *, **, and *** indicate significance at the 10%, 5%, and 1% levels, respectively. 15.  Evalution of Old Mutual’s On the Money program  ◾ 483 ANNEX B: BASELINE SURVEY Old Mutual FEIA Project Baseline Survey X Section X: Introduction X1 Interviewer name X2 Date of interview X3 Respondent details X4 Name X5 Surname X6 What is your sa id number? X7 Burial society or a WDB group? 1. Burial society (go to x7) 2. WDB group (go to x8) X8 What is your burial society name? X9 What is your WDB group number? X10 Date of birth (dd/mm/yyyy) X11 What is the highest level of education you have attained? 1. Grade 1 2. Grade 2 3. Grade 3/std 1 4. Grade 4/std 2 5. Grade 5/std 3 6. Grade 6/std 4 7. Grade 7/std 5 8. Grade 8/std 6 9. Grade 9/std 7 10. Grade 10/std 8 11. Grade 11/std 9 12. Grade 12/std 10 13. Tertiary 12. Grade12 13. Tertiary 16. No school 777. Other (specify) 888. Refused 999. Don't know X12 What is your marital status? 1. Married 2. Single 3. Divorced 4. Common law marriage 5. Widowed 888. Refused X13 How many people rely on you to provide financially for # them? A Income and assets A1 Are you currently employed by somebody who pays you 1. Yes (proceed to A2) a salary or wage? 2. No (skip to A4) 3. Retired (skip to A6) 888. Refused (skip to A6) A2 How much was your take-home pay from your job last # month? A3 If you were paid in a way besides money, what was the approximate cash value of your payment? A4 Do you earn an income from work you do for yourself? 1. Yes (proceed to A5) For example, you might buy and sell goods, be a 2. No (skip to A6) commercial farmer or work for yourself as a hairdresser. 484  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES A5 How much did you earn from your business last month? 1. # 2. Don't know A6 If you accepted payment in a way besides money, what was the approximate cash value of your payment? B Banking B1 Do you currently have a bank account? 1. Yes (skip to B4) This includes at the post office 2. No (proceed to B2) B2 Why don't you have a bank account? 1. Don't know how to open one (can choose multiple) 2. Too difficult to get transport to the nearest (do not prompt) bank 3. Worried about fees 4. Worried about documentation requirements 5. Don't have an income 6. Other B3 If the problems you mentioned were resolved, would 1. Yes (skip to B15) you be interested in getting a bank account? 2. No (skip to B15) B4 Do you have an mzansi account? 1. Yes (continue to B5) 2. No (skip to B6) B5 Do you have any other accounts? 1. Yes 2. No (skip to B14) B6 What type of account do you have? 1. Savings (can choose multiple) 2. Cheque account 3. Investment 4. Group account 5. Fixed deposit/call account 6. Other B7 Please list the reasons why you have this account 1. Security reasons -to keep money safe 2. Gain access to other financial services such as receiving payments 3. To earn interest/a return on your savings 4. Save money for a specific purpose 5. To be able to borrow money 6. Other B8 Do you earn interest from keeping money in this 1. Yes account? 2. No 999. Don't know B9 How often do you withdraw money from your account? Less than once a year Once a year Once a month Several times a month Once a week More than once a week B10 How often do you deposit money into this account? 1. Never 2. Once a year 3. Once a month 4. Two times a month 5. Once a week 6. Other B11 (if yes to A3 and respondent has a SME) 1. Yes (fill out questions b9-b13 for this do you have a separate bank account for your business? account) 2. No B12 Do you ever use your cell phone to carry out bank 1. Yes transactions (including Mpesa and take it easy)? 2. No B13 Would you be interested in using cell phone banking 1. Yes 2. No I am now going to ask you some questions about how you feel about banks. For the following statements please choose whether you agree or disagree. There is no right or wrong answer. 15.  Evalution of Old Mutual’s On the Money program  ◾ 485 B14 I could borrow money from the bank if I needed it 1. Agree 2. Disagree B15 Banks charge a lot of money for their services 1. Agree 2. Disagree B16 As soon as money is deposited into my account I 1. Agree withdraw it 2. Disagree B17 I feel like the people who work at the bank are there to 1. Agree help me 2. Disagree B18 I feel nervous going into a bank 1. Agree 2. Disagree C Saving C1 Do you save money in any way? (including with a 1. Yes (skip to C3) stokvel, at home, with friends and family, or with an 2. No (proceed to C2) insurance account?) C2 Why do you not save? 1. Not enough income (continue to next section) 2. Do not have access to a safe and convenient place to save (continue to next section) 3. Afraid family or friends will ask to borrow it (continue to next section) 4. Other (specify) (continue to next section) C3 What are the ways you have saved money in the past six 1. Deposit money in formal bank/post office months savings account (can choose multiple) 2. Save with a savings club (stokvel, umgalelo) (prompt for each) 3. Save with a burial society 4. Keep savings at home 5. Buy assets such as cattle or gold 6. Leave savings with friends or family for safe-keeping 7. MFI or SHG 8. Savings policy 9. Investment policy 10. Education policy 11. Other (specify) C4 How much you have saved this way? C5 Why do you save this way? 1. Earn interest 2. Convenient 3. Trust it will be safe 4. No other options 5. Didn't know about other options 6. Helps me to manage my money/save more 7. Other C6 How many years have you been saving this way? 1. # 2. Don't know C7 How much interest do you earn on your savings each 1. # month? (only ask for applicable savings methods) 2. Don't know C8 How many deposits have you made in the last 6 months? # C9 When you make deposits, how much do you normally deposit at a time? C10 What are you saving for this way? 1. A specific durable good, such as a TV or fridge 2. A wedding or other special celebration 3. Christmas 4. Saving for a funeral 5. School fees 6. To be able to put down a deposit on a property or renovate current home 486  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 7. For retirement 8. In case of emergency 9. Other (specify) C11 After you get your income, when do you put money 1. Immediately after getting income aside for saving? 2. After taking take of other costs 3. Varies C12 How much do you save each month C13 Have you changed your monthly savings in the past 6 1. Yes months 2. No C14 By how much each month? # (positive or negative) C15 Where did you change your savings? 1. Deposit money in formal bank/post office savings account 2. Save with a savings club (stokvel, umgalelo) 3. Save with a burial society 4. Keep savings at home 5. Buy assets such as cattle or gold 6. Leave savings with friends or family for safe-keeping 7. MFI or SHG 8. Savings policy 9. Investment policy 10. Education policy 11. Other (specify) C16 How much to you plan on saving each month in the # future? C17 Do you wish you saved more than you do? 1. Yes (continue to C14) 2. No (skip to C15) 888. Refused C18 Why don't you save more? 1. Afraid that friends/family will ask to borrow money 2. Afraid it will be stolen 3. Not enough income 4. Other (If checked yes to stockvel, continue. If not, proceed to next section) The following questions are about your participation with your stockvel C19 Is your stokvel a savings stokvel or an on-lend stokvel? 1. Savings (skip to next section) 2. On - lend (continue) C20 How much interest does your stokvel charge when it lends money? D Borrowing and debt management D1 Have you applied for a loan in the past 6 months? 1. Yes (proceed to D20) 2. No (skip to D23) D2 Was the loan application approved? 1. Yes 2. No D3 When you have borrowed money in the past 6 months, 1. Personal loan from a bank where did you get it from? 2. Microlender (can select multiple) 3. Credit card (prompt for each) 4. Home loan/bond 5. Credit with local grocer/ general dealer 6. Store card (e.g. Edgars, truworths) 7. Loan with mashonisa stockvel 8. Loan form family or friend 9. Loan from a family member or friend 10. Other (specify) 11. Did not borrow in past 6 months (skip to d19 loans and credit) 15.  Evalution of Old Mutual’s On the Money program  ◾ 487 D4 Why do you borrow from this lender? 1. Convenient/accessible (try not to prompt) 2. Low interest rate 3. Size of repayments suit budget 4. Flexibility of repayment schedule 5. Only provider(s) willing to lend to respondent 6. Loyalty to lender 7. Worried about the questions the bank might ask 8. Don't know what borrowing from other lenders would involve 9. Lender doesn't ask for documents or collateral 10. Didn't want to owe money to other lenders 11. Other D5 Do you currently have a loan with this lender? 1. Yes 2. No D6 What did you use the money from the loan for? 1. Consumption expenditure such as personal (can choose multiple) items, celebrations 2.investment expenditure such as building or renovating a house, buying a car etc. 3. Business expenditure such as enhancing a business or productive activity 4. To pay for unforeseen expenses, like a funeral 5. To pay off another debt 6. Other (specify) D7 How much did you borrow? 1. # 2. Don't know D8 How much did/do you have to pay back in total for you 1. # most recent loan from here? 2. Don't know D9 How long did you have to pay back the loan in months? # D10 How much did you have to pay in interest in rand? 1. # 2. Don't know D11 How much did you have to pay in service fees in rand? 1. # 2. Don't know D12 (only ask if yes to d5) 1. Yes are you currently paying off this loan? 2. No (skip to D18) D13 How many payments did you miss in the last 6 months? # D14 Do you borrow from this source to repay other loans? 1. Yes 2. No The following questions are about your past experiences with loans and credit D15 Have you ever compared different financing options to 1. Yes (proceed to D24) help you decide which is the best source to borrow 2. No (skip to next section) from? 3. Don't have multiple options (skip to next section) 4. Don't know how to compare loans with different terms and interest rates (skip to next section) 888. Refused D16 Which financing options did you compare? 1. Personal loan from a bank or microlender (can choose multiple) 2. Loan with mashonisa stockvel 3. Credit card 4. Home loan/bond 5. Credit with local grocer/ general dealer 6. Store card (e.g. Edgars, truworths) 7. Hire purchase 9. Loan from a family member or friend 488  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 10. Other (specify) E Hire purchase agreements We are now going to ask you some questions about how you feel about hire purchase agreements. For the following statements please choose whether you agree or disagree. There is no right or wrong answer E1 Hire purchase is a convenient way to buy things when 1. Agree you don't have enough money 2. Disagree E2 Hire purchase should only be used to buy things you 1. Agree absolutely need when you don't have enough money 2. Disagree E3 Hire purchase is a helpful product because it doesn't take 1. Agree very long and has few paperwork requirements 2. Disagree E4 It's a good idea to take out a longer hire purchase 1. Agree agreement if you have less money because your 2. Disagree payments will be smaller E5 I am now going to ask you about your past experience with hire purchase agreements E6 Have you bought a large item like furniture or a fridge in the last 6 months? E7 Were you offered a hire purchase agreement to pay for it? E8 Did you accept the hire purchase agreement or pay for the item in another way? E9 How many hire-purchase agreements have you taken out # since last winter? (ask all of the following for each item) E10 For the most recent purchase, did you attempt to pay for the item in another way before deciding upon hire purchase? E11 Why did you chose to take out your most recent hire 1. Convenient purchase instead of using money from savings or a loan 2. Low interest rate from a different source? 3. Size of repayments suit budget (can choose multiple) 5. Like structure of repayment schedule 4. Flexibility of repayment schedule 5. Only provider(s) willing to lend to respondent 6. Loyalty to lender 7. Other (specify): E12 Would you consider your purchase a necessity, or would 1. Necessity you have been fine without it or with a less expensive 2. Could have gotten less expensive item option? 3. Would have been fine without it E13 What was the cash price on the item? E14 What was the credit price? E15 How long is your plan to pay back the hire purchase? 1. 36 months 2. 24 months 3. 12 months 4. Other E16 How much do you pay in each instalment? E17 What initiation fee did you have to pay? E18 How much in monthly service fees does the retailer charge you as part of the agreement? E19 Did you agree to let the shop set up a debit order? E20 Did you get a detailed quote/contract when you took this 1. Yes (proceed to E19) hire purchase? 2. No (skip to next section) E21 Did you read the contract or have somebody besides the 1. Yes seller explain it to you? 2. No E22 Has somebody else in your household bought something on hire purchase over the last 6 months? 15.  Evalution of Old Mutual’s On the Money program  ◾ 489 E23 Nsako recently got married. He and his wife are 1. Big TV considering buying a TV. They do not 2. Small TV have enough savings and will need to take out a loan. Nsako has two options: (1) he can buy the TV on hire purchase and buy a big TV, or (2) he can take a loan only from a relative and buy a smaller TV. What would you advise Nsako and his wife? E24 Buying a TV or other appliance that will last a long time -1 agree from a shop and paying on instalments is a cheaper way -2 disagree to buy things because it spreads the payments out over a -3 don’t know long time. E25 Ntombi is in the market for a new bed. She visits two -1 agree shops. One shop offers her a 12 month hire purchase but -2 disagree she will have to pay back quite a lot each month. The -3 don’t know hire purchase from the other shop means that she will have to pay back less each month but over 36 months instead of 12. Ntombi decides to go with the option of paying less each month but over 36 months. Do you agree or disagree with Ntombi’s decision? F Financial planning The following questions are about your expected and unexpected costs. We are going to start with questions about your expected costs first. F1 In the past six months did you have to deal with an 1. Yes (continue to F2) expected large expenditure, for example a wedding; or a 2. No (skip to F3) drop in income, like a retirement? F2 Where did you get the money to deal with this from? 1. Used savings (can choose multiple) 2. Cut consumption 3. Borrowed from family members 4. Borrowed from friends 5. Borrowed from professional moneylenders 6. Claimed from insurance 7. Other (specify): F3 Are you expecting to have a large expenditure or loss of 1. Yes income in the next year? 2. No (skip to F5) 3. Haven't thought about it (skip to F5) F4 How are you going to deal with this? 1. Going to use savings (can choose multiple) 2. Currently cutting consumption 3. Will cut consumption afterwards 4. Will borrow from family members/friends 5. Will borrow from professional moneylenders 6. Haven't thought about it 7. Other (specify): The following questions are about your unexpected costs F5 In the past six months, did any unexpected event outside 1. Yes (proceed to F6) your control, like a funeral or a lost job, require a large 2. No (skip to F7) expenditure or cause loss in income? F6 How did you deal with the shortfall? 1. Used savings (can choose multiple) 2. Cut consumption 3. Borrowed from family members 4. Borrowed from friends 5. Borrowed from professional moneylenders 6. Claimed from insurance 7. Other (specify): F7 If you suddenly needed 1000r where would you get it? 1. Would use savings (can choose multiple) 2. Would cut consumption 3. Would borrow from family members/friends 4. Would borrow from formal credit sources 490  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES 5. Haven't thought about it 6. Other (specify) The following questions are about your finances F8 How much money do you need to cover your costs for # yourself and those you are responsible for in the average month? F9 How often do you manage not to spend more than your 1. Always income? 2. Mostly 3. Rarely 4. Never F10 Have you ever looked at your spending and income and 1. Yes written up a plan so they are balanced? 2. No 3. Want to but don't know how F11 (if yes to a3 and respondent has a sme) have you ever looked at your business's profits and costs and made sure that they balance, or that you make a profit? F12 Do you regret anything large that you bought in the last 1. Yes 6 months? 2. No F13 If you have insurance, what types do you have?? 1. Health insurance/medical aid (can choose multiple) 2. Funeral cover 3. Life insurance 4. Property insurance 5. Other (specify) 6. No insurance F14 If you could have one more in addition, which one 1. Health insurance/medical aid would you want? 2. Funeral cover 3. Life insurance 4. Property insurance 5. Other (specify) 6. No insurance F15 Do you have any financial plans for retirement beyond 1. Yes (proceed to F15) government grants? 2. No (skip to F16) F16 What are they? 1. My children will take care of me 2. Savings methods 3. Will rely on grant money 4. Provident funds 5. Retirement endowments 6. Other (specify) F17 Do you feel that you are on track to be financially secure 1. Yes as you get older? 2. No F18 What are your long term goals for your household that 1. Nicer home require a large amount of money? 2. Job training (can choose up to two) 3. More children 4. Get out of debt 5. Education 6. Marriage 7. Funeral 8. Other (specify) 9. None (end section here) 10. Don't know (end section here) F19 How likely do you feel that you will achieve this in the 1. Very likely next 10 years? 2. Somewhat likely 3. Unlikely 4. Definitely not 5. Haven't thought about it 15.  Evalution of Old Mutual’s On the Money program  ◾ 491 F20 How do you expect to pay for it? 1 savings 2. Loans 3. Remittances 4. Future income 5. No F21 It is better to combine the money for your personal, household and business affairs because it simplifies your money matters. Do you agree? F22 A friend of yours tells you that her expenses are more -1 borrow money from a friend to cover extra than her income. Which one of the following would you expenses recommend? -2 buy with store credit instead of cash -3 write down expenses and income G Financial attitudes and perceptions G1 Busisiwe has a very bright child who is currently in a. Buy child life insurance policy secondary school, but will probably do well in varsity. b. Borrow money from a moneylender She is worried how her family will pay for the child’s c. Open a savings account in a bank education. If busisiwe comes to you for advice what d. Save at home would you suggest? e. Discontinue education f. Other G2 Nonhlahla has two sons. Her husband and two sons are a. Open a savings account earning members of the household and contribute b. Start making a household budget towards household income. However nonhlahla does not c. Buy life insurance for her husband and sons know what is the household’s total income and a. Open a savings account expenditure. How do you think nonhlahla can track her b. Start making a household budget income and expenditure? c. Buy life insurance for her husband and sons G3 Only people who make a lot of money can be financially 1. Agree secure. 2. Disagree Do you agree? 98. Do not know 99. Refuse to answer G4 Once you are 40, it’s too late to improve your nancial 1. Agree situation 2. Disagree 98. Do not know 99. Refuse to answer G5 When i think about buying something, I also think about 1. Agree what i could use the money for instead 2. Disagree 98. Do not know 99. Refuse to answer G6 I tell my family and friends what I am saving for 1. Agree because I know they will help me save 2. Disagree 98. Do not know 99. Refuse to answer G7 Which is higher, the credit price or the cash price+b146 1. Credit 2. Cash G8 Saving at home instead of a bank is safer 1. Agree 2. Disagree 98. Do not know 99. Refuse to answer G9 Pumeza has just gone into town to collect money from 1. Uniforms her old age grant. Once she gets her money, she has 2. Groceries several errands to do in town. She wants to buy school 3. Savings uniforms for her grandson, buy groceries for the family, and put some money into her mzansi account for savings. Which should she do first? H Financial knowledge 492  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES H1 If you have r50can you open a bank account? H3 Do you know what a 'stop order' is? H4 Noncebo is preparing a budget for her household. Which a. Income only of the following needs to be included in the budget? b. Expenses only A. Income only c. Both b. Expenses only c. Both H5 Do you prefer to receive high or low interest rates on your saving? H6 Do you prefer to take a loan with high or low interest rates? I Insurance literacy I1 Assume you have purchased a medical insurance policy a. R. 3000 cover and R. 95 premium and suffer an accident which results in R. 3500 of b. R. 2800 cover and R. 80 premium hospital fees. Would you be better off if you had purchased an insurance policy with I2 Sibongile recently bought accident insurance with R. # 10,000 cover. The next day, he met with an accident and died. His funeral cost R. 5,000 how much do you think the funeral policy will pay for? I3 Tshepo does plastering on tall buildings. It is a a. Quit job dangerous job and he is worried that if he gets injured b. Purchase health/life/ accident insurance his family’s income will become inadequate to meet c. Increase savings their needs. If Tshepo comes to you for advice what would you suggest? I4 Vuyiswa is 20 years old and Noluthando is 30 years old. 1. Vuyiswa If they were to buy funeral insurance of r3000 for 20 2. Noluthando years, who between the two to your mind will have to pay higher premium? J Old Mutual Questions True/False J1 You can never be nancially secure if you don’t earn a 1. True lot of money 2. False J2 Earning well guarantees you will become financially 1. True secure 2. False Agree/Disagree J3 I always check my payslip 1. Agree 2. Disagree J4 I don't need to track how much i spend, as long as i have 1. Agree some left over at the end of the month 2. Disagree J5 I know exactly how much I owe every month 1. Agree 2. Disagree Multiple Choice J6 When you receive a windfall of such as a bonus or tax 1. Spend it refund etc., do you... 2. Save it 3. Pay off debt 4. Party J7 If you buy something like a fridge, bed, tv, do you 1. Use lay-by 2. Credit / hire purchase 3. Pay cash 4. Borrow from family or friends 5. Put it on credit to earn points but pay it off before the interest due date 15.  Evalution of Old Mutual’s On the Money program  ◾ 493 K Mathematical ability The following questions will test your mathematical ability. There is no penalty or reward for a right or wrong answer. K1 If you have r48, and I give you r58, how much will you # have? K2 If you have four friends and would like to give each one # (if wrong answer to both H1 and H2, skip to five sweets, how many sweets must you give away? next section) K3 What is 63 divided by 7? K4 If you have one half of 100, what percent of 100 do you have? K5 What is 10% of 400? K6 What is 400 divided by 10? K7 Suppose you have r100 in a bank account, and you earn 10% interest on what is in the account at the end of the year. If you do not spend it, how much do you have at the end of one year? K8 How much do you have at the end of two years? L Gambling L1 In the past 6 months (if yes to l1) ->number How much do you have you played: of times played in last spend each time you 30 days play on average? L2 The national lottery / 1. Yes lotto 2. No L3 Scratch cards 1. Yes 2. No L4 Fafi/Ichina 1. Yes 2. No L5 Dice games for money 1. Yes 2. No L6 Card games for money 1. Yes 2. No L7 (only ask if hasn't played in last 30 days) 1. Yes have you gambled in the past 6 months? 2. No L8 How often has somebody in your house gambled in the 1. Daily last 6 months? 2. Weekly 3. Monthly 4. Never L9 Fifi and Mpho are discussing the best way to play the 1. Fifi lottery. Fifi believes that the chances of winning the 2. Mpho lottery are so small that she should rather save her money. Mpho thinks that Fifi should play as she can win big and she would never know unless she tries. Do you agree with Fifi or Mpho? L10 Which would you prefer? 1. R50 guaranteed 1. R50 for sure 2. R125/r0 coin toss 2. Play game in which a coin is spun and if it lands on heads you get r125 but if it lands on tails you get nothing? M Wellbeing M1 How often do you feel stressed or worried about your 1. Very often finances? 2. Often. 3. Sometimes 4. Rarely 5. Never 494  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES M2 How often do you feel confident about your ability to 1. Very often handle your financial problems? 2. Often. 3. Sometimes 4. Rarely 5. Never M3 How often do you feel difficulties are piling up so high 1. Very often that you cannot overcome them? 2. Often. 3. Sometimes 4. Rarely 5. Never CHAPTER 16 F inancial management and vocational training in Uganda Impact and network effects in informal industrial clusters FRANCISCO CAMPOS, MARKUS GOLDSTEIN, OBERT PIMHIDZAI, MATTEA STEIN, AND BILAL ZIA T he Russia Financial Literacy and Education Trust Fund grant has cofinanced the ongoing impact evaluation of a comprehensive finan- cial management and vocational training program for small-scale industries in Uganda. With the contribution of this grant, we are also identifying the effects of the intervention on entrepreneurs’ business networks. The grant was a contributor for the first two post-intervention surveys on both firms participating in the training and the control group. We also inter- viewed business owners of firms that interact with our sample (treatment and control) to share information on new products, new designs, training programs, credit opportunities, etc. This chapter summarizes the work up to now—one baseline survey including enterprise, business owner households, and employee surveys; technical and financial management training interventions; and two follow-up surveys on both businesses that are part of the study and their contact networks. The impact evaluation will comprise an endline, replicating inter- views from the baseline. The analysis from the first of these follow-ups is also documented in this chapter but is still very preliminary. 16.1 INTRODUCTION Identifying the determinants of entrepreneurship is an important research and policy goal, especially in developing countries where lack of capital All opinions expressed in this chapter are those of the authors and do not necessarily represent the views of the World Bank, the Russia Financial Literacy and Education Trust Fund, or any other research partners or sponsoring institutions.   ◾ 495 496  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES and supporting infrastructure often imposes stringent constraints on business growth. While much of the literature has focused on access to physical capital and external finance (e.g., Banerjee et al. 2010; de Mel, McKenzie, and Wood- ruff 2008), a number of recent papers argue that “managerial capital” or busi- ness skills are another important driver of firm growth and a key determinant of productivity (e.g., Bloom et al. 2010; Bruhn, Karlan, and Schoar 2010). This interest has spawned a limited number of experimental evaluations of business training programs across different settings. For example, Karlan and Valdivia (2010) evaluate a business education program for female microentrepre- neurs in Peru and find that it improves recordkeeping, though not profits. In the Dominican Republic, Drexler, Fischer, and Schoar (2011) show that a basic rules- of-thumb-based training, but not formal business training, leads to improvements in business outcomes. Finally, Bruhn and Zia (2011) focus on medium-scale enter- prises in Bosnia-Herzegovina and find that formal business and financial training improves business practices and investments, although not business survival. These and other studies are summarized in McKenzie and Woodruff (2012). But businesses do not operate in a vacuum. They interact with neighboring businesses, with close associates operating in other areas, and with businesses above and below in the supply chain. Importantly, many businesses are part of larger networks, some of which make joint production and sales decisions, and share costs, revenues, and working capital. While the above-mentioned impact studies identify positive effects of business education on targeted business segments, an outstanding research question is how, if at all, such enhanced knowl- edge spreads to other businesses and across business networks. Also, are the diffusion paths of observable and unobservable skills different? This latter ques- tion can help us understand the nature of market and network interactions—e.g., are neighboring businesses considered competitors or partners? Are network members more likely to benefit from enhanced skills than others? These insights can help us better design and/or adjust training programs for target populations. Further, these questions have substantial policy relevance, as any identified positive externalities could justify scaling down spending on financial literacy programs while still being able to reap many, if not all, of the benefits. For example, if we find network effects are significant, then policy can be designed to provide financial literacy for businesses to only a subset of those targeted, which can then spread the knowledge to others. The resources saved in doing so could be spent on expanding business training in other areas or on other development projects. On the direct impact of training, we are primarily interested in the following questions: 1. Does the training increase profitability and growth of enterprises? Specif- ically, does it reduce costs and wastage of raw material? Does it improve sales and revenues, profits, worker efficiency, quality of products, and employment for workshops that benefit from the training? 16.  Financial management and vocational training in Uganda  ◾ 497 2. Does the training lead to better quality jobs and higher earnings for workers and workshop owners; and further, does it improve their access to credit, diversification of earnings, and job security? 3. Do the benefits of training translate into higher standards of living for beneficiaries and their households? For example, does the training and accompanying increase in income lead to improvements in dietary diver- sity, children’s health and education, asset accumulation, investments, and possibly a reduction in domestic violence? 4. How do the above impacts vary by gender and age? 5. Are there heterogeneous effects of the intervention across other aspects, particularly by the size of the enterprise (self-employed versus firms with employees; micro versus small) or the sector of activity (industrial versus services)? On network effects, our research questions are the following: 6. How are informational networks between small and medium enterprises formed? Is there assortative matching on enterprise characteristics (such as sector, size, location, years in business) and business owner character- istics (such as gender, education, experience, family ties)? What are the observable differences between network and non-network small and medium enterprises, or between those in large versus small networks? 7. Are business and technical skills diffused through business networks, and if so, through what mechanisms (e.g., direct observation, secondary training, temporary or permanent worker exchange, etc.) and across what lines (neighbor to neighbor, exclusive to network members, etc.)? Since tech- nical training is highly observable while managerial training is not, does the observability of skills matter in how (and to whom) knowledge is diffused? Questions 1–5 are being addressed through a comparison of baseline and follow-up outcomes between treatment and control groups. Question 6 is addressed through the baseline survey, where in addition to recording baseline outcomes for each firm, we recorded the names and contact details of the closest self-identified business associates. The baseline asked a series of questions on how information, knowledge, working capital (such as labor), and sales orders are shared with these businesses. This enabled us to map out the network tree of each entrepreneur in our sample. Finally, we addressed Question 7 through follow-up surveys not only on the sample of businesses in our study (both treatment and control), but also on network members identified in the baseline. We are then able to compare outcomes between these groups to identify the extent and magnitude of any network effects. 498  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES This is a large-scale project, and several key partners are providing vital finan- cial, procedural, and analytical support. These include the Enterprise Skills and Linkages subcomponent of the Uganda Second Private Sector Competitiveness Program, the Private Sector Foundation Unit, the World Bank Africa Region Gender Practice, the Africa Financial and Private Sector Development, the Development Impact Evaluation Initiative in Finance and Private Sector Development, and the Development Economics Finance and Private Sector Development Research Group. 16.2 PROGRAM AND INTERVENTION DETAILS Our main implementation partner is the Katwe Small Scale Industries Association (KASSIDA), an association comprising members from nine diverse and informal industrial clusters—from metal fabricators, carpenters, machinery workers to barbershop owners, tailors, and caterers. KASSIDA member workshops operate in the outskirts of Kampala, and are bunched together in physical clusters. Such industrial clusters are quite common in developing countries, and previous research has documented the benefits of localization economies such as favor- able access to market information, low transaction costs, ease of monitoring, and division and sharing of capital and labor (Nadvi and Schmitz 1994; Ruan and Zhang 2009; Sonobe and Otsuka 2006). In the Ugandan context, such small-scale operators constitute 13 percent of the share of employment and absorb 30 percent of new labor market entrants. Recent estimates suggest that nearly 40 percent of Ugandan households own and operate a nonfarm household enterprise, and the sector has grown at an average annual rate of 8 percent over the last 10 years. Women are active participants, with almost 46 percent of enterprises managed by women. Despite the fast growth, most employees and entrepreneurs operating in these sectors do not have formal technical or financial management training; instead, they rely on old and sometimes obsolete production methods and rarely maintain financial or transactional records. As a result, productivity is low, enter- prise management is weak, and the risk of failure remains fairly high. Through the support of the World Bank and the Second Private Sector Competitiveness Program, KASSIDA recently launched a Workers Apprenticeship and Training Skills Program, with the goal of improving skills, enhancing capacity, and boosting output among workshop owners and workers who operate in the Katwe area. KASSIDA is also introducing a comprehensive financial literacy and managerial training program for its members and plans to implement both tech- nical and financial skills training together. Our study is evaluating the second wave of training, which was conducted in the first half of 2012, and targeted both workshop owners and their employees. In all, 384 workshop owners and approximately 1,026 workers were offered 16.  Financial management and vocational training in Uganda  ◾ 499 training in this wave. The financial management component of KASSIDA’s training was administered to workshop owners only. Business owners participated in 12  working sessions for a period of three months. The management training focused on four main areas: (1) marketing of products or services, (2) finance and bookkeeping, (3) customer service, and (4) taxes and regulations. The curriculum included the following key components: ◾◾ Defining working capital ◾◾ Defining fixed versus variable assets ◾◾ Importance of keeping financial records ◾◾ External finance options ◾◾ Business investment ◾◾ Steps for participating in public procurement as a supplier of products ◾◾ Cluster-specific marketing ◾◾ Process of registering an enterprise in Uganda as a private limited company Both workshop owners and workers in the treatment group were invited to participate in the technical training. This training was sector specific and often located in model workshops where participants practice more efficient produc- tion techniques. The objective of the training is to improve firms’ productivity by intro- ducing new methods and products with increased value added to customers and expanding efficiency in resource management. The financial management and technical trainings seek to address market failures such as (1) the lack of availability of profitable training opportunities, (2) the difficulty on the part of firm owners in valuing the benefits of investment opportunities before actually pursuing them, and (3) credit constraints that prevent firms from pursuing training opportunities. These lead to firms operating under insufficient quality standards and lacking managerial skills to grow their enterprises. 16.3 EVALUATION METHODOLOGY The core methodology is a randomized evaluation of the KASSIDA training program. Among the 735 workshops that have applied to participate in the second wave of training, we randomly assigned firms to treatment and control groups. 16.3.1 Clustering of firms by sector and location When randomly assigning workshops to participate in the training exercises, we needed to take into consideration the fact that firms in Katwe often operate side by side, sometimes not even separated by physical walls. For instance, in the catering sector, women who own separate firms produce and sell meals every day in the food market sitting very close to each other. In the carpentry industry, 500  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES namely in handicrafts, self-employed individuals working in single-table work- shops are frequently located in premises where they share a common roof. Therefore, it is not possible to randomize at the workshop level, since spill- overs are inevitable. Although one of our evaluation goals is to measure spillovers through network effects, we have to maintain our ability to statistically detect primary effects of the trainings. Our evaluation design balances both of these goals. To avoid spillovers through direct observation of neighbors—which would in effect reduce the possibility of estimating the real impacts of the training program—we initially grouped workshops into clusters based on the location and sector where firms operate. Firms in the same sector located within a radius of each other are thus part of the same cluster. Each cluster was pair-matched with a similar one on observable characteristics including size of the cluster and socio- economic features of the firm and business owner. Pairs of clusters of firms were then randomly allocated to either the treatment or control group, ensuring that all firms have the same 50 percent chance of being part of the program. To group the firms into clusters, we conducted a listing exercise of all workshops that applied to the program and recorded their global positioning system (GPS) coordinates. We mapped the location of the workshops which, along with their industry, allowed us to lump together firms into clusters. We have created clusters by adding all firms in the same sector, until there is no firm that is 20 meters or less away from one of the cluster members. This 20-meter distance—our cut-off—is seen by the local KASSIDA team as sufficient to avoid direct copying of techniques between businesses (as well as being a politically acceptable distance at which to randomize access to training). With this design, where randomization is done at a cluster level rather than at the individual firm level, we hope to mostly eliminate the direct copying of new techniques learned in training—the first source of potential spillover effects— because those operating side by side are either all receiving training or are part of our control group. 16.3.2 Spillover effects We aim to assess the size of the spillover effects that may still be prevalent by mapping the full spectrum of interrelations between workshops that are part of our study (both within and across treatment and control groups) and also with those outside our sample. We analyze the existing networks on (1) information sharing about designs, techniques, skills, and tools; (2) sharing of employees; (3)  sharing of inputs/materials; (4) sharing of contacts and orders to suppliers; (5) sharing of contacts and orders by customers; and (6) partnership and sharing of information on usage of financial services. To assess the spillover effects, we are tracking through surveys the main enterprise outcomes of interest for up to five network contacts for both treatment and control groups and comparing 16.  Financial management and vocational training in Uganda  ◾ 501 the change in performance for the contacts of each group that are outside our sample. Any difference in the outcomes of interest between these two parallel groups of firms can be attributed to spillovers from the training program as a result of the selection design. The data for the impact evaluation is obtained from a baseline and multiple follow-up surveys for the treatment and control groups. The baseline and follow-up surveys collect data for all workshop owners (735) and workers (1,965), as well as for network members identified by each workshop owner. 16.3.3 Data We are conducting multiple rounds of follow-up surveys—a full survey similar to the baseline on enterprises, employees, and business owners’ households will be conducted 15 months after the end of training; and a set of enterprise/business owner mini-surveys were completed three to six and nine months after the end of the training. The short surveys (30 minutes long) have two objectives: (1) to improve the statistical power of the analysis on the main outcomes of interest including sales, profits, product range, number of employees, and household assets/consumption; and (2) to track the change in behavior on intermediate outcomes—e.g., bookkeeping, safety measures, business/technical knowledge— with the objective of ascertaining the erosion of learning over time. We have currently completely two rounds of follow-up surveys with both the enterprise owners and their networks of businesses and are preparing the final round. 16.3.4 Outputs and dissemination The main outputs of this impact evaluation are baseline and follow-up data, policy notes, and research papers. After the baseline, we examined the labor practices and business networks that prevail in the region. After the follow-up surveys, we will prepare a pair of research paper and policy notes to present the results of the impact evaluation. In addition to extensive discussions within KASSIDA, these outputs will be presented to policy makers in Uganda, as well as other countries worldwide. 16.4 RESULTS We have analyzed thus far the results of the first follow-up for the core members of the study. Any of the following findings are very preliminary and are solely representative of the initial short-term effects. We find some short-term effects of the training on the core sample of members on financial literacy and technical knowledge, optimism, and adherence to standards and procedures, but limited short-term effects on core business 502  ◾  ENHANCING FINANCIAL CAPABILITY AND BEHAVIOR IN LOW- AND MIDDLE-INCOME COUNTRIES outcomes. Subgroup analysis on the dimension/size of the business, gender of the owner, and baseline financial literacy provide some additional insights. Trained entrepreneurs are significantly more likely to think that their business will be growing in the next few months, which is driven by smaller enterprises (those with below-median baseline profits). There is also initial evidence that this may be linked to those with below-median baseline financial literacy scores. There are several positive effects of training on adherence to standards and procedures: trained entrepreneurs are more likely to say they have a valid quality standard certificate, and to be able to show it, and are more likely to have a written safety code of conduct. Training also seems to have a positive effect on having written a business plan in the past six months, but not on providing receipts to customers. There is some evidence that there may be an effect on providing receipts for female-operated firms and large firms. In terms of business outcomes, there is some evidence that trained entre- preneurs may be consolidating their activities—they have opened fewer busi- nesses in the past six months and own fewer businesses in total. They are more likely to have cut costs in that period, but also less likely to have developed new products. There is no overall effect on the number of workers, which is driven by differential effects by gender: men in the treatment group are more likely to have any employees than are control group men, while women in the treatment group are less likely to have any employees than control group women. We do not see a statistically significant positive short-term effect on revenues and profits, even though the coefficient on past month profit is relatively large. As said, these short-term results are based on the first follow-up survey of the core sample of members of the small industry association. Further results will become available once the second follow-up survey—already completed—is analyzed. The network dimension of the project, including the survey of network members of the core sample, will be analyzed next, to determine which effects spread through the core sample’s business networks. This impact evaluation led to a spin-off study of a particular group of women who operate in male-dominated sectors (which were identified at baseline in the KASSIDA sample). Gaps in profitability between male- and female-owned enter- prises are widely documented, and research points to the role of sector choice as a determinant of those differences, finding that women entrepreneurs tend to cluster in low value-added industries with less growth potential. This potentially constrains the effects of interventions such as financial literacy programs that have shown to have mixed results for this type of firms. We hence conducted additional qualitative and quantitative data collection to interview 183 women, part of them from the initial KASSIDA data. We find that women working in male-dominated sectors earn nearly double those working in traditional indus- tries, and also work fewer hours per week. At the same time, large informational gaps exist among women operating in traditional industries about the returns 16.  Financial management and vocational training in Uganda  ◾ 503 available in male-dominated sectors. Linking quantitative analysis with qualitative insights from focus groups, we conclude that both normative and informational factors are predictive of a woman’s decision to “cross over” to a male-dominated sector. We advocate that informational campaigns combined with mentorship interventions may facilitate the growth of female entrepreneurship in nontradi- tional industries. Coupling these interventions with financial literacy programs for instance can potentially have more sustainable effects in driving female entrepre- neurship growth. REFERENCES Banerjee, Abhijit, Esther Duflo, Rachel Glennerster, and Cynthia Kinnan. 2010. “The Miracle of Microfinance? Evidence from a Randomized Evaluation.” Unpublished. Bloom, Nicholas, Aprajit Mahajan, David McKenzie, and John Roberts. 2010. “Why Do Firms in Developing Countries Have Low Productivity?” American Economic Review 100 (2): 619–23. Bruhn, Miriam, Dean Karlan, and Antoinette Schoar. 2010. “What Capital Is Missing in Developing Countries?” American Economic Review 100 (2): 629–33. Bruhn, Miriam, and Bilal Zia. 2011. “Business and Financial Literacy for Young Entrepreneurs: Evidence from Bosnia-Herzegovina.” Presentation at the World Bank conference “New Ideas in Business Growth: Financial Literacy, Firm Dynamics and Entrepreneurial Environment,” Washington, DC, March 30. de Mel, Suresh, David McKenzie, and Christopher Woodruff. 2008. “Returns to Capital in Microenterprises: Evidence from a Field Experiment.” Quarterly Journal of Economics, 123 (4): 1329–72. Drexler, Alejandro, Greg Fischer, and Antoinette Schoar. 2011. “Keeping It Simple: Financial Literacy and Rules of Thumb.” http://www.mit.edu/~aschoar/KIS%20DFS%20Jan2011. pdf. Karlan, Dean, and Martin Valdivia. 2010. “Teaching Entrepreneurship: Impact of Business Training on Microfinance Clients and Institutions.” http://karlan.yale.edu/p/ TeachingEntrepreneurship_revision_jan2010.pdf. McKenzie, David, and Christopher Woodruff. 2012. “What Are We Learning from Business Training and Entrepreneurship Evaluations around the Developing World?” Policy Research Working Paper 6202, World Bank, Washington, DC. Nadvi, K., and H. Schmitz. 1994. “Industrial Clusters in Less Developed Countries: Review of Experiences and Research Agenda.” IDS Discussion Paper 339, Institute of Development Studies, University of Sussex, Brighton. Ruan, J., and X. Zhang. 2009. “Finance and Cluster-Based Industrial Development in China.” Economic Development and Cultural Change 58 (1): 143–64. Sonobe, T., and K. Otsuka. 2006. Cluster-Based Industrial Development: An East Asian Model. Basingstoke, England: Palgrave Macmillan. The Russia Financial Literacy and Education Trust Fund was established in 2008 at the World Bank with funding provided by the Ministry of Finance of the Russian Federa- tion. The work supported by the Trust Fund is jointly managed by the World Bank and the Organisation for Economic Co‑operation and Development (OECD) and is directed toward improving public policies and programs to enhance financial knowledge and capabilities in low- and middle-income countries. This effort has focused on the re- view of national strategies for financial education, the development of methods for the measurement of financial knowledge and capabilities, methods for evaluating the impact and outcome of programs, and research applying these methods to programs in developing countries. The products of this program of work can be found at the Trust Fund website at: www.finlitedu.org