W®!7(E)QB OLMB E Z I1d\¶?I0Joem T76Me)C,E' eoor invc r-ma(oonz DV nCOCD Sn(DOIDE 23a30nE2M BQVQD0E3e 52034 March 2003 af R$XL -i a ( i~~~~~~~~~~~i DAVM D3 smkRIy ,UMUR( JR.$ RE1 R n B0GM WORLD BANK LATIN AMERICAN AND CARIBBEAN STUDIES CLOSING THE GAP IN EDUCATION AND TECHNOLOGY by David de Ferranti Guillermo E. Perry Indermit Gill J. Luis Guasch William E Maloney Carolina Sdnchez-Pdramo Norbert Schady THE WORLD BANK Washington, D.C. (© 2003 The International Bank for Reconstruction and Development / THE WORLD BANK 1818 H Street, NW Washington, DC 20433 Telephone 202-473-1000 Internet www worldbank org E-mail feedback@worldbank org All rights reserved 1 2 3 4 04 03 02 01 The findings, interpretations, and conclusions expressed herein are those of the autlior(s) and do not necessarily reflect the views of the Board of 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 work is copyrighted Copying and/or transmitting portions or all of this work without permission may be a violation of applicable law The World Bank encourages dissemination of its work and will normally grant permission 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, wwwcopyright com. All other queries on rights and licenses, including subsidiary rights, should be addressed to the Office of the Publisher, World Bank, 1818 H Street NW, Washington, DC 20433, USA, fax 202-522-2422, e-mall pubrights@worldbank org About This Book David de Ferranti is vsce president and Guillermo E Perry is chief economist in the World Bank's Latin American and Caribbean Regional Office. J Luis Guasch is a regional adviser on regulation and competition, and William E Maloney and Carolina Sanchez-Paramo are econ- omists in the Latin American and Caribbean Regional Office. Indermit Gill is an economic adviser in the Bank's Poverty Reduction and Economic Management Department, and Norbert Schady is a sensor economist in the Bank's Vice Presidency for Development Economics For more information on publications from the World Bank's Latin America and Caribbean Regional Office, please visit us at www.worldbank.org/lacpublications (o en Espafiol: www.bancomundial.org/publicaciones). Cover design by Jeffrey Kibbler, The Magazine Group Cover photograph. The Explosion of the Green Chairs, by Pedro Meyer, is from the World Bank Art Program and can also be found in the book Truths and Fictions. A Journey from Documentary to Digital Photography by Pedro Meyer (1995 New York Aperture) The World Bank Art Program makes particular efforts to identify artists from developing nations and make their work available to a wider audience The art program organizes exhibits, education and cultural partnerships, competitions, artists' projects, and site-specific installations. ISBN 0-8213-5172-9 Library of Congress Cataloging-in-Publication Data Closing the gap in education and technology / David de Ferranti .. [et al.] with William F Maloney, Carolina Sanchez-Paramo p cm - (World Bank Latin American and Caribbean studies) Includes bibliographical references. ISBN 0-8213-5172-9 1 Education-Economic aspects-Latin America 2. Education-Effect of technological innovations on-Latin America I De Ferranti, David M II Series LC67 L3 C56 2003 338 4'7378'098-dc2l 2002034304 Contents Acknowledgments ........................................ ax Abbreviations ........................................ xi Chapter 1: Introduction and Summary: Skills Upgrading and Innovation Policies........1........ Productivity, Educational, and Technology Gaps in Latin America and the Caribbean . .. .2 From Financing Gaps to Productivity Gaps. Closing the Gap in Education and Technology in a Synchronized Way . . .7 Students, Workers, and Firms The Major Actors of Educational and Technological Progress .9 The Role of Government Policy ....... . ... ..10 The Critical Importance of Effective Innovation Networks ... . . .. ..10 Closing the Skill Gaps ..... . ........... . ...12 Closing the Technology and Productivity Gaps .... . .. ...15 Summing Up........... ...... ... . 20 Endnotes .. . ...... .. . .... . . . . .21 Chapter 2: The Gaps That Matter Most................................ 23 The Productivity Gap... ............. ..... .. 23 The Skill Gap . . . . .............. .... .... . . 24 The Technology Gap . . . ... . . . .. .... .. ..... .32 Conclusion. Benchmarking Latin American Performance ...........................40 Endnotes . . .. . ... ... .... . ..46 Chapter 3: How Technology and Skills Interact: The Evidence for Latin America and the Caribbean ......... 49 The Rising Demand for Skilled Workers in Latin America ... ...49 Skill-Biased Technological Change in Latin America...................... 55 The Relationship between Technology Transfer, the Rising Demand for Skills in Latin America, and Patterns of Integration into the Global Economy . . . . ..... 63 The Level of Skill in a Country Is an Important Determinant of Technology Transferred from Abroad . ..67 Long-Term Cycles in the Demand for Skilled Workers and Technology?'... ....... ..70 Conclusion ...... .. .. . ..... . 72 Endnotes .. .... . . . ...... 72 Chapter 4: Closing the Skills Gap: Education Policies..........................75 Educational Transitions The How . . ..... . . . . . . .. . 76 Educational Transitions- The (Uncertain) Why . .... .......80 Educational Transitions. Policy Implications ... . ..... .. .. . .. . 86 Understanding the Education Market- Secondary School..... . ... 86 Understanding the Education Market Universities . . .......... 95 Conclusion- Priorities for Education Upgrading in Latin America . . ............. 104 Chapter 5: Closing the Skills Gap: Training Policies......................... . . 109 Vocational Education and Training Systems in Latin America . . ..... .... .110 How Employers View Training.... ..... ...... . . .. 117 Correlates of In-Service Training Evidence from Mexico.... ... ...... ...120 III CLOSING THE GAP Policies to Promote Training ............125 Summary . . . .... .... ...... ................ 130 Endnotes . ... . ... . .. . . ... . .. . ..131 Chapter 6: Technological Transitions and Elements of Technology Policy ..................133 Rationale for Government Action: Market and Coordination Failures . ....... . ........ 134 Developing Technological Capabilities ............135 Policies for the First Stage of Transition: Spillovers, Diffusion Mechanisms, and Incipient and Selective R&D . ...138 LAC's Performance........ .....144 Advanced Technological Transition- R&D as the Key Driver................ .... . . 151 The Structure of R&D in LAC . . . ... . .. . . . . ... . . . 161 Conclusion ... .... . . .... ........ 167 Endnotes .. .... .......... ... ......... 168 Chapter 7: Networlks and National Innovation Systems.........................169 National Innovations Systems- Coherence and Incentives ....... ...... 170 Benchmarking the LAC NIS ... ..... ... .. .. ... . . . . . . 173 Indicators of Innovation Outcomes . ....173 Efficiency of Use of Resources and Collaboration .... ... .... ..... 178 University/Research Center/Firm Linkages The Core of the NIS. .... ...180 Policies to Improve Collaboration ... . ... ..... .. ....... . . . 187 Policies to Foment Interfirm Collaboration in Innovation . . . ......... .......190 Science Parks and Incubators in LAC .. . .... . 192 International Networks ............ ............ . . 194 Conclusion . . . .. ...... .... 198 A nnex .. . . . . . . . .. . . . . . . . . . . .. . . 200 Endnotes ... .......... . . ................ . . 202 Bibliography: ........................................203 Boxes Box 1. 1 Unlocking Productivity. The Knowledge Economy Decomposed ...............4 Box 1 2 Lessons in Contrast Exploiting Technology-Skill Complementarities .......... 8 Box 1.3 The Importance of Diversity and Leaders ..... ..... .. .. 20 Box 2.1 Benchmarking Latin America ..... ... ....... 24 Box 3 1 Data Sources .. . ........ .50 Box 3.2 Measuring Education, Relative Wages, and Relative Supply . .. ....... 52 Box 3 3 Constructing Relative Demand Shifts: What Needs to Be Assumed and Why ......56 Box 3.4 Scale, Technological Change, and Human Capital- Manufacturing and Development in Mexico . . 63 Box 3.5 Evidence of Skill-Biased Technological Change in the United States and other OECD Countries....66 Box 3.6 Trade Liberalization and Sector-Specific Skill-Biased Technological Change in Brazil .....68 Box 4 1 Expanding Secondary School: The Great Transformation of American Education . .... 87 Box 4 2 The Returns to Investments in School Infrastructure. Evidence from Peru and Indonesia . . . . 94 Box 4 3 Conditional Cash Transfers: The Impact of PROGRESA in Mexico ..........94 Box 4.4 How Information Technology Has Changed the Workplace- Evidence from the OECD ...100 Box 4 5 Student Loan Schemes ..... . . .. . .. . . ... .. . . . 103 Box 4.6 Minimizing Information Problems about Private University Providers: Experiences from Latin America . .. ........ 105 Box 5.1 Mexico's CONAILEP How to Make a Vocational Education Institution Work ..... 114 Box 5 2 Lifelong Learning. Concept, Evidence for LAC, and Implications .. ...........115 Box 5.3 World Business Environment Surveys . . ..... . ...... .... .. 117 Box 5 4 Enterprise Training and Productivity in Developing Countries .. ....... . 121 Box 5 5 Firm-Level Cross-Section and Panel Data for Mexico ................. 121 Box 5.6 Complementarities between Skills and Technology Training Interventions to Boost Firm Productivity in Malaysia ........ ......... ....126 Box 5.7 Well-Designed Levy-Grant Schemes Can Motivate Firms to Train . .... .....128 Box 5 8 Mexico's Proactive Approach to Small and Medium-Size Enterprise Support .. ... 129 IV CONTENTS Box 5 9 Recent Changes in Training Systems in Latin America and the Caribbean .. . 131 Box 6 1 The Dual Role of R&D Innovation and Learning . .. 140 Box 6 2 Supply-Side and Demand-Side Measures to Improve Technology Diffusion . . ... 141 Box 6.3 Singapore's Reliance on FDI . ... .147 Box 6.4 The Role of Intellectual Property Rights (IPRs) in Technology Transfer 153 Box 6 5 Managing Intellectual Property'Rights (IPRs) . . . ... . . 156 Box 6 6 TRIPS. Developing Countries' Concerns Regarding the Universalization of IPRs ...... 157 Box 6 7 Human Capital, IPRs, and R&D Policies as Determinants of Private R&D and Their Impact on TFP-Methodology and Results . . ... . . ... . 160 Box 6 8 The Contrasting Approaches to R&D by Korea and Mexico . . . ..... 167 Box 7 1 Venture Human Capital in the United States and Israel . .. 172 Box 7 2 Jump-starting a National Innovation Effort The NIS in Israel and Finland 179 Box 7 3 Build a Better Mouse and the World Will Beat a Path to Your Door (or Is Playing God a Low-Tech Activity') . . 180 Box 7 4 What Is a College Education For' .. . . .. . . . 183 Box 7 5 Did the Bayh-Dole Act Increase the Commercialization of Publicly Funded Research? . . 186 Box 7 6 The Cluster Approach of Finnish R&D Policy ..... . 190 Box 7 7 Research Consortia in Japan Did They Work' . . 191 Box 7 8 The Israel Magnet Program ... ..... . ... 193 Box 7 9 Migration as a Vector of Innovation . . .. . ... ... . 194 Box 7 10 Return Migration in Israel and Ireland . . . . . . ... 196 Box 7.11 Deficient Brain Drain in Brazil . .. . . .. 198 Box 7 12 Catalyzing an Innovation Cluster The Millennium Science Initiative in Chile . . . . 199 Box 7.13 The Perils of Intellectual Autarky, the Rewards of Trade . . .. 200 Tables Table 1.1 The Gaps That Matter Most ... ..... .. . 6 Table 1 2 First Stage. . . . . ..16 Table 1 3 Second Stage ........ . .. ...... ..... .... .. 17 Table 1 4 Third Stage . .... . . ....19 Table 2 1 TFP Growth Rates in Latin America Were Negative in the 1980s and Low in the 1990s ... 25 Table 2 2 In the 1980s and 1990s, Growth Rates of TFP Were High in Chile and Low in Most Other Latin American Countries ... . ... . .... 26 Table 2.3 Compared to East Asia and the Natural Resource-Abundant Countries, Latin America Has Low Levels and an Unequal Distribution of Educational Attainment among the Adult Population . 28 Table 2.4 Many Countries in Latin America Have Massive Secondary Enrollment Deficits, and Some Have Smaller Tertiary Enrollment Deficits .. . ... 30 Table 2 5 Countries in Latin America Have Low Import Penetration, Low Penetration of Capital Goods, and Low Levels of Royalty Payments . . .... . .. . .. .. 36 Table 2 6 Latin American Workers Have Small Amounts of R&D Expenditures ... ..... 40 Table 2 7 Residents of Most Latin American Countries Have Low Rates of Patent Registration in Their Own Countries as well as in the United States ... ..... ... 42 Table 2.8 The Gaps That Matter Most. .... ....... ....... .... . 45 Table 3 1 The Bulk of Changes in the Wage Bill for Tertiary Workers and Secondary Workers Occurred whtin Industries ..... ......... ..... . 60 Table 3 2 By and Large, Skill Upgrading at the Tertiary Level Occurred in the Same Sectors in Chile in the 1980s and in Other Latin American Countries in the 1990s . . . . ... . 61 Table 3.3 In Chile Firms That Were Exposed to New Technology from Abroad Also Upgraded Skills . .... 62 Table 3 4 There Is More Skill Upgrading, Measured by the Relative Wages, Relative Employment Shares, and Relative Demand for Tertiary Workers, in Countries and Industries That Have Higher Import Penetration, Especially of Imports That Are Intensive in R&D . .. . . ........... 67 Table 3 5 Countries in Latin America with More Skilled Workers Are Better Prepared to Adopt New Technologies .... ..... ... . .....68 Table 3 6 Wage Inequality Increased with Increases in Demand for Skilled Workers .... ... .... 71 Table 4 1 From Pyramid to Diamond in East Asia and Scandinavia .. . . 80 Table 4 2 From Pyramid toward Diamond in Some Latin American Countries .. . .... ... . 81 V CLOSING THE GAP Table 4.3 From Pyramid to Anvil in Other Latin American Countries ....................... 84 Table 4.4 Education Spending as a Fraction of GDP Is Not Low in Most Latin American Countries . . . 84 Table 4 5 The Opportunity Cost of Attending Secondary School Appears to Be Htgh in Most Latin American Countries .............................. .... . ....... 88 Table 4 6 High Rates of Return to Tertiary Education, and Modest Rates to Secondary School in Most of Latin America .. .. ..................... .............. . ... .. ....... 88 Table 4.7 There Is Generally Little Variation in Primary School Student Performance on Test Scores within Latin America, with the Exception of Cuba .. 89 Table 4.8 Tertiary Enrollments Have Been Growing since the 1970s ....... I................. ..... 95 Table 4 9 ... and Have Continued to Grow in the 1990s ....... .. . . . . . 95 Table 5 1 A Typology of Training Systems . . .. ..... . .... . . 112 Table 5 2 Differences between Lifelong Learning and Traditional Education . . .............. .... ... 115 Table 5.3 Estimated Productivity Effects of Formal Training . . ........................... .... .. .. 121 Table 5.4 Rationale and Policy Options for Public Intervention in Training ............... ........... . 126 Table 6 1 Impact of Trade on LAC Countries and Education Effect . .. .... .. ......... 142 Table 6.2 Impact through Trade of a 10 Percent Increase in Education on TFP Growth Rate, 1998 .......... 142 Table 6 3 Effects on Foreign Direct Investment .. ......................... . . . . 143 Table 6.4 Human Capital Threshold Required for Positive Impact of FDI ...... ........................ 144 Table 6.5 Non-Tariff Barriers Levels for LAC and East Asian Countries . . . . 145 Table 6.6 Index of Foreign Investment Barriers, 2000 . . . . 146 Table 6.7 Evolution of Total R&D Expenditures as a Percentage of GDP, 1960-99 . . . . 152 Table 6 8 Transitions of Strength of Patent Protection-IPR-for Selected Countries, 1960-2000 ........... 154 Table 6 9 Tax Credits versus Matching Grants . . . ... ...................... . 159 Table 6 10 Evolution of the Ratio of Private to Public R&D Expenditures, 1980-95 ... 159 Table 6.11 Determinants of Private R&D Per Worker in LAC and Other Countries. Elasticities ....... ....... 160 Table 6.12 Intellectual Property Rights Index, 1995 .. .... .. ... 162 Table 6.13 Fiscal Incentives Regimes for R&D as of 2000 .. . . . ..... 164 Table 6 14 R&D Expenditure by Activity (in percentages) . . . ............. . ... .. ........ 166 Table 6 15 Patterns of R&D Expenditures LAC versus OECD and East Asian Tigers, Averages for 2000 ....... 167 Table 6 16 Similarity of Education Patterns in LAC (1990s) versus Technology-Successful Economies (1980) ... . 168 Table 7.1 Potential Benchmarks for the LAC NIS. A Wish List .... .... ... .... . . . ....... 174 Table 7 2 Determinants of Patenting in the United States . . . . . . .. ... ............ . 181 Table 7 3 Tentative Benchmarks of Interaction of Higher Education (HE) and Public Laboratories (PL) with Firms ... .......... . . . . . ...... 185 Table 7 4 Guidelines for Sharing Royalties from IPRs . . . ................ 187 Table 7 5 Policies for Improving Academic/Private Sector Collaboration ... . ... . ...... 187 Table 7A 1 Social Rates of Return to R&D .. ....... .... ...... . ................ ..... 201 Figures Figure 1.1 The Knowledge Economy Simplified ........ . .............................. 3 Figure 1 2 The Latin America and the Caribbean Region's Deficits in Technology and Education Relative to East Asia (1980-99) .... ...5 .. 5 Figure 1.3 A Problem in Many LAC Countries: Massive Deficits on Secondary Education .... .... 6 Figure 1.4 A Simple National Innovation System (NIS) . ... ..................... .. ...... 11 Figure 1.5 Educational Transitions .. .................... . . . . . . . ... 13 Figure 1 6 Superstars in R&D ... .. ........ .. . . ........... ... 18 Figure 1 7 Returns to R&D and Physical Capital ...... ............................ .............. 18 Figure 2 1 Most Latin American Countries Have Low Levels of Educational Attainment of Their Adult Population .............. ..... ..... .. ..... 27 Figure 2 2 Many Countries in Latin America Have Massive Deficits in Secondary Enrollment ...... .......... 29 Figure 2.3 Some Countries Also Have Large Deficits in Tertiary Enrollment, although These Are Generally Smaller .............. . . . ... ............ .......... 31 Figure 2 4 Colombian and Chilean Secondary School Students Perform Badly on International Tests . .. .... 32 Figure 2.5 Chilean Adults Also Do Badly on International Tests of Literacy . .33 Figure 2 6 Some Latin American Countries Appear to Have Low Numbers of Scientists and Engineers for Their Income Levels, but Others Do Not ..... .......... .. .... 34 Figure 2 7 Most Countries in the Region Do Not Appear to Have an Important Deficit in the Fraction of University Students Enrolled in Science and Engineering . . . . . 35 VI CONTENTS Figure 2 8 In the 1980s East Asia Opened Up a Big Lead over Latin America in the Degree of Computer Penetration ....... . ... 37 Figure 2 9 The Digital Divide between East Asia and Latin America Did Not Shrink in the 1990s . . 38 Figure 2 10 Latin America Also Lags behind East Asia in Other Measures of ICT Infrastructure .. . 39 Figure 2.11 Latin American Countries Also Have Low Levels of R&D Per Worker ............... . .. ... 41 Figure 2.12 The Rates of Patent Registration by Latin American Inventors Are Low for Their Levels of Per Capita Income ..... ....... .... 43 Figure 2 13 The Widening of the LAC Technology Gap Regional Comparisons of Index Scores . . 44 Figure 2 14 Between 1960 and 2000, the East Asian Tigers Turned a Deficit in the Mean Years of Schooling into a Surplus, while the Deficit in Latin America Did Not Change . . .... 45 Figure 2 15 Between the 1960s and the 1990s, the East Asian Tigers Turned a Deficit in the Mean Number of Patents Per Worker Registered in the United States into a Surplus, while the Deficit in Latin America Did Not Change ..... .... . ...... 46 Figure 3 1 The Relative Wages and Relative Supply of Workers with Tertiary Education Increased Substantially in Most Latin American Countries ........... .. ... . .... 51 Figure 3 2 The Relative Supply of Workers with Secondary Education Increased Substantially in Latin America . . . 54 Figure 3 3 The Demand for Tertiary Workers Increased in Every Country except Brazil . . 57 Figure 3.4 The Demand for Secondary Workers Increased in Every Country except Argentina and Brazil .. 58 Figure 3.5 Tariffs Decreased and Import Penetration Increased in a Number of Latin American Countries in the 1980s and 1990s .......... .... 64 Figure 3 6 FDI Flows Increased Significantly in a Number of Latin American Countries .... ........ ...... 65 Figure 3.7 The Long-Term Cycles in Relative Demand in Latin America ...... . .. ......... 70 Figure 4 1 Education Transitions . ........ ...... 77 Figure 4 2 The Distribution of Educational Attainment Was Similar in Brazil, Spain, and Korea in 1960 78 Figure 4 3 Fast, Balanced Transitions in the East Asian Tigers and Finland . . . . . 79 Figure 4 4 Balanced but Slow Transitions in Some Latin American Countries . . . 82 Figure 4 5 Unbalanced and Slow Transitions in Other Latin American Countries . . . 83 Figure 4 6 There Is No Clear Relationship between Changes in Education Expenditure and Changes in Attainment in Latin America . .......... 85 Figure 4 7 There Is No Clear Relationship between Education Quality and Quantity in Latin America .. ....... 90 Figure 4 8 Patcerns of Educational Attainment in Guatemala, Nicaragua, and Peru . 92 Figure 4 9 Supply and Demand Factors in the Evolution of Educational Attainment in Colombia, 1990-2000 ..... .. 93 Figure 4.10 The Rising Share of the Private Sector in the University Market 96 Figure 4 11 Private Tertiary Education Is More Expensive to Households . . ........ 97 Figure 4 12 Conditional on Enrollment, Students from Poorer Households Are Less Likely to Attend Private Universities . . . . . . . . 98 Figure 4 13 The Wages of Workers with "Scientific" Degrees Relative to Those with "Humanistic" Degrees in Mexico Have Not Changed in the 1990s ............... . . ...... 101 Figure 4 14 Cost Recovery in Public Universities Is Low in Many Latin American Countries .... . . . 102 Figure 5 1 Education and Training Systems around the World . . .... .... . ...... ....... 111 Figure 5.2 Wage Profiles in Brazil, Colombia, and Mexico-Early and Late 1990s . . . 116 Figure 5 3 The Supply of Skilled Labor Is a Leading Constraint to Productivity in Latin America and the Caribbean ......... .. 118 Figure 5 4 Use of Mature Technologies and Availability of Skilled Workers Are the Main Reasons Why Firms in Latin America Do Not Invest in Formal Training ... .......... ..119 Figure 5 5 Knowledge of Training Providers and Resource Considerations Are More Important Constraints for Micro-Enterprises, Guatemala 2000 ...... ........ ... . . . .. . 120 Figure 5 6 Almost 50 Percent of Training in Latin America Is Internal to Enterprises, and Another 40 Percent is Privately Supplied . . .......... .. ....... .. .. 120 Figure 5 7 In Mexico, Educated Workers Get Trained More Often... ........... 122 Figure 5 8 and More Innovative Firms Train Their Workers More ....... . ......... 123 Figure 5 9 In Mexico, Two-Thirds of Training Is Internal .... .......... 123 Figure 5.10 .. and About Half of the Rest Is by Private Providers .... 124 Figure 5 11 In Mexico, Training Increases with Greater Openness, but Only if Firms Conduct R&D . . . 125 Figure 5 12 Training, Especially if It Is More Continuous, Facilitates TFP Increases ... . . ......... 125 Figure 6 1 R&D Effort, Licensing, and Developmentr Predictions from Median Regressions .. 136 Figure 6.2 Innovation and Development ...... .. ..... . . 136 Vil CLOSING THE GAP Figure 6.3 Mean Tariff Barriers (percent), 1990 and 1999 ... ............ .. 144 Figure 6.4 Foreign Direct Investment, 1989 and 1999 . ..... . . . . .. . 146 Figure 6 5 High-Achieving LAC Countries . ................... .... 149 Figure 6.6 Low-Achieving LAC Countries ...... ...... .... .... . . . ... . .. 150 Figure 6 7 Evolution of R&D/GDP Expenditures on GDP Per Capita 1980-1995 . . 152 Figure 6 8 Patent Rights and Privately Financed R&D (1980-95 average for 58 countries) ....... ...... . 161 Figure 6.9 Researchers, 1999 ..... . . ........ .. ..... 163 Figure 6.10 Financing of R&D Expenditures, 1999 ..... ...... ....... ... .... ...... 165 Figure 6.11 Gross R&D Expenses by Executing Sector, 1994-98 .. . .... . . . . . 165 Figure 6 12 Private R&D Performance and Experimental Development ... .... . .. 166 Figure 7 1 A Simple NIS ........ . ...... .... . .. .... . 170 Figures 7 2A-F Patenting, Licenses, and Royalties and R&D Benchmarking . . ... . . . 176 Figure 7 3 R&D Expenditure and R&D Financed by the Productive Sector. Taiwan (China), Korea, Finland, and Israel ..... .... . .... . .. ...... ... 177 Figure 7 4 Predicted and Observed R&D/GDP ...... ......... .... . .. . ... 177 Figure 7.5 Returns to R&D and Physical Capital . . ... .... . . . ...... . . . 178 Figure 7 6 Inefficiency of LAC Countries in Patenting in the United States .. . 181 Figure 7 7 Allocation of R&D Resources ...... ................. ..... . .. 183 Figure 7 8 Quality of Scientific Research Institutions and University Industry Research Collaboration .. 184 Figure 7.9 Brazil and Mexico. IRCA in Aircraft and Computing Equipment .. 189 Figure 7 10 Student Migration Rate 1993-98 . . ... . .... . . .... ..... .......... 195 Figure 7.11 Brain Circulation Migration/Population with Tertiary Education ....... .... .... .. 197 vill Acknowledgments TH HIS REPORT IS THE RESULT OF A COLLECTIVE EFFORT BY A WORLD BANK TEAM coordinated by the Office of the Chief Economist for the Latin America and the Caribbean Region and led by Guillermo E. Perry. The principal authors by chapter were Guillermo E. Perry (chapter 1), Carolina Sanchez-Paramo and Norbert Schady (chapters 2, 3, and 4), Indermit S. Gill (chapter 5), J. Luis Guasch (chapter 6), and William F. Maloney (chapter 7). Manjula Luthria and Prita Subramanian coauthored preliminary versions of chapters 6 and 7. Friends and colleagues from the World Bank made efforts that went well beyond the call of duty. We received important contributions from Harold Alderman, Ana-Maria Arriagada, Daniel Brehon, Soumya Chattopadhyay, Wendy Cunningham, Carl Dahlman, Eduardo Doryan, Marito Garcia, James Hanna, Barbara Larraneta, Daniel Lederman, Danny Leipziger, Ernesto May, Marcelo Olarreaga, Lant Pritchett, Jamil Salmi, Luis Serven, Wesley Yin, Shahid Yusuf, and others at the Chief Economist's Retreat in November 2001. The report benefited immensely from authors who pre- report together These contributors include Daron Acemoglu, pared background papers. These authors include Andreas Eduardo Bitran, Jose Joaquin Brunner, Paul David, Jean Blom, Magnus Blomstrom, Mariano Bosch Mossi, Martin Guinet, Alan Kreuger, Stephen Machin, Ramon Marimon, Carnoy, Ed Feser, Andrew Foster, Leonardo Garnier, Pinelopi Christina Paxson, Andres Rodriguez-Clare, and Gilles Goldberg, Claudia Goldin, Lauritz Holm-Nielsen, Caroline Saint-Paul Hoxby, Hanan Jacoby, Vanus James, Juan J. Jimeno, Larry Finally, we'd like to thank Lee Morrison, Anne Pillay, Katz, David Mayer, Walter Park, Nina Pavcnik, Maurice Patricia Soto, and the World Bank's Office of the Publisher Schiff, Simon Schwarztman, Emmanuel Skoufias, Maximo for production and dissemination support Torero, and Yanling Wang. Any errors or omissions in the report are the sole respon- The authors would like to thank those whose insight, sibiliry of the authors and should not be attributed to any of comments, and suggestions proved valuable in putting this the above individuals or the institutions they represent. Ix I Abbreviations CEPR Centre for Economic Policy Research CIMO Integral Quality and Modernization Program (Mexico) DHS Demographic and Health Surveys ECLAC Economic Commission for Latin America and the Caribbean ENIGH Encuesta Nacional de Ingresos y Gastos de Los Hogares (Mexico) EE Eastern Europe EPZ Export processing zone EU European Union FDI Foreign direct investment FTZ Free trade zone GDP Gross domestic product GRI Government research institute HE Higher education HRDC Human Resource Development Council (Malaysia) HRDF Human Resource Development Fund (Malaysia) IBRD International Bank for Reconstruction and Development ICT Information and communications technology ILO International Labour Organization IPRs Intellectual property rights LAC Latin America and the Caribbean MIT Massachusetts Institute of Technology MNC Multinational corporation NAFTA North American Free Trade Agreement NBER National Bureau of Economic Research NIC Newly industrialized country NIS National innovation system NPV Net present value OECD Organisation for Economic Co-operation and Development PL Public laboratories PPP Purchasing power parity PRO Public research organization PRR Private rate of return R&D Research and development RCA Revealed comparative advantage S&T Science and technology system xi CLOSING THE GAP SBTC Skill-biased technological change SITC Standard Industrial Trade Classification SMEs Small and medium-size enterprises SRR Social rate of return STPC Science and Technology Policy Council (Finland) STPS Department of Labor and Social Welfare (Mexico) TFP Total factor productivity TIMSS Third International Mathematics and Science Study TQM Total Quality Management TRIPS Trade-Related Intellectual Property Rights UNAM Universidad Nacional Aut6noma de Mexico UNCITRAL United Nations Conference on International Trade Law UNESCO United Nations Educational, Scientific and Cultural Organization WBES World Business Environment Surveys WTO World Trade Organization The authors are the sources for all figures and tables in this book unless otherwise noted xii CHAPTER 1 Introduction and Summary: Skills Upgrading and Innovatlon Policies * VVER THE LAST 50 YEARS, THE WORLD HAS INCREASINGLY BECOME DIVIDED in to two clubs-those of rich and of poor countries. What is most striking is that the increas- ing bimodal distribution of income is due not to concentration of the factors of production, such as capital, but rather of knowledge. The North has continued to generate new tech- nologies that statistical studies suggest drive at least half of economic growth, while most of the South has been unable to take full advantage of them. Latin America for the most part finds itself in the less desirable club. Between 1950 and 2000, the annual per capita income in OECD countries tripled from US$7,300 to $23,000 (hereinafter, all dollar values in US$). The Latin America and the Caribbean region's (LAC) income level grew much less during this period-just doubling from $3,000 in 1950 to $6,200 in 2000. So the ratio of LAC's average income to the developed-country average fell from more than 40 percent in 1950 to about 25 percent in 2000, with much of the divergence occurring during the last quarter century. This should perturb policymakers in the region. The trend is of special concern since the newly industrialized countries (NICs) of Asia-as well as other countries, such as Finland, that are closer to LAC in resource endowments-were able to achieve dramatic technology-driven increases in living standards across the Even as late as 1950, the United States was perhaps the same period. In fact, the central lesson of our flagship study only country where the median individual leaving school of 2001, From Natuiral Resources to the Knowledge Economy, was was a high school graduate. By the time countries in that what distinguished the successful resource-abundant Europe-seeing the rapid rise of the United States as an countries (Australia, Canada, and the Scandinavian countries) economic powerhouse-learned their lesson, the United from the disappointing LAC experience was precisely their States had expanded its lead over Europe in tertiary edu- ability to learn from abroad-their national innovative cation, and most European countries have only recently capacity. This report looks more carefully at what it takes begun closing this gap. The East Asian countries appear for countries, and more centrally, for firms, to learn how to to have learned the lesson much more quickly-the Republic learn The key ingredients in the success of these countries of Korea, for example, has secondary and tertiary educa- are, first, that they early on recognized the need for an tion indicators that are already better than those of many explicit, efficient, and sustained policy to move the private European countries. In most other countries-including sector to the technological frontier, and second, that they much of Latin America-this recognition has come more engaged in one of the most rapid and dramatic build-ups of slowly. Emerging evidence suggests that the knowledge national human capital in human history transfer benefits of foreign direct investment (FDI) and Education is vital for at least two reasons First, it has trade liberalization, for example, are enhanced by higher always been a critical complement to technological advance. stocks of human capital CLOSING THE GAP Technological change in the 20th century has been affected countries differently and at different points in time. increasingly biased in favor of skilled workers and appears Increases in the demand for skills associated with the to be the largest force driving the increasing skilled/unskilled adoption of new technologies have taken place primarily in wage differential in the industrialized countries. This has countries with levels of human capital above a minimum been true for so long that people now think that technical threshold. In addition, the patterns observed in most Latin change has always favored more skilled or educated workers. American countries in the 1990s are similar to those But this was not always the case: for example, technical change observed in Chile in the 1980s, suggesting the existence of in 18th-century Britain displaced skilled artisans. There is long-term cycles associated with technological progress in some evidence that the skill-biased character of technological the relative demand for more skilled workers, depending on change has accentuated in the last decades, partially as a when countries opened up to trade and FDI competition. result of the so-called information and communications tech- Such a process is thus creating both challenges and oppor- nology (ICT) revolution (de Ferranti and others, 2001). tunities. Demand for higher skills, and in particular for This report focuses not only on the gaps facing Latin workers with tertiary education, is rising faster than America in both education and technology, but especially demand for less skilled workers in most of the region. Hence on the interactions between the two. The central premise of the challenge inequality in wages tends to increase as long the report is that skills and technology interact in impor- as major inequalities in access to education remain, and the tant ways, and this relationship is a fundamental reason for lack of skills may become an effective constraint for fast and the large observed differences in productivity and incomes efficient technology transfer and thus for potential growth. across countries. Indeed, numerous studies have shown that But also the opportunity: the increased potential demand differences in per capita income depend more on differences for education offers the possibility to accelerate productiv- in total factor productivity (TFP) than on differences in pri- ity growth in the economy by closing the educational and mary factor accumulation I And this report argues that skills technological gaps that Latin American countries exhibit upgrading, technological change, and their interaction are with respect to their peers To achieve this, policy must major factors behind total factor productivity growth. respond swiftly so that this potential demand for higher and The report shows that skill-biased technological change better education is actually translated into higher and better is indeed being transferred today at faster speeds to LAC educational levels of the work force and higher productiv- countries, as elsewhere, insofar as they are now more open ity at the firm level. to trade and FDI flows and have increased the level of edu- Focusing on skills upgrading, technological change, and cation of their work force In chapter 2 we present empiri- their interactions helps us to discipline thinking about the cal evidence that technological change has been currently confused issues surrounding the so-called knowl- complementary with skill levels in Latin America in the last edge economy. In box 1.1 and figure 1.1 (elaborated upon in two decades, as it has been the case in more developed coun- Gill 2002) we present a simple graphic representation of the tries. We also show that, as a result, firms have substantially interaction between technology and skills that we follow to increased the demand for educated workers in the region, organize the discussion of skills upgrading and technologi- particularly workers with tertiary education, and this rise in cal change throughout the report. the demand for educated workers has bid up their relative wages, as has been happening in first world countries. Pboduci Eit ducational, and T'ecihnology Gaps This technological transformation appears to be inti- in LaWin Amedca and the Caribbean mately related to patterns of integration in the world econ- In chapter 2 we compare the performance of countries in omy. Trade and FDI have facilitated the transmission of Latin America in terms of productivity, skills, and tech- technological change across borders. Also, firms in sectors nology with the performance of an "average country in the with higher exposure to trade and FDI are subject to more world," of the East Asian "tigers," and of a group of suc- competitive pressures. Adopting and adapting more cessful natural resource-based economies Because the advanced technologies and hiring and training more edu- countries we want to compare are at different stages in cated workers is one way to respond to this pressure to development, we calculate their expected level of achieve- become more productive. Finally, these changes have ment given their per capita income levels and compare it 2 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES FIGURE I I The Knowledge Economy Simplified Prerequisite | _ ___ _ Prerequisites: Primary Eduication L KNOWLEDGE ABSORPTION Macroeconomic __- Stability and Incentives Regime GENERAL FOREIGN -Secondary Trade, FDI, Licensing - Tertiary - IPR Regime - Adaptive R&D TRAINING EDUCATION "- TECHNOLOGY NETWORKS . _____ 4 -_ _ __ _ _ _ _ _ SCIENTIFIC AND DOMESTIC TECHNICAL -Creative R&D - Tertiary __- - Intellectual - Postgraduare Property Rights Regime Catalyst. . - _ Catalyst: Labor Market KNOWLEDGE GENERATION Capital Market Development - _ _ _ _ _ Development Source Gill (2002) to their actual level of achievement so that the estimated (from 4.1 to 5.8 years). On average, however, Latin gaps are relative to the level of development of each country. American adults have 1.4 fewer years of education, and East We find that, in the aggregate, Latin America suffers from Asian adults have 0.4 more years of education than would significant deficits in productivity, skills, and technology. be expected for their income levels. This gap in the stock of (See figure 1.2 and table 1.1) educational attainment is a reflection of relatively slow and As mentioned, productivity differences between countries sometimes inadequate education investments in the past. It and between firms within countries are profoundly affected is therefore particularly worrying to observe that theflow of by differences in skills and technology 2 It is therefore no sur- new educated workers is also inadequate. The region has prise that the East Asian tigers, which exhibit well-above- large deficits in enrollment, particularly at the secondary average rates of TFP growth, also outperform Latin America school level (figure 1.3), as well as a problem with the qual- on measures of technology and skills. The same is true for ity of education. Latin America has an aggregate deficit of some of the successful natural resource-based economies around 20 percentage points in net secondary enrollment and Within Latin America, the best-performing country, Chile, 10 percentage points in gross tertiary enrollment given its concurrently had positive increases in productivity, sub- average income level, while East Asia has surpluses of more stanrial skill upgrading, and increases in all indicators asso- than 17 and 5 percentage points, respectively, and success- ciated with technology transfer and innovation ful natural resource-based economies of 6 and 14.3 Finally, Since there is no perfect measure of skills, we consider we discuss the quality of the students "produced" at each the evolution of educational attainment of the adult popu- level of education, as given by the performance of Latin lation (measured as average years of education) and the evo- American students and adults on international standardized lution of enrollment rates. Over the course of the last two tests. The evidence here is more limited, as most Latin decades, the mean years of education of the adult popula- American countries do not participate regularly in inter- tion aged 25 and older has gone up by 1.7 years in the region national tests, but the available data reflect yet another 3 CLOSING THE GAP |BOX I I Unlocking Poduciity: The LSnowledge Economy Decomposed The center of the action in figure 1.1 is the firm, where coor- cal obstacles for productivity growth only later in the edu- dinated decisions on skills recruitment, training, adoption of cational and technological transitions. technologies, research and development (R&D), and inno- For the system to work efficiently it is not only impor- vation are undertaken. At the left we have the formal tant to develop the individual components, policies, and education system progressing from basic to tertiary and post- instruments, but to ensure that they are properly "linked" graduate education. At the right, a set of channels for trans- through effective networks that help overcome coordination mission of foreign technologies and policy instruments and and informationat asymmetries problems. As an integrated institutions to foster the adoption, adaptation, and creation system builds up, societies transit from technological stag- of new methods of organization, production, and marketing nation, through a process in which individuals and firms by firms (what we call for short the "technology system"). engage increasingly m knowledge absorption (through skills Generalized basic education and macroeconomic stability upgrading, and adoption and adaptation of technologies), to appear as prerequisites for a meaningful and sustained process one in which a significant number of them engage also in of adoption, adaptation, and creation of modern technologies knowledge creation (development of new technologies and by firms. Skills acquired through secondary and tertiary basic science). As mentioned below, however, this process education enable firms to adopt and adapt existing tech- does not advance at the same pace in all sectors or firms. nologies in a more efficient manner and to train their work- Leaders are critically important (see box 1.3). ers. Skills developed at the postgraduate level-especially Our report is divided into two sections that conform but not exclusively in fields such as science and engineering- to this diagram. In the first section (chapters 2 and 3) we allow firms to create and develop new technologies. present evidence of the complementarities between tech- Openness to trade and FDI, and, more generally, com- nology and skills in Latin America, using household-, firm-, petitive pressures, create an environment that fosters inno- and country-level data, and measure the gaps of Latin vation and traimnig by firms and facilitates transfer of foreign America in skill levels and upgrading and in technologi- technologies. Licensing and circulation of brains are cru- cal inputs and outputs with respect to other countries, with cial channels for such a transfer. Well-designed government special reference to the East Asian Tigers and successful incentives and support to training and R&D activities by natural resource abundant countries (such as Canada, firms-intellectual property rights (tax incentives, com- Australia, and Scandinavian countries). Technology and petitive matching grants, ICT infrastruccure, and legal skills are important factors in development, and Latin framework)-on top of a supply of well-educated workers America is in short supply of both. In the second section and a competitive environment, are instrumental to bring (chapters 4 through 7), the policy implications of this these activities, which produce significant externalities, analysis are presented. Chapters 4 and 5 look at education closer to their social optimum. Well-regulated labor and and training policies, while chapters 6 and 7 focus on tech- capital markets are also important catalysts, but countries nology policies and knowledge networks, respectively. aiming to close productivity gaps should not wait to launch efforts until these conditions are met; these become criti- Somrc. Gill (2002). serious deficit. Not only do Latin American countries (other imports of capital goods, make payments to license foreign than Cuba) underperform relative to an income-adjusted technologies, and receive FDI. Import penetration as a share benchmark, they often underperform relative to much poorer of GDP is lower than expected, and import penetration of countries. capital goods, which often embody new technologies, is We also use three sets of factors to assess the technology about one-sixth of that in the East Asian tigers. While Latin gap. First, we examine the degree to which Latin American America enjoyed an advantage over East Asia in terms of countries are exposed to foreign imports, in particular computer imports in 1980, that position was reversed by 4 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVAIION POLICIES FIGURE 1 2 The Latin America and the Caribbean Region's Deficits in Technology and Education Relative to East Asia (1980-99) Income-adjusted comparisons 2 Education Technology Productivity 15 05 ** Latin America and the Caribbean 0 _ East Asian "tigers" 0 Natural resource-abundant countries -o 5 -1 5 -2 1990, and there are no indications that Latin America Is clos- vation nor the ability to look outssde for relevant tech- ing that gap FDI is, on average, higher in Latin America nologies. Partly for historical reasons and partly to com- than in East Asia as a proportion of FGDP. However, this pensate for the low private sector effort, governments have also reflects a greater preference for licensing-the contrac- tended to have a disproportionately high level of partici- tual transfer of technology, which provides technology in a pation in total financing and implementation of R&D (a more accessible manner than FDI-ln many Asian countries. large part of the latter being concentrated in public Second, we evaluate the development of the ICT sector universities). as an essential complementary instrument to develop tech- Finally, though far more difficult to benchmark, the use nological capabilities and we show that LAC's gap relative of innovation-related resources and human capital is highly to East Asia has more than tripled during the 1990s inefficient in the region. The overall coordination of uni- Third, we focus on domestic spending on research and versities, research centers, and the productive sector Is poor, development (R&D) and payments for licensing, relative to implying that the little R&D investment that is done is the benchmark for countries of their levels of income, but employed relatively inefficiently, with lower results in terms also with respect to the performance of superstars such as of patents and impact on growth than in the case of com- Finland, Korea, Israel, or Ireland, and to the extraordinary parable countries or the OECD. Thus, not only does Latin returns generated on innovation investments. The most America lag in terms of the total amount of R&D relative to striking result is the low level of R&D conducted by firms. GDP, but a relatively large share of that R&D is undertaken This is partly due to weak supporting institutions such as by the public sector and has less spillover on private R&D credit markets, ICT infrastructure, or government policies than in other latitudes As a consequence, R&D spending toward fomenting innovation (IPRs, competitive subsidies, in Latin America is not only small, but is less concentrated and tax incentives for private R&D, and so on) It also in applied research and development and has less effect on reflects that until the recent opening of LAC economies to patents and productivity growth than what we observe in competition and foreign ideas, firms had neither the moti- OECD countries and in the Asian tigers (see chapters 2 and 7), 5 CLOSING THE GAP TABLE I I The Gaps That Matter Most OBSERVED INCOME- INCOME- INCOME- EAST ADJUSTED OBSERVED ADJUSTED ADJUSTED ASIAN SURPLUS IN NRA SURPLUS OBSERVED DEFICIT TIGER EAST ASIAN COUNTRY IN NRA LAC MEAN IN LAC MEAN TIGERS MEAN COUNTRIES Productivity TFP growth per year (1990-99) 0 45 -0 21 1 42 0 76 078 0 12 Education Mean years of education, population aged 25+ 5 8 -1.4 97 1 0 11 1 1 4 Net secondary enrollment rate (percent) 46 7 -18 7 93 3 17 8 93 2 6 0 Gross tertiary enrollment rate (percent) 20 0 -10 0 47 3 5 1 67 0 14 0 Third International Mathematics and Science Study (TIMSS) score (points) 387 -81 2 584 86 3 524 + (n s Technology Capital goods imports as a fraction of GDP, 1999 (percent) 7 7 -3 6 29 5 18 1 126 20 Domestic R&D expenditures per worker (1995 US$) 35 6 -264 329 5 116 5 725 4 152 5 Patents registered in the U S, 1996-2000 average (per million) 08 -1 5 54 4 48 3 114 8 80 9 Mean number of computers per 1,000 workers 37 7 - (n s) 172 0 + (n s) 404 2 + (n s) Note The regional "deficits" or "surpluses" are calculated as the weighted sum of the difference between the observed value and the predicted value from an OLS ordinary lease squares regression on log per capita income for all outcomes except TFP growth and imports of capital goods For TFP, the means and the predicted value are given by the weighted world average, without a regression For imports of capital goods the explanatory variable is GDP The regional means, deficits, and surpluses for imports of capital goods are weighted by GDP, those for the mean years of education, net secondary enrollment rartes, gross tertiary enrollment rates, TIMSS scores, and the number of computers are weighted by population, the measures of domestic R&D and patents, finally, are weighted by the working-age population aged 25 and older "n s " indicates that the deficit or surplus is not statistically significantly different from zero (at the 5 percent level) The East Asian tiger economies included in the sample are Hong Kong (China), Republic of Korea, Malaysia, and Singapore Data for Hong Kong are not available for TFP growth, imports of capital goods, or R&D expenditures, data on Singapore are not available for nec secondary enrollment rare, and imports of capital goods, data on Malaysia are not available for R&D expenditures The natural resource-abundant countries included in the sample are Canada, Australia, New Zealand, Finland, Norway, and Sweden Soaure Data on TFP from Loayza, Fainzylber, and Calder6n (2002), data on educational attainment of adults from Barro and Lee (2002), data on enrollment rates from UNESCO, data on TIMSS scores from TIMSS, data on imports, imports of capital goods, GDP, GDP per capita, and population from World Bank databases, data on R&D and patent registration from Lederman and Saenz (2002) FIGURE 1 3 A Problem in Many LAC Countries: Massive Deficits on Secondary Education 100 - Repubhc of Sweden Repubic of Finland Malaysia Korea Canada New Zealand United States Au, Cco 80- Jamaica Trinidad and Tobago A cmx e Chl Hong Kong (China) E Guyana ' 60- Philippn m inican Republic Ecuador Colombia i Indo <,,,,/lndo nesia Paraguay Costa Rica 40 ~~~~~~~~~~El Salvador Brazil R B de Venezuela 20 5 6 7 8 9 10 11 Log of per capita GDP, 1998 6 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES even when benchmarked by their respective average per manner to low or erratic economic growth. The reason is the capita income. strong complementarity between technology and skills that The overall picture that emerges is thus one of LAC coun- lies behind skill-biased technological change because of three tries not having given innovation and science and technol- facts. First, skilled workers are needed to implement new ogy a central role in their development strategies until very technologies because skilled workers are more adept at recently. As a result, while their levels of indicators of tech- dealing with change. The second is the phenomenon of nology absorption and innovation have increased, these coun- "directed" technological change-the availability of more tries have fallen behind more aggressive economies such as skilled workers in developed countries has created incentives the Asian tigers, Israel, Ireland, and Finland. No doubt the for firms to develop new technologies that are more skill- opening of the LAC countries has given them greater access intensive (Acemoglu, 2001). Directed technical change to various technologies developed abroad. However, as we requires an even greater need for speed in the upgrading of argue throughout this report, merely having access to tech- education systems in developing countries, as the skill-biased nology need not be enough for rapid sustainable growth. The nature of technological change is poised to continue. Third, critical question is what capabilities and institutions these skilled workers, engineers, and scientists are required to pro- countries need to use these technologies efficiently and even- duce significant adaptations of existing technologies and even tually to steer themselves onto a path of innovation-based more to create new ones. Even plain adoption and diffusion development, which allows for sustainable long-term growth. of existing technologies require a minimum generalized level of education of the work force (at least some secondary edu- From Financing Gaps to Productivity Gaps: cation) and of training and R&D in firms. Closing the Gap in Education and Technology Thus, countries with low levels of education remain in in a Synchronized Way a trap of technological stagnation, low growth, and low The 1980s and 1990s in Latin America and the Caribbean demand for education (a reason why basic education must were marked by a struggle with twin deficits. a budget gap be highly subsidized everywhere). Conversely, countries may and a trade gap. Fiscal and monetary policies kept center strongly subsidize tertiary education, but if they do not open stage. It would not be an exaggeration to state that coun- to trade and FDI and firms are not subject to competitive tries of the region have mostly accepted the wisdom of bal- pressures that stimulate technological progress, and hence ancing the budget and ensuring external balance through demand for education, they will find out that a high pro- conservative monetary and exchange rate policies But, with portion of their educated emigrate and that they must keep few exceptions such as Chile, these decades have also been or increase the level of subsidization to compensate for weak marked by sputtering growth engines-which in turn effective demand. In a similar vein, low levels of R&D in reflects low annual average rates of productivity growth firms may easily become an effective limitation to adoption With productivity growth accepted as the sine quia non of sus- of new technologies and innovation and hence to the growth tained economic growth, Latin America's challenge is shift- of demand for education; while efforts by the government ing from closing the financing gaps to closing the to subsidize R&D may fail to achieve their goals when low productivity gaps. levels of education prevail and/or the economies are not open Just as the financing gap had two closely related com- to trade and FDI and firms face no competitive pressures to ponents (the so-called twin deficits-fiscal and current innovate. A background paper for this report (Gill 2002) account), productivity gaps can be thought of as having two compares a set of Latin American and non-Latin American equally symbiotic components-skills and technology. In the countries, some of which have been highly successful in syn- same way that the financing gaps must be closed together chronizing their educational and technological policies and to ensure sustainability, skill and technology gaps must be developing educational and technological capabilities in a reduced simultaneously in order to reduce productivity gaps balanced way, and some that have not. The paper's conclu- in an efficient way sions are summarized in box 1 2. The experience of countries in Latin America and around The implication for countries of Latin America and the the world appears to condemn countries that attempt to close Caribbean and elsewhere is that they can close skill and tech- education and technology gaps in an unsynchronized nology gaps more productively, that is, with stronger effects 7 CLOSING THE GAP BOX 1 2 Lessons in Contrast: Exploiting Technology-Skill Complementaritles A background paper for this report (Gill 2002) explored nology transfers also experience low productivity a comparison of a set of five Latin American (Brazil, Costa growth. However, it is easier for such countries to Rica, Chile, Peru, and Mexico) and five non-Latin American increase productivity-an opening up of the economy countries that, at some moment of their development, to foreign trade, investment, and knowledge flows shared both similarities and dissimilarities in educational results in increased productivity growth almost and technological policies (China, Korea, Finland, Spain, immediately. The contrasting experience of Peru in and Singapore) to illustrate how and why some policy the 1980s and 1990s is a case in point. regimes have proven to be more successful than others in o Countries that both increase education levels and capturing the complementary environment in which skills increase technology transfers, but do so mostly and technology can boost productivity. The conclusions of through direct public sector provision (of education, such an exercise reinforce the basic messages of this report: training, and R&D) also appear to do poorly in pro- | Countries that neither increase education levels nor ductivity outcomes. These countries face a difficult approach the technological transition in a sequenced challenge of increasing private sector participation in manner suffer poor productivity outcomes: developing new technologies and delivering education - Opening up to foreign technology will not help and training. Mexico is a prime example, where the much in aggregate productivity, if education main challenge today is to effectively link the high levels are kept low (or highly skewed), given the levels of public R&D with a boost to private R&D. complementarity between technology and skills. o Countries that have increased education levels and These countries run the risk of not only facing that have done so by building a broad base of pri- skill bottlenecks, but also of exacerbating earn- mary and secondary schooling, and at the same time ings inequality. have encouraged the adoption and adaptation of new Heavily investing and subsidizing R&D would technologies by private sector firms through open- not pay back in higher productivity growth either ness, should focus most of their attention on in closed economies (which do not provide com- strengthening their support to private R&D and petrtive pressures for firms to innovate and do not stimulating knowledge networks, through improved facilitate transfer of technologies through trade linkages between universities and firms and among and FDI) or in those with low levels of education firms, both within the country and internationally, of most of the working force (Brazil's past his- extending tertiary schooling, and supporting post- tory is a prime example of such an imbalanced graduate programs, particularly in sciences and engi- educational and technological policy that did not neering. Chile is a good example of a country in this pay off in the last decades.) stage o Countries that have increased their average education levels but do noc institute policies that facilitate tech- Soumce: Gill (2002) on productivity and hence economic growth-and indeed adaptations and creation of new technologies) should match with less effort-if their policies in these two areas are one another. Thus there is no "one size fits all" recommen- synchronized. dation: policy priorities should depend on where precisely The need for synchronization implies that phases of the the country is located, at a given moment of time, on the educational transition (from low basic and secondary edu- educational and technological transitions from "knowledge cation levels and quality to high levels of secondary and ter- absorbing" toward "knowledge creating" societies, and what tiary education) and the technological transition (from imbalances it shows in progress in those areas. The report reliance on adopting simple foreign technology to major thus offers a road map for policymakers and private sector 8 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES actors to identify policy priorities and actions to correct resources to public schools and universities generate perverse imbalances and accelerate educational and technological incentives in many countries in the region. Education is def- progress in a synchronized way. The section on Closing the initely not just a question of the number of schools and teach- Technology and Productivity Gaps and tables 1 2-1.4 pro- ers. We found too many cases in the region in which large vide a simplified version of such a road map. increases in expenditures in public education in the last decades were not matched by a corresponding increase in Students, Workers, and Firms: The Major Actors coverage or quality of Educational and Technological Progress The firm is at the center of the process of skill upgrad- In this report we focus our analysis on the major actors of ing and technological change. A firm makes decisions on the the educational and technological process. students, work- skills of its work force (through recruitment and on-the-job ers, and firms They are the ones that accumulate and use or external training) and the technology it uses (through knowledge to spur productivity and growth. Families, schools, adoption, adaptation, or creation) in an integrated way. At universities, research centers, and government institutions and the micro level the skill-technology complementarities are policies are just facilitators-though of critical importance- just too obvious. It is educators, economists, and policy- of the process of skill and technological upgrading in a soci- makers who sometimes miss the point Indeed, we found that ety Both the main actors and the facilitators must face the firms that are more aggressive in adopting, adapting, or cre- right incentives to contribute in an efficient way to thlis ating new technologies are also those that recruit higher process. skills and invest more in training Firms that do not change Students and workers will not accumulate enough technologies do not train. Firms that invest in R&D do more knowledge if they are not offered (or cannot identify) high- training. Not surprisingly, firms that both innovate and train quality education and training and see no prospects of prof- are the ones that show higher productivity growth (of course, itable employment for educated/skilled workers What they causality runs both ways here, as usual). Incentives are crit- need in today's world of rapid skill-biased technological ical. We found that firms more exposed to foreign compe- change are abilities to "learn," to adapt, to innovate, to work tition (whether of imports or in export markets) recruit in teams, and to relate to a wide variety of actors. It is not higher skills (and pay them better) and spend more in train- excessively specialized, "technical" education, at least not ing, licensing, and R&D until tertiary or postgraduate studies, and even then basic A critically important finding is then that the level of scientific knowledge in their areas and development of "prob- training by firms depends more on the competitive pressures lem solving" abilities are more important than mastering they face, and their incentives to adopt, adapt, and create specific techniques that may fast become obsolete. new technologies, as well as on the level and quality of edu- Students and workers need information about the qual- cation of the labor force they can hire, than on subsidies for ity of education and training offered by the diverse suppliers training (see chapter 5) In particular, firms in Latin America (hence the importance of state exams, sound accreditation and the Caribbean find little value in technical education and processes, labor observatories) and credit availability to over- training services offered by large public agencies, generously come liquidity constraints (and in the case of poor families financed by earmarked taxes This is an area in which reform outright subsidies to overcome high short-term opportunity has been under way in many countries (Chile, Mexico, and costs). Credit and information are especially important Uruguay are ahead of the curve in this area) in the region, in tertiary education, in order to convert the potential but still a lot remains to be done. Good educational and tech- demand created by rapidly rising wage premiums for nology policies are good training policies Existing training workers with tertiary education into effective demand and services must become more competitive And earmarked higher enrollment, particularly in the best private and taxes for training should either be allocated in a competi- public universities. tive way through matching grant schemes (to both in-firm Having the right incentives for public schools and uni- and external training) or to improved coverage and quality versities, as well as for teachers, is also crucial. Too often of general education. public resources follow the schools and the teachers, and not Flexible labor markets are important for firms to be able the students. Currently used criteria for allocation of to rapidly restructure the skill composition of their labor CLOSING THE GAP force in response to requirements of new technologies. Rigid 3 in developed countries and that R&D should probably be labor markets delay the pace of adoption of new technolo- increased by a factor of 2 to 10 in developing countries. gies and adaptations by firms. Hence, a variety of policies and instruments have been Similarly, access to risk capital is also critical for firms designed and implemented everywhere to support the process engaging in larger R&D and skill upgrading efforts. Hence of knowledge creation and diffusion: intellectual property the importance of deep and more sophisticated financial rights; tax incentives or subsidies to private R&D and/or markets-and, specifically, the need to develop venture training; and the creation of specialized public (or public/ capital funds-for countries wishing to step up efforts in private) educational, scientific, and technical institutions. adaptation and innovation. The report analyzes design issues for some of these instru- ments and where, when, and under which conditions these T'he RoI e oi Governlnent FoUcy interventions seem to have been more successful. It finds, The previous discussion highlights the fact that although for example, that tax incentives for private R&D are effec- students, workers, and firms are the main actors in the tive only in countries with a strong tax system character- process of accumulation and productive use of knowledge, ized by few exemptions and high enforcement. Transparent and that a variety of facilitating institutions or firms that and competitive matching grants schemes may be more effec- supply educational, training, and technological services pop tive in most LAC countries under present circumstances. The up spontaneously in the right environment, there is a critically most important conclusion, however, is that effective inter- important role for government policy in these areas. First, action (through so-called innovation networks) is as impor- governments must create the right environment through stim- tant as the individual design and development of institutions ulating openness to foreign trade and FDI, facilitating the and policy instruments to support the process of knowledge development of deep and sound financial markets and flex- creation and diffusion. ible labor markets and applying effective competition poli- cies. Only in such an environment are firms willing and able The CrIiDcal DmpporaSnce o [fedhve to engage in adoption, adaptation, and creation of new tech- Dnnovaiaon e3tworfl6s nologies and in recruiting higher skills and training. The importance of sectoral, national, and international "inno- Second, as indicated, credit and information constraints, vation networks" (or clusters) arises not only from the strong as well as opportunity costs for poor families, may inhibit links that exist between education and technology, but from potential demand for education to be converted into effec- the nonmarket character of many of the institutions and tive demand. Government support to create widespread organizations that are active participants in these areas and student loan systems; targeted educational vouchers, schol- the significant "informational asymmetries" that character- arships, and subsidies for students of poor families; and state ize the processes of knowledge creation, diffusion, and exams, labor observatories, and accreditation systems to deal absorption. We find that the Latin America and the Caribbean with asymmetric information problems are critical policy region not only lags in the level of stocks and flows of edu- actions to overcome imperfections in educational and train- cational and technological capital, but also often uses them ing markets. Appropriate governance and funding allocation poorly owing to a faulty incentive framework and to lack of rules for public schools and universities are essential to create integration of the countries' "national innovation systems" the right incentives in public educational systems. Supply (NIS) (see figure 1 4 and chapter 7) constraints are still important in many countries in basic edu- Linkages between universities and private sector firms are cation, especially in rural areas important even at the early stages of technological catch-up, Third, knowledge creation and diffusion is characterized when most firms are primarily engaged in the adoption and by imperfect appropriability and high spillovers, that is to minor adaptations of foreign technologies. (At this stage of say, the social value is much higher than the private value development, universities and public research institutions and, thus, in the absence of effective subsidies or protection, are likely to be the only institutions where scientific research firms and markets will produce less than what is socially and technological knowledge are acquired in a significant optimal In particular, the literature finds that the social value way, and thus they are critical supports for leading firms and of private R&D exceeds its private value for a multiple of sectors that engage in more significant adaptations or even 10 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES FIGURE 1 4 A Simple National Innovation System (NIS) Human Capital ;. | Th~~~~ink' lnks/ _ University Antenna Innovatin on Clusters- ' ,, / ,," p- | * | Firms Other Public Policies Rules of the Game .1+ Infrastructure (ICT) SubsLdies/Tax Incentives Innovation and TFP Growth] Coordination Initiatives in the creation of technologies.) Strengthening those links centers)/firms skills upgrading and R&D efforts, to help over- becomes even more important in the second phase, in which come some of the asymmetric information and coordination a large number of firms begin to be engaged in major adap- problems. tations and leaders engage more in creation. However, a More generally, the so-called National Innovation System number of institutional design problems can blunt the incen- can hardly be called a "system" in Latin American countries tives for firms and universities to interact, and public policy and it produces little innovation. It is not-nor should it can play a very important role in redressing them be-strictly national either. Sectoral and international Latin American countries may want to start by paying dimensions are critical. close attention to designing policies that strengthen incen- Firms interact with other firms in a variety of ways-and tives for such collaboration. Collaboration is ineffectual proximity (or virtual clusters) can reduce the transactions owing not only to problems of asymmetric information (firms costs of these interactions and improve the speed and qual- do not know well what universities can indeed offer, and uni- ity of the flow of information among them. More important, versities do not know well what firms need), but also because as there are significant spillovers arising from in-firm train- of bad incentives Of particular importance are issues of intel- ing and R&D efforts, sectoral clusters help internalize in lectual property rights of research results obtained in uni- many ways these externalities (by setting cooperative skill versitles and through public funding (individual researchers upgrading and R&D efforts and establishing "rules of the must benefit from a substantial part, but universities or inde- game" among private actors), thereby providing a boost to pendent research centers need to share to compensate for the these activities. They also help address the asymmetric infor- financial and institutional risks they incur in supporting mation and coordination problems that plague the rela- these activities) as well as other incentives embedded in the tionships of market and nonmarket actors in knowledge structure of benefits and promotions within universities and creation and diffusion activities. Indeed, sectoral innovation public research centers. In Latin America and the Caribbean systems or "clusters" can be developed in a rather indepen- some degree of cultural reticence must also be overcome, on dent way (the innovation system in copper mining may have the part of public universities as well as firms. Public policy very little that intersects with that of agricultural exports supporting private R&D and training might give some or of fisheries or forestry in Chile, or elsewhere). They require preference to joint university (or independent research the development of specialized skills and R&D efforts (which II CLOSING THE GAP is not to say that success stories may not spill over to other personal interaction, the international circulation of skilled sectors, as happened in Finland with the Nokia case or in workers and students is a critical aspect of a successful the U.S. mining Industry in the last century). Governments National Innovation System. Study-abroad programs and can and should support efforts of such emerging innovation temporary emigration of workers is a major and highly effi- clusters. However, attempts to create clusters from scratch cient source of technological transfer. Clearly brain drain have often backfired and there are no clear policy design hurts the country of origin in the short run because in most recipes that can be followed. Evidence shows that existing developing countries education is publicly funded and successful clusters seem to be the result of private entre- migration represents a drain on resources. But even brain preneurship that benefits from a positive business environ- drain can be turned into a gain, as witnessed by the critical ment: open trade regimes, strong domestic legal and financial role of the Indian engineers in Silicon Valley for the subse- institutions, and sound educational and technology policies. quent development of the domestic software industry or the Most successful experiences suggest that well-designed sup- Chinese or Jewish diasporas for the rapid technological port policies for R&D and skills upgrading, of a general catch-ups of China and Israel. nature and coverage, might be efficiently implemented through existing and incipient sectoral clusters and some- OwDiDng Qlltie MMII00 CSaps what tailored to their specific needs. As mentioned, new technologies have particularly increased As important, any successful National Innovation the productivity of educated workers, which means that skill System must have an international dimension. Knowledge upgrading now should have a larger effect on growth rates creation and diffusion is essentially an international than it did before. The region must then close the skill gaps phenomena. Most new knowledge will continue to be pro- as fast as possible. Chapter 4 shows that by 1960 the region duced in a handful of advanced countries, where the largest had levels of education comparable to those of East Asian stocks of human capital and innovation networks exist, and, countries and many peripheral European countries Those thus, both firms and national education and innovation sys- countries, however, engaged in more rapid educational tems must effectively link with them if they want to suc- transitions and thus we observe today major gaps in Latin ceed in adopting, adapting, and creating technologies in an America, as indicated above. efficient way Thus, national educational systems in Latin Most successful episodes of educational upgrading-the America and the Caribbean must be well connected with United States between 1850 and 1950, Korea and the the best educational systems abroad. These links are espe- Scandinavian countries since World War 11-have all fol- cially critical at the tertiary and postgraduate level. Link- lowed a sequential pattern of upgrading, first building up ing up local universities and autonomous research centers universal basic education, then broad secondary education, with world-class institutions can be relatively straightfor- and finally generalized access to universities. Educational ward and yield a high payoff. transitions are discussed in chapter 4 and represented in The ability of domestic firms to network with foreign ones figure 1.5, in which the top part shows the transition fol- will usually require that the domestic firm already have some lowed by most OECD and East Asian countries. The figure degree of indigenous technological capability This makes it shows that while most countries sought to gradually build difficult to come up with general policies that can jump-start the mean level of general education. others focused on ter- such linkages, other than removing regulatory hurdles to tiary education. FDI, or other export or import regulations that prevent domestic firms from integrating with global chains. How- Public Policy and Funding in Most Countries in Latin ever, successful sectoral innovation clusters have to be con- America Should Focus on Expanding Secondary nected with their peers abroad. Government policy might Education, while Facilitating Private Expansion especially support some skill upgrading and R&D initiatives of Tertiary Education of existing sectoral clusters that have such an international The increasing returns to secondary and tertiary education enhancing dimension. suggest that the technological transition will require more Finally, as a lot of knowledge is "disembodied" and can workers of this skill level. However, as mentioned before, hardly be transmitted (in spite of ICT progress) without the most distinctive feature of the Latin America and 12 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES FIGURE 1 5 Educational Transitions While most countries sought to gradually build the mean level of general education pi The pyramid (most countries, circa 1960)- - with a narrower base The diamond The inverted pyramid (e g, East Asia, 1980s) (e g , East Asia, current) (e g , North America) some countries focused on tertiary education The lopsided anvil (yunque) (e g , R B de Venezuela and Costa Rica) Highest level of education 7 Educated beyond high school I Secondary only Primary or no education Caribbean region's average educational landscape is a large dance to tertiary education In the other extreme, there are shortfall in secondary-educated workers, compared to mild laggards that still need to focus most of their efforts on deficiencies at the tertiary level given its level of develop- improving coverage and quality of basic education ment. In this sense, the region has followed an "unbalanced" educational path, when compared with OECD and Asian The Best Ways to Expand Access to Education Vary countries This pattern makes poor use of the intrinsic com- with Bottlenecks Faced plementarities within the educational system since it Are students completing primary school and not enrolling reduces the efficiency of tertiary education (as there is a nar- in secondary school, or are they dropping out midway rower talent pool from which to select students for tertiary through secondary school? Low enrollment levels may them- education). It also makes poor use of the complementary rela- selves reflect either insufficient effective demand by mdi- tionship between the educational and the technological tran- viduals, or insufficient supply of educational opportunities. sition of a country, as a widespread coverage of secondary The demand for more education by individuals should be education has been found to be a condition for countries to closely related to the demand for more educated workers by fully reap technological spillovers from FDI and trade. It is firms. as relative wages rise, we would expect young people also likely to be unsustainable since it exacerbates inequal- to seek more schooling. This demand for more schooling by ity. The lesser deficit, and the higher private returns to ter- individuals may fail to materialize, however, because of credit tiary education may suggest that public policy (and in constraints, uncertainty about the returns to education, and particular the allocation of public resources) should focus other information market failures. In other cases, the prob- more on increasing secondary education, while at the same lem may lie not with the demand for education by indi- time facilitating the private expansion of tertiary education. viduals, but with the supply in the education system. The This general pattern notwithstanding, some countries in number of slots in a public university may be fixed (as there the region-in particular Chile, Argentina, Uruguay, and is not enough cost recovery and most countries face fiscal con- many of the English-speaking Caribbean countries-have straints), and in rural areas in some countries there may done a better job on secondary education and their educa- simply be no secondary schools. The appropriate mix of tional push should now center more on broadening atten- policies will therefore be country-specific, and should be 13 CLOSING THE GAP determined after a careful analysis of bottlenecks, demand, given fiscal shortages in most countries. In many coun- and supply constraints in the education market tries, where costs per student are inordinately high by We find in several countries in the region indications of international standards, there are significant possibilities both supply constraints and insufficient effective demand for of increases in enrollment for a given budget. In most, secondary education. Such insufficient demand for secondary higher cost recovery, if coupled with widening credit and schooling may be related either to insufficient demand by targeted scholarship programs to relieve liquidity con- firms and to high opportunity cost of schooling. Policies that straints, would be an efficient way to increase coverage. promote technological change have the potential to increase This obvious course of action, however, appears to be con- the returns to schooling and the willingness by young men strained by the same political economy forces that led in and women to defer entrance into the labor market. So will the first place to unbalanced educational transitions and those that improve quality, and hence returns of secondary high subsidization of tertiary education for the elites and education. And, as an important motivation to finalize middle classes. In such cases, there is at least the need to secondary school lies in the possibility to proceed to tertiary transfer public resources according to performance, thus education and reap the benefits associated with the high creating appropriate incentives for efficiency, and for wage premiums for workers with higher education, students increases in coverage and quality. must know that they will have a fair chance to get a university education (see below). Programs that subsidize the cost of Training Policies Have in General Overemphasized secondary schooling for students from poor families, such as Public Provision and Underemphasized the Importance conditional cash transfers like those in the Progressa program of Good Technology and Education Policies in Mexico or the Bolsa Escolha in Brazil, may also be effec- Chapter 5 considers training and lifelong learning policies, tive in overcoming gaps between perceived moderate returns identifying several broad types of technical education and to secondary education and high opportunity costs of forgone training systems in Latin America and the Caribbean, employment. including the classic corporatist "Latin American Model" Insufficient supply of secondary school infrastructure (for example, the approach followed by Brazil, Colombia, appears to be a problem in rural areas in many Latin and Peru) and the approaches adopted by countries such as American countries and even in urban areas in some. In these Chile (which relies on training incentives to firms) and settings, programs to increase the availability of schooling Mexico (which relies more on public sector provision of facilities are required. training). We examine the relative benefits of these We also find that, in many countries in the region, the models-and other approaches found in East Asia and main constraint for private universities is effective demand, OECD countries-with the aim of drawing out the most while that for subsidized public universities is supply. appropriate policy mix for the region in the coming decades Although there is high potential demand for private univer- Using firm survey data for 22 Latin American countries and sities, because of the high returns, this does not necessarily comparing the findings to data for selected OECD and East translate into high effective demand because of liquidity con- Asian countries, we find that the public provision of tech- straints and information asymmetries. Student loan and tar- nology-oriented training is generally viewed as relatively geted scholarship programs hold promise, if they are well inefficient and irrelevant. designed. Providing households with information about the Evidence from evaluations in East Asia indicates that well- quality of private providers (through state exams for grad- run schemes such as tax rebates are effective in increasing in- uates, labor market observatories, and sound accreditation service training, but the evidence from within and outside programs) will help them make better choices and, by LAC overwhelmingly indicates that education levels of work- reducing uncertainty and bridging information gaps, may ers and introduction of new technologies are the most impor- increase effective demand. tant correlates of in-service training. The main lesson is that In contrast, public universities are heavily subsidized LAC countries should view "training policy" as not Just sub- in many countries and, as a consequence, face unmet effec- sidizing or providing training, but also increasing the demand tive demand Increasing aggregate budgets to public uni- for training through appropriate technology-related policies, versities is generally not a feasible or desirable solution, and increasing the trainability of workers through appropriate 14 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES education policy. Education seems to be the most important and innovations cannot be done in-house but in the largest enabler of training, and technology change its most impor- of firms and thus require effective "innovation networks," tant instigator. The main reform needed for explicit train- at least at a sectoral level ing policies is a change in the existing public-private balance However, it must be kept in mind that even in developed in training provision Many LAC countries (most notably countries most of the firms are fundamentally adopting or Chile, Uruguay, and Mexico) have been progressing in this adapting with just a few doing cutting-edge innovation. To direction, but much still remains to be done. wit: it has been estimated that 86 percent of France's tech- nological progress is purchased from abroad, the remainder Lifelong Learning Has Always Been Important being generated locally (Eaton and Kortum, 1999) and that Lifelong learning is necessitated by the obsolescence of gen- resources dedicated to "adoption" in the United States, eral skills acquired in schools, as compared with retraining, widely recognized as the innovation leader, exceed those ded- which is necessitated by the obsolescence of specific skills icated to innovation at the frontier by a factor of roughly acquired on the job or through technical education. The 30:1 (Jovanovic, 1997) report addresses the following question: Has the labor market For the sake of convenience, we offer a stylized view here changed so that the emphasis on lifelong learning is now of the technological transition that countries pass through. greater than it was before) The evidence from six LAC coun- The choice of policies and instruments, we argue, depends tries appears to indicate that it has not: the "new economy" on their levels of technological and educational achieve- may not be that different from the old one. There is no doubt ment-country conditions and endowments. Technological that rapid technological change, especially skill-biased tech- progress and innovation is a process and evolves over time nological change, requires constant retraining. But even if as conditions change. We thus attempt to classify the coun- lifelong learning is more important now than ever before, the tries of Latin America according to these rubrics, and draw main policy implications are in fact for education systems. And appropriate policy lessons for each stage of the technologi- the implications are that schools should teach how to learn, cal evolution. not occupation-specific skills, and the importance of secondary vocational education should decline over time as occupation- "Plugging In": The First Stage in Modern specific learning gets pushed up to tertiary levels. Technological Transitions We consider a country to have entered the "plugging in" Closing the Technology and Productivity Gaps stage when most of its firms are in a state of relative tech- The Progressive Stages of Technological Evolution nological stagnation and it takes the conscious decision to and the Synchronized Phasing of Educational accelerate the transfer of new technologies by achieving full and Technological Policies coverage of basic education, opemnng to trade, FDI and ITC, In chapter 6 we identify the instruments and policies for facilitating licensing and skilled labor circulation, and more technology acquisition and development, and evaluate their generally, creating a competitive environment for firms that impact on productivity. Firms usually begin their techno- induce innovation and skills upgrading. Transiting through logical catching-up by adopting more modern technologies this stage will unleash a significant increase in potential in a rather straightforward fashion, though in many cases demand for education by rapidly raising premiums for even this requires some skills upgrading and R&D to appro- secondary and tertiary education. Most countries in Latin priately select and adapt technologies to local conditions. America and the Caribbean have already gone successfully Later on many successful firms begin to do major adapta- through this stage. tions, which require stepping up efforts in skill upgrading Access to global markets exposes firms to newer products (through recruitment and training) and significant R&D and processes-which leads to a process of technological investments. Finally, the most successful firms, especially imitation and some adaptation-the first step toward any when competing internationally, require major innovations catch-up. However, technology adoption is not an easy and and development of new technologies to keep a competitive costless activity; even the first stage of adoption-selecting edge. For this, they require highly specialized skills and the right technology-implies an ability to gather relevant major investments in R&D Progressively, major adaptations information and make astute judgments And quite often 15 CLOSING THE GAP it requires at least some minor adaptations to the quality of TABLE 1 2 local inputs and other local conditions. First Stage For countries that find themselves farthest from the tech- EDUCATION nological frontier (such as Haiti), the cornerstone of their GENERAL CONDITIONS POLICY PRIORITIES technology policy should be their education policy-a push Basic Low coverage 1 Increase public spending to Increase toward getting to the threshold level of basic education cou- Secondary Low demand coverage of basic education pled with an open trade policy and deregulation of markets Tertiary Low demand 2 Institutional reform to increase efficiency and quality that enforces market discipline and discourages rent- seeking behavior, understanding that the lower the educa- POGY GENERAL CONDITIONS POLICY PRIORITIES tional levels the lower the benefits from trade and FDI. Trade Low Increase openness to trade by As countries transit through this stage, however, simplifying structure of protection improvements in the quality of basic education and FDI Low Promotion of FDI through EPZs increased coverage of secondary education begin to IPR Low CopyrLght and trademark protection R&D Low Regulatory reform to reduce increase in importance, as does beginning to step up efforts disincentives for R&D to attract FDI, particularly through institutional shortcuts Selective subsidies to private R&D ICT Low Telecomunications sector reform and such as special regimes for export promotion zones. Copy- friendly ICT legal framework right and trademark protection and some form of patent Networks Nonexistent Public support to emerging innovation protection do need to be considered at this stage to facil- clusters itate technology transfer from abroad. Specific support pro- grams to skill upgrading and R&D efforts in emerging innovation clusters, normally in sectors with traditional First priority for these countries, in educational policy, comparative advantage such as those based on natural is to close the large secondary education gap. As explained resource endowments, may have high payoffs. It is also earlier, to achieve this countries need to overcome remain- important to engage as early as possible in telecommuns- ing supply constraints (mainly in rural sectors), improve the cations sector reform, to increase competition and cover- quality of secondary education to increase expected returns age, and the development of a friendly Internet legal and effective demand, and target subsidies to poor families framework to increase connectivity. Countries at the upper to compensate for the high opportunity costs of schooling. end of this stage include Nicaragua, Guatemala, Bolivia, But, as an important motivation to finish secondary educa- Honduras, and Paraguay (see table 1 2) tion is to be able to proceed to tertiary education and reap the high wage premiums, the countries also need to put in Catching Up-Moving toward the Technological Frontier place a well-designed system of credit (and information) for The first stage is followed by a relatively long period of tertiary education, so that students from poor families know "catching up," in which most firms are busily engaged in they have a fair chance to proceed to tertiary education if technological transfer and catch-up, but where innovations they indeed finalize secondary school are still scarce A natural boost to making the transition to Countries in this bracket, in which a large number of this next stage is usually received when firms move from firms need to transition from simple adoption to more com- domestic to export markets and when they are exposed to plex adaptations, are likely to benefit from more active pro- significant levels of competition. Demand for workers with motion of technological transfer and adaptation. Further higher education is high, but these countries still face a large encouraging trade and FDI by engaging in bilateral and mul- secondary education gap. Skill needs become more special- tilateral agreements that ensure market access, and improv- ized, as does the policy environment. As a country transits ing the investment climate through active promotion through this stage a larger fraction of firms engage in skills policies (as practiced by Costa Rica) will accelerate the pace upgrading through training and in R&D for adaptation of of technology transfer. Technology diffusion and significant foreign technologies, reflecting the need to take better advan- adaptations, as well as creation by an emerging set of lead- tage of local resources or to modify the technology to better ers, should be fostered by implementing a patent protection suit local conditions. regime that is trade-related intellectual property rights 16 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES (TRIPS)-consistent, but not as strict as to impose high TABLE 1 3 short-run costs in terms of slower diffusion; by fully devel- Second Stage oping the ICT sector, and by establishing a broader pohcy EDUCATION of support for private R&D and training through competl- GENERAL CONDITIONS POLICY PRIORITIES tive grant systems, in lieu of the previously more focused Basic High coverage 1 Increase resources and efficiency support to a few innovation clusters. Secondary High demand in secondary education It is also important at this stage to improve incentives Tertiary Low coverage 2 Facilitate access to higher education IsoTertiary High demand (through student loans and state for university/firms cooperation in R&D and specialized skills Low coverage accreditation and exams) upgrading As mentioned earlier, chapter 7 finds that most 3 Improve cost recovery and allocation of resources to public universities of R&D in Latin America and the Caribbean is currently according to performance being done by the public sector and universities without TECHNOLOGY enough links and spillovers on R&D by firms and special- GENERAL CONDITIONS POLICY PRIORITIES ized private agencies, in sharp contrast with what happens Trade Increased Trade agreements to guarantee in more successful countries, and that many of them still have market access generously financed public or tripartite training institutes FDI Increased (Mauuilas Active promotion and natural Improvement of investment climate that tend to get behind the technology curve. resoLurces) Labor market flexibility and deep financial sectors (in par- IPR Low Increase protection of ineellecrual ticular, access to venture capital) become more important in R&D Low property rights (TRIPS consistent) R&D Low ~~~~~Establish competitive subsidy schemes this stage as an increasingly larger number of firms engage to promote private R&D in technological and skills upgrading. ICT Low Development of ICT infrastructure Networks Incipient Promote university-firm linkages Most Latin American countries, such as Brazil, Colombia, Fragmented Increase support to emerging Costa Rica, the Dominican Republic, El Salvador, Panama, innovation clusters the Repiblica Bolivariana de Venezuela, and the English- speaking Caribbean countries, are in different phases of this What is noteworthy is that newly advanced countries such stage. Thus, in summary, the central priorities for most LAC as Korea, Israel, and Finland find themselves in this first tier countries at present are closing fast the gap in secondary edu- also. Despite frequent references to such nations as "new cation whlte continuing to expand tertiary education, and entrants" to the league of innovative economies, the notion at the same time creating better incentives for private R&D of suddenly "arriving on the scene" is misconceived Firms and training, including an overhaul of present predominantly in these countries have been engaged in making product and publicly funded and executed R&D and training schemes process improvements for many years It is also true, how- (see table 1.3) ever, that there appear to be "turning" points in many cases, when investments in R&D by firms and patent registration Joining the League of Innovators (Just to mention two major indicators of technological inno- Firms that have been present in global markets-adapting vation) accelerate at a vary rapid pace and such cases seem existing technologies and selling the resulting products at to be related to a major push by governments and private lower cost than their competitors-find margins eroding as sectors to create a fully integrated "innovation system " Such new low-wage entrants take away market share. Sustaining countries appear as outliers in terms of their efforts in R&D a position in global markets then necessitates a leap forward and their outcomes in terms of patents when benchmarked into creating new products and processes (which is greatly with respect to their income level (see figure 1.6). Such efforts facilitated by the acquired knowledge), as well as capital are typically associated with very high overall productivity accumulation, from the adapt-and-export stage Countries growth and, as chapter 7 stresses, the rates of return appear in this tier normally have full coverage of secondary educa- to justify these investments (see figure 1 7) tion for new cohorts of workers and rapidly increasing enroll- Countries in this stage need to cope with the large frac- ments in tertiary education. Postgraduate programs begin tion of students who, having finished secondary education, to flourish As expected, the United States and Japan are far are demanding access to tertiary education. Student-loan pro- ahead of the rest of the world. grams must become massive, targeted scholarships need to 17 CLOSING THE GAP FIGURE 1 6 Superstars in R&D 5 0% 4 5% 4 0%o Israel 3 57 Finland 0% 205% co 1 0% China .orea 0 5%o India 0 0% Argentina 4 5 6 7/ 8\ 9 10 11 Mexico Costa Rica Log GDP Per Capita Source Lederman and Maloney (2002) FIGURE 1 7 Returns to R&D and Physical Capital 120% 5 45 100% 4 Investmet in R:&D should be 3 5 .80%6- 2 L _to15 times highe < NReturns toR&D -3 ., X ~~~~~~~~~~~~~~Optimal R&D/R&D1 0 60%o- 22 5 X ~~~~Returns to2 f Physical KO 0~~~~~~~~~~~~~~~~~~~~ 4 0%- w BRA AR~~~~~~~G US1 5 NIC J A'~~~~MX CHL KOR FN ~ _ _ 20% 9o u=0 0 5 0% 0 2,000 7,000 12,000 17,000 22,000 Constant 1995 PPP US$ 18 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES be increased substantially, and information systems (state Finland made when they had a similar level of income, or exams, labor observatories, accreditation systems) need to be that India and China appear on the way to making. Other substantially strengthened Postgraduate systems, particu- countries, such as Argentina and Uruguay, have not only not larly in sciences and engineering, must receive a strong push taken off in terms of innovation in spite of a high level of With the distance to the technological frontier narrowed human capital, but they are also far behind the "normal" even more, technology policy gets even more complex. The countries at their level of income. As chapter 6 notes, many objective now is to stimulate the creation of new technolo- LAC countries have today the level of tertiary education that gies by a large number of firms and enlarge the fraction of the NICs, Finland, and Israel had at the beginning of the those that engage in major adaptations of foreign technolo- 1980s during their takeoff period and hence it would appear gies (most will continue to transfer and adapt technologies timely for them to make a big push in policies to encour- from abroad) Advancing to the innovation stage requires a age private RD investment (see table 1.4). significant leap forward in private R&D. On the part of firms, The transition that we have sketched above should be it necessitates large investments in in-house R&D activity viewed as a continuous process, though there appear to be or close cooperation with universities and independent some significant nonlinearities in the path followed by many research centers Governments must support these efforts by successful countries, particularly when shifting to the inno- stepping up in a considerable way subsidies through com- vation phase: a concentrated push in innovation policies petitive matching grant systems, establishing tax incentives and a corresponding exponential increase in private R&D, for R&D, increasing the financing and provision of basic patents, and productivity growth. And progress through the research, strengthening intellectual property rights, deep- different stages of this transition may be quite uneven among ening credit markets, and continuing to upgrade ICT-related sectors, as well as among firms in each sector (See box 1.3.) infrastructure to world class levels. At this stage much of technological advance is tacit TABLE 1 4 (embedded in individuals) rather than coded (available on Third Stage "blueprints"), and so issues related to labor mobility and coop- erative networks of firms and institutions of higher learning EDUCATION and research (sectorally, nationally, and internationally) GENERAL CONDITIONS POLICY PRIORITIES assume even greater importance. A coherent and stage- Basic High coverage I Significantly increase resources Secondary High coverage for higher education (credit/ dependent innovation and R&D framework is essential to suc- Tertiary High demand scholarships) cessfully reach the creation stage Most of all, the incentives Medium coverage 2 Develop postgraduate programs, facing all actors have to be clearly aligned and a coherent, specially in sciences and engineering 3 Strengthen state accreditation integrated innovation system must be in place. This has not and exams happened in the Latin America and the Caribbean region. 4 Further improve cost recovery and allocation of resources to The evidence of the impact of a well-designed system is public universities clear Evidence presented in this report shows that patent TECHNOLOGY rights, public funding, and human capital stimulate private GENERAL CONDITIONS GENERALCONDITIONS POLICv PRIORITIES R&D in OECD and East Asian countries. Private R&D and Trade High Trade agreements to guarantee market public R&D in turn stimulate domestic patenting and for- access eign transfer of technology, which contribute to the stocks FDI High (including Active promotion :ndustrie of domestic and foreign-generated knowledge capital, intensiveIn R&D) which in turn lead to higher productivity. Such links, IPR Increased Increase protection of intellectual especially those from public R&D to private R&D, are found property rights to OECD levels R&D Increased Strengthen private R&D competitive to be weaker in Latin America and the Caribbean than subsidies and tax incentives, and elsewhere public R&D spending ICT Increased Advance to OECD levels Indicators for Chile and Mexico suggest that they are just Networks More developed Promote integration of a strong in the first stages of this phase, but have not yet engaged Better integrated national innovation system with in the kind of dramatic takeoffs that Korea, Israel, and international links 19 CLOSING THE GAP BOX 1 3 The importance of Dliversity and Leaders In this report we insist on the importance of balanced, sectoral innovation clusters should be part of educational and bottom-up, and synchronized educational and technolog- technological policies even in countries that are in the first ical transitions. Obviously, however, this does not mean stages of the educational and technological transition. that all students and workers must have the same level of Natural resource-based sectors are especially important, education and skills or that all firms in an economy will as even when technological transfers from abroad are pos- be in the same phase of the technological transition. sible, they normally require significant adaptations to the Diversity is always important and in the fields of knowl- characteristics of indigenous natural resources and their edge and technological progress, more than in others, lead- particular environmental conditions. Major examples can ers play a critical role. Thus, even in countries that still have be found in the successful development of fresh fruits or low coverage of their basic education there are a few uni- salmon in Chile or cut flowers in Colombia. Even more, versities that train the elites and conduct research-and there is very little R&D in tropical agriculture (or health) they would rather be good and play a useful social role. in OECD countries for obvious reasons, so in these areas There is, of course, the question of whether in such a case either we do it or productivity growth will likely stall. Our the state should subsidize higher education for the elites- 2001 flagship publication, From Natural Resources to the as most LAC countries do-and the answer is: of course not. Knowledge Economny (de Ferranti and others, 2001), of which But the state may well give loans or fellowships to the few the present report is a natural outgrowth, presented students from poor families who can complete secondary strongly the case that Latin American countries' histori- i school so that they have a fair chance of proceeding to ter- cally lackluster performance in comparison to other more tiary education; it may also support joint applied research successful natural resource-based economies (such as Scan- projects that effectively link these universities with firms. dinavian countries, Australia, New Zealand, Canada, and As important, even in economies in which the large the United States), hlad a lot to do with the fact that they majority of firms may be stagnant, in technological terms, did not invest-and still do not invest-enough in skills or just adopting foreign technologies in a straightforward and R&D in these sectors. way, there will be some doing major adaptations and even Something similar may happen in some service sectors significant technological innovations. This is most likely the such as marketing, finance, and others that require sig- i case in sectors in which the country has for long had a com- nificant adaptations to local tastes, culture, and institu- parative advantage, as is the case with those based on nat- tions Contrary to popular belief, one may expect more ural resources. It is important that government policies R&D in primary and service sectors, as compared to man- support such leadership efforts even in countries that are lag- ufacturing sectors in which straightforward transfer of ging overall in the technological transition. Hence, sup- technologies is more possible, at least in the initial stages porting skill upgrading and technological efforts in incipient of the technological transition Summing Up tries seek to accelerate the pace of technological progress, One of the abiding mysteries in development economics has ensuring that the right human capital is available and coor- been why poor countries have not aggressively exploited the dinated effectively with technological policies becomes immense global stock of knowledge to accelerate growth. central. Increasingly the literature focuses on shortcomings in The evidence clearly shows that the higher-performing national innovative capacity. High levels of human capital countries that made a transition to full partners in global and exposure to foreign technologies-for instance through innovation-Israel, Finland, Korea, for example-have dra- trade, FDI, licensing, and the international circulation of matically increased both their level of human capital and skilled workers-are critical not only in their own right, but their investments and policies for innovation in a concerted also vitally in how they complement each other. As coun- fashion. In addition to getting the basics right in terms of 20 INTRODUCTION AND SUMMARY SKILLS UPGRADING AND INNOVATION POLICIES plugging in to the global knowledge economy, they also said, the successful countries have consistently taken an active came to terms with two fundamental issues affecting every approach to integration in the world economy-upgrading the country following in their footsteps. learning and training capacity of firms, selectively financing First, knowledge as a commodity is plagued with extra- private R&D, encouraging the licensing of foreign technolo- ordinary market failures and hence the market will not gen- gies, protecting intellectual property rights, stimulating the erate the optimal level of innovation. To repeat, serious development and access to ICT and progressively deepening analysts argue that the United States should probably and tuning up their National Innovation Systems rather than increase its R&D by a factor of 4 and we offer evidence that passively waiting for multinational corporations or imports to LAC is also even further below optimal. Furthermore, the transfer technology. Thus engagement in the long process of institutions created to resolve these market failures- undertaking the necessary institutional reforms needs to start universities, government laboratories, intellectual property early in the development process. rights-lie, by definition, outside the market and hence are Though arguments for traditional "industrial policy" have not coordinated by the price mechanism. largely been discredited, a government's role in providing Second, a critical aspect of the process of development is the necessary innovation and skill upgrading-related com- that firms, and the country as a whole, "learn to learn." In plements to previous reforms provides a challenging policy particular, increasing the technological absorptive capacity agenda over the next decades An active and efficient of the firm has required a supportive set of policies and insti- "innovation policy" is required, and though many of its tutions ranging from well-designed fiscal incentives and sub- components and institutions should be neutral across sec- sidies to the active promotion of collaboration through tors, some need to be tailored to support emerging innova- incubators, technological parks, and clusters; to the creation tion clusters in particular sectors. As this report will show, and coordination of industrial consortia that share the costs most countries in the Latin America and the Caribbean and risks of R&D and skill upgrading and serve as learning region lag in almost every dimension of educational and tech- laboratories for less advanced firms; and to establishing nological achievement As a region, to rephrase Pasteur's antennas for identifying ideas from abroad. quote, our collective mind is not yet prepared to take advan- Both considerations demand an integrated approach and tage of the unpredictable technological opportunities that a coordinative and even leading role for government At a the new millennia will present us. minimum the state needs to ensure a consistent and coher- ent set of incentives to ensure that the institutions created Endnotes to address market failures collaborate fruitfully with firms. 1. See de Ferranti and others (2001, chapter 3), Hall and Jones In the highly successful countries, governments have not (1999), Parente and Prescott (2000) been shy about financing and undertaking R&D that has 2 Of course, differences in skills and "technology" do nor explain broad spillovers. away all differences in productivity levels. Differences in quality of Not all countries are at a stage where undertaking such poli- institutions, depth of financial sectors, macro/fiscal performance, among cies is feasible It probably does not make sense to contemplate others well known in growth studies, contribute in an important way b d e n n nn re to such differences See Loayza, Fa)nzylber, and Calder6n (2002) 3. It is important to note, however, that low net enrollment rates closed, if basic institutional integrity is in doubt, or as the at the secondary level are partly the result of high repetition rates and report has stressed, if the required human capital is absent. That late entry into primary school 21 CHAPTER 2 The Gaps That Matter Most T _ HE CENTRAL PROPOSITION OF THIS REPORT IS THAT SKILLS AND TECHNOLOGY interact in important ways, and are a fundamental reason for the large observed differences in productivity across countries. Focusing on gaps in productivity, technology, and education is a natural way of considering differences in welfare across nations, and goes back (at least) to the work of economic historian Alexander Gerschenkron. (Gerschenkron, 1962; see also Fagerberg, 1994). But estimating the size of gaps is not an easy matter for a number of reasons. First, data are not always available on appropriate measures of productivity, skills, and technology and, when they can be found, these data are often plagued with measurement error. Second, different countries are at different stages in development: a simple comparison of the educational attainment of adults in the United States and Peru is obviously not very informative, but one of Peru and Thailand may be. Much of the early work by Gerschenkron and others analyzed the gaps between "leader" and "technologically backward" countries. In this chapter, we take a different approach. Most of the evidence we present assesses whether countries in Latin America and the Caribbean have high (or low) levels of productivity, skills, and technology for their levels of per capita income. The estimated gaps are therefore gaps relative to the level of development. They provide a benchmark for countries in Latin America (box 2.1). Throughout this chapter and in much of the report, we The Productivity Gap particularly stress the comparison between Latin America Table 2.1 summarizes total factor productivity (TFP) growth and two sets of economies the successful East Asian "tigers," for a sample of countries in the OECD and all developing including Hong Kong (China), Korea, Malaysia, and regions for the four decades between 1960 and 2000 Singapore; and a group of countries abundant in natural (Loayza, Fajnzylber, and Calder6n, 2002). (The top panel resources, including Canada, Australia, New Zealand, presents simple averages across countries, while in the Finland, Norway, and Sweden The comparison with the East bottom panel the averages have been weighted by a coun- Asian tigers is a natural one because both sets of countries try's GDP.) Specifically, we present the results of a Solow started at a similar level of development in the 1960s. In decomposition of output growth into the contributions of last year's report, From Natural Resources to the Knowledge capital, labor, and productivity growth. Note that this is Economy, we stressed that many countries in Latin America the simplest possible Solow decomposition: it does not con- have a comparative advantage in natural resources For this trol for changes in the "quality" of labor (for example, if reason, the comparison with other countries abundant in nat- the mean education level of the labor force rises), or changes ural resources is also informative. in the "quality" of capital (for example, if new technologies 23 CLOSING THE GAP rates of TFP in the post-1980 period. By contrast, TFP BOX 21I iBenchmarking Latin Amefca growth in the last two decades has been high in East Asia and South Asia. Table 2.2 presents the same information, We benchmark countries by running a linear OLS regres- by country. Table 2.2 clearly shows the differences between sion of the outcome in question (for example, the net Latin America, East Asia, and the natural resource-abundant secondary school enrollment rate, or the number of countries: in Latin America, only Chile had robust TFP patents registered) on the log of per capita income. Coun- growth rates in both the 1980s and 1990s, while Argentina, tries that are significantly below the regression line are Bolivia, Costa Rica, and Uruguay had respectable growth then taken to have a deficit or gap More complicated rates of TFP in the 1990s. Korea and Singapore were good formulations, which control for time or year effects performers among the East Asian tigers, while Finland and affecting all countries, rely on more general functional Norway also did well forms and, when necessary, control for variables other than GDP, or use estimation methods other than OLS Thie MDfiMD Cap (for example, count-data techniques for regressions on Latin America is falling behind in the global TFP leagues. patents), are also possible. Some results from these more What explains this? We first examine differences in skills. complex approaches to benchmarking are presented in Since there is no perfect measure of skills, we consider three later chapters in the report. components. We start off with a discussion of the stock of The sample in the regressions in this chapter educational attainment of the adult population in Latin includes all of the countries in the world for which data America. We then consider theflow of skilled workers in the are available. Regressions are generally weighted by pop- region, as given by current enrollment at various levels in ulation or GDP, as detailed in the notes at the foot of the education system. Finally, we discuss the quality of the each figure. Unweighted OLS estimates provided simi- students "produced" at each level of education, as given by lar benchmarks. In this chapter, Latin America is there- the performance of Latin American students and adults on fore benchmarked against the "weighted average" international standardized tests country in the world, although we will often stress the comparisons between Latin American councries and Workers' Educational Attainment in Many Latin those in East Asia, especially the "tiger" economies American Countries Is Low (Hong Kong [China), Korea, Malaysia, and Singapore), Over the course of the last two decades, the mean years of and a group of countries that are abundant in natural education of the adult population aged 25 and older has gone resources (Australia, Canada, Finland, New Zealand, up by 1.7 years in Latin America (from 4.1 to 5.8 years). Norway, and Sweden). But educational attainment in many Latin American coun- ___ _ __ _ _ _ _ ' tries is still very low. Figure 2.1 presents a scatter-plot of the mean years of schooling of the labor force aged 25 and are embodied in capital). All increases in skill, increases in older relative to per capita GDP, as well as the correspond- technology, as well as any interaction effects between skill ing regression line. (Note that here, as throughout this chap- and technology on growth will therefore be captured by this ter, the regression includes all of the countries in the world measure of TFP. for which data are available-in this case 105 countries. We do not show all the country labels on the graph to make it Total Factor Productivity Growth in Latin America more easily legible) Figure 2.1 shows that many Latin Has Been Low in the Last Two Decades American countries-especially Venezuela, Colombia, Costa Table 2.1 shows that the last two decades have nor been good Rica, Brazil, El Salvador, Honduras, Haiti, and Guatemala- for Latin America Focusing on the GDP-weighted averages, have large education deficits. The comparison between Latin Latin America had the highest rates of productivity growth America and the more successful East Asian countries is of any developing region in the pre-1980 period and, in a revealing- Malaysia, Costa Rica, and Brazil all have approx- fashion similar to Eastern Europe, the Middle East and North imately the same level of GDP per capita: despite this sim- Africa, and Sub-Saharan Africa, very low or negative growth ilarity, the mean years of education of Malaysian adults (7.9) 24 THE GAPS IHAI MAITER MOST TABLE 2 1 TFP Growth Rates in Latin America Were Negative in the 1980s and Low in the 1990s A UNWEIGHTED AVERAGE ANNUAL GROWTH RATES OF TOTAL FACTOR PRODUCTIVITY 1960-69 1970-79 1980-89 1990-99 East Asia and Pacific 0 93 1 26 1 25 1 30 Eastern Europe -0 17 -0 17 0 43 0 07 Larin America and the Caribbean 1 16 0 47 -I 60 0 47 Middle East and Northern Africa 1 12 0 84 -1 24 0 58 OECD 1 92 1 10 0 88 0 94 South Asia 0 71 -0 08 1 49 1 38 Sub-Saharan Africa 0 40 0 37 -0 79 -0 14 East Asian tigers 1 32 1 09 1 32 1 83 Natural resource-abundant countries 1 75 1 12 0 83 0 90 B GDP-WEIGHTED AVERAGE ANNUAL GROWTH RATES OF TOTAL FACTOR PRODUCTIVITY 1960-69 1970-79 19S0-89 1990-99 East Asia and Pacific 1 14 1 28 2 27 2 01 Eastern Europe -0 17 -0 27 0 52 0 3 Latin America and the Caribbean 1 50 1 15 -0 93 0 45 Middle East and Northern Africa 1 14 -1 62 -I 12 0 85 OECD 2 09 0 85 1 10 0 56 South Asia 0 46 -0 39 2 27 1 72 Sub-Saharan Africa 1 54 -0 55 -0 88 -0 43 East Asian tigers 1 79 1 03 2 18 1 42 Natural resource-abundant countries 1 68 0 93 0 76 0 78 Source Authors' calculations using data from Loayza, Falnzylber, and Calderon (2002) is almost two years higher than that of Costa Rican adults In Latin America, the Distribution of Education (6.0), and more than three years higher than that of Brazil- among Adults Appears to Be Less Equitable than ian adults (4.6). All of the comparison natural resource- That in the East Asian Tigers and the Natural abundant countries outside Latin America are above the Resource-Abundant Countries regression line, having education levels above those that The mean years of educational attainment of adults is only would be predicted for their income levels There are also a summary statistic Two countries with the same mean could Latin American high performers: Peru, Thailand, and have very different underlying distributions: in country A, Colombia have very similar income levels, but Peruvian a "medium" level of attainment could be the result of large adults have more than one more year of schooling than their numbers of adults with primary and tertiary education, and Thai counterparts (7.3, compared to 6.1 years), and more only a small fraction with secondary education, whereas in than 2 years more schooling than Colombians (5 1). On aver- country B the same level of attainment could be the result age, however, Latin American adults have 1 4 years fewer of a very large number of adults with secondary education years of education, while adults in the East Asian tiger coun- Table 2.3 therefore presents estimates of the fraction of adults tries have 1.0 more years of education, and those in the nat- with no schooling, some primary schooling, some secondary ural resource-abundant countries have 1.4 mare years of schooling, and some university for countries in Latin schooling than would be expected for their income levels.2 America, the East Asian tigers, and the group of natural As we will discuss in chapter 4, these differences are a result resource-abundant countries in 2000. The table clearly of the patterns of much faster skill upgrading in East Asia shows that the distribution of education among adults is very and in some Scandinavian countries than in Latin America different in the three sets of countries: in 2000, Latin America since 1960. had much higher fractions of adults with no schooling, or 25 CLOSING THE GAP TABLE 2 2 In the 1980s and 1990s, Growth Rates of TIFP Were High in Chile and Low in Most Other Latin American Countries ANNUAL GROWTH RATES OF TFP 1960-69 1970-79 1980-89 1990-99 Latin America 1.50 1.15 -0.93 0 45 Argentina 1 37 0 21 -1 75 2 81 Bolivia -0 25 0.23 -1 63 1 88 Brazil 1 69 2 15 -0 21 -0 42 Chile 1 31 0 31 2 29 2 22 Colombia 1 66 1 76 -0 06 -0 39 Costa Rica 0.73 0 29 -1 13 1 80 Dominican Republic 0 74 0 92 -0 39 1 66 Ecuador 0 38 3 59 -1 03 -0 87 El Salvador 1 05 -0 82 -3 92 1 05 Guatemala 1 55 1 56 -1 39 0 70 Guyana 0 89 -0 65 -3 85 -3 91 Haiti -097 -0 29 -2 54 -1 99 Honduras 0 89 1 46 -0 91 -1 24 Jamaica 2 06 -1 46 0 22 -2 10 Mexico 1 66 1 13 -1 60 0 31 Nicaragua 1 66 -3.53 -3 53 0 22 Panama 2 32 -0 71 -2 18 1 12 Paraguay 0 78 1 86 -1 27 -1 46 Peru 1 35 0 34 -2 45 0 83 Trinidad and Tobago 2 86 1 75 -2 26 -0 25 Uruguay 0 66 1 71 -0 06 2 02 Venezuela, R B de 1 50 -1 30 -2 79 -0 00 East Asian tigers 1.79 1.03 2 18 1.42 Hong Kong (China) Korea 2 03 0 96 2 53 1 30 Malaysia 0 79 1 46 0 42 0 98 Singapore 1 82 0 89 2 01 2 62 Natural resource-abundant countries 1.68 0.93 0 77 0 78 Australia 1 67 0 30 0 96 1 31 Canada 1 45 098 0 32 0 15 Finland 1 70 1 44 2 08 1 15 New Zealand -0 34 0 70 Norway 2 24 2 10 0 96 2 09 Sweden 1 65 0 79 0 90 0 58 Note Calculations based on annual data, so that figures for each country are GDP-weighted decade averages Sourre Authors' calculations using data from Loayza, Fajnzylber, and Calderon (2002) only some primary schooling compared to the tigers or nat- Venezuelan adult is only one-half that of a Scandinavian ural resource-abundant countries. There are also differences adult. To a large extent, these dsfferences are a result of the in the distribution of educational attainment for countries highly inequitable distribution of educational attainment in with the same mean years of schooling. Malaysia and Panama some Latin American countries. both have 7.9 mean years of schooling, but in Malaysia 43.0 percent of the population has some secondary schooling only, Some Latin American Countries Appear to Be and only 7 5 percent has some tertiary education, whereas Undergoing "Unbalanced" Educational Transitions in Panama 28.5 percent of the population has some secondary In chapter 4 we show that the bulk of skill upgrading schooling only, and 19 8 percent has some university In between 1960 and 2000 in Latin America took place at the Venezuela, the fraction of adults with some university tertiary level, while the bulk of skill upgrading in East Asia education is not much lower than that in the Scandinavian took place at the secondary level. (The picture for the nat- countries, even though the mean years of schooling of a ural resource-abundant countries is more mixed, as many of 26 THE GAPS I HAT MATTER MOST FIGURE 2 1 Most Latin Amenican Countnies Have Low Levels of Educational Attainment of Their Adult Population United States 12 - New Zealand Canada Sweden + Korea Australia g Kong (China) o Malaysia Agn awn(ha > ~~~~~~~~~~~~~~ ~ ~~Pa-Iama Chile /S.ng.po,e rv ~~~~~~~~~~Ph,hlpp,nes Tp vyh .ad To bago 0 Pr o G ~~~~~~~~~~~Ecuado~r -x,co C Guana Thaland 6 China Cast. Rica o B3livi a tguay Veneuel Ja,maica IdnsaDominican Republic 0 In, onesia ~~~~~~~Colombiam Brail Hondcas El Salvador Guatemala 3 ~~~~~Haiti 5 6 7 8 9 10 11 Log of Per Capita GDP, 2000 Note The line corresponds to the predicted years of schooling from a weighted regression on log per capita GDP and a constant, with clie weights given by the population aged 25 and older in a country The sample size for rhe regression is 105 countries Some country names have been omitted to make the figure more legible Source Authors' calculations using data from Barro aisd Lee (2002) them were already further in their education transltion in Most Latin American Countries Do Not Have a Gap 1960.) As a result, some Latln American countries-espe- in Net Enrollment in Primary School cially Costa Rica, the Dominican Republic, El Salvador, and Primary school is the necessary building block for all edu- Venezuela-seem to be following "unbalanced" educational cational upgrading. Benchmarking Latin America in transitions. although most adults in these countries still have terms of net primary enrollment rates suggests that the primary education or less, so much upgrading took place at region as a whole does not "underperform" on this dimen- the tertiary level, and so little at the secondary level, that by sion 4 (The net enrollment level is given by the fraction of 2000 there were fewer adults with only secondary school edu- children of primary school age who are enrolled in primary cation than adults with tertiary education.3 As we will dis- school ) Still, there are countries such as some of the poorer cuss in detail in chapter 4, unbalanced educational transitions countries in Central America, and regions within other are likely to have negative consequences in terms of the long- countries, such as parts of the south in Mexico and the term sustainability of educational transition, inequality, and Northeast in Brazil, where primary enrollment rates are the ability to upgrade technologies. still low, and should continue to be a focus for government The educational attainment of the population in a coun- policy. try is the result of the accumulated flow of investments in schooling in earlier decades Present enrollments determine Most Latin American Countries Have Massive Deficits how stocks will evolve in the future. New cohorts will be in Net Enrollment in Secondary School... most of the labor force in the next decades. We now turn to Although the Latin American picture regarding net pri- a discussion of contemporaneous investments in education mary enrollment rates is encouraging, the same cannot be by looking at the enrollment rates in primary school, sec- said about net secondary enrollment rates Overall, the region ondary school, and tertiary institutions. has a massive deficit in secondary enrollment. Figure 2.2 27 CLOSING THE GAP TABLE 2 3 Compared to East Asia and the Natural Resource-Abundant Countries, Latin America Has Low Levels and an Unequal Distribution of Educational Attainment among the Adult Population FRACTION OF THE ADULT POPULATION YEARS OF SOME SOME SOME SCHOOLING NO SCHOOL PRIMARY SECONDARY TERTIARY Latin America 5.8 17 9 50.1 20.3 11.8 Argentina 8 5 5 8 49 6 24 9 19 7 Bolivia 5 5 1 4 39 8 46 9 11 9 Brazil 4 6 213 568 13 5 84 Chile 7 9 5 3 42 9 36 0 15 8 Colombia 5 0 19 8 48 9 214 9 9 CostaRica 60 94 607 11 3 186 Cuba 7 8 5 8 40 3 42 6 113 Dominican Republic 5 2 25 6 46 8 13 1 14 5 Ecuador 65 178 452 183 187 El Salvador 4 5 35 0 45 6 8 8 106 Guatemala 3 1 47 1 37 6 9 5 5 8 Haiti 2 7 544 32 3 12 3 1 0 Honduras 4 1 25 9 57 0 10 6 6 5 Jamaica 5 2 3 4 54 5 38 0 4 1 Mexico 6 7 12 4 47 3 29 0 11 3 Nicaragua 4 4 31 6 43 0 16 5 8 9 Panama 7 9 11 3 40 4 28 5 19 8 Paraguay 5 7 9 8 63.8 18 1 8 3 Peru 7 3 13 8 35 7 28 1 22 4 Trinidad and Tobago 7 6 5 1 46 3 44 1 4 5 Uruguay 73 32 522 321 125 Venezuela 5 6 15 7 56 6 9 7 18 0 East Asian tigers 9.7 10.2 23 5 47.4 18.9 Hong Kong (China) 9 5 10 7 26 6 47 4 15 3 Korea 10 5 8 0 16 7 49.5 25 8 Malaysia 7 9 13 9 35 6 43 0 7 5 Singapore 8 1 12 6 28 3 48 5 10 6 Natural resource-abundant countries 11 1 1.7 21.1 38 6 38.7 Australia 10 6 2 2 24 4 43 6 29 8 Canada 11 4 1 7 18.6 26 6 53 0 Finland 10 1 0 4 29 2 47 3 23 2 New Zealand 11 5 0 0 32 2 26 3 41 6 Norway 11 9 1 2 11 5 62 5 24 8 Sweden 11.4 2 0 17.7 57 2 23 1 Note Regional averages are weighted by the population aged 25 or older in a country Source Authors' calculations based on data from Barro and Lee (2002) presents a scatter-plot of the net enrollment rate in sec- 18 percentage points, and the natural resource-abundant ondary school. The figure clearly shows that many Latin countries have a surplus of 6 percentage points. Within American countries are well below the regression line. In Latin America, the biggest deficits are those of Brazil (36 table 2.4, we tabulate the secondary enrollment deficit, points), Venezuela (42 points), and Costa Rica (24 points). defined as the difference between the observed secondary By comparison, many of the countries in the English- enrollment rate and its predicted value, for the sample of speaking Caribbean, including Barbados, Guyana, Jamaica, countries The table shows that Latin America has a deficit and Trinidad and Tobago, have net secondary enrollment of almost 19 percentage points in secondary enrollment. rates that are substantially higher than expected for their By contrast, the East Asian tigers have a surplus of almost income levels. 28 THE GAPS THAT MATTER MOST FIGURE 2 2 Many Countries in Latin America Have Massive Deficits in Secondary Enrollment 100 Sweden Korea Finland Malaysia Canada New Zealand Unit d States cc Aus c 80 Jamaica X ~~~~~~~~~~~~~~~~~~~Arge La Ch Trimm Hong Kong (China) 5 Guyana - 60 Philippines Mexico W I hatand Chi iminican Republic Ecuador Colombia Indonesia Paraguay Costa Rica E 40 El Salvador Behze z Nicaragua Brazil Venezuela 20 I I I I I 5 6 7 8 9 to 11 Log of Per Capita GDP, 1998 Note. Secondary enrollment data for Korea and Nicaragua are for 1997, for Venezuela are for 1996, for Colombia, Guyana, Hong Kong (China) and the Philippines are for 1995, and for Indonesia are for 1994 All other data are for 1998 The net secondary enrollment for Brazil has been calculated from the PNAD household surveys Net enrollment rates for all other countries are from UNESCO Countries with net secondary enrollment ratios above 100 were not included in rhe regression The line corresponds to the predicted net secondary enrollment rate from a weighted regression on log per capita GDP and a constant, with the weights given by the population in a country The sample size for the regression is 116 countries Some country names have been omitted to make the figure more legible Source Authors' calculations, based on data from UNESCO ... Reflecting, in Part, Problems in the Primary secondary school in Latin America are a reflectson of inade- School System quate progression earlier on. As we discuss in chapter 4, Latin Because net secondary enrollment rates are a measure of the American countries have problems with both progression fraction of children of secondary-school age who are in sec- wsthmn primary school and continuation of primary school ondary school, countries with high rates of repetition or late graduates on to secondary school. entry intoprzmary school will generally tend to have low net enrollment rates in secondary school. A comparison of the net Large Deficits in Tertiary Enrollment enrollment rates with the gross secondary enrollment rate, In addition to the deficit in secondary enrollment, many given by the fraction of children of secondary-school age who Latin American countries also have relatively low enrollment are enrolled in school, regardless of the school level they rates at the tertiary level. Figure 2.3 presents a scatter-plot attend, can shed some light on this. In a population- of the gross enrollment rate at tertiary institutions for coun- weighted regression of the gross enrollment rate on the log tries in Latin Amersca, East Asia, and the group of natural of per capita GDP and a dummy variable for Latin America, resource-abundant countries. As a whole, the Latin American the coefficsent on this dummy is negative but insignificant. deficit at the tertiary level is about 10 percentage points. (The coefficient is -4.8, with a standard error of 4.2). This Table 2 4 shows that Belize, Brazil, and Trsnsdad and Tobago suggests that some of the observed problems of coverage in all have particularly large deficits in the region, while 29 CLOSING THE GAP TABLE 2 4 Latin American Students and Adults Perform Poorly Many Countries in Latin America Have MWassive Secondary on International Tests Enrollment Deficits, and Some Have Smaller Teriary How well youth in Latin American countries are prepared Enrollment Deficits to participate productively in the economy depends on both SECONDARY TERTIARY the number of years they stay in school and how much they NET ENROLLMENT GROSS ENROLLMENT learn in each year of school Most Latin American countries DEFICIT DEFICIT do not participate regularly in international tests, but there Latin America -18.7 -10.0 Argentina -I 2 6 0 are some exceptions. eighth-grade students in Colombia par- Belize -22 8 -27 4 ticipated in the 1995 Third International Math and Science Borvia -36 3 -20 9 Study (TIMSS) test, and eighth-grade students in Chile par- Chile 1 2 -I 7 ticipated in the 1998 TIMSS. Fifteen-year-olds in Brazil and Colombia -14 I -6 5 Mexico participated in the OECD's PISA international tests. Costa Rica -23 9 0 0 Cuba Chilean adults also participated in OECD tests of adult lit- Dominican Republic -36 0 1 eracy Finally, a large number of countries in Latin America Ecuador -8 4 -3 4 participated in the OREAL (OREAL 2001) tests of third- and Guatemala -12 5 fourth-graders in primary school. However, because the Guyana 17 6 -3 6 OREAL tests were not administered to countries outside the Haiti -2.8 Honduras I 0 region, they cannot be used to compare the performance of Jamaica 22 8 -13 9 Latin American students with that of their counterparts in Mexico -84 -126 East Asia or elsewhere. Nicaragua -6 3 5 8 Panarna 3 5 The sum of the evidence on the performance of Latin Paraguay -13 9 -12 5 American students on international tests suggests that, in Trinidad and Tobago 4 7 -28 1 addition to the gap in terms of the quantity of education, both Uruguay -6 0 -3.1 measured as the stock of educated adults and the flow of stu- Venezuela -42 2 -2 5 dents from various schooling levels, Latin America has a seri- East Asian tigers 17 8 5.1 ous quality deficit. Figure 2.4 shows that both Colombia and Hong Kong (China) -17 4 -24 9 Chile were among the worst performers on the TIMSS, per- Korea 19 3 24 1 Malaysia 26 2 -21 9 forming substantially worse than countries of similar Singapore -9 9 income levels, and at about the same level as much poorer countries. the mean math score in Chile was almost 140 Natural resource- abundant countries 6.0 14.0 points below the score of Hungary, and below the score of Australia 1 9 27 1 Indonesia, a substantially poorer country The mean math Canada 7 8 6 6 Finland 4 9 27 7 score in Colombia was also about 140 points below that of New Zealand 7 3 13 6 Thailand, and well below that of Iran, a country with lower SNwydn -0 71 21 0 per capita income. Brazil and Mexico were the two worst per- Note Regional averages are population-weighted formers in the PISA test of 1 5-year-olds, well below similar- Source Authors calculations based on data from UNdESCO income countries such as Poland and Latvia, and well below their predicted performance from a regression of scores on spending on education per student.5 Figure 2.5, finally, Argentina and Bolivia have high tertiary enrollment rates shows that Chile, the only participating country in Latin for their income level. Also noteworthy is the fact that the America in tests of adult literacy, had the lowest quantita- countries in the English-speaking Caribbean, which gener- tive scores, well below those of Eastern European countries ally did well in terms of enrollments in secondary school, of similar income levels.6 For any given level of education, often have large tertiary enrollment gaps By comparison, Chilean adults appear to have learned less than their coun- both the East Asian tigers and the natural resource-abundant terparts in Poland, Slovenia, Portugal, Ireland, the Czech countries have surpluses in tertiary enrollment. Republic, and a number of OECD countries. 30 THE GAPS THAT MATTER MOST FIGURE 2 3 Some Countries Also Have Large Deficits in Tertiary Enrollment, although These Are Generally Smaller Finland 80 - Australia United States Korea 0' New Zealand Sweden 60- Canada c ~~~~~~~~~~~~~~~Argetn / ° 40 / Singapore ,a 40 w ~~~~~~~~~~~Panama C Y Thallandv BarbadosT Philippines Per nezuela Hong Kong (China) Bolivia Dom n Republic Bahamas 20~~~~~~~D 2 20 - /Clmbia 0 Honduras I Salvador Mexico Nicaragua I BParaguay Mal uana Jamaica Malyi Guatemala Trinidad and Tobago 0- - ts Ch Surlname Belize 5 6 7 8 9 10 11 Log of Per Capita GDP, 1998 Note Tertiary enrollment data for Barbados, Belize, Bolivia, Colombia, Costa Rica, the Dominican Republic, Ecuador, Guatemala, Guyana, Haiti, Hong Kong (China), Indonesia, Korea, Malaysia, Nicaragua, Panama, Paraguay, Singapore, and Venezuela are for 1997 The line corresponds eo the predicted gross tertiary enrollment rate from a weighted regression on log per capita GDP and a consranc, with the weights given by the population in a country The sample size for the regression is 161 countries Some country names have been omitted to make the figure more legible Source Authors' calculations, based on data from UNESCO Most Latin American Countries Do Not Have a Low esther.8 The figure shows that there is no clear relationship Stock or Flow of Scientists and Engineers for Their between overall enrollment levels and enrollment in sciences Income Level and engineering. Latin American countries are scattered widely: As we discuss in chapters 6 and 7, the number of scientssts Chile and Costa Rsca both have high tertiary enrollment and engineers is an important input into the production of rates, but the fraction of students in science and engineer- domestic R&D and is therefore of interest in its own right ing is much higher in Chile. A similar comparison can be In figure 2.6, we present a scatter-plot of the number of sci- made for El Salvador and Colombia, both of which have entists and engineers per 10,000 workers as a function of per middle-level enrollment rates, with El Salvador having a much capita GDP. The figure suggests that some Latin American higher fraction of students in science and engineering. countries, particularly El Salvador, Colombia, and Venezuela, In sum, Latin America has a profound skill gap in rela- have relatively few scientists and engineers for their income tion to many of its competitors in East Asia, and in com- levels Others, including Bolivia, Peru, and Costa Rica, have parison with the many other countries with abundant natural a relative abundance of scientists and engineers resources. In most countries, the mean years of education of Latin American countries generally do not appear to have the adult population is low. The enrollment gap in the region too small a stock of scientists and engineers for their income is particularly severe at the secondary school level, although levels.7 Figure 2.7 shows that enrollment levels in science some countries also appear to have a gap at the tertiary level and engineering as a function of gross tertiary enrollment On the positive side, there does not appear to be a problem levels are not low for the 'average" Latin American country with the composition of university graduates as a whole, the 31 CLOSING THE GAP FIGURE 2 4 Colombian and Chilean Secondary School Students Perfonn Badly on International Tests 650 - Singapore Korea Hong Kong (China) i- 5 o0 mAustraia Log o PerCapit GDP 199 -hailand Sen R ~~~~~~~~~~~~~~~~New Zealand Nra o ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~~~~~~Nra u E 450- Colombia 8 9 10 11 650- Singapore ci2 Korea Hong Kong (China) X 550- Cs ~~~~~~~~~~~~~~~~~~~Canada su ~~~~~~~ ~~Malaysia l ausdO O / e w Zealand U ~~~~~~~Thailand 450 - E X S Indonesia Chile 350 Philippines 7 8 9 10 11 Log of Per Capita GDP, 1995 Note: The line corresponds to the predicted TIMSS score from a weighted regression on log per capita GDP and a constant, with the weights given by population The sample size is 39 for the 1995 TIMSS, and 37 for the 1998 TIMSS Some country names have been omitted to make the figure more legible Source Authors' calculations, based on TIMSS data region does not have a deficit in the stock or flow of scien- The TecGOoGoDog Gap tists and engineers. Finally, the gap in terms of the quan- We next turn to a discussion of the extent to which coun- tity of education is further aggravated by the poor quality tries in Latin America appear to have a technology gap. We of education, as made apparent by the low performance of first present evidence of the degree to which Latin American Latin American students on international tests. countries are exposed to foreign imports, especially imports 32 THE GAPS THAT MATTER MOST FIGURE 2 5 Chilean Adults Also Do Badly on Intemational Tests of Literacy Adult Literacy: Quantitative Scores (OECD), 1994-98 Czech Rep - - - Sweden __-- Germany Netherlands =-_ =_=_= Belgium E Australia __=_= = United Kingdom Canada =_ _ ___Age 15-25| Ireland - - = | = = =2 Hungary - - - Portugal United States _ Poland Slovenia Chile 0 50 100 150 200 250 300 350 Note Scores are an average of scores for individuals with less than upper secondary, complete upper secondary, and complete tertiary These three scores were equally weighted when constructing the average score in order to abstract from differences in labor force composition across countries Source Author's calculations, based on data from OECD (2000) of capital goods, make payments to license foreign tech- new technology for developing countries Table 2.5 shows nologses, and receive foreign direct investment (FDI). We that countries in Latin America generally have low import also pay special attention to imports of computer equipment, penetration: the GDP-weighted mean import penetration as these are likely to be particularly intensive in high tech- is more than four times as high in the East Asian tiger nology. This first part of the analysis therefore summarizes countries as in Latin America, and almost twice as high for the degree to which countries in Latin America are likely the group of natural resource-abundant countries. Import to have access to foreign technologies from the "leader" coun- penetration is particularly low in two of the largest Latin tries. In the second part of this section we focus on domes- American economies-Argentina and Brazil. Larger economies tic spending on R&D by the public and private sectors, generally have lower rates of import penetration, and it is which is necessary for the absorption and adaptation of for- conceivable that Latin America's lower import penetration eign technologies, as well as for the creation of new tech- is driven by the relatively large size of some economies. In nologies. Finally, we consider the numbet of patents practice, however, this does not explain the deficit in import registered by Latin American inventors in their own coun- penetration in Latin American countries 9 Also, the fraction tries and the United States. of imports that comes in the form of capital goods is much lower in Latin America than in the East Asian tigers or the Latin American Countries Receive Fewer Imports than natural resource-abundant countries. Much new technology East Asian Countries, and Fewer Imports of Capital Goods is "embodied" in imports of capital goods, so this shortfall Imports of goods from countries that are more advanced tech- may be particularly damaging to Latin America in terms of nologically have the potential to be an important source of technology transfer. 33 CLOSING THE GAP FIGURE 2 6 Some Latin American Countries Appear to Have Low Nlumbers of Scientists and Engineers for Their Income Levels, but thers Do Not Singapore Norway Sweden 4 - Korea Cwdana Naln Si Peru Costa Rica e Q c 2 Nicaragua Bolivia v) o NicaraguaEcus Venezuela u; / ~~~~~~Guatemala Thailand zsi Hong Kong (Chima) W C Colombia o X El Salvador C: is -2 6 8 10 12 Log of Per Capita GDP, 1998 Note The line corresponds to the predicted number of scientists and engineers from a weighted regression on per capita GDP and a constant, with the weights given by the population aged 25 and older in a country The sample size for the regression is 110 countries Some country names have been omitted to make the figure more legible Source Authors' calculations, based on data from Park (2002) for the number of scientists and engineers, and World Bank databases for per capita GDP levels There is a Great Deal of Variation in FDI Penetration across countries within Latin America This is potentially Within Latin America an important consideration, as different forms of FDI may Table 2.5 shows that the levels of FDI in a number of Latin contribute more or less to technological upgrading in the American countries are comparable to those in the East host economy. Asian tiger or natural resource-abundant economies else- where. However, there is a great deal of variation. In Latin Latin American Countries Also Spend Relatively Little America, Bolivia, Chile, and Nicaragua all attract a great on the Licensing of Foreign Technologies deal of FDI relative to the size of their economies, while FDI Licensing involves the contractual transfer of knowledge penetration is low in Colombia, Peru, and Uruguay. Within between firms. Given that licensing provides knowledge in East Asia, FDI penetration is very high in Singapore and a more accessible manner than FDI, many countries such Malaysia, but negligible in Korea, which has actively pur- as Brazil, India, Mexico, and Japan have in the past sued high-technology imports and technology licensing but actively discriminated against direct investment by firms and has not similarly encouraged FDI. By contrast, Singapore favored technology licensing. In practice, as table 2 5 shows, has relied a great deal on FDI as the source of its techno- there is considerable variation in the amounts that logical miracle. Foreign direct investment is also high in countries spend on technology licensing. Worryingly, table Finland and Sweden. One thing that is not clear from the 2.5 shows that by this measure of technology transfer Latin FDI penetration statistics in table 2.5 is whether the com- America is also far behind many of its potential economic position of FDI varies across the three sets of countries, or competitors. 34 THE GAPS THAT MATTER MOST FIGURE 2 7 Most Countries in the Region Do Not Appear to Have an Important Deficit in the Fraction of University Students Enrolled in Science and Engineering Jamaica Barbados 60 - El Salvador Trinidad and Tobago J C\ China Chile - 40 Indonesia Finland Hong Kong (China) Sweden f a C' . G ,, Dominican Republic § Q ~~Nicaragua Mexico Korea " - Colombia Panama Argentina Esi Brazil Norway Honduras Australia ' 20 Paraguay Costa Rica New Zealand Cuba Canada Qi Philippines 0 0 50 100 Gross Tertiary Enrollment Rate, 1997 Note Tertiary enrollment data for Barbados, Belize, Bolivia, Colombia, Costa Rica, Dominican Republic, Ecuador, Guaremala, Guyana, Haiti, Hong Kong (China), Indonesia, Korea, Malaysia, Nicaragua, Panama, Paraguay, Singapore, and Venezuela are for 1997, science and engineering data for Barbados, Indonesia, Jamaica, Nicaragua, Philippines and Trinidad and Tobago are for 1996, for Argentina, Brazil, China, Honduras, and Panama are for 1995, and for Hong Kong (China) is for 1994 The line corresponds to the predicted percentage of students in science and engineering from a weighted regression on the gross tertiary enrollment rate and a constant, with the weights given by the population In a country The sample size for the regression is 117 countries Some country names have been omitted to make the figure more legible Source Authors' calculations, based on data from UNESCO In the Late 1980s, East Asia Opened Up an ... Which Did Not Shrink in the 1990s Important Lead in Computer Penetration The data used by Caselli and Coleman end in 1990. To inves- over Latin America ... tigate whether the digstal dsvide between East Asia and Latsn Imports of computer equipment are a clear sign of transfer Amersca shrank sn the 1990s we turn to data on the number of advanced technologies. In recent work, Caselli and of personal computers per 1,000 population for the 1988-99 Coleman (2001) explore cross-country differences in the period. These data, whsch are only available for a handful of adoption of computer technology Fsgure 2 8 uses these data countries in Latin America, confirm that some countries in to calculate population-weighted means of the country- East Asia-particularly Korea, Hong Kong (China), and specific values of computer imports (in 1995 U.S. dollars) Singapore-already had a serious lead on Latin American for a large sample of Latin American and East Asian coun- countries in the sample in 1988: the number of computers tries. The main message of figure 2.8 is abundantly clear: per 1,000 population was roughly 10 times as high in until 1980, Latin America had an edge over East Asia in Singapore as in Mexico, Argentina, or Chile, and 15 times computer penetration By 1990, however, this situation had as high in Hong Kong as in Brazil. By 1999, this number been reversed, so that the population-weighted mean com- was 10 to 20 times as high as at the beginning of the period purer penetration in East Asia was about three times that in virtually every country in the sample. However, there found in Latin America. is no consistent evidence of digital convergence between 35 CLOSING THE GAP TABLE 2 5 Countries in Latin America Have Low Import Penetraffon, Low Penetration of Capital Goods, and Low Levels of Royalty PaymenXt PERCENTAGE OF GDP IMPORTS CAPITAL GOODS ROYALTY GOODS AND SERVICES IMPORTS NET FDI PAYMENTS Latin America 19.10 7.69 3 33 0.15 Argentina 1142 416 2 72 016 Belize 63.78 14 45 1 86 Bolivia 25 09 988 10 81 0 06 Brazil 1213 4 18 4 39 017 Chile 30 79 745 246 0 07 Colombia 20 42 4 31 1 62 0 07 Costa Rica 46 08 11 74 3 56 Dominican Republic 39 33 5 65 0 19 Ecuador 30 80 4 58 4 28 0 40 El Salvador 42 71 6 90 3.75 0 19 Guatemala 27 88 8 74 1 84 Honduras 56 37 15 35 3 44 0 22 Jamaica 55 05 8 84 4 21 0 94 Mexico 33 23 14 75 2 52 0 16 Nicaragua 81 17 18 79 7 53 Panama 38 93 13 14 7 43 0 33 Paraguay 35 37 8 86 2 00 0 04 Peru 17 86 4 59 2 72 010 Uruguay 20 71 4 97 0 93 0 05 Venezuela 17 03 6 64 3 39 0 00 East Asian tigers 83.77 29.51 2.95 0.47 Hong Kong (China) 145 26 41.60 -2040 Korea 42 20 10 75 0 74 0 47 Malaysia 104 36 49 95 4 94 0 47 Singapore 161 45 78 29 6 93 Natural resource-abundant countries 34.09 12.59 5.07 0.37 Australia 22 31 7 44 1 99 0 29 Canada 40 78 17 95 5 62 0 43 Finland 29 27 10 50 6 81 0 31 New Zealand 32 75 1114 3 75 064 Norway 32 83 9 34 3 91 0 18 Sweden 37 83 10 67 8 91 0 50 Note Data on imports and capital goods imports are from 1999, data on royalties are for 2000, data on FDI are the mean for the 1998-2000 period The East Asian tiger average for FDI does not include Hong Kong (China) because of the high volume of FDI investments flowing from Hong Kong to mainland China in the late 1990s Regional averages are weighted by GDP in the corresponding year Source World Bank and UNESCO databases the East Asian and Latin American countries in the sample. The gap in information and communications technology Chile narrowed the gap with Singapore somewhat, but (ICT) between Latin America and East Asia is also apparent Malaysia did so even more rapidly. In 1999 there were about in other measures. Figure 2.10 presents a "summary" measure six times as many computers per person in Hong Kong as of ICT, which is a composite ofthe number ofcomputers, Inter- in Mexico and Argentina-just as in 1988. And the growth net hosts, main telephone lines, and cell phones, all normal- in computer penetration was particularly impressive in China ized by the population of a country.'0 The figure shows that (albeit starting from a very low base). The two panels in the high performers in East Asia-Singapore, Malaysia, figure 2.9 clearly show that the digital divide between East Korea-are well ahead of countries in Latin America. Within Asia and Latin America did not narrow significantly in the Latin America, countries can be grouped into those with rea- 1990s. sonably "high" levels of ICT infrastructure (Chile, Argentina, 36 THE GAPS THAT MATTER MOST FIGURE 2.8 In the 1980s East Asia Opened Up a Big Lead over Latin America in the Degree of Computer Penetration 30 25 u,20 t __ _ __ _ _ _o___ _ _ __Latin__ _ __ _ _ __ _ __ _ _ America a15 East Asia d 10_ E 1970 1975 1980 1985 1990 Year Note: Regional averages are population-weighted, and are in 1995 US$. The Latin American sample includes Argentina, Barbados, Bolivia, Brazil, Chile, Colombia, Costa Rica, the Dominican Republic, Ecuador, El Salvador, Guatemala, Guyana, Haiti, Honduras, Jamaica, Mexico, Nicaragua, Panama, Paraguay, Peru, Trinidad and Tobago, Uruguay, and Venezuela; the East Asian sample includes Fiji, Hong Kong (China), Indonesia, Malaysia, Papua New Guinea, Philippines, South Korea, Singapore, and Thailand. Source: Authors' calculations, based on data from Caselli and Coleman (2001). Costa Rica, Colombia, Venezuela, Brazil, and Mexico), those technology gap. Table 2.6 shows that Latin American with "medium" levels (Peru, Trinidad and Tobago, Uruguay, workers have less R&D capital available than those in and Ecuador), and those with relatively "low" levels (mostly Korea and Singapore, as well as those in natural countries in Central America, as well as Bolivia and Paraguay). resource-abundant countries. Within Latin America, Domestic technology is a complement to access to for- Brazil, Chile, and Peru all have relatively high ratios of eign technology-especially (but not only) for countries that R&D to GDP (although these are still considerably below are relatively close to the global technology frontier. But as those in Korea and Singapore, as well as the Scandinavian we discuss in chapters 6 and 7, even countries that rely countries). mainly on technology developed abroad need some domes- The amount of R&D capital per worker rises with per tic R&D to adapt imported technologies. We next turn to capita income, and it is therefore important to benchmark a discussion of two measures of domestic technology: R&D R&D expenditures. Figure 2.11 graphs the log of R&D expenditures and patent applications. expenditures per worker as a function of the log of per capita GDP for a sample of Latin American, East Asian, and nat- Workers in Latin America Have Access to Less Domestic ural resource-abundant countries. The figure shows a wor- R&D Capital Stock than Those in Korea or Singapore, rying picture for Latin America: all of the countries in the or in the Natural Resource-Abundant Countries region except Brazil aie below the regression line. By contrast, R&D expenditures are a measure of the domestic "effort" all of the East Asian tigers and the natural resource-abundant the public and private sectors are making to close the countries are on or above the line. 37 CLOSING THE GAP FIGURE 2 9 The Digabl Divide bteen East Ads and LaEn America Did Mot Shrink In the I191D Latin America 450 400 c z 350 | 300 g 250 entina ,9 200 - ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ ---BaIl U 150 100 50 0 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Year East Asia 450 400 350 300 z 250 - - Hong Kong (China) 01) 250 - ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~~-Indonesia - -- Korea 200 - - ~~~~~~~~~~~~~~~~~~~~~~~~Malaysia - - Singapore E ---- Thailand 150 100 z 50 - A - -~ ~ ~ - --- --- 0 4 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 Year Sourc Authors' calculations based on data from the International Telecommunications Union 38 THE GAPS THAT MATTER MOST FIGURE 2 10 Latin America Also Lags behind East Asia in Other Measures of ICT Infrastructure Chile Argentina - _ Costa Rica - 1 Colombia - _ _ Venezuela Brazil _ U Mexico _ Peru Trinidad and Tobago Uruguay Ecuador Bolivia Panama Paraguay Guatemala El Salvador NLcaragua Honduras Haiti Singapore Malaysia Korea Thailand China Philippines Indonesia -2 -1 0 1 2 3 4 5 6 7 8 Source Lederman and Xu (2001) The Number of Patents Registered by Residents of Most presents two sets of statistics: the number of patents regis- Latin American Countries Is Low tered by residents of a country in their own country, as well R&D expenditures are geared toward innovation, and the as the number of patents registered by residents of this coun- number of patents registered by the residents of a country try in the United States (both per million inhabitants). The provides some measure of the "success" of these expenditures. advantage of the measure of patents registered in the United Patent registrations are therefore a critical measure of domes- States is that it abstracts from differences in the stringency tic inventive activity in a country l In chapters 6 and 7 we of examInation and granting rates of patents across countries. show that patent stock capital, a measure of the depreciated In so doing, it arguably provides a measure of innovative number of inventions available in an economy at any given activity of patents of similar "quality" (within a given tech- point in time, has a positive effect on TFP Table 2.7 presents nology field) across countries. The shortcoming of this mea- summary statistics on the number of patents registered per sure is that it also reflects, in part, the expected likelihood million inhabitants in a sample of Latin American, East that inventors of a given country will sell their product in Asian, and natural resource-abundant countries. The table the U.S. market, and may therefore be contaminated by trade 39 CLOSING THE GAP TABLE 2 6 have patent registration rates that are lower than expected Latin American Workers Have Small Amounts of R&D for their levels of per capita income. Expenditures TOTAL R&D R&D Latin American Countries Do Poorly by Aggregate PER WORKER (PERCENTAGE OF GDP) Measures of Technology Latin America 35.6 0.5 Given all of the evidence on the deficit in individual mea- Argentina 54 5 0 4 BolAvia 54 0.3 sures of technology, it is no surprise that Latin America does Brazil 58 0 0 6 poorly on aggregate measures as well. Figure 2.13 shows the Chile 44 2 0 7 evolution of a technology index constructed on the basis of Colombia 10 0 0 2 Costa Rica 15 6 0 3 19 standard indicators for Latin America and other regions.13 Ecuador 2 2 0 1 The figure shows that the technology gap between Latin El Salvador 15 8 0 3 Guatemala 0 2 ~~~~~~~America on the one hand, and the OECD and East Asian Guatemala 0 2 Mexico 20 7 0 4 tiger economies on the other more than tripled in the 1990s. Nicaragua 1 1 0 1 Even Eastern and Central Europe did better than Latin Panama 18 2 0 4 Peru 2 3 0 6 America, a fact that can be seen more clearly in the lower Uruguay 31 1 0 1 panel, which graphs the same measures without the OECD Venezuela 18 7 0.5 and East Asian tigers. East Asian tigers 329.6 2.2 In sum, compared to East Asia, Latin American countries Hong Kong (China) 0 3 appear to have less access to technologies from abroad: import Korea 434 0 2 8 Malaysia 22 8 0.2 penetration is lower, the fraction of imports that are capital Singapore 628 2 1 1 goods is also lower, licensing payments are smaller, and Natural resource-abundant computer penetration and ICT infrastructure is below that countries 725.4 1.9 of the more successful East Asian and natural resource- Australia 521 6 1 7 abundant countries. Taken together, this suggests that Latin Canada 507 4 1 6 Finland 1300 9 2 5 American countries are receiving less technology transferred New Zealand 291 2 1 1 from "leaders" close to the global technology frontier than Norway 956 8 1 8 Sweden 1709 7 3 4 their competitors Latin American countres also have low Note Regional averages of private and public R&D are weighted by the number of levels of domestic R&D and they register few patents-even workers in a country, R&D as a fraction of GDP is weighted by GDP size controlling for their income levels In practice, this means Source R&D data from Park (2002) and Lederman and Saenz (2002) that a worker in Finland, Sweden, Korea, or Singapore will be operating with more advanced technologies, both domes- and other flows across countries 12 Moreover, the relatively tic and foreign, than his or her counterpart in Brazil, high cost may further discourage inventors from poor coun- Argentina, or Mexico. tries from registering a patent in the United States. Table 2.7 shows that patent registrations in Latin American Conc1 sion' Ienchmarking Latin American countries are low by comparison with those of the East Asian Perorman ace tiger and natural resource-abundant countries. The four Latin America has profound deficits in productivity, skills, countries with the highest rates of patent activity in Latin and technology Table 2.8 summarizes some of the most America-Argentina, Chile, Costa Rica, and Venezuela- salient gaps we have discussed in this chapter. The table pre- all have registrations in the United States that are very far sents the mean level of a given outcome in Latin America, below those of Singapore, Korea, and Hong Kong (China), and the gap in this outcome-where the gap is defined as as well as in relation to countries such as Finland and Sweden. the difference between the observed value in the region and Are the levels of patent registration in Latin American coun- the expected value for its income level. For comparison, we tries low for their level of GDP per capita? Figure 2.12 sug- also present similar statistics for the East Asian tiger gests that this is indeed the case: most Latin American economies (Korea, Hong Kong [China], Malaysia, and countries, including Argentina, Brazil, Mexico, and Chile, Singapore) and the natural resource-abundant countries 40 THE GAPS THAT MATTER MOST FIGURE 2 11 Latin American Countries Also Have Low Levels of R&D Per Worker 8 8- 0 O ~~~~~~~~~~~~~~~~~~~~~~~~Sweden rs Finland SO~~~~~~~~~~~~~~~~~~~~~~~~~ a 6 - 0 v. 4- - < Agnia Mexico Uruguay P anama Malaysia r Kvador ~~~~~~~~~~Venezuela Cosa Rica v. 2 - /Trinidad and Tobago Ecuador Peru O Nicaragua 6 7 8 9 10 11 Log of Mean Per Capita GDP, 1996-2000 Note. The line corresponds to the predicced log of the stock of domestic R&D in 1996-2000 from a weighted regression on the log of per capita GDP and a constant, with the weights given by the number of working-age adults in a country The sample size for the regression is 73 countries Some country names have been omitted to make the figure more legible Source AuLthors' calculations based on data from Lederman and Saenz (2002) (Canada, Australsa, New Zealand, Finland, Norway, and East Assan tigers, whsch had well-above-average rates of TFP Sweden) Table 2.8 presents a stark picture indeed. Latin growth, also outperform Latin America on measures of AmerEca has low TFP growth relative to the East Asian tiger technology and skills. Within Latin America, the best- countries, and relative to other natural resource-abundant performing country, Chile, concurrently had positive countries. Moreover, controlling for their levels of per capita increases in productivity, substantial skill upgrading, and income, Latin America also has profound defrits in all four increases in foreign and domestic technology. As we will measures of educational attainment, and all four measures show in chapter 3, there is also evidence that firms that of technology. By contrast, both the East Asian tigers and increase their productivity in Latin America often do so in the natural resource-abundant countries have surpluses in all a context of technological upgrading and increases in the of these measures. demand for skilled workers. Table 2.8 shows the depth of the productivity gap in the The deficit in educational attainment in Latin America region: annual TFP growth in Latin America between 1990 is profound. Latin American adults of working age have and 1999 was a mere 0.45 percent, compared to the global almost one-and-a-half fewer years of schooling than would average of 0.66, the East Asian tiger average of 1 42 percent, be expected. This gap in the stock of educational attainment and the natural resource-abundant country average of 0.78 is a reflection of inadequate investments in the past. It is percent. therefore particularly worrying to observe that theflow of new Productivity differences between countries and between educated workers is inadequate. Latin America has very large firms within countries are profoundly affected by differences deficits in enrollment, particularly at the secondary school in skills and technology. It is therefore no surprise that the level, as well as a problem with the quality of education In 41 CLOSING THE GAP TABLE 2 7 in Latin America, the East Asian tigers, and the set of nat- Residents of Most Latin Ameuican Countries Have Low Rates ural resource-abundant countries, by decade, between 1960 of Patent Registration in Their Own Countries as Well and 2000. The figure clearly shows that Latin America has as in the United States. had a deficit since 1960-a deficit that became particularly DOMESTIC PATENTS US PATENTS acute between 1970 and 1980. In 1960, adults in the East TO RESIDENTS TO RESIDENTS Asian tiger economies also had almost three-quarters of a (PER MILLION (PER MILLION INHABITANTS) INHABITANTS) year of schooling fewer than would be expected for their Latin America 2.07 0.80 income level. The difference is, however, that the tigers Argentina 4 65 1 83 gradually reduced this deficit, so that by 2000 they had a Bovla 2 433 0 72 surplus of almost a full year of schooling The natural Chile 2 43 1 09 resource-abundant countries, finally, have always had a sur- Colombia 0 44 0 27 plus in educational attainment-a surplus that has grown Costa Rica 2 10 Ecuador 0 15 0 19 over time. In chapter 4 we discuss in detail the differences El Salvador 0 10 0 11 in the educational transitions among Latin America, East Honduras 0 30 0 48 Mexico o0 90 1.0 Asia, and the natural resource-abundant countries, focus- Nicaragua 0 41 0 08 ing on both changes in the mean and the distribution of edu- Panama 2 17 0 46 cational attainment. Peru 0 38 0 19 Trinidad and Tobago 1 33 0 70 Access to technology, both foreign and domestic, is low Uruguay 1 66 1 17 in Latin America. Import penetration is lower than Venezuela 1 58 2 02 expected, and import penetration of capital goods, which East Asian tigers 187 0 54.4 often embodies new technologies, is about one-quarter of that Hong Kong (China) 1 53 27 8 in the East Asian tigers. The average Latin American worker Korea 283.6 81 0 Malaysia 1 54 1 80 has only one-tenth as much domestic R&D capital available Singapore 113 1 484 as the average worker in the tiger economies, and one- twentieth as much as a worker in the natural resource- Natural resource- abundant countries 83.8 114.8 abundant countries. And residents of Latin America regis- Australia 96 1 49 0 ter far fewer patents domestically and in the United States Canada 26 8 138 3 Finland 133 6 160 0 than their East Asian or Scandinavian counterparts Work- New Zealand 80 4 38 3 ers in Latin America also have access to fewer computers. Norway 71 7 6620 With the exception of the measure of computer penetration, Sweden 184 3 208.6 Note The figures are the mean number of patents registered per year in the 1996-2000 these differences in innovation activity among the three sets period of countries can be explained only very imperfectly by the Source Authors' calculations, based on data from World Bank databases (domestic parents to residents), and dara fmm Lederman and Saenz (2002) (U S patents to residents) difference in their per capita incomes. In chapters 6 and 7 we return to these differences in the level of technology chapter 3 we show that there have been very sharp increases between the two regions. in the demand for more skilled workers in Latin America What can we say about changes in the technological gap in the last two decades. In order to compete with the East in Latin America over time? Figure 2.15 graphs decade aver- Asian tigers or with other countries with a comparative ages in the gap in the number of patents per worker regis- advantage in natural resources, both of which have high tered in the United States for Latin America, the East Asian levels of educational attainment, enrollment, and quality, tigers, and the set of natural resource-abundant countries inter- governments in Latin America need to put in place policies vals between the 1960s and the 1990s. As with the compara- that will allow young people to acquire more high-quality ble figure for the mean years of schooling, figure 2.15 shows education. that the East Asian tigers turned a deficit into a surplus, while Did Latin America always have such a profound deficit the natural resource-abundant countries increased the size in educational attainment? In Figure 2.14 we graph the gap of their surplus in patents. In Latin America, meanwhile, in the years of completed schooling of the adult population the deficit in patents has remained essentially unchanged. 42 THE GAPS THAT MATTER MOST FIGURE 2 12 The Rates of Patent Registration by Latin American Inventors Are Low for Their Levels of Per Capita Inconne 6 - Sweden I-~~~~~~~~~~~~~~~~~~~~~ a)Korea 4- X rs ,/ H~~~~~~~~~~~~~~~~~~~~~~~~~n Kong iChna) 50 3 ION/ Oz Vene Argenrina -k v)0 J caPaam rjleruguay Z C Honduras Do an Republic Nicaragua a Haiti Boivia Colombia S / ~~~~~~~~~~~~~~Peru 0 -3 6 7 8 9 10 11 Log of Mean Per Capita GDP, 1996-2000 6 SiSiga 25 4 6 ~~~~~~ ~ ~~~~~~~~~~~New Zeal d/oa C / Canada ~~~~~~~Argentina v Afama ~~~~~~~~~~~~~~~Brazil China enezucla Uruguay Hong Kong (China) ; / ~~~~~~~~~~~~~~~Thailand e 0 Mexico Nicaragua olornba Barbados XV2 ie Guatemala 0 -4 0 6 7 8 9 10 11 Log of Mean Per Capita GDP, 1996-2000 Note The number of patents has been multiplied by one million Source Authors' calculations based on data from World Bank databases and Lederman (2002) 43 CLOSING THE GAP FIGURE 2 13 The Widening of the LAC Technology Gap: Regional Comparisons of Index Scores Regional Index Scores 4 3 5 Z 3 25 2 15 0 - ---- 10 11 ~~192 _ I4 15.A3 8 1 -0 5 Year | +-Other Asian Pacific Countries - East Asian Tigers --*- Middle East and North Africa Sub-Saharan Africa | ---Organisation for Economic Co-operation and Development -o-- Latin America and the Caribbean - Eastern Europe (Without OECD and East Asian Tigers) 03 02 01 0 1990 1991 1992 1993 1994 1995 1996 1997 1 2000 -01 -0 2 -0 3 -04 4 -0 5 t -06 Year -r - Other Asian Pacific Countries -o-- Middle East and North Africa Sub-Saharan Africa -|-- Eastern Europe +--- Latin America and the Caribbean Source Guasch and Brehon (2002) 44 THE GAPS THAT MATTER MOST TABLE 2 8 The Gaps That Matter Most OBSERVED INCOME- INCOME- INCOME- EAST ADJUSTED OBSERVED ADJUSTED ADJUSTED ASIAN SURPLUS IN NRA SURPLUS OBSERVED DEFICIT TIGER EAST ASIAN COUNTRY IN NRA LAC MEAN IN LAC MEAN TIGERS MEAN COUNTRIES Productiviry TFP growth per year (1990-99) 0 45 -0 21 1 42 0 76 0 78 0 12 Education Mean years of education, population aged 25+ 58 -14 9 7 1 0 11 1 1 4 Net secondary enrollment rate (percent) 46 7 -18 7 93 3 17 8 93 2 6 0 Gross certiary enrollment rate (percent) 20 0 -10 0 47 3 5 1 67 0 14 0 TIMSS score (points) 387 -81 2 584 86 3 524 + (n s Technology Capital goods imports as a fraction of GDP, 1999 (percent) 7 7 -3 6 29 5 18 1 12 6 2 0 Domestic R&D expenditures per worker (1995US$) 356 -264 3295 1165 7254 1525 Patents registered in the UnLted States, 1996-2000 average (per million) 0 8 -1 5 54 4 48 3 114 8 80 9 Mean number of computers per 1,000 workers 37 7 - (n s ) 172 0 + (n s) 404 2 + (n s) Note The regional "deficits' or "surpluses' are calculated as the weighted sum of the difference between the observed value and the predicted value from an OLS regression on log per capita income for all outcomes except TFP growth and imports of capital goods For TFP, the means and the predicted value is given by the weighted world average, without a regression For imports of capital goods the explanatory variable is GDP The regional means, deficits, and surpluses for imports of capical goods are weighted by GDP, those for the mean years of education, net secondary enrollment rates, gross tertiary enrolment rates, TIMSS scores, and the number of computers are weighted by population, the measures of domestic R&D and patents, finally, are weighted by the working-age population aged 25 and older n s means that the deficit or surplus is not statistically significantly different from zero (at the 5 percent level) The East Asian tiger economies included in the sample are Hong Kong (China), Korea, Malaysia, and Singapore Data for Hong Kong are not available for TFP growth, imports of capital goods, or R&D expenditures, data on Singapore are not available for net secondary enrolment rare, and imports of capital goods, data on Malaysia are nor available for R&D expenditures The natural resource-abundant countries included in the sample are Canada, Australia, New Zealand, Finland, Norway, and Sweden Somre Data on TFP from Loayza, Fajnzylber, and Calder6n (2002), data on educational attainmenc of adults from Barro and Lee (2002), data on enrollment rates from UNESCO, data on TIMSS scores from TIMSS, data on imports of capital goods, GDP, GDP per capita, and population from World Bank databases, data on R&D and patent registration from Lederman and Saenz (2002) FIGURE 2 14 Between 1960 and 2000, the East Asian Rigers Turned a Deficit in the Mean Years of Schooling into a Surplus, while the Deficit in Latin America Did Not Change 2- 1 5 - I 04 .0 Natural Resource-Abundant Countries *17 East Asian Tigers -1 -2 -2 5 Source Authors' calculations based on data from Barro and Lee (2002) 45 CLOSING THE GAP FIGURE 2.15 Between the 1960s and the 1990s, the East Asian Tlgers Turned a Deficit in the Mean Number of Patents Per Worker Registered in the United States Into a Surplus, while the Deficit in Latin America Did Not Change 50 40 - , 30 01 4) U ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1960s C U~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1970s 20 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 1980s X a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1990S 10 0 Latin meric Natural Resource-Abundant Countries ffi ii ; z Latin America i L A ia :e East Asian Tigers -10 Source: Authors' calculations based on data from Lederman and Saenz (2002). Chapters 6 and 7 discuss these technological transitions and 2. In a population-weighted regression of the mean years of school- the implications they have for countries in Latin America. ing of the population aged 25 and older on the log of per capita GDP In order to reduce poverty, Latin American countries must and a dummy variable for Latin America for a sample of 105 coun- grow. Firms must increase their productivity to compete in tries in the world, the coefficient on the Latin American dummy is -3.82, with a standard error of 1.66. In a corresponding regression global marketplaces. As we will show in the following chap- with an East Asian dummy, the coefficient is 0.80, with a standard ters, the interaction between technology and skill is critical error of 0. 28. to setting countries in Latin America on a higher growth 3. The same pattern can also be observed in Thailand. path, fueled by a virtuous cycle of technological upgrading, 4. In a weighted regression of net primary enrollment rates on the increases in the demand for skill, skill upgrading, and fur- log of per capita GDP and a dummy variable for Latin America, the coefficient on the Latin American dummy is 4.3, with a standard error ther technological improvements, of 3.7. 5. The PISA results also show that students scoring in the top 10 Endnotes percent of all students taking the mathematics test in Mexico and 1. The 1980s were nor a good time for Latin America, with many Brazil score below the OECD mean, and even the top 5 percent of stu- countries mired in the recessions associated with the Debt Crisis. dents in Brazil only score at the OECD mean. See Carnoy (2002). Loayza, Fajnzylber, and Calder6n (2002) concede that their estimates 6. PISA separately reports scores for individuals with less than may underestimate TFP growth in the 1980s, and overestimate TFP upper secondary, complete upper secondary, and complete tertiary edu- growth in the 1990s-because adjustments of the true rate of uti- cation (OECD 2001). To abstract from differences in the distribution lization of labor and capital stocks, using rates of unemployment and of educational attainment in the population, we construct a mean, with hours worked, may be imperfect. equal weight given to these three scores. Of course, to the extent that 46 THE GAPS THAT MATTER MOST there are differences across countries in the distribution of educational Taiwan (China), Malaysia, Hong Kong (China), and Singapore are attainment within these large categories, so that the mean Chilean adult above the line Economies in Latin America and the Caribbean with with less than a completed secondary education has fewer years of higher-than-expected rates of import penetration are Barbados, Belize, education than his or her counterpart in (say) Poland, this is only an Domsnica, Guyana, Honduras, Jamaica, Nicaragua, St Kitts and imperfect adjustment for underlying differences in educational Nevis, St Lucia, St Vincent and the Grenadines, and Trinidad and attainment across countries Tobago 7 When we include a Latin American dummy in the global regres- 10 The four variables are aggregated by factor analysis, after imput- sion of the number of scientists and engineers per 10,000 workers on ing missing values with other variables such as GDP per capita In per capita GDP, the coefficient on the Latin American dummy is clearly 1997, the resulting index has a range between 8 83 (Norway) and the insignificant (the coefficient is 0 85, with a standard error of 4 52) Democratic Republic of Congo (-1 17), in Latin America, the lowest 8 In a population-weighted regression of enrollment in science and value is for Haiti (-1 0), and the highest for Chile (4 0) See Lederman engineering as a function of overall enrollment rates and a dummy vari- and Xu (2001) able for Latin America, the coefficient on the Latin American dummy 11 See especially the review by Griliches (1990) is -4 1, with a standard error of 3 4 12 See Grossman and Helpman (1991) In practice, both measures 9 In an OLS regression of import penetration on log GDP and a of patent activity are highly correlated in the data set we use The constant, for a sample of 152 economies, most Latin American GDP-weighted correlation coefficient is 0 89, significant at the 1 economies, including (but not only) Argentina, Brazil, Venezuela, percent level Colombia, Chile, and Peru, are below the regression line, while Korea, 13 Guasch and Brehon (2002) 47 CHAPTER 3 How Technology and Skills Interact: The Evidence for Latin America and the Caribbean HE CENTRAL MESSAGE OF THIS REPORT IS THAT THE INTERACTION BETWEEN technology and skill is critical in determining growth, productivity, and the distribution of earnings. This is of particular relevance today, given the wave of rapid technological change in the United States and other OECD countries-technological change that has been com- plementary with skill. Chapter 2 showed that Latin America has profound deficits in both education and technology-in this chapter we turn to a dynamic picture of changes in Latin America, focus- ing primarily on the 1980-2000 period. We make use of the macroeconomic, firm-level, and household- level data in the region (box 3.1), and show that technology and skills interact in important ways in Latin America. The chapter makes the following four points: *There is a great deal of evidence of a wave of technological change that is complementary with skill in Latin America. As a result, firms have substantially increased the demand for educated workers in the region, especially workers with university education. The rising demand for educated workers has bid up the relative wages of these workers. * This technological transformation appears to be int- Latin American countries in the 1990s are similar to mately related to patterns of integration in the world those observed in Chile in the 1980s. economy. Trade and foreign direct investment (FDI) have facilitated the transmission of skill-biased tech- The Rising Demand for Skilled Workers nology across borders. Also, firms in sectors with in Latin America higher exposure to trade and FDI are subject to more competitive pressures. Using more advanced tech- In Many Countries in Latin America, the Relative nologies and hiring more educated workers is one Wages of Workers with Tertiary Education Have Been way to respond to this pressure to become more Rising over Time... productive. We start off with an analysis of data from labor force sur- * Increases in the demand for skills associated with the veys. The six panels in figure 3 1 show measures of relative adoption of new technologies have primarily affected supply and relative wages for workers with tertiary and sec- countries with levels of human capital above a mini- ondary education for Argentina, Bolivia, Brazil, Chile, mum threshold level. Colombia, and Mexico (see box 3.2). All series have been nor- * There appear to be long-term cycles in the relative malized, so that the relative wages and relative supplies are demand for more skilled workers associated with tech- given a value of 100 in the first year for which data are avail- nological progress. The patterns observed in most able.i Figure 3.1 shows that in all of the countries except 49 CLOSING THE GAP BOX 31 Daba Sources Three different sources of data are used in this chapter: Information on trade, foreign direct investment and household-level data, firm-level data, and macro-level data. other variables of interest is only available at a more aggre- Each of them has its strengths and shortcomings, hence gate level-for example, by country or industry These data informing different aspects of the questions we are inter- do not therefore have the richness associated with micro- ested in. This box briefly discusses what these strengths level data, but they do cover a larger number of countries and shortcomings are and time periods. Many countries in Latin America regularly collect large Most of the original work presented in this chapter uses household or labor force surveys. These surveys provide household-level data, while the analysis based on firm-level reliable, detailed information on the "skills" of the labor data draws from the work of other authors. The table below force, in particular on education levels, and these are fairly provides more details on these household data, their sam- comparable across countries. For our purposes, the main plmg frequency, and the periods covered by each of them. disadvantage of the household-level data is that they do not collect information about the level of technology in WouselOil4-Le3u Data Seomces the economy or across different firms. However, because PERIOD relative wages are a signal, they can be used to make infer- SURVEY FREQUENCY COVERED ences about the demand for workers with different amounts Argenitina Survey of Greater of skill, and about how these differences in demand relate Buenos Aires Yearly 1986-2000 Bolivia Encuesra Integrada| to changes in technology. Bolivia Eiicude Hogares Yearly 1989-95 Firm-level surveys frequently contain direct informa- Encuesca Nacional tion on technology upgrading by firms, including variables de Empleo Yearly 1996-97 Enscuecia Nacional such as investments in different forms of capital equip- de Hogares Yearly 1999-2000 ment, import of foreign materials, and use of patents. Brazil Pesquisa Mensal de Emprego Monithly 1982-99 Unfortunately, these surveys only collect information on Chile Encuesra de la firms in the manufacturing sector, and cannot therefore be Universidad de Chile Yearly 1966-99 used to analyze technology and skill upgrading in other Colombia Encuesta de Ingresos de los Hogares Quarterly 1982-2000 sectors, such as services. Moreover, there is often very little Mexico Encuesta Nacional information on the skill level of the workforce employed de Empleo Urbano Quarterly 1987-99 by firms, which IS generally divided into "production" and Ns'oe Some of these surveys use a roractng panel scheme for data collection by firms, which is genemUy divided into "production" and (Argentna and Mexico), bur we do not make use of this feature and treat the data as "non-production" workers, rather than by education level. a series of cross-sections Chile the wages of workers with tertiary education rise ... in Spite of Their Increasing Relative Abundance steadily relative to those of workers with secondary educa- The second striking feature of figure 3.1 is the fact that in tion. In Chile, meanwhile, the relative wages grew by 27 2 all of the countries except Brazil, this increase in relative percent in the 1980s, and fell by 13 5 percent in the 1990s wages took place in spite of an increase in the relative supply In spite of these common patterns, however, figure 3 1 also of workers with tertiary education. The biggest increase in shows that the magnitude of changes in relative wages varies relative supply in the 1990s occurred in Mexico (34.0 percent), a great deal by country. Calculating growth rates for the which makes the magnitude of the change in relative wages series, and focusing on the 1990s, the relative wages of work- there even more remarkable. The relative supply of work- ers with tertiary education increased by an incredible 72 9 ers with tertiary education increased by little in Colombia percent in Colombia, by 48.3 percent and 45.4 percent in (7.2 percent), and moderately in Argentina (25 percent), Mexico and Bolivia, and "only" by 19.7 percent and 11.7 Bolivia (20.9 percent), and Chile (16.9 percent). In Brazil, percent in Argentina and Brazil, respectively. the relative supply of workers with tertiary education fell 50 HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN FIGURE 3 1 The Relative Wages and Relative Supply of Workers with Tertlary Education Increased Substantially in Most Latin American Countries Argexntina Bolivi 20020 260 160 140 4140 1 20 '- ~ ~ ~~ ~ 120- __ _ _ __ _ _ __ _ _ __ _ _ 40 40 20 2 1986 1988 1990 I992 1994 1996 1998 2000 1989 1990 1991 1992 1993 1994 1995 1996 1997 1999 Year of Survey Year of SurveY Urban Brazil Ch.l. 200 200 100 -0, 60 60 40 40 6 060 140 .140 , " 8 20 ,S~~ '8120- S~ / 0 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 1982 1984 1986 19S8 1990 1992 1994 1996 1998 20 1 18 19 Year Of Survey Year of Survey 160 200 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~200 C 1- 00~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~1- so 40 41. 20. 20. 1982 1984 1986 (988 (990 (992 1994 1996 1994 2000 1987 1988 (9S9 1990 (991 1992 (993 199 (995 (996 (997 1 990 1999 Year of Survy Yea of Survy Wq -Ee~tve - Relativ S.pply Note Relative wages and supply series are three-year moving averages Source SAinchez-Pdramco and Schady (2002), based GD the labor force surveys 51 CLOSING THE GAP BOX 3 2 Illeasuring Education, Relative Wages, andl Relative Supply Much of the analysis in this chapter is concerned with dif- An alternative method of classification is proposed in ferences in the relative wages and supply of workers with dif- this chapter that, in a sense, combines the two described ferent amounts of education. This box briefly discusses how above. Specifically, individuals are assigned to two differ- workers are assigned across education groups and explains ent education groups- The group corresponding to the how relative wages and supply measures are constructed. highest education level they have completed and the group Education measures: Two issues need to be consid- corresponding to the next education level they can com- ered concerning the assignment of workers to different plete. Weights are then constructed that reflect the indi- education groups and the interpretation of such a classi- vidual's relative progress between both education levels. fication. First, education systems vary across countries, so This procedure can best be illustrated with an exam- that a person who has completed secondary school in Chile ple. Consider again an individual with seven years of edu- has not necessarily spent the same number of years in cation in Mexico. As described above, this person will have school as a person who has completed secondary school in completed her primary studies plus one year of secondary Mexico. The table below provides a schematic summary studies, and will therefore be included in both the "pri- of the education system in the countries of interest. mary education" and "secondary education" groups. Notice that secondary school requires a total of six years Education Levels and Years of Schooling of schooling in Mexico, out of which she has only com- pleted one. This implies that she has completed one-sixth YEARS of secondary school. She is therefore given a weight of 1/6 ,COUNTRY PRIMARY SECONDARY TERTIARY *. * COUNTRYiPRIMARY SEcONDARY TERTIARY when including her in the secondary education group and Argentna` 7 5 3-5 Bolivia 8 4 3-5 a weight of 5/6 (1 minus 1/6) when including her in the Brazil 8 3 3-5 primary education group. Chileb 8 4 3-5 There are two reasons why this seems a preferable way Colombia 5 6 3-5 Mexico 6 6 45 of assigning workers to schooling categories. First, in many a The Argentine system changed in 1997-98 to primary (6 years), lower secondary Latin American countries large shares of the population (3 yeats), and upper secondary (3 years). This change does niot affect the analysis have only a few years of education. In that context restrict- since we focus exclusively on individuals aged 25 and above b Individuals born prior to 1950 were educated under a dufferexit system in Chile ing the analysis to only those individuals who have actu- (prnmary, 6 years, secondary, 6 yeas, post-secondary, 1-5 years) This change waJs taken into account vvhen assigning individuals to d Tfferen educarion greups baed ally completed at least one education level would miss a on their years of schooling significant fraction of the population. Similarly, assigning I these individuals to the no-education group would also be Second, workers can be assigned across education groups misleading, since most of them do have some education. according to different criteria. The two most common cri- Second, a substantial amount of the educational progress teria in the literature have been those of partial or full com- that takes place in developing countries over time occurs pletion of a certain education level. For instance, a person within rather than between levels. This skill upgrading with seven years of education in Mexico would be con- would be missed altogether if workers were assigned to sidered a primary school graduate and hence assigned to education categories corresponding to the levels they have the primary school category if the full completion of a level completed or begun. The proposed procedure, by contrast, criteria were used for classification. On the other hand, accounts for these within-category changes with changes since she has some secondary school, she would be assigned in the weights that are assigned. to the secondary school category if the partial completion Relative wages and supply measures. Once workers have criteria were used instead. been assigned to different education groups, relative wage and 52 HOW TECIINOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMfERICA AND TIIE CARIBBEAN supply measures are constructed as follows Relative wages (2) S'J = (% education level t in labor force in year t)l for workers in education groups i and ; in year t are calcu- (% education level ; in labor force in year t) lated as the difference in their average (log) hourly wages: Finally, we restrict the sample to males and females aged (1) W' log(hourly age)1 log (horly wag)It' 25-60. Within this group labor force participants- where z = tertiary or secondary, and j = secondary or pri- employed and unemployed workers, defined in the stan- mary, respectively. Similarly, the relative supply in year t dard way-are considered when constructing labor supply is calculated as the ratio of the percentage share of groups measures, while only salaried workers are considered when z and ; in the labor force: constructing wage measures. by 7 3 percent, a change that can be explained by increases ative supply of workers with secondary education went up in the number of workers with secondary education in by more than 18 percent in every country, with particularly the absence of comparable increases in those with tertiary sharp increases in Bolivia (39.6 percent), Colombia (32 8 education percent), Mexico (29 2 percent), and Argentina (29 1 per- cent). In Brazil, there has been a notable increase in the rel- Which is Strong Evidence of Demand-side Changes ative supply of workers with secondary education since 1994 Favoring Workers with Tertiary Education (26.7 percent). In Chile, the most dramatic skill upgrading The relative wages of skilled and unskilled workers are a at this level took place in the 1980s, with an increase of 63 1 signal of the relative demand for and supply of skills in an percent. These changes reflect widespread public efforts to economy In the absence of changes in relative demand favor- increase secondary school enrollments in virtually every coun- ing skilled workers, we would expect the relative wages of try in Latin America skilled workers to fall because of their increasing relative abundance. In all of the countries in our analysis except ... and a Decrease in Their Relative Wages in a Brazil we find increases in both the relative wages and the Number of Countries relative supply of workers with tertiary education. This is Figure 3 2 also shows that the relative wage of workers with strong evidence of demand-side changes favoring workers secondary education fell over the period in a number of coun- with tertiary education in the region. Indeed, these patterns tries. As with figure 3 1, and focusing on the 1990s, the suggest that the demand-side shifts towards workers with magnitude of these changes varies a great deal by country- tertiary education would have to be so large as to more than from a very small decrease in Colombia (-0.4 percent), to offset the downward pressure on wages that results from their much larger decreases in Argentina (-32 3 percent) and increasing relative abundance Separate graphs that break Brazil (-20.3 percent). Only in two countries did the rela- down these changes by sector suggest that the increase in tive wage of workers with secondary education actually relative demand for workers with tertiary education has taken increase over the period-by 8.9 percent in Mexico, and by place in both manufacturing and services. a staggering 375 percent in Bolivia. A combined look at figures 3.1 and 3.2 makes obvious that the wages of work- There Have Been Increases in the Supply of Workers ers with secondary education in every country but Bolivia with Secondary Education Relative to Workers with and Mexico therefore fell relative to both workers with pri- Primary Education in Latin America ... mary education and those with tertiary education. The evi- Turning next to workers with secondary and primary edu- dence on changes in relative demand for workers with cation, figure 3.2 shows clear increases in the relative supply secondary education by firms is therefore unclear in every of workers with secondary education. In the 1990s, the rel- country except Bolivia and Mexico 53 CLOSING THE GAP FIGURE 3 2 The Relative Supply of Workers with Secondary Education Increased Substntially in Latin Ameiica Argentia Bobvia 200 140... 350 , 120. g Oo,, los- , f n °>so ... .. ,,00 60. ,,,..-- 150- 100~~~~~~~~~~~~~~~~~~~~~~~~~~~~~0 20 ... 0 01 1986 1988 1990 1992 1994 1996 1998 2000 1989 i9go 1991 1992 1993 1994 1995 1996 1997 1999 Yen of Su.vey Yea- of Survey Urboo Bra-il Chil. 20 010 4 ,_,,,__, tSO 350, 16030 140 250 60. 0 20 o. 1902 904 1906 1900 1990 1992 1994 1996 1998 00 o 1997 198 193 400o 199 199 199 40' 195 1,9 199 1 Yealr of Survey YeaDr of Survey 1100 - - - - - - - - - - - - - - - - - - 200 140 150 a - -. r- a.. -~~~~~~~IO - - - -- - - 54 0 20 OF~~~~~~~~~~~~~~~~~~~~~~~~~~~~~2 1902 1984 4 198 1988 1990 1992 19 94 im 190 1990 800 i9 1990 90 190 99 I99 199 104 5 1 1995 190 le Year of Survy Yea of Survy C.I.-bi. ~ ~ -RltieW0o rlov Spl Note RIaiewSsadOppysre r he-ya oigaeae Sourc I~ce-rm n cay(02,bsdo h ao oc uvy 40 40~~~~~~~~~~~~~5 HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LAI IN AMERICA AND THE CARIBBEAN Changes in Relative Wages Can Be Decomposed into gies flow primarily to countries with skill levels above a mm- Changes in Relative Demand and Relative Supply... imum threshold level The absence of demand shifts for more Because relative wages are determined by relative demand educated workers in Brazil may itself have been determined and supply, changes in relative wages can be decomposed into by the fact that education levels in the country were very low. changes in relative demand and changes in relative supply (box 3.3). To do this, assumptions have to be made about A Similar Decomposition also Shows that There Have the value of a, the parameter for the elasticity of substitu- Been Increases in Relative Demandfor Workers with tion between workers with more or less education. This elas- Secondary Education in Most Countries ticity of substitution is a measure of how sensitive the Because the relative supply and relative wages of workers with quantity demanded of workers with a given level of educa- secondary education over those with primary education have tion is to changes in their relative price, and is related to generally moved in opposite directions in Argentina, Brazil, the ease with which one kind of worker can be substituted Chile, and Colombia, it is not clear apriori whether there have for another in production. Figure 3.3 presents a series of been demand-side shifts favoring workers with secondary edu- graphs of the evolution of relative demand for workers with cation. Figure 3.4 presents the results of a decomposition of tertiary education for values of the elasticity of substitution the relative wages of workers with secondary education The between 1 and 3-the values that are commonly thought results vary substantially from country to country In Bolivia, to be "reasonable" for developing countries. The value of rel- there was a massive demand-side shift favoring workers with ative demand at any point in time is not a meaningful con- secondary education over workers wich primary education in cept, so all of the series have been normalized to zero for the the 1990s. In Argentina and Brazil, the pattern of demand- first year in which data are available (See Murphy, Riddell, side changes is quite sensitive to the choice of a However, and Romer (1998), especially pp. 27-28) On the other hand, abstracting from two crisis periods (1988-90 in Argentina, changes in log relative demand over time, given by the slope and 1990-92 in Brazil), the overall pattern seems to be of a of the lines, measure shifts in the demand schedule, and can moderate increase in relative demand for workers with sec- be compared across time periods and countries.2 ondary education. The time series for the other four countries appear to be closely related to those found for tertiary-educated ... Which Generally Confirms the Sharp Increases in workers Figure 3 4 suggests a consistent increase in demand Relative Demandfor Workers with Tertiary Education for workers with secondary education in Colombia In Bolivia, Figure 3.3 confirms that, no matter what the chosen value for Chile, and Mexico, a period of increasing relative demand for the elasticity of substitution, there is evidence of increases in secondary-educated workers (1990-94 in Bolivia, 1974-85 relative demand for workers with tertiary education in the late in Chile, and 1988-93 in Mexico) is followed by a flattening 1980s and in the 1990s in Argentina, Bolivia, Colombia, and or fall in relative demand This is precisely the pattern found Mexico In Chile, the steep demand-side shifts favoring skilled for workers with tertiary education. workers took place in the late 1970s and in the 1980s, with the overall pattern in the 1990s being essentially flat In Bizil, Skill-Biased Technological Change the estimated demand-side shifts are much more sensitive to in Latin America the choice of a, the parameter for the elasticity of substitu- The evolution of relative wages and supply make clear that tion. As with the relative wage series in figure 3.1, figure 3.3 firms in Latin America sharply increased the demand for more shows that the magnitude of the demand shifts favoring work- educated workers. We now show that the household-level ers with tertiary education varied a great deal by country, and data, firm-level data, and aggregate country-wide data are all over time. In Colombia and Mexico, firms appear to have most consistent with an explanation of the demand shifts that increased the demand for workers with tertiary education stresses the complementarity between technology and skill particularly sharply. In a background paper prepared for this report, Sanchez-Pdramo and Schady (2002) show that these Demand-side Shifts Favoring Skilled Workers Could Be demand changes are as large or larger than those that occurred the Result of Any One of a Number of Economic Reforms in the United States.3 In Brazil, meanwhile, demand shifts were The 1980s and 1990s were a period of profound changes in anemic. As we discuss later on in this chapter, new technolo- much of Latin America: Formerly closed economies opened 55 CLOSING THE GAP BOX 33 Constructing Relative iDemand Shifts: What Needs to Be Assumted and Why Changes in the relative wages of workers with different edu- where all of the parameters on the right-hand side are now cation levels are a function of changes in the relative demand "given." Unfortunately, because all of the series for the for and the relative supply of these workers, as well as of countries we analyze in this report are rising (or falling) the elasticity of substitution between workers with differ- almost monotonically, estimates of a conditional on a time ent education levels. This elasticity is a measure of the ease trend are highly imprecise (with t-statistncs of one or with which one kind of worker can be replaced for another lower), and often wildly improbable, rendering this first in production. For instance, if the elasticity of substitution approach invalid. between engineers and technicians is high, a small increase The second approach relies on assumptions about aT in the relative wage of engineers would lead to a large sub- rather than on assumptions about the behavior of D(t), thus stitution out of engineers and into technicians, whereas if avoiding econometric estimation. In particular D(t) is then the elasticity is low changes in the wages of engineers would estimated directly from (3) under plausible values of aT. In lead to little substitution between both types of workers. this chapter r is assumed to vary between 1 and 3, which More formally, if a, the parameter for the elasticity of sub- is the range of international estimates summarized in Katz stirution, is equal to x, a 1 percent increase in the relative and Autor (1999). wage of engineers is associated with an x percent decrease An additional concern with simple estimates of a and in the quantity demanded of engineers relative to techni- D(t) obtained from equations (2) and (3) is the changing cians. These ideas can be modeled simply as follows. Under composition of the workforce. The five countries in the the assumption of a common elasticity of stubstitution (CES), sample are relatively advanced in their demographic tran- relative wages have to satisfy the condition: sitions. The fraction of workers who are older is therefore higher in later than in the earlier years. Older workers gen- (1) Log[w,Qt)w2t)] = (1Ia)(D(t) - log [X1(t)/X2(t] erally earn more than younger workers, and a life-cycle where w,(t)IW2(t) is the ratio of relative wages, x,(t)1x2(t) model of earnings determination suggests that the wage is the ratio of relative supplies, cr is the elasticity of sub- gap between more and less educated workers should also stitution between workers in the two education levels, and increase with age (Mincer 1974; Heckman, Lochner, and D(t) is the time series of relative demand shifts (Katz and Todd 2001). The observed increase in the relative wage of Murphy 1992). the skilled could, therefore, be a product not of the chang- Since both D(t) and a are unknown parameters, addi- ing supply or demand for educated workers, but rather of tional assumptions about either the behavior of D(t) or the the changing age profile Many countries in Latin value of a are necessary in order to construct a time series America are also witnessing important changes in the par- of relative demand shifts using (1). Two different approaches ticipation of women in the labor force. If the difference in have been proposed in the literature for this purpose. First, wages by education is larger (or smaller) for women than a working value of a can be estimated if relative demand for men, this too could distort uncorrected estimates of shifts are assumed to follow a linear trend by running a D(t). Again following Katz and Murphy, compositional regression of the (log) relative wages on the (log) relative changes in the labor force were controlled for by holding supplies and such a trend as follows: age and gender distributions constant over time in each (2) = a + (1/a) time trend country. Specifically, for any education group, average share 2 Log[w1(t)1W2(t of total employment were calculated for 14 age-gender _ (1'Ia) log [x1(t)1x2(t)1 cells over the entire period and these weights were then used The value of a can then be recovered from the coeffi- to construct mean wages for an education group in any cient on relative supply and plugged into (1) after rear- given year. ranging terms to obtain: (3) D(t) = a log[w,(t)1w2(t)] + log [x,(t)1x2(t)1 Source- Katz and Murphy (1992). 56 HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LAI IN AMERICA AND 1 1E CARIBBEAN FIGURE 3 3 The Demand for Tertiary Workers Increased in Every Country except Brazil Mexico _Bolivia 16 16 1 4 _4 1 02 - oO _ . 1982 1984 966 1988 199 1992 1994 996 199. ........'. _ o = -402 198 19804192 19 96 98 20 o 06 2 o. 296 -- --- La 0 1987 1988 1909 1990 1991 1992 15999 1994 1995 1996 1997 1998 1999 1989 9911 1991 1992 1999 1994 1995 i999 1997 1 999 Year of Surey Year -f Survey B-dzi Chile 1 2 0 6~~~~~~~~~~~~~~~~~~~~~i 04 ~~~~~~~~~~~~~~~~~~.0 1 NoeD-mn 02 le are rhe-ya leln averages'l e elelel -0 2~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ -046 Snbb h-oS.-O - f S._.5 -9.. -*61~~~~~~~~~~~~~~~~~~~~~Ar-. 14~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ 1 2~~~~~~~~~~~~~~~~~~~~~~~1 - 12~~~~~~~~~~~~~~~~~~~~~~~~~~1 00 E o 02 -0 . 1982 1984 1986 1988 1990 1992 1994 1996 1998 25000. -92 1~~~ ~ ~~~~ ~ ~~~~~ ~ ~~~~~ ~~~~~ ~ ~~~~900 1980 1990) 1992 19941 1996 1990 20)00 Yea of Survey Yea of Survy --El.astcy I . .. Elxs-cii 2 - Elastic,ty S Note Demand series ore three-year mnoving averages Source Sgnchez-PAramo and Schady (2002), based on the labor force surveys 5 7 CLOSING THE GAP FIGURE 3 4 The Demand for Seconda,y Workers Increased in Every Country except Argentina and Brazil Artgentn, Boh-ix r O23ol18 90 19 9 99 9495' 96 ........... V 089 04 CD 07 - - - ------ - 1 xlxoi Chll -02 .- 1996 1988 1191 1994' I!t6.. 1998 2000 ~~~~~0 03 -o 4 ~~~~~~~~~~~~~~~~~~~~~~~~1909 1990 001i 19092 1993 1994 1995 1996 3007 1999 -0 4 -0 Yetx of Survey Y" of Surey 8-oml1 Mex,k 006 0 6 .0 -0~~~~~~~~~~~~~~~~~~~~~~ - 0 i , o2'ai°2 ,.3 .3. -219S2 Is sN 1E3 1s 92 1964 1986 19s 2e0 3-2 97tis1s 1990 i9119 X 9419 996 1998 4, 19 i.0 2-0 Y..u of S-r7y Yeir of Su-a7y 05 ,. 04 8 0 ~0 -0 2. -~~~~~~~~~~~~~~~~~~~~0 02. . . . I .. .. . 5 A 3 -024-0 Yea of Survy Y-e. of SurveY 0 .0~~~~~~~~~~~~~~~~~~~~~~~ HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN up to trade and FDI; Latin American firms were forced to skill-intensive sector. The accompanying Stolper-Samuelson raise their productivity to compete in the global marketplace; theorem predicts that this change in product prices will be in some countries, modest attempts were made to reform translated into an increase in the wages of workers in the labor markets All of these changes could potentially explain skill-intensive sector of the economy. Since the skill-inten- the rising demand for skilled workers in the region. As it sive sector, by definition, employs more skilled workers, turns out, the patterns we observe are not consistent with changes in product demand should therefore lead to an an explanation based on product demand shifts, such as those aggregate rise in the wages of skilled labor in a skill-intensive predicted by the Heckscher-Ohlin (H-O) model of interna- country. It is not clear a priori whether Latin American coun- tional trade, or with changes resulting from labor market tries are relatively skill-abundant or relatively skill-scarce.4 reform. Rather, the evidence is most consistent with the Also, in many Latin American countries tariff protection was effect of skill-biased technological change on the relative highest in sectors that employed the highest fraction of demand for workers with different amounts of skill. unskilled workers.5 As a result, tariff declines were largest in sectors that were relatively skill-unintensive.6 Trade lib- Labor Market Reforms Do Not Explain the Patterns eralization therefore reduced the relative price of goods pro- Observed in Latin America duced by the skill-unintensive sector, and this could have The impact of labor market reforms on the relative demand translated into a reduction in the relative wages of the for skilled workers is likely to have been modest simply unskilled through Stolper-Samuelson effects because the magnitude of these reforms in Latin America was The second effect of trade reform is that it facilitates the modest Proposals to increase labor market flexibility in Latin transmission of technology across borders Decomposing skill America in the 1980s and 1990s often faced substantial upgrading into within-sector and between-sector compo- opposition, making them difficult to implement. In fact, nents provides an empirical test to distinguish between while most countries in the region carried out profound trade H-O-Stolper-Samuelson and skill-biased technological and financial reforms, only a few actually introduced sub- change The critical prediction of an H-O-Stolper-Samuel- stantial changes to their labor legislation during the 1990s son type explanation is that economy-wide skill upgrading (Gill, Montenegro, and Domeland 2002). Moreover, although is a result of an increase in the relative size of skill-intensive generally intended to increase labor market flexibility, sectors. This, in turn, increases the aggregate demand and reforms were not always successful at achieving this. A paper the relative wage of skilled labor. Because skilled workers by Behrman, Birdsall, and Szekely (2001), which graphs the are now relatively more expensive, however, we would expect evolution of various indexes of reform in Latin America, to see within-sector substitution away from skilled labor and including a composite measure of labor market reform, sug- into unskilled labor. By contrast, skill-biased technological gests that labor markets were, on average, no more flexible change is consistent with within-sector increases in both the in the 1990s than in the 1980s. relative wage and employment share of skilled workers. A standard way of considering the importance of these two Trade Reforms Could Explain Many of the Patterns competing explanations for the increasing demand for skilled We Observe workers is to decompose the change in the aggregate wage bill, Product demand shifts associated with trade reforms in Latin given by the product of employment and the wage, into America seem like a more serious contender for the between-industry and within-industry components 7 Within- observed increases in relative demand. Trade liberalization industry changes in the wage bill can then be attributed to skill- in a country will generally have two effects First, it will biased technological change or skill-biased capital accumulation, change the relative price of goods produced in different sec- while between-industry changes in the wage bill can be attrib- tors This, in turn, will have an effect on the relative size of uted to increases in the relative size of skill-intensive sectors 8 sectors, and on the relative wages of workers with different levels of education Specifically, in a two-sector economy with The Bulk of Skill Upgrading in Latin America different skill intensities, the H-O theorem of international Has Taken Place within Industries trade predicts that trade reform in a skill-abundant coun- We turn again to the household data for Argentina, Bolivia, try will increase the relative price of goods produced in the Brazil, Chile, Colombia, and Mexico, and disaggregate the 59 CLOSING THE GAP wage and employment data into the greatest number of sec- of the observed skill upgrading can be explained by within- tors possible-from 13 in Bolivia to 75 in Mexico The sector changes. In Chile, the labor force survey data shows results from this decomposition are presented in table 3.1. that just over 90 percent of the skill upgrading in the 1980s The top panel considers changes in the share of the wage bill took place within sectors.9 The picture for changes in the for workers with tertiary education (as a fraction of the wage secondary wage bill in table 3.1 is, if anything, even clearer: bill for workers with secondary and tertiary education), while 90 to 150 percent of the increase in the wage bill for work- the bottom panel considers changes in the share of the wage ers with secondary education took place within industries. 1 bill for workers with secondary education (as a fraction of the wage bill for workers with primary and secondary edu- ... Which Is Consistent with Skill-biased Technological cation). Table 3.1 shows clearly that the bulk of the increase Change or Capital-skill Complementarity, But Not in the aggregate wage bill for workers with tertiary educa- with H-O-Stolper-Samuelson Effects of Trade Reform tion took place within industries. The fractions vary by Within-industry upgrading is consistent with skill-biased country-from 72 percent in Argentina to 122 percent in technological change, or with a complementary relationship Bolivia (within-industry changes larger than 100 percent are between capital and skill together with a falling relative price possible if there are decreases in the relative size of the skill- of capital. Empirically, skilled labor and capital do appear intensive sectors). In Mexico, despite the greatest number to be complements (Berman, Bound, and Griliches 1994). of sectors in the data, which should allow for a more accu- Capital is likely to have become relatively more abundant rate measurement of shifts across sectors, fully 95 percent and cheaper in Latin America as a result of trade liberalization, TABLE 3 1 The Bulk of Changes in the Wage Bill for Tertiary Workers and Secondary Workers Occurred within Industries TERTIARY CHANGE IN S, CHANGE DUE TO WITHIN- CHANGE DUE TO BETWEEN- NUMBER (ANNUALIZED - INDUSTRY CHANGES INDUSTRY CHANGES PERIOD OF SECTORS INITIAL S, % POINTS) (ANNUALIZED - % POINTS) (ANNUALIZED - % POINTS) Argentina 1986-89 and 1997-99 25 53 0 1 8 1 3 0 5 Bolivia 1989-91 and 1997-99 13 59 7 18 2 2 -0 4 Brazil 1982-83 and 1987-89 48 60 9 -0 1 -01 00 1987-89 and 1997-99 48 60 6 -0 2 -0 2 00 Chile 1977-79 and 1987-89 39 510 1 2 1 1 0 1 1987-89 and 1997-99 39 60 8 0 1 02 -0 1 Colombia 1982-83 and 1988-89 34 486 -0 1 00 -0 1 1988-89 and 1998-99 34 482 11 09 02 Mexico 1987-89 and 1997-99 75 36 6 2 1 20 0 1 SECONDARY CHANGE IN S, CHANGE DUE TO WITHIN- CHANGE DUE TO BETWEEN- NUMBER (ANNUAUZED - INDUSTRY CHANGES INDUSTRY CHANGES PERIOD OF SECTORS INITIAL S, % POINTS) (ANNUALIZED - % POINTS) (ANNUALIZED - % POINTS) Argentina 1986-89 and 1997-99 25 45 9 1 0 09 01 Bolivia 1989-91 and 1997-99 13 48 2 3 7 3 6 0 1 Brazil 1982-83 and 1987-89 48 43 7 1 6 15 0 1 1987-89 and 1997-99 48 49 4 06 09 -0 3 Chile 1977-79 and 1987-89 39 63 5 1 0 1 1 -0 1 1987-89 and 1997-99 39 72 4 06 06 00 Colombia 1982-83 and 1988-89 34 513 1 3 1 4 -0 1 1988-89 and 1998-99 34 56 7 1 4 1 3 00 Mexico 1987-89 and 1997-99 75 48 8 06 07 -0 1 Note The wage bill for tertiary workers, S,, is the wage bill of tertiary workers as a fraction of the wage bill for secondary and tertiary workers, while the wage bill of secondary workers, $,, is the wage bill for secondary workers as a fraction of the wage bill for primary and secondary workers Source Sanchez-Paramo and Schady (2002) 60 HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN rising FDI, and other developments that have increased cal change is pervasive, we would expect to find a positive cor- investment A careful analysis of the changes In the use of relation between the degree of skill upgrading across sec- skilled labor by Colombian firms between 1983 and 1998 tors in different countries. In table 3.2, we present weighted (which takes into account the effect of the changing price correlations of the changes in the employment share of work- of capital) suggests that skill upgrading has been driven ers with tertiary education (upper panel) and that of work- both by capital-skill and technology-skill complementarity ers with secondary education (lower panel). Because the (Kugler 2002) In general, however, the changes in the price patterns we observe in Chile in the 1 980s resemble those in of capital in Latin America and in other developing coun- other countries in the 1 990s, we correlate employment shares tries appear to be much too small to account for the observed for Chile in the 1980s with those for Argentina, Brazil, changes in the demand for skilled workers (Berman and Colombia, and Mexico in the 1990s.11 (We do not include Machin 2000). the data for Bolivia because the classification for sectors and Given the rising relative cost of skilled workers, table 3.1 industries was not comparable.) If there were no relation- leaves little doubt that the observed increases in demand ship between the pattern of skill upgrading in different coun- for skilled workers in Latin America are inconsistent with tries, we would expect half of these correlations to be H-O-Stolper-Samuelson explanations. This is not to say that positive, and the other half negative. Instead, 9 of the 10 trade has not had an important influence on the evolution correlations for workers with tertiary education, and (less of relative wages in Latin America. Rather, as we show below, impressively) 6 of the 10 correlations for workers with sec- trade has worked as a vehicle whereby skill-biased tech- ondary education are positive 12 Consider the implications nologies have been transferred to Latin America, rather than of this. Despite profound differences in the productive struc- through H-O-Stolper-Samuelson effects ture of the five countries, skill upgrading at the tertiary level appears to have taken place in the same sectors. There was Patterns of Skill Upgrading at the Tertiary Level particularly large skill upgrading in professional and finan- Are Similar across Different Countries in Latin America cial services, education and welfare, and government.l3 These Another piece of evidence suggestive of skill-biased tech- results are highly suggestive of technological changes that nological change comes from the similarity of patterns of affected some sectors more than others, and which were trans- skill upgrading across countries, especially among workers mitted to Chile in the 1980s and to the other countries in with tertiary education Insofar as skill-biased technologi- the region in the 1990s. TABLE 3 2 By and Large, Skill Upgrading at the Tertiary Level Occurred in the Same Sectors in Chile in the 1980s and in Other Latin American Countries in the 1990s... ARGENTINA BRAZIL CHILE COLOMBIA Brazil -0 251 Chile 0 314 0 508 Colombia 0 154 0 409 0 543 Mexico 0 115 0 315 0 322 0 298 ... while the Picture for Upgrading at the Secondary Level Varied a Great Deal More from Country to Country ARGENTINA BRAZIL CHILE COLOMBIA Brazil -0 186 Chile -0 028 0 190 Colombia -0 020 0 138 0 163 Mexico -0 376 0 085 0 611 0 291 Azote These are cross-country correlations of within-industry changes in the proportion of workers with tertiary education as a percentage of workers with tertiary and secondary edu- cation (top panel), and corresponding correlations for the proportion of workers with secondary education as a percentage of workers with secondary and primary education (bottom panel) Changes for Argentina, Brazil, Colombia and Mexico are measured between 1988/89 and 1998/99 Changes for Chile are measured between 1978/79 and 1988/89 Bolivia is not included because data by industry are nor disaggregated enough Observations are weighted by industry employment shares averaged over time and across all countries Source S5nchez-Parmo and Sclhady (2002) 61 CLOSING THE GAP Firm-level Data Show That Plants Using Foreign workers in Chile. This is highly suggestive of technology- Technology Increased the Demandfor Skilled Workers skill complementarity. Firms with access to foreign tech- in Chile... nology demanded more skilled workers, and increased the We continue our analysis of the complementary relationship share of skilled workers in the wage bill-all of this, in spite between technology and skill in Latin America by turning of their rising relative wages." The Chilean data also show to the firm-level data. These data have the advantage of often that the relationship between technology and skill varied including direct measures of technology, but the disadvan- considerably with the skill distribution of plants Specifi- tage of covering only manufacturing, and not the service cally, the impact of technology was negligible for firms that sector. Pavcnik (2002) shows that between 1979 and 1986 originally had few skilled workers, but important for skill- the relative wages, employment shares, and wage bill of intensive plants. As we discuss in the final section of the skilled workers all went up in Chile. These patterns closely chapter, similar differences can also be found across coun- mirror those observed in the labor force survey data, and pro- tries in Latin America, with new technologies flowing pri- vide further strong prima facie evidence that firms were marily to countries with higher mean levels of education increasing the relative demand for workers with tertiary edu- of their population. cation Moreover, skill upgrading in Chile took place at the same time as heavier investments and increased use of for- ... as Well as in Mexico, Colombia, and Brazil eign technology by firms (measured by foreign technical The patterns observed in the firm-level data in Chile are assistance, patent use, and import of foreign materials). The reasonably similar to those observed elsewhere in Latin results in table 3 3 show that plants that used foreign tech- America.iI The relative wages of skilled to unskilled work- nical assistance, patented technology, and imported mate- ers are higher in Mexico among firms that upgraded their rials all had a significantly higher wage bill share of skilled technology through licensing arrangements (Harrison and Hanson 1999). A background paper for this report by Mayer TABLE 3 3 and Foster (2002) shows that the returns to education were In Chile Firns That Were Exposed to New Technology higher in municipalities that had larger firms-firms that from Abroad Also Upgraded Skills are often thought to be more technologically advanced com- pared to smaller ones (see box 3.4). In Colombia, the firm- level data show increases in both the relative cost of skilled SKITL UPGRLADING DEPENDENT VARIABLE SHARE OF SKILLED REGRESSIONS, WORKERS IN THE WAGE BILL labor, and the intensity of use of skilled labor In production CHILE, 1979-86 (1) (2) in the 1980s and 1990s (Kugler 2002). In Brazil, finally, Ln (Capital/Value Added) 021** 022** firms with a highly schooled labor force were more likely to (001) (001) be exporters. Insofar as exporters are more technologically Ln (Value Added) 046** 049** sophisticated than non-exporters, this result supports the (001) (001) Foreign Technical conclusion that more skilled labor is needed to competitively Assistance Indicator 021** use sophisticated technology (Corseuil and Muendler 2002). (005) Patent Indicator 016** (002) Firms with Exposure to Foreign Technology Are Also Imported Materials Indicator 047** More Likely to Complement This with Their Own R&D (003) Foreign Technical Assistance Firms often have to adapt foreign technology to make it Cost/Value Added 196** appropriate to their particular (skill) endowments. Increased ( 063) Patent Cost/Value Added 004** competition from imports and potential access to foreign (001) markets that require higher technical sophistication also pro- Imported Materiali/Materials 087** vide strong incentives for firms both to adapt the latest avail- ( 007) R' (Adjusted) 480 480 able technology and to innovate As a result, exposure to Number of Observations 26,513 26,513 foreign technology is an important incentive for firms to Note All regress,ons include area, industry, and year indicators H-luber-White standard invest in their own R&D activities. Recent work compar- errors are in parentheses ** indicate significance at a 5 percent level Source Pavcnik (2002) ing R&D activities in Mexican firms before and after the 62 HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN The Relationship between Technology Transfer, Scale, Technological Change, and Human Capital: the Rising Demand for Skills in Latin Ameiica, Manufactunng and Development in Mexico and Patterns of Integration into the Global Economy In a seminal paper, Foster and Rosenzweig (1996) show Firms using foreign materials and technology invested in that during the Green Revolution in India the returns complementary R&D efforts, and increased their demand for to primary and secondary education increased in areas skilled workers. We now present more evidence on the effect where technological change, measured by adoption of of patterns of integration of Latin America into the global new, more productive seeds was largest and fastest, and economy on the observed changes in relative demand that those increases in turn induced private investments in education. Following a similar approach, Mayer and Integration into the Global Economy Facilitated the Foster (2002) combine municipal-level information on Transfer of Skill-biased Technologies to Latin America manufacturing firms in Mexico for 1989 and 1993 with The 1980s and 1990s were a period of increasing trade lib- demographic information from the 1990 Population eralization, rising foreign direct investment (FDI), and more Census and the 1995 Population Count to study these flexible licensing arrangements for many countries in Latin Issues In order to do this they use scale, defined as the America Figure 3.5 shows very large decreases in tariffs for average number of workers per manufacturing firm in some of the countries in our analysis-Colombia and each municipality, as a proxy for technology-that is, Argentina in the 1980s, Brazil in the early 1990s. By and larger-scale firms are assumed to be more technologically large, Chile and Mexico had liberalized tariffs earlier Figure advanced firms-after arguing that "scale and its rate of 3.6 shows dramatic increases in FDI in the mid-1990s in change are natural technological indices in manufac- Brazil and Mexico. As we discuss in box 3.5, the patterns turing." we observe in the evolution of relative wages in Latin Their main findings can then be summarized as fol- America are quite similar in character to those found in the lows First, scale appears to be highly correlated with United States and other OECD countries. It thus seems likely municipal levels of skills and net migration flows, sug- that some of the technological changes that account for the gesting that firms that use more advanced technologies rising wage of skilled workers in OECD countries, in par- do indeed locate in skill-abundant areas. Second, the ticular in the United States, were transmitted through trade, returns to human capital (education) are higher where FDI and licensing, and helped determine the evolution of scale is higher, suggesting that increased demand for relative wages in Latin America. Since the technological skills by these firms translates into higher wages for changes in the United States were complementary with skill, skilled workers. Finally, higher returns to education we would expect new technologies adopted in Latin Amer- are positively and significantly correlated with ica also to favor more educated workers. increases in schooling and school attendance, sug- gesting that individuals acquire more education when ... as Can Be Seen from an Analysis of the Relationship it pays to do so. Hence, to the extent that scale does between Changes in Demand and Changes in Import in fact proxy for technology, there seems to be evidence Penetration of a virtuous relationship between technology and To analyze the relationship between the skill upgrading and human capital. trade in Latin America, we disaggregate the data for the man- ufacturing sector in a country into industries, and pool these Source Mayer and Foster (2002) data to create an unbalanced panel. We then regress relative demand, the relative wage bill, relative wages and the rel- ative employment share on measures of import penetration, North American Free Trade Agreement (NAFTA) finds that the foreign R&D stock, and a decomposition of this foreign the export orientation of a firm and the tariff levels in a sector R&D stock into a measure of volume and one of composi- were both important determinants of a firm's own R&D tion of the trade flows. (We focus on the comparison between efforts in 1999, but not in 1992 (Meza and Mora 2001) workers with tertiary education and those with secondary 63 CLOSING THE GAP FIGURE 3.5 Tariffs Decreased and Import Penetration Increased in a Number of Latin American Countries in the 1980s and 1990s Aogoarioo. . ~~~~~~~~~~~~~~~~~~~~~~~BoOivi8 "0 40 12 0 1 30.00 dO ,20 30.00 I D D40.00 40~~~~~~~~~~~~~~~~~~~~~~~~~~4 190 199 19 -ss 193 19- 199 -99 -9 199S 2091S0 45.09S0ss 1S .49 199 - 99 - 3 -s 2200 0 35,~~~~~~.*= ,,,, -................ - -- - 20. '0 .....,-"" 9 4b 30 \ S0 *4 8 5 X 320 .0 -~~~~' '-' ; 4 -/ \ _ _ .I X .0 0 b S ~ , . . 4 ,-'.e 20 8.0 10 . o 2o o..00 19S 1"- 19S 'gm 192. 19 Mu 19R 19f 1 99. 2- s o M2 .qs I 6 99 Iq, 1q2 .k qdu q Y.. Dtbs Mar3co.s-. 2. 30 .340 20 200 0 6o. . \ , . ., . 40 12 D9 . ~\ .. , ,"' . .4000 20 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~200.00 10. . - 20 5.0, 0 0.00 5~~~~~~~~~~~~~~~~~~~~.00 2000 2082 1984 9806 196S boo 1boa oog4 19oo3 too. ooo 2080 2082 2904 2066 lOSS 000 1002 2004 1906 2008 2000 Yto of Soroy YtrofSO,Stvy BoooO thus~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~loo 20 p40 25000 . . . . . . . 5~~~~~~~~~~~~~~~~~~~~~20.0, .1 2~~ ~~~~~~~~~~~~~~~~~0 to0 500 . . . . . . . . . . . . . . . . . .0~* 64~~~~~~~~~~~. c D ~~~~~~~~~~~~~~~~~~~~~~~~0 4w 0 120 60n ID - - - - - - - - Do~~~~~~~~~~~~~~~~~ 5.0 3.~~~~ C 3.0 20 40 u to~~~~COOh. 0000 00 200 10 200 0 D~~~~~~~~~~~~~0 1900 0920 96 9 00 tN M t D 92 'm .m 9 u .94 1 6 t9 20 . . ~~~~~ ~~~~~~ ~~. .. 40 d- -i-9b.GD0 - -. 0 ~ ~ _rP-i.18010 Sore:GPdt20mteWrdDveomn niaos h ol ak Tr0d a0 taif 20afo h eerhGop;TeWrdBn uigWO IDBCDROMdaabae;an TrdePolcyReiewContt Rpot, arousisue, 190200; NCAD,Hadbok f rad Cntrl eaure oDev0oin Sounrces:D daplmeta from theeWorld Devlpopmet Rndimatos, The4 World Bank. Trade andic RefrminDataelompthe Researchs Group; The5 Worl Bankc(usiongPpe WT67, 1994; The Uruguay Round: Statistics on Tariffs Concessions, Given and Received, 1996 and World Development Indicators, 1998-2000; OECD, Indicators of Tariff and Non-Ta riff Trade Barriers, 1996; IDB, Statistics and Quantitative Analysis Data, 1998. 64 HOW TECHNOLOGY AND SKILLS INTERACT: THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN FIGURE 3.6 FDI Flows Increased Significantl in a Number of Latn American Countries Argentina Bolivi- 6l hss 0s2 1s s6 ls sn 19 94 I6 Is 0 z etX-t E 3 61.00 4s sz Iu lf ss 90 l9 94 19 I s a 00- 4".'-. *'',8~~~~~~~2. f~~~~~~~~~~~~~~~~8 obbi ' dic o b oe o~# o. 0- : ¶ :1,390 0 3984 1900 ioeC 1 90 89 994 1996 090s 200 1930 '00 IOU 8986 1988 1090 1992 1994 3996 I899 2000 YearefSry Y-ofs-y Brazil chile 30 20~~~~~~~~~~~~~~~~~~~~~~~~~~~~~o~ 30 8 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~20 25 -~~~~~~~~~~4 -411 Por1s dirsx inesme0 ns, in0o (6 le leDPI 3~~~~~~~~~~~~~~~~~Gl as3 m>o 6oGP 0 g0 *0 1 6 2' ,t.,--.~~~~~~~~~~~~~~~~~~~~~~~~~~~2 2 304 20 20~ 8 6 9 I~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~' a 4 2 a~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~ A n -.J-.. 8980 8982 3984 8980 3908 1990 8992 8994 3996 3999 000 390 18 94 18 9 "D '9 94 19 98 2 Year o Survy Vearfourvy 0 ~~~~~~~~~~~~~~~~~~~~~~~65 CLOSING THE GAP F BOX 3.5 Evidence of Skill-Biased Technolocal Change in the United States and other OECD Countries There is substantial evidence for the United States and Autor, Katz, and Krueger (1998) study the same phe- other OECD countries that the increases in the relative nomenon using data on computerization as a proxy for wages of more educated workers observed in the last 25 skill-biased technological change. They find evidence that years are to a large extent associated with skill-biased tech- the growth of demand for more educated workers is asso- nological change (SBTC). This box reviews this evidence, ciated with rapid skill upgrading within detailed indus- | paying special attention to those papers whose approach tries, and that this upgrading is greater in more and methodology we follow in this chapter to assess the computer-intensive industries. I extent of SBTC in Latin America. Berman, Bound, and Machin (1998) and Machin and The literature on SBTC and its effect on the demand Van Reenen (1998) discuss the extent to which there is for skilled workers has followed two approaches. First, esti- evidence of SBTC in countries other than the United mates of changes in the relative demand for skilled work- States. The first paper investigates whether R&D inten- ers are often constructed using a simple model of (relative) sity is related to increases in the demand for more skilled supply and demand for skilled and unskilled labor, where workers in the United States and other OECD countries, relative wage changes that cannot be explained by changes and concludes that SBTC is an international phenome- in relative supply are assumed to be associated with changes non. The second paper analyzes within-industry changes in relative demand (see box 3.3). Second, the degree to in the wage-bill for skilled workers. The authors argue which skill upgrading has taken place within sectors and that not only did the proportion of skilled workers within firms is then used to assess the extent to which these increase in most industries in most countries, but the changes in relative wages are associated with SBTC. magnitude of these increases was also correlated across Katz and Murphy (1992) apply the demand-and-supply countries. Specifically, increases in the demand for skilled framework described above to study changes in the struc- workers were particularly large in capital-intensive ture of wages in the United States during 1963-87. They industries, and in industries with the highest degree of report that increases in the wages of more educated and computer usage. more experienced workers are driven by changes in the demand for these workers associated with skill-biased tech- Source: Katz and Murphy (1992); Autor, Katz, and Krueger (1998); Berman, nological change. Bound, and Machin (1998); Machin and Van Reenen (1998). education, as this is where we observe the clearest patterns cell, lagged by one period. The results are summarized in in relative wages, employment shares, and demand.) table 3.4. Data on imports, and on the R&D content of these In the first panel, the explanatory variable is import imports are only available for three countries in our sample- penetration-where import penetration is defined as total Chile, Colombia, and Mexico-and only for the manufac- imports in an industry over value added. Table 3.4 shows that turing sector. Moreover, since all our series are trending increases in import penetration are consistently associated upwards, we have to de-trend the data to avoid a spurious with increases in skill upgrading-regardless of whether this correlation between our measures of relative demand, or rel- upgrading is measured in terms of relative wages, demand, ative wages, or relative employment, and the foreign R&D wage bill, or employment share of workers with tertiary edu- stock. We do this by removing a country-industry specific cation. In the second panel, the explanatory variable is a linear time trend from all variables. We then regress the measure of the foreign R&D stock available in an industry- dependent variable (relative wages, relative demand, the country-year cell. This measure, developed by Schiff, Wang, wage bill, or the employment share) in industry i in coun- and Olarreaga (2002) can be thought of as a measure of try c at time t on the measure of imports, foreign R&D, or import penetration by industry, when the imports are the decomposition of the imports in this industry-country weighted by their R&D content.'6 Once again, the results 66 HOW TECIHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN TABLE 3 4 There Is More Skill Upgrading, Measured by the Relative Wages, Relative Employment Shares, and Relative Demand for Tertiary Workers, in Countries and Industries That Have Higher Import Penetration, Especially of Imports That Are Intensive in R&D EMPLOYMENT EMPLOYMENT RELATIVE WAGES RELATIVE DEMAND WAGE BILL SHARE (HOURS) SHARE (WORKERS) MODEL I Imports (% VA) 0 049... 0 087- 0 018os. 0 006"' 0 005" (0 011) (0 023) (0 006) (0 002) (0 003) N 375 375 375 375 375 Prob > Ch12 0 000 0 000 0 000 0 000 0 000 MODEL 2 Foreign R&D stock 0 138--- 0282"' 0051"' 0011" 009- (0 012) (0 027) (0 007) (0 003) (0 003) N 375 375 375 375 375 Prob > Ch12 0 000 0 000 0 000 0 000 0 000 MODEL 3 Trade volume (imports) 0 041- 0 065'-- 0 016--- 0 002 0 001 (0 011) (O 023) ( 0006) (O 002) (O 002) Trade composition (R&D index) 0044--- 0119- 0015- 0019- 0021- (0 012) (0 028) (0 007) (0 003) (0 003) N 375 375 375 375 375 Prob > Chi2 0 000 0 000 0 000 0 000 0 000 Source S6nchez-Pgramo and Schady (2002) show that increases in foreign R&D stock are consistently background paper commissioned for this report, Pavcnsk related to increases in skill upgrading within industries. and others (2002) show that there is also evidence of sector- Moreover, the relatively larger magnitude of the coefficients specific skill-biased technological change transferred through suggests that the composition of imports, not lust the volume, trade (box 3.6). is important In the third panel, we disaggregate changes in the foreign R&D stock in an industry into changes in volume The Level of Skill in a Country Is an Important and changes in composition (increases in the R&D content of Determinant of Technology Transferred imports, holding constant the volume), and regress skill from Abroad upgrading on both of these measures.17 The results show that Up to this point, this chapter has focused on the extent to both measures, but particularly the measure of trade com- which technology has determined the demand for workers position, are important determinants of skill upgrading in with more or less education in Latin America. We now con- Chile, Colombia, and Mexico. Put differently, it is not just sider the reverse question: Can low levels of skill be a con- the amount of trade but whom you trade with that matters- straint to technology transfer through trade and FDI7 We so that imports from countries and industries that are inten- find a great deal of evidence that the answer to this ques- sive in R&D are particularly important as an impetus for skill tion is "yes," suggesting a circular relationship between skill upgrading.18 Earlier work by Machin and Van Reenen (1998) upgrading and technology transfer. Technological innova- shows that there is a strong association between industry-level tion increases the demand for skilled workers and, as we will R&D intensity and the wage bill for skilled workers in a show in chapter 4, can result in increases in educational sample of OECD countries (including Denmark, France, attainment. This, in turn, can further stimulate the demand Germany, Japan, Sweden, the United States, and the United for new technology by firms. Some countries may therefore Kingdom) The results in table 3 4 are consistent with a pat- enter a virtuous cycle of upgrading of both technology and tern whereby foreign R&D, transmitted through trade, skill-a cycle that results in higher productivity and higher increases the relative wages, employment shares, and demand growth rates Countries with very low skill levels, by con- for workers with tertiary education in Latin America In a trast, may be unable to attract sufficient levels of FDI, and 67 CLOSING THE GAP BOX 3 6 Trade Liberalization and Sector-Spaclfic Skill-Biased Technological Change in ova&A I Recent work by Pavcnik and others (2002) investigates the increase in the wage premium paid to educated workers. relationship between trade liberalization and the relative These are comparatively large effects, as more than half the wages of workers with different amounits of education in industries in the sample cut tariffs by more than 50 points Brazil, a country that experienced a profound reduction between 1987 and 1998 in Brazil. in tariffs in the late 1980s, Because of differences in tariff What explains these surprising findings? First, they levels across industries before the reform, this reduction point to rigidities in the Brazilian labor market: if edu- in tariffs affected some industries more than others, and cated workers were perfectly mobile across industries, we at different times, which makes it possible to convincingly would expect them to switch out of the industries with estimate the effect of changes in import penetration or tar- falling wage premia for skill and into those with rising iffs on changes in wages across industries. premia. Second, the results are consistent with a pattern In Brazil, as in many other countries, workers in cer- of sector-specific skill-biased technological change. There tain industries are paid a wage premium that cannot be is considerable evidence from around the world that tech- explained by their education or experience. The underly- nological innovations from a "leader" country transmitted ing reasons for this premium are not clear, although it may through trade to a "follower" country have the largest effect be related to differences across industries in working con- on productivity in firms that are in the same industry in ditions, nonwage compensation, or unionization affecting both countries (see, for example, Keller 2002). It there- relative wages. In addition, there is a wage premium to fore seems plausible that in Brazil new technologies were skill that varies across industries: highly educated work- adopted primarily in those industries that faced the biggest ers are paid more in some industries than in others, and reductions in tariffs. These technologies likely had two these patterns do not necessarily correspond to the patterns impacts on firms in these industries: they raised produc- in overall industry wage premia. How did trade liberal- tivity, and some of these productivity increases were passed ization affect the industry-specific wage premium to skill? on to workers in the form of higher wages. However, Pavcnik and others show that while the overall industry because new technologies were skill-biased, they particu- wage premium was not affected by trade liberalization, the larly increased the productivity and wages of skilled work- industry wage premium for skill was significantly related ers, rather than of the average worker in an industry. to reductions in tariffs. Specifically, a one percentage point reduction in tariffs is associated with about a 0.2 percent Soarce: Pavcn,k and others (2002) the amount of technology transferred through trade may also Differences in the Degree of Computer Penetration be low. Moreover, what little technology is transferred may between Latin America and East Asia Are Explained be quite inappropriate because it was developed for the by Differences in Education and Trade Exposure... highly skill-intensive workforce of the "leader" countries As We begin our discussion of the effect of human capital on a result, expected productivity improvements may not technology transfer with an analysis of the relationship materialize (Acemoglu and Zilibotti 2001). These countries between computer penetration, trade, and human capital. To may also be unable to generate technological upgrading do this, we return to the data from Caselli and Coleman we through domestic R&D, as this is by its very nature a skill- referred to in chapter 2. The main finding of Caselli and intensive process In the absence of policy reforms, countries Coleman's analysis is that computer adoption is strongly asso- with very low levels of skill may therefore be trapped in a ciated with high levels of human capital (even after con- vicious cycle of little technology transfer, no domestic inno- trolling for differences across countries in income per vation, low returns to education, low productivity, and stag- worker), and with manufacturing trade with the OECD. To nant growth prospects. estimate the extent to which these two variables can explain 68 HOW TECHNOLOGY AND SKILLS INTERAC'I rHE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN differences in the pattern of computer penetration between results show that differences in computer penetration Latin America and East Asia, we perform a simple calcula- within Latin America are explained by differences in two tion: We predict what the change in computer penetration factors, and two factors only. First, the level of education, in Latin America between 1980 and 1990 would have been specifically the fraction of the population with primary school if Latin America had had the mean changes in education and and (especially) secondary school education. At the means manufacturing trade with the OECD observed in East Asia. of the other variables, a one-standard deviation (13 5 per- The results from these calculations suggest that computer centage point) increase in primary education increases com- penetration would have increased by a greater amount in Latin puter imports by about 39 percent, while a one-standard America than it did in East Asia during the 1980-90 period deviation (10.5 percentage point) increase in secondary edu- if Latin America had had the East Asian patterns of skill cation almost doubles computer penetration (an increase of upgrading and trade.i9 95 percent). These effects are much larger than those for the world as a whole, suggesting that primary and secondary ... as Can Differences within Latin America schooling levels are better predictors of the adoption of new In chapter 2 we showed that computer imports within Latin technologies in Latin America than elsewhere The second America varied a great deal-even for countries with the variable that is a significant predictor of computer penetra- same level of income. For example, the value of computer tion is the amount of trading in the manufacturing sectors imports per worker in 1990 was more than twice as high in with OECD countries. Countries in Latin America that trade Panama as in Ecuador, despite their comparable per capita with the technological leaders, especially in manufacturing, incomes, while the level of computer penetration in Brazil have significantly higher propensities to upgrade their tech- was the same as in Honduras, despite the very large income nologies. This effect is also larger in the Latin American differences.20 sample than in the global sample. At the means of the other What explains these differing patterns of computer pen- variables, a one-standard deviation increase in manufactur- etration within Latin America? Table 3.5 reproduces some ing imports from the OECD increases the degree of com- of the regressions in the Caselli and Coleman study, when puter penetration in Latin America by about 90 percent 21 the sample is limited to countries in Latin America. The None of the other variables-the level of income or investment TABLE 3 5 Countries in Latin America with More Skilled Workers Are Better Prepared to Adopt New Technologies DEPENDENT VARIABLE LOG OF COMPUTER IMPORTS PER WORKER, 1970-90 COEFFICIENT STANDARD ERROR Log income per worker 0 350 0 437 Log investment per worker 0 269 0 337 Fraction with primary education 0 025* 0 012 Fraction with secondary education 0.063** 0.016 Fraction with tertiary education 0 004 0 035 Log manufacturing imports from OECD per worker 0 636* 0 303 Log non-manufacturing imports from OECD per worker -0 117 0 189 Log manufacturing imports from non-OECD per worker -0 192 0 341 Log non-manufacturing imports from non-OECD per worker -0 125 0 177 Fraction speaking English -0 997 0 585 Property rights (1-10) 0 098 0 075 Agriculture share in GDP -0 013 0 014 Manufacturing share in GDP 0 004 0 016 Government spending share in GDP -0 015 0 018 R-squared 0 772 Number of countries 23 Number of observations 82 Note All regressions include year fixed effeccs Estimation rechnique is Random Effects (RE) * significant at the 5 percent level, ** significant at the 1 percent level Source Schady (2002), based on data from Caselli and Coleman (2001) 69 CLOSING THE GAP per worker, measures of the composition of the economy, the ability to rapidly absorb new technology transferred property rights, and trade other than manufacturing trade through trade, FDI, and licensing. with OECD countries-is a significant predictor of tech- nological upgrading through increased computer imports in ... with Chile Moving a Decade or More before Other Latin America during the 1970-90 period. Latin American Countries In this context, the case of Chile is instructive. Differences There Is Also a Complementary Relationship Among in the patterns of relative demand, technological change, FDI, Trade, and Human Capital integration in the world economy and productivity between The analysis of computer penetration shows that high- Chile, on the one hand, and other countries in Latin America, technology imports flow primarily to countries with high on the other, are noteworthy as Chile undertook many levels of secondary school attainment. Empirical work on FDI "Washington Consensus" type reforms well before other finds a similar pattern Technologies transferred through FDI countries in the region. Of the six countries we analyzed increase productivity in the host country only in cases involv- extensively with the labor force survey data, an "ad hoc" ing countries where the mean level of education is about two ranking of the degree of global integration and structural years of secondary school.22 Other studies have found a strong transformation of the economy would put Chile highest, then (positive) interaction between imports of machinery and sec- Mexico, followed by Argentina, Bolivia, Brazil, and Colombia ondary school enrollment in determining growth rates.23 (these last four, in no particular order). This ranking of coun- There is also evidence specific to Latin America suggesting tries by macroeconomic reforms-reforms that should have that FDI is directed primarily at countries and regions with a critical influence on the degree to which countries are able high levels of human capital. For every additional year of to adopt technologies from abroad, for example-can be schooling in 1985, countries in Latin America received mapped reasonably well onto the series of changes in rela- almost half a percentage point more FDI as a fraction of GDP tive demand for workers with tertiary education for the five during the 1985-98 period In Mexico, states with higher countries in our sample. We pursue this line of thought in levels of human capital receive more FDI-even after figure 3 7, in which we superimpose the demand series for controlling for distance to the United States border (de Fer- co= 2 for Argentina, Bolivia, Brazil, Colombia, and Mexico ranti and others 2002, pp. 93-103). The maquiladoras in on the longer Chilean series. Figure 3.7 suggests that coun- northern Mexico and the Dominican Republic employ work- tries in Latin America are at different stages in a long-term ers with higher mean levels of education than the popula- cycle of changing relative demand for workers with differ- tion average, especially women with secondary schooling ent amounts of skill. (The "years" on the x-axis are not the Cunningham (2002). Finally, trade with developed countries same across countries, precisely because our hypothesis is that has a particularly large impact on TFP in Latin America different countries are at different stages in this cycle of rel- when trade takes place in R&D-intensive sectors in the ative demand.) This is more apparent if we visually "smooth exporting country, and when there is a high degree of human out" deviations from long-term trends that are caused by capital in the recipient country (Schiff and Wang 2002). macroeconomic crises. Understanding long-term patterns in relative demand, Long-Tewnn CycWy c n3 the Demand for $SMed productivity, and growth has profound implications for the Worhers and lfechndoWo ? development of countries in the region In the early 1970s, Chile started off with reasonably low education and tech- The Experiences of Different Countries in Latin nology levels, limited access to international technology, and America over the Last 20 Years Were "Determined" low returns to education. Between 1975 and 1990, it opened by Their Potential Capacity to Receive and Effectively up to trade and FDI, both of which served to transmit tech- Use New Technologies ... nology from the "leader" countries As this technology was Because new technologies are complementary with skills, the applied by firms, these increased their demand for skilled initial levels of technology and education, as well as the speed workers and the wages they were willing to pay them. In with which these levels changed over time, have played an consequence, wage inequality increased-temporarily important role in determining the returns to education, and (table 3.6). In response to the sharp wage differentials 70 HOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND I HE CARIBBEAN FIGURE 3 7 The Long-Term Cycles in Relative Demand in Latin America (ar= 21 Implied Relative Demand Time Series - Tertiary and Secondary Education 12 0 8~~~~~~~- ° 0 6 o44 TABLE~~~~ Memc 0 06 1980-85 ~ ~ 9 1 345 9 l05'21'4i5 16 174892 2 22 4 226728930 3 392334 i-0 1985-90 ~ 049 055 Ch0 07 35 1990-95 ~ ~ 02 377 0- 4560 A70469 nti2a40 1995-99 ~ ~ ~7 9 3088 12 1354 10 1647 18 1901234256272829303132333 Years Source Senchez-Phramo and Schady (2002) TABLE 3 6 Wage Inequality Increased with Increases in Demand for Skilled Workers ARGENTINA BOLrVIA BRAZIL CHILE COLOMBIA MEXICO GINI COEFFICIENT OF MONTHLY WAGES (FIVE-YEAR AVERAGES) 1965-70 0 447 1970-75 0 419 1975-80 0 453 1980-85 0 545 0s484 0 392 1985-90 0p396 0w559 0g504 0 371 0 352 1990-95 0 377 0 456 0 557 0 469 0 402 0 403 1995-99 0 388 0 458 0 548 0 461 0 416 0 430 Source Sgnchez-Piramo and Schady (2002), based on household and labor force survey data from each country associated with education, young men and women in Chile educated workforce, a much larger fraction of the workforce acquired more education before entering the labor market, can benefit from the high returns to educatson than before Mean education levels increased and this, in turn, kept the Where does the cycle go from hered Although Chile was returns to education from increasing further Productivity able to import much of the technology it needed in the 1980s also increased substantiallys Between 1985 and 2000 value and early 1990s from abroad, it may now need to place an added per worker grew by 71 percent. Within a 20-year increasing amount of effort into developing its domestic period, Chile made the transition from a country that was technological capacity. Undoubtedly, Chile is closer to the arguably in a low technology-low education equilibrium to world's technological frontier now than it was 20 years ago- one at the medium technology-high education equilib- even though the frontier has itself shifted farther out. Indeed, rium-at least by regional standards. Indeed, by 2000 it had in some sectors, Chile may itself need to push out the fron- both the highest educational attainment and the highest tier in order to stay competitive. These distinctions between degree of technological sophistication in the region.24 This countries that are more or less far away from the global tech- transformation helps explain the increases in firm produc- nological leaders have important implications for the tivity, the steady growth, and the recent improvements in design of policies for domestic innovation structures We wage inequality in Chile. They also explain why, with a more return to these in chapters 6 and 7. 71 CLOSING THE GAP (oncDuslol$ ondary and tertiary education, if they are to remain as com- In this chapter we have made a number of points about the peritive players in global markets. Chapter 4 discusses dif- relationship between technology and skills in Latin America. ferent policy interventions aimed toward achieving these The most important include the following- goals, as well as some successful experiences in the region that can serve as guidelines for the reformers to come. o There has been an increase in the demand for skills in Latin America, in particular the demand for workers Edo with university education. This increase in demand has 1 This section is based on Sanchez-P§ramo and Schady (2002) not been matched by equivalent increases in the supply Since our sample sizes are relatively small, in particular in Argentina, of highly educated workers, so that overall relative Bolivia, Chile, and Colombia, and especially when we consider work- wages of workers with tertiary education have gone up ers with university education, we graph three-year moving averages relative to those of workers with secondary education. for the values of relative supply and wages in a given year 2. For example, the fact that the demand series for CT= 3 for Colom- bia is steeper for the 1990-93 period than for the 1993-96 period pro- of the transfer of skill-biased technologies to Latin vides evidence that the relative demand for workers with tertiary America. education grew more quickly in the first three years of the decade. Sim- o Improved access to these technologies, as well as grow- ilarly, the fact that the demand series for a = 3 for Colombia for the ing competitive pressure to use them, can be linked 1990-93 period is steeper than the corresponding demand series for Argentina, Bolivia, Brazil, Chile, and Mexico shows that demand grew to trade opening and new FDI flows. However, new bymrm ombatnelwhe by more in Colombia than elsewhere technologies were transferred primarily to countries 3 Sanchez-Paramo and Schady (2002) calculate changes in a with reasonably "high" levels of human capital. manner comparable to those found in Autor, Katz, and Krueger (1998) o Many of the changes observed in Argentina, Bolivia, 4 For one thing, the degree of skill-abundance varies a great deal Colombia, Mexico, and (less clearly) Brazil in the from country to country. Insofar as most trade of most Latin American 1990s closely resemble the experience of Chile in the countries is with the United States or other OECD countries that are relatively more skill-abundant, we would expect trade reform to lead 1980s. As we showed in chapter 2, Brazil has by far the lwsleeofsilofnoftecutito a decrease in the wage premium associated with education in Latin the lowest level of skills of any of the countries in our America. sample. This may be the reason why skill-biased tech- 5. On Brazil, see Pavcnik and others (2002); on Colombia, Attanasio, nologies have not been adopted to a significant degree Goldberg, and Pavcnik (2002), on Mexico, Hanson and Harrison in Brazil, and may be a plausible explanation for the (1999) differences between the Brazilian and (more general) 6 In Brazil, for example, an industry with 10 percentage points more unskilled labor could expect to see, on average, a four point bigger reduction in tanffs between 1987 and 1998 A regression of the annual change in tariffs against the share of unskilled workers in 1987 yields The Relationship between Technology and Skills and a coefficient of -3.97 for the share of unskilled workers (t-statistic = the Patterns It Generates Has Important Implications -1.88) See Pavcnik and others (2002, p 13). for Education Policy 7 As proposed by Berman, Bound, and Macbin (1998), changes inthe wage bill share of workers with tertiary education are decom- Countries need a broad base of workers with secondary edu- posed as follows- (1) ASn= YASn,W,+XAW Sn, where Sn is the cation to be able to access new technologies. Given that most I I I I countries in the region exhibit a large gap at the secondary wage bill share of tertiary (skilled) workers defined as Sn = S level, substantial improvements in secondary enrollment and w graduation rates will be requwred in order to create such a and WI is the wage bill share of industry t defined as = WB, , base. Also, new technologies appear to be complementary wWoe with skills, in particular with tertiary education. Although w ith secondary education (or unskilled), W IS wages of workers with countries in the region do not exhibit as large a gap at the tertiary education, W, is wages of workers with secondary education, tertiary level as they do at the secondary level, the job is far WB is the wage bill, i is an industry index and an overstrike indicates from done. As a result, Latin American countries will have an average between the initial and final periods considered Industry wage bill shares are then used as industry-specific weights in (1), so to concentrate heavily on improving the education level of that the first term captures within-industry changes in the wage bill their population, increasing the number of workers with sec- share and the second term captures between-industry changes In other 72 HIOW TECHNOLOGY AND SKILLS INTERACT THE EVIDENCE FOR LATIN AMERICA AND THE CARIBBEAN words, the first terms holds industry-specific weights constant and the 18 These results are from Sanchez-Paramo and Schady (2002) second term holds within-industry wage bill shares constant 19. Specifically, we regress the difference sn the log of computer 8 See Auror, Karz, and Krueger (1998), Katz and Autor (1999), penetration between 1980 and 1990 on changes in the education van- Berman, Bound, and Machin (1998), and Berman and Machin (2000) ables between 1970 and 1980 and changes in manufacturing trade with 9 Similar estimates based on firm-level data for Chile for the the OECD between 1980 and 1990 for the sample of Latin American 1979-86 period suggest that 66 percent of the changes in the wage bill countries, and then multiply the coefficients from this regression by occurred within very detailed four-digit ISIC industries (Pavcnik 2002). the mean change in these covariates in East Asian countries over the 10 Calculations based on a United Nations classification of indus- period. (Changes in human capital are lagged to limit the extent to trial employment into "non-production" (skilled) and "production" which the changes in education could themselves have been caused (unskilled) workers for the 1980s show a similar pattern for the six Latin by changes in technological penetration ) This calculation, which is American countries for which data are available The proportion of the similar in character to the standard Oaxaca-Blinder decomposition change in the wage bill of non-production workers that takes place used by labor economists, gives a sense of the changes in computer within industries is 51 percent in Uruguay, 62 percent in Venezuela, penetration we would have expected to see in Latin America if there 69 percent in Guatemala, 84 percent in Colombia, 103 percent in Peru, had been changes in education and trade of the magnitude of and 153 percent in Chile See Berman and Machin (2000, p 29) those observed in East Asia The results should be treated with cau- 11 Observations are weighted by the sector employment shares tion, as the regression in changes only has 22 Latin American obser- averaged over time and across countries, in a similar fashion to the vations, while the mean changes in East Asia are an average of only 9 correlations for industrialized countries presented in Berman, Bound, observations and Machin (1998) 20 According to these data, Jamaica, Panama, Chile, and Costa 12 The fact that none of the correlation coefficients is significant Rica are all positive outliers within the Latin America region, having is probably not surprising given that we only have 11 comparable sec- higher levels of computer penetration than would be predicted for their tors across countries (regional) income levels 13 To identify the sectors most affected by skill upgrading at the 21. Note that, as Caselli and Coleman point out, the signifi- tertiary level, we pooled the data for Chile in the 1980s with the data cance of this result is not driven by the inclusion of computers in for other countries in the 1990s, and regressed the within-sector change the measure of manufacturing imports, as computing equipment in the employment share on a set of country and sector indicator varn- is a miniscule fraction of total trade with both OECD and non- ables See Berman, Bound, and Machin (1998) for a similar approach OECD partners Also, Caselli and Coleman (p.7) point out that for 14 Pavcnik's results for Chile are not robust to the inclusion of the global sample the effect of manufacturing imports from the plant fixed effects, raising the possibility that some unobserved dif- OECD remains significant even after a host of related variables- ferences across plants explain both skill upgrading and technology bilateral-trade-weighted measures of distance from the leading adoption However, it is quite possible that measurement error and world exporters, measures of FDI inflows, the black market pre- the comparatively small number of years in the plant-level panel are mium, and the Sachs-Warner openness measure-are included in responsible for the lack of significance of the technology measures in the regression. Moreover, none of these additional variables is itself the plants fixed-effects regressions a significant predictor of computer penetration. Unsurprisingly, 15 Firm-level data for Mexico between 1992 and 1999 also show given the obvious time trend in computer penetration, all of the that firms that adopted new technology paid higher wages to work- year dummies are highly significant Still, the results in table 3 5 ers from all skill groups, but especially semi-skilled workers The Mex- are remarkable, as much as anything else, when we consider all of ican data also show that there is a positive, significant correlation the variables that do not have an effect on computer penetration in between a firm-level measure of Total Factor Productivity (TFP) and Latin America the employment share of high-skilled workers See Lopez-Acevedo 22 Xu (2000) Earlier work by Borensztein, De Gregario, and Lee (2002) (1998, p 126) reaches similar conclusions The authors write that there 16 Schiff, Wang, and Olarreaga (2002) define North-foreign R&D are "strong complementary effects between FDI and human capital on in industry i of developing country c, as NRDa, = YaIRD, = the growth rate of income," a result that is "consistent with the idea that the flow of advanced technology brought along by FDI can increase Mc)R,] where c-(k) indexes developing (OEICD) coun- the growth rate of the host economy only by interacting with that k VA country's absorptive capability," although the threshold level of sec- tries,) indexes industries, M, VA and RD denote imports, value added, ondary school in Borensztein, De Gregario, and Lee (1998) is some- and R&D, respectively, and a" is the input-output coefficient (which what lower than that estimated by Xu. measures the share of imports in induscryy that is sold to industry i 23 Romer (1993), cited in Borensztein, De Gregario, and Lee in country c) Data on import input-output flows are not available, so (1998, p. 127) they are proxied by domestic input-output flows in the estimation 24 Chile had the largest amount of private R&D and the largest Sales of imports for final consumption are excluded from the estimation number of computers available per worker of any country in the region, 17 This methodology closely follows that in Coe, Helpman, and and was second-highest in the number of patents registered in the Hoffmaister (1997) United States, after Argentina 73 CHAPTER 4 Closing the Skills Gap: Education Policies I N EARLIER CHAPTERS WE ARGUED THAT LATIN AMERICA HAS A PROFOUND GAP in skills. Latin American adults have almost one-and-a-half fewer years of education than would be expected for their levels of per capita income, and enrollment rates at the secondary and tertiary levels are both low as well. Moreover, the quality of Latin American education is deficient, and students from the region (with the exception of those from Cuba) do poorly on international tests. Chapter 3 showed that technological upgrading, such as that which is transmitted through trade and FDI, is a key reason why firms have increased the demand for skilled workers, and have bid up their rela- tive wages. We also argue that policies enabling young men and women in Latin America to meet this increase in demand by seeking more schooling should be a key objective of government policy. In this chapter, we consider policies to close the skill gap in Latin America. The chapter moves from an analysis of the labor market, which determines the relative wages of individuals with differ- ent amounts of schooling, to an analysis of the education market. Does the education system in a coun- try produce an adequate amount and mix of skilled workers? Specifically, the chapter considers two sets of issues. First, we consider the speed and distrzbution of skill upgrading for countries in the region. Second, we discuss appropriate policies for secondary school and for the university system. We leave aside concerns about primary education-both because most Latin This means that skill upgrading should have a larger American countries do nor appear to have a deficit in pri- effect on growth rates now than before. Positive exter- mary enrollment for their income level, and because skill- nalities associated with education may further augment biased technological change in the region has particularly the impact of schooling on growth raised the demand for workers with secondary and tertiary * There is very limited potential for "leapfrogging" edu- education.1 cation levels The most successful episodes of educa- The key messages of this chapter are three. tional upgrading-the United States between 1850 and 1950, Korea and the Scandinavian countries since * Latin America needs to close the skills gap quickly if it World War II-have all followed a pattern of bottom- is to be competitive in the global marketplace. New up upgrading, first building up basic education, then technologies have particularly increased the produc- secondary education, and finally university A balanced tivity of educated workers, and employers are passing education transition ensures that the most gifted stu- on some of these productivity increases in the form of dents go to university, is likely to minimize inequal- higher wages There is a direct link between this rela- ity, and takes advantage of the complementarities tionship of education and wages at the individual level, between the phases in an educational transition and and the relationship of education and aggregate those in the technological transition of a country. A income at the country level (Krueger and Lindahl 2001). number of Latin American countries appear to have 75 CLOSING THE GAP been following "unbalanced" educational transi- look like a pyramid. The base, corresponding to the tions-increasing the coverage of university without fraction with no schooling or with some primary ensuring a large pool of secondary school graduates. schooling only, is wide. The middle section, corre- Expanding the coverage of secondary school is there- sponding to the fraction with some secondary school- fore the biggest priority for most countries in the ing, is thin but fatter than the top, which corresponds region However, policies that enable students to to adults with some tertiary education. Virtually all attend university are also important, both because of the countries we consider in this report found them- there are high returns to tertiary education, and selves at this phase in 1960. because the option value of going to university is one o As they moved from the first stage of the transition of the main incentives for completing secondary to the second, some countries upgraded from the school. bottom up. These countries gradually increased the o Both the quality of education and the nature of the proportion of adults with some secondary schooling skills learned in school have important consequences by decreasing the number who had no schooling or for wages and productivity Higher quality education only primary education. The resulting distribution still will make workers more productive, increase the rate looks like a pyramid-albeit one with a much thin- of return to education, and provide an incentive for fur- ner base and a much fatter middle. This stage corre- ther private investments in education. In a rapidly sponds to East Asian tigers and Finland in 1980, and changing environment, schools and universities should to many Latin American countries in 2000. teach general, problem-solving skills that enable o When countries move from the second phase of the young men and women to "learn how to learn" transition to the third, so much mass is shifted from throughout their lives. the base to the middle that the distribution resembles a diamond. More adults have some secondary school- The rest of the chapter proceeds as follows. We first con- ing than those with only some primary schooling or sider the evolution of educational attainment in the region, less. Finland, Norway, Sweden, and all of the East focusing in particular on countries with different transition Asian tigers except Malaysia (but none of the Latin paths. The second half of the chapter then applies a supply- American countries) have diamond distributions. demand framework to the analysis of the secondary school a Some Latin American countries, including Costa Rica, and university systems. In the final section we discuss policy the Dominican Republic, Ecuador, El Salvador, and implications. Venezuela, as well as Thailand in East Asia, have evolved along a different path. These countries Iducatlonab ransMOIDn: Tle Noviw "squeezed from the middle" in moving beyond the first phase: The base got thinner and the top wider, while Educational Transitions: Pyramids, Anvils, the middle was relatively unaffected. The resulting and Diamonds distribution resembles a lopsided anvil rather than In this chapter we refer to the process of skill upgrading a pyramid, with the bulk of the population with through the formal school system as an "educational tran- only some primary schooling or less, but more sition." This transition entails changes in both the mean and individuals with tertiary education than secondary the distribution of educational attainment. In figure 4.1 we schooling only. present some stylized shapes for the underlying distribution a Finally, access to university in some countries has of schooling for countries at different stages in their educa- become so widespread that more adults have univer- tional transitions. As we will show, these correspond to the sity education than either secondary or primary school paths followed by countries in Latin America, East Asia, and only. For example, in the United States and Canada Scandinavia. more than half the adult population has some uni- versity education. The distribution of educational o When the mean years of schooling is low, the distri- attainment in these countries resembles an inverted button of educational attainment in countries tends to pyramid. 76 CLOSING THE SKILLS GAP EDUCATION POLICIES FIGURE 4 1 Education Transitions While most countries sought to gradually build the mean level of general education The pyramid (most countries, circa 1960)- - with a narrower base The diamond The inverted pyramid (e g , East Asia, 1980s) (e g , East Asia, current) (e g, North America) some countries focused on tertiary education The lopsided anvil (e g , Venezuela and Costa Rica) Highest level of education [F Educated beyond High School Secondary Only u Primary or No Education In 1960 the Distribution of Educational Attainment The Story of the East Asian Tigers and Finland: in Latin America, East Asia, and the Countries of the Fast, Bottom-up Upgrading European "Periphery" Resembled a First-phase Pyramid Although the starting points were similar, the evolution of In 1960, the first point at which we have comparable edu- educational attainment was not. Figure 4.3 and table 4 1 cation statistics for a large cross-section of countries, edu- show that between 1960 and 2000 the mean years of school- cational attainment was low in Latin America, East Asia, and ing went up by more than 4.5 years in Hong Kong (China), some countries on the European periphery such as Finland Korea, Malaysia, Singapore, Taiwan (China), and Finland In and Spain. The mean years of schooling in 1960 among the the most remarkable transformation, Korea more than tripled population aged 25 or older was 5 or below in every East the mean years of schooling of adults to 10.5-a level well Asian country in the sample, and less than 3.5 in Korea, above those of most European countries Malaysia, Singapore, and Taiwan (China). In Latin America, How did these countries achieve such a remarkable per- mean educational attainment was also 5 years or fewer in formance? In all of them, skill upgrading began with every country, and 2.5 years or fewer in Mexico, Venezuela, improvements at the bottom Take the examples of Korea and a number of countries in Central America. In Spain, and Finland. Between 1960 and 1980, the fraction of adults meanwhile, the mean years of schooling was 3.6, and in with only primary education or less fell by more than 30 per- Finland it was 5.4. In every one of these countries in East centage points in both Korea and Finland Meanwhile, the Asia, Latin America, and the European periphery, more than corresponding fraction of adults with only some secondary three-quarters of the population had only primary school- education increased by more than 25 percentage points in ing or less.2 Figure 4 2 illustrates these similarities by graph- both countries In terms of the schematic distributions in ing the distribution of educational attainment in 1960 in figure 4 1, Korea and Finland both moved from the first to Korea, Brazil, and Spain The figure shows that both the the second stage of their educational transitions by squeez- mean years of schooling and the distribution of schooling ing the base of the pyramid, fattening the middle section, were similar in all three countries. and leaving the top largely unchanged. 77 CLOSING THE GAP FIGURE 4.2 The Distribution of Educational Attainment Was Similar in Brazil, Spain, and Korea in 1960 100 90 80 60 _ U No Schooling or Some Primar 50 _ Some Secondary 40 - - 0~~~~~~~~~~~~~~~~~~~~~~~~~E Some Tertiary 30-- 20~ 10 0 0 -- Brazil Korea Spain Source: Authors' calcularions, based on data from Barro and Lee (2002). Between 1980 and 2000, the fraction of adults with only small: In Brazil, Bolivia, and Guatemala the educational level some primary education or less fell by an additional 32 of adults went up by fewer than 2 years, and in another 13 percentage points in Korea, and by 22 points in Finland. This countries changes were 3 years or fewer. second wave of upgrading above the primary school level was Not only did Latin American countries fall behind the more evenly divided between increases in the fraction of East Asian tigers and Finland in terms of the mean educa- adults with some secondary education only, and increases in tional attainment of the workforce, the pattern of transitions the fraction of adults with some tertiary education. That is, was also different. In 1960, there was (on average) one worker both Korea and Finland moved from the second to the third with tertiary education for every 5 to 10 workers with sec- phase of their educational transitions by again squeezing the ondary education in most countries in Latin America and base of the pyramid, further fattening the middle, and increas- East Asia. Between 1960 and 2000 this ratio of workers with ing the size of the top. As a result of this 40-year process, by tertiary education to those with secondary education 2000 there were more adults with secondary education in quadrupled or more in Bolivia, Chile, Costa Rica, Ecuador, both countries than adults with primary schooling or less. El Salvador, Guatemala, Panama, and Venezuela, and Upgrading skills from the bottom, Korea and Finland had tripled or more in Argentina, Brazil, Dominican Republic, made the transition from a pyramid to a diamond. As table Honduras, and Peru. The bulk of Latin American countries 4.1 shows, so had Hong Kong (China), Singapore, and upgraded skills by making the distribution of education Taiwan (China)-and Malaysia was not far behind. more skewed. The only exceptions to this pattern were Mexico, Cuba, and the English-speaking countries of the The Story of Most Latin American Countries: Caribbean (Guyana, Jamaica, and Trinidad and Tobago). By Slow, Top-down Upgrading contrast, skill upgrading was much more egalitarian in the Overall, skill upgrading between 1960 and 2000 was much East Asia tiger countries. The ratio of tertiary-educated less impressive in Latin America. Some countries did have workers to those with secondary education was essentially large increases in attainment: In both Mexico and Peru the unchanged in Hong Kong (China), Malaysia, and Taiwan mean years of schooling of adults went up by 4.3 years. But (China). In Singapore, this ratio increased, but was still lower table 4.2 shows that in many countries the changes were in 2000 than anywhere in Latin America except Haiti and 78 CLOSING THE SKILLS GAP EDUCATION POLICIES FIGURE 4 3 Fast, Balanced Transitions in the East Asian Tigers and Finland Korea 100 - 90 10 80 Change in years of eucation =7 2 70 - 60 6 50 40 30-- 20 10~~~~~~~~~~~~~~~~~~~~~~~~~~ 1960 ~~~~1980 2000 | Primary or Less Secondary Tertiary Years of Education Finland 100 90- .. --10 Change In years of education = 4 8 60-. - 6 40 _1.- 30 - - 20- : -2 10_ _ , |:* 1lI. - 1960 1980 2000 l Primary or Less l Secondary _ Tertiary -Years of Education in the English-speaking countries of the Caribbean, anc in some countries, such as Mexico, one of the region's star per- Korea it went up only after 1990-once the country had formers in terms of skill upgrading, the mean and the dis- education levels that were higher than those found in most tribution of education attainment in 2000 were similar to OECD countries. In Finland, finally, the ratio of tertiary- those of Korea two decades earlier, while in others, such as educated workers to those with secondary education actu- Colombia, which had done comparatively little skill ally decreased between 1960 and 2000 Mean educational upgrading, the mean and distribution in 2000 were simi- attainment thus went up by more, and the dispersion of edu- lar to those in Korea three decades earlier. Over the course of cational attainment went up by less in the East Asian tigers four decades, Mexico had fallen behind Korea by 20 years, and Finland than in Latin America. and Colombia had fallen behind by 30 Where does this leave Latin American countries in their Other countries, including Costa Rica, the Dominican educational transitions in 2000? Figure 4 4 shows that in Republic, Ecuador, El Salvador, and Venezuela, upgraded 79 CLOSING THE GAP TABLE 4 1 From Pyramid to Diamond in East Asia and Scandinavia YEARS OF NO STUDIES+ COUNTRY YEAR SCHOOLING PRIMARY SECONDARY TERTIARY Korea 1960 3 2 86 5 10 9 26 2000 10 5 24 7 49 5 258 Taiwan (China) 1960 3 3 84 8 110 42 2000 8 5 39.9 411 19 1 Singapore 1960 31 76 6 23 4 00 2000 81 410 48 5 10 6 Hong Kong (China) 1960 47 77 9 17 4 47 2000 9 5 37 3 474 15 3 Finland 1960 5 4 89 9 60 41 2000 10 1 296 47 3 232 Sweden 1960 7 6 58 6 33 9 7 5 2000 114 197 57 2 23 1 Note Countries in East Asia and Scandinavia are sorted (separately) by the change in mean years of schooling between 1960 and 2000 Source Authots' calculations, based on data from Barro and Lee (2002) primarily at the tertiary level, so that their educational tran- while East Asian countries do not. Production processes sitions were unbalanced. As figure 4.5 shows, in terms of the involving natural resources may favor workers with tertiary stylized transitions in figure 4.1, these countries had made education-mining or petroleum engineers, scientists in the transitions from a fat pyramid to an anvil. Table 4.3 biotechnology-over secondary school graduates. By con- shows that in all of them, there were fewer adults with only trast, firms in East Asian countries, building on more tra- secondary school than those with only primary school or with ditional manufacturing processes, may have demanded a some tertiary education. workforce with only secondary schooling. This idea has some appeal, although it is hard to test because of the lack of labor diuCationaD TranS DIons- The (Ulnceria2t) Wh3y force data from many countries prior to the 1980s. Also, it What accounts for the differences between Latin America, does not explain the similarities in the education transition Finland, and the East Asian tigers, as well as differences patterns between Venezuela, where oil extraction dominated within Latin America? We next turn to a discussion of the the economy, and Costa Rica, a country with fewer natural way in which demand, supply, and institutions may affect resource endowments. the speed and distribution of the education transition. The "unusual" amount of skill upgrading at the tertiary level relative to the secondary level in some Latin American The Demand Side ... countries may itself be a response to the rapidly rising Whether technologies favor workers with primary school, sec- returns. In Chile and Colombia gross tertiary enrollment ondary school, or university education can obviously have an rates more or less doubled between 1980 and 1995, at a time effect on the patterns of educational transition. Households of rising relative wages of workers with tertiary education. respond to price signals-including the signals sent by rel- However, there are also counter-examples: between 1980 and ative wages. In a well-known study on India, Foster and 1995 tertiary enrollments did not increase by much in either Rosenzweig (1996) show that the Green Revolution increased Mexico or Brazil, despite the fact that these countries also the returns to education and this, in turn, prompted house- saw rising returns to university. Moreover, in some Latin holds to seek further schooling in those areas most suited for American countries such as Argentina and Mexico the the new technologies. A background paper for this report by biggest increases in tertiary enrollment rates took place Mayer and Foster (2002) shows that households in Mexico between 1970 and 1985, well before the increases in the are also responsive to the rising rates of return to schooling. demand for educated workers associated with the transfer of Latin American countries generally have a comparative skill-biased technologies through trade and FDI. One way advantage in natural resource extraction and processing, of reconciling these varying patterns within the region is by 80 CLOSING THE SKILLS GAP EDUCATION POLICIES TABLE 4 2 From Pyramid toward Diamond in Some Latin American Countries YEARS OF NO STUDIES+ COUNTRY YEAR SCHOOLING PRIMARY SECONDARY TERTIARY Malaysia 1960 2 34 912 7 2 15 2000 7 88 49 5 43 0 75 Spain 1960 3 64 92 7 3 9 3 4 2000 725 530 307 162 Mexico 1960 2 41 94 3 4 3 14 2000 6 73 59 7 29 0 113 Peru 1960 3 02 87 7 9 7 2 6 2000 7 33 49 5 28 1 22 4 Cuba 1960 3 79 93 7 4 3 19 2000 7 78 46 0 42 6 11 3 Panama 1960 4 26 82 1 15 4 2 6 2000 7 90 518 28 5 19 8 Argentina 1960 4 99 85 4 116 3 0 2000 8 49 55 4 24 9 19 7 Trinidad and Tobago 1960 419 88 0 10 9 10 2000 7 62 514 44 1 4 5 Chile 1960 4 99 75 5 22 4 2 1 2000 7 89 48 2 360 15 8 Jamaica 1960 2 46 94 1 5 5 0 5 2000 5 22 578 380 4 1 Guyana 1960 3 50 97 0 2 5 0 5 2000 6 05 60 0 36 0 40 Honduras 1960 1 69 95 8 3 5 06 2000 408 829 106 65 Paraguay 1960 3 35 91 8 69 12 2000 5 74 73 6 18 1 8 3 Nicaragua 1960 2 09 92 7 4 7 2 5 2000 442 747 165 89 Uruguay 1960 5 03 78 1 16 8 5 1 2000 7 25 55 4 32 1 12 5 Colombia 1960 2 97 86 4 11 9 1 9 2000 5 01 68 7 214 9 9 Haitr 1960 0 70 95 6 4 2 0 2 2000 267 86 7 12 3 1 0 Brazil 1960 283 867 113 20 2000 4 56 78 0 13 5 8 4 Guatemala 1960 1 43 95 9 3 6 0 5 2000 3 12 847 95 58 Bolivia 1960 4 22 68 7 27 5 3 7 2000 5 54 70 7 14 9 14 4 Note Countries are sorted by the change in mean years of schooling between 1960 and 2000 Source Authors' calculations, based on daia from Barro and Lee (2002) underlinng that the increase In demand for tertiary educa- the Supply Side tion by individuals could only translate into more school- Even the most casual glance at the comparison between Latin ing where institutions, especially the university system, were America and East Asia suggests that both the speed and bal- responsive. As we discuss below, it is likely no coincidence ance of transitions can be affected by government "effort " that Chile and Colombia, two countries with a sharp increase All of the East Asian tiger economies made education a pri- in enrollments concurrent with the increase in the wage premia ority. Korea is a good example: The country's government to tertiary graduates, had also reformed their university emphasized the sequential expansion of the education system. system, focusing first on enrollments in primary schools 81 CLOSING THE GAP FIGURE 4.4 Balanced but Slow Transitions in Some Latin American Countries Colombia 1000- 90 10 80 8 70 Change in years 60 of education 2.0 6 50 40 4 30 20 2 10 0 0 1960 1980 2000 | Primary or Less _ Secondary Tertiary -Years of Education Mexico 100- 90- 10 80 | Changein 70 - t _ years of -8 70 ~~~~~~~~~~~~~~~~~~~education =4.2 60 6 50 40-4 30- 20 2 10~~~~~~~~~~~~~~~~~~~~~~~ 1960 1980 2000 | Primary or Less Secondary = Tertiary -Years of Education (between 1950 and 1970), then on enrollments in secondary through a comparably rapid expansion of the secondary schools (between 1970 and 1985), and finally on access to school system. university (especially, since 1990) (Gill and Patrinos 2002; More "effort" does not always mean more money. In Kim 2001). Within Latin America, the clearest example of describing the success of the Cuban school push, Carnoy how a supply push can make a difference is Cuba. In 1950, (2002) underlines the change in "the level of expectations the gross secondary enrollment in Cuba was about 5 percent; in the society as a whole regarding what students are by 1965, it had reached 23 percent, and continued to rise expected to learn in school." Indeed, aggregate public to 42 percent in 1975; by 1985 it had reached 82 percent, resources do not seem to explain much of the difference the highest level in Latin America at the time, bar none between Latin America and East Asia. Table 4.4 shows that, (Reimers 2001). No other Latin American country has gone on average, Latin American governments spend a higher 82 CLOSING THIE SKILLS GAP EDUCATION POLICIES FIGURE 4 5 Unbalanced and Slow Transitions in Other Latin American Countries Costa Rica 100 90 10 80 70 --Change in years of education =2 2 60 50 - - 40 4 30 20 --2 10- 0-- 0 1960 1980 2000 1= Primary or Less _ Secondary _ Tertiary -Years of Education Venezuela 100 90 -10 80 8 70 60 Change in 6 50 years of 6 education = 3 1 40 4 30-- 20 2 10_ 0 - 0 1960 1980 2000 J=Primary or Less 1 Secondary Tertiary -Years of Educatrion fraction of GDP on education than those in East Asia Broad public expenditure aggregates are notoriously dif- (although private expenditures on education may be higher ficult to compare. In some countries, they include spending in East Asia than in Latin America). Moreover, within Latin by the central government and regional governments only, America there is no clear relation between high levels of which will be misleading in heavily decentralized systems. expenditure and good education outcomes-again with the Also, they may or may not include items such as retirement exception of Cuba, which both spends a lot on education and pay for teachers, transfers to private institutions, and expen- has some of the most impressive outcomes in terms of both ditures by entities outside the Ministry of Education, such the quantity and quality of schooling. as Social Fund programs which repair or build school 83 CLOSING THE GAP TABLE 4 3 From Pyramid to Anvil in Other Latin Amenican Countries YEARS OF NO STUDIES+ COUNTRY YEAR SCHOOLING PRIMARY SECONDARY TERTIARY Thailand 1960 3 4 94 5 49 06 2000 61 79 5 9 3 11 3 Ecuador 1960 2 9 915 7 1 1 4 2000 65 629 183 187 Venezuela 1960 2 5 93 2 5 3 1 5 2000 56 723 97 180 Dominican Republic 1960 2 4 96 5 2 7 0 8 2000 5 2 72 5 13 1 14 5 El Salvador 1960 1 7 95 0 4 6 0 5 2000 4 5 80 6 8 8 10 6 Costa Rica 1960 3 7 90 5 6 8 2 7 2000 6 0 70 1 11 3 18 6 Note Countries in Latn America are sorted by the change in mean years of schooling between 1960 and 2000 Source Authors' calculations, based on data from Barro and Lee (2002) TABLE 4 4 infrastructure in many Latin American countries. For all of Education Spending as a Fraction of GDP Is NIot Low in Most these reasons, comparisons of broad expenditure aggregates Latin American Countries should be treated with caution. 1980 1985 1990 1995 If there are large differences across countries in terms of Latin American Economies 3.73 3.53 2.94 4.37 how they account for education spending, and if these dif- Argentina 2 67 1 47 1 12 3 33 ferences remain more or less constant within a country over Belizvea 4 428 4 71 24 47 5 93 time, estimating the effect of changes in spending on changes Brazil 3 60 3 78 5 07 in enrollment or attainment may provide more accurate Chile 4 63 4 38 2 67 3 10 results.3 Figure 4.6 presents a scatter-plot of the changes in Colombia 1 86 2 89 2 52 3 67 Costa Rica 7.76 4 45 4 58 4 64 educational attainment of adults (in years of education) Cuba 7 19 6 27 6.55 6 78 between 1980 and 1995 as a function of changes in expen- Dominican RepubILc 2 16 1 59 1 92 Ecuador 5 59 3 70 3 08 3 44 ditures (as percentage poits of GDP). The figure shows no El Salvador 3 90 1 98 2 19 clear relationship between the two. Colombia increased its Guatemala 1 81 1 56 1 39 1 70 education spending by two percentage points, while Guyana 9 79 9 79 4 83 4 29 Haiti 1 48 1 18 1 46 Ecuador decreased it by two points. In spite of these differ- Honduras 3 19 4 17 3 62 ences, both countries had small increases in mean years of Jamaica 6 96 5 40 5 37 6 35 schoolmg (just over one year). Mexico and Guatemala both Mexico 4 73 3 94 3 73 4 87 Nicaragua 3 38 5 93 3 44 3 71 left their levels of expenditure unchanged, but in Mexico Panama 4 90 4 55 4 93 5 22 the mean years of schooling grew by three years, and in Paraguay 1 51 1.50 112 3 36 Peru 3 09 2 86 2 29 2 94 Guatemala by one year only. Assuming there were no Trinidad and Tobago 3 97 6 05 3 97 3 64 major changes in the quality of the schooling delivered, Uruguay 2 29 2 83 3 08 2 81 Mexico and Ecuador would seem to be getting a lot more Venezuela 4 40 5 08 3 14 for their additional education spending than Guatemala Ease Asian Economies 2 45 3.42 2 24 2 60 Indonesia 1 72 1 04 1 40 and Colombia. Korea 3 73 4 45 3 45 3 68 Perhaps the most fair assessment of the relation between Malaysia 6 04 6 61 5 45 4 68 expenditures and changes in mean educational attamment Philippines 1 72 1 35 2 90 2 96 Singapore 2 84 4 40 3 01 2 98 in Latin America is that increases in expenditure, whether Thailand 3.42 3 79 3 59 4 14 public or private, can only make a difference if additional Vietnam 2 06 2.99 pulcoprvt,cnolmaeadfeecifdiinl monies are used well. One consideration is the level- Note Regional averages are population-weighted Source Authors' calculations using World Bank databases primary, secondary, tertiary-that receives the additional 84 CLOSING THE SKILLS GAP EDUCATION POLICIES FIGURE 4 6 There Is No Clear Relationship between Changes in Education Expenditure and Changes in Attainment in Latin America 3s - r4 Mexico 0 C1E 0 C0 0s = 2 Chile e Panama Chile Peru ~~~Argentina Dominican Republic : ~~~~~~~~~~~~~~~~Honduras Cuba oi XJamaica NBrazBil Niaragua Bolivia 0 Uuguay >1 Costa Rica c El Salvador Ecuador Paraguay Trinidad and Tobago Colombia X Guatemala -Y I -4 -2 0 2 Changes in Education Spending, Percentage Points of GDP, 1980-95 Source Authors' calculations based on World Bank and UNESCO data resources. Cross-country comparisons of the composition of equality as an explanation for the differences in educational education expenditures across levels are even more hazardous attainment between Latin America and North America.5 than comparisons of expenditure aggregates. But it IS likely This clearly has some appeal in accounting for differences no coincidence that within Latin America the ratio of expen- between Latin America and the East Asian tigers as well, ditures on tertiary education to primary and secondary school given the much higher levels of inequality of income and expenditures is highest in Venezuela and Costa Rica, two other assets in the former in 1960. The focus on inequality countries with the largest secondary school gaps (together may also do well in explaining the patterns found in some with Brazil) but essentially no gaps in tertiary enrollment.4 countries, such as Brazil, with high levels of inequality, and A second consideration is how resources are used within a an unequal distribution of education Brazilian public uni- given education level-specifically if programs are structured versities, in particular, have often been described as a system so that they relieve a clearly identified constraint to the acqui- of "elite formation " Yet other countries are a puzzle, sition of further schooling. This is the subject of the second including Costa Rica (where income inequality is low, but half of this chapter, where we discuss policies for secondary the distribution of education is skewed), or Chile (with rel- school and universities atively high levels of income inequality, but reasonably broad access to mass secondary schooling) (Reimers 2001). ... and Institutions Ultimately, the dynamics of educational transitions seem No matter what the ultimate effect of additional expendi- to be highly country-specific. In some countries (such as tures on outcomes, the deeper question is why governments Argentina, Chile, and Uruguay), reasonably broad coalitions in some countries have been willing to tax their citizens to emerged which supported mass schooling in the first half provide a majority of people with education, and why some of the 20th century; in others, such as Cuba, Korea, or have spent additional resources on egalitarian interventions, Taiwan (China), authoritarian governments saw mass such as increases in secondary school enrollments, rather than schooling as integral to the construction of a new national elite interventions, such as high-quality university educa- identity; and in yet others, no such consensus emerged, so tion for a privileged few. There is no simple answer to this that the speed of educational transition was slow, and the question. Some authors have stressed higher levels of pattern often unbalanced. 85 CLOSING THE GAP Educational rranmsons: Policy Ompcati$ons lesson is that the countries that upgraded skills most suc- Differences in educational transitions pose a public policy cessfully did so from the bottom-up Like Korea a century challenge. What is the appropriate role for government in later, the United States went through three sequential stimulating the pace and the pattern of transitions? We argue 'waves" of skill upgrading, involving large increases in pri- below that public investments should focus on expanding mary enrollment in the mid-19th century, in secondary the coverage and quality of the secondary school system in enrollment between 1910 and 1940, and in university enroll- most countries in Latin America. In addition, to facilitate ment after World War II. It is hard to overstate the impor- the transition to university, government policy should help tance of these "transformations," especially the transition to individuals overcome market failures. mass secondary education It gave the United States a huge lead in educational attainment over other industrialized A Balanced Education Transition countries, and provided the essential building block that The case for a balanced education transition for Latin Amer- made the United States the most successful economy of the ica rests on three arguments: ensuring that the most gifted 20th century (see box 4 1). The second lesson is that forg- students go to university, minimizing inequality, and taking ing a broad consensus is a sine qua non for the reforms nec- advantage of the complementarities between the phases in essary for a fast, balanced educational transition. Governing an educational transition and those in the technological tran- elites in Latin America have traditionally viewed education sition of a country. as elite formation, and have limited access to education as a means of controlling the distribution of economic o A small base of secondary school graduiates means that the resources and wealth. The complementary relationship most able are often excluidedfrom university. Education is between technology and education means that educated cumulative. one must graduate from secondary school workers are ever-more important as the key to high pro- in order to go to university Having few secondary ductivity and high growth rates. Insofar as this more closely school graduates therefore means a small pool of uni- binds the economic well-being of elites to access to educa- versity candidates When university students are tion by non-elites, it may broaden the support for policies that drawn from this small pool, many of the most gifted increase the coverage and quality of the education system students cannot go to university The third lesson is that, even if the most necessary reforms o Unbalanced transitions may perpettuate inequality. Latin to education for many Latin American countries involve America is the most unequal region in the world, and those that raise enrollment levels in secondary school, coun- unbalanced transitions can perpetuate inequality. tries cannot afford to disregard tertiary education There is The bi-modal "anvil" distributions of educational considerable interdependence across education levels Parents attainment relegate the bulk of the workforce to low- send their children to secondary school in part because of skill, low-wage jobs, while a privileged few earn the the option value of a university education Reforms that help high wages that come with a university education.6 more children from disadvantaged backgrounds enter umn- Because of the close relationship between schooling versity may therefore also raise secondary school enrollment and earnings, an unequal distribution of edutcation tends levels We now turn to a discussion of precisely what poli- to perpetuate an unequal distribution of income. cies are most necessary at the secondary school and univer- o Sequential technological tupgrading requires a high frac- sity levels in Latin America tion of the workforce to have secondary schooling. As we dis- cussed in chapter 3, mass secondary schooling is an UnlldeIlsi3ndlng thDe EducalGon arkete important determinant of the ability of countries to $eacoiNd1lgy SChcoD attract trade and FDI with a high technological con- Understanding whether the bottleneck in secondary educa- tent, and to benefit from the spillovers from these tion is on the demand or supply side of the market is an investments. important first step to designing appropriate policies. In addition, careful consideration must be given to the extent In education, as elsewhere, history holds important lessons to which there are incentive problems that result in waste- for Latin America Three are particularly important. The first ful education spending. 86 CLOSING THE SKILLS GAP EDUCATION POLICIES BOX 41 Expanding Secondary School: The Great Transfonnation of American Education In education, as elsewhere, history holds important intend to go on to college. Second, school administrators lessons for Latin America. Between 1910 and 1940, sec- and others waged a successful campaign to convince ondary schooling expanded greatly in the United States. youths and their parents that high schools offered them In 1910, 10 percent of youths were high school graduates, something of value. Third, the low levels of inequality and but by 1930 the median youth had a high school diploma. the egalitarian features of the schooling system in the This transformation gave the United States a large lead United States, which did not encourage early, rigid | over other industrialized countries in the level of educa- tracking of students, facilitated mass secondary enroll- tion of its work force-a lead that persisted for a long time: ments. Lastly, the Great Depression substantially reduced As late as 1962, for example, only 15 percent of British the employment prospects of youths, and thus reduced the 17-year-olds were in school, while the high school grad- opportunity cost of schooling As Goldin explains, "the uation rate in the United States was almost 70 percent. Great Depression may have had one positive effect: it What explains this "Great Transformation" of Amern- enticed youth to stay in school." Differences in the oppor- can education? Recent work by Goldin (1999, 2001) pro- tunity cost of schooling also help explain why the "high vides a number of insights. First, prior to mass secondary school movement" lagged in the agricultural areas of the enrollment, the rate of return to secondary school was very South, as well as the industrial areas of the North and high in the United States. This rate of return was high, Midwest. in part, because high schools in the United States suc- The massive expansion of secondary school enrollments cessfully transformed themselves from institutions that in the United States substantially brought down the high only prepared youths for college, to programs that taught school wage premium As the premium fell, employers in their students skills-such as accounting, typing, short- turn hired more high school graduates in occupations that hand, and algebra-that were valued in the workplace. had previously been filled by unskilled workers. Such high rates of return provided youths with the incen- tive to seek further schooling-even those who did not Source: Goldin (1999, 2001) Direct Private Costs, Opportunity Cost, and figures are probably low by regional standards, but direct Demand-side Constraints to Secondary School private costs are unlikely to be a critical reason for the low Demand-side constraints to secondary education can arise for secondary enrollments in many countries in the region. three reasons First, the direct private costs of schooling may While the opportunity cost of sending a child to primary be an impediment. Second, the opportunity cost of going school is generally thought to be low, this is not the case for to school, in terms of foregone wages, may be high Third, children of secondary school age. Child labor is a reality in the "rate of return" to secondary schooling may be low all but the wealthiest Latin American countries, and teenage In general, direct private costs are much less of an imped- children can supplement household income by working in iment at the secondary school level than at the university the informal sector, or as agricultural laborers Table 4 5 pro- level. Calculations based on the ENIGH survey in Mexico, vides estimates of the ratio of the hourly wages earned by for example, suggest that direct private cost, including school 15- to 19-year-olds in urban areas in 12 Latin American fees, school supplies such as textbooks and pencils, uniforms, countries to those of median wages for the population aged and transportation, are less than 10 percent of household 15 to 70, as well as the ratio of unemployment rates among income per capita-both for households in the poorest youths to those in the overall population. The table shows income quartile, as well as for those in the richest income that the wages earned by youths are generally between one quartile The fact that the fraction of expenditures is rela- half and three quarters of median wages. The probability of tively constant across quartiles is indirect evidence that the unemployment of 15- to 19-year-olds is much higher than poor are receiving a worse-quality education. The Mexican those observed in the overall population, and this will also 87 CLOSING THE GAP TABLE 4 5 TABLE 4 6 The Opportunity Cost of Attending Secondary School Appears High Rates of Return to Teriary Education, and Modest Rates to to Be Hig* in Most Latin American Countries Secondary School in Most of Latin America WAGE RATIO UNEMPLOYMENT RATIO 'RATE OF RETURN" TO "RATE OF RETURN TO MALES FEMALES MALES FEMALES SECONDARY SCHOOL TERTIARY EDUCATION Argentina 061 0 53 2 67 2 54 Argentina 0 05 0 11 Brazil 044 0 54 2 80 2 74 Bolivia 008 0 14 Chile 0 52 0 62 3.12 3 12 Brazil 0 19 0 19 Colombia 0 72 0 73 2 68 2 35 Chile 008 0 22 Dominican Republic 0 65 0.72 Colombia 005 0 18 El Salvador 0 86 0 71 1 38 1 14 Mexico 0 06 0 13 Honduras 0 72 0 72 1.93 1 84 Note The rare of return is approximated by che wage increment divided by the number Mexico 0 66 0 69 of years in a cycle Nicaragua I1 1 0 74 Source Aurhors' calculations based on data from the labor force surveys Paraguay 0 59 0 46 Peru 0 55 0 58 discount rate households use when they compare investments Venezuela 0 64 0 62 2.40 2 89 in schooling with other investments. For one thing, this Note The wage ratio is the ratio of median wages of 15- to 19-year-olds to the median discount rate is likely to be higher for poorer households. wages of 15- to 70-year-olds, the unemployment ratio is the ratio of rhe unemploy- ment rate of 15- to 19-year-olds to the unemployment rate of 15- to 70-year-olds Table 4.6 suggests, however, that investments in secondary Source Authors' calculations based on data in Cunningham (2002) school may not be an economically sensible decision for chil- dren in some Latin American countries unless there are be a factor in any reasonable calculation of the opportunity improvements in quality, additional increases in the demand cost of secondary schooling.7 Still, table 4.5 suggests that for secondary school workers by firms, or high transition rates the opportunity costs of schooling are not low, at least in from secondary school to university. urban areas. Further evidence of the importance of the opportunity Access and Supply-Side Constraints to Secondary School cost of schooling comes from careful analyses of patterns of Inadequate access to school and teacher shortages can both school enrollment and work of children during macroeco- be supply constraints to schooling. In some areas, especially nomic crises in Latin America. The theory of human capital rural areas, there may simply be no school available The suggests that macroeconomic crises will have counter-acting extent to which lack of school infrastructure explains low effects on child schooling: On the one hand, crises limit enrollment levels can often be approximated with adminis- opportunities in the labor market, and thus depress the trative data-for example, by comparing the distribution of opportunity cost (which should lead to more schooling), on the school-aged population from a census with the distrib- the other hand crises reduce household income and may ution of school facilities. Many household surveys also have make borrowing constraints more binding (which should community modules that ask informants about the presence lead to less schooling). In practice, the negative effects of of certain facilities, including schools, in a community. crises on schooling in Latin America are, at most, moder- Focusing on rural areas, 79 percent of communities with 500 ate: In Brazil, schooling decisions are largely unaffected by people or fewer in Peru reported having a primary school, aggregate economic conditions, while children in Mexico and but only 17 percent had a secondary school. In those com- Peru are less likely to drop out of school during a crisis. The munities where there was no secondary school the mean time counter-cyclical relationship between schooling and aggre- to walk to a neighboring school was about one hour and 15 gate income provides further evidence that the opportunity minutes, and in 10 percent of communities the time was cost of schooling is an important determinant of secondary three hours or more.9 Clearly, lack of school infrastructure school enrollments in a number of Latin American countries.8 could be a constraint to secondary school enrollments in rural The high opportunity cost of secondary schooling in Latin areas in Peru. America is a serious concern because the rate of return to secondary schooling is low in many countries. Table 4.6 The Problem of Low Education Quality shows that this rate of return is 6 percent or lower in The low performance of Latin American students on inter- Argentina, Colombia, and Mexico. It is not clear what national tests is evidence of the low quality of education 88 CLOSING THE SKILLS GAP EDUCATION POLICIES TABLE 4 7 There Is Generally Little Variation in Primary School Student Perfornance on Test Scores within Latin America, with the Exception of Cuba MEDIAN SCORES ON UNESCO - OREAL TEST, 1997 THIRD GRADE THIRD GRADE FOURTH GRADE FOURTH GRADE LANGUAGE MATHEMATICS LANGUAGE MATHEMATICS Cuba 343 351 349 353 Argentina 263 251 282 269 Chile 259 242 286 265 Brazil 256 247 277 269 Venezuela 242 220 249 226 Colombia 238 240 265 258 Bolivia 232 240 233 245 Paraguay 229 232 251 248 Mexico 224 236 252 256 Peru 222 215 240 229 Dominican Rep 220 225 232 234 Honduras 216 218 238 231 Source Carnoy (2002, table 7) throughout the region. School quality is important in its own bringing a high proportion of their young population to right, since it determines how much children learn for a complete levels of schooling available to only a few in the given year of education. Higher quality schooling may also past." Still, there is some agreement that education systems reduce repetition rates and free up resources for further in Latin America have generally spent too much on recur- expansions in coverage. rent expenditures, especially teacher salaries, and too little The OREAL-UNESCO exams tested third- and fourth- on capital expenditures. Filmer and Pritchett (1999), for graders in 12 Latin American countries on language and example, argue that school inputs are systematically biased mathematics. These tests show that there appears to be rel- in favor of those valued by teachers over those valued by stu- atively little variation in the education quality of primary dents This would suggest the need for policymakers to re- school across countries in the region, with the exception of allocate spending from the wage bill to complementary Cuba, which outperforms every country in Latin America by inputs such as textbooks and classroom material. Most stud- a wide margin Table 4.7 shows that in language as well as ies also find that teachers are critical in the learning process, mathematics, and among third-graders as well as fourth- but the relation between "observable" teacher characteris- graders, Cuban students score at least 70 points more than tics (such as education, experience, and salary levels) and students in the next-highest scoring country. By contrast, learning outcomes (such as performance on tests) is weak the difference between the mean score in this next-highest (Hanushek 2002). Teacher motivation and ability are criti- scoring coimtry and the worst performer in the region is never cal, but how best to improve them is unclear. more than 50 points. There are no comparable tests for sec- Should educators give higher priority to policies that seek ondary school students covering a large number of countries to improve the quantity or quality of secondary schooling' in the region, but the equally low scores of Chilean and Fortunately, there may be no quantity-quality trade-off. Colombian students on the TIMSS suggests a widespread Higher quality schools attract more students, and countries problem with quality at the secondary school level. in Latin America with higher mean enrollment levels do not There is little consensus among academics or policy- appear to have lower quality. Figure 4.7 graphs the mean makers about the most effective interventions to improve test score on OREAL-UNESCO tests as a function of enroll- school quality-in Latin America, or elsewhere.10 In a back- ment levels. The figure shows that there is no clear rela- ground paper prepared for this report, Carnoy (2002) writes tionship between these two outcome measures: For example, that "historically, nations have not been successful in improv- Argentina and Peru both have high enrollment rates, but ing the average amount that students learn in, say, fourth Argentina outperforms both Peru and Venezuela, a country or eighth grade, but they have been very successful in with a relatively low enrollment rate, in terms oftest scores. 89 CLOSING THE GAP FIGURE 4 7 There Is No Clear Relationship between Educafton Qual;ty and Quanftit in Latin America 350 - Cuba u- g K3 300 z Chile Bzil Argentna 3 250 - Colombia ij) Mexico Paraguay Bolivia I enezuela Dominican Republic Peru 200 - 85 90 95 100 Net Primary Enrollment Rate, 1998 Within countries, too, there IS some evidence that school evaluation needs to be conducted to establish the conditions access and quality are complements rather than substitutes under which these policies can be most effective. (Hanushek 1995) Capitation schemes transfer resources to schools in direct proportion to the number of students they enroll (rather than Incentives and Institutional Reforms in proportion to the number of teachers, or in accordance A recent review of the literature concludes that "develop- with historical budgeting rules). In Brazil, a substantial por- ing countries spend hundreds of billions of dollars each year tion of public resources for education for subnational levels on education, and there is ample evidence that these funds of government have been distributed on the basis of student are used inefficiently" (Glewwe 2002, p. 475). In Latin enrollment since 1998, under the financing mechanism America, too, education outcomes in many countries con- called the Fzndo de Manutenfdo e Desenvolvimento do Ensino Fun- tinue to be poor, both in terms of quantity and quality, in damental e de Valorizafdo do Magist'rio (FUNDEF). State and spite of substantial expenditures. This has led many poli- municipal governments contribute an earmarked proportion cymakers to consider deep reforms of the incentives in the edu- of their revenues, and these resources are then distributed cation system as a possible remedy. Many of these reforms on the basis of the number of primary education students are designed to hold schools and teachers accountable for enrolled. As each subnational government contribuites to learning outcomes. FUNDEF, but obtains receipts from the fund only to the Reforms to align incentives generally fall into one of two extent that it has enrolled students, the new legislation cre- categories: (i) those in which the government conditions ated an incentive to subnational governments to enroll as budget transfers on outcomes, including capitation schemes, many primary education students as possible. This new and schemes in which schools or teachers with superior per- financing scheme is credited with substantially increasing formance are rewarded with bonuses; and (ii) those in which primary enrollment levels in Brazil (World Bank 2002). greater control is given to parents, either in the form of par- Performance-based financing has been implemented in a ticipation and decisionmaking power over the allocation of number of countries in Latin America In Mexico, under the school budgets, or with a voucher that allows them to select Carrera Magisterial program, teacher pay is linked to several a school. All of these have been tried in Latin America, and measures, including performance on the job (as evaluated by all of them hold promise-although more systematic the principal), scores on a test taken by the teacher, student 90 CLOSING TH e SKILLS GAP EDUCAI ION POLICIES test scores, and teacher education and experience The pro- gave private (non-tuition-charging) subsidized schools the gram is voluntary, but more than 60 percent of eligible teach- same per-student payment as public schools. The Chilean ers have participated in it. A recent evaluation of the program experience has been the subject of intense analysis, but little suggests that students in classes taught by teachers partic- consensus exists about the extent to which the reforms as a ipating in the Carrera Magisterial program have about 2 whole Improved student outcomes.1' More recently, careful percent higher scores on Mathematics tests (Lopez-Acevedo analysis of a randomized experiment with a voucher program 2002, p. 28) In Chile, the Sistema Nacional de Evaluaci4n del in Colombia (Programna de AnplhaciAn de Coberttura de la Edu- Desempeno de los Establecimientos Edzucacionales Subvenczonados caci6n Secuindaria, or PACES) suggests that the program (SNED) gives monetary bonuses to schools chosen for their resulted in a reduction in repetition rates among participants outstanding performance-whereby schools are selected pri- (See Angrist and others 2002). marlty in terms of student performance on a nationwide test Misaligned incentives are pervasive in the provision of The SNED is innovative in a number of ways: It only com- education services. Principal-agent problems can be found pares schools serving similar populations (to control for the at every link in the delivery chain-between Ministry of fact that children from better-off economic backgrounds gen- Education officials and school administrators, between prin- erally do better on tests); it considers both the performance cipals and teachers, and (most important) between parents level and changes in performance of schools; and it rewards and teachers. Reforms that seek to find innovative solutions all teachers within a high-performing establishment While to these problems in Latin America, be they those attempt- its nationwide coverage makes it difficult to convincingly ing to link finance with performance or those giving par- estimate the impact of the SNED on student outcomes, it ents greater control over resources, should continue to be appears to have improved the accountability of schools and tried and carefully evaluated teachers (See particularly Mizala and Romaguera 2002) Other reforms have sought to directly empower parents Putting the Pieces Together: Patterns of Secondary In El Salvador, schools in the Community-Managed Schools School Attainment in Latin America Program (Educacz6n con Participactzn de la Comunidad, or How can we assess the relative importance of demand, EDUCO) are managed autonomously by community edu- supply, quality, and incentive problems in explaining the low cation associations (Asociaciones Comunalespara la Educacitn, levels of secondary school attainment in a countryp To con- or ACES) The Ministry of Education contracts these ACES sider this, we turn to data from the Demographic and Health to deliver a given curriculum to an agreed number of stu- Surveys (DHS). The advantage of the DHS is that they apply dents The ACES are then responsible for hiring and firing the same questionnaire in all countries Moreover, and unlike teachers, closely monitoring performance, and equipping and most labor force surveys, they collect data for both urban and maintaining schools. The EDUCO program is credited with rural areas. a rapid expansion of schools to rural areas, with diminished In figure 4.8, we graph the probability that persons aged student absences, and with a significant improvement in stu- 20 to 29 have completed at least a given number of years dents' language skills (Jimenez and Sawada 1999). A simi- of schooling, ranging from 1 to 9 years, separately for urban lar program in Nicaragua is also thought to have positive and rural areas in Guatemala (1999), Nicaragua (1998), and effects on the learning outcomes of children (King and Ozler Peru (2000).I2 The figures include a vertical line corre- 2000) sponding to completion of primary school (The sample Vouchers are perhaps the most radical solution to the includes those who are still in school as well as those who incentive problem in the delivery of education services have dropped out ) A comparison of the figures shows large Vouchers make education resources portable, and thus allow differences across countries In Peru more than 90 percent parents to "vote with their feet" by sending their children of youths in urban areas have completed at least primary to a school of their choice Chile has had a nationwide school, while in Guatemala and Nicaragua the numbers are voucher program since 1981 The Chilean voucher program below 80 percent The graphs also show that attainment Is was implemented in the context of a larger reform that trans- much lower in rural areas than in urban areas in all of the ferred schools from the Ministry of Education to munici- countries. In the rural areas of Guatemala and Nicaragua a palities, gave municipal schools a per-student transfer, and quarter of youths have no schooling whatsoever. 91 CLOSING THE GAP FIGURE 4 8 Patterns of Educational Attainment in Guatemala, Nicaragua, and Peru Guatemala, 1999 1 °° - 0 90 - 0 80 .X a °~ ' 07 0 - 060 ON . 05 a 0 40- 0 0 30- : E o 20 - End of Primary u 020 School iZv 010- 000 - 1 2 3 4 5 6 7 8 9 Nicaragua, 1998 Peru, 2000 1 00 1.00 0 90- 090 0 80 0 80 070- 070 060- 060- 0 50 0 50- 040 - 040° 030- 030- 0 20 - End of Primary 020 End of Primary 0 20 - School ^ 0 20 - School 010 010 000 000 1 2 3 8 9 1 2 3 4 5 6 7 8 9 1 2 3 4 5 6 7 8 9 Grade Grade Note The rural schedule is shaded in lighter color, and is always below the urban schedule Source Authors' calcularions based on data from Filmer (2000) Two features of the graphs are helpful in determining level cannot therefore be a result of the lack of access. On whether the low attainment levels of youths are driven by the other hand, this decline could be explained by the rising problems of inadequate demand or inadequate access: The opportunity cost of schooling. To see this, consider two facts. steady decline in the number of youths who have completed First, the number of years of schooling a child can complete at least a given number of years of schooling, and the sharp is related to his or her age. Put differently, 20- to 29-year- kinks associated with transition from one schooling level to olds who have completed only one year of schoolsng will, another observed in some countries. on average, have dropped out of school at a younger age than Schools generally teach all grades within a level-for those who have completed two. Second, the wage that chil- example, a primary school will teach all grades in primary, dren can command also rises steadily with age. The steady while a secondary school will teach all grades in secondary decline in attainment may therefore reflect the rising oppor- The steady decline in attainment within a given schooling tunity cost of schooling: It may not pay off to drop out after 92 CLOSING THE SKILLS GAP EDUCATION POLICIES completing only one or two years of schooling, because these that both the urban and rural schedules pivoted counter- children would be very young, but it may pay off to drop clockwise: A higher fraction of 20- to 29-year-olds have com- out for older children who have completed a few more pleted more years of schooling in 2000 than in 1990 On years-especially, if the quality of education is low. the other hand, conditional on completing the primary cycle, What about the kinks in attainment that coincide with the transition to secondary school does not appear to have one's completion of primary school? These are particularly become much easier in 2000 than in 1990, especially in rural obvious in the rural areas of Guatemala and Peru. As we areas. Figure 4.9 therefore suggests that increases in the showed in our analysis of the Peruvian data, primary schools demand for schooling, rather than improvements in access, are generally available in most settings, both urban and rural, explain the bulk of the Improvements in educational attain- but secondary schools may not be. As a result, a dispropor- ment in Colombia in the 1990s tionate number of children drop out of school in the tran- sition from primary to secondary school These kinks are Using Demand-Supply Diagnostics to Inform therefore highly suggestive of supply-side constraints. the Design of Policies Alternatively, they could be the result of particularly large These diagnostics provide a useful first step to considering wage premia to completing the last year of schooling within policies to improve schooling outcomes. In many countries a given level-graduation or so-called "sheepskin" effects. 13 in Latin America, there are limited (if any) secondary school This seems unlikely, however, as sheepskin effects are gen- facilities in rural areas. Under these circumstances, invest- erally thought to be lower in rural than in urban areas ments in school infrastructure can have very high returns (see because of the higher fraction of the rural population in non- box 4.2). More innovative solutions such as long-distance wage employment education may also hold promise The TeleSecundaria program Comparing graphs in a country for two years can also give in Mexico is an interesting example of this approach. Finally, one a sense of how demand and supply constraints have innovative reforms to better align incentives with learning evolved over time In figure 4.9, we present graphs for outcomes within the education system should be imple- Colombia using the 1990 and 2000 DHS. The figure shows mented, and carefully evaluated. FIGURE 4 9 Supply and Demand Factors in the Evolution of Educational Attainment in Colombia, 1990-2000 Colombia 0 9 > 0 8 o ? 0 6 -~- ., \ \ Urban 1990 *~Rural 1 990 c, 3 0 5 - -Urban 2000 Rural 2000 04 1 2 3 4 5 6 7 8 9 Grade Source Authors' calculations based on the DHS data available in Filmer (2000) 93 CLOSING THE GAP BOX 4.2 BOX 4.3 The Retums to Investments in School Infrastnucture: Conditional Cash Transfers: The Impact of PROGRESA Evidence from Peru and Indonesia in Mexico A number of studies have assessed the impact of spend- Conditional cash transfer programs can be an effective ing on school infrastructure in the developed and education intervention when high opportunity cost keeps developing world. The majority of studies in develop- children from attending school. There are a number of ing countries have found a positive impact of these conditional cash transfer initiatives in Latin America, investments in school facilities on learning outcomes including programs in Brazil and Colombia, but perhaps (Hanushek 1995). Two recent papers on Indonesia and the best known is PROGRESA in Mexico. The educa- Peru confirm that there can be high returns to invest- tion component of PROGRESA provides monetary I ments in school infrastructure. In addition, both stud- transfers to families that are contingent on the regular ies make obvious the value of impact evaluations in attendance of children at school. So as to offset the oppor- assessing the effectiveness of education interventions, and tunity cost of schooling, the benefit levels rise with the in designing future programs and policies. grade a student is attending, and are highest in secondary Between 1992 and 1998, the Peruvian Social Fund school. Benefits are also higher for girls than for boys, (FONCODES) spent in excess of US$100 million build- in recognition ofthe lower secondary school attendance ing and renovating school infrastructure, mainly primary I rates among girls. PROGRESA has been shown to have schools, in rural areas in Peru. Paxson and Schady (2002) significant effects on enrollment: on average, poor chil- show that these funds were generally well targeted to dren who benefited from PROGRESA completed almost poor communities, and had a large impact on the enroll- one-half more years of schooling than they would have ment levels of young children aged 6 to 11. They esti- in the absence of the program, with somewhat larger mate that every sol (about 60 cents of a U.S. dollar, at impacts among girls. the December 1992 exchange rate) spent on school infra- PROGRESA also shows the importance of evaluation I structure increased enrollment rates by between 0.75 and in program design. In its first phase, once a list of eli- 2.5 percentage points in districts with an initial enroll- gible poor communities had been developed, commu- ment rate of 85 percent. These are large impacts, con- nities were selected to receive PROGRESA transfers by sidering that about a quarter of children lived in districts randomization. This design allowed for highly credible with enrollment rates equal to or less than 85 percent evaluation of the impact of PROGRESA on schooling in 1993. outcomes. Duflo (2001) estimates the impact of a program that built about 61,000 schools in Indonesia between 1973 Source: Schultz (2001); Behrman, Sengupta and Todd (2001). and 1978. She finds large improvements in school attain- ment and in the wages of adults exposed to the program. much incentive to seek more schooling. The demand for edu- Duflo estimates that each primary school constructed per cated workers by firms is often determined by the degree to 1,000 children led to an average increase of between 0.12 which there is technological change, so reforms to liberal- and 0.19 years of education, and a 1.5 to 2.7 percent ize trade regimes, encourage FDI, and facilitate licensing of increase in wages. technologies may therefore unleash the demand for school- ing.14 Firms will also be willing to pay secondary school Source: Hanushek (1995); Paxson and Schady (2002); Duflo (2001). graduates higher wages if there are improvements in school quality-improvements that raise the productivity of sec- ondary school graduates.'5 And where the opportunity cost Often, however, the problem appears to be that the of schooling is high, conditional cash transfer programs such demand for secondary schooling is low. If firms do not as PROGRESA in Mexico and Bolsa Escola in Brazil can be demand workers with a secondary school education, the used to subsidize education and successfully raise enrollments returns will be low, and young men and women will not have (see box 4 3). 16 94 CLOSING THE SKILLS GAP EDUCATION POLICIES Understanding the Education Market: suggesting a large unmet demand by firms for these most Universities skilled workers. This section discusses how Latin American Enrollments in tertiary education in Latin America have countries can best put in place policies that will allow more grown steadily since the 1950s, and very rapidly in some students to access university education. countries since the 1980s. Tables 4.8 and 4 9 show gross ter- tiary enrollment rates for a sample of Latin American coun- A Segmented Market tries Table 4 8 shows that between 1980 and 1995 Until mid-century, university enrollment was essentially an enrollment rates more than doubled in Chile and Colombia, affair of the public sector for most countries in Latin Amer- and increased substantially in Argentina, Panama, and Peru ica. Since then, there have been important changes 17 Figure Despite these increases in enrollment, the relative wages of 4 10 shows that the increase in overall enrollments has been workers with university education have continued to grow, accompanied by a rising market share of private institutions TABLE 4 8 Tertiary Enrollments Have Been Growing since the 1970s... LATIN AMERICA, GROSS RATES OF ENROLLMENT IN TERTIARY EDUCATION, 1970-97 COUNTRY 1970 1975 1980 1985 1990 1995 1996-97 Argentina 13 4 26 6 218 35 7 38 1 36 2 - Brazil 4 7 10 I 11 1 103 112 14 - Chile 91 148 123 156 213 282 315 Colombia - - 86 10 9 13 4 15i - Jamaica - - 6 7 44 68 7 8 - Mexico 54 102 143 159 145 160 - Nicaragua - - 12 4 88 8 2 115 - Panama - - 20 8 24 5 21 5 30 0 - Paraguay - - 8 6 9 1 8 3 10 I - Peru - - 17 3 22 4 30 4 27 1 - -Data not available Source UNESCO and World Bank database TABLE 4 9 ... And Have Continued to Grow in the 1990s LATIN AMERICA, GROSS RATES OF ENROLLMENT IN TERTIARY EDUCATION, 1990-97 COUNTRY 1990 i99i 1992 1993 1994 1995 1996 1997 Argentina - 381 - - 36 2 - Bolivia 213 217 - - - - BraLil 112 112 109 11 1 113 - 14 5 Chile - 213 24 2 26 5 27 4 28 2 30 3 315 Colombia 13 4 14 0 14 6 14 7 154 15 5 167 Costa Rica 26 9 27 6 29 4 29 9 303 - - Cuba 209 198 181 167 139 127 124 Dominican Republic - - - - - - 22 9 Ecuador 200 - - - - ElSalvador 159 I68 172 170 182 189 178 Guatemala - - 8 3 8 1 8 4 8 5 - Haiti - - - - - - Honduras 8 9 8 9 9 2 9 0 10.0 Mexico 14 5 14 1 13 6 13 9 14 3 15 3 160 Nicaragua 8 2 8 1 8 9 - - 11 5 11 5 11 8 Panama 21 5 23 4 25 3 27 3 27 2 30 0 315 Paraguay 8 3 - - 10 3 10 1 10 1 10 3 Peru 30 4 32 0 315 28 0 26 8 27 1 25 7 25 8 Uruguay 29 9 30 1 27 2 - - - 29 5 - Venezuela, RB 29 0 28 5 - - -Data not available Sounoe World Bank 2001a 95 = 0 (A~~~~~~~- Share of Market Share of Market Share of Marker 0 0 0 0 0 0 0 0 00 0 C 0 0 0 0 0 0 C 0 0o0 0 0 o 0 0 o 0 o u w 0 - C° o _ - _ >a a 0 ~~~- N) - . C,'- 0-J 0_________________ 1950 - 1950 :r 1950 2 1956 1955 ~ 1955 1956 1960 1960 > 1962 1965 1965 r 1968 x 97- 19750(: in 1980 \ 1970- >> 1980 v = 0 - 1968 1975 1970 1980 ~ ~ ~ ~~~~~~01980 a 1980: 19741780195d D C) 1986 1985 \ 1985 3 ° 1992 V 1990 1990 o L~~~~~~~;J~~~~~JO ~~~1995 1995 E ° 1998 tr 2000 2000 _ Share of Market Share of Market Share of Market 0 00 0 0 0 00 00 0 0 0 0~ 0 0 ° - sJ w v _- a C o 0 0 0 0 0 0 0 D 9 .0 w as -s Oa 1950 0-- C ) z Go 1950 1950 1955 1956 1956 1960 \ 8 1968 \ 1962 1975 ) a 1962 1965 1 2 ~~~~~~~~~~~~~~~1968 1970 a 1968 1975 1974 17 1980 1980 a 1980 1985 - ' 1986 1986 1990 19 1992~~~~~~~~~~~19 1995 1998 1998 2 0 0 0 _ _ _ _ _ _ _ _ _ _ _ _ _ __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ CLOSING THE SKILLS GAP EDUCATION POLICIES FIGURE 4 11 Private Tertiary Education Is More Expensive to Households Direct Costs for User (Tuition) .t 100 < __ X80- _i _M r0 a 7 //bhc o 60 1<|i 0° 50 / o Cil I- t, 40-/ 1 X30-// i1 20 */i|l1 10- H O iX . . - - ~~~~~~~~~~~~~Private Argentina- Public Colombia Source Schwartzman (2002), World Bank (2002) for Colombia in every country for which data are available. In many coun- demand for a limited number of seats in public universities, tries, the rising demand for university education prompted along with afallang demand for seats in some private uni- governments to deregulate the university market, abolish- versities. In Brazil, the number of applicants per seat in ing what had been a public sector monopoly, and allowing public universities grew from 5.7 to 8 9, while that in pri- for some competition between private and public providers. vate universities fell from 2.9 to 1.9 between 1990 and 2000. The private sector now accounts for more than half of the By some estimates, more than half the seats in private umn- university market in Brazil and Colombia, and for more than versities in Colombia are unfilled. These numbers are simple a third in Chile and Peru. averages. In practice, there is a great deal of heterogeneity in Private providers generally charge much higher fees than providers, especially those in the private sector, in terms of public universities In figure 4.11, we use administrative data the types of institutions, diplomas and disciplines, as well to calculate these differences. The figure shows that tuition as in the quality of the services they offer. As a result, the costs in Argentina and Brazil are essentially zero in the public demand for university slots in the most prestigious private sector, but large in the private sector. Tuition payments in universities in virtually all Latin American countries con- public universities are substantial in Chile and Colombia, tinues to be high. two countries that have reformed their university systems, When there is excess demand, rationing becomes neces- although they are still much lower than in the private sector sary. Private universities have relied on a combination of high (especially in Colombia).`8 Similar estimates based on the fees and entrance exams, and public universities on a com- household survey data for Peru, Guatemala, and Mexico also bination of entrance exams and quotas to assign seats i9 All suggest that the difference in cost between the public and of these make it more difficult for poorer students to attend private sectors is large, especially once costs above and university: Poorer students have fewer resources, are less beyond tuition are taken into account. likely to have access to credit, will generally have attended In most countries in Latin America, then, there are large lower quality secondary schools, and will therefore do worse differences in fees between the public and private univer- on university entrance exams. Despite the increase in aggre- sity systems. These differences in costs have led to a rising gate enrollments, university education is often something 97 CLOSING THE GAP FIGURE 4.12 Conditional on Enrollment, Students from Poorer Households Are Less Likely to Attend Private Universities 0.7- 0.6 0.5- ,04 - 0.4 -~0.3 0 ~0.2- 0 'u 0.1U 0 Peru 1997 Guatemala 2000 Mexico 1998 U Quartile 1 U Quartile 2 E Quartile 3 10 Quartile 4 Sourcesr: Peru: Encuesta Nacional de Hogares (ENAHO); Guatemala: Encuesta de Condiciones de Vida (ENCO VI); Mexico: Encuesta Nacional de Ingresos y Gastos de Los Hogares (ENIGH). that only the better-off can afford in Latin America. Seventy suggests it is a good economic investment in Brazil, Chile, percent of all university students in Brazil, and 50 percent and Colombia-even with a high discount rate of 10 per- of those in Colombia, belong to the highest income quin- cent. In Argentina, where the returns to university educa- tile; a further 20 percent in both countries comes from the tion are lower, the picture is less clear, and is sensitive to second-highest income quintile. Moreover, poor students the choice of discount rate. who attend university are less likely to attend private uni- Low university enrollments by poor households in spite versities. Figure 4.12 shows that, of those who enroll, vir- of the high returns to tertiar-y education suggest that liq- tually no students in the poorest consumption quartile attend uidity constraints may be binding. A background paper for private university in Mexico, compared to about 30 percent this report by Jacoby and Skoufias (2002) uses household in the richest quartile. In Guatemala and Peru there is also survey data to estimate the extent to which liquidity con- a clear relationship between household welfare and the prob- straints are an impediment to university enrollment in ability that enrolled students attend private university. Mexico. The authors distinguish between 'transitory" Youths from poorer households are much less likely to go income (proxied by current income), 'permanent" income to university and, for those who do, are much less likely to (proxied by measures of family background and education), attend a private institution, and the opportunity cost of schooling (approximated by vari- ation in local labor market conditions, both across states, and Direct Private Costs, Opportunity Cost, Information over time, such as before and after the Mexican Peso Crisis). Flows, and the Demand-side Constraints to University Jacoby and Skoufias show that, conditional on graduation Expansion: the Story ofthe Private Sector from secondary school, current parental income is a good pre- Unlike secondary school, the pay-offs to tertiary educa- dictor of tertiary attendance. In other words, households with tion are high in most Latin American countries. Of similar permanent income, and facing similar labor market course, the direct costs of private university to individ- conditions, are less lively to send a child to university if they uals are also high. But a calculation of the net present had a bad year. These differences in tertiary attendance rates value of tertiary education that combines the data on costs can then plausibly be attributed to the effect of binding in figure 4.11 with estimates of the returns in table 4.6 credit constraints. 98 CLOSING THE SKILLS GAP EDUCATION POLICIES In addition to credit constraints, information problems Science, Humanities, and the Content of Tertiary may help explain the low demand for private universities rel- Education in Latin America ative to public universities in the region. Some households, In addition to concerns about coverage, there is considerable perhaps especially households with lower mean education debate about the content of instruction at tertiary institutions levels, may only be partially aware of the returns to educa- in Latin America-in particular, about the extent to which tion In the United States, a public information campaign universities are producing "too many" students in the about the economic benefits associated with education humanities and social sciences, and "too few" students in the appears to have been one of the reasons for the expansion in hard sciences and engineering There is no evidence that the high school enrollments in the first half of the century (see "knowledge economy" has particularly increased the demand box 4.1). The large increase in the number of private for scientists and engineers-in Latin America or elsewhere. providers in tertiary education may also make it difficult for Rather, it is the returns to general problem-solving skills, households to discriminate between high quality and low as well'as to (harder to measure) attributes such as flexibil- quality establishments and programs.20 The quality of uni- ity, adaptability, and "client skills" that have increased sub- versity education is hard to measure without information on stantially wage differentials by university, or an exit exam for univer- sity graduates Faced with uncertainty about the quality of General, Problem-Solving Skills are Most Valuable providers, risk-averse households may sensibly under-invest in the "New Economy" in private tertiary education. In the United States, increases in the demand for skilled workers have been linked to the profound effects of com- Limited Resources, Low Efficiency, and the Suipply-side puterization and other technological changes on workplace Constraints to University Expansion: the Story of the organization In these countries, the "new" skills demanded Public Sector entail the ability to perform complex tasks-tasks that limited budgets are the main constraint to expanding require non-algorithmic reasoning, cognitive work, and man- enrollment in public universities. The gap in secondary agerial skills, rather than increases in the demand for very school attainment, the need to upgrade skills from the specialized "scientific" skills (see box 4.4) The wages of com- bottom up, and the fact that the bulk of university entrants puter programmers and others with scientific or "new econ- in Latin America are (and will continue to be) relatively well- omy" skills have not gone up disproportionately in the off all make it hard to justify an aggregate expansion in the United States (Card and DiNardo 2002). We find a similar budget of public universities. pattern in Mexico, one country in which the labor force More money may not be the solution, but cost recovery survey asks workers about the subject they studied in uni- and better use of existing funds would help. In many Latin versity. There is a small skill premium to scientific studies American countries, public universities are inefficient in at the secondary level, and a small premium to non-scien- terms of the amount of time it takes a given student to grad- tific studies at the tertiary level. As figure 4 13 shows, how- uate This is a serious drain on resources, as every spot occu- ever, these premia did not change between 1987 and 1999, pied by a student who takes more than the required years suggesting that the increase in demand for workers with ter- to complete a program could have been filled by another tiary education in Mexico has favored scientists and human- applicant who was turned away from university In ists alike. Colombia, only somewhere between 50 and 60 percent of Evidence that the demand for "scientific" skills has not an entering class graduates on time in both the public and gone up disproportionately can also be found in employer private sectors. This "internal efficiency" of the public sector surveys. Recent surveys in Thailand and Costa Rica both will almost certainly be lower in countries where tuition at suggest that employers are much more concerned with the public universities is essentially nonexistent Anecdotal evi- general education level and the ability of their employees dence suggests that many students at the Universidad to learn on the job than with specific skills such as math- Nacional Aut6noma de M'lxzco (UNAM) and similar free-of- ematics and computer proficiency-even among firms in the charge universities are often enrolled for eight years or technology sector in Costa Rica."i In Thailand, fully 80 longer percent of employers in the manufacturing sector and 99 CLOSING THE GAP BOX 4.4 Now IOfnnaflon Technologl Has Changed the Woriplace: IEvidence from the CIECD In a background paper commissioned for this report, by job rotation, integration of tasks, and learning across Carnoy (2002) considers the changes in production that tasks). Computerized access to information has provided have taken place as a result of the "new information econ- employees with better information about the work of others omy" and the relevance they have for Latin American coun- within an organization, as well as about customers. As a tries One of the features he stresses is changes in workplace result, employees have become more involved with each organization: "The organization of production and of eco- others' work, and have been better able to respond to nomic activity.., has changed from mass standardized pro- rapidly changing customer needs. Also, the introduction duction to flexible customized production, and from of flexible machine tools and programmable equipment has vertically integrated large organizations to vertical disin- made the capital equipment itself more versatile, and better tegration and horizontal networks between economic capable of performing a variety of tasks. In response, work- units." Moreover, Carnoy reasons that "the effect of mdi- ers using this capital equipment have become more versa- vidualization and differentiation (of production) is to sep- tile as well. Finally, a steady increase in the human capital arate more and more workers from 'permanent' ftill-time of workers in OECD countries has resulted in "human cap- jobs in stable businesses that characterized post-World ital deepening" (allowing workers to better perform par- War II development in Europe, Japan, the United States, ticular tasks), as well as "human capital widening" and other industrialized countries." He concludes that "as (allowing firms to reorganize workers so that they multi- post-industrial economies and governments adjust to new task). Lindbeck and Snower argue that a variety of realities, employment growth, not displacement, domi- managerial innovations-for example, Total Quality Man- nates. There will be plenty of jobs in the future, and many agement (TQM) and Just-in-time production-exploit of them will be high-paying jobs." these changes in workplace organization. The earliest research on the effect of ICT on workplace Timothy Bresnahan, Erik Brynjofsson, and Lorin Hitt organization in the United States was an analysis of Tam- (2002) also focus on the profound impact of information many Bank by Levy and Murnane (1996). Since then, technology on workplace organization: "Firms do not much of the econometric evidence for the United States simply plug in computers or telecommunications equip- has focused on the increasing returns to worker charac- ment and achieve service quality or efficiency gains. Instead teristics other than education, experience, gender, and they go through a process of organizational redesign and industry affiliation in the past two decades. These char- make substantial changes to their product and service mix." acteristics, unobservable to the econometrician but Based on a combination of formal modeling, econometric observed by employers, are often though to capture traits work, and case study interviews with firm managers, they such as "people skills," the capacity to work in teams, argue that there are profound complementarities between multi-task, work successfully without supervision, show information technology, workplace organization, and initiative, and be entrepreneurial. Two recent papers go product innovation-and that it is these complementar- back to Levy and Murnane's focus on the complementar- ities that drive skill-biased technological change. Fre- ity between ICT and organizational changes within firms. quently, investment in new ICT involves a second round They suggest that recent technological changes in infor- of innovation. Unlike direct investments in ICT, which can mation and communication technology have profoundly (to a large extent) be bought ready-made off the shelf, these redefined the ways in which firms employ workers in dif- second-order "organizational co-inventions" are highly ferent tasks. firm-specific. The result of this is much greater hetero- Assar Lindbeck and Dennis Snower (2000) argue that geneity and turnover of firms, a greater degree of market firms in industrialized countries are restucturing to change instability, and profound changes in worker demand. from "Tayloristic" organizarions (characterized by special- Simple tasks (record-keeping, simple calculating, indi- ization of tasks) to "holistic" organizations (characterized vidual transactions) are automated, while the premium on 100 CLOSING THE SKILLS GAP EDUCATION POLICIES complex tasks (tasks that require non-algorithmic rea- enabled organizational change, and computer-enabled soning, cognitive work, and managerial skills) increases forms of output. dramatically. Bresnahan and his co-authors conclude that the information revolution has actually resulted in a "clus- Source: Carnoy (2002); Levy and Murnane (1996), Llndbeck and Snower ter of inventions," including computerization, computer- (2000); Bresnahan, Brynjoifsson, and Hirt (2002) FIGURE4 13 The Wages of Workers with "Scientific" Degrees Relative to Those with "Humanistic" Degrees in Mexico Have Not Changed in the 1990s Real Hourly Wages by Type and Level of Education Mexico 3 75 ~~ ~~~ -- - - - - --. - ------- --.---.- So2 21 - Seondary - Other : 15 --- Secondary - Scientific 0 -- - Tertiary - Other - Tertiary - Scientific l. 1 05 87 88 89 90 91 92 93 94 95 96 97 98 99 Year of Survey Source S§nchez-P§ramo and Schady, based on the ENEU surveys 100 percent in the service sector considered the "ability to Putting it Together: Supply, Demand, and Policies learn" important, compared to only 7 percent in manufac- to Reform Tertiary Education in Latin America turing and 60 percent in the service sector who stressed High quality tertiary education in Latin America pays off- computer literacy. Many more employers listed reading as for individuals, who will earn higher wages, and for coun- an important skill than mathematics, again in both man- tries, which will enjoy higher growth rates. How best to ufacturing and non-manufacturing jobs. In Costa Rica, reform tle university system in a given country will employers in the technology sector were twice as likely to (obviously) depend on its institutional setup-but some stress "learning speed" as "specific knowledge." In an evolv- policy reforms apply to a large number of countries ing workplace, with a rapid pace of technological change, The fundamental challenge for the public university schooling that teaches workers to "learn how to learn" will system is rnaking better use of existing resources This entails be most valuable. a combination of policies: cost recovery, increasing the fraction 101 CLOSING THE GAP FIGURE 4 14 Cost Recovery in Public Universities Is Low in Many Latin American Countries Ireland B-- Chile r 'r - - United States Jamaica '.' Ecuador . Korea - _ inu _ - Vietnam - - 1 Colombia F 7 I. - Spain - _ Costa Rica L .C Councre bdoesi I Other Countries lndoneaia _ _ _U China Thailand Japan Honduras Guatemala Bolivia France Argentina Brazil 0 5 10 15 20 25 30 Percentage of Costs Borne by Student Source World Bank (2002) of students admitted to university, improvements sn efficiency, egregious example is Brazil, where the unit costs of the and better allocation of resources across publsc universities. public university system are staggering. Per-student costs Cost recovery through fees is very low in some Latin Amer- in 1998 are estimated to have been above US$8,500, about ican countries. Figure 4.14 shows the fraction of costs borne 6 tsmes the per-student costs in other "expenssve" public sys- by students in public universities in a cross-section of coun- tems such as those in Venezuela and Costa Rica, and almost tries. The figure shows that in some Latin American coun- 10 times those in Argentina.22 An alternative summary mea- tries, such as Chile, Jamaica, Ecuador, Colombia, and Costa sure of the real economic cost of education is expenditure Rica, this fraction is comparable to that in the United States, per student as a proportion of GDP per capita By this mea- Ireland, or Korea. In other countries, such as Honduras, sure, educating students is 10 times as expensive in federal Guatemala, and Bolivsa, it is very low, and in Argentina and universities in Brazil as it is in Spain, and 5 times as expen- Brazil it is essentially zero. Low cost recovery has serious sive as in the United States.23 Reducsng per-student costs implications for the coverage, quality, and sustainability of should clearly be a priority for the public university system the publsc umnversity system. Where politically possible, Latin in Brazil-even if this entails some drop in quality. Simply American countrses with low degrees of cost recovery in the opening the doors of the federal universities to a larger public system should attempt to introduce fees. Increases in number of students may be the single most important reform tuition and other fees should be implemented hand in hand of the university system in Brazil. with a program of student loans or well-targeted scholarships. The third set of reforms that are promising for the public Reforms to increase tuition tend to be politically diffi- university system are those seeking to improve internal effi- cult to implement (as shown by the crippling strike at the ciency. Performance-based financing, which ties transfer of UNAM in Mexico). Reducing unit costs is an alternative way public monies to measures such as per-student cost, gradu- of increasing the coverage of the public university system. ation rates, time to graduation, and others hold promise. In Per-student costs are alarmingly high in the public univer- Chile, for example, the funding formula whereby resources sity system in some Latin American countries. The most are transferred to public and private universities sncorporates 102 CLOSING THE SKILLS GAP EDUCATION POLICIES the fraction of the 27,500 highest-scoring students in a survey suggest that, of those enrolled, no student in the national academic test which a given university can attract poorest income quartile received a scholarship, compared The fundamental challenge for an expansion of the pri- to 8.3 percent of students in the richest income quartile vate university system in most Latin American countries These figures are noteworthy because the probability of is overcoming the liquidity constraints faced by poor house- enrollment was itself about 40 times higher for households holds. The scholarship programs that exist in most Latin in the richest income quartile than those in the poorest American countries cover only a very small fraction of stu- In many Latin American countries, well-targeted schol- dents, and are often regressive. In Mexico, the probability arships for able, poor students and student loan programs that a student enrolled in university receives a scholarship both hold considerable promise. Box 4.5 provides some rises with income Estimates based on the 1998 ENIGH recommendations for setting up effective student loan BOX 4 5 Student Loan Schemes Many hopeful graduates from secondary school are keen has been in operation since 1981 (Blom and Paqueo 2002). on continuing on to university. In countries where uni- SOFES is a relatively new program, less than five years old, versity education is not free of charge, however, they may which provides loans to about 10,000 students per year. be unable to do so because of the upfront living and tuition In the scheme, private universities buy shares into the stu- costs. University education is clearly a good investment dent loan company, which is capitalized by the government for many secondary school graduates. One might there- of Mexico and the World Bank. Participating universities fore expect the -private sector to offer student loans. But on-lend SOFES funds to students on unsubsidized terms. information asymmetries between the lender and borrower The iunversities are responsible for all interaction with the (for example, about a student's ability, or the effort he or students. If more than 10 percent of the portfolio-is non- she will put into his studies), and the absence of seizable performing, the university either replenishes SOFES with collateral (unlike loans on a car, or a house, which can be the lost amount or becomes ineligible for further funds. repossessed by a bank) limit private sector provision in the This insiitiutlonal setup features several aspects that should absence of some government intervention This has resulted help achieve sustainability, including (i) clearly stated in provision by the public sector in many countries. objective of financial sustainability, which allows admin- Public sector efforts to provide student loans can be a istrators to provide loans without subsidy, and provides serious drain on the public purse, for at least two reasons. them with credible corrective measures in case of non-ser- The first is that interest rates are often subsidized, and the vicing of debt; (ii) strong incentives for collection in insti- second is that repayment rates are often low. A systematic tutional setup, (ill) frequent information-gathering about (if by now outdated) review of student loan programs in borrower s' prospects for repayrnbent; (iv) frequent interaction Latin America by Albrect and Ziderman (1991) concluded with the borrower, which helps build a culture of repayment; that only 25 percent of costs are recovered, compared to (v) monetary incentives for the education institutions to pro- over 70 percent ih Scandinavian countries. Low-cost recov- vide relevant, quality education to the student; and (vi) infu- ery provides a substantial subsidy to beneficiaries. Since sion of private capital, which is critical given the high student loan schemes are generally not targeted, student potential demand for student financing in Mexico. loan schemes in many countries, essentially work like To date, SOFES has had single-digit default rates. The poorly targeted scholarships. SOFES prograrn has only been in operation for five years, Although the bulk of the experience with student loan and is therefore relatively new, and perhaps untested. Yet schemes in Latin America has not been a success, there are SOFES holds important lessons for other countries devel- exceptions. The Soczedad de Fomento a la Educari6n Superior oping student loan schemes in Latin America. (SOFES) in Mexico is an interesting case. Mexico has a long-standing -tradition with student loan schemes, Souricr Albrecht and Ziderman (1991}, Blom (2002), Blom and Paqueo including a program in the northern state of Sonora which (2002). 103 CLOSING THE GAP schemes, as well as a description of an innovative program designed with a clear understanding of the nature of in Mexico. constraints-specifically, whether these are primarily The second problem of the private university market, on the demand or supply side of the education market. namely the information asymmetry between providers and Institutional reforms that better align incentives with consumers, could be minimized with an accreditation system outcomes, including capitation, teacher compensation for private universities, and an information campaign that schemes that reward better teachers, parent partici- provides data to the public about the relative performance pation in schools, and school competition may also of various tertiary institutions, both public and private, in hold promise. terms of learning and labor market outcomes. The challenge o Low demand for more education by individuals is a of an accreditation system is providing minimum standards serious concern at the secondary school level. Insuffi- without limiting entry into the market by legitimate cient demand for schooling is related to insufficient providers Chile is a good example of a well-functioning demand by firms, to the low quality of education, and accreditation board in Latin America (see box 4.6). The chal- to the high opportunity cost of schooling. Policies that lenge of a public information campaign about university per- promote technological change have the potential to formance, whether in terms of test scores of graduates or their increase the returns to schooling and the willingness labor market performance is to disseminate comprehensible by young men and women to defer entrance into the information widely The Provio system in Brazil tests every labor market Improvements in the quality of sec- graduating student in over a dozen disciplines and publishes ondary school should also increase the demand for edu- mean results by institution. This system has allowed poten- cation. Finally, programs that subsidize the cost of tial students and their families to make reasonable inferences schooling, such as conditional cash transfers, can also about the quality of instruction (see box 4.6). be effective. o Insufficient supply of secondary school infrastructure CDo¢nTcsionI: I ori3es for Education UJpgrading is a problem in rural areas in some Latin American Din La2un Amerca countries. In these settings, programs to increase the This chapter has discussed the main constraints to skill availability of schooling facilities can hold promise. upgrading through the formal schooling system in Latin o The main constraint to an expansion of the private America, as well as the broad direction of policies to over- university system is on the demand side of the market. come them. Some of the most important conclusions of the There is high potential demand for private university chapter include the following because of the high returns, but this does not neces- sarily translate into high effective demand because of o Skill-biased technological change means there is a liquidity constraints and information asymmetries. bigger premium to education in terms of wages, pro- Student loan and targeted scholarship programs hold ductivity, and growth now than before. Latin American great promise, if they are well designed. Providing countries therefore need to close the education gap households with information about the quality of with East Asia and the OECD qickly. private providers may also help them make better o Countries that have most successfully closed the skill choices. gap have done so from the bottom up. Many Latin o The main constraint to an expansion of the public uni- American countries have given insufficient attention versity system is on the supply side of the market. to secondary education in their education transitions. Increasing aggregate budgets to public universities is This is perhaps the most serious constraint to an "intel- generally not a feasible or desirable solution. Cost ligent integration" of Latin American countries into recovery, increases in enrollment for a given budget, the global economy and policies that tie the transfer of public resources to o The main problem with Latin American education sys- performance should all be considered. tems is not insufficient aggregate spending, but the quality of spending. Appropriate policies for sec- In much of Latin America, firms have increased their ondary school and tertiary education can only be demand for more educated workers in the last two decades. 104 CLOSING TIHE SKILLS GAP EDUCATION POLICIES BOX 4 6 Minimizing Infornation Problems about Private University Providers: Experiences from Latin America Lack of information about the quality of private univer- about quality in tertiary education. This standard, nation- sity providers is a serious constraint to expansion of the wide exam measures the performance of graduates in over coverage of the university system in Latin America. There a dozen disciplines. The results are disaggregated by insti- are two approaches to minimizing information asymme- tution, and published. As such they serve as a defacto com- tries between providers and consumers accreditation of parative indicator of the quality of graduates, and, by private providers and the provision of information about inference, the quality of instruction and education. learning or labor market outcomes. Since its Inception in 1996, the Provao has grown both Chile has a well-functioning accreditation system devel- in coverage and influence. The first exam covered only three oped as a response to an increasingly heterogeneous offer disciplines (Administration, Engineering, and Law). The of private tertiary education. This quality assurance system very existence of the Provao provoked strong opposition has proved successful in two ways: (i) establishing a qual- from segments of both the students and the professoriate, ity stamp through external institutional and programmatic including boycotts and thrcats of dlsruptions at exam sites. evaluation procedures, and (ii) creating a culture of self- Such opposition has not continued, especially given the evaluation in which institutions continuously benchmark interest of the press and the general public in the results teacher qualifications, curricula, and learning facilities with In its fourth year, the Provao is now widely accepted. competitors and thus stimulate competition. The accred- The most notable effect of the Provao has been to pro- itation is voluntary but more than 35 universities and 106 vide much greater information on the quality of individ- programs have or are currently applying for accreditation, ual degree programs to potential students, thereby which points to the importance of being accredited. creating more savvy educational consumers. Private insti- The reason for the success of the Chilean accreditation tutions, many of which felt wrongly deprived of prestige scheme appears to be related to the fact that the system by the wealthier, research-oriented public universities, now is voluntary, not excessively onerous, and relatively free of have an objective means of demonstrating the quality of bureaucracy. Unfortunately, other accreditation schemes in their course programs. Several well-known public uni- the region have sometimes been "captured" by entrenched versities have degree programs whose Provao scores were interests such as the administrators of public universities disappointing; these are now struggling to save their rep- seeking to protect their market position In Venezuela, utations as the leaders in the field. IESA, a dynamic private institute of business administra- Students are voting with their feet thanks to the Provao tion, had to wait several years to receive the official Applicants now routinely inquire about Provao perfor- approval from the Council of Rectors for a new MBA pro- mance, and schools that do well highlight this information gram designed and delivered together with the Harvard in their informational literature. Those private institutions Business School. The Brazilian Institute of Applied Tech- whose scores have been consistently high have almost uni- nology (IAT), the most prestigious private engineering versally reported increasing applications Also, private uni- school in the country, faced similar problems getting versities, which have now proven their quality, are accreditation. And in Nicaragua, the University of Mobile attracting talented professors away from public instiLtutions. in Alabama was denied a license to operate by the Coun- "Labor market observatories" provide an interesting cil of Rectors keen on protecting the public universities complement to graduation tests. The Chilean government, from competition. with support from the World Bank, is in the process of The second way countries have attempted to correct developing such an observatory to regularly provide con- information asymmetries is through standardized gradua- sumers with information on the labor market performance tion tests, or information about the wages and employment of graduates from different tertiary institutions. prospects of university graduates. In Brazil, the Exame Nacional dos Cursos, or Provio, has raised pubhc awareness Source Blom and Hoim-Nielsen (2002), World Bank (2000, 2002) 105 CLOSING THE GAP Having a more educated workforce is a sine qua non of com- More resources may help, but often they make little differ- petitiveness and high productivity. Designing policies that ence. This is not to say that education expenditures do not will allow young men and women to seek more schooling, matter, but rather that how money is spent matters a great and to take advantage of the higher wages paid to educated deal. Policies should focus on alleviating a given constraint workers is the challenge facing Latin American education to the acquisition of more schooling, and on improving systems. incentives. Sometimes, effective policies may be under the How best to upgrade the skill levels of a population means purview of the ministry of education-for example, pro- different things for different countries. Haiti and some coun- grams to build school infrastructure, provide teaching mate- tries in Central America should focus on increasing the cov- rials, or regulate the university system. At other times, erage of the primary schooling system. For Brazil, continued reforms to the safety net could be important, as with the cre- increases in primary education and secondary education ation or expansion of conditional cash transfer programs that should be the priority. Secondary education should also be reduce the opportunity cost of schooling. And in yet other the priority for most other Latin American countries- circumstances policies that increase the demand for more especially countries such as Venezuela, Costa Rica, Colombia, educated workers, including trade reform and legislation that and others with alarmingly low secondary attainment for encourages the flow of technologies from abroad may hold their income levels. Jamaica, Trinidad and Tobago, and other the biggest promise. countries in the English-speaking Caribbean have managed Impressive secondary school attainment, but comparatively Dd $ I This is not to say that Latin American countries do not have low enrollments at tertiary institutions. Chile is at or very problems with their primary school systems-in terms of both cov- close to the world technology frontier in a number of sec- erage and quality Indeed, many countries, especially some of rhe poorer tors such as fruit cultivation, vineyards, and salmon farm- counrries in Central America, are still far from achieving 100 percent ing, yet it has only 6 percent as many Ph.D. scientists per primary enrollment levels The point is, rather, that the region as a million people as Korea, and 2.5 percent as many as Fin- whole does not have a primary enrollment deficit for its income level, land.24 In addition to necessary Improvements i secondary whereas it has deep deficits in secondary and tertiary education school and university, the comparatively low number of 2 In Brazil, Mexico, Peru, and Venezuela, 40 percent or more of the population had no schooling whatsoever, while the comparable fig- Ph.D. students, especially In the hard sciences, may be a real ures in Korea, Malaysia, Singapore, and many Central American coun- source of concern in Chile. tries were in excess of 60 percent The relative emphasis on different policies will vary 3 This is equivalent to sweeping out any fixed effects in a regres- depending on the stage at which countries find themselves sion However, if there is a lot of measurement error in these data, the in their education transitions. Because there is a lot of inter- first-differenced results could be less, nor more, accurate than the results from a regression in levels dependence across education levels, however, governments frmaegsioinlvs dependence across educaionlvelsh4 Authors' calculations, based on data from UNESCO and the must often work at many levels in the education system World Bank These ratios are an average of the ratios for the 1980-95 Consider a poorly performing university system. Part of the period problem may be that secondary school graduates do not have 5. Mariscal and Sokoloff, in Haber (2000). the content knowledge and learning ability to be successful 6 Bolivia, Costa Rica, Ecuador, and Venezuela all have higher mean in university. Policies that improve secondary school qual- years of schooling than Jamaica, but much higher fractions of the pop- ity may therefore decrease repetition at the university level ulation with no schooling whatsoever In all of the East Asian tiger countries, the "diamond" distribution of educational attainment is Consider also a poorly performing secondary school system, bunched around the middle, with large fractions of the workforce with many dropouts before graduation. Part of the problem having secondary schooling may be that children leave school because they consider it 7 In a simple formulation, the opportunity cost would be given highly unlikely that they will be accepted by a university by the expected wage, defined as the product of the observed wage (or that they will be able to pay for it). Policies that increase and the probability of employment 8 A summary of evidence can be found in de Ferranti and others (2000), on Brazil, see also Neri and Thomas (2000), on Mexico, see therefore reduce dropouts throughout secondary school. Cunningham and Maloney (2000); on Argentina, see Cunningham and A careful understanding of the education market within Maloney (2000), and Espaina, Parandekar, and Savanti (2002); on Peru, a given level is the key to improving schooling outcomes see Schady (2002). 106 CLOSING THE SKILLS GAP EDUCATION POLICIES 9 Authors' calculations, based on the 2000 ENAHO survey 17 Much of this section draws on background notes by Blom and 10 See, for example, Hanushek (1995), Kremer (1995), and Yin (2002), and Frydman (2002) Glewwe (2002) 18 A breakdown of the direct private cost of a year of university I I See, among others, Hsieh and Urquiola (2002); Contreras education by low-priced, middle-priced, and high-priced public and (2001), Gauri (1998), McEwan (2000), and Mizala and Romaguera private universitles in Brazil, Chile, and Mexico shows very large dif- (1999) ferences in these costs in Brazil and Mexico, and much smaller dif- 12 The graphs we present are based on the tabulations in Filmer ferences in Chile See Frydman (2002) (2000) 19 There are exceptions In Argentina, for example, access to public 13 On sheepskin effects, see Lam and Schoeni (1993) and Strauss universities is essentially unrestricted See Frydman (2002, p 8) and Thomas (1996) on Brazil, and Schady (2003) on the Philippines 20 In Brazil, for example, the mean quality is lower and the dis- 14 In some countries, the possibility of international migration persion higher in private than in public universities may also increase the demand for schooling even when the domestic 21 On Thaltand, see Abelmann and others (2001), on Costa Rica, returns to education are low This may be part of the explanation for Garmier (2002) the high observed rates of secondary schooling in countries in the 22 IPEA (2000), Schwartzman (2002, p 25), World Bank (2000, English-speaking Caribbean pp 30-31) 15 Policies that increase the demand for educated workers by firms 23. The high per-student costs in federal universities in Brazil are, or improve quality will become particularly important if the expan- in part, a reflection of very low ratios of students to academic staff sion of secondary school enrollments places downward pressure on the This ratio (9 2) is lower than that in any other country for which data relative wages of secondary school graduates are available There are more than twice as many studencs per acade- 16 For a review of demand-side financing initiatives in Latin mic staff in Italy, the Netherlands, Switzerland, or Ireland as in Brazlt America, see Patrinos (2002) 24 These figures are from Brunner (2001, p 12) 107 CHAPTER 5 Closing the Skills Gap: Training Policies HE PRECEDING CHAPTER DISCUSSED POLICIES TO CLOSE THE GAP IN RELATIVELY general skills acquired in schools, colleges, and other institutions of learning. This chapter focuses on skills acquired in the workplace, referring to them as "training," though it also addresses vocational-technical education in schools. What is equally important is the shift in the unit of analysis from the household to the firm. A quick look at figure 1.1 in chapter 1 can help in understanding why training is treated specially: education-technology complementarities are fundamental to an understanding of the "knowledge economy," and training and networks play a critical role in this regard. In the most general terms, training may be seen as efforts by firms and workers to cus- tomize the skills acquired in schools and colleges to the specific technologies in use. This emphasis on actively increasing the complementarity between education and technology explains why training is positioned between education on the left and technology on the right in the figure. Gill (2002) explains this approach to the knowledge economy, and the role of education, technology, training, and networks in more detail. This chapter addresses the issue of training policy in Latin America and the Caribbean, using this inter- pretation of training as a device to increase the complementarity between skills and technology. It is not a comprehensive analysis of training in the region; there are competent surveys that meet that criterion, such as CINTERFOR/ILO (2001), Callart (2001), Zuniga (2001), and firm-sponsored training, and between technology and Guerrero (2002). Like many of these assessments, this and firm-sponsored training chapter concludes that the movement of reform in many * Matching this evidence of complementarities to the countries of the region is indeed in the right direction. But design of training systems in Latin America and the there remains a considerable unfinished agenda. Caribbean indicates that in many of the countries there Interpretations of training as a link between education and is considerable potential for increasing the effectiveness technology conform well with the evidence, and have impor- of training systems. The proposed tenet is that "training tant implications for the design of training policy. Using this policy" should be seen as consisting of three components. approach, this chapter makes three main points: eclucation policy, technology policy, and incentives for firms to formally train their workers Education and tech- * A summary of the evidence gleaned from surveys of nology policies are treated in other chapters in this report, firms and households in Latin America and che but policymakers interested in improving training out- Caribbean indicates that firms invest in formal train- comes should view these three systems as symbiotic. ing more when they adopt new technologies and- * In assessing training policy, this chapter argues for an when they do so-more educated workers are more increased emphasis on providing learning rather than likely to receive this training That is, the evidence vocation-specific skills in schools, and a shift away from points to strong complementarities between education publicly sponsored vocational education in schools or 109 CLOSING THE GAP industry-level vocational training institutes to well- general secondary education, from there, they go either designed training incentives to firms. The evidence to firms that provide training upon initiation, or to summarized in chapter 3 and the associated imper- tertiary education. ative to have a workforce capable of facilitating tech- o The American system is similar in that there is no nological progress calls for (a) a shift of vocational "streaming" until after secondary education, but is dif- education from the secondary to the post-secondary ferent in its reliance on postsecondary education for level; (b) a shift in the balance between public and facilitation of transition to work. Students complet- private providers of training toward the latter; and, ing secondary education go on to community colleges, based on reform experience, (c) a shift away from polytechnic institutes, and universities, which provide coercive training legislation toward fiscal incentives both general and professional training. to firms. This chapter also contains general guidance o The French system is marked by vocational stream- on how this can be done, but the specifics are better ing that starts at the secondary level. Students going left to more detailed country-level training policy into (school-based) vocational-technical education are assessments prepared for entry into the labor market, and those going into the humanistic-scientific stream are pre- This chapter first summarizes the training policy design pared for higher education. in countries of Latin America and the Caribbean, contrast- o The German system is based on a long tradition of ing the "Latin American model" of vocational training with apprenticeships; for a (diminishing) majority of sec- other approaches. We then summarize the evidence on skills ondary school students, instruction is "dual," in that and training in Latin America and East Asia, drawn from it consists of a school-based general instruction, and enterprise surveys The determinants of in-service training firm-based occupation-specific training regulated by are then analyzed, drawing mainly on a paper commissioned industry guilds. The German model is distinguished for this report that utilizes firm panel data from Mexico (Tan by a system of qualifications that are regulated by and Lopez-Acevedo, 2002) The issue of lifelong learning is guilds, which provide a broad equivalency between also briefly examined, with illustrative evidence from Brazil, graduates of the academic and the dual subsystems Colombia, and Mexico using household survey data. The con- o The Latin American system of training is a hybrid of cluding section summarizes the tentative policy priorities the French and German systems. In the most simple that, when augmented by country-level analysis, could pro- terms, it is identifiable by its reliance on autonomous vide useful guidance to policymakers engaged in the diffi- vocational training institutes for a set of students com- cult task of improving training systems in the region. pleting basic education as a bridge to the labor market, general (humanistic-scientific) education at the sec- Vocational Education and Training Systems ondary level as a bridge to tertiary education for in Latin America another, and school-based vocational education for the Training policies recognize that training is the most impor- others (Callart 2000) Most secondary-level students nor tant linkage between education and work, and generally aim streamed into the academic tract are expected to receive to strengthen this link. But government efforts to do so have vocational instruction in training institutes under varied considerably. tripartite-employer, labor union, and government- auspices, funded by an earmarked tax on enterprises. Education and Training Systems around the World The streaming thus takes place at an early stage, and Figure 5.1 illustrates different education and training systems there are few opportunities to go back into a stream around the world, based on ILO (1999). At the risk of over- that would allow going on to tertiary education. simplification, and emphasizing that all of these systems are constantly changing, five basic approaches can be distinguished As can be seen in figure 5.1, training systems have to be assessed in relation to education systems, and not in isolation. o The Japanese system may be the most simple in Examining training systems more closely, three types can be design. Students completing basic education go on to distinguished (see table 5 1) cooperative, enterprise-based, 110 CLOSING THE SKILLS GAP TRAINING POLICIES FIGURE 5 1 Education and Training Systems around the World The German System The French System | Basic Education | asic udcation Apprenticeship Secondary Secondary an Secondary and Tech nological Education Vocational- Education Workshop Techmca Schools Tertiary Trir Education Education Labor Market Labor Market The North American System The Japanese System Basic Education ucation Secondary Education, Secondary Educanon including Vocational Subjects Tertiary Education Colleges Education lF4 Vocational Training by Corporations Labor Market Labor Market The Latin American System The Mexican System | ai ducation | asic duco | Vocational | econdary Training Education Vlocationa B svalen c | Secondary _S.hool _ hools | Education I Internship Teriar Education Labor Market Labor Market CLOSING THE GAP TABLE 5 1 A Typology of Training Systems TYPE EXAMPLE COUNTRIES MAIN FEATURES Cooperative Employer-led Germany, AustrLa, Switzerland, Brazil Pressure to train workers resulting from cooperation between employer federations, the state, and labor unions State-led Barbados, Colombia, Jamaica, Paraguay, Pressure to train workers resulting from tripartite cooperation Peru, Venezuela Enterprise-based Corporate Japan Firms take responsibility for training workers, who stay with firms for long periods Voluntarist United States, United Kingdom Little or no institutional pressure on firms to provide training State-driven Demand-led Korea, Singapore, Hong Kong (China), State plays a leading role in coordinating training, but operates in an open Taiwan (China), Malaysia, Cbile and competriive environment Supply-led France, East European countries, Mexico State plays a leading role in providing training, with little or no pressure on employers to train Source LO (1999) and state-driven. Within each of these three categories, there education (regulated or run by the Ministry of Education) are two categories, depending on whether it is the employer or these vocational training institutes (run by employer fed- or government that leads the training effort, whether there erations). For the latter, the prospects for reentry into the are societal or government pressures to train, and whether education system were slim. the role of government is largely provision of training, or Over the next three decades, many countries followed its finance. What is noteworthy is that there are no Latin Brazil's example with similar enterprise-funded vocational American countries where training systems may be classi- training institutes, but with the difference that they were run fied as enterprise-based by governments; in other words, they were public institutions. It is important here to distingwsh between vocational During the 1950s and 1960s, six countries introduced such education and vocational training (see Gill, Fluitman, and institutes: SENA in Colombia (1957), INCE in Venezuela Dar 2000, for more details). While vocational education is (1959), SENATI in Peru (1961), INA in Costa Rica (1963), classroom-based, and should be considered part of the INACAP in Chile and SENAP in Ecuador (1966). In the (government-sponsored or regulated) schooling system, 1970s, more countries introduced such institutes: SNPP in vocational training takes place outside of the formal school- Paraguay (1971), INFO in Honduras and INTECAP in ing cycle. The clearest example of the latter is workplace Guatemala (1972), and IFARHU in Panama (1973) training of workers, sponsored by enterprises. The Latin By the 1990s, vocational training institutes in 14 Latin American system-with its reliance on taxpayer-funded train- American countries that had such systems (Argentina, ing institutes managed by governments and employers- Barbados, Brazil, Colombia, Costa Rica, the Dominican attempts to establish a hybrid of these two forms of skill Republic, Ecuador, Guatemala, Haiti, Honduras, Jamaica, accumulation. Paraguay, Peru, and Venezuela) enrolled more than 5 million people (about half of this enrollment was in Brazil), equiv- Main Features of the Latin American Model alent to about a third of secondary enrollment in the same The prototype of the autonomous vocational training insti- countries. As a result of the dominance of this mode of train- tute was Brazil's SENAI, founded in 1942 for industry, and ing, the ratio of enrollments in vocational training and voca- the SENAC, SENAR, and SENAT-for services, agriculture, tional education courses to those in general secondary education and transport, respectively-which were founded some years in Latin America is more than double that in other developing- later. A training levy-collected by the government on country regions, with more students enrolled in training behalf of employer federations-was earmarked for financ- institutes and centers than in vocational schools. This ratio ing these training institutes. Students completing basic edu- has fallen, as secondary enrollment in schools has increased cation in schools were streamed into general secondary considerably during the last decade in LAC. But in most of 112 CLOSING THE SKILLS GAP TRAINING POLICIES these countries, this system remains an important part of the education than other Latin American countries; almost apparatus to close skill gaps between LAC and industrial- 40 percent of students leaving primary school are streamed ized countries into vocational and vocational-general hybrid (called Over time, three main changes have taken place. "bivalent") schools Public institutes such as those in the CONALEP system provide vocational education (see box * In some countries, the definition of "basic education'" 5 1). Over time, private providers of training are becoming has been changed to include more years of formal more important, especially as the importance of in-service schooling. In Brazil, for example, basic education (after training has grown. which streaming into vocational institutes and sec- ondary schools takes place) has been increased from six A System in Trouble? years to eight. During the half century of its existence, the Latin American * In other countries, vocational education has been intro- model has also not spread outside the region; only the Demo- duced in secondary schools, leaving open the prospects cratic Republic of Congo and Morocco have adopted simi- of going on to postsecondary education institutions. lar systems. This is in sharp contrast to the Latin American In Argentina and Chile, for example, a sizable fraction model of old-age income security, for example, which has of secondary school enrollment is vocational. spread to every corner of the world While extraregional dis- * In countries where vocational training institutes were interest in the model is scarcely definitive proof of its failure, run by governments, some have introduced an element it should lead us to question the optimality of the approach of competition (see CINTERFOR/ILO 2001 for a even for Latin America. detailed account of the changes introduced in most As the economies in the region have become more and countries). The most extreme version of this is Chile, more complex, and as the pace of technological change has where INACAP lost its special status and became Just picked up, the demand for workers who are trainable (that another training institution that had to compete with is, those who can adapt and retool quickly and cost- other public and private training institutions. effectively) appears to have increased relative to those who have been trained-that is, who have job- or technology- Because of differences in original design and the changes specific skills. This has put systems following the Latin that have taken place over the last two decades, while the American approach to vocational education and training 12 countries that are classified as falling into the "Latin under strain for the following reasons: American mode" of training have elements in common, there are important differences in the roles of government, employ- * The systems were designed to channel a large fraction ers, and workers among these countries Today, the most of students at an early stage in the schooling cycle into notable exceptions to the "basic Latin American model" in vocational training institutes, which emphasized job- the region may be Mexico, Uruguay, and Chile, which have specific skills rather than developing strong learning over time adopted approaches that can be better classified faculties. As the economy demanded workers with as variants of, respectively, the American and French models learning skills, the training and education system could of vocational education and training. respond in two ways either by increasing the fraction Chile's reforms since the mid-1970s are especially illus- of students going into the general secondary stream trative. It broke away from the Latin American mold in after completing basic education, or by increasing the 1976, when it created a system in which the government education levels of students going into vocational train- replaced free training in the national training institute ing institutes. Both solutions require close coordina- (INACAP) with tax rebates for training in enterprises. tion between the ministries of education that run the More than 2000 private training agencies (universities, education system and the ministries of labor or indus- schools, centers, consulting firms, and others) compete to try that run the vocational training system. In Latin sell their services to private firms and government programs America, this coordination is rarely present I (see Espinoza, 1997; and Cox Edwards, 2000). Mexico's * The changes in the nature of work and the pace of system of vocational training relies more on vocational technical change required a shift in focus from "the 113 CLOSING THE GAP BOX 5.1 V12XICo'3s W NW t5o 2l Jale a V 3alI UivuLc&dJn JjmALiTh W3on% The Mexican government introduced CONALEP as an Moreover, graduates from the post-reformed program alternative technical education system to traditional have a 45 percent greater probability of finding upper-secondary education. CONALEP has undergone sig- a job than those from the pre-reformed program. nificant structural changes in the past decade A major Furthermore, the 1994 Survey cohorts earned higher transformation took place in 1991, when the system hourly earnings than the 1998 Survey cohorts. A reduced the number of specializations offered from 146 to plausible explanation is that since 1994, real wages 29 and introduced modular courses, the forerunner of the have decreased in Mexico by almost 40 percent. competency-based education and training model (CBET) o The third part of the evaluation examines CONALEP's now adopted in Mexico. cost-effectiveness. The results indicate that CONALEP Examining CONALEP's performance compared to that is a cost-effective program. In addition, CONALEP of a well-designed control group, a recent evaluation spon- has had spillover effects on the rest of the technical sored by the World Bank found the following: education system by stimulating other educational institutions to be more efficient and to adapt to the • CONALEP graduates search longer for a job, but they changing economic and social situation. are more likely to find jobs that they were trained for. As in previous studies, this evaluation found that It is difficult to discern the relative contribution of the CONALEP increases graduates' earnings. However, different factors responsible for the good overall perfor- the order of magnitude of eamings increase differs mance of CONALEP, but it is safe to conclude that the greatly from those in previous studies: on average, special features of CONALEP as a whole have made this CONALEP increases graduates' earnings by 22 possible. These are autonomous organizational structure, percent-not the 30 to 40 percent found in previous decentralized operation, strong links to industry, industry- studies-compared to a control group. experienced instructors, and modular courses. However, o The second part of the evaluation examined the further challenges remain, notably more rapid curriculum benefits of the 1991-92 CONALEP reforms. Results adjustment to changing market circumstances and the indicate that graduates from the pre-reformed pro- improvement of external and internal efficiency. gram (1994 Survey) search longer for a job compared to those of the post-reformed program (1998 Survey). Source Lopez-Acevedo (2002). individual trainee to firms, a broadening of activities Chile have made the shift from cooperative to state- to include other productivity-related services and an led but demand-driven training systems, even such increasing recognition of the importance of small transformations of approach to training are rare in the firms" (ILO, 1999, p. 72). It is far from clear that gov- region. ernments in the region have recognized this, and attempts to systematically restructure training insti- The shift from cooperative to more market-oriented tutions and incentives to conform to these principles approaches to training is under way in many Latin American have usually been slow or half-hearted. countries; the question is whether these reforms are quick o The decline of employment in public enterprises rel- enough and go far enough. ative to private sector firms and self-employment It is important here to briefly address a concern that is required a commensurate shift in emphasis from public occupying the attention of policymakers in the region: the to private training institutions and, even more impor- need for development of systems for lifelong learning. Box tantly, from state-led to enterprise-based, market- 5.2 provides a brief summary of the evidence in favor of these oriented training. While some countries such as arguments and their implications for training policy design. 114 CLOSING THE SKILLS GAP TRAINING POLICIES BOX 5.2 Lifelong Learning: Concept, Evidence for LAC, and Implications WVhat Is Lifelong Learning? workers during the 1980s and 1990s, suggesting that there Lifelong learmng (LLL) has become a buzzword armong edu- were differential shifts in the experience'profiles of differ- cators and increasingly enters the agenda of policymakers. ent age groups (Card and Lemieux, ?001; Card and DiNardo, Lifelong learning preaches that human beings should learn 2002; and Heckman and others, 2001). These. changes "from cradle to grave" or "from womb to tomb." The cen- have sometimes been interpreted as evidence that, in times terpiece of the concept is the idea of a self-motivated and of rapid technological change, skills are more likely to self-reliant person who learns based on his or her interests depreciate, or even become obsolete faster. In other words, and needs. It is believed to constitute a paradigm shift from when technology is changing rapidly only young workers a traditional supply-driven method with a centrally may be able to offer the kind of skills that employers decided curriculumri and class teaching of children toward demand, and thus they may be the ones who reap most of a demand-driven process taking place in all aspects and ages the benefits from skill-biased technological change. If this of life with the individual learner in the driver's seat. is so, we would expect to see a steepening of the experi- Table 5.2 provides a sumnmary of the differences between ence profile, as most of the returns to education are col- lifelong learning and traditional education systems lected early on in one's career. Unlike the United States, Latin America shbws little TABLE 5 2 evidence of such patterns. Figure 5.2 presents experience- Differences between Ufelong Learning and Traditional Educafton earnings profiles for relatively educated workers in Brazil, Colombia, and Mexico in the first and second half of the TRADFiTIONAL LIFELONG 1990s. If skills were rapidly depreciating for older work- ASPECTS EDUCATION SYSTEM LEARNING SYSTEM ers, we would expect the 1996-99 profiles to be steeper What Facts Learning to learn for low levels of experience and flatter for high levels of Who Child and youth All age groupa How Class reaching Individual learning experience than those for the early 1990s. As it turns our, Where Schools and colleges Forman School and colleges both profiles are almost identical for the two periods, and Nonformal. Private courses and self-study this pattern is found to hold for all three countries. While Informal. Workplace and this Implies that the need for LLL has not increased in these Delivery Supply friends countries during the last decade, it does not rule out the possibility that there has always been an unmet need for While LLL is not a well-defined concept, we attempt LLL opportunities. here to make a distinction between the need for LLL from the need for retraining workers. One interpretation couild he What Are the Implications? that while retraining is neressitated by the obsolescence of skills In any case, the main implications of LU appear to be for acquzired on thejob (informally through learning by doing or for- educatidn, not training systems. mnally throuzgh training), LLL is made necessary by the obsoles- cence of skullf acquired during formal schooling. This o The shift in emphasis from knowing vocation- interpretation squares well with the observed increase in specific skills to having better-developed faculties for interest in LLL as the pace of technical change has accel- learning new skills implies that vocational educa- erated during the last few decades. ti*n should be%increasingly postponed to higher levels of education, for example, from secondary to What Is the Evidenzce? postsecondary levels in countries where secondary There is considerable evidence for the United States that education enrollments are high. the increase in the wage gap between education groups expanded faster among younger workers than among older (conitinutes on next page) _ _ _ _ _ _ _ _ _ _ fi_1 CLOSING THE GAP BOX S 2 I Continued FIGURE 5 2 tWage Profiles in Brazil, Colombia, and MexEco-Eaiy and Late 19M Experience Profile: Tertiary-Educated Workers Brazil 0.8 07 0.6- 0 5 i 04- ,,<4"' t---99-95 03 0.2 0 1 0 0to4 5to9 lOtol4 15to19 20co24 25to29 Experience Experience Profile: Terutary-Educated Workers Colombia 0.7 06 0 5 0 4- 03 02 - 01 / , , 0 0to4 5to9 10rol4 15to19 20to24 25to29 Experience Expericoce Profile: Tertiary-Educated Workers Mexico 0.700 0 600 0.500 " 0.400 - 0300 - 0200 - 0100 0 000 0to4 5to9 10tol4 15lo19 20to24 25to29 116 CLOSING THE SKILLS GAP TRAINING POLICIES o There are implications for curriculum in schools, certificatio n mechanisms, so that passage between with an increased need for higher-order cognitive the formal education system and the workplace is skills that aid learning, problem-solving, and analy- srnooth and in both directions-from school to work sis, rather than rote-memorization, simple literacy and from work to school. or nurneracy, and knowledge of facts. o Perhaps the most important training-related implication is the heightened need for senstble Sor,tce. Andreas Blom and Norbert Schady (World Baank staff) Now Employers View Training BOX 5.3 Since training is an important instrument for firms in making World Business Environment Surveys the skills available in the workforce better complement the technologies they employ, we examine how employers in The -World Business Environment Surveys (WBES) Latin America and the Caribbean regard training This inclucle information on firms' size (employees, sales, and section is based on a note commissioned for this report by assets); years in operation; sales, debt, and growth per- Barra (2002), which analyzes the skill- and technology-related formance; sources of finance; and a mix of qualitative and findings of enterprise surveys in 16 Latin American coun- quantitative evaluations of the business environment, tries conducted by the World Bank to assess the business inclucing governance, predictability of economic policy, environment The results of these surveys are comparable the judicial systems, financing, and general constraints across both countries and regions (see box 5.3 for details). to business operations. For Latin America, East Asia, and Using these data, we provide tentative answers to the fol- the OECD, WBES included a module on competitive- lowing questions: Do employers in the region regard the ness with detailed information on enterprise training, shortage of skilled workers a serious constraint for produc- technology, and productivity. tivity increases7 What do employers do to address this short- The WBES sample aimed to accurately represent the age of skills, that is, when and whom do they train? Where relative importance of manufacturing, service and com- do employers that train workers obtain training services? mercial firms in each economy. To ensure comparability Briefly, the results of these surveys indicate the follow- across countries, a sampling frame was developed using ing First, employers in Latin America (and East Asia) rely the distribution of private firms in each country, by sector, overwhelmingly on private suppliers of training, even though size, numbers of employees, and location. Because the sur- policymakers have generally been fixated on public training veys oversampled larger firms relative to their weights, institutions. Second, "skill shortages" appear to be less serl- and because the results presented are not weighted, aggre- ous in Latin America than in East Asia, which seems some- gate training estimates should be viewed only as illus- what surprising until one considers the possibility that the trative of broad patterns of training in sample countries. - demand for more skilled workers is greater the more rapid the rate of adoption of new technologies by firms, and the |wire: liatra (2002) higher the education levels of workers. As seen in the ear- lier chapters of this report, Latin America lags behind East and training incentives that recognize that much of train- Asia in both education levels and the pace of technical ing provision in the region is private, not public. change Finally, these findings suggest that while training policy reform has been seen in the region largely as syn- Are Skills a Serious Constraint? onymous with the reform of public training institutions, it The supply of skilled labor-both professionals and pro- would be more appropriate to see it as a package consisting duction workers-is cited as a constraint to productivity in of efforts to improve education and adopt new technologies, Latin America and the Caribbean (see figure 5.3). Almost 117 CLOSING THE GAP FIGURE 5 3 The Supply of Skilled Labor Is a Leading Constraint to Productivit in Latin America and the Caribbean (Percentage of firms citing factor as a constraint to productivity 2000) Access to and cosc of working capital Quality and supply of skilled personnel Government regulations and caxes Lack of high-quality local suppliers Quality of infrastruccure services Lack of machinery and equipmenc suppliers Quality and supply of production workers High cost of labor Restrictions on hiring and firing practices 0 10 20 30 40 50 60 Souirce Batra (2002) 60 percent of firms in the region cite quality and quantity training providers. Responses to questions in the WBES pro- of skilled personnel as a constraint, and about 45 percent cite vide some clues. The reasons are somewhat different in the the quality and quantity of production workers as a barrier two regions (see figure 5 4) to higher firm productivity The main finding is that quality and supply of skilled per- o The use of mature technologies is easily the most impor- sonnel was ranked as a major obstacle to firm productivity tant reason for not training in both regions, indicating and competitiveness by almost 60 percent of firms in Latin strong complementarities between training and tech- America nology. This is supported by the responses of firms sug- gesting that informal tramnng is adequate in many cases. What Are the Firm-Level Correlates of Training? o A related finding: Latin American and Caribbean firms Virtually all employers report providing some form of infor- that do not train state that they are likely to find ade- mal on-the-job training to their employees, especially new quate skills in workers, indicating either that the edu- hires, so informal training is not a useful discriminator for cation system is supplying workers with appropriate our purposes-identifying efforts by firms to improve the skills, or that the demand for skilled workers is rela- match between skills and technology-and about 60 percent tively low in LAC, or some combination of the two. of surveyed firms provide some formal training. Consider- o In general, information asymmetries or financing con- ing the productivity gains from training, why do almost half straints appear to be a less serious problem in Latin of firms in Latin America not train? America than in East Asia. The explanation could be related to technology, educa- tion, labor market functioning, or training supply: (a) The The rationale for public subsidies for training appears to use of mature technologies obviates the need to train for- be weaker in Latin America and the Caribbean than in East mally; (b) the lack of adequately educated workers makes Asia: labor turnover and resource constraints are less likely it costly to train them; (c) high labor turnover makes it to be impediments to training in LAC than in East Asia. unprofitable to do so; or (d) there are weaknesses in the These rankings are consistent by firm size as well. How- supply of training, or employers have poor information about ever, small firms are more constrained by resources, lack of 118 CLOSING THE SKILLS GAP: TRAINING POLICIES FIGURE 5.4 Use of Mature Technologies and Availability of Skilled Workers Are the Main Reasons Why Firms in Latin America Do Not Invest in Formal Training Reasons for Firms Not Training Workers, 2000 Skeptical about benefits I I I Lack knowledge about providers 7pl Resouirce limitations Labor turnover _ Informal training adequate Mature technologies - Skilled workers available 0 5 10 15 20 25 30 35 40 45 50 Percentage of All Firms I0 East Asia a LAC Source: Bacra (2002). knowledge about training benefits, and labor turnover than What does appear to be a general finding is that small are large firms. The WBES fielded in Guatemala in 2000 and medium-size firms should be the primary targets for provides some evidence of this (figure 5.5). incentives to encourage training. What also seems to be true Failures in the market for training-roo low private is that training incentives are often ineffective in reaching demand because of labor turnover, or too low supply, or poor precisely these firms (see box 5.6, however, for a more effec- information about demand or supply-pose more difficult tive training incentive scheme in Malaysia). questions. If market failures are responsible, what kinds of training policies are effective in addressing these failures? Where Do Firms Get Training Services? Do the low incidence of formal training and the striking dif- The survey distinguished between formal training provided ferences in training by firm size reflect the weak training and in-house by the employer and formal training provided technological capabilities of enterprises or market failures? by external training institutions, both public and private. Are firms constrained by poor access to financing for train- Reliance on internal training is strong across all East Asian ing or do they lack the interest, know-how, or capability to countries, accounting for about 40 percent of all training; design and implement training programs? How important private providers account for more external training than a constraint is "labor poaching," the hiring away of employ- do public providers. In Latin America, 50 percent of enter- ees trained at another employer's expense, which prevents prises provide in-house training (see figure 5.6). Of the exter- firms from recouping the returns to their sunk investments nal providers, public training institutions are not a popular in training? Answers to these questions are crucial when choice-only 16 percent of enterprises use public institutes. designing public policy. Insights into some of the questions These ratios may come as a surprise to many engaged in are provided by firm respondents in the competitiveness training policy reform in the region, which is regarded as module; policymakers need to know why firms train and synonymous with the reform of public training provision. why they do not. But they require country-specific analysis Employers rely most on private external training sources in order to arrive at sound training policy design. in Chile and Uruguay, both countries that do not follow the 119 CLOSING THE GAP FIGURE 5 5 Knowledge of Training Providers and Resource Considerations Are More Important Constraints for Micro-Enterprises, Guatemala 2000 Reasons for Not Training Mature technology requires little training _ _ _ W-_ Skilled workers readily hired,__ ,; High labor turnover i.iO makes training costly _ ° Dissatisfied with current training providers i Lack knowledge about X0 training _EIml Medium ] l l . 2~~~~~~~~~E Small Limited Resources __ Micro Skeptical about benefits of I X training _ 0 10 20 30 40 50 60 70 80 90 100 Source Batra (2002) FIGURE 5 6 Almost 50 Percent of Training in Latin America Is Internal to Enterprises, and Another 40 Percent is Privately Supplied Training by Source 80%7 0 ii lmN 209' .20%....% C~~~~~~~~~~~~~~~~~C Country .. Public training a Internal training Source Barra (2002) Latin American mode of training Public training suppliers Countar i PublsSvueTraing [idcanice play a relatively large role in Mexico, Ecuador, Belize, ifrolm l;eSECc Trinidad and Tobago, and Nicaragua Nevertheless, given Recent research indicates that training significantly in- the importance of internal training by enterprises, it would creases firm-level productivity. For example, in Malaysia and not be far from the truth to assert that even in these coun- Colombia, firms that train are more than 25 percent more tries, about three-fourths of training provision is private.2 productive than firms that do not. Training effects of this 120 CLOSING THE SKILLS GAP TRAINING POLICIES BOX 5 4 BOX S 5 Enterprise Training and Productivity in Developing Countries Finn-Level Cross-Section and Panel Data for Mexico Using a simple production function approach, the pro- In a paper prepared for this report, Tan and Lopez- ductivity effects of formal training can be estimated. Acevedo (2002) use the 1992, 1995, and 1999 These estimates, shown in table 5.3, indicate that enter- ENESTyC surveys; Ciecon (2001) uses the 1995 and 199 prise training is associated with much higher levels of surveys to examine pre- and post-NAFTA incidence of worker productivity in private firms. trainirg. The survey was also fielded in 2001, but data were riot available at the time this paper was commis- TABLE 5 3 sioned. The ENESTyC surveys a large, stratified random Estimated Productivity Effects of Formal Training sample of manufacturing establishments and contains PRODUCTIVITY questions on employment, wages, worker characteristics, COUNTRY YEAR EFFECT (PERCENT) technology adoption and use, and formal and informal Nicaragua 2000 564 training. The ENESTyC also allows a cross-sectional Guatemala 1999 49 0 Mexico 1992 444 analysis of the impact of training on wages. Colombia 1992 266 Tan and Lopez-Acevedo also link firm panel data from Indonesia 1992 71 1 Malaysia 1994 28 2 the ainual survey of manufacturing (EIA) to the ENESryC to more rigorously estimate the impact of In contrast, no significant productivity effects were training on wages and productivity over time. The EIA, found for in-service training provided by public insti- available annually from 1993 to 1999, contains detailed tutions. Most of the productivity gains from training in and comparable information on the productivity and private firms came from private training institutions and wage outcomes of interest, but not in-firm training. So in-firm training programs. Finally, productivity effects training information from the ENESTyC surveys is of in-service training seem to be larger the poorer the brought to bear on a subsample of EIA firms at three country, possibly reflecting the scarcity of skills in lower- points in time. The firms in the subsample differ from income countries the population in two important respects-they tend to be larger and, as survivors over a turbulent period Soarer: Tan and Barra (I 995), Batra (I999, 2000). between 1993 and 1996, are probably more productive ____ ___________________________________________ on average than the typical firm. magnitude are not unusual for developing countries-some Sorurce Tan and Lopez-Acevedo (2002), Ciecon (2001) are even higher; in Guatemala and Nicaragua, the training premium is estimated to be twice as much (see box 5.4). Given these estimates, training is (legitimately) seen as a aggregated sectoral data Box 5.5 presents the characteris- potent instrument for raising productivity, and governments tics of the data used, and the main findings are presented in intervene pervasively to increase in-service training. But to figures 5.7 through 5 10. The findings confirm the results design effective interventions, one has to understand why so found in the WBES for Latin America and the Caribbean at many firms pass up opportunities for increasing productiv- a more aggregate level, and using a smaller sample size per ity, and not train their workers. This section examines the country. experience of Mexico, for which available data allowed rela- tively rigorous analysis of the determinants of training. o First, regression estimates using probit techniques point to education levels of workers as the most impor- Complementarities Confirmed: Evidence tant correlate of training incidence, and this finding from Firm-Level Cross-Sectional Data is iobust for all subsectors of the Mexican economy- Ciecon (2001) estimates the correlates of in-service training from agriculture to manufacturing to services (see in Mexico during the second half of the 1990s, using dis- figure 5 7) 121 CLOSING THE GAP FIGURE 5 7 In Mexico, Educated Workers Get Trained Move Oten ... Marginal effect in the probability of taking a training course, computed at the mean values of the variables inatnaentto lovd Group of age (Seder Fan-g, fa ey, -d lug hJ Fd. bt-g., ad eb.no_ T-dt,le, pp.e, ted 1-ahre Lattl. ned erd pedacre P.pee, pt teg. ted upblisits __ Cheminet, p,eeneit. ebbeer,-d pl.t- Stone, tiey, ted gbus pr dmt Pertty -eI ad---eta M l pmdcr., _eah-oe,y, td r parnt Elanecu eiy Wht ol -ed ertal -ade R -rruanu ad h-oelC T-epa-r-aeeone, amd c an FPa_ee -a-mace, ad ral estte eece, S-eal ad p -e deeti Pablhe edeuneeesronI 0 00 0 01 0 02 0 03 0 04 0 05 0 06 -0 02 -0 01 0 00 0 01 0 02 0 03 0 04 -0 30 -0 20 -0 10 000 0 10 0 20 Job -embe Ntbeir of -mplo.ye Onnap-toaIl gw-up F.-n. fotey, .ed fathmg Mteeng]_= Fran1. brntna0et ted teltaceo 2] Te-ee,,appre ted tle t Leebee antd tvod pnd-ter P.per pe--rg, td pelehlneg Ch-re-ol, pnerele n, tbre. ted pl.t t Stone., dy, td gInt pedanet Metal peodetet, eeadtnesy, ted re1eepnenr Od.aetllan,net ecaafaeean- tdetusc- Cnmun,ttae Eles--rnery Whblara and _nntI trdeI Ilataunanen ted bmots _ Ttnnspeer, stanage. sad rommumnteat,nt 1 F -axe, rnane, ted teal estate tcnI Soc.1 md p-.1nlSlc All _ _ _ -001 000 001 002 003 004 -005 000 005 010 o015 020 -015 -010 -005 000 005 Note White bars indicate that the estimate is nor significantly different from zero Source Ciecon (2001) o Second, for a subsample of manufacturing firms for small role as providers of external training for firms (Tan and which measures of innovativeness were available, R&D Lopez-Acevedo 2002) spending is the second most important correlate of What was driving the rise in training in the 1990s7 Two training (see figure 5.8). hypotheses have been advanced in the literature as possible o Third, for the full sample of all sectors, about two- explanations for the growing demand for training: first, the thirds of training is in the workplace (see figure 5.9). effect of globalization and, in particular, the post-1994 grow- About half of external training is from private ing integration of Mexico into the North American market providers (see figure 5.10). These numbers are very under NAFTA; and second, the effect of technological change similar to those found for Latin America as a whole and adoption of new information and communication tech- in figure 5.6. nologies (ICT), both of which are believed to be relatively skill-intensive (or skill-biased). Tan and Lopez-Acevedo Particularly noteworthy in Mexico is the high frequency (2002) provide causal evidence that both forces may have with which respondents-both small and large-identified contributed to the rising trend in training incidence. Over private companies, industry associations, and to a lesser time, training incidence trends upwards irrespective of tech- degree public centers for worker training as the principal nology or trade status, but these trends are especially pro- providers of external training. Also noteworthy is the grow- nounced for firms conducting R&D or exporting ing role, between 1992 and 1999, of private training con- Simulations suggest that the observed rising training sultants and equipment suppliers In both years, public and trend over the 1990s is attributable principally to the chang- private universities and technology centers only played a ing relationships between training and its covariares, rather 122 CLOSING THE SKILLS GAP: TRAINING POLICIES FIGURE 5.8 ... and More Innovative Firms Train Their Workers More Marginal effect in the probability that an establishment provides training to its workers, computed at the mean values of the variables R&D expenditure Introd. of much, and eq. Domes-ic competitivenes Food, beve-aen, and tobacco _ Te-tiles, opparel ,ond leather Lumber and s-nod poductsI Paper, pinting,. and publithing Chemicals, petrleum, robber, rnd plastic Stone, clay, and glas products= Primary -mlr industies I MetaI products, machinery, and ecuipment Miscellaneous manouf creing industries A ll _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ 000 0.20 0.40 000 005 010 0.15 0.20 025 -0.30 -0.20 -0.10 0.00 0.10 0.20 Internation-al roes tiivenets Inteodu-son of o snaional chunten Year of op-aton Food, beveeages, and tobacco Textiles, apparel, and leather Lumb-rand mood poducts- Paper, printing, and publishing Chtemirol, petroleum, robber, and plastic Stoae, clay, and glass prodacts Prmary metal industries Me-tl products, machinery, and equipment Miscellanenas manufacturing industries All, -0.20 0.00 0.20 000 010 020 0.30 0.40 0.00 005 010 015 020 Total sales Food, beve-oges, and tobacco Textiles, apparel, and leather Lumber and wood products Paper, printing, and publishing Chemicals, petroleum, robber, and plastic Stone, clay, and glass products Primary meral industries MetaI products, machinery, and equipment Miacellaneous manufacturing industries All 0.00 0.02 0.04 0.06 0.08 0.10 0.12 0.14 0.160.18 Note: White bars indicare that the estimate is nor significantly different from zero. Source: Ciecon (2001). FIGURE 5.9 In Mexico, Two-Thirds of Training Is Internal ... Percentage of employed population that has taken a training course whose latest course was taken in the workplace Farming, forestry, and fishing Mining Food, beverages, and tobacco Textiles, apparel, and leather Lumber and wood products Paper, printing, and publishing Chemicals, petroleum, rubber, and plastic Stone, clay, and glass products Primary metal industries Metal products, machinery, and equipment Miscellaneous manufacturing industries Construction Electricity Wholesale and retail tnade Restaurants and hotels Transport, storage, and communicarions Finance, insurance, and real estate services Social and personal services Public administration Total 0 20 40 60 80 100 Source: Ciecon (2001). 123 CLOSING THE GAP FIGURE 5 10 ... and About Half of the Rest Is by Pnvate Providers Percentage of employed population that took their latest course outside their workplace who did it in a private institution Farming, forestry, and fishing Mining Food, beverages, and tobacco Texriles, apparel, and leather Lumber and wood products Paper , printing, and publishing Chemicals, petroleum, rubber, and plastic Stone, clay, and glass producrs Primary metal industries Metal products, machinery, and equipment Miscellaneous manufacturing industries Construction Electricity Wholesale and retail trade Restaurants and hotels Transport, storage, and communications Finance, insurance, and real esrate services Social and personal services Public administration Total l 0 20 40 60 80 100 Source Ciecon (2001) than the covariates themselves. Changes over time in the dently increases the likelihood of employer invest- means of these covariates-for example, changing industrial ments in new technology through R&D. structure, size distribution of firms and geographic location, o R&D has a much greater impact on the likelihood of R&D, and exports-are less important. external training than on in-house training programs. The marginal effects of R&D on training more than Productivity Effects of Education, Training, and R&D: doubled during this period of growing trade contacts Evidence from Panel Data with the rest of the world. For in-house training, the Using regression techniques and firm panel data, in a paper impact of R&D rose from 1.6 to 3.7 percent, and for commissioned for this report, Tan and Lopez-Acevedo external training it rose from 3.1 to 7.8 percent. (2002) analyze the interaction among skills, technology, and o Consistent with findings that export-oriented firms are productivity during a period of increasing openness in more likely to train their workers, greater openness is Mexico between 1992 and 1999 Their main findings are associated with a higher incidence oftraining. But the as follows: results also show that training incidence has increased only in firms that do some R&D (figure 5.11). o An increase in average years of schooling raises the o While a substantial number of(especially large) firms train likelihood of training, a relationship that persists but report no R&D, the likelihood that firms do R&D and becomes quantitatively more important over the without training is extremely low. In 1999, no more than 1990s. This provides more evidence in favor of the grow- 7 percent of firms in any given size category reported doing ing complementarity between education and training. R&D without training. In short, while firms may train o The educational distribution of the workforce plays a for a variety of reasons, they seldom engage in R&D with- key role in both technology and training decisions The out complementary investments in worker skills. share of the workforce with a college education appears o Training-especially if it is continuous-increases total to drive training by employers, here, the results sug- factor productivity growth. Similar results hold for gest that a highly educated workforce also indepen- technology transfers and R&D conducted by the firm 124 CLOSING THE SKILLS GAP TRAINING POLICIES (see figure 5.12) R&D, and technology transfer through lyzes the effects of a training intervention-the Human licensing and patents, are statistically significant only for Resource Development Fund (HRDF) scheme-in a frame- the firms that reported investing in these innovative work that recognizes the importance of technology adoption activities for at least half of the intervening seven years. in traininig decisions of enterprises These results confirm, using panel firm-level data for one Policies to Promote Training country, what the WBES suggest at a region-wide level. The evidence on the correlates of training indicates that two Box 5.6 presents evidence from Malaysia presented in Tan of the three components of an effective training policy pack- (2002), which reinforces these findings. Box 5.6 also ana- age are measures to increase education levels of workers-since training and education appear to be strong complements- PIGURE 5 11 and to increase the rate of adoption of new technology. An ear- In Mexico, Training Increases with Greater Openness, but Only If Finns Conduct R&D h~~~~~~~~ler chapter dealt with education policy; here it should suffice Firnns Conduct R&D to point out that this may involve changes in the education In Mexico,Training Increases with Grearer Openness, and training systems so as to shift the balance between gen- but Only if Firms Conduct R&D eral and vocation-specific skills acquired in the formal edu- cation system. Two later chapters discuss policies to promote technology adoption in Latin America and the Caribbean. This chapter concludes with a discussion of the third component- incentives for firms to train their workers. a.N / _ - 1 _Rationale for Public Intervention /V _ _ ,The policy intervention that is appropriate depends on the nature of the market failure. Table 5.4 provides an o _ |- -- application of this approach for training markets. When poor 1992 1999 information is the constraint, for example, the appropriate oFirms with R&D oFirms wich no R&D policy response is to disseminate best practices in training FIGURE 5 12 Training, Especially if It Is More Continuous, Facilitates TFP Increases Mexico Productivity Effects of Training and Technology, 1993-99 05 0 45 04 035 03 0 25 . 02 - 0 0 1 [U 005 L _ _ _ _ T-ninS I Tr-unng 2 Tra-nig 3 Technology Technolo-y Tec-inoigy Techoology R&D 1-2 R&D 30 -O 05 Period Periods Perods Transfer 1 Teansfer 2-3 Transfer 4-5 Transfer 6. Periods Periods Period Periods Periods Periods 125 CLOSING I HE GAP BOX 5.6 Complementarities between Skills and Technology: Training Intervenions to Reast Ann PrduciVity in Malaysia In 1993, Malaysia replaced its scheme for promoting and medium-size firms train more frequently, and training (double deduction for approved training-related that participation in the HRDF scheme is associated expenses in calculating taxable income), which had been with more freqluent training. in operation since 1987 but found to be largely ineffec- o Second, simulations indicate that training increased tive, with a training levy administered by a council of rep- more owing to the enactment of HRDF as compared resentatives of the private sector and selected government with the effect of introducing new technology. The agencies. To start the scheme, the government provided a exception is for small firms, for which these two effects sizable grant that matched company levies. are roughly similar. The impact of HIRDF is most pro- Was the HRDF scheme effective in raising training and nounced for medium-size firms, while the impact of firm productivity? To answer this question, Tan (2002) technical change is greatest among small firms. uses firm-level panel data from 1988, 1994, and 1997, o Third, the productivity impacts of training are siz- namely, comparing the training decisions and productiv- able; even more important, there is evidence of strong ity outcomes before and after the intervention. However, complementarity between training and technology. this was a period of considerable dynamism for Malaysia, The productivity impacts-of 50 and 23 percent, and other things that would affect training and firm pro- respectively-are twice as high in firms with new ductivity changed as well. In particular, the percentage of technology as compared with firms without new firms introducing new technology in the panel rose from technology, providing "dramatic confirmation of the 53 to 75 percent during the decade, and those using key intermediating role that training plays in real- advanced process IT rose from 48 to 83 percent. The edu- izing the productivity potential of new technology." cation level of the workforce also rose rapidly. The analy- o Finally, productivity impacts of training are about sis considers these changes in assessing the effectiveness half as large in small firms as in medium-size and of the intervention. large firms, suggesting an explanation for the fre- By reimbursing some of the approved training expenses quent observation that training is less common in of participating firms and hence lowering the cost of train- small firms. Training may simply be less productive ing, the HRDF creates incentives to train on the cost side. in small firms because they do not make comple- In contrast, technological change creates incentives to train mentary investments in new and potentially more and to hire educated workers on the revenue side. The productive technologies. To address this, the HRDF analysis yields the following findings. has introduced group training schemes for small firms, and the national SME agency has stepped up o First, probit estimates show that training is signif- its efforts to provide integrated business, training, icantly higher in firms that introduced new tech- and technology development services. nology and that employed a higher proportion of skilled workers The estimates also show that large Source, Tan (2002) TABLE 5 4 Rationale and Policy Options for Public Intervention in Training KIND Of INTERVENTION REASON FOR INTERVENTION FINANCE TRAINING PROVIDE TRAINING COMPLEMENTARY POLICIES External benefits Preferred Not justified None Market imperfections Second-best Not justified Preferred deal with source of market imperfection Weak private training capacity Not )ustified Second-best or preferred Preferred build firm training capacity Equity Second-best Not justified Preferred selective scholarships Source Middleton, Ziderman, and Adams (1993, p 116) 126 CLOSING THE SKILLS GAP TRAINING POLICIES know-how and information about the availability and cost collective agreements (Belgium, Denmark, France, Greece, of services High rates of labor turnover may suggest that the Netherlands, Sweden), or national insurance levies for dis- there are externalities in training to the extent that firms placed workers (France, Sweden) Levy exemption schemes are unable to internalize the benefits of training because have also been used as a means of subsidizing smaller firms skilled workers can be hired by other firms, there will be and adult training providers (Austria and Germany), simi- underinvestment in training. Mandates, collective action, or lar levy schemes have been implemented in Morocco and incentives can help to internalize some of these externali- Turkey. Levy-rebate schemes have been used in Jordan, Korea, ties This section focuses on some of these public policy Malaysia, Mauritius, Singapore, and Taiwan (China) Some instruments employed in various countries and provides countries have modified payroll levies Peru, for example, has some examples of good practice lowered the levy and diversified the sources of finance Many countries, both advanced and developing, have put Argentina, Brazil, and Colombia have cofinancing arrange- into place different policies designed to foster increased in- ments with enterprises, communities, and vocational schools service training among their enterprises, including payroll- Singapore and Taiwan (China) have used matching grants. levy training funds and tax incentives for employer-sponsored Experience with levies yields the following lessons (see training. These employer-targeted training policies take also box 5 7): many forms: (i) levy-grant schemes, in which fund admin- istrators use earmarked levies to make grants to employers * Keep employers in charge. Industries need to own the for approved training, as in Singapore and previously in the levy Argentina, Brazil, and Chile have vested super- United Kingdom, (ii) training levy rebate schemes, in which vision of levies in industrial bodies, Malaysia's expe- employers are partially reimbursed for approved training out rience provides the clearest lesson in this matter. of their payroll levies, as in Malaysia, Nigeria, and the * Increase competition in provision. Levy funds are not Netherlands; (iv) levy exemption schemes, in which employ- cost-effective when they support only government ers are exempt from levy payments provided they spend a training providers. given percentage of their payroll on training, as in France, * Earmark funding strictly. Funding levels are better Korea, and Morocco; and (iv) tax incentives for approved maintained with levies than with government grants, training paid out of general revenues, as in Chile and pre- which tend to decline with shrinking budgets. But viously in Malaysia. levy funds should not be used for other government expenses, as has happened in the past in Costa Rica The Payroll Levy Grant System * Provide support for smaller enterprises These The payroll levy is a common instrument in Latin America schemes have typically been used by large firms and for overcoming the underprovision of training. Argentina, enterprises that already have a high skills base-support Barbados, Brazil, Chile, Colombia, Costa Rica, Ecuador, mechanisms are needed for small and medium-size Honduras, Jamaica, Paraguay, Peru, and Venezuela have firms and farms to participate. Mexico's CIMO pro- implemented levies, with mixed success. Brazil's levy, one gram has been assessed as effective in supporting small of the oldest, suggests that factors influencing success include and micro-enterprises (see box 5 8). administrative independence of the levy fund, a combina- tion of private ownership and public mission, and a man- National Training Councils agement structure that includes industry and government Many countries have established national training councils In Chile small and medium-size enterprises have overcome (Argentina, Australia, Brazil, Chile, Cote d'Ivoire, Malawi, the tendency for levy funds to subsidize only large firms, by Mauritius, South Africa, and the United Kingdom) Their grouping together on a sectoral basis Sectoral centers have experience suggests that locating country management of also been established in Argentina, and small and medium-size training with the social partners (business, unions, and govern- enterprises have negotiated a sectoral rate with the national ment) can improve the quality, relevance, and flexibility of supervisory agency training. Training funds managed by training partners have Levies can also be set and controlled through industry tended to become more diversified in how the money is col- sector bodies (as in Mexico and South Africa); industry lected and spent, including their use in the informal sector 127 CLOSING THE GAP BOX 5 7 Well-Designed Levy-Grant Schemes Can Wloiva4e Finns to Train Several East Asian economies have used direct reimburse- developing training plans, organizes regional courses on ment of approved training expenses, funded through pay- training need assessments, and adminiscers a variety of sub- roll levies, to encourage firms to train their employees. sidized programs targeting small enterprises. A recent Successful schemes are flexible, demand-driven, and often analysis indicates that the scheme has significantly accompanied by an information campaign and technical increased the incidence of training. assistance to smaller firms. In Malaysia, the HRDF was established in 1993 with a The introduction of such a scheme in Taiwan (China) matchlng grant from the governmenc. The act created a led to dramatic increases in training, which continued after council (HRDC), with representatives from the private I the program ended in the 1970s. Singapore uses a levy on sector and from responsible government agencies, and a sec- the wages of unskilled workers to upgrade worker skills retariat to administer the HRDF schemes. Eligible employ- through the Skills Development Fund. The fund's aggres- ers with 50 employees and above are required to contribute sive efforts to raise awareness of training among firms, to I percent of payroll to the HRDF. Those who have con- support development of company training plans, and ro tributed a minimun of six months are then eligible to claim provide assistance through industry associations have led a portion of allowable training expenditures up to the limit to a steady rise in training, especially among smaller firms. of their total levy payments for any given year. The HRDC However, such schemes can also create disincentives to set rates of reimbursement, varying by type of training and train when rigidly administered. In Korea, which required generally lower for larger firms. Box 5 5 provides evidence that training last a minimum of six months or that firms of the impact of this intervention on firm productivity. pay a fine, many firms paid the penalty rather than train to this standard. In addition, the fund provides grants for Soarce: Batra (2002) In Argentina, Brazil, Chile, Mauritius, and Peru, indus- tributed through open tender, with the state as purchaser try associations have assumed responsibility for administer- rather than supplier of training (Australia, Chile). A more ing the levy. It has been important for these industry bodies radical measure has distributed funds to the user or trainee to have responsibility for the bulk of the funds and to work through voucher schemes, such as the United Kingdom with both training providers and enterprises The Japan Training Credits scheme Industrial and Vocational Training Association is an associ- Because different types of firms and workers require dif- ation of employers that provides training programs for indus- ferent types of training, it is important that the training try trainers. The association receives no funding but charges market not be too inhibited through institutional con- membership and course fees. The semi-autonomous Voca- straints. In Colombia, despite an extensive and well-funded tional Training Corporation in Jordan is an industry body training system, many skilled workers have not used the that works closely with government and industry in pro- formal training system. viding in-house and external training. In the United Kingdom, industry-administered training Matching Grants Schemes funds at the regional level, through Training and Enterprise Some countries use matching grants schemes to increase Councils, are being replaced by regional Learning and Skills training. The most successful schemes are demand-driven, Councils that will combine a broader range of education and with private sector implementation, and aim to facilitate training functions. the creation of sustained markets for service provision There are many mechanisms for distributing training (Crisafulli 1998) Programs in Chile and Mauritius use pri- funds. Financing can go to state-run training institutions, vate sector agents to administer the matching grants schemes it can be directed selectively to enterprises on the basis of Both have reported positive results-and so has Mexico's training plans (Germany, Korea, Singapore), or it can be dis- program. 128 CLOSING THE SKILLS GAP TRAINING POLICIES BOX 58 Mexico's Proactive Approach to Small and Medium-Size Enterprise Support The Integral Quality and Modernization Program (CIMO), programs and other consulting assistance. CIMO is established in 1988 by the Mexican Secretariat of Labor, expanding its support in two directions: assisting groups has proven effective in reaching small and medium-size of small and medium-size enterprises along specific sec- enterprises and assisting them to upgrade worker skills, toral needs, and providing an integrated package of ser- improve quality, and raise productivity. Set up initially as vices, including information on technology, new a pilot project to provide subsidized training to small and production processes, quality control techniques, and mar- medium-size enterprises, CIMO quickly evolved when it keting as well as subsidized training. became apparent that lack of training was only one of many Evaluation studies in 1995 and 1997 found CIMO to factors contributing to low productivity. By 2000 CIMO be a cost-effective way of assisting small and medium-size had provided an integrated package of training and indus- enterprises. The study tracked two groups of small and trial extension services to over 80,000 small and medium- medium-size enterprises over three years, one with firms size enterprises each year and training to 200,000 employees. that participated in CIMO in 1991 or 1992, and another Private sector interest has grown, and more than 300 busi- with a broadly comparable control group of enterprises that ness associations now participate in CIMO, tip from 72 in had not participated in the CIMO program. CIMO firms 1988. tended to have lower performance indicators than the con- All states and the Federal District of Mexico have at trol group prior to participation in the program, but by least one CIMO unit, each staffed by three or four pro- 1993 labor productivity had either caught up or exceeded moters, and most units are housed in business associations, that of the control group. Other performance indicators which contribute office and support infrastructure. These showed similar improvements: increased profitability, sales, promoters organize workshops on training and technical capacity utilization rates, and wage and employment assistance services, identify potential local and regional growth--and reduced labor turnover, absenteeism, and training suppliers and consulting agents, and actively seek rejection rates for products. The most dramatic impacts out small and medium-size enterprises to deliver assistance of CIMO interventions were among micro-size and small on a cost-sharing basis. They work with interested small firms. and medium-size enterprises to conduct an initial diag- nostic evaluation of the firm, as the basis for training Source: STPS (1999), Tan (2000). An increased investment in training has been matched regularly engage workers in informal training. The Basic Law by a reduction in enterprise failure. A side benefit has been for Vocational Training in Korea is designed to encourage the development of a network of industry management train- in-company training Strong training cultures have been ing consultants that are available to enterprises that want established in some Asian countries (Tapan, Korea, Singapore), to invest in enterprise-based training Singapore has under- some northern European countries (Germany, Nether- taken a program to build up irs stock of industry trainers, lands, Scandinavia), and, judging on the basis of levels of and Japan's Industrial and Vocational Training Association in-company training, some Latin American countries (Brazil has trained over 30,000 industry trainers in the past 30 years and Chile). Matching grant schemes can support the development of Matching grants schemes can also link educational and a training culture by providing both an incentive and a human resource development policies. The Singapore Skills means of investing in training. It is important to build a Development Fund was designed and successively modified training culture with a high level of training capacity in to provide an incentive for enterprises to increase the skill enterprises and a high propensity for workers to undertake and pay level of their workers. But a matching grants training, so that enterprises continue to invest in training. scheme, by itself, will not lead to an expansion of the train- In Japan most managers have a training function, and ing market. And grants should not be restricted to state-run 129 CLOSING THE GAP training institutions. Funds should support strengthening o Increase general education levels. Employers' decisions to and diversifying the supply of training and stimulating train and the productivity outcomes of training depend demand. Mexico's Integral Quality and Modernization on the stock of education and technical skills that mdi- Program concentrates on the productivity of small enter- viduals bring to the labor market. The maintained prises, using both private training consultants and govern- hypothesis is that it is more cost-effective to impart ment and private training institutions (see box 5.8). training to workers who are adept at learning. Even Training for the informal sector is typically provided as the effectiveness of public vocational training increases informal apprenticeships, often through nongovernmental as education levels increase as education has improved, organizations, which help to diversify the funding for vocational training institutes in many Latin American training programs for the poorer sections of the economy. countries have reverted to their original mission of Argentina, Costa Rica, and Peru have successful programs training for productive employment instead of imple- of this type. menting remedial adult education programs o Postpone vocational education. More rapid technical suD0mla1g change necessitates a stronger education base: the main Productivity analyses in developing countries find that implication for vocational education and training sys- investments in firm-sponsored training-especially in-house tems is that vocational education should be pushed to training-have large payoffs. Yet a sizable fraction of man- post-secondary levels in countries that have attained ufacturing and services firms do not provide any formal train- universal primary education. ing for their employees. This is especially pronounced for o Promote adoption of new technologies. Technology policy small and micro-size firms-over half of them provide no to improve access to new technologies and stronger structured formal training. A significant number of large incentives to adopt them is also an important com- firms also report no training in Latin America and the ponent of an effective approach to training. The most Caribbean innovative experiences in Latin America in training What accounts for this lack of interest in an activity that "conceive the latter as part of a set of technology trans- appears to have high payoffs in terms of productivity? The fer actions" (CINTERFOR/ILO, 2001, p. 29). evidence indicates that firms are more likely to train when o Adopt more effective trainingpolicies. Training policies need they employ an educated and skilled workforce, invest in to be well designed and implemented, and targeted R&D and technology, are relatively large, emphasize qual- firms need to be made aware of the policies. Several con- ity control methods, or have foreign capital participation or straints on training-poor information about benefits, export to foreign markets. Thus there are strong comple- high training costs from the inability to exploit scale mentarities between training and schooling and critical links economies in training, weak managerial capabilities, among firms' training, technology, and competition in prod- absence of competitive pressures, or market imperfec- uct markets. Firms that train use a variety of in-house and tions-may be operative and policy initiatives to external providers. Private sector providers are as important address these constraints should be explored. as-if not more important than-government-run training o Establish coordinated and proactive small and medium- institutions These findings have implications for education, size enterprise policies. A high percentage of small and technology and training policy, and policies to promote small medium-size enterprises do not train. Such firms face and medium-size enterprise development. a variety of training constraints, from high labor Since firms view training as an instrument to increase the turnover to poor information and finance Proactive complementarity between education and technology, gov- measures are needed to seek out and deliver a pack- ernment policies to encourage in-service training must consist age of integrated services to small and medium-size of a package of actions and incentives to increase educa- enterprises-including consultancies, training, and tion levels, adoption of new technologies, and incentives to technology information and incentives. firms-especially small and medium-size enterprises-to train workers. Important recommendations include the There is reason to be optimistic in Latin America and the following: Caribbean. Education levels have been increasing, firms have 130 CLOSING TIIE SKILLS GAP TRAINING POLICIES 'BOX 5 9 Recent Changes in Training Systems in Latin America and the Caribbean Countries in the Latin America and Caribbean region have occupations or jobs, is increasingly being replaced by changed their approach to vocational training during the last the view that training is workplace- or firm-centered, few decades. In general, the changes have been in the right and is a series of competencies. direction. A recent assessment summarizes these changes o Mure private than public. Private training and voca- as follows: tional education provision, which used to be small o Stroniger links with technology. Diverging from the relative to public providers, has grown enormously "notion oftraiing as asupplementary field of activ- both because of economic changes and because of iotyivorc o ftromina tecnlo aly finnovati andiv- encouragement by government policy. ity; divorced from technological innovation and o More diversifiraton in finance. Starting with a reliance development," many countries are now integrating are train ing liti from the two sets of activities. ~largely on training levies and appropriations from the two sets of activities. o Stronger Iznks wi$th edecation Setting aside the notion the general budget, the financing of vocational train- of vocational training as separate-though theoreti- ing has evolved to also rely on incentives, rebates, of vcatinal raiing s searat-thugh heorti-and other government support. cally complementary-from education systems, coor-anotegvrmntspr. cally complementary-from educationsystems,coor-Mo More variation in approach. In place of a relatively uni- dination efforts are being increased to build national r education and training systems, capable of meeting form approach in different countrles that featured a growmg and diverse demand. national institutions that monopolized vocational a growing and diverse demand. tanjgatvte,tergo o a ieydvr o More firm- than institution-centered. The notion of a trainingactivities,trheuregionnowhaswdelydver- school- or institution-based activity, concentrated in gent organzational structures. an early part of life to prepare individuals for specific Source CINTERFORIO1T (2001) increasingly more reasons and ability to institute new tech- enterprise-based training systems like those of the United nologies, and recent diagnoses of training systems in the States Such a move may require a transitional phase of state- region point to noticeable improvements in their design and led but demand-driven systems such as those of East Asian functioning (see box 5.9). These changes in training systems countries. Chile's training reforms provide perhaps the most can be seen as evidence of a growing recognition of train- important clues on how this can be done in Latin America, ing as a link between education and technology, and of firms but policy reforms in other countries also provide the ratio- as the principal agent for strengthening this link in the most nale and guidance for continued efforts to make training sys- productivity-enhancing manner tems respond better to the needs of private enterprises In conclusion, exploiting the complementarities among training, education, and technology implies that these sys- tems have to be better integrated with one another. For train- E CNdutotes and~~~~ tehooy thsitgainiIotntrlydn CINTERFOR/ILO (2001) proposes rhat "vocational rraining ing grew in the region as a separate system from regular education schemes at the enterprise level The very nature of production implies Although theoretically complementary to each other, they showed in that employers must assess their skill and technology needs reality a disparity of objecrives, little coordination between themselves " simultaneously. Thus "enterprise-based" training systems 2 A related and relevant question is whether firms that are better may be best suited for countries that have reasonably well- linked to world markets, through either product or factor markets, developed education systems, and where economic pressures train their workers more In particular, are foreign-owned firms, which and incentives to acquire new technologies are present. Even ostensibly bring new technologies from abroad, more or less likely to train workers) Again, are enterprises that export goods more pressured developed countries such as Germany that have cooperative to stay competitive and hence invest in new technology and skills training systems-variants of those that proliferated in Latin WBES from around the world indicate that the answer to both these America-have introduced elements that resemble those of questions appears to be yes 131 CHAPTER 6 Technological Transitions and Elements of Technology Policy ECHNOLOGICAL INNOVATION IS A KEY SOURCE OF ECONOMIC, INCOME, AND employment growth. This premise has been validated by theory and empiricism and by now has been accepted by most countries. As a consequence, countries are trying to secure tech- nological and knowledge capabilities so as to reap their associated benefits. For the LAC coun- tries, this is a matter of utmost urgency since, as we showed in chapter 3, the technology gap-however it is measured-between them and the East Asian countries, or between them and the nat- ural resource-abundant countries, is quite significant, and has been widening through the 1990s. The issue for the LAC countries is how, given where they are now, they can best develop the capability to generate and manage technological change and innovation. That development is a process that does require the involve- ment of literally all actors in the economic environment, the public and the private sectors; and, in order for that development to be effective, it has to be coherent, coordinated, and tailored to country endowments and characteristics-as it was done in most of the successful East Asian and natural resource-abundant countries. While the government cannot mandate innovation, it can and should provide the leadership, coordina- tion, commitment, and incentives to induce the desired response from the drivers of the process-the firms and individuals. Individuals make educational and skill of foreign direct investment (FDI), from supplying original acquisition choices and firms contract skills and make inno- equipment manufacturers, from imports of capital equip- vation, technology, and production choices; and those selec- ment, or from licensing. The key to absorbing technology, tions are not independent of each other. These choices and a task that generally rests on a country's indigenous R&D decisions respond to expected returns and competitive pres- effort, is the availability of human capital and the readiness sures and are shaped by country conditions and by the to fully master new products and means of production incentive structure provided by the umbrella of the legal, (Forbes and Wield 2000). Without the human capacity to institutional, and fiscal framework provided by the govern- act on new technologies, any leaps made in the supply of, ment. And, as that environment evolves, so do the firms' and access to, new technologies are unlikely to translate into choices. A key premise of this report is that the availability significarit productivity gains. Yet, accepting that, we have and quality of skills constrain the firms' choices and possi- to recognize, as the evidence here presented shows, that with- ble innovations out a proper incentive and institutional framework the Knowledge is the composite of technology and skills. The expected private-led response to develop technological supply and quality of skills available influence the search for capabilities generated by large investments in R&D-the new technology, the monitoring of information about sci- key driver of technological advancement-will not happen. entific discoveries around the world; and the assimilation of Many LAC countries by the mid-1990s had reached struc- relevant local technology, whether from the spillover effects tural conditions-GDP/capita, years of education of labor 133 CLOSING THE GAP force, educational attainment, access to foreign technology, o Investments in R&D are long-term and risky Financial open trade regime and foreign investment laws, and buoy- markets in LAC seldom provide the right instruments ancy of the private sector-that were very similar, if not and term structure for financing R&D expenditures. better than, those that today's technologically successful o There are indivisibilities in innovation expenditure. Inno- countries had in the late 1970s when they took off Yet, prac- vation is lumpy, requires critical mass of investments, tically all of the LAC countries, as of 2002, have still to take and in a broader sense is subject to economies of scale. off on the path of technological advancements. o The process of innovation involves bringing diverse agents The purpose of this chapter is to understand the process together and that process is costly-it is subhect to coordina- of developing technological capabilities, the patterns of suc- tion failures andfree-riding behavior. The cost of bringing cessful transitions, to identify the main drivers and policy and coordinating the various agents and institutions levers, and to consider which ones are the most appropriate involved in the production of knowledge or concerted for specific country conditions, to advance in this process. actions is costly and not a single agent has incentives The chapter also evaluates the progress of the LAC countries, to bear those costs, since the benefits of such coordi- provides evidence of the determinants of technological efforts nation are often not exclusive and thus it leads to free and of the impact on the productivity of the respective driv- rider-type problems and results in coordination ers, and provides policy recommendations. In the course of failures. this exposition, the links between developing technological o Diffusion is not appropriable. No individual agent has capabilities and skills are highlighted; however, a detailed any incentives to widely and quickly diffuse knowl- discussion of these linkages is left for Chapters 5 (training) edge and innovation, since knowledge is not appro- and 7 (networks). priable and it is costly, and once diffused it can be accessed by anyone. Ra Donal,e for Government Action: Market and Coordination FaSiures Recognition that the process of knowledge acquisition The pursuit and management of knowledge is distinguished and diffusion is plagued by market and coordination fail- by a number of market and coordination failures that inhibit ures has led to an emphasis on the public interventions nec- the optimal level of private effort and investment and argues essary to ensure the socially optimal level of innovation. for government intervention to address those failures. Rely- Governments throughout the world have developed a long ing on the market alone will not induce appropriate levels tradition for intervening in the science and technology (S&T) of investment for innovation or optimal development of system in various ways. Practically all countries that have technological capabilities. Thus, government should play a successfully managed to advance on innovation have broader role in correcting those market failures and shap- addressed those market failures and taken the leadership of ing the incentives so that firms and individuals will make the process. First of all, governments have taken on the role the "efficient" decisions. The main market failures here are as overall coordinator of S&T activities. This entails gov- induced as a result of the following ernments leading the effort, setting up institutions (public or private), and implementing and coordinating policy pro- o Knowledge is not appropriable, it is a quasi-public good. grams to maximize net social returns. Moreover, govern- Firms making the investment are not always able ments intervene to correct market failures by supplying to keep the resulting rents to themselves-unlike funds, directly and indirectly, to public and private agents physical property, knowledge cannot be "owned" alike, subsidizing R&D expenditures; providing intellectual exclusively-and when agents are not compensated for property rights (IPR) protection to innovation; and facili- the effect of their actions, they do not invest optimal tating access to foreign technology. Finally, governments amounts have also undertaken in recent years to correct so-called sys- 0 Knowledge and innovation generate significant positive exter- temic failures (economies of scale, public goods, externali- nalities and spillovers. The social rate of return on R&D ties, and coordination issues) by setting up training programs expenditures is often three times as large as the pri- and specialized research centers (cooperation schemes vate rate of return (Griliches 1992) designed to promote joint public-private R&D efforts and 134 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHiNOLOGY POLICY improve the dissemination of information in society), and widely recognized as the world leader in innovation, exceed facilitating the development of networks for effective and those dedicated to innovation at the frontier by a factor of fast diffusion of knowledge. These latter types of interven- roughly 30 to 1. So, in some sense most of the world's firms tions are dealt with at length in chapter 7. are fundamentally adopting/adapting with a few doing little The timing and extent of these interventions matter. and a few doing cutting-edge innovation Countries differ in terms of initial conditions of technolog- Furthermore, such typologies are rendered less useful by ical capabilities, educational attainment, institutions, and the fact that they are difficult to measure. Unlike the case endowments As a result, and to increase effectiveness-since for education-for which it is easy to tabulate progress in those interventions are costly-the policies and types of terms of enrollments and attainments of primary, secondary, intervention have to fit each country's initial conditions, and tertiary education-there are not natural or easy mea- institutions, endowments, and characteristics For example, sures to classify firms into adoption, adaptation, and creation it might not make sense for a government to offer extensive modes We are left with imperfect aggregate measures benefits to stimulate R&D investments if the country has a of technological development. We can use proxies, albeit very poorly trained labor force As conditions evolve, so imperfect-such as patents, R&D investments, ICT tech- should the types and extent of government interventions. nology index, or even GDP per capita or educational levels, Failure to address this issue in a coherent way-as has been all of them highly correlated-to classify countries into the case for most LAC countries, as we will argue-condemns stages of technological development. countries to a slow and inefficient path of technological Figures 6.1 and 6.2 provide the estimated "standard" tra- advancement. jectories of countries-on measures of technological devel- opment--across the process of development. In all cases, we Developing Technological Capabilities plot a standardized measure of an index of the variable in The key drivers for firms to develop technological improve- question against the log of per capita GDP The x-axis is ments and capabilities are competitive pressures and expected logged to "spread out" and cover the poorer countries in the returns. These pressures force firms to constantly reevaluate sample (Latin America spans Haiti, with a value of 6, and and improve their production and organizational techniques Argentina, with a value of 8) The y-axis is not, since log- in search of efficiency gains and new processes and products. ging effectively minuscule numbers still carries the appear- Firms secure those improvements by adopting and adapt- ance of a significant quantity, even when it is not. This ing existing, usually foreign, technologies, to fit their needs, scaling does, however, overstate the degree of "takeoff" appar- bringing in minor product and process innovations, and ent, for instance, in R&D-which is partly, although not developing new ideas and creating new technologies. The exclusively, an artifact of logging income These trajectories traditional induscrial organization theory used those modes are as is, not necessarily at all optimal In fact, all of the to classify countries into stages of technological development successful-in terms of technological advancement- Today, however, these stages are not as sharply delineated countries have significantly departed from those trajectories as they might have been in the past, particularly because of as we show later on in this chapter the information revolution and globalization effects, and- Several factors merit mention. First, purchases of foreign for any country-considerable overlaps exist. In each coun- technologies, as measured by royalties payments, continue try, at any given time there are firms adopting, firms adapting, across the development process and are a complement to and firms creating. Further, the categories themselves are not R&D expenditures Similar arguments can be made-albeit exclusive Most purchases of foreign technology involve some less strongly-for FDI. That is, as mentioned before, coun- adaptation to local conditions and hence the number of cases tries never stop adopting/adapting foreign technologies In of pure adoption may be few. Finally, countries never stop fact, as recent surveys in OECD countries show (OECD adopting and adapting technologies (OECD 2002). Eaton 2002), many more firms do adoption than creation. Among and Kortum (2000), for example, argue that 86 percent of the richer countries, however, the relative effort in innova- France's technological progress is purchased from abroad, the tion does steeply rise. Second, these "innovation inputs"' are remainder being generated locally Jovanovic (1997) argues largely reflected as intermediate "output" measures as well. that resources dedicated to "adoption" in the United States, Patenting is low for the poorer of the LAC countries, but 135 CLOSING THE GAP FIGURE 6 1 R&D Effort, Licensing, and Development: Predictions from Median Regressions 8 7 - 6- 5 /S 4v 4 3- 0 - 47 57 67 77 87 97 107 Log GDP Per Capita R&D over GDP --- RoyYral erries o GDP| Source Iederman and Maloney (2002d) FIGURE 6 2 Onnovation and Development 5- 4 -0 3- 2 C 1~ ~ ~ ~ ~~~~~~~3 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY rises across time It is important to remember that patent- in chapter 7 shows that the effectiveness and efficiency of ing in general is not 'blue sky" innovation but often repre- R&D depend on both the education of the populace and sents process modifications associated with the adaptation trade openness) of existing technologies. In fact, we can think of the pur- The initial dependence on FDI or trade should not be seen chase of foreign technologies, the application of R&D to as passive. The overall goal is for firms to learn to innovate adapt to local conditions, and patenting as complementary over time and this implies active approaches toward ensur- activities that, over the course of development, move toward ing spillovers and learning. This, however, will be limited innovating at the frontier of a discipline and introducing to the degree of human capital, the coherence of the over- entirely new products all incentives to learn, and the overall functioning of the Finally, these stages may carry a normative implication economy. The process usually begins with firms beginning that somehow a firm or country should be adopting or adapt- to get exposed to newer products and process through trade ing at a given level of development. However, what is clear and FD], which become vehicles for the diffusion and adop- from the Asian experiences (see chapter 7) is that govern- tion of technology This exposure leads to imitation, and the ments sought to leapfrog stages of product development in process of technological mimicry-the first step toward any trying to move to the frontier of innovation as soon as pos- catch-up-begins. sible. Ideally a poor country would get the highest return Formal methods of technology transfer involving licens- to innovating at the frontier. In fact, last year's study de ing and royalty payments may be, on average, less common Ferranti and others (2001) argued that to maintain the at these early stages. Little formal R&D on the part of firms dynamism of its natural resource industries at the turn of is expected at this stage, although universities and public the 20th century, LAC needed to innovate at the frontier- research institutions and some firms may be engaged in con- in mining, in forestry-despite being far poorer in per capita ducting research in selected areas, such as in agriculture and terms than many LDCs today. natural resources Yet even at that stage some R&D is nec- What we find more useful in this report is classifying essary for its learning effects Having access to a labor force countries by the obstacles to their being an innovating that possesses general education and basic technical skills is country-and hence, by what policy packages make the most a critical determinant of absorptive capacity. While some sense. We propose the following phases: specialized skills may be called for, selective on-the-job training for secondary-educated workers can adequately fill Phase I: Plugging in to the Global Knowledge Stock- the gap and remove some of the obstacles to the adoption Efficiency and Reconnaissance of technology. Evidence from firms in Taiwan (China), Korea, Countries (and firms) in this phase concentrate their efforts Colombia, Mexico, and Malaysia, among others, supports the in seeking access to foreign knowledge and technologies- view that firm-led training and adoption of technology go through trade, FDI, and licensing-rather than developing hand in hand and that firms that do both reap the biggest their own. The argument is that substantial private sector productivity gains (see chapter 6). R&D efforts, or efficiently used public efforts, will be foiled if the basic conditions for efficient functioning of the econ- Phase 11: Catch-up and Consolidation of Innovative omy are not in place-which at this stage seldom are. Trade Infrastructure barriers and other market distortions impede not only com- Countries at this stage have reached a minimum adequate petitive forces that prompt firms to improve but also pre- level of education of their labor force, with well-functioning vent knowledge of what knowledge could be brought to bear. universities. As firms begin to face increased competition, An absence of basic human capital means not only that either domestically or by imports penetration, or when they spillovers from trade or FDI will be unrealized and, in fact, try to move from domestic to export markets, the response FDI is less likely to come. Poor communication and ICT is to improve quality, primarily involving new technology infrastructure also impedes knowledge of the outside world for agriculture and other primary sector goods and/or pro- and the ability to take advantage of even basic fragments of ducing relatively standard products, but at reduced cost This the global production chain (see Lederman and Xu 2002). is driven by low wages as well as the lowered cost of tech- (In fact, econometric evidence discussed in this chapter and nology through modifications All this should be facilitated 137 CLOSING THE GAP by beginning to develop innovation-related institutions and of highly skilled labor both become crucial. Whereas the by increasing linkages with the private sector. R&D begins nature of innovations at earlier stages is essentially "coded" to play a significant role here and thus skill needs become or embodied in products (hence making its transfer easier), more specialized at this stage. However, it is unlikely that it tends to become more "tacit" and disembodied during this a large pool of scientists and engineers can be fully utilized phase. When new knowledge is not completely embodied by firms Most tertiary-educated workers at this stage of the in a product, spillovers from the sale and use of a product economy find suitable employment in research institutions leading to imitation through reverse engineering also become or universities, and firms call upon them from time to time restricted. Human capital takes center stage; more than the to assist in making some of these adaptations to existing mere existence of skilled workers, their proximity to one technology. Links between firms and some institutions begin another and their mobility become the medium of technol- to form, albeit in an unpredictable and spotty manner ogy transfer and the fuel for innovation. This proximity and because of the information and cultural gaps that exist mobility are critical to connecting workers between firms between these entities. The government here is beginning as well as with researchers in university research institutions, to play a major role leading the process, establishing the and is often found to characterize places such as Silicon main components of a National Innovation System (NIS) and Valley, which have become synonymous with cutting-edge addressing market and coordination failures by providing innovation While R&D is the main element for the creation fiscal and economic incentives for private R&D, and incen- of knowledge, large amounts of private R&D is now the key tives for the collaboration of public and private institu- factor and driver. tions-such as setting up specialized sectoral scientific institutions or laboratories. In this phase, well-targeted Policies fr the First Stage of Transition: public support for specific sectors-shown to be both impor- Spillovers, Diffusiion Mechanisms, tant for the country and exhibiting signs of potential excel- and IUncipient and Selective R&D lence-is warranted, as can be seen from the experience of Governments and firms have choices on the policies and successful (transition) countries Increasing the financing of instruments-and extent of use-to build their domestic R&D and setting intellectual property rights, albeit at rel- technological stock and capabilities. These choices depend atively lower levels of protection, are essential at this stage. on a number of factors, and a critical one is the skill level of the labor force Predictably, the availability of technical Phase III: Technological Leaders and managerial skills becomes the natural limit and often At this stage, firms already involved in global markets- the most pressing constraint facing firms in advancing in the adapting existing technologies and selling the resulting prod- development of technological capabilities. Without an ade- ucts at lower cost than their competitors-find margins eroding quate skilled labor force, the impact of a full-fledged R&D as new low-wage entrants take away market share. Sustaining program is bound to be ineffective and wasted. Then tech- a position in global markets then necessitates a leap forward nology policy should aim to facilitate profitable exposure into creating new products and processes. This requwres firms to foreign technology and on the leveraging via impact on to make large investments in in-house R&D activity consist- facilitating diffusion and spillovers, and in incentivating ing of expensive plant and equipment for advanced laborato- incipient and selective R&D programs. ties and testing facilities. Given the externalities in knowledge Foreign technology can be accessed through trade, FDI, creation arising from the less-than-full appropriability of the and licensing. The rationale for trade, FDI, and licensing as returns to such investments, this R&D push is likely to be com- instruments for the transfer of technology comes from endoge- promised unless mechanisms that correct for this market nous growth theory (Romer 1986, 1990, Lucas 1988). That failure-such as strong intellectual property rights (IPRs) or theory posits that the returns to the accumulation of knowl- significant fiscal and economic incentives and a coherent incen- edge capital and human capital do not diminish at the aggre- tive and institutional framework-are in place. gate level because of positive spillover effects, and that policies The role of the government here is crucial in creating can have a permanent impact on the rate of economic growth. strong incentives and an institutional framework, and in con- Grossman and Helpman (1991) extend the analysis by explor- solidaring a highly effective NIS. The quantity and quality ing endogenous growth theory in an open economy setting 138 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY The basic idea is that goods embody technological know- limits to the share of foreign investment and restrictions on how and therefore countries can acquire foreign knowledge investing in certain sectors. A more complicated factor is the through imports. There are two principal mechanisms for extent of fiscal benefits, such as preferential tax treatment. technology diffusion through trade: One is through direct Two other key factors that impact FDI flows-market learning about foreign technological knowledge from the economic opportunities and the intellectual property rights imports, and the second is through employing specialized and framework-are discussed separately below. advanced imported intermediate products, which involves the Finally licensing involves the contractual transfer of implicit usage of the knowledge that was created with the knowledge between firms. Given that licensing provides foreign investors' R&D investment. knowledge in a more accessible manner than FDI, many Thus, not surprisingly, more than any other imports, countries such as Brazil, India, Mexico, and Japan have in imports of capital goods are especially important for tech- the past actively discriminated against direct investment by nology transfer, owing to their rich knowledge and poten- firms and favored technology licensing Multinational tial productive applications. (Eaton and others 2000). The firms, on the other hand, have an incentive to prevent a best-known example of a country following a deliberate strat- reduction in the value of their knowledge-based assets. This egy of acquiring technology through trade, and doing so can arise because the licensed firm becomes a competitor or successfully, is Korea. The government encouraged imports because the licensed firm allows knowledge to leak This of capital goods and at the same time encouraged master- seems to have led multinational firms to use licensing or jomit ing of the technologies embodied in these imports through ventures to transfer older technologies and use FDI for the various incentives, including R&D for its learning effect latest technology (Mansfield and Romeo 1980). In addition, FDI can provide technology transfer if local Direct purchases of technology from abroad have been at firms can copy or imitate technology used by affiliates of the center of the make-versus-buy debate in the 1970s and foreign firms operating in the local market, or if backward 1980s. The evidence is that, for countries far from the tech- and forward linkages with the foreign firm lead to inter- nology frontier, and with low skill levels, the "buy" option industry technology upgrading (vertical spillovers), or by translates into bigger productivity gains for firms. India, for labor turnover from the multinational firms to local firms, example, tried to force an R&D push at early stages of devel- bringing acquired technological know-how (horizontal opment when adoption would have been more appropriate; spillovers). Foreign direct investment can also lead to tech- Brazil may have a similar story. In the case of India, the rate nology upgrading by introducing competitive pressures of return to technology purchase is about 44 percent higher While in principle FDI could have a more direct impact on than the rate of return to domestic R&D. When the sectors technology transfer, because of its in situ production, usu- are divided into two categories as scientific and nonscien- ally the FDI volumes relative to the trade volumes are much cific, even more interesting results emerge. In the scientific smaller (with rare exceptions), limiting their impact. The sectors, the rate of return to technology purchase is 166 per- best examples of a country acquiring technological capabil- cent, whereas the rate of return to R&D is merely 1 percent; ities through FDI have been Singapore and Ireland in nonscientific sectors, the rate of return to technology pur- The institutional and policy framework for FDI is usu- chase is 95 percent, as opposed to 64 percent for domestic ally embodied in specific legislation crafted for that purpose, R&D (Basant and Fikkert 1996). and obviously in the effectiveness of its implementation. The These findings support the premise that the more uni- more open, transparent, and fiscally favorable and the less form or standardized (what we call "embodied" and "codi- administratively cumbersome economic conditions, the larger fied") knowledge is, the greater its transferability, and hence the flows of FDI. A complementary jump-starting instru- the greater possible returns from its transfer through pur- ment that has been reasonably effective in Latin America and chasing The phenomenal differences in the rate of return the Caribbean in attracting FDI is the export-processing to these two modes of upgrading discussed above clearly zones (EPZs) or free trade zones (FTZ), which function as imply that restrictions on technology purchases come at great an institutional shortcut to address broader problems in cost in terms of reduced productivity to firms. countries with institutions that impede FDI (de Ferranti While the role of trade, FDI, and licensing in pulling and others 2001). The most common obstacles have been domestic firms closer to the frontier is acknowledged, it must 139 CLOSING THE GAP also be noted that this process is far from automatic. And B BOX 61 that is reflected by the empirical evidence. A plethora of lhe 3 >o3 of & lnOViOn Ead Lrning results on the subject shows mixed results (Saggi 2002). Three things make technology diffusion non-automatic. The endogenous growth literature casts industrial inno- First are the requirements of a threshold level of absorptive vation as a key driver of growth (see Romer 1990; capacity, often captured in the quantity and quality of skills Aghion and Howitt 1992). In the literature, R&D, along available, as we show below Second is the need for com- with human capital, plays an important role in the cre- plementary R&D expenditures at the firm level for at least ation of new technologies and innovations. However, if their learning effects for adopting and adapting (box 6 1). R&D were only a source of new innovations, its relevance Third are the information asymmetries and strategic to developing countries would be much smaller. Case interests and the extent of networks in the economy that studies show that much indigenous effort is required for impact the degree and speed of diffusion. Potential users technological change even in industrializing economies face uncertainty, information and learning costs, and other (see Pack and Westphal 1986). Thus, R&D has a second externalities that may result in underinvestment in avail- role-in developing a "firm's ability to identify, assim- able technologies Similarly, potential suppliers of infor- dlate, and exploit knowledge from the environment- mation and assistance also face learning costs, may lack what we call a firm's "learning" or "absorptive" capacity" expertise, or face other structural barriers in promoting the (Cohen and Levinthal 1989) diffusion of rewarding technologies. System-level factors Empirical studies have typically tried to measure the such as the lack of standardization, regulatory impedi- impact of R&D on economic growth, without discin- ments, weakness in financial mechanisms, and poorly orga- I guishing between the two roles of R&D. These studies nized interfirm relationships may also constrain the pace have seldom focused on Latin America because of a lack of technology diffusion. These constraining or facilitating of information. Overall, these studies show a significant factors in technology adoption can be placed in four contribution of R&D to productivity growth, with social categories: returns to R&D exceeding private returns. Private returns have been estimated in the range of 10-20 percent by o Firm level. At the firm level, managers face problems i OECD, and Griliches (1992) estimated social returns of time, cost, absorptive capacity, and technical exper- i ranging from 20 to 60 percent (Yusuf and Evenett 2001). tise in accessing information on available technologies A few empirical scudies have tried to distinguish and available solutions. between innovation and learning, the two roles of R&D. o Institutional infrastructure. The degree and effectiveness (Examples are Griffith, Redding, and Van Reenen 2001, of innovation-related institutions and their linkages for a panel of industries across 13 OECD countries; to the private sector and the degree of networks play Parisi, Schiantarelli, and Sembenelli 2002, for a sample significant roles in aiding technology diffusion. of Italian firms; and Ray and Bhaduri 2001, for Indian o Business infrastructure. The organization of industry and electronics and pharmaceutical firms.) While these stud- intensity of competition; the strength and nature of ies use different methodologies, by and large they show interfirm relationships, such as closeness to customers, that R&D has an important role to play in the learning suppliers, vendors, and subcontractors; and the asso- and absorption of new technologies, in addition to its ciations within firms via trade, industry, or regional more traditional role of innovation. associations also play a role in aiding or impeding tech- nology diffusion. A variety of policy measures have been pursued to address o Social infrastructure. Education, skill level, and training these constraints. A convenient way to organize thinking about systems can again play a critical role in either enhanc- them is as supply-side and demand-side measures (box 6.2). ing or severely limiting a firm's capability to upgrade Finally, to facilitate access to the global storehouse of technology. Most firms devote too few resources to work knowledge, and to disseminate knowledge effectively, the force training, and educational institutions may not development of the ICT sector is most warranted at this always be responsive to their training needs. stage The quicker the access to information and diffusion 140 TECHNOLOGICAL TRANSI riONS AND ELEMENTS OF TECHNOLOGY POLICY BOX 62 Supply-Side and Demand-Side Measures to Improve Technology Diffusion Supply-side tools: augmnent sources of information and assistance Demand-side tools: increase willingness and absorptive powoer available tofirms. Examples of supply-side instruments are of firns to adopt technologies. Examples of measures that information search and referral services, such as those offered attempt to increase the absorptive power of firms to adopt by Valencia Institute of Spain, and the Awareness build- new technologies are programs that offer training assistance ing technology demonstrations offered by Center for Man- or personnel exchanges. Because the effective deployment ufacturing Information Technologies at the Georgia of technology and improved operational techniques Institute of Technology in the United States. Information involve changes in human capital requirements, training search services aim to reduce the information search costs in the form of classroom training, management seminars, associated with technology diffusion. Such services add fur- team-building workshops, and distance learning serves as ther value by often qualifying information requests and a useful tool. These measures address the tendency of users matching user needs,with appropriate resources. Technol- to underinvest in human capital development, which not ogy demonstrations seek to make potential users more only hincders the initial decision to deploy a technology but knowledgeable about available technologies, their possible can also lead to subsequent inefficiencies once in use. Two applications, and their benefits and costs. Such demon- programs that serve as best practice are the Australian stration services are offered by the Georgia Institute of Tech- cooperative research centers that offer focused training on nology for potential users to see and try. Similarly, Japan's specific industry needs in specific technological areas, and municipal technology center (Kohsetsushi) demonstrates the Local Enterprise Councils in Scotland, which identify new technologies to firms, often extending the assistance industry training needs and support training initiatives to hands-on training and pilot production. Similarly, with private and public resources. When firms lack the Baden-Wiirttemburg's model for technology transfer is car- resources to apply (or train) their existing personnel to new ried out through the work of the Steinbeis Foundation in research and technology projects, measures intended to Germany. The main aim of the foundation is to help local support r he secondment of personnel to technical centers firms keep up with the advent of new technologies via by or other firms where such technologies are developed or identifying potential partners and assisting them in their in use come in very handy. For example, in Japan, public negotiations; and offering expertise in special technologies, technology centers accept staff from firms to receive train- management, and marketing of the acquired technology ing in new technologies. of existing knowledge across the country, the larger the integration within the supply chain, so spillover effects are impact on the development of technological capabilities and maximized. (This latter factor is developed in the next the faster the transition The starting point is an efficient chapter on networks.) and as competitive as possible telecommunications sector, Coe and Helpman (1995) provide an empirical imple- followed by the implementation of the legal and regulatory mentation of the open economy endogenous growth model framework for ICT as described below But as mentioned ear- They construct an index of foreign R&D as the trade- lier, the labor force's skill level is most important to secur- weighted sum of trading partners' stocks of R&D. They find ing the greatest benefits that for a sample of developed countries, both domestic and "foreign" R&D have a significant impact on total factor pro- Trade as a Channel for Knowledge Transfer ductivity (TFP), and that the latter increases with the general Economic evaluation studies consistently confirm trade as a degree of openness of the economy and with openness toward significant channel for technology transfer. Yet for any the larger R&D-producing countries. amount of trade, the impact of productivity is not fixed, ir Coe, Helpman, and Hoffmelster (1997) examine the same depends on a number of factors. The education and skill level issue for developing countries. They find that developing coun- of the labor force is a main factor and so is the degree of tries benefit more from foreign R&D spillovers, the more open 141 CLOSING THE GAP they are and the more skilled their labor force is. These find- countries and the openness of the countries in the study that ings provide support for the hypothesis that trade is an impor- provided this stock of knowledge. tant mechanism through which knowledge and technological The study shows that, for the LAC countries, "foreign progress are transmitted across countries. The idea is that R&D" has a positive impact on R&D-intensive industries, importing countries (firms) learn from the knowledge embed- but this impact is dependent on the level of education in ded in the inputs that they import. Cross-country studies that the country In other words, the greater the educational have attempted to measure technology transmission through attainment of the country, the greater the impact of foreign trade from developed to developing countries have found the R&D on TFP through trade Educational attainment is mea- elasticity of total factor productivity to some measure of for- sured as the share of the population aged 25 years and above eign R&D capital stock' to vary between 0.02 to 0.10. In other that completed secondary education The study thus shows that words, a 100 percent increase in the R&D capital stock of education (skills) andforeign knowledge (R&D) are mutually rein- developed countries raises TFP in developing countries by an forcing in their effet on TFP in R&D-intensive induistries. Second, average of 2 to 10 percent. Coe, Helpman, and Hoffmeister from the construction of foreign R&D, the results for LAC (1997) estimate that in 1990 the total spillover effects from imply that education and openness are also mutually rein- R&D in the industrial countries may have increased output forcing in their impact on TFP in R&D-intensive industries. in the developed countries by about $US22 billion. Education reflects the capacity of the LAC countries to absorb A background study for this report examines the impact knowledge from the North and transform it into higher of trade with OECD countries in raising the productivity productivity. And absorptive capacity is clearly more of 9 LAC and 16 other developing countries through important in high R&D-intensive industries than in low spillover of knowledge created by R&D (Schiff and Wang R&D-intensity industries. 2002). The study covers the period 1976-98. The variable Tables 6.1 and 6.2 show that education helps in the "foreign R&D" is a weighted average of R&D in the OECD absorption of foreign technologies-the complementarity countries by industry, where the weights are a measure of TABLE 61 bilateral openness with the respective country by industry. Tact r The measure of the stock of foreign R&D obtained by an importing country at the industry level explicitly incorpo- A I PERCENT INCREASE IN EDUCATION* IN LAC rates the production structure of the economy as reflected Increases TFP by 5 78% in Low-Intensity R&D Sector in the input-output relationships. The study specifically Increasei TFP by 10 93% in High-Intensity R&D Sector * This refers to an increase in the share of the population aged 25 and above thac examines the impact on TFP of international technology dif- completed secondary education fusion, education, and governance for the LAC region. It thus Note Interaction effect between foreign R&D-trade related-and LAC education is positive and scrongly significant captures both the stock of knowledge available in the OECD Source Schiff and Wang (2002) TABLE 6 2 Impact through Trade of a 10 Percent Increase in Education on UP Growth Rate (%), 1998 SECONDARY SCHOOL IMPACT COMPLETION RATIO THROUGH ABSORPTION WITH 10 PERCENT COUNTRY DIRECT OF FOREIGN R&D TOTAL 1998 INCREASE OVER 1998 Bolivia 0 036 0 034 0 07 6 2% 68% Chile 0 087 0 075 017 15 0% 16 5% Colombia 0 049 0 041 0 09 8 4% 9 3% Ecuador 0 046 0 041 0 09 8 0% 8 8% Guatemala 0 015 0 013 0.03 2 5% 2 8% Mexico 0 075 0 065 0 14 12 9% 14 2% Panama 0 094 0 086 0 18 162% 17 8% Trinidad and Tobago 0 071 0 065 0 14 12 39% 13 6% Venezuela 0 025 0 021 0 05 4 3% 4 8% Note The R&D-intensive industries in the study are (1) 382-Non-Electrical Machinery, Office and Computing Machiner5, (2) 383-Electrical Machinery and Communicarton Equipment, (3) 384-Transportation Equipment, (4) 385-Professional Goods, (5) 351/2-Chemicals, Drugs, and Medicines, and (6) 353/4-Petroleum Refineries and Products Source Schiff and Wang (2002) 142 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY effect-and that this effect, shown in the second column of developed countries, but not weaker than those of many of table 6 2, is almost as big as the direct impact of education Asian countries One reason is that not all types of FDI on productivity. For example, raising the stock of adults with necessarily involve significant amounts of knowledge trans- secondary school education by 10 percent raises the TFP fer Balasubramanyam and others (1996) find that the growth growth rate from a low of 0 015 percent for Guatemala to effect of FDI is weaker in countries that pursue policies of a high of 0.094 percent for Panama In addition, this increase import substitution rather than export promotion. This in education also raises TFP growth rates by aiding the could be because export-oriented FDI is more likely to be absorption of foreign technologies by 0 013 percent and internationally competitive. Another reason is that, even if 0.086 percent, respectively the technology component of the FDI is important, foreign firms may take actions to limit spillovers to the host coun- Impact and Determinants of Foreign Direct Investment try's domestic firms. Furthermore, even though technology Most studies that deal with the impact of FDI on the domes- spillovers may be significant, there may be other negative tic economy agree that recipient firms benefit from foreign effects of FDI that counteract the positive impact of the investment The few studies that deal with inter-industry spillovers. For instance, FDI may reduce the share of the spillovers usually find these to be positive However, results market left for domestic firms, and under economies of scale, on mntra-industry spillovers are more ambiguous owing ro a a smaller output implies a higher average cost. Also, the com- variety of factors (Schiff 2002). These studies typically esti- position of FDI matters. In a number of countries, a large com- mate, at the firm or plant level, the statistical impact of some ponent of FDI has been in rhe service sector and retail measure of FDI on a measure of productivity, most often total distribution, both low R&D intensive (for example, Brazil and factor productivity (TFP).2 A study commissioned for this Mexico). There the level of technology and knowledge trans- report (Park 2002) shows very clearly the impact of the insti- fer is reasonably small Finally, while FDI can provide access tutional regime on FDI flows as reported in table 6.3. to technology, the domestic firms' ability to effectively absorb However, whether spillovers to domestic firms increase knowledge spillovers can limit the usefulness of FDI. This last with stronger intellectual property rights (IPRs) is unclear point resonates strongly with the main theme of this report because, even though technology transfers to licensed firms or recipients of foreign investment increase, the amount of Minimum Threshold of Human Capital Required leakage per unit of technology transfer is likely to fall because to Reap FDI Benefits of the stronger IPRs (such as the ability to go to court to As we have shown for trade, the skill level of the labor force prevent leakage or get compensation) While there has been plays a major role in determining the extent of benefits to a strengthening of intellectual property regimes in most LAC be captuied through FDI. Borensztein, De Gregorio, and Lee countries over the last decade, they are still weaker than those (1998) study the effect of FDI flows from industrial coun- tries on economic growth from a sample of 69 developing TABLE 6 3 countries over the 1980s and 1990s. They find that FDI has Effects on Foreign Direct Investment a positive overall effect on per capita GDP growth, and that IMPACT ON FOREIGN DIRECT INVESTMENT (FDI) INFLOWS the effect increases with the level of human capital in the (AS A PERCENT OF GOP) OF A I PERCENT INCREASE IN host country, but it is subject to the existence of a minimum PERCENT IMPACT ON FDI INFLOWS threshold level of human capital in the host country. They ONE PERCENT INCREASE IN AS PERCENT OF GDP Private R&D per worker 0 492* estimate this threshold level to lie between 0.52 and 1.13 Public R&D per worker 0 553* years of average secondary school education in the male pop- Patent rights strength 1 578*** ulation aged 25 years or older (table 6.4) Enrollment in tertiary education -0 030 Scientists and engineers 0 884*** In an attempt to identify the technology diffusion effect Adjusted R2 0 28 of multinational corporations (MNCs) among all their Number of observers 155 productivity-enhancing effects, Xu (2000) considers the Note Estimation is by fixed effects regression for 58 countries, including 19 LAC countries ***, **, and * denote significance levels of I percent, 5 percent, and 15 impact of technology transfer spending of U S. MNCs on percent, respectively Since all variables except the constant (not shown above) were in productivity growth in 40 developed and developing log, the above numbers are just the coefficients from ehe regression equation Source Park (2002) countries from 1966 to 1994 The results from this study 143 CLOSING THE GAP TABLE 6 4 Human Capital Threshold Required for Positive Impact of FDI HUMAN CAPITAL THRESHOLD (MINIMUM AVERAGE YEARS OF SECONDARY SCHOOL EDUCATION STUDY WHAT THE STUDY MEASURES IN MALE POPULATION >25 YEARS) Borensztein, De Gregorio, Irnpact of FDI flows from industrial countries on per capita GDP and Lee (1998) growth of 69 developing countries 0 52-1 13 years Xu (2000) Impact of technology transfer spending of U.S MNCs on TFP growth of 40 countries 1 4-2 4 years FIGURE 6 3 iWean Tariff Barriers (percent), 1990 and 1999 30- 20 10 0 sit 40 g7 o o< +s .tt I 0 1990 0 1999 * 1992 & 1999 ** 1993 & 1997 *** Note External tariffs for the European Union have been used Source World Bank (2001c) support the previous one (table 6.4). The presence of MNCs high levels of protection. The 1990s have seen significant has a positive and statistically significant effect on TFP growth lowering in the barriers to trade in most LAC countries, with for countries with average male secondary school attainment mean tariff rates coming down dramatically (figure 6.3). of at least 0.52 years in the population above 25 years of age Brazil has the highest rate (13.6 percent) among the LAC In addition, by distinguishing the technology diffusion effect countries while Costa Rica has the lowest at 7.2 percent. of MNCs from other productivity-enhancing effects, the These numbers compare favorably with those for Korea at study shows that a much higher human capital threshold is 8.7 percent. The European Union (EU) and the United States required to benefit from technology transfer. They estimate had slightly lower rates, at 5.6 percent and 4.8 percent, this threshold to lie between 1 4 and 2.4 years. In other respectively, in 1999 Since openness is a critical instrument words, technology transfer from MNCs does not have a sig- for developing technological capabilities at all stages, LAC nificant impact on the productivity of countries with male countries have done well at least with regard to tariff levels. secondary school attainments below 1.4 to 2.4 years. This opening up of the trade and investment regimes through tariff reductions and deregulation has caused a sig- R.,'S Pe13 onnDrace nificant rise in the amount of goods, services, and invest- ments flowing into the LAC region as was shown in chapter Trade Regime 3. However, Brazil and Argentina remain fairly closed LAC countries have made significant progress since the economies compared to some other LAC countries such 1980s, when most countries were practically closed with as Costa Rica, and countries such as Korea, Spain, and 144 TECHNOLOGICAL TIRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY especially Ireland. Also, there has been a decrease in non- tariff barriers, safeguards, countervailing duties, and tariff barriers in most LAC countries. Yet in some coun- antidumping framework While all are important, the main tries-Brazil, Mexico, and Venezuela-they still remain one clearly is the external tariff regime. The lower those bar- high (table 6.5). Continued efforts there are needed to fully riers and tariffs, the larger the trade imports Fiscal reasons secure the benefits of increased trade, mostly through reduc- and claifns for the protection of domestic industries are usu- tions in non-tariff barriers, particularly for the aforemen- ally the countervailing factors for not lowering tariffs. tioned countries There is little more that most LAC countries can do in Imports of capital goods are another measure of openness the trade area to facilitate technological transfer or the tran- to foreign technologies and perhaps also a measure of the sition to increased technological capabilities. While there are readiness and capacity of domestic firms to absorb foreign always improvements that can be made they are marginal technology. The numbers for LAC countries remain low as and unlikely to provide a major boost to technology transfer. shown in chapter 3 Among East Asian economies, China, There are some exceptions such as Brazil and Argentina. Of Korea, Malaysia, and Thailand have all increased capital course thie benefits of trade can and will increase as the other goods imports (Indonesia being the exception). Among LAC pieces of a coherent science and technology system- economies, only Argentina and Venezuela have increased improved skills of the labor force, stronger network linkages, capital goods imports, the others staying nearly the same selected increases of R&D efforts, and so on-fall into place, over two decades (Brazil, Uruguay) or even falling (Chile, and that is where the efforts on LAC should be placed. In Columbia, Mexico, Peru) summary, trade policies in LAC are not any longer a lever to The broad institutional and policy regime defining the increase technological capabilities, however. While avoiding trade regime rests on a country's external tariff regime, non- reversal on trade policies should be a given, the efforts should be placecl elsewhere-that is, on capturing an increased share TABLE 6 5 of the benefits of the trade volume, through expanding edu- Non-Tariff Barrier Levels for LAC and East Asian Counitnes cation, and facilitating spillovers by strengthening network CORE NTMs developrnent and linkages. LAC COUNTRY 1989-94 1995-98 Argentina 3 1 2 1 Foreign Direct Investment Bolivia 0 0 - Countries in Latin America and the Caribbean compare very Brazil 16 5 21 6 favorably to newly industrialized countries (NICs) in secur- Chile 5 2 5 2 Colombia 55 2 10 3 ing FDI. In Latin America and the Caribbean, FDI flows Costa Rica _ 6 2 increased from 0.5 percent of purchasing power parity (PPP) Dominican Republic - 6 2 El Salvador - 5 2 GDP in L989 to 3.2 percent of PPP GDP in 1999 on average. Mexico - 13 4 Brazil is the most significant host of FDI in the region in Paraguay 27 8 0 0 absolute terms (US$34 billion in 1999), while in relative Peru Uruguay 6 3 0 0 terms, Chile receives the largest fraction of FDI in the region Venezuela 32 3 17 7 (10.3 percent of PPP GDP in 1999). The overall increase in FDI largely reflects the liberalization of investment CORE NTMs regimes throughout the region and fiscal incentives designed ASIAN COUNTRY 1989-94 1995-98 explicitly to attract foreign investors. The index of foreign Fiji 5 2 investmfnt barriers for LAC compares well with that for HonKong (China) 53 6 3213 high-income countries, and is the lowest among all regions Korea 1 0 25 0 (table 6 6) Mexico's position as the second largest reservoir Malaysia 1 19 6 of FDI in the region is largely a result of NAFTA, which Philippines 2 Singapore 2 0 2 1 boosted U.S. investments into Mexico. Costa Rica is a case Thailand 36 5 17 5 in point with regard to the effects of fiscal incentives on the ANote Non-rariff measures (NTMs) are calculated as frequency ratio in percent of all HS flow of FDI. Also, a number of countries in LAC have used 2-digit product categories Core NTMs include licensing, prohibitions, quotas, and administered pricing quite effectively the EPZ concept to attract FDI (de Ferranti 145 CLOSING THE GAP and others 2001). However, despite liberalizing their FDI The index of foreign investment barriers ranges from 1 regimes, a number of LAC countries retain barriers to FDI. to 5. The exact assessment of values is based on the follow- These relate more to the cost of doing business and apply ing criteria: (1) having very low barriers to foreign invest- to both domestic and foreign investors ment means that there is open treatment of foreign TABLE 6 6 investment and an accessible foreign investment code; (2) Index of Foreign Investment Barriers, 2000 having low barriers means that there exist certain restric- SCOR-E tions on sectors such as utilities and natural resources, but High-income countries 2 1 that there is a limited, efficient approval process for new Developing countries 3 0 projects; (3) having moderate barriers means that there are Europe and Central Asia 2 9 restrictions on many investments and a bureaucratic South Asia 3 0 East Asia and the Pacific 3 1 approval process, but at least there is an official policy that Sub-Saharan Africa 3 2 conforms to an established foreign investment code; (4) Latin America 2 2 Argentina 2 0 having high barriers means that foreign investment is per- Bolivia 2.0 mitted on a case-by-case basis with a bureaucratic approval Brazil 3 0 process that may be marked by some corruption, (5) having Chile 2 0 Colombia 2 0 very high barriers means that the government actively Costa Rica 2 0 tries to prevent foreign investment and there is rampant Ecuador 2 0 El Salvador 10 corruption i the process. Guatemala 3 0 Ireland is an extreme case with regard to FDI. Based on the Guyana 3 0 evidence presented in figure 6.4, FDI represented approxi- Honduras 3 0 Mexico 2 0 mately 25 percent of the country's GDP (PPP) in 1999 (with Nicaragua 2 0 a large share originating in the United States), compared to Panama 2 0 0.2 percent only 10 years earlier. This remarkable increase is Paraguay I 0 icnie, peec Peru 2 0 partly a result of fiscal incentives, the presence of a skilled work Uruguay 2 0 force proficient in English, and a location within the Common Venezuela 3 0 Source Interregional Cooperation in Trade and investmenr in Latin America European Market providing access to that market. Spain has (STtESCAP/2069) also seen a notable increase over the past 10 years, but to a much FIGURE 6 4 Foreign Direct Investment, 1989 and 1999 25 . 20- 15 l - 0 10 1 (~5 0 1989 0 1999 Source. World Bank (2001a) 146 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY smaller extent. In the case of Korea, FDI takes up a fairly low negligible Importance in India, it was a key instrument for share of GDP (PPP). This, in turn, is partly testimony to Singapore (box 6 3, and see Pack 2000, p. 74). government policies, which consistently have preferred import A final point on the interaction between FDI and trade of capital goods to FDI (see Viotti 2001). Last, while FDI has policy is worth bearing in mind. Much of the FDI flowing BOX 63 Singapore's Reliance on FDI Singapore's technological development can be attributed choilce of becqming outward oriented and to rely on FDI, to reliance on international markets and FDI, with active there is rauch less room for policy mistakes and successful government intervention during the last 40 There are some lessons from Singapore's policies years. Historical circumstances forced Singapore to rely on toward FDI: export orientation rather than on import substitution, and on FDI rather than on indigenous companies. The lack of o Singapore managed to get access to foreign knowl- a sizable domestic market after the separation with edge and technology through a laissez-faire policy Malaysia necessitated the export orientation, and the polit- toward multinational corporations (MNCs) rather ical turbulence made the government mistrust the domes- than by various requirements focusing on their oper- tic business community. Given these circumstances and ations. policy choices, the government acted with pragmatism and o Since foreign MNCs started to locate in Singapore, efficiency to maximize the economic benefits of the nation. the government has constantly been encouraging The ability to act on opportunity has been a characteris- them to upgrade their activities. tic of the Singaporean government, as witnessed by the ini- o The government has repeatedly used strong eco- tial efforts to attract foreign electronics companies when nornic incentives rather than legislation or require- they became interested in outsourcing labor-intensive pro- ments to achieve this upgrading. The economic I duction, or the formation of an international currency trad- incentives have ranged from tax exemptions to the ing market when international actors were looking for the training of workers and local suppliers, and the local opportunity of 24-hour-a-day trading. MNCs have responded positively. There are reasons why the Singaporean government has been able to make politically difficult policy choices and Recently, Singapore has tried to make the second tran- why it has been development oriented. The government's sition to an economy that builds upon its own innovations autonomy from important interest groups has been crucial and technology rather than importing ready-made tech- for its ability to shape the economic policies without much nologies through MNCs The policies have, once again, domestic interference. When the domestic business com- been based on financial incentives, including tax exemp- munity and the labor movement became marginalized tions for the training of workers and local suppliers, to because of political conflicts, it made the government increase local knowledge, create a domestic business com- autonomous from the strongest interest groups in the munity in technology-intensive sectors, and make the society Still, there is nothing that guarantees that an MNCs locate R&D in Singapore. Whereas the domestic autonomous government will always act in a growth- business community is still rather weak, the amount of enhancing way rather than enriching itself through vari- R&D conducted in Singapore increased substantiallyI ous rent-seeking activities The reason why the Singaporean during the 1990s. Much of this R&D took place within government became development oriented is twofold First, the MNCs Most R&D that takes place in Singapore is its public legitimacy has largely rested on the provision of advanced development rather than basic or even applied increased living standards-most Singaporeans seem to research., have accepted restrictions on personal freedom as long as there is a steady increase in incomes. Second, given the Sorce: Blomscrdm anid orhers (2002) 147 CLOSING THE GAP to inward-oriented developing countries is undertaken to on average, have done reasonably well on access to foreign substitute for imports from richer countries, which often con- technologies, accomplishing most of what is required. While sist of relatively advanced products The production tech- some specific countries should improve their regimes (such nologies used by MNCs in import substituting economies as Argentina and Brazil), most of them ought to focus on are therefore often relatively capital- and skill-intensive com- increasing the benefits of that exposure to foreign technol- pared to those employed by local firms. MNCs investing in ogy, by fostering education, network development, and outward-oriented economies, by contrast, can be expected selected R&D efforts. Without adequate human capital or to employ technologies that are more in line with the host investments in R&D, and integrated networks, spillovers country's factor endowments because these MNCs are from those vehicles are likely to be limited, as has consis- exposed to tough competition from the world market tently been shown (Saggi 2002). (whether or not they are export-oriented) and tend to use production technologies that fit the host country's compar- The TelecommunicationsllCT Sector: A Critical ative advantage. To the extent that the trade regime then Instrument for Diffusion and Advancing the has implications for the technological composition of FDI, Development of Technological Capabilities it also has very important implications for the amount of The telecommunications/ICT sector today more than ever spillovers that can be expected underpins the process of technological development Most of the countries that managed successful and rapid transi- Licensing tions make a special effort to enhance the ICT sector. It can Korea ranks at the top in license payments, taking up be a powerful instrument for access to the global storehouse approximately US$55 per resident in 1999, which in turn of technological information and for diffusion of knowledge, represents a doubling from the 1990 level. Similar devel- as well as for facilitating competition through fast and cheap opments have taken place in Spain and the United States. access to information. While ICT in itself will not generate However, nowhere is the flow of licensed technologies greater or increase global knowledge, it does increase local knowl- than in Ireland, which paid close to US$2,000 per resident edge and is the key instrument for a broad and speedy dif- at the end of the 1990s, approximately 40 times as much fusion of information and knowledge ICT makes it possible as Korea. That complemented Ireland's FDI strategy nicely. to share knowledge in real time over great distances. Using During the period 1985-98, license payments per capital it, researchers in Chile's wine industry can continuously learn from East Asia went up from 2.7 to 26.6 (not including from and coordinate their activities with the University of China). These economies clearly outstrip the LAC region, California, Davis. And ICT is the key element that acceler- where the average license payments only took up US$5 per ates the transition of knowledge. For example, computer resident by the end of the 1990s, doubling from slightly usage is linked to technology adoption potential, and the above $2 in 1990. India's royalty payments remained neg- number of Internet hosts to adaptation potential. Thus ICT ligible over the decade is an essential component of a country's technology and at It should be noted that trade, FDI, and licensing ought the core of the innovation system to be evaluated as a package since there is a fair amount of ICT is also increasingly becoming a dominant factor in substitutability, particularly between the latter two. The the acquisition and transfer (through distance learning) of same foreign company is unlikely to both set up a subsidiary skills. It can impact firm productivity in various ways. These in a host country and provide a license to another host com- include increasing competitiveness and market discipline pany. The choices by the firms often depend on the favor- through better and faster information on prices and markets, able treatment of foreign investment laws, distribution costs reducing transactions costs, and providing better monitor- and strength of patent protection, and overall long-term firm ing of the production process ICT also enables the creation strategy (Harvard University study 2002). Also, Blonigen of more complex analytical tools that support decision- (1999) has analyzed the relationship between trade and FDI making, provides firms with valuable knowledge about cus- and showed convincingly that imports of intermediate goods tomer demand, serves as a medium for technology transfer and sales of affiliates-are complements while imports and and communication of best practices, improves and increases sales of final goods are complements. Overall LAC countries, information and communication, and diffuses existing 148 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY knowledge Practically all countries that have been success- deliver the goods transacted on the Internet These compo- ful in technological transitions placed a strong effort in mod- nents should also be complemented by a labor force that has ernizing and developing, through competition and special the proper skills to use the technology effectively (see Guasch interventions, their telecommunications and ICT sectors. and Subramanian 2002) The most immediate impact on LAC and other develop- Latin America and the Caribbean lags behind East Asia ing countries has often been through alleviating information- and OECD in the development and usage of ICT, as was related constraints such as finding cheaper sources of required shown in chapter 3. Yet there is a large variance among LAC inputs, linking producers to nearby markets, uncovering and countries' performance as shown in figures 6.5 and 6.6, where opening new markets, acquiring information about new the countries have been classified as high and low perform- forms of production, and so on. The pure information impact ers on ICT on account of levels and improvements through of ICT seems to be larger and more pervasive in developing the 1990s Chile, Uruguay, Mexico, Colombia, and El Salvador countries mainly because their information deficiencies are have shown the highest improvements, while Panama, larger and more pervasive. Honduras, and Nicaragua have barely advanced (Guasch and An effective ICT framework consists of five well-developed Brehon 2002). components (i) connectivity: the physical wiring, bandwidth, The government has a key role to play in facilitating the spectrum allocation, and associated technology needs to be development of ICT, particularly in the legal and regulatory in place and extensive, (ni) content the available informa- arena. The ICT legal framework encompasses laws on elec- tion online has to be significant, access-friendly, and prefer- tronic data interchange, consumer privacy, protection of ably largely in the local language; (iii) transparency and intellectual property rights, and online security. Most coun- security. the legal and regulatory framework for transacting tries have been quite late in introducing a law on electronic online should be secure; (iv) a competitive telecommunica- data interchange. That law is based on a model proposed by tions sector, so as to induce extensive coverage and low prices: the United Nations Conference on International Trade Law the online prices should be as low as possible so as to facili- (UNCITRAL). It (i) gives digital signatures the same legal tate use by a large share of the population, and (v) physical status as handwritten ones; (ii) recognizes electronic docu- distribution network the country should have an effective ments and contracts as having the same legal value as con- physical distribution network with low logistics costs to tracts executed on paper, and (iii) defines the obligations of FIGURE 6 5 High-Achieving LAC Countries 0 8 0 6- 04- 0 2 - -0 - -0 6. -o 8- Year | Argentina Brazil : Chile -Colombia + Costa Rica| |El Salvador -Mexico --Peru Uruguay Venezuela| Source. Guasch and Brehon (2002) 149 CLOSING THE GAP FIGURE 6 6 Low-Achieving LAC Countries 08 - 06 - 04 - x 02- 1990 1991 1992 1993 1994 1995 __9,96 -19)k 1998 199 -2- 0 c0 2 -_v Z_ ._ (,- -> C~~~~~~~~~~~~~~~~ -0 4- -0 6 - i }+ -08 - Year Bolivia - Ecuador Guatemala --- Honduras ----Nicaragua ---- Panama ---- Paraguay Source Guasch and Brehon (2002) institutions providing electronic certification. Although lack below developed-country expectations. On the other hand, of interest and initiative from businesses in the e-commerce on the issue of software piracy Chile is significantly below field has likely been the main reason for the slow takeoff of the rest of the region in this area. online sales in LAC, the slow approval process of the elec- Security is a complex topic that encompasses several areas, tronic data interchange law may have also played a role. including the following: authenticity (evidence that a com- Although a law governing electronic contracts is neces- munication actually comes from the purported sender); sary, there is more that a country can and must do to facilitate integrity (proof that the content of a communication has not e-commerce. Electronic contracts can be effective only if they been altered during transmission); non-repudiation (pre- are complemented by protections for privacy (of consumers), vention of later denial of the transmission of the message); property rights (of suppliers), and system security. Chile has and confidentiality (evidence that the contents of a trans- been ahead of other Latin American countries in enhancing mission have not been disclosed to third parties) (Van Dam online privacy protection, through a data protection law 1999; Feldman and Meranus 2000). For example, ad hoc cer- enacted in 1999. The act includes rules governing the pro- tification technologies are typically used to ensure the con- cessing and use of personal data, the rights of individuals, fidentiality of transmissions to and from a Web site. This is the use of information by government agencies, and the use an issue in practically all LAC countries. Only 17 percent of financial, commercial, or banking data. A recent plan of Chilean e-commerce sites were "secure" in 2002, versus would also amend the law to include a bill containing a spe- 20 percent in Brazil and 25 percent in Argentina, accord- cial section relating to e-commerce. ing to a recent survey by the Santiago Chamber of Commerce. Most LAC countries are lagging in approving a law that Although Brazil was only slightly ahead of Chile in terms addresses protection of intellectual property rights on the of secure Web sites, it is much more advanced in terms of Internet. (This is proving a difficult issue for all countries- e-commerce in general. Another legal requirement that is see the White House 1997 and CEPAL 2001 ) At present, critical for promoting use of ICT among businesses (and gov- trademarks are covered by the same laws and international ernment) is approval of electronic invoicing, which most agreements on the Internet as they are in traditional media. LAC countries do not have. Electronic invoicing will give Offline protection of intellectual property tends to be better businesses the opportunity to keep their invoices in electronic enforced in Chile than elsewhere in the region but still falls format and send them electronically to the government 150 TECHNOLOGICAL I RANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY authorities. This would facilitate exchanges between gov- expenditures enable domestic firms to generate new tech- ernment and business, reduce businesses' and government's nologies and innovation for local needs and to better absorb inventory costs (SI1 estimates that thanks to electronic invoic- and adapt existing advanced technologies from abroad. It is ing, a government's costs would decrease from $1.2 to $0.4 the key instrument for creation and adapting. However, the per invoice, generating total savings of about $300 million strong public good characteristic of knowledge and the per year-SII 2002), and support online invoicing. Finally, spillovers it generates demand particular government inter- LAC also lags behind in credit card penetration rate, the vention First, there is the issue of the degree of spillovers, and essential form of payment for electronic transactions, and its hence appropriability, varying by the type of research, and that mat- use of Internet-based payment systems is still behind the ters because the degree of intervention has to match the best-practice income comparator country. Supporting devel- extent of those spillovers. Basic research consists of experi- opment of medios de micropagos (prepaid cards, debit cards, mental or theoretical work conducted to obtain new knowl- payment systems with direct charges to fixed or mobile edge about the origin of fundamental phenomena and phones) would increase e-commerce penetration, especially observable facts, without a specific application in mind. by giving lower-income populations the opportunity to buy Given the fundamental nature of this research, spillovers are online, provided they have access to computers. high and long-lasting. Overall, it is important to underline that while govern- Appropriability of benefits in a free market, correspond- ments should take certain regulatory steps to improve the ingly, is lower. Applied research, which consists of original policy environment for improving competition in the sector work undertaken in order to acquire new knowledge with and doing business online, the focus must ultimately be on a specific goal in mind, is narrower in scope and hence has implementation. Governments should improve the pro- less spillover than basic research. Finally experimental devel- tection of property rights on the Internet, ensure that the opment, which consists of systematic work based on exist- privacy law is amended to include a special section on e-com- ing knowledge and is directed toward the production of new merce, support the development of a secure infrastructure materials or products, or toward improving products that when needed (promote an electronic invoice law, and inter- already exist, has the least spillover. While R&D is mostly vene to increase access in selective areas such as in educa- associated with the creation of knowledge, it also has an tion and rural areas). Beyond that, however, the initiative important role to play in facilitating the adaptation of exist- must be taken by the private sector-the government's role ing technologies, as was shown in box 6.1 In addition to con- should be to encourage and enable the private sector to take tinued education expansion-now with particular emphasis this initiative. on the completion of secondary education and increasing ter- tiary enrollments-and openness to FDI and licensing, to Advanced Technological Transition: R&D reach the advance stage of technological capabilities requires as the Key Driver efforts to cultivate creativity through strengthening patent To reach advanced stages of technological capabilities the key protection regimes, increase public funding (through tax instrument is significant R&D expenditures, particularly pri- Incentives and subsidies) to secure significant private R&D vate No country has ever secured advanced technological investments; continue, perhaps even expand, public provi- capabilities without significant private R&D expenditures sion R&D aimed at basic research; facilitate and target R&D Thus the key policy issue is how to induce significant in special areas relevant to the country and showing signs amounts of private R&D. To address this issue we first pre- or potential of excellence, and continue the advancement and sent the instruments governments have and then proceed to deepening of ICT. But the key focus at this stage should be impute the determinants of private R&D from theory and securing private investment in R&D. empirical analysis. A casual glance at trajectories of productive sector R&D, Formal technological effort is captured by R&D expen- shown in figure 6.1 and confirmed in table 6.7, suggests that ditures on plant, equipment, and personnel. R&D is both numerous countries, such as Korea, Israel, and Finland, have capital- and skill-intensive and hence not typically conducted had "takeoffs" in productive sector R&D at roughly the same in any significant amount in economies at the very early level of income as Mexico, Argentina, and Chile, which have stages of technological catch-up. Research and development made far fewer efforts (figure 6.7). Thus the central issue of 151 CLOSING THE GAP TABLE 6 7 Evolution of Total R&D Expenditures as a Percentage of GDP, 1960-99 1960-64 1965-69 1970-74 1975-79 1980-84 1985-89 1990-94 1995-99 Australia 1 117 0 862 1 027 1 181 1 450 1 578 Spain 0 179 0 193 0 259 0 330 0 454 0 624 0 841 0 821 Finland 0 713 0 849 0 978 1 321 1 737 2 136 3102 Ireland 0 420 0 591 0 706 0 683 0 699 0 824 1 062 1 342 Korea 0 240 0 366 0 357 0 529 0 833 1 692 2 098 2565 New Zealand 0.779 0 814 0 883 1 007 1 051 Singapore 0.060 0 214 0 408 0 871 1 069 1553 Sweden 1 180 1 270 1 463 1 735 2 361 2 835 2 965 3 663 Taiwan (China) 0 829 0 879 1.168 1 739 1 896 Group average 0.764 1.329 1.615 1.957 Argentina 0 595 0 560 0 704 0 936 0 392 0 382 0 315 0 413 Brazil 0 344 0 608 0 481 0 380 0 828 0 844 Chile 0319 0430 0428 0574 0564 Colombia 0014 0081 0110 0270 Costa Rica 0 300 0 222 0 261 0 261 0 160 0 273 Mexico 0 168 0 194 0 543 0 285 0 255 0 354 Peru 0 177 0 270 0 261 0 102 0 025 0 058 Venezuela 0 094 0 257 0 478 0 318 0 308 0 422 0 358 LAC Average 0.350 0.392 Soyrce Guasch (2002) with data from Lederman (2002) FIGURE 6 7 Evolution of R&D/GDP Expenditures on GDP Per Capita 1980-1995 2 76524 - 2164 - land~~~~~~~~~~~~~~~~~,n Kore Ca ada Aq-- Australia Eoudor-.. _ a 018 0 Haaco 0- Paaay Veia 882 18391 6 1294 31 7400 91 GDP Per Capita GDP Per Capita Source Bosch and Maloney (2002) thss and the next chapter is how LAC can increase Its par- public execution of R&D. Having in place intellectual prop- ticipation in this area. Below, we discuss several policies erty rights increases the appropriability of benefits to inno- employed to accomplish this. vation. Fiscal incentives, such as tax credits, work by reducing the cost of R&D to the performing firm, thus better Government Policies for R&D aligning them with the private benefits of R&D. Public The most widely used policies for correcting for market fail- funding works by either entirely or partially (as in the case ures are (i) intellectual property rights, (ii) fiscal incentives of matching grants) providing public funds to compensate and public funding of research and development, and (iii) for the shortfall in private funding for R&D. 152 TEC14NOLOGICAL TRANSITIONS AND ELEMENTS OF TECHINOLOGY POLICY Intellectual property rights BOX 6'i Intellectual property rights work by giving the innovating The Role of Intellectual Propery Rights (IPRs) firm the right to temporarily exclude others from using the in Technology Transfer new idea commercially so that it can appropriate some of the benefits of its R&D effort-particularly those related to Intellectual property rights have traditionally been patent protection, which are also a key determinant in tech- regarded as a means to encourage research and develop- nology transfer, trade, and FDI (box 6.4) Thus, copyrights ment. There is, however, another role that IPRs may protect the rights of authors (books, music, software); trade- play: that of encouraging technology transfer. Since pro- mark registration protects trade logos and symbols, and prietary knowledge is valuable to foreign firms, they are patents protect inventions with industrial applicability likely to try to minimize its leakage to domestic firms. (products as well as processes). For technology development, Although relatively new to the research, some studies patents are most relevant; hence, this section focuses on the show that weak intellectual property protection may unique issues surrounding this instrument. deter investors in high R&D sectors Similarly, some Patents are intended to spur innovation in a number of studies argue that IPRs in the host country may affect ways. First, they award exclusivity of use, sale, and manu- the amnount of licensing. It is possible that the strength- facture, for a limited period of time, to the owners of the ening of IPRs in LAC may have had a role in the increase intellectual property, thus compensating them for under- in FDi and licensing in the region. Smith (2001) finds taking expensive and risky innovative activities. In exchange that strong IPRs are positively correlated with U.S. affil- for this benefit the owner must disclose the invention on the iate sales and licenses, especially when host countries patent document for anyone "skilled in the art" to be able posses strong imitative capacity. Since evidence of large to replicate. Thus, patents are a tradeoff. A market distor- sales implies a relatively large-scale operation, the scope tion in the form of a temporary monopoly is created in for spillover is higher than with smaller MNE presence. exchange for disclosure of the information relating to the Similarly, with strong IPRs, licensing cost is much lower, technology. Disclosure is intended to benefit society by dis- facilitating more licensing from the United States to host seminating new technologies and indeed encouraging com- countries. See also Glass and Saggi (2000) for a theo- petitors to invent around the new technology in a second retical treatment of the IPR and FDI relationship with round of innovation. a product cycle model. For relationship between IPR and The link between IPRs, innovation, and productivity has licensing, see Yang and Maskus (2001). Smarqynska stirred some controversy. Intellectual property rights have two (2002) finds that the composition of FDI depends cru- opposing effects on R&D. On the one hand, they increase the cially on the strength of IPR in transition economies. degree of appropriability of the returns to R&D, thus pro- She finds that weak intellectual property protection viding incentives for R&D expenditures; on the other hand, deters FDI in high technology sectors. Furthermore, the that protection prevents the use and diffusion by others and type of FDI is also affected. She finds that weak IPR leads increases the cost of R&D for rivals and/or follow-on inven- to more FDI in distribution rather than local produc- tors, which can impact R&D and productivity adversely. tion or goods for fear of imitation. But presumably, In setting up an IPR framework, which is essential to fos- spillovers from FDI would be quite limited in the case tering private R&D, the issue is how strong the framework of distribution compared to local production In a study should be and how it should evolve. Strong IPRs come at a of 58 countries, Park (2002) shows that patent rights cost to society and the argument can be framed in terms of do indeed positively impact the amount of FDI coming a bargain between static and dynamic concerns. Advocating into the region. stronger intellectual property rights therefore must presume that the combined positive impact of the appropriability incentive for the innovator and the disclosure element for diffusion for producers). This is nearly impossible to test peers (resulting in greater innovation) outweighs the nega- empirically and remains a highly debated subject to date. tive Impact of the temporary market distortion (resulting The above discussion points toward the importance of the in higher prices for consumers and slower technology timing of strong IPR protection. Protecting IPRs very early 153 CLOSING THE GAP TABLE 6 8 Transitions of Strength of Patent Protection-IPR-for Selected Countries, 1S90-2000 ECONOMY 1960 1965 1970 1975 1980 1985 1990 2000 Argentina - - - - 2 26 2 26 2 26 3 19 Brazil - - - - 1 85 1 85 1 85 3 05 Chile - - - - 2 41 2 41 2 41 3 07 Mexico - - - - 1 40 1 40 1 63 2 86 Sweden 2 04 2 04 2 04 2 04 4 33 4 33 4 33 4 75 India 1 85 1 85 1 42 1 42 1 62 1 62 1 48 - Finland 1 90 2 04 2 04 2 04 3 95 3 95 3 95 4 35 Japan 2 85 3 18 3 32 3 61 3 94 3 94 3 94 3 94 Korea 28 28 294 2.94 328 361 394 420 Malaysia 2 37 2 37 2 37 2 37 2 57 2 9 2 37 2 85 Singapore 2 37 2 37 2.37 2 37 2 57 2 57 2 57 3 90 United States 3 86 3 86 3 86 3 86 4 19 4 52 4 52 4 86 - Data not available Note The patent protection index ranges from 0 co 5, with 5 being the strongest protection of patent Source Guasch (2002) with data from Park (2002) in the technological catch-up process can choke off an impor- reduce sales in a foreign market because of greater market tant channel for technological learning. Failing to protect power in an imitation-safe environment Whether market IPRs when private firms are launching R&D programs is expansion or market power effects will be stronger is not likely to weaken incentives to conduct R&D. Table 6.8 shows known and empirical estimates have gone both ways. Survey the evolution of IPR strength for selected countries. Note- data have shown that FDI is not especially sensitive to the worthy is the case for Korea and Singapore, where we see IPR regime. This is because the importance of IPRs varies the gradual strengthening of IPR, fitting country techno- greatly by industry and because relatively little FDI is in logical conditions as they evolve. R&D-intensive activities (particularly for nations at the ear- Also, IPRs can improve the effectiveness of public R&D. lier stages of technological catch-up). The data show that Much of the criticism of government R&D has been that it most R&D is done at the parent firm, with production or is disconnected from the needs of the private sector, and that sales facilities being set up in major markets. the private sector has difficulty in using innovation produced The case might be clearer for copyrights, trademarks, and by public R&D. That has been the case for LAC countries. geographical indications where no real market distortion is In the United States the Bayh-Dole Act allowed universi- created because the underlying idea behind copyrighted ties and government institutes to patent their inventions, materials can still be adopted. For trademarks and geographic thus increasing their collaborative efforts with the private indicators as well, an infinite supply potentially exists; hence sector-although some would argue that this comes at the the distortion is minimal In fact, for countries at the early cost of compromising basic research. And if IPRs are clearly stages of adaptation, firms may be in dire need of mecha- defined and portable (that is, if a worker can take all or part nisms to signal quality or otherwise differentiate their prod- of the rights to the invention with him or her when he or ucts to build consumer loyalty. For instance, Chinese she switches employers), firms are likely to fiercely compete producers have reported difficulties in promoting their own for skilled labor and workers are likely to move in favor of brands of soft drinks, processed foods, and clothing. Having more competitive terms The mobility of labor is a key driver established some brand recognition through costly invest- of innovation at the advanced stages and is discussed in more ments, enterprises find their trademarks applied to coun- detail in the chapter on networks. terfeit products quickly, which damages the reputation of Potentially, learning at the early stages of technological high quality producers and sometimes forces them to aban- catch-up through trade and FDI could also be influenced by don their trademarks or even close down altogether (OECD IPRs The effect of strong IPRs on trade is unclear-while 2001). In consequence, trademark and copyright protection it could increase imports as foreign intellectual property appears more relevant at the adoption stage The evidence owners face increasing net demand for their products owing on their impact on TFP shows that it is little in contrast to to displacement of "pirates," a title holder may choose to the impact of patents (Park 2002). 154 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY A question raised by critics of IPRs is this Given the * Two legal attributes of thepatentprocesses could be tweaked. small domestic R&D sector, will not the strengthening of First, public disclosure of the patent application early IPRs shift economic rent to foreign multinationals, which on disseminates the invention (unlike in the United own most of the technologies? Also, since the LAC region States, where it is disclosed very late in the process) is predominantly an imitator of Northern technologies, will to competitors so that they can begin the "invent not technology diffusion, and hence overall welfare in the around." Second, narrowing the scope of the patent region, suffer as a consequence of improved property rights? (one claim, one patent) also reduces the extent of the Though valid, these questions ignore the feedback between monopoly awarded to a single owner. Japan is an R&D activities and policies: that is, R&D sectors may be excellent example of a country that followed such an small in part precisely because the incentives to do R&D are IPR regime during its technology-importing days, weak, and the "switch from imitation to innovation" may and revised it recently to suit its role of technology be hampered by the lack of institutions supporting it. creator. Another practical issue is that if individuals or firms in a * Two leeways in TRIPS could be exploited. First, countries developing nation have the capacity to infringe, they typi- are allowed to determine their own criteria for com- cally have the capacity to innovate. Thus, nations such as pulsory licensing of patented technologies so long as Brazil or India that succeed in producing and distributing the "national treatment" clause (in other words, treat- patent-infringing goods usually reveal a capacity to inno- ing foreigners and nationals equally) is not violated. vate but also reveal an environment where the relative eco- Second, the decision to allow parallel imports is also nomic rewards of infringement are greater than those from within national boundaries because individual coun- innovation Thus it is most important to have the right tries can determine when the "exhaustion" of patent incentive structure, and part of that is to strategically manage rights has occurred. IPR (see box 6 5) * Two extra-TRIPS instruments could be instituted. First, Theoretically, there is also another reason to worry about these countries are strongly encouraged to institute IPRs at the earlier stages of the transition. Given that these utility models Utility models are in principle like countries are essentially technology importers from the patents, but they differ in several important ways: they North, the development of appropriate technologies in the have much lower standards of novelty in order to pro- North for use in the South can only be facilitated when tech- tect minor innovations; they are not subject to the nology creators see the South as markets where their invest- rigor of patent examinations (where absolute novelty ments can be recouped. While isolated and individual is stressed); they are not nearly as expensive to file for; markets may not constitute a sufficient incentive, collec- ancl they are valid for three to five years only. Thus, tively their market influence on changing the focus of tech- utility models are particularly suitable for boosting the nology efforts in the North could be sizable. However, since domestic adaptation effort. Second, improving the gen- the payoff to any single country from "free riding" is poten- eral business environment to reduce costs through tially quite high, a uniform upgrading of patent regimes is competition policy, improving logistics, and supply mandated and perhaps most effective when enforced chain management become imperative for countries through an international institution such as the World Trade that are faced with enduring higher prices for patented Organization (WTO). The Trade Related Intellectual Prop- goods in the short run erty Rights (TRIPS) agreement, which has been signed and is in different stages of implementation in different coun- Tax incentives and subsidies tries, claims to deliver this benefit to developing countries As mentioned, the public good nature of knowledge and in the long run (see box 6.6 for details). associated positive spillovers make knowledge's social ben- In the short run, however, there are costs for developing efits far larger than its private benefits, leading to under- nations in the form of higher prices and slower access to investment and argues for government intervention.3 patented technologies borne by domestic producers, and Further and more complicated are the imperfections of higher-priced goods borne by consumers. What might gov- capital markets Credit constraints and mismatch of terms ernments do to mitigate these costs in the short term? of lending reduce the demand for R&D by firms. Credit 155 CLOSING THE GAP BOXX65 eRSianaging intellectual Property Rights (IPRs) Intellectual property is increasingly recognized as a global property often vary within a country according to asset that needs to be strategically managed. Many OECD researcher status, the type of performing institution, countries have recently broadened and strengthened the or the source of the funding. Governments have tried laws and regulations covering intellectual property pro- to make patent and licensing procedures more umn- tection in order to increase incentives for innovation and forn in order to simplify the transfer of research improve returns on these investments. The changes in results to thc private sector by standardizing own- intellectual property regimes have influenced the behav- ershlip rules for publicly funded and publicly per- ior of firms and research organizations. Throughout the formed research. 1990s, the number of patent applications and patents o Increase incentives to commtercialize. Exploiting intel- issued in Europe, the United States, and Japan grew sub- lectual property is a time-consuming and costly stantially. These increases may reflect a real rise in inno- business. Public research organizations (PROs) vative activity, but they are also linked to the explosive cannot afford investments in the needed infrastruc- growth of patent-intensive fields, such as biotechnology ture and skills if they cannot be assured of sufficient and ICT, and increased recognition of the strategic role remuneration (for example, through licensing or of IPRs as a currency that allows firms to compete and other fees). Some countries provide subsidies for local cooperate. or institution-based technology transfer and licens- Industries vary tremendously in the type of intellectual ing offices. Other countries let PROs elect title to property protection that is relevant to their activities. their innovations in exchange for the promise to seek Patent protection is considered essential for pharmaceu- protection and eventually exploit the invention, thus ticals, medical technologies, and biotechnology. Software letting the PRO profit from its research results. developers use a mix of patents and copyrights to protect o Decrease costs ofprotection and exploitation. For univer- code, but open source software is gaining in popularity. sities, public laboratories and small and medium- Such new technologies continuously challenge govern- sized enterprises (SMEs), the cost of applying for, ments to search for the correct balance between the com- maintaining, and defending patents represents a mercial need for incentives to innovation and the public burden that governments are seeking to reduce. need for disclosure and access to innovations. Governments Application and maintenance costs can be reduced I should attempt to maintain stability in intellectual prop- or waived for these institutions, and medcation rather erty regulations, but must determine whether a new bal- than litigation is encouraged. ance needs to be struck between public and private o Limit restrictions on publication and scientifc enquiry. imperatives for research and innovation as new rechnolo- Licensing agreements and research contracts can con- gies emerge and eventually mature. tain clauses that hamper the ability of scientists to To improve the diffusion and commercial impact of continue their exploration in a field or slow the dif- publicly funded research results, many OECD countries fusion or access to fundamental research results by have experimented with novel approaches to the regula- other researchers. To minimize clashes between the tion and institutional management of intellectual prop- research and teaching mission of PROs and their erty resulting from publicly funded research. Policy commercial activities, governments can play a role measures aim to accomplish the following: in limiting the contractual demands placed on PROs. o Increase predictability and reduce transaction costs. Reg- ulations governing the ownership of intellectual Source: OECD (2001) 156 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY or R&D tax credit is based on a percentage of R&D BOX 6 6 TRIPS: Developing Countries' Concerns Regarding the expenditures Under a voltume-based system, the credit Universalization of IPRs applies all of the R&D expenditures incurred during the period. Under an increment-based system, the credit As part of the trade deal hammered out in 1994, coun- applies to incremental R&D over a particular period. tries joining the WTO also signed on to TRIPS, an inter- In the event that the total credit exceeds taxable national agreement that sets out minimum standards for income, there may in some jurisdictions be provisions the legal protection of intellectual property. These for carrying forward (or carrying backwards) the excess include extending intellectual property rights to include (unused) credit. The kinds of firms that would bene- computer programs, integrated circuits, plant varieties, fit from carry-forward provisions would be start-ups or and pharmaceuticals, which were unprotected in most firms that suffer temporary losses. developing countries until the agreement came along. o Matcbing grants. Another type of government inter- According to the terms of the agreement, most of the vention to stimulate R&D is matching grants, a cost- poor world had until the beginning of 2000 to bring sharing instrument that potentially avoids the moral their legal protections up to scratch. Developing coun- hazard problem that plagues outright subsidies. Recip- tries were given an extension until the end of 2005. lerits put up a certain percentage of research costs, thus TRIPS was pushed onto the trade agenda by the signaling a precommitment to do a research project. United States, Europe, and Japan, which together hold Few countries have provided such programs and the the lion's share of the world's patents and whose com- understanding of these programs is still incomplete panies wanted more protection abroad. Meanwhile the poor countries still have misgivings about the agreement Various design issues need to be considered when coming as they await the promised benefits of patent protection. up with incentives. In the context of LAC, revenue consider- Much of the poor world's anxiety about TRIPS is focused ations and sustainability issues are important, given LAC's on two issues access to medicines and protection of tra- experiences with budgetary debt and deficits The overall tax ditional resources. Many developing countries, includ- system may also significantly impact the effectiveness of fiscal ing India and some sub-Saharan states, would like incentives. When the prevailing tax system is distorted, the clarification on the agreement's provisions and exceptions marginal effects of new fiscal measures are likely to be weak to protect public health and the environment, and or offset. Furthermore, in LAC, where the R&D sector is small, amendment of its articles on the patenting of life-forms. some of the new R&D investments may come from the ranks of start-up firms, whose initial startup costs are high but whose From "The Right to Good Ideas," The Economist, June 23, 2001 revenues are low They are thus not likely to have sufficient taxable income early on to take advantage of R&D tax incen- constraints for R&D activity are exacerbated by the fact that tives, such as an R&D tax credit Allowing for "carry-forwards" R&D is risky and requires long-term financing. The most will help them use the credits later on but, owing to the time common type of intervention to counter these negatives is value of rnoney and inflation, these credits will be worth less. through fiscal incentives and subsidies. Governments pro- Defining R&D will be important in light of the practice of vide a number of these to stimulate firms' R&D investments: relabeling expenditures as R&D, as has been witnessed in OECD countries. A major design issue is whether a volume- o Depreciation rates. By providing high and accelerated based system or an incremental one is preferred. One needs depreciation rates for current and capital expenditures to distinguish between inframarginal (what would have been on R&D, governments can favor R&D by allowing done anyway) and supramarginal (what would not have been firms to avoid tax burdens on "assets" whose revenue- done without the tax incentive) R&D. However, with an incre- generating capability is active. mental tax credit system, the marginal effective credit rate may o Tax credits. A second type of fiscal incentive is the R&D be "negative," thus discouraging R&D investment. tax credit, which enables firms to obtain deductions While matching grants can be a useful way to link public from income taxes The precise amount of deduction and private R&D, governments need to be careful not to 157 CLOSING THE GAP "crowd out" private funding. In order for public subsidies levels of private R&D can be seen as a measure of the effec- or grants to stimulate private R&D effort, the subsidies or tiveness of public R&D. grants must prompt firms to undertake projects they would While the capability of firms or governments to engage not have otherwise found profitable (but which are socially in R&D at the early stages of technological catch-up is bound valuable). To the extent that the government funds infra- to be constrained by the limited learning that has been accu- marginal R&D (which would have been undertaken anyway), mulated as well as the lack of highly skilled workers, some there is little additional stimulus to overall R&D. However, pockets of highly successful R&D activity are likely to occur politically, there can be great temptation for bureaucrats to in some enclaves, and the record from East Asia shows a high support commercially attractive projects in order to demon- level of direct public involvement, as well as the record of strate "success." These commercially attractive projects are Brazil with EMBRAER and Chile with CODELCO (de also the ones that are more likely to have been undertaken Ferranti and others 2001). It is generally accepted now that in privately Thus funds spent may actually be a rent to private the early stages of technological catch-up governments or firms. Best practices here are based on "competitive" grants universities should pursue selective research activity on their and a transparent process that involves the participation of own and with selected firms-or sectors-in shop-floor tink- experts and the private sector in the allocation. Finally, the ering, facilitated by contractual relationships with public organization structure to allocate and oversee competitive research institutions. As the technological gap narrows, the allocation of funds-subsidies-is far different, and proba- importance of indigenous research grows and firms are best bly less costly, than the one to provide for tax incentives. placed to conduct R&D in an effective manner-public R&D Table 6.9 summarizes some pros and cons of tax credits is largely focused on basic research, and growing private versus matching grants. R&D takes the lead, becoming the key determinant of suc- cessful technological transitions, as we have seen it in NIC Public Provision of R&D and natural resource-abundant countries. Hence, rather than Governments also provide direct-and often full-funding the R&D to GDP ratio, the percentage of R&D financed by to a number of R&D activities, in principle as a measure to the private sector becomes central in explaining the transi- correct for market failures. In particular the public sector is tion into the creation stage as shown in table 6 10 supposed to finance "basic" R&D, which has more proper- ties of a public good than does applied R&D, and which is How Do These Factors Affect R&D Expenditures? unlikely to be financed by the private sector since it does Based on a review of both the theory and the evidence, it not usually have direct commercialization use yet, recog- seems accessing increased technological capabilities and suc- nizing that basic R&D generates significant positive exter- cessful transitions will only come through increased private nalities and spillovers. Moreover, at the earlier stages of R&D. While spending on research and development does technological development, as a way to Jump-start the process, not, in itself, ensure innovation or rapid growth rates, such governments tend to finance specific programs and sectors, spending is a necessary condition at all stages of develop- for potential economic impact and strategic reasons. ment. The magnitude and the focus depend on the stage of R&D is conducted by both private firms and public insti- a country's development. Without an adequate volume of tutions, and the distribution between the two has received R&D expenditures a country will most likely not ascend the great attention. The public sector role in the provision and technology ladder. No country has managed successful techno- execution of R&D evolves as countries advance in acquiring logical transitions without significant private R&D expenditures. technological capabilities and in increasing the skills of the Thus being able to guide any policy for LAC countries to labor force. The share of public R&D in total R&D expen- stimulate private R&D is essential for understanding the ditures declines as countries technologically advance, and determinants of private R&D investments. Those invest- that is another indicator of successful transitions While all ments are done by firms, and their level depends on their countries start with a high share of public R&D, the suc- expected returns, which in turn depend on (i) the costs of cessful ones have managed to elicit increasingly higher levels R&D, which are affected by fiscal policies and subsidies to of private R&D, so as countries evolve, the share of public R&D; (ii) the probability of success, which is affected by the R&D keeps decreasing (see table 6.10 ). Thus, eliciting high skill of the labor force; (iii) the degree of appropriability of 158 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY TABLE 6 9 Tax Credits versus Matching Grants TAX CREDITS MATCHING GRANTS Advantages * Less government interference * Coordinates public and private R&D spending * No need for governments to choose teclnologies * In principle, could target areas with large gap between * Tend to be open-ended social and private rate of return Disadvantages * Start-up firms may be too small to qualify * Government may need to choose technologies * Firms may relabel expenditures as R&D * Pressures to show visible success may lead to financing * Tax structure may be complex of projects with high private rates of return, thus * Projects thar generate greater short-run profits are "crowding out" private investment likely to be favored * Need to account for governance issues-transparency, * High intramarginal costs, small marginal effect incentive, and competitive granting ("crowding out" may be larger than in grants) Preconditions * Well-functioning tax system None TABLE 6 10 (although their results are mixed), in favor of findsngs of com- Evolution of the Ratio of Prnvate to Public R&D Expenditures, plementarsty between public and private R&D investments 1980-95 The evidence from five programs in the United States and 1980 1985 1990 1995 Israel reviewed in Park (2002) provides mixed results-some Argentina 0 1765 0 1765 0 1765 0 1806 studies showing the positive impact of these programs on Brazil 0 4749 0 4749 0 4749 0 4749 stimulating private R&D, with others highlighting some Chile 0 4493 0 4493 0 4205 0 4598 Cost Rica 0 0893 0 0891 0 0894 0 0892 shortcomigs in actual practice. Mexico 0 3850 0 1765 0 1111 1 3042 The experience in Korea, Chile, and Colombia, however, Finland 2 1738 2 1854 2 3337 2 4819 seems to show that competitive allocation of funds-subsidies Ireland 0 8019 1 0618 1 2675 2 5971 such as matching grants-has a higher impact to stimulate Israel 0 5674 0 5674 1 6596 1 6954 Korea 1 6225 4 0136 4 4336 4 7961 private R&D than tax incentives In a background paper for Source Guasch (2002) with data from Park (2002) this report, Park (2002), using recent data, shows that fiscal incentives have a positive effect on private R&D even when successful innovation, which is affected by IPRs; (iv) the patent rights and public R&D funding are controlled for extent of and access to supportive knowledge available, which However, while public R&D is a strong determinant of pri- is affected by public R&D and the quality and extent of vate R&D for non-LAC countries, it does not have an impact research institutions, and (v) the degree of competition faced for LAC countries That result hints at the inefficiencies in by the firms (Tirole 1996; Guasch 1985) LAC of public R&D and is corroborated by Lederman and There has been a fair amount of research testing those Maloney (2002), who show that indexes for quality of public theory implications Earlier research (conducted in the research institutions and extent of private-public collabora- 1980s) tends to find that fiscal incentives have a weak effect tion are very low in LAC They show that the effectiveness on R&D Later research, conducted during the 1990s, tends of public R&D as a determinant of private R&D-and to find more favorable evidence of the impact of R&D tax patents--is strongly affected by those two indexes, a fact that credits Hall (2000) attributes the difference in results to the explains the ineffectiveness of public R&D in LAC. The study fact that the earlier elasticity estimates were based on a time also shows that for both LAC and non-LAC countries, pri- period when firms were uncertain as to whether the R&D vate R&D is a function of an economy's level of intellectual tax credits were permanent or temporary. Given the costs of property rights, and patent protection is the most important ad)usting R&D capital, firms appeared to choose to wait. area of rights protection. Copyrights and trademark rights A problem with these studies is that they compare the positively explain R&D, but not when patent rights are con- tax credits to the amount of R&D generated rather than to trolled for, which indicates that these variables likely picked the amount of benefits from the R&D generated. David, up the effect of the omitted variable (patent rights). Thus, Hall, and Toole (1999) review evidence from micro- and despite the fact that IPR violations have occurred in LAC, macro-level studies, mostly for the United States, to conclude patent protection has been important to private research in 159 CLOSING THE GAP Latin America. But even more important in LAC, is the emer- TABLE 6 11 gence as a determinant of R&D the education or skill level Determinants of Private R&zD Per Worker in LAC and Other of the labor force. This is emphasized by the high elasticity Countnes: Easticities as seen in table 6 11, where a summary of the results is pro- ALt COUNTRIES LAC vided, as well as in box 6.7. Elasticity private R&D/patent rights 1 77 1 Elasticity private R&D/public R&D 0 58 n s The positive correlation between IPRs and private R&D Weak positive effect of copyrights protection Weak + Weak + can be seen in figure 6.8. Weak negative effect of trademark rights n s Weak - In summary, a coherent R&D framework is essential to Elasticity private R&D/fiscal incentive -2 60 -I Elasticity pcivate R&D/education 0 33 1 32 successfully reaching advanced technological capabilities. Elasticity private R&D/competiton 11 0 9 Overall, we find that patent protection, public funding, and n s not significant Note All figurei are significanE at the i percent level human capital are important to private R&D, and ultimately, Source Park (2002) and Guaich (2002) BOX 67 Human Capital, IPRs, and R&D Policies as Determinanuts of Private RQD and Uhekr Ompect on 1P-iethodology and Resuts Based on an R&D supply and demand model, Park (2002) tection stimulate TFP indirectly by stimulating the deter- derives the steady-state stock of private R&D capital per minants of TFP-for instance, domestic patent capital worker (Rp/L) as a function of public R&D capital (R /L), through private R&D spending. physical capital per worker (KIL), human capital per Regressions performed for LAC countries alone are worker measured by the scientists and engineers per much less robust owing to lack of adequate data. How- 10,000 workers (SE) and tertiary enrollment rates (E), and ever, one result that the data do seem to confirm is that the policy environment (for example, level of intellectual public R&D spillovers to private R&D are negligible in property protection and fiscal incentives) (IPR). LAC, which is confirmed by case studies. R/L = f(R /L K/L SE E IPR) The following table provides results of some simula- R = fR IL, KL, SE, E, IPR) tion analyses for LAC countries based on the regression The technology stock is measured by the stock of domestic results. We use the estimated equations to predict the patent capital per 1,000 domestic workers (A) and is a finc- effects of changes in the independent variables on R&D tion of public and private R&D capital, and human capital. and TFP. These results should be interpreted as a steady- Ad = fiR /L, Rg/L, SE, E, IPR) state comparison (since the model was solved for in steady P g state). Since we are not able to estimate a complete set of Finally, total factor productivity (TFP) is a function of the R&D and TFP equations with all the R&D policy vari- domestic and foreign (A) stocks of patent capital and ables included, primarily because the number of observa- human capital. tions for the fiscal incentive variable was small, we mix coefficients from different regressions. TFP = flAd, A~ SE, E) PRETCAG ECN The stocks of public and private R&D capital, and domes- IN PIUVATE R&D CHANGE tic and foreign patent capital, are determined by the per- SPENDING PER WORKER IN TFP petual inventory method (assuming a 10 percent rate of i0 percenr increase in patent depreciation). right strength (Index)* 4 00 0.08 10 percent increase in public Empirical analysis for the pooled sample of 20 devel- R&D funding per worker 6 60 0.22 oped, 19 LAC, 7 East Asian, and 12 other countries shows 10 percent increase in that patent rights and the human capital variables are a scientilts and engineers per 10,000 workers 2.03 0.95 significant determinant of private R&D. Fiscal incentives 10 percent increase in tax (as measured by the B-index) are also weak determinants incentives (decrease in B-index)* 11 6 0 20 of private R&D Hurnan capital also directly increases TFP *These estimates hold for Brazil and Mexico only, since those were the only two whereas fiscal incentives, public funding, and patent pro- LAC countres with B-index data 160 TECHNOLOGICAL TRANSI'I IONS AND ELEMENTS OF I ECHNOLOGY POLICY FIGURE 6 8 Patent Rights and Privately Financed R&D (1980-95 average for 58 countries) C: 80- , 60- is ;,40- e 20 0 -~0 0 1 2 3 4 5 Patent Rights Index I Cl LAC * Others Note The Patent Rights Index varies between 0 and 5 It is obtained by aggregating five subindices - extenc of coverage, membership in international treaties, duratiorl of protection, absence of restrictions, and enforcement See Park (2002) for details Source Park (2002) as we show below, to total factor productivity We find that countries are shown in table 6.12. Among the few LAC coun- the key determinants with the stronger impact on stimulat- tries that had IPRs in 1995, there was stronger copyright ing R&D, as measured by elasticities, are education-tertiary, and trademark protection than patent protection This is an in particular science and engineering disciplines-financial important issue, since the evidence shows stronger TFP assistance, and patent protection strength The causal mecha- effects from patent protection nisms are as follows patent rights, public funding, and human capital stimulate private R&D. Private R&D and public R&D High Cost of R&D Financing in turn stimulate domestic and foreign patenting, which in turn Not only are venture capital markets underdeveloped in LAC, contribute to the stocks of domestic and foreign-generated so are regular capital markets. Thus, unless a firm is big enough knowledge capital, which in turn are important determinants to finance its research and development expenditures through of total factor productivity-as we see below internal funds, it is likely to find financing very costly R&D is a long-term endeavor and thus requires long-term financing. The Structure of R&D in LAC The current paradigm in the LAC financial sector of very high Given the identified determinants of private R&D, the interest rates and practically no long-term financing is clearly limited development of technological capabilities in LAC a factor in the limited private investment in R&D. Also, macro- countries is not surprising, since most of them lack or are economic instability hampers incentives for long term invest- deficient in the provision and levels of those determinants, ments. Moreover, R&D investments are lumpy, a reasonably as we show below. large amount of R&D is needed just to start, and the relatively small size of firms in LAC makes securing or committing that Weak Intellectual Property Rights critical mass not feasible Venture capital is usually, then, the LAC countries typically have had weak IPR regimes, alternativc to secure funding for existing or new, innovative although there has been some strengthening in recent years, firms. However, venture capital markets are extremely under- prodded by the developed countries. Nevertheless, IPRs are developed in LAC, in contrast to OECD and NIC countries weaker in LAC than in most OECD and some East Asian In any event, as Hall (2002) argues, venture capital may be a countries, with Chile, Argentina, Panama, and Brazil in the limited solution to financing innovation. Given the informa- lead; Mexico might be the outlier, with relatively weak IPRs. tion and administrative requirements placed on venture capi- Indexes for three basic IPR instruments-namely, patents, tal, it tends to focus on a few sectors and to make investments copyrights, and trademarks-for selected LAC and non-LAC with a minimrum size that may be too large for start-ups in some 161 CLOSING THE GAP TABLE 6 12 Weak Skill Supply of Labor Force: Not "Enough" Intellectual Property Rights Index, 1995 High-Quality Research Personnel PATENT Human capital is an important determinant of research and RIGHTS COPYRIGHT TRADEMARK development activities in a country. The level of education INDEX INDEX INDEX in LAC countries as shown in chapter 4 (as well as the level Nicaragua 0 92 and quality of research workers) is a cause for concern in LAC. Costa Rica 1 80 Guatemala 1 80 The supply of researchers is highly concentrated in a few Turkey 1 80 2 42 2 75 'high-quality" institutions For instance, in Mexico, the Honduras 2 10 Thailand 2 24 2 85 3 42 National University of Mexico (UNAM), the Instituto Politec- Bolivia 2 31 nico Nacional (IPN), and the private Instituto Tecnologzco de Mon- Dominican Republic 2 41 Colombia 2 57 3R42 3 17 terrey (ITM) produce most of the graduate-level scientists and Hong Kong (China) 2 57 engineers in the country. The University of Chile and the Uruguay 2 60 2.88 3 00 Catholic University of Chile produced almost 90 percent of Philippines 2 67 2 68 2 50 Ecuador 2 71 all Chilean scientific publications at the end of the 1990s Peru 2 71 3 30 2 83 (Wolff and Albrecht 1992) Argentina's major universities- Paraguay 2 80 especially the University of Buenos Aires-receive the bulk Malaysia 2 85 El Salvador 2 86 of research fundmg going to higher education These uni- Mexico 2 86 2 99 2 33 versities turn out relatively few research-trained scientists and Venezuela 2 90 3 12 2 92 Portugal 2 98 3 83 4 00 engineers because of relatively small graduate programs in Argentina 3 19 these areas. In addition, high-quality graduate programs may Ireland 332 2 88 2 70 have trouble expanding because of the limited number of Brazil 3 50 3 59 3 80 high-quality trained undergraduates (those that graduate Panama 3 52 from the few nhigh-qualty" istitutions). Moreover as we Canada 357 3.63 3 75 Israel 3 57 3 12 3 78 show in the next chapter the linkages between research cen- United Kingdom 3 57 3 80 4 33 ters and the private sector are extremely weak. Chile 3 70 2 83 2 50 In 1997, 140,500 researchers were engaged in R&D in Australia 3 86 3 87 4 80 Germany 3 86 4 54 4 17 the LAC countries, a number that corresponds to only 7 New Zealand 3 86 3.48 3 12 researchers per 10,000 in the labor force. The difference Belgium 3 90 4 50 4 67 Norway 3 90 3 66 4 80 between LAC and countries such as Spain and the United Singapore 3 90 2 60 3 17 States is tremendous-the relative number of researchers in Switzerland 3 91 3 70 4 67 Japan 3 94 4 10 3 50 the latter examples IS about 10 times as high as the LA Finland 4 19 4 28 4 67 average (figure 6.9). Even Chile, which is usually a star Italy 4 19 4 42 4 42 among LAC countries, shows significant deficiencies there. Korea 4 20 3 80 3 25 Sweden 4 24 4 67 4 67 The country's stock of human capital for the knowledge econ- Netherlands 4 38 4 10 4 67 omy is still thin and uneven Chile had only 369 scientists Denmark 4 50 4 33 4 33 France 4 50 4 00 4 00 and engineers per one million inhabitants in 1999, compared Spain 4 50 4 65 4 75 with 2,132 in Ireland (World Bank, SIMA database, 2002). Austria 4 57 4 33 4 42 The recent figure for Chile is roughly the same as that in USA 486 4.35 417rogl the early 1990s. But having said that, the comparison of LAC Sourre Park (2002) numbers-at least for the most advanced countries-in 2000 fields. Furthermore, in order for venture capital investors to have with those of the successful technological transition coun- an exit strategy that allows them to move on to financing new tries in the late 1970s, when they took off, is rather favorable start-ups, an active stock market in new and young stocks is Few scientists and engineers are employed in productive critical. Hall thus concludes that the effectiveness of govern- activities. Two out of three S&T researchers are located in ment incubators, seed funding, loan guarantees, and other such universities, with weak links to business, and only 6 per- policies for funding R&D deserves further study. cent operate in private companies. This compares to more 162 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY FIGURE 6 9 Researchers, 1999* Bolivia Colombia Mexico LAC Cuba Chile Costa Rica Argentina Spai n USA 1_110M _"il 0 2 4 6 8 10 Researchers Per 1,000 Residents (EAP)** * Or latest year available ** Economically active population Source RiCyT (2001) than a third in leading OECD countries with a tradition of monopoly rents unchallenged. Thus firms have less incentive close firm-university collaboration in research Only to pursue technological innovation in order to ward off or stay Argentina, which features high gross tertiary enrollment ahead of competitors The historic protectionist practices in ratios and high levels of enrollment in science-related fields, LAC cotntries and the levels of concentration and relation- has fairly high concentrations of researchers compared to the ship among domestic producers in the same sector have not average LAC level favored or induced the desired market discipline that forces Yet, there may be some circularity in this argument. It firms to seek technological upgrading. As trade barriers have may be counterproductive to turn out a lot of research- fallen in LAC during the 1990s, we have seen episodic and trained scientists and engineers if there are no jobs for them sectoral responses but, with some notable exceptions, these in the economy. Unless private and public industries expand have been slow-perhaps owing to a lack of credibility in the their research activities, research-trained scientists and engi- reform program-and not as extensive as elsewhere. As of neers will be limited to working in the education sector or today, according to the standard, albeit imperfect competitive in jobs that do not require their research skills. Changing indicators, most LAC countries score below the average. industrial culture requires public policies (for example, Brazil's attempt to develop its information technology incentives, as in Korea or the United States) that are out- industry provides an illustration. Brazil tried its hand at pro- side the education sector. tecting and developing its domestic personal computer industry through the Informatics Law (in the 1980s). It pro- Lack of Market Discipline vided various fiscal incentives to firms in the IT sector to Firms in LAC traditionally-up until the 1990s-have not undertake R&D Personal computers and peripherals were faced competitive pressures for technology upgrading. A crit- reserved exclusively for national firms, and foreign par- ical element of the incentive regime and a key determinant ticipation was prohibited. However, by October 1992 it of technological advancement is the extent of competitive pres- was clear that this policy had not created a vibrant, innov- sure that firms face in their domestic markets (foreign mar- ative domestic computer industry in Brazil, and the domes- kets for the export-oriented firms). As firms face increased tic market protection on informatics officially ended. A major competition, their response is to fold over or seek productiv- problem with the policy was that local firms had little incen- ity gains, usually through technological upgrading. In the rive to engage in technological innovation in the absence of absence of competition, firms can continue to reap quasi- foreign competition. In an environment of protected markets, 163 CLOSING THE GAP public R&D funding may be used to further the market power and so on. For most of the LAC countries, which are farther of incumbent firms. Furthermore, by disallowing imports and away from the technology frontier and have distorted tax foreign participation, the government discouraged technol- systems, it probably was the right decision ogy transfer in an industry that is characterized by a high rate LAC governments have, however, mounted significant of technological innovation, and the users had to work with R&D efforts in universities and think tanks. This, along with costlier and lower-quality computers (Mani 2002) the relatively low private sector effort, has led to a situation in which, by global standards, the region has a relatively high Limited Fiscal Incentives and Subsidies public/productive R&D ratio (see figures 6 10 and 6.11) in Most countries in Latin America and the Caribbean do not offer both the financing and the execution of R&D. any fiscal incentives. Table 6.13 presents a brief comparison of This is not intrinsically bad It has generally been the case R&D fiscal incentives across countries. Brazil and Mexico are that countries jump-start the innovation process with public the only two countries that provide some fiscal incentives for efforts. R&D among the LAC countries, and even this support has been marginal and volatile. One way to see that is the B-index, which The low private participation in R&D is reflected in low is an index of fiscal generosity toward R&D. It tells us what experimental development benefit, B, is needed to cover the cost of an initial outlay of The path to industrialization Korea followed-importing cap- R&D of $1. The smaller the B-index, the more generous the ital, learning how to use it, making further expansions, and R&D tax support system.4 As pointed out earlier, LAC coun- then exporting it to others-provides a stark example of the tries have rarely used matching grants to stimulate private importance of experimental development in firms to build- R&D (there is some limited experience through Fundacion Chile ing technological capability Experimental development is and Colciencias in Colombia). Overall, LAC countries have not defined as systematic work based on existing knowledge and made a great effort to provide fiscal incentives for R&D to firms. is directed toward the production of new materials and prod- The reasons are varied, and include fiscal austerity and short- ucts or toward improving products that already exist-a key age of public funds, limited demand from the private sector, component to the adapting stage and for the transition to the induced by macroeconomic stability issues; shortage of skills creation stage. By 1995 more than 65 percent of all Korean and own funding; lack of credibility for sustained assistance, R&D was undertaken by the private sector More than TABLE 6 13 Fiscal Incentives Regimes for R&D as of 2000 CURRENT R&D CAPITAL R&D CARRY-FORWARD TAX CREDIT BASE FOR B-INDEX DEPRECIATION METHOD DEPRECIATION METHOD PROVISION RATE TAX CREDIT 1995 Australia 150% 3 years 3-10 years None - 0 893 India 100% 100% ? None - N/A Ireland 100% 100% ? >100% 1 000 Korea 100% 20% 10-25% Incremental 0 893 Spain Capitalize 100% 3-5 years 15-30% Incremental 0 658 Sweden 100% 30% 7 None - 1 015 Taiwan (China) 100% Like invest 4 years 15-20% Volume N/A USA 100% 3 years 3-15 years 20% Incremental 0 893 Brazil 100% Like invesr 4 years None - 1 030* Mexico 100% 3 years None - 1 015 Peru - - - None - N/A Argentina - - - None - N/A Chile - - - None - N/A Colombia - - - None - N/A Costa Rica - - - None - N/A *1980 ")" refers to unclear information, N/A and - refer to nonexisting Note Depreciation methods of 100 percent (or more) indicate that R&D is fully expensed Under "Capital R&D,` the number of years refers to the duration of the straight-lne depreciation method Under "Base for tax credit," "Volume" refers to the fact that the tax credit applies to total R&D, while 'Incremental" refers to the fact that the tax credit applies to the excess of R&D over some base (usually past R&D or a moving average of past R&D) Source Hall and Van Reenen (2001), OECD (1996) In Park (2002) 164 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF I ECHNOLOGY POLICY FIGURE 6 10 Financing of R&D Expenditures, 1999 100% 40%7 20%- * Public U Private El Foreign U1 Other *1996, **1997, ***1998 Note For purposes of international comparison, public general university funds (GUF) are included under public " These are funds that higher education establishments allocate to R&D from the general grant they receive from the Ministry of Education or from the corresponding provincial or local authorities in support of their overall research/teaching activities "Other" includes funds from the private nonprofit sector (PNP) Source RICyT (200 1), OECD (2002) FIGURE 6 It Gross R&D Expenses by Executing Sector, 1994-98 80 0% 70 0% 60 0%7 50 0% 40 0% 30 0%o 20 0%o 10 090 0 0%0 Argentina Bolivia Brazil* Chile Colombia Mfexico Spain Korea USA U Government U Private Il Higher Educatio UOte Note. "Other" includes funds from the private nonprofit sector Source RICyT (2001), OECD (2001), MOST, Korea, http //www most go kr/index-e html 60 percent of all R&D expenditure in 1999 was in experi- The Latin America and the Caribbean region, by contrast, mental development (table 6.14). The United States similarly puts much less effort into development research, a-s a result directed more than 60 percent of its research and develop- of the low share of private R&D, since that type of R&D is ment effort toward expersmental development in 1999 generally done by the private sector (figure 6.12). In 1999, 165 CLOSING THE GAP TABLE 6 14 R&D Expenditure by Acttiviy (in percentages) BASIC RESEARCH APPLIED RESEARCH EXPERIMENTAL DEVELOPMENT COUNTRY 1995 2000 1995 2000 1995 2000 Argentina 286 28 0 481 44 8 23 3 27 2 Bolivia 49.0 39 0 12 0 Chile 58 6 553 36 5 316 67 13 1 Colombia 39 7 22 3 215 49 7 38 8 28 0 Ecuador* 30 1 64 0 6 0 El Salvador* 58 8 31 8 9 4 Mexico 35 9 23.3** 33 1 47 7** 31 1 29 1** Panama 51 6 47 6* 24 2 45 9* 24 1 6 5* Peru 38 3' 48 3' 13 4- Korea 12 5 12 6 25.7 24 3 62 5 63 1 Portugal 24 9 26 2' 45 0 42 6- 30 1 31 2 Spain 25 3 22 0' 37 0 36 9- 37 7 41 1- USA 158 183 208 229 618 611 'Data for 1999 *Data for 1998 **Da.a for 1997 Note "Basic research" consists of experimental or theoretical work in order to obtain new knowledge about the origin of fumdamental phenomena and observable facts, without any specific application "Applied research" consists of original and new works undertaken in order to acquire new knowledge with a practical and specific goal in mind "Experimental development" consists of systematic work based on exiscing knowledge that came from research or practical experience This is directed toward the production of new materials and products or toward improving products that already exist Note that the Korean dataser calls the category of experimental developmenr `development research " Source Red lberaer,rana de Indiadore en Ceenca y Tecnolog,a (RICyT) Pr:nrcpales ,ndcadore de r,cnrca y ternolog:a, hrrp.//wwwricyt.edu sr/, Korean Ministry of Science and Technology (MOST), http.l/www.most. so.kr/index-e.html FIGURE 6 12 Private R&D Performance and Experimental Development 80 --- -- --_____ _ _ __ __ ___ 5. 70- r 8&0. 70~~~~~~~ '6; 60 -O E-. 5- , 40 - : 30 - -0O' 20 - 0 ZR 10 _ 0 0 0 10 20 30 40 50 60 70 Experimental Development as a % of Total R&D Expenditures Note Data are for 1999 or the latest prior year available including and after 1995 Data are for 13 OECD and 8 LAC countries for which data were available For Argentina, Ecuador, and Peru, data are for science and technology The Latin American countries included are Argentina, Bolivia, Chile, Ecuador, Colombia, Mexico, Panama, and Peru Source OECD Basic Science and Technology Statistics (2001), RICyT (2002), available at http //www ricyt edu.ar/ Argentina and Mexico were among the largest spenders on The low involvement of the private sector in LAC in development research, with approximately 29 percent of research and development reflects thss deficiency in experi- R&D expenditures. Uruguay and Brazil, not shown in table mental development activities (see figure 6.12). 6 14, also appear to do well. The rest of the LAC countries Box 6.8 presents the contrasting approaches to R&D in lagged far behind, with Ecuador and Panama devoting only Korea (private and performance-based R&D focus) and around 6 percent of their research and development expen- Mexico (public and non-competitive R&D focus). In sum- ditures to experimental development. mary, table 6.15 presents the R&D predicament in LAC 166 TECHNOLOGICAL TRANSITIONS AND ELEMENTS OF TECHNOLOGY POLICY BOX 6.8 The Contrasting Approaches to R&D by Korea and Mexico Two opposite poles of R&D systems are represented by the every three years shows good progress By 2000, 36 SRCs, Korean and Mexican systems in terms of spread of R&D 47 ERCs, and 37 RRCs had been selected and funded. capabilities. Acquisition and development of technology in The structure of R&D in Mexico is quite different. Korea results mainly from R&D undertaken in firms, By Except for state-owned companies, such as Pemex and the 1999, more than 70 percent of R&D was funded and under- electrical utility company, the government does not pro- taken by the private sector. Research institutes still did about vide research and development funds or incentives for R&D one-fourth of the R&D in the late 1980s. The institutes are in firms. The research and development effort is centered specifically organized to support industry and received the in federal and state research projects (primarily in agri- bulk of public monies for research. By 2001, there were 27 cultural and medical research), and in the public univer- government research irnstitutes (GRls) supporting research sities About 40 percent of all public university R&D is capabilities in various strategic areas. The government also at the National University of Mexico (UNAM). One-third financially encouraged, and still does, private firms to estab- of all the researchers registered in the late 1980s -by the - lish research centers and research unions with other firms. national system of researchers (SNI) worked at UNAM. By 1995 there were 2,210 research institutes and 63 research Thus, outside of agricultural, petrochemical, and med- unions in the Korean private sector.5 Within the university ical research carried out in specialized public institutions, sector, there are various universities with well-developed publicly financed R&D is highly concentrated in public.uni- research capabtlities. To provide an illustration, the AsiaWeek versities, and much of that research in one university. Indeed, ranking of Asia-Oceania universities by research had 13 UNAM is not only the main institution for scientific Korean universities in the top 50 as compared to 8 for Japan research in Mexico, but perhaps in Latin America as a whole and 10 for Australia.6 Starting in 1989, the government (Gutevara Niebla 1990, pp. 445-52). According to the Pres- introduced Centers of Excellence (COE) in Korea, namely ident of Mexico's 1990 state of the union message, from Science Research Centers (SRCs), Engineering Research 1976 to 1990, UNAM carried out almost 37,000 research Centers (ERCs), and Regional Research Centers (RRCs), to projects and 1,500 technology development projects- encourage basic research in major universities. These cen- 16,000 were in exact and natural sciences, 10,000 in social ters were selected on the basis of creativicy and research capa- sciences' and humanities, 3,500 in engineering, and more bility, and in the case of RRCs, also based on their than 6,000 in medicine (Martinez and Ordorica 1991, p. contribution to the regional economy and comrnunity. Once 91) The research is done in 24 institutes and 13 research the centers are selected, they receive government funding centers. Clearly, UNAM represents one of the single largest for nine years, provided that the interim evaluation done reseach systems in the developing world. TABLE 6 15 countries and contrasts it with OECD and newly industri- Patterns of R&D Expenditures: LAC versus OECD and East Asian alized countries, illustrating the key differences and the weak rigers, Averages for 2000 position of LAC countries in terms of their transition to advanced stages of technological development. R&D EXPENDITURES LAC OECD AND EAST ASIA Total expenditure (as percentage of GDP) 0 5% 2 1% Financing Public sector 75% 23% ¢Conclusion Private sector 20% 66% Executing Public sector 45% 12% In sum, much of Latin America has gotten through the "plug Private sector 14% 70% in" phase of opening to foreign technology through trade Theme of R&D Basicpp and 18% and FDI and getting the basic institutions of property rights experimental 17% 63% and infrastructure in the form of ITC in place. However, few Source Guasch (2002) have experienced a takeoff in productive sector R&D that 167 CLOSING THE GAP appears to be the requisite for being an innovative and high lEndnotes growth economy. Part of the reason is that some of the basic 1 For example, Coe, Helpman, and Hoffmeister (1997) tools that countries have used to boost R&D-such as fiscal construct a foreign R&D capital stock variable for every incentives offered to the private sector-have not been used developing country in their sample as a weighted average in the region. Governments have not addressed the perva- of the domestic R&D capital stocks of the country's 22 indus- sive market and coordination failures. trial country trading partners, with bilateral machinery and But the problem in fact lies deeper. LAC has made large equipment import shares of each developing country in rela- investments in higher education, and in research in univer- tion to the respective industrial country serving as weights. sities and government research centers, yet these have not 2. See Aitken and Harrison (1999) for Venezuela; Aitken, had the effect of either stimulating investments in innova- Hanson, and Harrison (1998) for Mexico, Girma and Wakelin tion by the private sector, or moving the region to the inno- (1998) for the United Kingdom; Hoekman and D)ankov (2000) vation frontier. As of 2000, a number of LAC countries have for the Czech Republic; and Kugler (2000) for Colombia. equal, if not better, educational numbers than the countries 3. Rates of return inclusive of such inter-industry spillover that today are technologically successful had in the late effects-that is, social rates of return (SRR)-have been esti- 1970s, when they took off (table 6.16). So there is no reason mated by a few investigators. In agricultural R&D projects, why those countries could not take off in the successful path the SRR was 150 to 300 percent greater than the private of technological advancement rate of return (PRR). Studies for industrial countries have The next chapter argues that the failure of the LAC coun- also estimated that a 1 percent increase in spillover caused tries to take off on the technology path is because the insti- average costs of other firms to fall by 0.2 percent. Across the tutions, when created, charged with resolving the market major industries that were studied, non-electrical machin- failures inherent to innovation are both internally dysfunc- ery, chemical, rubber and plastics, and petroleum were found tional and uncoordinated among themselves. Addressing this to be major spillover sources. Their SRRs ranged from four issue is the challenge faced by the more advanced countries times the PRR in non-electrical machinery, to three times of the region. the PRR in chemical products, and to twice the PRR in the other two industries (Bernstein 1989). 4. The B-index makes a number of simplifying assump- tions. It assumes equal interest rates across countries (that is, TABLE 6 16 the cost of finance is the same for all countries). It also assumes Similarity of Education Patterns in lA4C (1990s1 versus Technology-Successful Economies (1980) away international differences in the defimtion of R&D for tax purposes. Other taxes that may indirectly affect the motiva- YEARS OF EDUCATION % OF POST-SECONDARY tion to do R&D are ignored (for example, commodity taxes, OF LABOR FORCE EDUCATION property taxes, capital gains taxes) This omission would be Fnland (1980) 33 11 rather serious in jurisdictions without corporate income tax- Hong Kong (China) 6 73 7 1 ation but which rely on a host of other types of taxes Another Ireland 7 60 7 9 drawback is that countries with high corporate tax rates and Korea 6 81 8 9 Singapore 3 65 3.4 high R&D fiscal incentives will score high on the B-index, as Spain 5 15 7 1 opposed to countries with low levels of both. Finally, another Taiwan (China) 6 37 9 3 simplifying assumption is that the firms are large corporations, LAC (2000) rather than small businesses. The methodology assumes that Argentina 8 49 19 7 Brazil 4 56 8 4 firms have sufficient taxable income and thus does not incor- Chile 7 89 15 8 porate explicitly any carry-forward or carry-back provisions. Colombia 5 01 9 9 Costa Rica 6 01 186 5. Ministry of Science and Technology, Republic of Mexico 6 73 11 3 Korea, http://park.org/Korea/Pavilions/PubllcPavilions/ Peru 7 33 22 4 Government/most/policye3 html. Uruguay 7 25 12 5 Venezuela 561 180 6. http://www.asiaweek.com/aslaweek/features/ Source Barro and Lee (2000) universities2000/schools/multi.research.html. 168 CHAPTER 7 Networks and National Innovation Systems T _ s HE PREVIOUS CHAPTERS OF THIS BOOK DEALT WITH THE STOCKS OF FACTORS that are essential to the innovation process and policies to increase them. This chapter focuses on how these factors interact-in other words, on the networks of public, private, and inter- national actors and institutions that form the "National Innovation System." As chapter 6 detailed, knowledge is especially susceptible to rnarket failures and these have given rise to a variety of nonmarket institutions and legal protections: Appropriability problems impede innovation and diffusion. Given the difficulty of excluding others from ideas, innovators will not appropriate the full social benefits of their work. This has given rise to intellec- tual property rights (IPRs) protections, such as patenting that creates a market for new ideas. However, the social value of the innovation is itself a function of the degree and speed of diffusion, which in some dimen- sions is impeded by the IPR system-the cost of using the innovation is higher than the marginal cost of its diffusion.' This contradiction can be resolved to a degree by public universities or think tanks explic- itly set up to provide the public good, or by research consortia in which numerous firms together bear the cost and reap the benefits of research. The same market failure that impedes investment in innovation slows its transmission as well. A firm that incurs the costs of tap- at resolving the non-appropriability problem in "pure sci- ping into the global stock of knowledge by, for instance, a ence," provides a rationale for institutions dedicated to R&D study tour or making contact with foreign universities will and innovation efforts, such as research centers and univer- soon find its discoveries adopted by other firms that free-ride sities Innovation and knowledge developed by those insti- on the investment. This basic failure extends from the sim- tutions tend to be nonexclusive and are made available to plest problem of identifying market niches to, as Hausman all and any interested parties, guided by an appropriate allo- and Rodrik (2002) argue, a country discovering its com- cation of property rights. The issues that emerge then are parative advantage. Historically, this has given rise to insti- how those institutions are developed and how they are tutions ranging from agricultural extension services to public financed. Even more relevant, because the knowledge is pro- research institutions designed to act as "antennae" for new duced by some who are nonusers and used by others, is ideas at the sectoral and national levels. ensuring that what the institutions produce is in tune with what the users want and need-in other words, an incen- Lumpiness and scale economies dictate specialization. R&D tive issue. and innovation are characterized by economies of scale and lumpiness To be effective, resources need to be concentrated Innovation, diffusion, and application require collabo- in a manner beyond the capacity of the individual firm. This, ration among many institutions andfirms. Although inno- combined with the fact that even patents are not effective vation is sometimes the product of one firm alone, the more 169 CLOSING THE GAP common pattern is one of oint efforts, among various firms, which may be too cumbersome for an individual firm to among firms and R&D-related institutions, or among var- attempt. ious R&D-related institutions. Furthermore, progress does not proceed linearly from pure science to applied technolo- Much knowledge is difficult to codify. Often, knowledge is gies, but moves in both directions (Nelson and Rosenberg tacit-not easily codified in blueprints or documents-and 1993); and feedback from frontline users of technology to hence can only be communicated through the interaction of researchers is essential for the refinement of products and pro- individuals This slows the diffusion process and provides a duction processes (Rosenberg 2001). Finally, as a result of rationale for initiatives facilitating the interaction of these specialization, the full supply chain of knowledge is not fully individuals, ranging from conferences, study-abroad programs, integrated, either vertically or horizontally. Technological and faculty exchanges to the promotion of geographic advance is therefore not necessarily evenly diffused through- agglomerations of firms using similar technologies. out the supply chains. In each example, success requires coor- dination and cooperation across all necessary actors-a iatOaDDIa MDnavatDions Systems: $ oD erence process that is subject to coordination failures and transac- and Oncentives tion costs. In many industrial countries, these issues have The development and necessary interaction and coordination given rise to national institutions devoted to fomenting or of these market and nonmarket institutions has led to the eliminating impediments to technological collaboration concept of National Innovation Systems (NIS) and an exten- among different institutions sive literature that we can only touch on here. (See Nelson Information asymmetries: The very long-run and risky nature 1993; OECD 1998, 2001; Lundvall and others 2002.) of R&D further compounds standard problems of informa- Figure 7.1 sketches a simple NIS capturing the ideal inter- tion asymmetries A firm may have a hard time knowing action of these various institutions and, where applicable, whether a researcher conducting basic research is incompe- the human capital employed in them. In theory, the networks tent or whether the problem is simply extremely difficult. of public institutions and private firms interacting in a con- This again argues for specialization of functions-universities certed way to generate and adopt technologies form the have systems for evaluating the quality of their members, mechanisms through which nations learn. This "national FIGURE 7 1 A Simple NIS Human Capital w~~~~~~~~~~~~~~~~~~~. . ...... Universit: -. 4 Think".Tanks/ < i + | Univesity | v DAntennia < o ~~~~~Clusters.. \/ Firm s Other Public Policies Rules of the Game Infrastructure (ICT) Subsidies/Tax Incentives 4 | Innovation & TFP Growth| Coordination Initiacives 170 NETWORKS AND NATIONAL INNOVATION SYSTEMS learning capacity," as numerous observers have called it, is A complementary effort to improve the depth and func- what permits economies to adopt and innovate in their ini- tioning of the NIS is also essential. tial areas of comparative advantage and helps create new ones. In fact, the term NIS must be treated with some caution. (See Furman, Porter, and Stern 2002; Romer 1990, Nelson First, many critical interactions may, in fact, not be national 1993; Wright 1999.) but rather of a primarily regional character: many of the key Examples of the importance of such a learning capacity innovation institutions in Chile's southern forestry industry abound in both manufacturing and in natural resource sec- may have little interaction with those of the northern mining tors. In Taiwan (China), for example, a network of firms reverse- sector. Such subsystems are often termed knowledge or inno- engineered an IBM AT personal computer and produced a vation "clusters." Far more critical, however, is that these successful clone, in the process learning enough to move to "national" systems-and even regional clusters-are and the frontier of notebook computer production (Mathews must be fundamentally international. The critical element 2001) As discussed in From Natural Resources to the Knowledge of successful growth for developing countries is the ability Economy (de Ferranti and others 2002), the Finnish firm Nokia to take advantage of the stock of knowledge abroad and apply began as a highly innovative forestry firm embedded in a it domestically and then work toward becoming contribu- network of institutions that continually introduced new tors to the international flow of new innovation. This implies productivity-enhancing techniques and continues to make that national learning capacity is fundamentally a function the sector very dynamic But what was striking was that the of how plugged-in the economy is to the global centers of same capacity to learn was transferable to the telecommu- innovation. nications sector as Nokia diversified into information tech- A second caveat is that the innovation system depends nologies. This suggests that a key characteristic of the NIS heavily on factors that may or may not be strictly innova- is its flexibility in the face of new innovation possibilities. tion related. Since the center of the innovation process must The nature of the next innovation is unpredictable, but, as necessarily be the firm, all factors that impinge on its per- Pasteur noted, "Chance favors only the prepared mind" or, ception of gains from R&D and the barriers to innovating by analogy, a country with a well-functioning NIS.2 must be considered part of the "extended" NIS. Working the in tradition of endogenous growth, Peter Howmtt and David Mayer of Mexico's CIDE (2002) offer a * In trade policy, access to technologies embodied in formalization of Pasteur's insight in their notion of coun- imported intermediate or capital goods and coopera- tries falling into distinct "convergence clubs" based on their tive R&D projects with foreign investors are proven level of innovative capacity. When faced with the arguably channels for transferring technology. Competitive pres- near-random advent of a new technology, countries with a sures also appear to be critical factors encouraging high level of capacity will be able to use it and then inno- innovation. Japan and the Republic of Korea saw the vate upon it. Countries with lower levels of capacity will be temporary granting of excess profits through trade pro- able only to use it, but their inability to take it further will tection of fledgling industries to be an incentive to limit their ultimate level of development-they will never innovate. However, prolonged protection and entry converge with innovators Finally, those with truly deficient restriction to "targeted" industries led to the late entry capacity will not be able to absorb new technologies, and of Korean firms into international R&D competition stagnate The absence of this complementary capacity embod- and slow growth (Cho and Sakakibara 2002). An open ied in a well-functioning NIS partly explains why LAC has trade policy can be a useful disciplining device in the had lower TFP growth in both agriculture and manufac- product market to complement efforts to get firms to turing than those at the frontier, despite the existence of tech- collaborate in innovation Kim (1993) argues that the nologies for the taking, and why it became dependent on market power of the Chaebols in Korea reduced the foreign technologies over the course of the last century.3 This overall dynamism of the NIS logic also implies that the greater exposure to ideas- * Since R&D is fundamentally a risky and longer-term whether embodied in FDI, imported intermediate goods, or project, well-functioning capital markets are essential. otherwise-through trade agreements such as NAFTA may In fact, Trajrenberg argues (2002) that in Israel tech- not be enough to drive convergence with the richer partners. nology policy was, to a large degree, credit policy 171 CLOSING THE GAP BOX 71 Venture Human Capital in the UnIted States and Ismel Innovation is fundamentally a long-term and risky prospect options-were most important, 17 percent of receiving and hence, particularly for start-ups, venture capital funds firms reported recruiting managers as the most important are necessary complements to other elements of the NIS. activity and 54 percent listed it in the top three. In the United States Microsoft, Cisco, Compaq, Federal This was recognized by Israel in its design of the Yozma Express, Apple Computers, Genentech, and Amazon.com prograrn to jump-start the country's venture capital mar- all counted on venture capitalist participation and 30 per- kets (Trajtenberg 2002). In 1992, Yozma set up a number cent of companies going public relied on it. These mar- of venture capital funds with a total of $US 200 million- kets are distinctly absent in LAC. initially funded by the government as well as domestic and However, venture capital is about more than capital. foreign entrepreneurs from the United States, France, In fact, as de Carvalho, Gledson, and Amaro de Matos Germany, and Taiwan (China). Foreign investors were espe- (2002) argue in "Venture Capital as Human Resource cially valued precisely for their management expertise and, Management," the volume of financing in the UJnited as an incentive, were given the option in five years trme States is relatively small, averaging roughly $US 7 bil- to buy Yozma's shares at a fixed price. lion per year of $US 1 trillion in fixed investment. But Yozma was expected to be phased out after seven years as important as the provision of credit is the access to a but was fully privatized within five. The venture capital network of management expertise that venture capital- market now has 80 funds in operation and has raised ists offer. When asked which of the functions of venture $5 billion. The self-sustaining ability of the venture cap- capitalists-providing financing, benchmarking perfor- ital funds means the government can now focus its sub- mance, providing industry knowledge, developing a busi- sidies on aligning social and private benefits and leave ness strategy, recruiting managers, and two blank financing per se to the private sector. o Entrepreurial and managerial capital is cited by technological policies. Leipziger and Thomas (2000) numerous authors as critical to converting ideas into and Noland and Pack (2002) stress this as central to value added (see, for example, Nelson and Pack 1999). the "Asian Miracle De Carvalho and De Matos (see box 7.1) argue that venture capital markets are as much about manage- In sum, a heightened focus on the NIS does not imply ment networks-linking firms developing new tech- that the first- and second-generation reforms to date are in nologies with experienced managers-as the finance any way less important. function. The final caveat, and the central focus of this chapter, is o Labor market regulations may impede the adoption of that the idea of a "system" connotes a deliberateness of design new technologies (see Hirschman 1958; Parente and and coherent functioning that is often not the case. In fact, Prescott 2000) or the flow of workers from universi- a critical feature of the NIS is that there are few market forces ties to firms and back again, and hence restrict the that guarantee that the various components in themselves communication of tacit knowledge. The OECD (2001) will remedy the market failures they were inspired by, or finds that the tax wedge, unemployment benefits, and engage the actors they were designed to serve. By their very employment protection legislation were all negatively nature, many of the institutions developed to remedy market correlated with specialization in R&D-intensive failures are not market based, and as a result they do not industries. respond to price signals that would ensure that they func- o All the usual factors cited as important to firm tion well. Hence, it is necessary to think "systematically" growth-stable macro-economies, access to interme- about the incentive structures that govern the interactions diate inputs-all appear again as sine qua non for within the NIS to ensure that they are internally coherent encouraging firm innovation, regardless of explicit and generate the desired results. These include those 172 NETWORKS AND NATIONAL INNOVATION SYSTEMS embodied in national laws or administrative codes that act for by the growth of physical factors of production, is per- directly on private firms, or indirectly through their inter- haps most relevant here and, as discussed in earlier chapters, actions with universities, think tanks, and other institutions. has been low in Latin America. As chapter 2 notes, LAC's Ensuring the effective interaction of the available performance in the 1990s has been subpar for most coun- innovation-related factors poses Latin America's second great tries with the exception of perhaps Chile Other "macro" challenge after remedying the gaps in its stock of innovation- performance variables such as growth more generally or related factors In fact, the LAC region has evolved many of unemployment are probably not related specifically enough the "boxes" found in the more sophisticated versions of to the innovation process to be useful. More relevant figure 7 1; however, case study evaluations of its function- although still highly imperfect are measures of intermedi- ing are fairly consistently depressing.4 The families of insti- ate outputs, such as patents, or academic articles per GDP. tutions are often dysfunctional: not only is the available Figures 7 2a-f suggest that while comparator economies- human capital not organized in a fashion most conducive to Korea, Israel, Finland, Sweden, Canada, India, and Taiwan innovation, but balkanized institutions inhibit the desired (China)-have performed above predicted levels, and often interactions. In sum, the region needs to work toward estab- dramatically so, Brazil, Mexico, and Argentina have fallen lishing a coherent set of institutions and incentives that progressively behind while Chile and Venezuela have per- encourage firms to grow, innovate, and look outward. formed at an "average" level.6 Benchmarking the LAC NIS Levels of R&D and Licensing Quantitative benchmarking of the functioning of the NIS, The measure most cited as an indicator of the functioning and hence of the region's performance, is difficult. The mea- of the NIS is the level of productive sector financing of inno- sures of stocks of innovation-related factors presented in ear- vation since the firm, as figure 7 3 stresses, is the vehicle her chapters are relatively easily available but standardized through which knowledge is converted into productivity- measures of the productivity of interactions among the var- and growth in particular Two of the high performers, Israel ious actors in the system are nor 5 Table 7.1 offers a tenta- and Finland, viewed targeted R&D as the critical bench- tive list of indicators that might be employed in the region marking measure and the means to improve the quality of and also helps structure much of the rest of the chapter The the NIS. Arguably the analogous level of licensing payments, first section of the table measures the stocks of innovation- often although not always associated with foreign invest- related factors, especially human capital, and section III of ment, is also a measure of the firm's awareness of techno- the table, the supporting elements of the "greater" innova- logical possibilities and willingness to adopt them. Cohen tion system discussed above such as credit markets, labor and Levirithal (1989) among others stress the importance of markets, and the like. technological absorptive capacity in firms and how this is Section II of table 7 1 corresponds to the core of this chap- highly related to their investment in R&D. R&D functions ter and deals particularly with measures of innovative activ- as the firm's learning unit, helping to benchmark the firm; ity per se and the functioning of the NIS proper The it sets the "technological tone" and eventually builds an quantitative indicators suggested are available for the OECD independent design and, more generally, innovative capac- or elsewhere and hence are comparable. Unfortunately, at this ity (Forbes and Wield 2000) However, much of the litera- point most measures are not available on a consistent basis ture suggests that firms fundamentally need to learn how for the region and in the following sections we are forced to to learn and that this does not happen naturally. At any rely on fragmentary or merely related indicators to get a moment in time firms of differing levels of technological rough picture A central element of the diagnosis of a NIS awareness or capacity coexist. must, therefore, also include the quality information avail- Ideally comparable firm-level surveys of absorptive able on the system. capacity would permit cross-country comparisons, but they are not available At the aggregate level, figure 7.4 presents Indicators of Innovation Outcomes a striking pattern of R&D expenditures versus the fitted As a measure of the overall functioning of the system, Total value for t-he sample as a whole. What is clear is that Korea, Factor Productivity (TFP), the share of growth not accounted Israel, and Finland have experienced dramatic takeoffs. 173 CLOSING THE GAP TABLE 7 1 Potential Benchmarks for the LAC NIS: A Wish List SOURCE OF MEASURE COMPARATOR DATA I Stocks of Innovation-Related Inputs Basic Competencies Avg Education, Secondary, Tertiary Enrollment UNESCO-# Science Personnel # # Scientists, engineers UNESCO_# Education Quality Standardized test scores TIMSS_# Training relevant to private sector Share of students interning in industry OECD02a_Q Incidence of Training Share of firms that train OECD02a_# II Functioning of NIS Output Overall Productivity Growth TFP Growth LFC_# Intermediate Outputs Patents, Articles per GDP LS_# Input Levels R&D (Private, Public), Royalties LS_# Private Sector Absorptive Capacity Measures of awareness, effectiveness _Q Efficiency of Resource Use Patenting Patents/$R&D or # Researchers, adjusted LS#, L & M# Articles Academic articles in journals/Researcher LS Collaboration Contract and collaborative research R&D financing by industry of HE and PL OECD02a_# Faculty consulting with industry R&D consulting with firms by HE, PL OECD02a_Q Cooperation in innovation projects Innovative firms cooperating with HE, PL OECD02a_#/ GCR_Q HE, PT as science information source Innovative firms that use HE, PL OECD02a_# Quality of collaboration partner Perceived quality of R&D institutions GCR_Q Degree of collaboration Perceived level of interaction GCR_Q Incentives to bring ideas to market Royalties sharing guidelines OECD02a_# Gov't incentives to collaborate Promotion programs/gov't R&D financing OECD02a_# Incentives to interact with private sector Presence of tech transfer offices Q Administrative burden Degree of centralization, accounting Q Mission of universnties/PL Stated mission/organization Q Cluster innovative capacity Revealed innovative comparative advantage LM_# Profitability of collaboration initiative DNPV of intervention Mobility of researchers Researchers at HE, PL moving to industry per year OECD02a_#,Q Antennas Brain circulation Number of tertiary-ed workers abroad CD_# Student studying abroad Share of tertiary student with U S F-I visas IEE_# Interactions with foreign universities Agreements Interactions with foreign innovators Perceiitage of scientific publicatLons/patents with foreign co-author OECD02b_# Joint ventures Q Technology outposts Q 111. Extended NIS Icr Connectivity # Internet hosts/1000 ITU_#, Participation # PCs/capita, computer imports ITU_, CC_# Potential Telecom regulations ITU_# Investment ICT investments as % GDP OECD02b_# Credit Markets Credit availability Private sector credit/GDP IBRD_# Venture capital availability Venture capital financing/GDP OECD02b_# Regulatory environment Taxation, minority shareholder rights Q Intellectual Property Rights Index 174 NETWORKS AND NATIONAL INNOVATION SYSTEMS TABLE 7 1 Continued SOURCE OF MEASURE COMPARATOR DATA Labor Markets Mobility among firms Researchers in HE, PL moving tci industry OECD02b Framework Pension portability, hiring/firing costs, and so on ILO, HP_# IV. Data Coverage of above Key # = Quantitative measures available Q = Qualitarive measures available HE Higher education PL Public laboratories Source CC = Caselli and Coleman (2001) CD = Carringron and Detragiche (1999) GCR = Global Competitiveness Report HP = Heckman and Pages-Serra (2000) IBRD = World Bank ILO = International Labor Organization ITU = International Telecommunication Union LFC = Loa)za, Fainzylber, and Calderon (2002) LM = Lederman and Maloney (2003, 2003b) LS = Lederman and Saenz (2003) OECD = Organisation for Economic Co-operation and Development OECD02a (2002a) = Benchmarking Industry-Science Relationships OECD02b (2002b) = Science and Technology Outlook TIMSS = Third International Machematics and Science Study UNESCO = United Nations Educational, Scientific and Culcural Organization For four countries about which we are confident in the dis- and the annex) Calculated at the estimated rate of return aggregation, figure 7 3 suggests that this was largely driven on physical capital, developing countries should be invest- by the "productive" sector's (in other words, firms') R&D. ing over twice what they presently do. At global interest Similar efforts are found in the much poorer China and India, rates of 7 percent, the number is closer to 10 times. Though so this is not purely a function of income Figures 7.2a-f show estimates for licensing are slightly lower, the orders of mag- the results of a more careful benchmarking and confirm the nitude are similar, suggesting that for a while increases in takeoff in total R&D expenditures of many of the compara- both R&D and licensing as shares of GDP should be seen tor countries and that LAC has consistently underperformed. as moves toward the optimum level, even for countries that This is less severe in licensing, where, again, many of the com- have diverged so dramatically from the norm parator countries have maintained above-median licensing There are a range of explanations for why LAC under- expenditures, thus complementing their R&D takeoffs LAC's performs, ranging from a short planning horizon brought licensing has been below median and declining over time. on by persistent macro volatility, to defective credit markets, It is reasonable to ask whether the over-performance of to a rentzer mentality due to a long history of passive nat- some countries is not, in fact, due to well-functioning NIS ural resource exploitation, to the intrinsic conservatism of that encourage the development of innovative firms, but the Catholic religion (although Ireland is apparently an rather indulgence in high-tech white elephants not justified exception) There has, however, been little systematic study by expected returns The equation of returns to investments of firm behavior to date An aspect that probably merits more in innovation to the cost of investment should be the true emphasis is simply that when Scandinavian, Israeli, and benchmark. The existing evidence suggests that these invest- newly industrializing countries' (NIC) firms were being ments are justified Most of the estimates to date of the impact encouraged to compete and to learn how to innovate (see of R&D spending on TFP in selected U.S. firms and indus- box 7.2), LAC compounded its weak educational and tech- tries are astronomical-ranging from 30 to 120 percent- nical background by walling itselfofffrom outside influence. which, compared to a return on capital of 7 percent, implies Yet, as Chudnovsky (1999) notes, even for countries such as that the United States should invest more by a multiple Argentina where education levels are quite high, opening of around at least 4 (see Jones and Williams 1998 and the to trade and competition has not been enough Technolog- annex). ical progress in the 1990s was largely due to ease of access On the other hand, developing countries may invest less to imported inputs and massive FDI, but innovative activ- in R&D precisely because returns are lower However, ity by nationals as measured by patents stagnated across the Lederman and Maloney (2003b) find rates of return that are period while that by nonresidents doubled It is likely that higher for developing countries-principally in LAC and the weak supporting elements of the NIS did not provide Asia-than those for the advanced countries (see figure 7.5 fertile ground for innovative private firms to evolve.7 175 CLOSING THE GAP FIGURES 7 2A-F Patenting, Licenses, and Royalties and R&D Benchmarking Figure 7 2a Patenting in Beta Countries Figure 7 2b Patenting in LAC Pa.eni..g Patenting 12 0 002 08~~~~~~~~~~~~~~~~~~~~~~~~~~ .-0 04 0 6 E -0 06 A 0 04 C o 0~~~~~~~~~~~~~~~~~~~~~~-0 08 ' 02 "l * 0 2----- o 0 -0 12 -0 2 -0 14 Y-a | Koreal o israel -_ *Pinhnd o Sweden oCanada indadl Taiwaa (Chins) Y eto.....*he Bad rgnm Vne| F,.I..d S-d- - C.-d. Id ~ ~ ~ ~ ~ 0Me-i -Chile -Bra-i OAg-nes0 eeul Figure 7 2c Licenses and Royalties in Beta Countries Figure 7 2d Licences and Royalties in LAC Royait-es and License Payments Royalties and LIcense Payments 200 C 150 2 5 10 1 50 2 .8 C 19 iSSi 1982 t903 i9M 198~~~~~~~~~~~~~~~~19&6905i7 1981 OOS otw _ I -ISR N -iFo -150 -2 -200 A-3 ~~~~~~~~~~~~~~~~~~~~250 -4 -300 Yea 0 nio0Chiie 0 Bra-i 0 Argesoo r-n-O -iO NOR a PIN --ORWE -W-CAN -IND -IRL Figure 7 2e R&D Expenditures in Beta Countries Figure 7 2f. R&D Expenditure in LAC 1 5 ~~~~Tona1 R&D Expend--iens Total R&D Expendit-.se 176 100 0~~~~~~~~~~~~~~~~~~~~~~~~~ NNN N~~~~~~N ~ ~ ~ ~~'~~~N - ~~-105 -0 5 -1 50 -1 ~~~~~~~er-200 --CAN -F-IN -5-IND -]-SR OKOR -0--WE --TWN IR lEt CHL -erBRA -i-VEN -MEX O-ARG Source. Lederman and Maloney (2002b) 176 NETWORKS AND NATIONAL INNOVATION SYSTFMS FIGURE 7 3 R&D Expenditure and R&D Financed by the Productive Sector: Taiwan (China), I(orea, Finland, and Israel Taiwan (China) Korea 300 3 00 2 50 2 50 200 2 00 150 150 1 00 1 00 0 50 0 50 0 00 0 00 1960-64 1965-69 1970-74 1975-79 1980-84 1905-89 1990-94 1995-99 1960-64 1965-69 1970-74 1975-79 1980-84 1985-89 1990-94 1995-99 M R&D E.pod,r- (96 of GDP) * R&D Epend-ioe F,n-ced by Prod-c-ve Secoir (96 of GDP) [a R&D Exp&nd-rur (5% of GDP) * R&D E.peod,iur- F-M,ced by Prd---ive .ror 9(% f GDP) Finland Israel 4 00- 350- 350 300- 300 - 250- 250 100 1 00 0o 0 50 1960-64 1965-69 1970-74 1975-79 1980-84 1985-89 1990-94 1960-64 1965-69 1970-74 1975-79 1980-84 1985-89 1990-94 1995-99 | R&D E.prnd::re ,F,.,ced b3 Pnodu-v . S .ror 5(%9 of GDP) | R&D hbp-,d-ore F--nced by Prdoo--or Orror (95 of ODE) | R&D ERpe-durr (% of GDP) II R&D toqp-oohio- (% of GDP) FIGURE 7 4 Predicted and Observed R&D/GDP 4 5% - 4 0% Israel Finland 35%- 3 0%o 255t 2 0%, Korea i 1 0%0 China 0 5% India ArgentIna 0 0% 4 5 6 7 M 8 9 10 11 Mexico Costa Rica Log GDP Per Capita Source Lederman and Maloney (2002) 177 CLOSING THE GAP FIGURE 7 5 Retums to R&D and Physical Capital 100% . 4 lnvestment in R&D should be 3 5 80% - 2 to 15 timeshigher Returns to R&D 3 Optimal R&D/R&D 60% - -2 5 Returns to E Physical K O BRA ARG 1 5 C JA RI ~MEX CHL KORFN 20% 9 or--=0 0 5 0% 0 2,000 7,000 12,000 17,000 22,000 Constant 1995 PPP USS Source Lederman and Maloney (2000b) Before looking at thss aspect in more detail, it is worth and its present preoccupation with being excessively depen- mentioning two variables not included as benchmarking the dent on it. But an export structure that consists entirely of successful functioning of the NIS. assembled microchips cannot be considered intrinsically First, there is a temptation to use the share of "high tech" preferable to a diversified basket of dynamic natural exports, defined by the Standard Industrial Trade Classifi- resource-based industries. cation (SITC) categorization, as a proxy for entry into the Second, information and communication technologies high innovation leagues. As argued in From Natural (ICT) comprise a particularly important element of infra- Resources to the Knowledge Economy, the literature to date is not structure that is revolutionizing information flows and accel- supportive of this view. Agricultural products show higher erating the dissemination of ideas with vast consequences rates of TFP growth than manufacturing (Martin and Mitra for the organization of the economy. Furthermore, the sec- 2001) globally, and recent econometric research suggests that tors themselves-computer and peripherals, for example- countries with high natural resource exports grow faster have shown tremendous productivity growth. That said, ICT partly because of higher TFP growth (Lederman and Maloney is to the NIS as nerves are to the brain-while vital, they 2003c).8 Hunt and Tybout (1998) argue that TFP growth are not to be confused with the cognitive organ itself and in "lower tech" industries grew at roughly the same rates. the mere fact that an economy has numerous Internet nodes More generally, any product can be produced either with a does not in itself signify capacity for innovation. For the highly innovative process or not. Intuitively, maquila assem- purpose of benchmarking, we include them as a vital input bly of computers will require a far less sophisticated NIS than in the extended NIS. advances in forestry in Scandinavia or mining in Australia have, or, for that matter, advances in wine production, salmon EffiGeIDCy oi Use of Resouirces and Collaboration harvesting, or genetic engineering (see boxes 7 3 and 7.11) How a country uses its resources dedicated to R&D, whether would What is a disincentive to growth is excessive spe- public or private, offers additional insight into the func- cialization in any one product, and this partly explains tioning of the NIS. Simple indicators such as patents or Finland's desire to move into the telecommunications sector articles per researcher or per R&D offer a good first cut and, 178 NETWORKS AND NATIONAL INNOVATION SYSTEMS BOX 7 2 Jump-starting a National Innovation Effort: The NIS in Israel and Finland Both Israel and Finland had per capita incomes similar to a variety of services not available from the market-start- those of Mexico, Costa Rica, or Chile and below that of up capital, managerial skills, and so on. The government Argentina when they began their spectacular takeoff in also signed a variety of bilateral agreements with foreign R&D. In both countries, governments targeted a level of governments designed to encourage cooperation with for- R&D and recognized that the entire NIS had to be exam- eign firms and cofinance R&D projects with foreign gov- ned to achieve this goal. ernments to foment research in both countries. Both the Israel established the Office of the Chief Scientist (OCS) R&D subsidies and the Yozma venture capital start-up in 1968 with the mandate to subsidize commercial R&D funds (see box 7 1) were designed to remedy failures in the projects Previously, government support had been capital markets. restricted to national R&D labs, especially those dedicated Finland targeted a rise in R&D spending from 1.5 per- to defense and agricultural sectors. Industrial R&D expen- cent of GDP in 1983 to 2.7 by 2000. From'the begin- diture grew 14 percent a year for the next 20 years. The ning, the government took a "systems" approach, viewing next key innovation was the Law for the Encouragement all elements of the NIS as essential to the R&D goal. A of Industrial R&D in 1985. The goal was to develop National Technology Agency, Tekes, was established in science-based, export-oriented industries. The centerpiece 1983 to finance applied and industrial R&D, especially of the program was a system of financial incentives, par- through clusters. In 1987 the Science and Technology ticularly matching funds of up to 50 percent of R&D Policy Council (STPC) was expanded to include repre- expenses in established companies and 66 percent for start- sentatives from industry and the research community, its ups. These subsidies are thought to have "crowded in" an status raised within the government, and its mandate set additional 41 cents of R&D spending for every dollar. Suc- to include ensuring collaborating across ministries. STPC cessful firms are required to repay royalties of 3 percent reviews science and technology policy every three years, of annual sales up to the dollar-indexed amount of the identifies main policy challenges facing the country, and grant. Paybacks now account for 32 percent of the OCS makes recommendations for all actors. The deficient ven- budget. R&D must be executed by the applicant firm ture capital market was addressed partly by financial lib- itself, products developed must be manufactured in Israel, eralization, and partly by Tekes policy of financing 30-40 and the know-how acquired may not be transferred to third percent of total R&D costs in product development proj- parties. ects it runs These programs require cooperation and net- The magnate program, discussed in box 7.8, sought to working between business enterprises and research integrate the fragmented industrial landscape and encour- institutes and promote technology transfer and interna- age cooperation among firms and between firms and Israel's tionalization Universities were allowed to collaborate with world-class universities. The Incubarors program sought industry in a reversal of previous policy. See box 7.7 for to provide support for fledgling entrepreneurs and provide discussion of consortia policy in East Asia. to a great degree can be standardized across countries. How- Table 7.2 reports estimates of the determinants of patent- ever, more careful work requires controlling for a variety of ing activity in the United States (Bosch-Mossi, Lederman, factors affecting a country's propensity to structure For exam- and Maloney 2003). As expected, the greater the R&D effort, ple, the individual components of the manufacturing the higher the trade volume with the United States; and the processes are much more heavily patented than those of other lower the natural resource component of exports, the higher industries. Furthermore, countries that trade with the United the level of patenting. Interacting the R&D variable with a States are more likely to patent there (see, for example, LAC dummy yields a strong negative coefficient suggesting Trajtenberg 2001) that LAC: underperforms in converting innovative resources 179 CLOSING THE GAP __ education. But neither these variables nor ICT infrastructure BOX 7 3 Buid a BEter (iouse anXd the Wo3d Wifl Beat a Ptto Your knocks out the LAC dummy. What does is either a measure | 11eeor (or is Playing God a Low-Tech lAcvily?) | of the quality of research institutions or a measure of col- laboration between the private sector and universities taken In 1988, Scientists at Harvard's medical school patented from the Global Competitiveness Report. In the complete a mouse genetically altered to develop cancer. The specification, it is the quality of research institutions that "Onconomouse" is now used throughout the world for I drives the LAC dummy. These factors are the focus of the next cancer research and enjoys patent protection in the sections. United States, Japan, and several European countries., The patents give Harvard University exclusive rights to Unvemr$y/R$seaurcOt Ceudi2!eir/1ThDrt Linkages: create the mice and charge licensing fees. Con aiD B! NO'SbJD In December 2002, the Canadian supreme court ruled Critical to the development and continuation of the R&D that higher life forms cannot be patented and rejected capacity of firms in the high-performance countries has been Harvard's application, although it would permit patent- the close interaction of government agencies and universi- ing the process by which the mouse is altered. Canadian ties. As chapter 6 showed, LAC has a very large presence of law does permit the patenting of bacteria, single-celled both, executing the bulk of the R&D budget. A partial organisms, plants, and agricultural crops. The issue is explanation for this pattern is found in low private sector hotly debated, with religious groups praising the deci- investment and in the existence of state-owned public agen- sion on ethical grounds and others agreeing with the cies in the postwar period that developed R&D capacity in court that Harvard was not really inventing the mouse, 1 order to better exploit the available natural resources and to only tinkering with it. Nonetheless, industry observers jumpstart industries to supply local demand For less clear argue that the inability to receive life-form patents could reasons, universities play the dominant role in the region in create a chilling effect on scientists doing research in both government funding for innovation and attracting most Canada, especially since for many emerging companies, of the skilled researchers. patent royalties are their only source of income. There There may be nothing intrinsically wrong with a high are 403 registered biotechnology companies in Canada share of the public sector or university sector in the execu- with a total of 58,000 employees and $US 3.21 billion tion of R&D. The previous section stressed the importance in annual revenue (Wall StreetJournal, December 6, 2002, of the productive sector making an increasing effort over p. B6). time. But given the extraordinarily high rates of return to All biotech products, including the Onconomouse, R&D, it would seem perverse to view a strong comple- are classified under agriculture in the SITC categories mentary public sector effort, particularly in the area of basic and hence would not count as "high-tech" exports. science, as somehow bad.9 However, consistent with the ! regression results above, the literature on Brazil, Mexico, Chile, and Argentina referred to below stresses lack of coor- into actual innovation. Figure 7.6 presents graphically the dination among public sector institutions and the private estimates of the interactive parameter for a variety of sector as one of the major obstacles to improving their sci- countries of the region. What is immediately apparent is that ence and technology capacities. LAC, with the exception of Costa Rica and Venezuela, is LAC has few measures with which to benchmark coor- among the worst performers, with its coefficient roughly 5 dination. One is simply to "follow the money." In Mexico, percent below that of the OECD and far below that for for example, there is virtually no cross-financing: Only 20 Korea, Finland, or Israel. That is, not only does LAC under- percent of the private sector's R&D expenses are financed out- invest in R&D, but it uses it ineffiiently. side, and mostly by foreigners One hundred percent of the The next rows of table 7.2 pursue the possible causes of governments' spending is self-financed. And in the higher this underperformance. Sachs' measure of openness and gen- education sector only 7.8 percent is financed by the private eral market functioning does show up as a significant deter- sector. From this Bazdresch (2002) concludes that the three minant of efficiency in many specifications as does years of sectors function more or less autarkically. In effect, Mexico, 180 NETWORKS AND NATIONAL INNOVATION SYSTEMS TABLE 7 2 Detenninants of Patenting in the United States Observations 512 512 512 512 419 419 Countries 53 53 53 53 50 50 In (Average Patents 1963-64) 0 37 *** 0 25 *** 0 32 *** 0 27 *** 0 37 *** 0 29 In (R&D Expenditure) 0 50 *** 0 24 ** 0 36 * 0 33 *** 0 49 *** 0 24 In (Ustrade) 0 32 *** 0 31 *** 0 29 *** 0 36 0 17 0 22 Natural Resources -0 03 ** -0 07 *** -0 08 *** -0 10 *** -0 05 *** -0 10 LAC*ln (R&D) -0 05 *** 0 00 -0 01 -0 02 *** -0 02 *** 0 00 Opennessln (R&D) 0 02 ** 0 02 * 0 01 0 00 0 01 0 00 In (Quality)*1n (R&D) 0 17 *** 0 10 In (Collaboration)Yln (R&D) 0.11 *** -0 03 In (years of Education)*ln (R&D) 10*** 0 08 ICT X In (R&D) 01 *** 0 00 Time Trend 0 03*** 0 03** 0 03 -0 01 -° 07 ° 03 Pseudo R-squared 0 19 0 20 0 20 0 21 0 23 0 26 x implies an interactive rerm between two variables * indicates significance at the 10 percent level i* indicates significance at the 5 percent level *** indicates significance at the 1 percenc level Negative binomial estimates, presample mean period for fixed effects, control 1963-64, see annex for details Soarce Bosch, Lederman, and Maloney (2002) FIGURE 7 6 Inefficiency of LAC Countries in Patenting in the United States Interactive Effects of LAC Countries with R&D Expenditures and People in a Negative Binomial Estimation of Patents on R&D Variables Compared to the OECD 15 00% 1000%_ 5 00% 0 00%0 -5 00°% -10 00% Source Bosch, Lederman, and Maloney (2000) like Latin America more generally, systematically set up the Government-Financed Research Institutes institutions for innovation without institutionalizing an As chapter 6 suggests, these institutes enjoy a dispropor- innovation system. Ensuring the effective interaction of the tionate share of the national research budgets. The logic has available innovation-related factors poses LAC's second great tended to be that fledgling industries often lack the in-house challenge after remedying the gaps in its stock of innovation- capability to undertake the necessary research. Alterna- related factors. tively, the same problems of appropriability and lumpiness 181 CLOSING THE GAP suggest that industry-level institutions can address an logico da Aeronautica, and the Empresa Brasileira de Pesquisa important market failure. However, as Rosenberg (2000) Agropecuaria (EMBRAPA), they conclude that most of the argues, in practice, government institutions established for 18 primary public research institutes have had limited these purposes may have limited impact for two reasons. impact, precisely because of their limited contact with First, in general, government researchers have relatively industry. A survey of sources of technology found that less little understanding of the specific needs of industry and that 2.5, 5.1, and 3.9 percent of product designs, tool hence the necessary feedback is absent. Second, it is diffi- designs, and manufacturing processes, respectively, origi- cult to provide researchers as public institutes with strong nated in research institutes (Braga and Matesco 1989). incentives to be responsive to industrial needs. In the LAC Again, benchmarking with comparable numbers is difficult. region, the latter factor emerges as especially critical, com- OECD numbers on "Innovative Manufacturing Enterprises pounded by the competition of other non-innovation- That Use Public Laboratories as an Information Source in related agendas. Innovation" are also somewhat low, with the maximum in The development of think tanks during the import sub- Ireland at 7.4 percent. Such comparisons are rendered dubi- stitution period, in which competition was stymied and ous by critical definitional issues-such as what is consid- outward orientation discouraged, led to inward looking ered an "innovative firm." Further, though the data are not and bureaucratized institutions with little connection available, in Korea, Japan, or Taiwan (China), where public with the private sector. In Mexico (Bazdresch 2002), the laboratories are credited with providing the big push toward roughly 150 Public Centers of Investigation are dependent the knowledge frontier, the number would presumably be on the secretariat to which they belong and frequently much higher. oppose any efforts of the secretariat to contract firms or Chudnovsky (1999) finds a similar distortion of mission outside universities that might be more qualified than in Argentina, where the National Atomic Energy Com- they are to investigate a particular question. They thus mission absorbs substantial government R&D resources often monopolize the possible incentive that government while the National Institute of Industrial Technology procurement can have Further, the lack of competition has (INTI) receives less than 5 percent of the budget despite had the usual depressing effects on quality and created the manufacturing sector generating 25 percent of GDP. obvious disincentives to work with other institutions that Lack of coordination among the programs of different orga- might be potential rivals. Proposed reform laws foresee nizations, lack of precise objectives and evaluation objec- greater autonomy for the centers and further offer the pos- tives, and lack of long-run projects prevent effective sibility that research funds will be allocated by competi- technological transfer and innovation to critical productive tion and not automatically to the particular center of sectors. investigation. For unclear reasons, the incentives in the Caribbean have led to vital regional agencies, such as the Universities Caribbean Tourism Organization, to be relatively slow in The university as a collection of scientists, interacting with monitoring basic trends, compiling and disseminating the private sector, emerges as a critical element in stories of data for reference and analysis, or collaborating with other the relative success or failure of the industrial countries (see agencies working in what remains a critical sector in the box 7.4). This is also true historically in Latin America, and region. universities claim a large share of the region's overall research But the existence of national agendas competing with that budgets (see figure 7.7)10 of jump-starting a dynamic private sector has also taken a There are several channels through which universities toll. In Brazil, Dahlman and Fischtak (1993) attribute the enrich the innovation network. First, the production of low interaction with the private sector to a preoccupation tertiary-educated workers is the lifeblood of the NIS and per- with foreign exchange generation in the 1980s, combined haps its most important function Interviews with high-tech with weak incentives to adapt to the changing needs of the companies in Costa Rica highlight the issue of generation private sector over time. With the exception of the Instituto of quality human capital in a country as an order of mag- de Pesquisas Tecnologzcas (IPT) in Sao Paolo, the Centro Tecno- nitude more important than other measures discussed in 182 NETWORKS AND NATIONAL INNOVATION SYSTEMS I BOX 7 4 What Is a College Education For? The incegration of universities with industry emerges as and utilization operating efficiently at a national a central explanation of the disparate performance of U.S level because it was self-contained, internalizing and U.K. economies. To quote Nelson and Wright (1996): the resource base and market demands of the national economy (p. 147) As with education more generally, what is impor- | tant is not the sheer number of students or the quan- This can be compared to Lazonik's (1994) diagnosis of tity of their training, but the effectiveness with the very different role of U.K universities, a central factor which that training is integrated into the process of in explaining the failure of British enterprises to take improving the technology of operating firms In advantage of the technological and market opportunities interwar America that coordination was advanced to of the second industrial revolution: a high state of refinement, as the curricula of edu- cational institutions came to be closely adapted to the requirements of the "positions" that graduates Lacking industrial roots, the aristocracy who con- would be taking; and vice versa ... in 1919, for exam- trolled these elite institutions during the era of the ple MIT launched its Cooperative Course in electrical second industrial revolution had no need for an edu- engineering, a program that divided the students' cational system that developed technologists. They time between courses at the Institute and at valued the study of science as a branch of sophisti- General Electric, which hired half of the students cated knowledge but had no interest in its applica- after graduation. The program was later joined by tion to industry Indeed, the British elite positively ATT & Bell Labs, Western Electric, and other firms resisted the notion that a concern with technology Whatever the merits of .. reservations about the had any place in an aristocratic education. They close links between universities and private firms, wanted education to set them apart from the lower what [emerged] is an effective network of training classes (p. 170) FIGURE 7 7 Allocation of R&D Resources Percent 70- 60 50 Cot 40 Bra Chi 30 Bol |/ -I Bol 20 -T~ lo X H- 3D ~Chi R&D $ in the Scientists in the Basic R&D/total University Private Sector Source Hansen, Agapitova, Holm-Nielsen, Vukmirovic (2002) 183 CLOSING THE GAP chapter 6-IPRs, R&D incentives, and so forth. Second, as (2002) argue that this high fraction of universities in total discussed earlier, universities are better suited to carrying out R&D effort translates into a heavy emphasis on basic research, larger, long-term, less immediately applicable research. which accounts for approximately 60 percent of total R&D Third, they are often more likely to maintain contacts with spending in Chile and 50 percent in Bolivia In the United research centers in industrial countries and hence perform States, where higher education amounts to only 15 percent the search for new technologies In all cases, the degree to of total R&D expenditure, basic research accounts for approx- which they remedy the underlying market failures is depen- imately 16 percent of total R&D spending (figure 7.7). The dent on the tightness of the links to the private sector-that appropriateness of this mix remains debated. Pavitt (2001), is, the degree to which they help solve problems, train stu- for instance, argues that if, in fact, small countries could dents, or identify foreign knowledge 'that is relevant to the simply take the basic research of the advanced countries, the production process in some way. extraordinary efforts in basic research in Scandinavia, or even Today, the higher education sector plays the dominant role Korea and Taiwan (China) would appear irrational. Not so, in the region in both government funding for innovation and however, when the importance of basic research in training providing the most skilled researchers. As the previous chapter human capital and in identifying and facilitating the absorp- suggests, at least 50 percent of total R&D expenditures in tion of foreign technologies is acknowledged. In general, basic Brazil, Bolivia, Chile, and Colombia is through the univer- research is not as easily communicated across countries as sities. Further, generally, most researchers are found in the often assumed and developing countries cannot confine them- universities. In Chile, for example, two of every three R&D selves to simply adapting shelf technologies professionals are found in academia, and only 6 percent are Nonetheless, a measure of the value of this nonapplied in private enterprises. By contrast, in more industrial coun- research to the economy can be captured by the degree tries such as Finland and the Netherlands, a third or more of interaction with firms. Figure 7.8 suggests that the of scientists and engineers are found in firms (Brunner 2001). perception of entrepreneurs of the degree of collaboration is The high concentration of R&D in the universities has an far below that in other parts of the region and this is borne impact on the nature of the research done. Hansen and others out by country and industry studies as well For example, FIGURE 7 8 Quality of Scientific Research Institutions and Universit Industry Research Collaboration 7 65 6 5 5 5 4 5 3 5 3 25 u u E X x c U c El Quality of scientific research institutions Cl University/industry research collaboration Source: Based on World Economic Forum (2001/02) 184 NETWORKS AND NATIONAL INNOVATION SYSTEMS TABLE 7 3 Tentative Benchmarks of Interaction of Higher Education (HE) and Public Laboratories (PL1 with Firms FINLAND GERMANY IRELAND SWEDEN UK Innovative manufacturing firms that cooperate with HE 47 3 104 13 8 26 1 Innovative manufacturing firms that cooperate with PL 38 0 13 6 6 3 16 3 Innovative manufacturing firms that use HE as information source 6 9 6 7 5 0 4 5 3 9 Innovative manufacturing firms that use PL as information source 5 3 2 9 7 4 - I 9 - Nor available Sourca OECD (2002a) in Finland, at least 40 percent of firms have collaborative tutions and the private sector in LAC. These are summarized arrangements with universities (Brunner 2001 and table 7 3), below and as Blomstrbm, Kokko, and Sjoholm (2002) document, these interactions have been vital to the continued dynamism * Real or perceived low quality of scientific institutions. Sur- of both the high-tech and more traditional forest industries. veys suggest that, in the LAC Region, the private Brunner notes that in Chile, 32 percent of firms acknowledge sec-tor's opinion of the quality of scientific institutions benefiting from innovations from universities and 25 percent is not very high, with the exception of Costa Rica- actually have signed contracts Nonetheless, surveyed firms and this seems highly correlated with the level of inter- did not report that external interactions were crucial to the action (figure 7.6) and explains why collaboration firm and most new ideas came from within In a similar vein, disappears as a critical variable in the patent regres- in his study of Avimex, a veterinary pharmaceutical com- sions above when quality is included More concretely, pany in Mexico, Mayer (2002) notes that despite a high common measures such as academic publications, or 10-15 percent of sales spent on R&D and world-class inno- patents per university researcher, can offer an indica- vations in joint pro)ects with U S research institutes, the tion of the quality of the university. major disadvantage Avimex and Mexican firms in general * Lack of incentives for universities to link and address pri- face Is the lack of research institutions available for joint vate sector knowledge needs. The incentives within the research projects. universities are generally biased away from collabora- There are success stories of university-private sector col- tion with business laboration. Rodriguez-Clare (2003) argues that the most crit- * University culture. Arguably in universities there is a ical element in the development of the software industry in more "liberal arts" as opposed to a "technical" culture, Costa Rica was the strength and dynamism of the Universi- with deep historical roots that resonate with Lazonik's dad de Costa Rica and the Instituto Tecnol6gico de Costa Rica, analysis about the inappropriateness of the U.K which have pushed the country to the top of Latin America system (see also Maloney 2002). Agapitova and Holm- an the perceived quality of scientific research institutions and Nielsen (2002) argue that, overall, the university men- an "impact index" of its publications. This has translated, tality in Chile is not geared to solving problems on a for instance, into the setting up of a center for software devel- business time scale, and Mullin (2001) argues that opment and semiconductor design by Intel. The Chilean overall academic interests tend to be narrow and unap- fruit sector provides another example of the importance of plied. The experience of Mexico suggests that the universities to the diffusion of technology and the develop- inward orientation of academia to a large extent fol- ment of innovations appropriate to local conditions. How- lows from the rules guiding promotion, which put ever, a generalized picture is one of monasteries, where the extensive stress on academic performance that is of most talented of the scientific brotherhood are cloistered litt-le relevance to society at large. As the exception away from the applied innovation needs of the economy. that appears to prove the rule, observers of Costa Rica's There are a number of factors-along the themes of poor two excellent technical schools stress not so much design and faulty incentives-responsible for the dearth of incentives but the "foundational impulse"-a desire effective linkages and collaboration between scientific insti- to be patterned more on the Massachusetts Institute 185 CLOSING THE GAP of Technology (MIT) or other technical schools of BOX 7.5 excellence than on those with a liberal arts bias Did the l3ayh-lole idt Oncrease the Commercialization Intellectual property rights. Beyond this bias, a rich lit- i P3uIblicsy londed Research? erature discusses how the incentives in academia- intrinsic, reputation, and financial-are distinct from From 1980 to 1984, a series of laws, the most famous those in industry, and hence there is no guarantee that being the Bayh-Dole act, allowed universities to patent the two sets of organizations will interact productively. the results of federally funded research. The main argu- Globally, the incentive to innovate and then contribute ment for the act is that intellectual property rights would to bringing products to market has been a source of help accelerate the commercial exploitation of labs and debate (see box 7.5). The overall allocation of rights universities' discoveries because of the additional invest- to the benefits of innovation, the faulty incentive ments needed to turn university and laboratory R&D regime to appropriate the benefits of public-financed into commerctally viable products for the marketplace. R&D that exists in LAC countries, does not favor coop- Mowery (1999) and Mowery and Ziedonis (2002) eration. (A U S survey finds that at least 71 percent argue that the Bayh-Dole Act did not stimulate uni- of the inventions require further involvement by the versity patenting and licensing-the trend had already academic researcher if they are to be successfully com- begun before the act. Other factors, such as the increase mercialized Forty-eight percent of the ideas are in in the number of Technology Transfer Offices, stimulated proof-of-concept stage, 29 percent have a prototype universities to patent more-as did the emergence of available on a lab scale, and manufacturing feasibil- the biotech and software fields. Henderson, Jaffe, and ity is known for only 8 percent.) (See Goldfarb, Traltenberg (1998) found that the explosive growth of Henrekson, and Rosenberg 2001.) Although the num- patents was probably supported by the legal changes but bers for Latin America are likely to be different, there that after the mid-1980s there was a decline in the qual- is no reason to believe that commercialization of inven- Ity of the patents as measured by citations. tions will require less involvement of the researchers, Jaffe and Lerner (2001) examined the behavior and and the incentives for that are lacking in LAC. Because quality of patenting by U.S. Department of Energy lab- researchers cannot appropriate the benefits of innova- oratories in reaction to both Bayh-Dole and the earlier tion, they have little incentive to undertake innova- Stevenson-Wydler Act (1980), which specifically encour- tions and link with the private sector. Various aged the commercialization of their research They found developed countries allow ownership rights to gov- that, until 1980, national labs had lagged behind uni- ernment-funded R&D and in some instances, such as versities in terms of patents per R&D dollar. But post- in the United States and Japan, explicitly in the 1980, national labs reached parity with universities in national patent laws (see box 7.5 and table 7.4). The patenting activity The citation rates for national labo- allocation of property can also influence whether the ratories either rose or remained constant. researcher stays affiliated with a university or must leave to pursue the commercialization of the product. Arocena and Sutz (2001) argue that these reforms had important impacts on stimulating collaboration may be less germane to the Latin American case High Universities also often have liberal policies on faculty costs of patenting abroad, and low enforcement of leaves of absence and consulting privileges that allow patenting domestically, means a focus on fostering faculty members to pursue commercial opportunities university-firm relations less intermediated by patents. while keeping their position at the university intact In fact, this implies that direct collaboration between The roots of Sweden's lower rate of university the two sectors must be even better than in the indus- spillovers are found in the absence of such policies trialized countries (Goldfarb, Henrekson, and Rosenberg 2001). o Bureaucratic disincentives. In the United States univer- In Latin America, the barriers are much more mun- sities and government think tanks have established dane Mayer (2002) and Brunner (2001) argue that, technology transfer offices (TTOs) that appear to have in Mexico and Chile, bureaucratic rigidities make it 186 NETWORKS AND NATIONAL INNOVATION SYSTEMS TABLE 7 4 Guidelines for Sharing Royalties from IPRs SHARE OF ROYALTIES LABORATORY/ APPLY TO INVENTOR DEPARTMENT INSTITUTION NO SHARING Australia Universities 33% 33% 33% Austria General Practice 100% to owner Belgium Flemish Universities 10 to 3096 50% 20 to 30% Canada Federal Research 35% by law variable variable France Public Labs 25% 25% 50% Germany Max Planck and HGF Centres 33% 33% 33% Hungary 0% undetermined up to 100% Israel Hebrew University 33% 33% 33% WelLmann Institute 40% 0% 60% Japan Universities 100% to owner Korea KIST Institute up to 60% 0% 40% Netherlands Public Labs 100% to owner Poland no general rule United Kingdom BBRCs [CK] Sharing encouraged by institute guidelines United States Universities Sharing required by law Stanford 33% 33% 33% LAC Chile General Practice No general rule, researcher <30% Mexico Public Labs 100% to owner Soerce Table from OECD (2002a) updated by authors TABLE 7 5 Policies for Improving Academic/Private Sector Collaboration INTERACTION EFFECTIVENESS OF MODES CONSTRAINTS FACED POLICY INSTRUMENTS INSTRUMENTS/CAVEATS Licensing Universities not allowed to "own" research Modify IPR law May reduce incentives for basic research? Consulting Culture gap between universities and firms Improve contractual cei tainty (IPRs) May reduce incentives for basic research' Bureaucratic barriers Reduce public funding to encourage May "corporatize" academic collegiality? Weak infrastructure private contracts May mistakenly patent basic research' Decentralize systems and offer greater autonomy Improve accounting systems to allow for contracts Collaborative Information asymmetries Incubators May turn into subsidized real estate venrures Coordination failures Technoparks May lapse into old Industrial Policy Agglomeration externalities Science parks Distract from necessary horizontal measures Consortia Cluster policy difficult to write contracts and get access to the use of Policies to Improve Collaboration laboratories and equipment from the university. Aside from the necessary horizontal policies related to edu- Arocena and Sutz (2001) cste the inadequacy of the cation, IPRs, and internal incentives within government university accounting systems as slow, heavy, and agencies, and those that are essential elements of the unsuitable for handling resources and contracts with extended NIS-ICT-well-functioning credit markets, and the private sector rapidly. Numerous universities have outward trade and investment orientation, governments generated pseudo-TTOs although these tend to be pas- have experimented with a variety of policy measures to sive, waiting for the private sector to discover relevant actively improve the interactions among elements of the NIS university research rather than "selling" it. (table 7 5). 187 CLOSING THE GAP Industry Clusters: Subsystems of the NIS the issue is broader. New technology generation is The best-known network concept is that of clusters that, highly capital intensive, and having producers of in the present context, can be seen as subsystems of the NIS. machinery intensively interface with users increases Well-known examples in the United States are Silicon the chances of a "better fit." Similarly, owners of Valley and Boston's Route 128 (IT and biotechnology), specialized equipment may want to spread out the greater Seattle (software and aircraft), and North Carolina's costs of new investments by renting it out to other Research Triangle (electronics, pharmaceuticals, biotechnol- firms. The same applies to labor, which also becomes ogy). Similar examples can be found in Sweden and Finland highly specialized In fact, co-specialization of labor (forestry), and Australia (mining). Recent studies of the LAC via joint training programs is a common feature of Region have identified emerging innovation clusters of dif- clusters attempting the transition into the innova- fering varieties and size in Argentina, Brazil, Costa Rica, tive stages. Another example is information and Cuba, Peru, and Mexico-although none that could be con- communication technologies, where innovative clus- sidered "mature" (Bortagaray and Tiffin 2000). As in inno- ters rely on timely market intelligence and efficient vation more generally, the high potential sectors may be logistics management to ensure on-time delivery in high technology-related activities, but may as well be and tracking to far-flung markets through high- in fruits, vegetables, tourism, wine, and so forth as well as quality facilities When the provision of infra- manufacturing. structure entails high fixed costs, significant savings The rationales for cluster formation are the same market per user can be obtained when consumption of these failures discussed in chapter 6. infrastructure services is spread out over many firms in one place. Furthermore, financing of R&D by sup- • Clustering may further ensure that technological pliers of risk capital is extremely risky, and traditional spillovers are, in fact, reaped by other related firms banking instruments are unable to bear this risk through greater flows of relevant information. Clus- because they are ill equipped to value the collateral tering encourages nonmarket transactions, which at the early stages-often an idea in the heads of reduce transaction costs and permit a greater level researchers. Venture capitalists specialize in this and of transparency and trust, which is important in concentrate near high-tech businesses because prox- managing the appropriability problems of knowl- imity to these intangible assets is the best way to edge. The punitive consequences of free riding are evaluate quality. immediate, and this lowers the transaction costs of interaction, which translates into efficiency gains The discussion has thus far focused largely on innovation for all. clusters as opposed to the more traditional notion of back- o Tacit knowledge is more easily diffused through offer- ward and forward linkages often stressed As From Natural ing the possibility to be able to observe firsthand and Resources to the Knowledge Economy stressed, in an age of rapidly share its application and the learning-by-doing process. decreased transport costs and the fragmentation of global Access to skilled labor, the principal vector of tacit production, the latter focus is perhaps less central than knowledge, is automatically made easier for firms previously and, to the degree that it creates industrial because labor itself chooses to locate in areas where sev- policy working against a country's comparative advantage, eral employment options exist, lowering job search counterproductive. costs arising from acquiring, processing, and acting on That said, a concern with improving conventional inter- information about relevant jobs Labor mobility then actions among firms and sectors can be seen as a common becomes the key to building the dynamism of tech- strategy of promoting industry dynamism. Guasch (2002) nology clusters. identifies several stages of cluster evolution: o Indivisibililiites of all kinds justify coordination among firms. Clearly, technology's lumpiness and o Diagnosis phase. Identification and elimination of riskiness dictates cooperation as discussed earlier, but problems and bottlenecks. 188 NETWORKS AND NAI IONAL INNOVATION SYSTEMS * Economies-of-scale phase. Facilitating coordination to over the long run they will need to develop an innovative capture economies of scale, transportation, marketing, capacity that will allow further product differentiation, along licensing, commercialization, purchasing, and access- with quality and productivity increases Korea, they argue, ing markets lost its original edge in computers precisely because it rested * Spillover phase. Capturing knowledge spillover too long on the cost advantage and simple turnkey plants. effects and disseminating current knowledge within Taiwan (China), on the other hand, reverse-engineered the the supply chain-forward, backward, and lateral IBM AT, assigned the component parts to individual firms linkages for local production, and not only had a functioning clone * Knowledge creation phase. Joint learning and inno- within one year but had "learned" enough as a country to vation. innovate at the frontier of laptops within five years (Mathews 2001). The lesson is that neither trade opening The relatively quick and easy-to-secure benefits of the nor a focus on the first two phases of cluster formation unac- diagnosis and economies-of-scale phases are persuading some companied by an aggressive technology program will lead LAC governments to promote cluster development focusing to sustainable growth primarily on three broad policy areas export promotion and One way to benchmark the success of a clustering effort attraction of inward investment, value chain integration, and is to measure the development of "innovation revealed com- networking/SME policy However, few clusters reach or are parative advantage" (IRCA), calculated as the flow of patents intended to reach the knowledge creation phase (Bortagaray in industry k in countryj over total new patents in country and Tiffin 2000) This raises the concern that a lack of atten- j relative to the same ratio for the world as a whole. A value tion to the innovation dimension will lead to an inability over 1 suggests that a country has a comparative advantage brought on by the initial stages. As an example, Lee and Lim in innovation in that industry. Figure 7.9 looks at two promi- (2001), among others, argue that countries may initially nent experiences in the region. The first is the aircraft and compete in derivative industries on the basis of low cost, but parts industry dominated by Embraer in Brazil. Though a FIGURE 7 9 Brazil and Mexico: IRCA in Aircraft and Computing Equipment 14 -+-- Brazil Aircraft -U-Mexico Office Computing and Accounting 12 S..A I 0 - U 06 i) 04 02 1980 1984 1988 1992 1996 2000 Year 189 CLOSING THE GAP strong competitor in the global market, the high import con- BO7 tent of the aircraft raises some concern about whether, in fact, BOX 7 6 the industry has the innovative potential to continue as a I dynamic player. The IRCA since privatization in 1994 sug- With limited domestic resources for R&D in Finland, gests a slow and steady trend toward 1, which suggests that it was imperative that the available public funds were it does have the potential to remain a dynamic sector over used efficiently-so the government sought to ensure time. Somewhat more worrisome is the computing sector that all possible spillovers would occur. The degree to in Mexico, which, despite the presence of IBM and HP for which potential spillovers are realized depends on the over 20 years, does not and will not have IRCA in the near intensity of contacts between innovators and related future. This raises the question about whether Mexico really firms, as well as the ability of related firms to absorb the has the capacity to be a participant in the sector over the potential knowledge externalities-factors that led to a long run. "cluster" focus in Finnish R&D policy. By subsidizing i R&D in industries and companies that belong to major clusters-where input and output linkages are strong- Pokcies ts Io Domen Deur&m ColDaboiration it was possible to concentrate resources in sectors where fon oDflOvDatol I the likelihood of spillovers is relatively high while at the Approaches to encouraging interfirm collaboration in inno- I same time addressing the non-appropriability issue in vation range from the surgical to the promethean Feser both invention and diffusion. Furthermore, by explic- (2002) argues for an approach that seeks to "leverage" inno- itly including small and medium-size firms in publicly vative synergies among existing or emerging businesses funded R&D projects, which are often dominated by rather than seeking to initiate particular clusters. Leverag- large MNCs such as Nokia, it was also possible to ing implies using the most important insights of the clus- strengthen the absorptive capacity of these firms and ter literature-the notions of cooperative competition and achieve a faster rate of technology diffusion. Thus, the collective efficiency, supported by standard policy analyti- basic idea was to support the development of the clus- cal tools (public goods and market failure) to strengthen ter as a whole rather than supporting individual com- innovative synergies. This appears to have been a large part panies and "improving integration, collaboration, and of Finland's (box 7 6) approach-ensuring that all related division of labor within the whole research system." This firms took advantage of possible spillovers. This may or may last emphasis is important since the intense focus of the not imply a regional dimension, but it allows key formal and Finnish government on improving the functioning of the informal linkages to occur at all spatial scales and between NIS, especially through eliminating bottlenecks caused different kinds of firms. by shortages of skilled workers and venture capital, sug- Encouraging firms to collaborate in innovation has also I gests that necessary horizontal policies were not sidelined characterized the Finnish and Asian approaches-and for by the cluster approach. reasons that are perhaps even more appropriate in the In this sense, Finnish cluster policy, as with Israeli developing-country context. Resolving one market's failures policy, attempted a kind of "informed neutrality." The can often exacerbate another Strengthening intellectual long-run goal is to ensure a well-functioning NIS that property rights reduces the non-appropriability problem but i will allow the emergence of new sectors while at the same can lead to sub-optimal diffusion of new innovations since time boosting those sectors in which the country clearly the owner of a patent charges more than the almost-zero dif- has comparative advantage. fusion cost. Socially optimal innovation is thus at odds with socially optimal diffusion. Research consortia, in theory, Source: Blomstrbm, Kokko, and Sjoholm (2002) internalize the externalities and solve both problems. They may also help alleviate the problem of costly patenting and economies, and avoid duplication. But where firms differ in weak enforcement in developing countries discussed earlier, abilities or capacity, they have also importantly served as which make market intermediation of ideas so expensive. In vehicles for one partner to upgrade by internalizing others' theory, consortia share risks, the fixed costs, benefit from scale learning (see box 7 7) 190 NETWORKS AND NATIONAL INNOVATION SYSTEMS BOX 7 7 Research Consortia in Japan: Did They Work? Japan's experience with cooperative R&D gained attention and Sakakibara (2002) found that what appeared more | among close follower countries, such as Korea and Taiwan important than the level of financing by the government (China), and more recently among academics seeking to was the organization of the consortium and, in particular, evaluate their impact The most celebrated example was whether it was more oriented to basic research. the VLSI (Very Large-Scale Integrated Circuit) project that Surveys of participants found the most important ben- involved all of the major Japanese semiconductor manu- efit to be gaining access to complementary knowledge, a facturers, led to a worldwide leadership in the field after- finding similar to those from Germany, and one of the least wards, and arguably spurred the U.S. formation of the important to be the sharing of fixed costs often empha- analogous SEMATECH and the European EURECA and sized in the literature Firms also benefited moderately ESPRIT projects. from researcher training, a quicker pace of R&D devel- The logic varied among different types of consortia. opment, and an increased awareness of R&D in general. "Technological Research Consortia," introduced in Japan In a global market where product cycles grow shorter over i In 1 961. and broadly patterned on British research associ- time, this is potentially important. ations, were a reaction to the postwar trade liberalization Although studies of the U.S. semiconductor consortium and were intended to promote R&D as a means of increas- have yielded mixed results, Branstetter and Sakakibara ing the productivity of Japanese industries. Others have (1998) and Sakakibara (2001) suggest that the legendary included petroleum-alternative energy development pro- Japanese consortia at least partly deserve their fame: par- |ects (the Sunshine Project), energy conservation initiatives ticipatiorl in a consortium led to a 2-9 percent increase (the Moonlight project), and large-scale projects that the in R&D, and patenting per R&D dollar of between 4 and private sector could not have initiated. 8 percent, amounts which, in the aggregate, can be sub- Government subsidies depended on the mission of each stantial. In addition Sakakibara (2001) finds that a greater consortium. The original TRCs benefited from tax bene- diversity of knowledge in the consortium leads to higher fits, accelerated or instant depreciation of fixed R&D assets, R&D spending, a conclusion also found for the United and discounts of R&D-related property. VLSI was subsi- States. dized 20 percent, the Sunshine and Moonlight projects 100 Somewhat counterintuitively overall, however Sakakibara percent, and the Key Technology Development project 70 (1997) finds that firms did not find the consortia critical percent The most active companies in various consortia to their competitive position. It is doubtful that this find- between 1959 and 1992 have been household names such ing would hold up in Taiwan (China), or even Korea, where as Httachi, Toshiba, and Mitsubishi. Overall, Japanese gov- the consortia appear to have served a nurturing role in ernment spending has been low by U S standards. In a creating a taste for research and defining a mission for study of the impact of consortia on patenting, Branstetter sectors that have since become dominant. The literature is ambiguous on what the actual impact Evaluations of consortia have been few and preliminary (see of research consortia on R&D should be. If the consortium box 7.7). members compete in the same output market, increased Despite the obvious benefits of collaboration, numerous variety or quality of competition may actually lead to lower authors note the apparent difficulty of initiating it, perhaps returns and R&D, a problem presumably ameliorated by because, in practice, some firm or agent needs to bear the an aggressive export strategy. On the other hand, as Cohen coordination costs and this provides a role for government and Levinthal (1989) argued, if the firm's ability to appro- intervention Intel in Costa Rica was able to work with local priate spillovers is a function of the amount of R&D, then universities to bring the curriculum in line with their needs, participation in consortia may lead to higher total R&D. but it is not clear that a group of loosely linked industries 191 CLOSING THE GAP could organize itself to do the same. Partially for these rea- their winning horses. There are the usual attendant risks and sons, in Japan-and in Korea and Taiwan (China), close concerns about government administrative capacity or imitators-government agencies were charged with coor- simply foresight. As Noland and Pack (2002) argue for Chile, dinating industries (particularly in the highly diffuse "It is unlikely that government officials considering export Taiwanese industrial structure), serving as a gateway for the promotion measures would have had agriculture, no less diffusion of external research and interaction with foreign peaches, grapes, apricots, and plums grown in November firms. Not only did government laboratories identify and through March, on the list of potentially profitable export provide original research to nourish an emerging research sectors" (p. 52). The case of Curitiba, Brazil, illustrates the consortium, they set target technologies to catch up to and, particularly troublesome political economy issues. Not only most important, sought to upgrade firms, creating within was the government unable to persuade members of the city's them "a taste for research" (Cho 2001; Sakakibara and Cho advanced IT cluster to cooperate in an effort to work on a 2002; Mathews 2001).11 In fact, they can be seen as encour- software project with Japan, but by playing favorites and aging and enabling firms and consortia of firms to move up fomenting rivalries among supporting institutions, it actu- the technological ladder, not only by providing the basic ally discouraged collaboration.12 ingredients but also providing a focus and mission for emerg- Further, Feser notes that even where governments are ing firms Returning to the earlier discussion of the role of very capable, highly specialized cluster-building initiatives public think tanks, Taiwan's Industrial Technology Research can unintentionally help lock a region into an industrial spe- Institute (ITRI) was founded in 1973 and its laboratories cialization or innovation competency that will eventually served a similar role to those in the United States, Europe, face decline. An economic policy regime with a narrow set and Japan as a core vehicle for R&D collaboration, but also of competencies targeted to a particular cluster is less nimble more specifically focused on rapid adoption of new standards, when economic and technological conditions change, as they products, and processes and rapid diffusion to as many firms must. Furthermore, targeting-and in fact applied cluster as possible, rather than extending the envelope of R&D. The analysis more generally-are heavily data intensive on real initiative for the formation of early alliances came exclusively innovative or entrepreneurial strengths and may be unre- from public agencies, with the private sector taking an alistic in many contexts (Feser and Luger 2002). Hence, in increasingly active role. both the Finnish and Israeli cases, the focus is on ensuring Such consortia are also found in Europe and the United that the NIS in general functions well so that potential States One of the most famous is SEMATECH, a research industries that are not on the government's radar will find consortium of semiconductor manufacturers set up in 1987 the conditions to grow. In short, cluster-type policies cannot by 14 U.S. semiconductor firms with the financial assistance be a substitute for a well-functioning NIS over the long run. of the U.S government, which has been given credit for reviving the industry in the face of Japanese competition Science Parks and Incubators in LAC (Irwin and Klenow 1995) But similar success stories are The literature is skeptical about the public sector's capacity found in the Instituto de Capacitaczin e InvestzgaciMn del Plas- to build innovation clusters from scratch even in advanced tco ydel Caucho in Colombia, which provided the focal point countries (Feser 2002). Again, the canonical example of for technology transfer and development in the revived Brazil's failed informatics industry provides a dramatic exam- Colombian plastics industry. Similarly, the Mexican Uni6n ple. On the other hand, reforming the NIS so that it pro- Nacional Avicola partly concerns itself with raising the qual- vides a fertile and level field for cultivating new ideas is a ity of technological inputs into the production process- long-run process. Science parks and incubators have been again, with important interest in importing foreign employed to help firms and universities overcome informa- technologies (Mayer 2002). In all three cases, the dominant tional and technical barriers to collaborative R&D efforts, private sector presence ensures relevance. and, to some degree, to act as microcosms of a well- Clearly, the Asian experience in fomenting coordination functioning NIS. They are managed spaces, typically on the involved targeting industries, which to some degree makes campus of a university or technology institution, where firms sense-given scarce innovation resources and vast unrealized of varying size and age can operate in proximity to scien- returns to innovation, governments probably should back tific personnel, although increasingly they are virtual. 192 NETWORKS AND NAi IONAL INNOVATION SYSTEMS Science parks establish formal cooperation links among BOX 7 8 universities, research instrtutions, and high-tech enterprises. The Israel Magnet Program Silicon Valley, for example, is an offspring of the Stanford Technological Park. Such institutions can be useful in the *Notwithstanding the rapid growth of the high-tech early stages of development, when firms do not conduct large sector in Israel since the late 1960s, it became clear by amounts of in-house R&D but draw upon a common pool the early 1990s that the country's industrial landscape of scientific personnel for "hand-holding " was fragmented and, with few notable exceptions, Israeli Technological incubators provide a supportive environ- industrial companies were too small to be able to meet ment that offers fledgling entrepreneurs an opportunity to the escalating costs of developing new technologies in develop innovative new ideas and commercialize them. Israel cuttinm-edge fields. Moreover, Israel boasted world- introduced this program in the early 1990s to help Soviet class-research universities, but they operated largely in immigrants who had valuable human capital and ideas but isolation from surrounding industrial developments lacked basically all necessary skills, ranging from basic and needs, and hence the vast economic potential Hebrew to managerial training (Trajtenberg 2002 and embedded, in both the highly qualified academic man- box 7.8). The incubator seeks to increase the entrepreneurs' power and in university research remained largely prospects of raising capital, finding strategic partners, and untapped. developing a self-sustaining business on their own. Incuba- Against this background, in 1993, the Office of the tors represent the first vital stage for starting a business, after Chief Scientist established the Magnet Program to sup- all the initial planning has been done. They may provide a port the formation of consortia of industrial firms and I physical infrastructure for entrepreneurs to set up their first acadermic institutions and develop generic, precompet- office, they may also help with marketing and searching for itive technologies. These consortia are entitled to mul- funding (Agapitova and Holm-Nielsen 2002) tiyear R&D support (usually three to five years) The literature is moderately positive about the impact of consisttng of grants of 66 percent of the total approved incubators. Colombo and Delmastro (2002) find in a survey R&D budget, with no recoupment requirements. The of new technology-based firms that on-incubator firms conisortia must comprise the widest possible group of showed higher growth rates, higher levels of human capi- industiial members operating in the field, together tal, and higher rates of adoption of advanced technologies with Israeli academic institutions doing research in and participation in international technology programs, scientific areas relevant to the technological goals of the higher levels of collaboration with universities, and easier consortia. access to financing. They argue that the science parks are Mindful of possible conflict with antitrust provisions, especially useful in countries such as Italy where the NIS is consortium members must pledge to make the products weak. Mian (1996) finds substantial value added in a study or services resulting from the joint project available to of six University Technology Business Incubators, specifically any interested local party, at prices that do not reflect in terms of university image, laboratories and equipment, the exercise of monopoly power. and student employees. Less conclusively, looking at 273 firms on and off science parks in Sweden, Lindelof and SoU*rce, Tralrenberg (2002) Lofsten (2002) find higher rates of job creation and sales growth, although not profitability-and higher links with universities, although not higher patent rates. economy, an entrepreneurial business culture, a consolidated Cluster innovations, science parks, and incubators all share legal framework, and a highly developed financial system the same caveat: shell institutions cannot substitute for poor As a final caveat, the difficulty of measuring the university or research institution quality, institutions that spillovers associated with innovation clusters should not lead are coherent and consistent with the goal of innovation, and to completely abandoning the basic net present value cal- a well-functioning NIS more generally Silicon Valley, for culations applied to any project or justifying any level of example, was not created through top-down planning, but financing. Despite Embraer's successes, Fajnzylber (2001) was rather emerged spontaneously in the context of an open unable to find any evaluations of the total costs of the project, 193 CLOSING THE GAP inclusive of government subsidies. Though it is not clear B0X79 what the math was, in general, the governments of Japanese BOX 7 9 ' ' ~~~~~~~~~~~~fim4tion as a Vctov of Onnovation and Korea funded only about 20-30 percent of the costs of research consortia. In Taiwan (China), the government con- History is clear on the extraordinary dependence on tributed roughly 30 percent of the costs of the 20 consor- immigrants as innovators and entrepreneurs in Latin tia involving hundreds of firms over two decades (Mathews America. Industrialization in Mexico in the late 19th 2001). Benchmarking cluster performance must necessarily century was almost entirely undertaken by the resident include such calculations. foreigners (Hansen 1971). Using machinery from their homeland, the French started the textile industries in DngiematIfionaD letw ibrks Veracruz and Puebla (Buffington and French 1999), and The final theme of this chapter is the international dimen- foreigners also started Mexico's first iron and steel plant sion of the NIS. Because the vast stock of knowledge is in 1903, the Funtdidora de Fierro y Acero de Monterrey, largely found in the advanced countries, closing the tech- which would build on the region's ore deposits and nology gap requires going beyond the technological trans- anchor its industrial development. Hansen argues that fers through trade and investment, as discussed in chapter 6, there were entrepreneurial spillover effects that drew moving to substantial augmentation of international net- many Mexicans into the capitalist ranks, but the initial works among researchers, entrepreneurs, business service impulse came from foreigners. providers, and firms, as well as institutions such as think Collier and Sater (1996) also note the influence of tanks and universities. It should be pointed out, however, immigrants in introducing new industry and technolo- that given the tacit nature of much knowledge and the gies in Chile. Immigrants set up many of the industrial importance of personal contact, the actual movement of enterprises of the 1860s and 1870s 36 of the 46 dress- people through either short- or long-term migration makers counted in 1854 were French; Americans assumes an importance that is higher than is often installed the flour mills; Americans and British built the acknowledged. In fact, these various types of linkages are j railroads. Loveman (1979, p. 193) notes that the list of extremely complementary and need to be discussed officers and members of the executive committee of together. SOFOFA, the principal organization of industrialists, showed the disproportionate influence of immigrants: Migration and Study Abroad "Only three Spanish surnames accompanied those of Moving individuals among countries has traditionally been the other members of the directorate: Edwards, one of the most effective means of transferring knowledge Subercasseaux, Hillman, Tupper, Tiffou, Mitchell, among them. As box 7.9 shows, the impulse for manufac- Gabler, Lanz, Klein, Muzard, Lyon, Bernstein, Crichton, turing sectors in Chile and Mexico came primarlty from for- Osthous, Stuven" (Maloney 2002). eigners immigrating to those countries. Arguably, the lower - rates of in-migration to Latin America, compared with the United States or Canada, explain to some degree the slower diffusion of new European technologies. the region has more students as a fraction of tertiary enrollees Graduate study in the centers of innovation is effectively in the United States than any other region in 1998. Asia has temporary migration for the same purpose of communicat- far more tertiary-educated workers; however, the logic for ing tacit knowledge. Finland relied on this to the exclusion study in the United States for LAC is much greater than for of FDI as a means of transferring know-how (Myllyntaus any other region, and before its crisis, Asia actually had more 1990)-and in Taiwan (China) by the mid 1980s more than students studying 20 percent of executives had studied either in Japan or the The move toward establishing local graduate programs United States and all employees with graduate degrees in is a double-edged sword. On the one hand, costs of domes- manufacturing had studied abroad (Pack 2000). Compara- tically trained students are perhaps 20 percent of external tive data on LAC are ambiguous Figure 7.10 suggests that training, and returning academics have a ready supply of 194 NETWORKS AND NATIONAL INNOVATION SYSTEMS FIGURE 7 10 Student Migration Rate: 1993-98 Students Abroad/Tertiary Enrollment 0 06 005 - o 04 E19 0 03- 0 02- 00 1 _ 0 Africa Asia Europe Latin America Source Angel-Urdinola and others (2002) graduate students for research assistants who can help start technical and business trends and helped design policies to a local research program. On the other hand, what is lost is take advantage of them. Similarly, the Chinese Institute of the contact with the most recent generation of thinking at Engineers, founded in 1979 by a small group of Taiwanese the frontier, and a certain amount of intellectual inbreed- engineers working in Silicon Valley, organizes an annual sem- ing is possible without very stringent standards at interna- mar in collaboration with their counterpart organization in tional levels Taiwan (China), and provides consultative services to the Local doctoral programs are also seen as a way of avert- Taiwanese government. Private firms have established ing "brain drain," which implies a deadweight loss to the research centers abroad precisely with the goal of industrial developing country after years of public spending on the edu- reconnaissance. cation of these individuals But what has been less well con- Diasporic networks of emigrants also stimulate capital sidered is the importance of emigrants to establishing flows and business opportunities more generally because emi- contacts and serving as the advance guard of the NIS. grants are better placed to evaluate investment opportuni- The last decades of the 20th century have provided some ties and retain contacts in their home countries, helping dramatic examples of the importance of return migration as foreign direct investment sources find trustworthy and com- a stimulus to innovation, as box 7 10 notes Arguably, the petent partners. As Saxenman (1999) observes, "The scarce software industries in Ireland and India and the technology resource in this business environment is the ability to locate sector more generally in Israel would have been far less foreign partners quickly and manage complex business rela- dynamic or even nonexistent were it not for return tionships across cultural and linguistic boundaries. This is migrants. But as important may be keeping individuals particularly a challenge in high tech industries in which abroad. In fact, both Korea and Taiwan (China) leveraged products, markets and technologies are continually being their scientists and engineers abroad very well to bridge the redefined-and where product cycles are routinely shorter knowledge gaps that existed between their countries and the than nine months." Saxenian's view corresponds very much United States. U.S.-based Korean engineers were designated with the Indian IT success story, in which immigrants have 'R&D outposts" who advised the Korean government on not returned to India as much as facilitated outsourcing of 195 CLOSING THE GAP their career but planned eventually to return to Colombia BOX 7 10 Retum iMigration in Israel and Ireland (Angel-Urdinola, Takeno, and Wodon 2002). The Millen- nium Science Initiative in Chile proved successful in attract- In both Israel and Ireland, the impact of recently ing back Chilean world-class researchers and establishing returned migrants on innovation has been dramatic. science nuclei-the core of an incipient knowledge cluster- Carmel and de Fontenay (2001) note that from 1989 in several promising fields (box 7.12) to 1991, Israel experienced a return migration of per- The critical problem seems to be using diasporas most haps the quintessential diaspora, on the order of a mil- efficiently. The Colombian Network of Scientist and Engi- lion immigrants from the former Soviet Union. The neers Abroad (Red Caldas, named after one of Colombia's Soviet Union's emphasis on theoretical sciences made first notable scientists) was established in 1991 when Israel a mathematical superpower by the end of 1991: COLCIENCIAS, the government agency in charge of 5,300 newly arrived scientists registered with the gov- national research, sought to complement the commercial ernment. These immigrants are now heavily represented opening with a higher profile for science and technology. in most technical and engineering groups. Trajtenberg COLCIENCIAS appointed a staff member to create the net- (2001) argues that this inflow is largely responsible for work, which began as an e-list of Web-linked professionals the sharp jump in patenting after 1994. in 21 nodes around the world; a meeting with expatriates Irish software industries benefited from returning in Paris, Madrid, and Mexico; a study of the profile of the expatriates who went abroad during the 1980s and emigrant community; and an ongoing upscale electronic 1990s to meet a chronic U.S. shortfall in IT workers. "chat room" accommodating scientific dialogue. Unfortu- Along with the United Kingdom, Ireland constituted nately, enthusiasm for the ultimate goal of Red Caldas-the the largest European Union (EU) source of emigration realization of joint projects-has waned as the limitations to the United States, with California as a principal des- of this low-budget virtual approach became clear. The expe- tination In the 1990s, however, net immigration has rience of the Bio-2000 project to launch a joint U.S.-Europe- been positive and more than half are returning emi- Colombian effort in the biomedical applications of physics grants. Twenty-five percent of male emigrants with com- 1 offered an example in this regard. Despite efforts on the part pleted tertiary education returned during the 1990s. of the expatriate community to include exceptional foreign More than 40 percent of Irish graduates under the age scientists and line up financing commitments with financ- of 40 have worked abroad for at least one year. ing organizations, COLCIENCIAS' low rate of response and l~ --- ------ ------------ - ---- -- coordination ability led to a radical downscaling of goals over time to several joint studies linking universities in Spain, service development to their homeland. Return migrants Italy, Brazil, and Peru-but, perversely, none from Colombia. have been a source of knowledge for their home countries, The absence of strong accompanying institutional affiliations a source of contacts within their countries for U.S. firms, and has also led to problems whereby individual expatriate sci- links to foreign capital markets, filling in gaps in domestic entists want to participate in the Colombian project but credit markets. (Arora, Gambardella, and Torrissi 2001). effectively have to do it pro bono since the home institu- Figure 7.11 suggests that no LAC country, with the tion's agenda in, say, Sweden is unlikely to coincidentally exception of the islands, has had as much tertiary-educated overlap much with that of a researcher in Colombia 14 migration as Korea and Taiwan (China).'3 Box 7.11 presents the view of a prominent Brazilian commentator, who cites Institutional Linkages the deficient brain drain as a major failing of his country's More successful than networks of individual scientists have NIS. Brazil, perhaps more than most Latin American coun- been international linkages among institutions and firms. tries, has invested massively in sending students to study Numerous modalities appear, but it is again the case that such abroad. Yet measures to attract expatriates of the kind that linkages suffer from the usual market failures affecting Korea and Taiwan (China) employed yielded few returnees- diffusion, and hence may have a role for government inter- most were already back Surveys of Colombians abroad found vention. Korea's Hyundai, Daewoo, and Samsung have estab- that 75 percent saw their foreign work as only a phase in lished think tanks as outposts in the centers of their 196 NEIWORKS AND NATIONAL INNOVATION SYSTEMS FIGURE 7 11 Brain Circulation: Migration/Population with Tertiary Education Brain Circulation: Migration/Population with Tertiary Education 30 0 25 0 20 0- 150- L 10 0 - 5 0 0 0 Source Carrington and Derragiache (1998) respective industries'5 and Finland's Nokia has 55 research a university research center in Paris joined forces with another centers in 15 countries (Blomstrom, Kokko, and Sjoholm in Cali, jointly funded by COLCIENCIAS and the French gov- 2002). Other governments have explicitly financed the exter- ernment to gain know-how in the production of robotics Even nal antenna. In Taiwan (China), whose less concentrated in economics the tremendous links developed over the last industrial structure makes coordination more difficult, the decades, involving both public and private universities and government research agency actively searched for new tech- particularly with the United States, have almost certainly con- nologies and served as a clearinghouse for interactions with tributed to narrowing the parameters within which the debate large foreign firms. In Israel's bilateral agreement with the over economic policy occurs and hence helping to avoid some United States, both governments subsidized collaborative of the more adventurous experiments of the past. research among firms. However structured, the clear lesson is that the best strat- Several examples suggest the importance of connections to egy to reach the knowledge frontier is to actively engage it foreign universities and think tanks. Chile's legendary revo- rather than turn inward to learn by doing before opening lution in fruit production traces its roots to a 10-year agree- to the world. One illustration is offered by the Costa Rican ment signed in 1965 between the University of California, and Brazilian strategies in the "informatics" area (box 7.13). Davis and the public University of Chile, the agreement is As discussed above, in Asia public laboratories working credited with helping to establish a first-rate department in within consortia of private firms served a critical role as anten- fruit-related sciences, with spillovers to other agencies and um- nas for foreign ideas in technology. In Latin America, one of versities (Jarvis 1992) This pairing highlights how foreign the most successful parallels is Fundacz6n Chile, originally a institutions may be more appropriate than other potential joint effort between the Chilean government and the U.S. firm national or Latin American partners: the ecosystems of the cen- ITT, but now largely autonomous. Fundaci6n Chile uses four tral valleys in California and Chile share many similar char- main techniques in its technology transfer and dissemination acteristics When Colombia's plastics industry sought to work. (1) It creates innovative enterprises, almost always in modernize, it sent graduate students to the "frontier" insti- association with companies or individuals; (2) It develops, tutions in the field in Germany and its industry-wide think adapts, and sells technologies to clients in the productive and tank maintains active connections, largely funded by foreign public sectors, both domestically and abroad, (3) It fosters aid One of Red Caldas more successful projects occurred when institutional innovations and incorporates new transfer 197 CLOSING THE GAP ii BOX 711 { The creation of "demonstration" companies by Fundaci6n 7Deficien rain Drain in bazii Chile has undergone successes and failures, but overall has proved effective as a method for disseminating new tech- A prominent Brazilian social scientist, Claudio De nologies. The companies are transferred to the private sector Moura Castro, argued in the weekly Veja that despite the once the technologies have been proven in practice and their evidence that the vast majority of Brazilians returLned, 1 economic profitability has been established The organiza- Brazil could not be complacent: tion has created more than 40 enterprises in numerous sec- "So then, does this mean that there is no problem of ... tors, of which about 30 have been sold, with the cost of brain drami Unfortunately, this is not the case. We have resources invested recovered and then reinvested in new pro- a grave problem. Science is internationalized. Firms need jects. One of the most successful cases, which exhibits many to operate globally (at the very least to export). The best elements of the successful development of a knowledge clus- business is undertaken between actors who have a global j ter, is that of the salmon industry, which in a period of 10 view. We are at a double disadvantage. We still have years grew to become a dynamic export sector ancient vices of self-sufficiency and isolation (which a visit to Disney World doesn't fix). The second defect is coKduCS5cn that we don't have a diaspora to function as Brazil's One of the abiding mysteries in development economics has window on the world, like the thousands of Koreans been why poor countries have not aggressively exploited the who, from the outside, provide their country with the immense global stock of knowledge to accelerate growth. I technologies they need. Increasingly the literature focuses on shortcomings in "There are almost no Brazilians in the exterior in uni- national innovative capacity, and the discussion here of the versities, international organizations or the business NIS draws together the various themes developed through world. For Brazil, this diaspora would be of vital impor- the report. High levels of human capital and exposure to tance to generate collaboration in research, to get and foreign technologies, for instance through trade or FDI, are give invitations for seminars, to receive the latest paper, critical not only in their own right, but also vitally in how to identify consulting work, and to vaccinate our science they complement each other. As countries seek to accelerate against intellectual isolation. Korea and Taiwan more or the pace of technological progress, ensuring that the right I less are equal to Brazil in scientific production, but many human capital is available and used effectively becomes central. of their authors are abroad, airing and internatlonaliz- The evidence is clear that the higher-performing coun- I ing their research. tries that were able to make a transition to full partners in "For this reason, we have a problem of brain drain, global innovation-Israel, Finland, and Korea, for example- but it's the opposite. Our drain is insufficient and neg- have dramatically increased their level of human capital, their ligible. We produce 5,000 doctors each year. We need investments in innovation, and the quality of their policies to lose some of these so that, this way we can irrigate in a concerted fashion. In addition to getting the basics right our science and technology with the knowledge of the in terms of plugging into the global knowledge economy, great production centers abroad. We need to lose engi- gaining exposure to international competition and ideas, pro- neers and businessmen so that, later, they return with moting openness to interactions with foreign capital, experience that you can't learn in school. The big drama fomenting ICT, establishing property rights, and address- of the Brazilian brain drain is its absence" ing credit market deficiencies, they also came to terms with two fundamental issues affecting every country following in Source: De Moura Castro, Claudio Veja. November 28, 2001 "O drama their footsteps da evasao de cerebros." First, knowledge as a commodity is characterized by extra- ordinary market failures and hence the market will not gen- erate the optimal level of innovation. To repeat, serious mechanisms; and (4) It captures and disseminates technolo- analysts argue that the United States should probably increase gies to multiple users (as a technological antenna) through its R&D by a factor of 4 and we have offered evidence in this seminars, specialized magazines, project assistance, and so on. chapter that LAC's level is also far below optimal. Further, 198 NETWORKS AND NATIONAL INNOVATION SYSTEMS BOX 7 12 Catalyzing an Innovation Cluster. The Millennium Science Initiative in Chile The Millennium Institute for Applied and Fundamental quality (Science Nuclei). The idea was to provide adequate Biology (Instituto Milemizo de Biologia Fundamental y levels of medium-term funding to enable the best scien- Aplicada, MIFAB) was established in 2000 as one of three tists to return to Chile in a situation in which they could Millennium Science Institutes in Chile, and it has already compete with their international peers, perform pioneer- made remarkable contributions in genome research and in ing scientific research, and provide training to graduate research on infectious biological agents. During 2001 and postgraduate fellows. The selection of beneficiaries was MIFAB analyzed and patented the DNA sequence of a bac- placed entirely in the hands of a group of international teria (Pisczrickettsia salmonts) that causes infections of the academics headed by Dr Phillip Griffiths of the Prince- Chilean salmon. This is critical for identifying and diag- ton Institute of Advanced Study to ensure quality and nosing infectious diseases, but particularly for developing transparency. vaccines Important advances have also been made on diag- Several facto4is merit note First, the MSI catalyzed this nostic tools and preventative measures aganst the bac- emerging clustei from a primordial soup of world-class sci- terium Alexandrium catenella (red tide). MIFAB has also entists already working in fields strongly relevant to Chile's been studying a vineyard virus (Vitis vinmfera), and comparative advahtage. The necessary high quality is guar- progress has been made toward diagnosis and improvement anteed by the coirnpetitive selection process. Second, the of viral resistance in the gene pool of the grape. growing interest, and potential investment by the private The MIFAB has 11 senior scientists, 6 associate scien- sector-mnost obviously the critical Chilean wine and tists, 12 post-docs, 39 Ph.D. students, 42 master's degree salmon sectors-in the output of the Millennium Science students, 10 technicians, and administrative personnel- Institutes, attunes their research agenda to the needs of the a total of about 120 people. Its existence is due to the Chilean economy over the long run. The MSI effort there- efforts of the Millennium Science Initiative (MSI), a Joint fore played strongly to Chile's strengths. Third, the teams initiative by the government and the World Bank. Under and their leadership are guaranteed autonomy to pursue the oversight of a high-level consultative body with inter- science at global best-practice levels with minimal inter- national participation, $US 15 million were partly allo- ference of other agendas. cated to ftinding three world-class research groups (Science Institutes) and 10 emerging groups of high Soarce. Hol_n-Nielsen and Norsworthy (2002) the institutions evolved to resolve these market failures- Both considerations suggest an integrated approach and universities, government laboratories, and intellectual prop- a coordinated, even leadership role for government. At a min- erty rights-lie, by definition, outside the market and hence imum the state needs to ensure a consistent and coherent are not coordinated by the price mechanism set of incentives to ensure that the institutions created to Second, a critical aspect of the process of development is address market failures collaborate fruitfully with firms. But, that firms, and the country as a whole, "learn to learn." In in the highly successful countries, governments have not particular, increasing the technological absorptive capacity been shy about stimulating private R&D and financing it of the firm has required a supportive set of policies and insti- where it has broad spillovers The U.S government remains turions ranging from well-designed fiscal incentives and sub- the largest financer of research and development in the world sidies to the active promotion of collaboration through (Jaffe and Lerner 2001) It is also the case that virtually no incubators, technological parks, clusters, to the creation and country has managed the big innovation push without gov- coordination and direction of industrial consorria that share ernment playing a leadership role. the costs and risks of R&D and serve as learning laborato- Not all countries are at a stage where undertaking such ries for less advanced firms, and to promoting antennas for policies is feasible. It probably does not make sense to identifying ideas from abroad. contemplate broad government financing of R&D if the 199 CLOSING THE GAP BOX 7 13 The Perils of Ontellectual Autarky, the Rewards of Trade Costa Rica and Brazil offer two alternative paths to estab- years developed "Freedom," which had almost complete lishing high-tech industries. The success of Costa Rica's code recovery rates compared with the next best program's software industries rests on two first-class universities rate of 70 percent. Venture capital from Intel aided fur- whose roots were firmly in the exterior. Two brothers, ther development, and Microsoft eventually signed an Alberto and Daniel Cafias, received their doctorates from alliance with ArtinSoft to include Freedom as part of its i the University of Waterloo and the University of Texas at software development package for e-business and the Austin and returned to Costa Rica to set up the Instituto Internet (Rodriguez Clare 2003). Tecnol6g:co de Costa Rica (ITCR). Then they both went back As the inevitable contrast, we can invoke the canoni- to the United States, where they hold professorships at cal case of Brazil's experience with the informatics indus- Wake Forest and Pensacola-now foreign outposts of Costa try, begun in 1976 and managed by the powerful Special Rica's NIS. The desire to compete with the ITCR allegedly Secretary for Informatics. As Dahlman and Frischtak inspired the Universidad de Costa Rica to send a small team (1993) note, the attempt to develop domestic firms' tech- I of mathematicians to study toward graduate degrees in the nological capabilities while blocking foreign investment and Umted States and Europe. One of the graduates of the influence led to the development of an uncompetitive 1 ITCR's first class also studied for a doctorate abroad, then industry with adverse effects on downstream user indus- returned to head the Center for Computer Research at the tries forced to use second-class inputs. The attempt to ITCR and made breakthroughs in automaring the process achieve "autonomy" and displace multinational firms from of migrating software from one language to another- key industrial segments deterred the attraction of best- an important and lucrative advance, given the stock of practice technology. The combination of weak domestic valuable code with applications in other languages. He and rechnological efforts and restricted access to the best of | several students received seed money from Oracle of Cen- foreign technologies impeded the modernization of tral America to set up a firm, ArtinSoft, and within seven Brazilian firms. economy remains closed, if basic institutional integrity is Though arguments for traditional mechanistic "industrial in doubt, or as the report has stressed, if the required human policy" have largely been discredited, an active and efficient capital is absent. That said, the successful countries have "innovation policy" is called for to provide the necessary consistently taken an active approach to integration in the innovation and education-related complements to previous world economy by upgrading the learning capacity of firms, reforms. This poses a challenging policy agenda over the next selectively financing R&D, encouraging the licensing of for- decades. But as this report has shown, Latin America lags eign technologies, and extending intellectual property rights in almost every dimension of educational and technological and ICT infrastructure-in short, progressively deepening achievement. As a region, to rephrase Pasteur's quote, our and tuning up their NIS rather than passively waiting for collective mind is not yet prepared to take advantage of the MNCs or imports to transfer technology Thus engagement unpredictable technological opportunities that the new mil- in the long process of undertaking the necessary institutional lennia will offer us. reforms needs to start early in the development process Fur- ther, as chapter 6 showed, many of our countries show levels hDnneX of income and education similar to those of the innovation superstars when they began their big innovation push. A How Much Should LAC Be Spending on R&D? concerted effort to raise the level of especially private R&D As table 7A.1 shows, it is common in the literature to find should start now using the instruments shown to have pos- social rates of return to R&D of 30 to 100 percent Lederman itive impacts in the industrialized world. and Maloney (2003b) estimate comparable rates of return 200 NETWORKS AND NATIONAL INNOVATION SYSTEMS TABLE 7A I regressed on initial income Investment/GDP and labor growth Social Rates of Return to R&D R&D/GDP. Additionally, R&D/GDP and Investment/GDP RETURN SOCIAL OPTIMAL have been interacted with the initial level of GDP per capita (OWN) SPIUIOVERS RETURN R&D/ACTUAL in an attempt to study the way the returns to different input Sveikauskas 1981 17 2 4 accumulation vary as the countries become richer As Griliches 1994 30 4 pointed out in the text, the returns to R&D and licensing Griliches and Lichtenberg 1984 34 4 9 are high relative to the investment in physical capital. The Terleckyj 1980 25 82 1 07 117 negative signs on these iteractive terms idicates the Scherer 1982 29 74 1 03 10 6 Griliches and decreasing nature of the returns to R&D and physical cap- Lichtenberg 1984 30 41 71 5 9 ital. These results are robust to alternative specifications Jones and Williams 1998 35 5 0 icluding additional regressors Lederman and Maloney 2003b 1 02 2-10 Endnotes Source Jones and Willilams (1998), Authors Edoe I Consider, for example, the well-known case of medicine to fight AIDS Patenting, by guaranteeing property tights, does encourage the for both R&D and licensing using a panel of emerging and development of the drugs and facilitates the diffusion of the architec- industrialized countries ture of the drugs themselves But it does substantially raise the cost of They begin with a simple production function the innovation to the user and thus restricts the benefits that could accrue a 1-cs A 2. Our thanks to Daniel Lederman for bringing this quote to Y = KAL S our attention http llwww lucidcafe com/lucidcafe/library/95dec/ where S is the stock of accumulated R&D. This can be pasteur html rewritten as 3. Maloney (2002) argues that a decrease in the relative capacity of LAC over the last century left it dependent on the advanced coun- AInY = r. - +(1-a)AInL tries for technology The Chilean copper industry, for example, once Y Y dominatecl the copper market, yet by the beginning of the 20th cen- tury, it had virtually disappeared, largely because it had not made the using the fact that rAln(X) = ri(x) Incremental advances in technology that were required to stay at the usingn X x y frontier 4 See Da Motra e Albuquerque (1999), Dahlman and Frischtak where rx is the rate of return on factor X, dx is the share of (1993) for Brazil, Bazdresch (2002) for Mexico, Chudnovsky (1999) investment in X over Y and D the output elasticity of X from for Argentina, and Mullin (2001) for Chile a standard log production function If we remove the influ- 5 In fact, there is little literature on the subject-the OECD's ence of physical factors to get TFP, the social rate of return (2002a) recent Benchmarkzng Industry-Science Relationships offers some to R&D IS pilot studies of benchmarking in France, the United Kingdom, and Japan Given how recent this literature is in the OECD, it is not sur- r5 = A In TFPls prising that there would be no work on LAC See also Niosi (2002) 6 Benchmarking from Lederman and Maloney (2003b) Condi- where s is the share of R&D spending in income Follow- cioning variables for patents are GDP, number of workers and their ing Jones and Williams (1998), the optimal level of R&D squared terms, year dummies, exports to the United States, and nat- expenditure occurs where r = r, the real interest rate. So, the ural resource endowments e o 7. Patents by nonresidents doubled from 1,955 in 1990 to 4,012 ratio of the optimal level of R&D investment to actual R&D in 1996 However, patents by local residents fell from 1,000 per year investment along a balanced growth path can be expressed in the 1980s and 900 in the early 1990s to less than 700 in 1994 and as the ratio of the social rate of return to R&D over the real 1995 Lederman and Maloney (2003a) similarly argue that NAFTA interest rate is not enough to put Mexico on a dynamic growth path s r 8 Lederman and Maloney (2003c) find using the Systems GMM -=- panel estimator that not only does Leamer's measure of net exports of natural resource-intensive products per worker correlate positively with A GMM system estimation has been employed using five- growth, but Sachs and Warner's measure of NR exports/GDP does as year averages ofthe variables from 1975 to 2000. Figure 7.5 has well if Singapore and Trinidad and Tobago are not substituted with been constructed using column one where Growth of GDP is Net NR/GDP Lederman and Xu (2001) also show that countries with 201 CLOSING THE GAP a high knowledge index were more likely to exporr forestry products helped make Mexico one of the primary exporters of technical than average knowledge on the continent and to occupy the vice presidency of 9 Preliminary work with relatively sparse panel data suggests that the World Mining Association at the turn of the nineteenth cen- within countries, the ratio of productive sector to non productive sector tury (See Maloney 2002 ) R&D first rises with development, but then the very rich countries see 11. See also Leipziger and Thomas (2000), Dahlman and again a relative rise in the non-productive sector component This would Sananikone (2000), Kihman and Leipziger (2000), Kim (1993) be consistent with the public sector trying to kick start the private sector 12 Quale (2001), cited in the World Economic Forum, Global by fostering public research institutes, the private sector coming in, Competitiveness Report (2002) but then at high levels of income near the innovation frontier, very basic 13 This is an extremely crude measure that is effectively the science performed in universities or think tanks again increases number of tertiary-educated individuals over 25 years of age registered 10 In Latin America, universities have also historically anchored in the U S census who report having been born in another country many of the growth miracles of the region The Antzoquza Escuela de Some ad)ustment is made for the possibility that these "immigrants" Minas in Colombia was founded to serve the mining sector, but grew are in fact presently studying to provide talent for emerging industry in Colombia's most dynamic 14 This section based on Angel-Urdinola, Takeno, and Wodon manufacturing regions (see, among others, Safford 1976) In Brazil, (2002). Baer (1969) sees the critical event for the development of the native 15 Numerous successful innovating firms in emerging economies steel industry as the foundation in 1879 of the Escola de Minas at Ouro have physically located part of their research operations in the centers Preto, Minas Gerais, which led to the establishment of the first new of knowledge generation. Korea's LG has laboratories in Tokyo, blast furnace since the failures of the beginning of the century Grad- Sunnyvale, Chicago, Germany, and Ireland; Samsung in San Jose, uates of the Escola de Engenharia do Exercoto, established in 1930, Maryland, Boston, Tokyo, Osaka, Sendai, London, Frankfurt, and would lead the steel industry as it developed through the 1960s In Moscow; Daewoo in France and Russia, Hyundai in Michigan, San Mexico, the precursor to the Unoversodad Nacaonal, the Real Seminarzo Jose, Frankfurt, Singapore, and Taipei Samsung, LG, and Hyundai de Mineria, was founded in 1792 and taught higher mathematics, have all bought firms in Silicon Valley to monitor advances there physics, chemistry, topography, dynamics, and hydraulics, which (Forbes and Wield 2000). 202 Bibliography The word processed describes informally reproduced works Airken, Brian, Gordon Hanson, and Ann Harrison 1997 that may not be commonly available through libraries. "Spillovers, Foreign Investment, and Export Behavior." Journal of International Economics 43(1/2): 103-32. Abelmann, Charles, Lee Klan Chang, and Pinchuda Aitken, Brian, and Ann Harrison. 1999. "Do Domestic Tinakorn Na Aydhaya 2001 "Thailand. Secondary Firms Benefit from Direct Foreign Investment? Evi- Education for Employment." Report #22660-TH dence from Venezuela " American Economic Review 89(3) World Bank, Washington, D.C. 605-18 Acemoglu, Daron 2000. "Technical Change, Inequality, Albrecht, Douglas, and Adrian Ziderman 1993. "Student and the Labor Market." National Bureau of Economic Loans' An Effective Instrument for Cost Recovery in Research Working Paper 7800, Cambridge, Mass. Higher Education " World Bank Research Observer 8(1)' 2001. "Directed Technical Change." National 71-90. Bureau of Economic Research Working Paper 8287, Altbach, Philip G, Cheong Siew Yoong, Margaret S Cambridge, Mass. Gremli, and Hui-Ling Pan 1989. Scientific Development Acemoglu, Daron, and Joshua Angrist. 1999 "How Large and Higher Eduication: The Case of Newly Industrialized Are the Social Returns to Education? Evidence from Nations. New York' Praeger Compulsory Schooling Laws." Department of Econom- Andersen, Esben S , Bent Dalum, Bjorn Jonson, and ics Working Paper 99/30, Massachusetts Institute of Bengt-Ake Lundvall. 2002 "National Systems of Technology, Cambridge, Mass Production, Innovations and Competence Building" Acemoglu, Daron, and Fabrizio Zilbotti. 2001. "Produc- Research Policy 31(2). 213-31 tivity Differences " QuiarterlyJournal of Economics 116(2): Angel-Urdinola, Diego, Taizo Takeno, and Quentin 563-606 Wodon. 2002. "Policy Interventions for Maximizing Acemoglu, Daron, Philippe Aghlon, and Fabrizio Zili- Brain Gain: Framework and Case Studies." Processed botti. 2002. "Distance to Frontier, Selection, and Eco- Angrist, Joshua David, Eric Bettinger, Erik Bloom, Eliza- nomic Growth." National Bureau of Economic Research beth M King, and Michael Kremer 2002. "Vouchers Working Paper 9066, Cambridge, Mass. for Private Schooling in Colombia: Evidence from a Adams, James. 2001. "Comparative Localization of Acade- Randomized Natural Experiment." American Economic mic and Industrial Spillovers " National Bureau of Eco- Review 92(5): 1535-58. nomic Research Working Paper 8292, Cambridge, Arocena, Rodrigo, and Judith Sutz. 2001. "Changing Mass Knowledge Production and Latin American Universi- Agapitova, Natalia, and Lauritz Holm-Nielsen. 2002. ties " Research Policy (30)8: 1221-34. "Chile-Science, e-Science, Technology and Innova- Arora, Ashish, Alfonso Ganbardella, and Salvatore Torrisi. tion." World Bank, Washington, D.C. Processed. 2001. "In the Footsteps of Silicon Valley? Indian and Aghion, Philippe, and Peter Howitt 1992. "A Model of Iris Software in the International Division of Labour." Growth Through Creative Destruction." Econometrica Stanford Institute for Economic Policy Research, Stan- (March)' 323-51. ford, Calif. 203 CLOSING THE GAP Attanasio, Orazio, Pinelopi K. Goldberg, and Nina Pavc- Incentives and Capabilities. Washington, D.C.: World nik 2002 "Trade Reforms and Income Inequality in Bank. Colombia." Paper prepared for the 2002 IMF Confer- Baumol, William J , and J Gregory Sidak. 1994 Toward ence on Macroeconomic Policies and Poverty Reduc- Competition in Local Telephony. Cambridge, Mass.: MIT tion, March 14-15, Washington, D.C. Press. Autor, David H., Lawrence F. Katz, and Alan B. Krueger. Bazdresch, Carlos 2002. "Consideraciones generales sobre 1998. "Computing Inequality: Have Computers el sistema mexicano de clencia y tecnologsa." CONA- Changed the Labor Marketr" Quarterly Journal of Eco- CyT (National Council for Science and Technology of nomics 113(4): 1169-1213. Mexico), Mexico, D.F. Processed. Baer, Werner. 1969. The Development of the Brazilian Steel Behrman, Jere, Nancy Birdsall, and Miguel Szekely. 2001. Industry Nashville, Tenn Vanderbilt University Press. "Economic Policy and Wage Differentials in Latin Balasubramanyam, V. N., M. A. Salisu, and D. Sapsford. America." Washington, D.C.: Inter-American Develop- 1996 "Foreign Direct Investment and Growth in EP ment Bank. Processed. and IS Countries." EconomicJournal 106(434) 92-105. Behrman, Jere, Piyali Sengupta, and Petra Todd. 2001. Barro, Robert J., and Jong-Wha Lee 1996 "International "Progressing Through PROGRESA: An Impact Assess- Measures of Schooling Years and Schooling Quality." ment of a School Subsidy Experiment." Processed. American Economic Association Papers and Proceedings: Berman, Eli, and Stephen Machin. 2000. "Skill-Biased Economic Reform and Growth 86(2). 218-23. Technology Transfer around the World." Oxford Reviewv .2001. "International Data on Educational Attain- of Economic Policy 16(3)' 12-22. ment: Updates and Implications." Oxford Economtc Papers Berman, Eli, John Bound, and Zvi Griliches. 1994. 3(2001): 541-63. "Changes in the Demand for Skilled Labor within U.S. 2002 "International Measures of Schooling Years Manufacturing: Evidence from the Annual Survey of and Schooling Quality" [data set] Available at http:// Manufactures." Quarterly Journal of Economics 109(2)' www.worldbank.org/research/growth/ddbarle2.htm. 367-97. Basant, R., and B. Fikkert. 1996 "The Effects of R&D, Berman, Eli, John Bound, and Stephen Machin. 1998. Foreign Technology Purchase, and International and "Implications of Skill-Biased Technological Change: Domestic Spillovers on Productivity in Indian Firms." International Evidence." Quarterly Journal of Economics The Review of Economics and Statistics 78(2): 187-99. 113(4): 1245-79. Batra, Geeta. 1999. "Skills Upgrading and Competitive- Bernstein, J I. 1998. "Inter-Industry and U S. R&D ness in Guatemala." World Bank, Private Sector Advi- Spillovers, Canadian Industrial Production and Produc- sory Services, Washington, D.C. tivity Growth." Carleton University and NBER. Work- . 2000. "Private Sector Training and Competitive- ing Paper 19. National Bureau of Economic Research, ness in Nicaragua." World Bank, Private Sector Advi- Cambridge, Mass. sory Services, Washington, D.C. Bishop, John H., Ferran Mane, and Michael Bishop Forth- 2001. "Skills Upgrading, Technology and Produc- coming. "Secondary Education in the United States' tivity: Evidence from the WBES." Working Paper, Private What Can Others Learn from Our Mistakes?" Sector Advisory Services, World Bank, Washington, D.C. Processed. 2002. "Training, Technology, and Firm-Level Black, Sandra E., and Lisa M. Lynch. 1996 "Human- Competitiveness-Evidence from the World Business Capital Investments and Productivity." American Economic Environment Survey for Latin America and the Association Papers and Proceedings: Technology, Human Caribbean " Background paper, World Bank, Washing- Capital, and the Wage Structure 86(2)' 218-23. ton, D C. Processed. Blom, Andreas, and Lauritz Holm-Nielsen. 2002. "The Batra, Geeta, and Hong Tan. 2002. "Upgrading Workforce World Bank in Tertiary Education in Latin America and Skills to Create High-Performing Firms." In Ijaz Nabi the Caribbean." Information Note, World Bank, Wash- and Manjula Luthria, eds. Building Competitive Firms ington, D C. 204 BIBLIOGRAPHY Blom, Andreas, and Vicente Paqueo. 2001. "Summary of Industrial Policy." Journal of Industrial Economics 0022- Evaluation of the Impact of Student Aid in the State of 1821, No. 2 Sonora, Mexico " World Bank, Washington, D C. 2002. "When Do Research Consortia Work Well Processed. and Why? Evidence from Japanese Panel Data" The Forthcoming. "Do Student Loans Increase Enroll- American Economic Review, March 2002. ment' Evidence from Sonora, Mexico" World Bank Bresnahan, Timothy, Alfonso Gambardella, AnnaLee Sax- Working Paper, World Bank, Washington, D C. enian, and Scott Wallsten 2001. "'Old Economy' Blom, Andreas, and Wesley Yin. 2002. "Higher Education Inputs for 'New Economy' Outcomes: Cluster Forma- in Latin America- Unmet Demand in a Segmented Mar- tion in the New Silicon Valleys." Working Paper, Stan- ket " World Bank, Washington, D.C. Processed. ford Institute for Economic Policy Research. Stanford, Blomstrbm, Magnus, Arn Kokko, and Fredrik Sjoholm. Calif 2002 "Growth and Innovation Policies for a Knowl- Brunner, Jose Joaquin. 2001. "Chile: Informe e Indice edge Economy. Experiences from Finland, Sweden, and Sobre Capacidad Tecnol6gica." Universidad Adolfo Singapore." Background Paper for the LAC flagship Ibafiez, Santiago de Chile. Processed. report, World Bank, Washington, D C. Buffingron, Robert M., and William French 1999. "The Blonigen, B A. 1999. "In Search of Substitution between Culture of Modernity." In Michael C. Meyer and Foreign Production and Exports " Working Paper William Beezley, eds , The Oxford History of Mexico. W7154. National Bureau of Economic Research, Cam- Oxford: Oxford University Press bridge, Mass Calder6n, C6sar, Pablo Fajnzylber, and Norman Loayza Blyde, Juan 2001. "Trade and Technology Diffusion in 2002 "Economic Growth in Latin America and the Latin America." Washington, D C.: Inter-American Caribbean: Stylized Facts, Explanations, and Forecasts Development Bank. World Bank, Washington, D.C Processed. Borensztein, E.,J. De Gregorio, and J-W. Lee. 1998. "How Callart, Maria Antonia. 2001 "Job Training in Latin Does Foreign Direct Investment Affect Economic America: Past, Present and Future " Paper presented at Growth?" Journal of International Economics 45(1). the UNESCO Seminar on Prospects for Education in 115-35. Latin America and the Caribbean, Santiago, August Bortagaray, Isabel, and Scott Tiffin 2000. "Innovation 23-25. Clusters in Latin America." Fourth International Con- Card, David, and John E. DiNardo. 2002 "Skill Biased ference of Technology Policy and Innovation, Curitiba, Technological Change and Rising Wage Inequality Brazil Some Problems and Puzzles." National Bureau of Eco- Bosch Mossi, Mariano, Daniel Lederman, and William F nomic Research Working Paper 8769, Cambridge, Mass. Maloney. 2003. "Patenting and R&D Efficiency." Office Card, David, and Alan B. Krueger. 1996. "School of the Chief Economist for Latin America and the Resources and Student Outcomes: An Overview of the Carribean, World Bank, Washington, D.C Processed. Literature and New Evidence from North and South Braga. H., and Matesco, V 1989. "Desempenho Techn6- Carolina "Journal of Economic Perspectives 10(4): 31-50. logico da Industrial Brasileira: uma Analise Explorato- Card, David, and Thomas Lemieux. 2001 "Can Falling na" Textos Para Discussao Interna No 162: Rio de Supply Explain the Rising Return to College for Janeiro. INPES/IPEA Younger Men? A Cohort-Based Analysis." Quarterly Bresnahan, Timothy F., Erik Brynjolfsson, and Lorin M. Journal of Economics 116(2)' 705-46. Hitt 2002. "Information Technology, Workplace Orga- Carmel, Erran, and Catherine de Fontenay. 2001. "Israel's nization, and the Demand for Skilled Labor. Firm-Level Silicon Wadi: The Forces behind Cluster Formation." Evidence " Quarterly Journal of Economics 117(1)' SIEPR Discussion Paper No. 00-40. Stanford Institute 339-76 for Economic Policy Research, Stanford, Calif Branstetter, Lee, and Mariko Sakakibara. 1998 "Japanese Carnoy, Martin 2002. "Is Latin American Education Research Consortia: A Microeconometric Analysis of Preparing Its Workforce for 21st Century Economies?" 205 CLOSING THE GAP Background paper for the LAC flagship report, World Colombo, Massimo G., and Marco Delmastro 2002. "How Bank Washington, D C. Effective Are Technology Incubators? Evidence from Carrington, William, and Enrica Detgragiache. 1998 Italy." Research Policy 31(7): 1103-22. "How Big Is the Brain Drain?" International Monetary CONACyT (National Council for Science and Technology Fund Working Paper WP/98/102. Washington, D C. of Mexico). 2001 Annual Report 2001. Mexico, D.F. Caselli, Francesco, and Wilbur John Coleman II. 2000. http://www.conacyt.mx/ "The World Technology Frontier." National Bureau of Contreras, Dante 2001. "Evaluating a Voucher System in Economic Research Working Paper 7904, Cambridge, Chile: Individual, Family and School Characteristics." Mass. Universidad de Chile/Yale University. Processed. . 2001. "Cross-Country Technology Diffusion. The Corbucci, Paulo Roberto. 2000. "As Universidades Fed- Case of Computers." American Economic Association Papers erais: Gastos Desempenho, Eficiencia e Produtividade." and Proceeding: Technology, Education, and Economic Growth Texto para discussao No. 752, IPEA (Instituto de 91(2): 328-35 Pesquisa Econ6mica Aplicada). Brasilia, Brazil CEPAL (Economic Commission for Latin America and the Corseuil, Carlos Henrique, and Marc-Andreas Muendler. Caribbean). 2001 Informe Anual Santiago, Chile 2002. "Skills, Innovations and Wages in Brazilian Man- Cetto, Ana Maria, and H Vessuri. 1998. "Latin America ufacturing." Background paper for the LAC flagship and the Caribbean." In World Science Report UNESCO, report 2002, World Bank, Washington, D.C. Paris. Cunningham, Wendy. 2001. "Gender in the Maquila Sec- Cho, Dong-Sung, and Mariko Sakakibara. 2002. "Cooper- tot in Mexico and the Dominican Republic." Unpub- ative R&D in Japan and Korea A Comparison of Indus- lished note, LCSPG/World Bank, Washington, D.C. "srPi 7-. 2002. "The Well-Being of Latin America's Men trial Policy" Research Policy 673-92 and Women 1990-2000" World Bank, Washington, Chudnovsky, Daniel. 1999. "Science and Technology Pol- ' icy and the National Innovation System in Argentina." C .nProcessed. Cunningham, Wendy, and Willilam F. Maloney. 2000. CEPAL Review 67: 157-76. "Child Labor, School Attendance, and Credit Con- Ciecon (Centro de Investigaci6n sobre Economfa Laboral y straints in Argentina and Mexico: A Dynamic Panel Gesti6n del Conocimiento). 2001. "The Incidence of Approach." World Bank, Washington, D C Processed TrainingnMeo " Background Paper for Mexico: Dahlman, Carl J., and Jean Eric Aubert. 2001. China and Training Mechanisms. Washington, D.C. World Bank the Knowledge Economy. Seizing the 21st Century" World CINTERFOR/OIT. 2001 "Modernization in Vocational Bank, Washington, D.C. Education and Training in the Latin America and the Dahlman, Carl J., and Claudio R. Frischtak. 1993 Caribbean Region." Paper presented at the Regional "National Systems Supporting Technical Advance in Workshop on the Modernization of VET in LAC, Rio de Industry: The Brazilian Experience." In R. R. Nelson, Janeiro, December 2000 ed., National Innovation Systems. New York: Oxford Uni- Coe, D. T., and E Helpman 1995. "International R&D versity Press. Spillovers." European Economic Review (39) 859-87. Dahlman, Carl J., and Ousa Sananikone. 1997. "Taiwan, Coe, David, Elhanan Helpman, and Alexander Hoffmeis- China Policies and Institutions for Rapid Growth." In ter 1997. "North-South R&D Spillovers." Economic Danny Leipzinger, ed, Lessons from East Asia. Ann Journal 107 134-49. Arbor, Mich University of Michigan Press Cohen, Wesley, and Daniel Levinthal. 1989 "Innovation Da Motta e Albuquerque, Eduardo. 1999 "Domestic and Learning. The Two Faces of R&D." EconomicJournal Patents and Developing Countries: Arguments for their 99: 569-96. Study and Data from Brazil (1980-1995)." Research Collier, Paul, and William F. Sater. 1996. A History of Chile Policy 29: 1047-60. 1808-1994. Cambridge, U K.: Cambridge University David, Paul A., Bronwyn H. Hall, and Andrew A. Toole. Press. 2000. "Is Public R&D a Complement or Substitute for 206 BIBLIOGRAPHY Private R&D' A Review of Economic Evidence." Fernandes, Ana M. 2002. "Trade Policy, Trade Volumes Research Policy 29. 497-529. and Plant-Level Productivity in Colombian Manufac- De Ferranti, David, Daniel Lederman, William F Maloney, turing Industries." Yale University, New Haven, Conn. and Guillermo E Perry 2001 From Natural Resources to Processed. the Knowledge Economy: Trade and Job Quality World Feser, Edward. 2002. "The Relevance of Clusters for Inno- Bank Latin American and Caribbean Studies, Washing- vatiori Policy in Latin America and the Caribbean." ton, D C. Background paper for the LAC flagship report, World De Ferranti, David, Guillermo Perry, Indermit Gill, Luis Bank, Washington, D.C. Serven, Francisco H G Ferreira, Nadeem Ilahi, William Feser, Edward, and M. Luger. Forthcoming "Cluster Maloney, and Martin Rama 2000 Securing Our Future in Analysis as a Mode of Inquiry- Its Use in Science and a Global Economy. Washington, D.C.: World Bank Technology Policymaking in North Carolina." European Duflo, Esther. 2001. "Schooling and Labor Market Conse- Planning Studtes. quences of School Construction in Indonesia. Evidence Filmer, Deon 1999-2001 (continually updated). "Educa- from an Unusual Policy Experiment " American Economic tional Attainment and Enrollment Profiles A Resource Review 91(4) 795-813. 'Book' Based on an Analysis of Demographic and Easterly, William, and Ross Levine. 2001. "It's Not Factor Health Surveys Data." Available at http.//wwwworldbank. Accumulation: Stylized Facts and Growth Models " org/research/projects/edattain/edbook htm. Develop- World Bank Economic Review 15. 177-219. ment Research Group, World Bank, Washington, D.C Eaton, Jonathan, and Samuel Kortum 1999 "Interna- World Bank. tional Technology Diffusion: Theory and Measure- Forbes, Naushad, and David Wield. 2000. "Managing ment" International Economic Review 40(3)- 5 37-70. R&D in Technology-Followers." Research Policy 29(9): - ~ 2000. "Technology, Geography and Trade." 1095-1109 Boston University and NBER Processed. Foster, Andrew D , and Mark R Rosenzweig 1996. "Tech- Espania, Sergio, Suhas D. Parandekar, and Maria Paula nical Change and Human-Capital Returns and Invest- Savanti 2002 "The Impact of the Crisis on the Argen- ments: Evidence from the Green Revolution " American tine Educational Process." World Bank, Washington, Econormic Review 86(4): 931-53. D C. Processed Frydman, Carola. 2002 "Evolution of Public and Private Fagerberg, Jan 1994. "Technology and International Dif- Higher Education in Latin America: Cost Sharing and ferences in Growth Rates." Journal of Economic Literature Financial Aid." Background paper for the LAC flagship 32(3). 1147-75 report World Bank, Washington, D.C. Processed Fajnzylber, Pablo. 2001 "Aircraft Manufacturing in Latin Furman, Jeffrey L., Michael Porter, and Scott Stern America? Notes on Brazil's EMBRAER." World Bank, 2002."The Determinants of National Innovative Capac- Washington, D.C. Processed. ity." Research Policy 31: 899-933. Feenstra, Robert C., and Gordon H. Hanson. 1997. "For- Garmier, Leonardo. 2002. "Costa Rica within the 'New eign Direct Investment and Relative Wages Evidence Economy': The Role of Education, Training and Inno- from Mexico's Maquiladoras." Journal of International vation Systems." Background paper for the LAC flag- Economics 42(1997): 371-93. ship report, World Bank, Washington, D.C Processed. Feldman, K, and J. Meranus. 2000. "Electronic Identity Gauri, Varun. 1998 School Choice in Chile: Two Decades of and Authentication in the 21st Century The Role of Educational Reform. Pittsburgh, Pa University of Pitts- Government as a Certificate Authority." STP 305 Final burgh Press. Paper Processed. Gerschenkron, Alexander 1962. Economic Backwardness in Feldman, Maryann P, and Frank R. Lichtenberg. 1997 Historical Perspective, a Book of Essays. Cambridge, Mass: "The Impact and Organization of Publicly Funded Harvard University Press. Research and Development in the European Commu- Gill, Indermit. 2002. "An Economic Approach to the nity" Working Paper 6040 National Bureau of Eco- 'Knowledge Economy' Technology-Skill Complemen- nomic Research, Cambridge, Mass. tarities and Their Implications for Productivity and 207 CLOSING THE GAP Policy." Background paper for this report, World Bank, Past." American Economic Association Papers and Proceed- Washington, D.C. Processed. ings: Technology, Human Capital, and the Wage Structure Gill, Indermit S, and Harry A. Patrinos. 2001. "Human 86(2): 252-57. Capital Policies for Growth in a Middle-Income Coun- 1998. "The Origins of Technology-Skill Comple- try." World Bank, Washington, D.C. Processed. mentarity." Quarterly Journal of Economics, August: . 2002. "Labor: Investments in Human Capital" 693-732. World Bank, Brazil: Vol. 2 of "The New Growth . 1999. "The Shaping of Higher Education: The Agenda" (World Bank Report 22950-BR). Processed. Formative Years in the United States, 1890 to 1940." Gill, Indermit, Fred Fluitman, and Amit Dar. 2000. Voca- Journal of Economic Perspectives 13(1): 37-62. tional Education and Training Reform: Matching Skills to Gordon, Robert. 2000. "Does the 'New Economy' Measure Markets and Budgets. New York Oxford. Up to the Great Inventions of the Past?" Journal of Gill, Indermit S., Claudio E. Montenegro, and Dbrte Economic Perspectives 14(4): 49-74. Dbmeland. 2002 Crafting Labor Policy. Washington, Griffith, Rachel, Stephen Redding, and John Van Reenen. D.C.: New York: Oxford. 2001. "R&D and Absorptive Capacity: Theory and Girma, Sourafel, and Katharine Wakelin. 2000. "Are Empirical Evidence." Institute for Fiscal Studies Work- There Regional Spillovers from FDI in the UK?" Uni- ing Paper, London, England. versity of Nottingham, Centre for Research on Global- Griliches, Zvi. 1990 "Patent Statistics as Economic Indi- ization and Labour Markets. Research Paper (UK); No. cators: A Survey." Journal of Economic Literature 28(4)- 2000/16: 1-19. 1661-1707. Glass, Amy Jocelyn, and Kamal Saggi. 2002. "Intellectual . 1992. "The Search for R&D Spillovers." Scandina- Property Rights and Foreign Direct Investment." Jour- vianJournal of Economics 94: 29-47. nal of International Economics, (56): 387-410. Grossman, Gene M, and Elhanan Helpman. 1991. Innova- Gledson de Carvalho, A , and Joao Amaro de Matos. 2002. tion and Growth in the Global Economy. Cambridge, Mass. "Venture Capital as Human Resources Management." and London MIT Press. Universidade de Sao Paulo/Universidade Nova de Grossman, Gene, and Edwin Lai. 2002. "Intellectual Prop- Lisboa. Processed erty Rights, the WTO, and Developing Countries." Glewwe, Paul. 2002. "Schools and Skills in Developing Presented at the Annual Bank Conference on Develop- Countries: Education Policies and Socioeconomic Out- ing Economics, World Bank, Washington, D.C. comes." Journal of Economic Literature (U.S.) 40(2): Guasch J. L. 1985 "A Primer on Innovation and Compe- 46-82 tition" Lecture notes, University of California at San Goldfarb, Brent, Magnus Henrekson, and Nathan Rosen- Diego. La Jolla, Calif. berg. 2001. "Demand vs Supply Driven Innovations: . 2000. Argentina: Labor Market in the New Millen- U.S. and Swedish Experiences in Academic Entrepre- nium. Washington, D.C.: World Bank. neurship." SIEPR Policy Paper 00-035, February. Stan- . 2002. "A Conceptual Framework for Cluster ford Institute for Economic Policy Research, Stanford, Development and Analysis." World Bank, Washington, Calif. D.C. Goldin, Claudia. 1999 "Egalitarianism and the Returns to Guasch, Jose Luis., and D. Brehon. 2002. "The Technology Education during the Great Transformation of Ameri- Gap for LAC Countries: A Principle Component Analy- can Education." Journal of Political Economy 107(6): sis." The World Bank, Washington, D C. Processed S65-S94 Guasch, Jose Luis, and Prita Subramanian. 2002 "Infor- . 2001 "The Human-Capital Century and Ameri- matron and Communication Technology: A Key Instru- can Leadership: Virtues of the Past." Journal of Economic ment for Knowledge." Background paper for the LAC History 61(2) 263-92. flagship report, World Bank, Washington, D.C. Goldin, Claudia, and Lawrence F. Katz. 1996. "Technol- Guerrero,Jaime Ramirez. 2002. "The Financing of Vocational ogy, Skill, and the Wage Structure Insights from the Training in LAC," Paper presented at the Inter-American 208 BIBLIOGRAP HY Tripartite Seminar on Training, Productivity, and Decent Henderson, Rebecca, Adam B. Jaffe, Manuel Trajtenberg. Work, Rio de Janeiro, May 15-17, 2002 1998 "Universities as a Source of Commercial Technol- Guevara Niebla, Gilberto. 1990. "La UNAM y la Nacion." ogy: A Detailed Analysis of University Patenting, In Jose Blanco and Gilberto Guevara Niebla, eds., Uni- 1965-1988." Review of Economics and Statistics 80(1). versidad Nacional y Economia Miguel Angel Porrua, pp. 119-27. 429-52. Hoekman, B , and Simeon Djankov 2000. "Foreign Hall, Bronwyn. 2002. "The Financing of Research and Investment and Productivity Growth in Czech Enter- Development." National Bureau of Economic Research prises." World Bank Economic Review (International) Working Paper 8773, Cambridge, Mass 14(1): 49-64. Available at http://www.worldbank.org/ Hall, Bronwyn H., and John Van Reenen 2000. "How research/journals/wbermast.htm. Effective Are Fiscal Incentives for R&D? A Review of Hsieh, Chang-Tai, and Miguel Urquiola. 2002. "When the Evidence." Research Policy 29. 449-69. Schools Compete, How Do They Compete? An Assess- Hall, Robert E., and Charles I. Jones. 1999. "Why Do ment of Chile's Nationwide School Voucher Program." Some Countries Produce So Much More Output Per Princeton, N.J., and Ithaca, N.Y: Princeton University Worker than Others)" Quarterly Journal of Economics and Cornell University. Processed. 114(1)- 83-116. Hunt, Julie, and James Tybout. 1998 "Does Promoting Hansen, Roger D. 1971. The Politics of Mexican Development. High-Tech Products Spur Development'." Georgetown Baltimore Johns Hopkins University Press. University. Processed. Hansen, Thomas Nikolaj, Natalia Agapitova, Lauritz ESCAP (United Nations Economic and Social Commission Holm-Nielsen, and Ognlenka Goga Vukomirovic. for Asia and the Pacific). 2000. Interregional Cooperation 2002. "The Evolution of Science and Technology: Latin in Trade and Investment: Asia-Latin America, Trade and America and the Caribbean in Comparative Perspec- Investment Division, Studies in Trade and Investment tive " Background paper for the LAC flagship report, 43 (ST/ESCAP/2069). Bangkok, Thailand. World Bank, Washington, D.C. ILO (International Labour Organization) 1999. World Hanson, Gordon, and Ann Harrison. 1999. "Who Gains Employment Report 1998-99' Employability in the Global from Trade Reform? Some Remaining Puzzles."Journal Economy-How Training Matters. Geneva, Switzerland. of Development Economics 59(1999): 125-54. Irwin, Douglas A and PeterJ Klenow. 1995. "High-Tech Hanushek, Eric A. 1995. "Interpreting Recent Research R&D Subsidies Estimating the Effects of Sematech" on Schooling in Developing Countries." World Bank Journal of International Economics 40 323-44. Research Observer (International) 10: 227-46 Jacoby, Hanan, and Emmanuel Skoufias 2002. "Financial -- . 1996. "Measuring Investment in Education." Constraints on Higher Education' Evidence from Mex- Journal of Economic Perspectives 10(4): 9-30. ico." World Bank and International Food Policy . 2002. "The Failure of Input-Based Schooling Research Institute, Washington, D.C. Processed. Policies." National Bureau of Economic Research Jaffe, Adam 1989 "The Real Effects of Academic Working Paper9O4O, Cambridge, Mass. Research " American Economic Review 79(5)' 957-70. Hausmann, Ricardo, and Dani Rodrik. 2002. "Economic Jaffe, Adam B., and Josh Lerner 2001. "Reinventing Pub- Development as Self-Discovery." National Bureau of Eco- lic R&D: Patent Policy and the Commercialization of nomic Research Working Paper 8952, Cambridge, Mass National Laboratory Technologies." RAND Jouirnal of Heckman, James, and Carmen Pages-Serra. 2000. "The Economics 32(1): 167-98 Cost of Job Security Regulation: Evidence from Latin James, Vanus, Roland Craigwell, Rosalea Hamilton, and American Labor Markets." Economia:Journal of the Latin Winford James. 2002 "Knowledge in the English- American and Caribbean Economic Association 1(1) 109-44. Speaking Caribbean: Reconsidering Education, Train- Heckman, James J., Lance J. Lochner, and Petra E Todd ing, and Technology Policies." Background paper for 2001. "Fifty Years of Mincer Earnings Regressions." the LAC flagship report, World Bank, Washington, Processed. D.C. Processed. 209 CLOSING THE GAP Jimenez, Emmanuel, and Yasuyuki Sawada. 1999. "Do Klhman, Kim, and Danny S Leipzlger. 1997. "Korea: A Community-Managed Schools Work ? An Evaluation of Case of Government-Led Development." In Danny El Salvador's EDUCO Program." World Bank Economic Leipzinger, ed, Lessons from East Asia. Ann Arbor, Review (International) 13(3): 415-11. Mich.: University of Michigan Press. Jimeno, Juan F. 2002. "Education, Innovation, and Com- Kim, Gwang-Jo. 2001. "The Expansion of the Education plementarity between Skills and Technological System in Korea." Processed. Progress: The Spanish Case." Universidad de Alcala and Kim, Linsu 1993. "National System of Industrial Innova- FEDEA, Spain. Processed. tion: Dynamics of Capability Building in Korea " Sys- Johnson, George E. 1997. "Changes in Earnings Inequal- tems of Innovation: Growth, Competitiveness and Employment ity: The Role of Demand Shifts." Journal of Economic 1' 391-417. Perspectives 11(2). 41-54 . 1997. Imitation to Innovation: The Dynamics of Jones, C. I., and John C. Williams. 1997. "Measuring the Korea's Technological Learning. Boston Harvard Business Social Return to R&D." Working Paper 97002, Stan- School Press. ford University, Department of Economics. King, Elizabeth M, and Berk Ozler. 2001. "What's . 1998 "Measuring the Social Return to R&D." Decentralization Got to Do with Learning' Endogenous Quarterly Journal of Economics 113: 1119-35. School Quality and Student Performance in Nicaragua." Jovanovic, Boyan. 1997 Advances in Economics and Econo- World Bank, Washington, D C. Processed. metrics: Theory and Applications: Seventh World Congress. Kremer, Michael R. 1995. "Research on Schooling: What Volume 2. Cambridge: Cambridge University Press We Know and What We Don't-A Comment on Kane, Thomas J., and Cecilia Elena Rouse. 1995. "Labor- Hanushek " World Bank Research Observer (International) Market Returns to Two- and Four-Year College." Amer- 10 247-54 ican Economic Review 85(3)' 600-614. Krueger, Alan B 1993 "How Computers Have Changed . Kane, Thomas J., and Cecilia Elena Rouse. 1999. the Wage Structure- Evidence from Microdata, "The Community College Educating Students at the 1984-1989 " Quarterly Journal of Economics 108(1). Margin between College and Work."Journal of Economic 33-60. Perspectives 13(1). 63-84 2002 "Inequality, Too Much of a Good Thing." Katz, Jorge. 2001. "Structural Reforms and Technological Princeton University and National Bureau of Economic Behavior: The Sources and Nature of Technological Research. Processed. Change in Latin America in the 1990s." Research Policy Krueger, Alan B., and Mikael Lindahl 2001 "Education (Netherlands) 30(1): 1-9. for Growth Why and for Whom." Journal of Economic Katz, Lawrence F, and David H. Autor. 1999. "Changes in Literature 39(4): 1101-36 the Wage Structure and Earnings Inequality." In Orley Kugler, Maurice. 2002 "Market Reform, Technology Ashenfelter and David Card, eds., Handbook of Labor Adoption and Skill Formation: Evidence from Colom- Economics. New York: Elsevier Science, North-Holland. bia." Background paper for the LAC flagship report Katz, Lawrence F, and Kevin M. Murphy. 1992. "Changes 2002, World Bank, Washington, D.C. Processed. in Relative Wages, 1963-1987: Supply and Demand Lam, David, and Robert F. Schoeni. 1993. "Effects of Fam- Factors." QuarterlyJournal of Economics 107(1). 35-78. ily Background on Earnings and Returns to Schooling. Keller, Wolfgang. 2002. "Trade and the Transmission of Evidence from Brazil." Journal of Political Economy Technology."Journal of Economic Growth 7: 5-24. 101(4): 710-40. Kenny, Charles. 2002. "Information and Communication Lederman, Daniel, and William F Maloney. 2002 "Open Technology-A Role in the Development of Argentina, Questions About the Link between Natural Resources Chile, and Uruguay?" World Bank, Washington, D C. and Economic Growth: Sachs and Warner Revisited." Processed World Bank, Washington, D.C. Processed. Kenny C , and Marialisa Motta. 2002. "The ICT Frame- . 2003a. "Innovation in Mexico: NAFTA Is Not work in Chile." World Bank, Washington, D.C. Processed. Enough." Office of the Chief Economist for Latin America 210 BIBLIOGRAPHY and the Caribbean, World Bank, Washington, D C Lopez-Acevedo, Gladys. 2002a. "Technology and Firm Processed Performance in Mexico." World Bank, Washington, 2003b "R&D and Development " Office of the D C Processed Chief Economist for Latin America and the Caribbean, . 2002b "Technology and Skill Demand in Mex- World Bank, Washington, D.C. Processed. ico.' World Bank, Washington, D C Processed. . 2003c. "Trade Structure and Growth." Office of . 2002c. "Determinants of Technology Adoption in the Chief Economist, Latin America and the Caribbean Mexico " World Bank, Washington, D.C. Processed. Region, World Bank, Washington, D.C. Processed Loveman, Brian. 1979. Chile, the Legacy of Hispanic Capital- Lederman, Daniel, and Laura Saenz. 2002 "A Database of ism. New York. Oxford University Press. Innovation Indicators for the World, 1960-2000" Lucas, Robert E., Jr. 1988 "On the Mechanics of Eco- Office of the Chief Economist, Latin America and the nomic Development."Journal of Monetary Economics 22 Caribbean Region, World Bank, Washington, D.C 3-42. Processed. Lundvall, Bengt-Ake, Bjorn Johnson, Esben Sloth Ander- . 2003 "Innovation around the World A Cross- sen, and Bent Dalum. 2002. "National Systems of Pro- Country Data Base of Innovation Indicators." Office of duction, Innovation and Competence Building." the Chief Economist, Latin America and the Caribbean Research Policy 31(2002). 213-31 Region, World Bank, Washington, D.C. Processed Machin, Stephen, and John Van Reenen. 1998. "Technol- Lederman, Daniel, and Lixin Colin Xu 2001. "Compara- ogy and Changes in Skill Structure: Evidence from tive Advantage and Trade Intensity Are Traditional Seven OECD Countries " Quarterly Journal of Economics Endowments Destiny?" World Bank, Washington, 113(4). 1215-44. D.C. Processed Maloney, William F. 2002 "Missed Opportunities Inno- Lee, Keun, and Chaisung Lim. 2001 "Technological vation, Natural Resources and Growth in Latin Amer- Regimes, Catching-Up and Leapfrogging. Findings ica " Economia 3(1) 111-150. from the Korean Industries." Research Policy 30(2001) Mani, Sunil 2001 "Government, Innovation and Technol- 459-83 ogy Policy, an Analysis of the Brazilian Experience dur- Leipziger, Danny M., and Vinod Thomas 2000. "An ing the 1990s" INTECH Discussion Paper Series Overview of East Asian Experience." In Danny #2001-11 Maastricht, the Netherlands Leipziger, ed., Lessons from East Asia, Studies in Interna- Mansfield, E., and A Romeo. 1980 "Technology Transfers tional Trade Policy. Ann Arbor, Mich.. Michigan Univer- to Overseas Subsidiaries by U S.-Based Firms." Quar- sity Press terlyJournal of Economics, Vol. 95, pp. 737-50 Levy, Frank, and Richard J Murnane 1996 "With What Mariscal, Elisa, and Kenneth L Sokoloff. 2000. "School- Skills Are Computers a Complement?" American Eco- ing, Suffrage, and the Persistence of Inequality in the nomic Review 86(2): 258-62. Americas, 1800-1945 " In Stephen Haber, ed., Polit- Lindbeck, Assar, and Dennis J. Snower 2000. "Multitask ical Institutions and Economic Growth in Latin America. Learning and the Reorganization of Work: From Tay- Essays in Policy, History, and Political Economy. Stan- loristic to Holistic Organization "Journal of Labor Eco- ford University, Stanford, Calif.. Hoover Institution nomics 18(3). 353-76. Press Lindelof, Peter, and Hans Lofsten 2002 "Science Parks Martin, William, and Devashish Mitra. 2001 "Productiv- and the Growth of New Technology-Based Firms- ity Growth and Convergence in Agriculture and Manu- Academic-Industry Links, Innovation and Markets." facturing." Economic Development and Cultural Change Research Policy 859-76. 49(2): 403-22. Loayza, Norman, Pablo Fajnzylber, and Cesar Calder6n. Martinez, Salvador, and Imanol Ordorika. 1991. "UNAM: 2002. "Economic Growth in Latin America and the Espejo del Mejor Mexico Posible " Processed. Caribbean' Stylized Facts, Explanations, and Forecasts." Mathews, John A. 2001 " The Origins and Dynamics of World Bank, Washington, D.C Processed Taiwan's R&D Consortia" Research Policy 31: 633-51. 21 1 CLOSING THE GAP Mayer, David. 2002 "Liberalization, Knowledge, and Industrializing Knowledge: University-Industry Linkages in Technology: Lessons from Veterinary Pharmaceutics and Japan and the United States: 269-306. Cambridge, Mass.: Poultry in Mexico " Background paper. Office of the MIT Press. Chief Economist, Latin America and the Caribbean Mowery, David C., and Arvids A. Ziedonis. 2002. "Acade- Region, World Bank, Washington, D.C. Processed. mic Patent Quality and Quantity Before and After the Mayer, David, and Andrew Foster. 2002. "Scale, Techno- Bayh-Dole Act in the United States." Research Policy logical Change and Human Capital. Manufacturing and 31(31). 399-418 Development in Mexico." World Bank, Washington, Muendler, Marc-Andreas. 2002. "Trade, Technology, and D.C. Processed. Productivity: A Study of Brazilian Manufacturers, 1986- McEwan, Patrick J. 2000. "The Effectiveness of Public, 1998." University of California, Berkeley. Processed. Catholic, and Non-Religious Private Schools in Chile's Mullin, James. 2001. "Science, Technology, and Innova- Voucher System." Education Economics 9(2): 103-28 tion in Chile." International Development Research Melitz, Marc. 2002. "Exports versus FDI." Working paper, Centre, Ottawa. Harvard University, Cambridge, Mass. Murphy, Kevin M., W. Craig Riddell, and Paul M. Romer. Meza Gonzalez, Liliana, and Ana Belen Mora Yagiue. 2001. 1998 "Wages, Skills, and Technology in the United "Why Mexican Manufacturing Firms Invest in R&D?" States and Canada." National Bureau of Economic Universidad Iberamericana and Georgetown University. Research Working Paper 6638, Cambridge, Mass. Processed Myllyntaus, Timo. 1990. "The Finnish Model of Technol- Mian, Sarfraz A 1996. "Assessing Value-Added Contribu- ogy Transfer." Economic Development and Cultural Change tions of University Technology Business Incubators to 38(3): 625-43. Tenant Firms " Research Policy 25(3) (May): 325-35. National Center for Education Statistics. 1995. "Third MICIT (Ministerio de Ciencia y Tecnologfa). 1999. Annual International Mathematics and Science Study." Office of Report, 1999. San Jose, Costa Rica. Educational Research & Improvement, U.S. Dept. of Middleton, John, Adrian Ziderman, and Arvil van Adams. Education, Washington, D.C. Available at http.// 1993. Skills for Productivity New York. Oxford Univer- nces.ed.gov/timss/. sity Press. . 1999. "Third International Mathematics and Sci- Mincer, Jacob 1974. Schooling, Experience and Earnings. ence Study-Repeat." Office of Educational Research New York: National Bureau of Economic Research and Improvement, U.S. Dept. of Education, Washing- Press. ton, D C. Available at http://nces ed gov/timss/ Mizala, Alejandra, and Pilar Romaguera. 1999. "School Nelson, Richard R. 1993 National Innovation Systems: A Performance and Choice: The Chilean Experience." Comparative Analysis. New York: Oxford University Journal of Human Resources XXXV(2): 392-417. Press. . 2002. "Evalucion Del Desemperno E Incentivos Nelson, Richard, and Howard Pack. 1999. "The Asian En La Educaci6n Chilena." Caudermos de Economia, ano Miracle and Modern Growth Theory " The EconomicJour- 39, No. 118. nal 109(457): 416-36. Moretti, Enrico. 2000. "Estimating the Social Return to Nelson, Richard R., and Edmund S. Phelps. 1966. Education: Evidence from Longitudinal and Repeated "Investment in Humans, Technology Diffusion, and Cross-Sectional Data " Department of Economics, Economic Growth." The American Economic Review 56(1/2): UCLA. Processed 69-75. MOST (Ministry of Science and Technology, Republic of Nelson, Richard, and Nathan Rosenberg. 1993. "Technical Korea). n.d "Promotion of Basic Science and Develop- Innovation and National Systems." In National Innova- ment of High-Quality Manpower." Available at http:// tion Systems: A Comparative Analysis. New York' Oxford park.org/Korea/Pavilions/PublicPavilions/Government/ University Press. most/policye3.html. Nelson, Richard, and Gavin Wright. 1994. "The Erosion Mowery, David C. 1999 "The Effects of the Bayh-Dole Act of U.S. Technological Leadership as a Factor in Postwar on U.S University Research and Technology Transfer." Economic Convergence." In William J. Baumol, 212 BIBLIOGRAPHY Richard R. Nelson, and Edward N. Wolff, eds., Conver- Parisi, Maria Laura, Fabio Schiantarelh, and Alessandro gence of Productivity: Cross-National Studies and Historical Sembenelli. 2002 "Productivity, Innovation Creation Evidence. New York: Oxford University Press. and Absorption, and R&D: Micro Evidence for Italy." Neri, Marcelo, and Mark Thomas. 2002 "Household Universita di Padova, Boston College, Universita di Responses to Labor-Market Shocks in Brazil 1982- Torino, Working Paper. 1999 " World Bank, Washington, D.C. Processed. Park, Walter G. 2002. "Institutions and Incentives for Niosi, Jorge. 2002. "National Systems of Innovations Are R&D Implications for LAC Economies." Background 'X-Efficient' (and X Effective) Why Some Are Slow paper for the LAC flagship report, World Bank, Wash- Learners " Research Policy 31 (2002): 291-302 ington, D.C. Noland, Marcus, and Howard Pack 2002. "Industrial Patrinos, Harry Anthony 2001 "A Review of Demand- Policies and Growth: Lessons from International Expe- Side Financing Initiatives in Education " Background rience " Working Paper 169. Central Bank of Chile. paper for the World Development Report 2003, World OECD (Organisarion for Economic Co-operation and Bank, Washington, D C Processed. Development) 1998. Technology, Productivity and Job Pavcnik, Nina. 2002a. "What Explains Skill Upgrading in Creation. Best Policy Practices. Paris: OECD Publications. Less Developed Countries?" Dartmouth College and . 1999. Managing National Innovation Systems. Paris: National Bureau of Economic Research Processed OECD Publications. . 2002b. "Trade Liberalization, Exit, and Produc- . 2000. Literacy in the Information Age: Final Report tivity Improvement Evidence from Chilean Plants." of the International Adult Literacy Survey Paris' OECD Review of Economic Studies 69(1)' 245-76. Publications. Pavcnik, Nina, Andreas Blom, Pinelopi Goldberg, and 200 la. Basic Science and Technology Statistics. Paris. Norbert Schady. 2002 "Trade Liberalization and Labor OECD Publications. Market Adjustment in Brazil." World Bank, Washing- .200 lb Innovative Clusters: Drivers of National Inno- ton, D.C. Processed. vation Systems Paris. OECD Publications. Pavitt, Keith. 2001. "Public Policies to Support Basic 2001c. Knowledge and Skills for Life. Paris. OECD Research: What Can the Rest of the World Learn form Publications. U.S Theory and Practice? (and what they should not 2001d "The Cross-Market Effects of Product and learn) " Industrial and Corporate Change 10:3 pp. Labour Market Policies." In OECD Economic Outlook No 761-779. 70, December 2001 Paris. OECD Publications. Paxson, Christina, and Norbert R Schady 2002. "The . 2002a Benchmarking Industry-Science Relationships. Allocation and Impact of Social Funds Spending on Paris: OECD Publications. School Infrastructure in Peru " World Bank Economic 2002b. Science, Technology and Industry Outlook Review 16(2). 297-319. Paris OECD Publications. PREAL (Partnership for Educational Revitalization in the Pack, Howard. 2000 Technology, Learning, and Innovation: Americas) 2001 Lagging Behind: A Report Card on Edu- Experiences of Newly Industrializing Economies: The End of cation in Latin America. The Inter-American Dialogue, the Road? Commentary. Cambridge (U.K.): Cambridge Washington, D C and the Corporation for Develop- University Press ment Research, Santiago, Chile. 2001. "The Role of Acquisition of Foreign Tech- Pritchett, Lant, and Deon Filmer. 1999. "What Education nology in Taiwanese Growth." Industrial and Corporate Production Functions Really Show: A Positive Theory Change (10)2. of Education Expenditures " Economics of Education Pack, Howard, and Larry Westphal 1986. "Industrial Review 18(2). Strategy and Technological Change: Theory versus Quale, Addison. 2001 "The Culture of Competition and Reality." Journal of Development Economics 22(1986): Cooperation in Curitiba, Brazil' A Barrier to Competi- 87-128 tiveness." Harvard University, Cambridge, Mass. Processed. Parente, Stephen L., and Edward Prescott. 2000. Barriers to Ray, Amit, and Saradindi Bhaduri. 2001 "R&D and Tech- Riches Cambridge, Mass.: MIT Press. nological Learning in Indian Industry. Econometric 213 CLOSING THE GAP Estimation of the Research Production Function." Workers in Latin America." World Bank, Washington, Oxford Development Studies 29(2): 15 5-71 D.C. Processed. . 2001. "The Politics of Learning: The Struggle for Saxenian, AnnaLee. 1999. "Silicon Valley's New Immi- Educational Opportunity in Latin America." Draft for grant Entrepreneurs." http://www.ppic.org/publications/ Victor Bulmer-Thomas, John Coatsworth, and Roberto PPIC120/pplcl20.ch4.html#h3 (p. 1) Cortes Conde, eds, Cambridge Economic History of Latin Schady, Norbert R 2001 "Convexity and Sheepskin America. Effects in the Human Capital Earnings Function: Rodriguez-Clare, Andres. 2003. "Innovation and Technol- Recent Evidence for Filipino Men." World Bank Policy ogy Adoption in Central America." World Bank, Wash- Research Working Paper 2566. Washington, D C. ington, D.C Processed (Forthcoming in Oxford Bulletin of Economics and Romer, Paul M. 1986. "Increasing Returns and Long-Run Statistics.) Growth" Journal of Political Economy 94: 1002-37. 2002a "Caselli and Coleman in Latin America October. Computing Technology Diffusion, Human Capital, and . 1990a "Human Capital and Growth- Theory and Trade Openness " World Bank, Washington, D C. Evidence." Carnegie-Rochester Conference Series on Public Processed. Policy 32(0): 251-86. . 2002b. "The (Positive) Effect of Macroeconomic . 1990b. "Endogenous Technological Change." Crises on the Schooling and Employment Decisions of Journal of Political Economy 98(5): 71-102. Children in a Middle-Income Country." World Bank . 1993. "Endogenous Technological Change" In Policy Research Working Paper 2762. Washington, Edwin Mansfield and Elizabeth Mansfield, eds., The D.C. Economics of Technical Change. United Kingdom: Elgar Schiff, Maurice, and Yanling Wang. 2002. "Education, Rosenberg, Nathan. 2000. "Why Do Firms Do Basic Governance and Trade-Related Knowledge Spillovers in Research (With Their Own Money)?" The Economics of Latin America." Background paper for the LAC flagship Science and Innovation 2: 197-206. report, World Bank, Washington, D C . 2001. "The Role of the Private Sector in Facili- Schiff, Maurice, Yanling Wang, and Marcelo Olarreaga. tating the Acquisition of Technology in Developing 2002. "Trade Related Technology Diffusion and the Countries: A New look at Technology Transfer." Work- Dynamics of North-South and South-South Integra- ing paper, Stanford University. Processed. tion." Policy Research Working Paper Series 281, Rosenzweig, Mark R. 1995. "Why Are There Returns to World Bank, Washington, D.C. Schooling?" AEA Papers and Proceedings. 85(2): 153-58. Schultz, T. Paul. 2001. "What Are the Returns to Poverty Safford, Frank. 1976 The Ideal of the Practical: Colombia's Reduction: Evaluation of PROGRESA in Mexico." Struggle to Form a Technical Elite. Austin, Texas Univer- Processed. sity of Texas Press. Schwartzman, Simon. 2002. "Higher Education and the Saggi, Kamal. 2002. "Trade, Foreign Direct Investment Demands of the New Economy in Latin America." and International Technology Transfer: A Survey." Background paper for the LAC flagship report, World World Bank Research Observer (International) 17(2): Bank, Washington, D.C. 191-235. Fall. SeCyT (Secretaria de Ciencia, Technologia e Innovation Sakakibara, Mariko. 1997. "Evaluating Government- Productiva) 2001. Annual Report, 2001. http://www. Sponsored R&D Consortia in Japan: Who Benefits and secyt.gov.ar/. How'." Research Policy 26(4-5): 447-73. Shimer, Robert. 1999 "The Impact of Young Workers on . 2001."The Diversity of R&D Consortia and Firm the Aggregate Labor Market." National Bureau of Eco- Behavior: Evidence from Japanese Data." The Journal of nomic Research Working Paper 7306, Cambridge, Industrial Economics 49(2). 181-96. Mass. Sanchez-Paramo, Carolina, and Norbert R Schady. 2002 SII (Servicio de Impuestos Internos). 2002. Annual Report, "Off and Running? The Rising Demand for Skilled 2002. 214 BIBLIOGRAPHY Smarzynska, Beata K. 2002. "Composition of Foreign Trajtenberg, Manuel 1992. "Innovation Policy and the Direct Investment and Protection of Intellectual Prop- Economy, National Bureau of Economic Research, Gov- erty Rights. Evidence from Transition Economies." ernment Support from Commercial R&D " In David, World Bank Policy Discussion Paper (Feb). World Paul A., Lessons from the Israeli Experience by Knowledge, Bank, Washington, D.C Property, and the System Dynamics of Technological Change. Smith, Pamela. 2001. "How Do Foreign Patent Rights Proceedings of the World Bank Annual Conferences on Affect U.S Exports, Affiliate Sales, and Licenses."Jour- Development Economics nal of International Economics 55 411-39. . 2001. "Innovation in Israel 1968-1997: A Com- Solow, Robert. 1956. "A Contribution to the Theory of Eco- parative Analysis Using Patent Data " Research Policy nomic Growth." QuarterlyJournal of Economics, 50 65-94 30(3): 363-89. STPS (Department of Labor and Social Welfare, Mexico). 2002. "Government Support for Commercial 1995. La Capacitacion y Asistencia Tecnica en la Micro, R&D: Lessons from the Israeli Experience." In Adam B Pequena y Mediana Empresa- Evaluacion del programa Jaffe, Josh Lerner, and Scott Stern, eds., Innovation Policy CIMO. Mexico City and the Economy. Cambridge, Mass.: MIT Press. Strauss,John, and Duncan Thomas. 1996. "Wages, School- Viotti, E 2001. "National Learning System: A New ing, and Background- Investments in Men and Woman Approach to Technological Change in Late Industrializ- in Urban Brazil." In Nancy Birdsall, Barbara Bruns, and ing Economies and Evidence from the Cases of Brazil Richard Sabot, eds., Opportuinity Forgone: Education, and South Korea." Technological Forecasting and Social Growth, and Inequality in Brazil. Baltimore' Johns Hop- Change 68: 1-28. kins University Press. White House (Office of the Press Secretary). 2000 "Infor- Swan, Trevor. 1956 "Economic Growth and Capital Accu- mation Technology, Research and Development: Infor- mulation." Economic Record 32 334-61. mation Technology for the 21st Century." Washington, DC. Tan, Hong 2000 "Evaluating SME Training Programs: Wolff, Lawrence, and Albrecht Douglas. 1992. "Higher Some Lessons from Mexico's CIMO Program " World Education Reform in Chile, Brazil, and Venezuela: Bank Institute, World Bank, Washington, D.C. Towards a Redefinition of the Role of the State." LAC . 2002. "Do Training Levies Work? Malaysia's Division, World Bank, Washington, D C HRDF and Its Effects on Training and Firm-Level Pro- World Bank. 1997. Enterprise Training, Technology, and Pro- ductivity" World Bank Institute Working Paper, ductivity in Malaysia. World Bank Country Study. World Bank, Washington, D.C. Washington, D.C. Tan, Hong, and Geeta Batra. 1995. Enterprise Training in . 200 la. World Development Indicators. Washington, Developing Countries. Washington, D.C. World Bank D.C. Tan, Hong, and Gladys Lopez-Acevedo 2002. "Mexico: . 2001b. "Brazil: The New Growth Agenda " Vol- In-Firm Training for the Knowledge Economy." Back- ume II. Washington, D.C. ground paper for this report World Bank, Washington, . 2002a. "Brazil Municipal Education Resources, D.C. Incentives, and Results." Report No. 24413-BR. Wash- Tirole, J 1989. The Theory of Industrial Organization. Cam- ington, D.C. bridge, Mass MIT Press. 2002b "Colombian Tertiary Education Paving Topel, Robert. 1999 "Labor Markets and Economic the Way for Reform." Green Cover Report 23935-CO. Growth " In 0 Ashenfelter and D Card, eds., Handbook Washington, D.C. of Labor Economics, Volume 3. Amsterdam- North- 2002c SIMA Database Washington, D C Holland. World Economic Forum. 2002. The Global Competitiveness Torero, Maximo 2002. "Structure of Innovation and Report, 2001-2002. Harvard University, Center for Inter- Research and Development in Peru Relative to Other national Development. Geneva. World Economic Forum Latin American Countries." Background paper for the Wright, Gavin 1999. "Can a Nation Learn? American LAC flagship report, World Bank, Washington, D C Technology as a Network Phenomenon." NBER 215 CLOSING THE GAP Conference Report series. Chicago: University of Yusuf, Shahid, and Simon Evenett. 2002. Can East Asia Chicago Press Compete? Innovation for Global Markets. New York: Xu, 2000. "Multinational Enterprises, Technology Diffu- Oxford University Press. sion, and Host Country Productivity Growth "Journal Yusuf, Shahid, Simon Evenett, and Welping Wu, eds. of Development Economics 62(2): 477-93. 2001. "Facets of Globalization: International and Local Yang, Gulfang, and K. E. Maskus. 2001a "Intellectual Dimensions of Development." World Bank Discussion Property Rights and Licensing: An Empirical Investi- Paper 415. Washington, D.C. gation." Weltwzrtschaftslzches Archzv 137: 58-79 Zuniga, Fernando Vargas. 2001. "Competencies-Based . 2001b. "Intellectual Property Rights, Licensing, Training in Latin America and the Caribbean: Recent and Innovation in an Endogenous Product-Cycle Developments, A Few Experiences." CINTERFOR/ILO, Model."Journal of International Economics 53: 169-187. Montivideo, Uruguay. 216 Edward A. Strudwicke 00803 ISN 1 MC C3-301 WASHINGTON D(. i9 9 178081 51727