77743 Global Development Horizons Capital for the Future: Saving and Investment in an Interdependent World Global Development Horizons Capital for the Future: Saving and Investment in an Interdependent World THE WORLD BANK Washington, D.C. © 2013 International Bank for Reconstruction and Development / The World Bank 1818 H Street NW, Washington DC 20433 Telephone: 202-473-1000; Internet: www.worldbank.org Some rights reserved 1 2 3 4 16 15 14 13 This work is a product of the staff of The World Bank with external contributions. Note that The World Bank does not necessarily own each component of the content included in the work. The World Bank therefore does not warrant that the use of the content contained in the work will not infringe on the rights of third parties. The risk of claims resulting from such infringement rests solely with you. The findings, interpretations, and conclusions expressed in this work do not necessarily reflect the views of The World Bank, its Board of Executive Directors, or the governments they represent. The World Bank does not guarantee the accuracy of the data included in this work. The boundaries, colors, denominations, and other information shown on any map in this work do not imply any judgment on the part of The World Bank concerning the legal status of any territory or the endorsement or acceptance of such boundaries. Nothing herein shall constitute or be considered to be a limitation upon or waiver of the privileges and immunities of The World Bank, all of which are specifically reserved. Rights and Permissions This work is available under the Creative Commons Attribution 3.0 Unported license (CC BY 3.0) http:// creativecommons.org/licenses/by/3.0. Under the Creative Commons Attribution license, you are free to copy, distribute, transmit, and adapt this work, including for commercial purposes, under the following conditions: Attribution —World Bank. 2013. Capital for the Future: Saving and Investment in an Interdependent World. Global Development Horizons. Washington, DC: World Bank. doi:10.1596/978-0-8213-9635-3. License: Creative Commons Attribution CC BY 3.0 Translations —If you create a translation of this work, please add the following disclaimer along with the attribution: This translation was not created by The World Bank and should not be considered an official World Bank translation. The World Bank shall not be liable for any content or error in this translation. All queries on rights and licenses should be addressed to the Office of the Publisher, The World Bank, 1818 H Street NW, Washington, DC 20433, USA; fax: 202-522-2625; e-mail: pubrights@worldbank.org. ISBN (paper): 978-0-8213-9635-3 ISBN (electronic): 978-0-8213-9955-2 DOI: 10.1596/978-0-8213-9635-3 ISSN: 2221-8416 Cover photo: Richard H. Fox/Getty Images Cover design: Marc Pekala, FCI Creative Contents Foreword . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xi Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xiii Abbreviations. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xv OVERVIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1 Outlooks under the two scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 Modeling the global dynamics of investment, saving, and capital flows . . . . . . . . . . . . . . . . . . . . . 10 Note . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14 CHAPTER 1: The Emerging Pattern of Global Investment . . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 Changing patterns of investment worldwide . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 19 Long-run drivers of investment around the world. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 33 The changing face of future investment opportunities . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 Transformations in sectoral investment in developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 Conclusion and policy directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 54 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 56 CHAPTER 2: Global Saving in 2030 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61 Saving at the aggregate level: Past, present, and future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 A more nuanced view of the future of saving: Distribution, public finance, old-age support, and other impacts . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 Conclusion and policy directions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109 CHAPTER 3: Capital Flows in the Third Age of Financial Globalization . . . . . . . . . . . . . . . . .115 Rates of return around the world . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .119 Changing patterns in net capital flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120 Expanding volumes of gross capital flows to and from developing countries . . . . . . . . . . . . . . . . 125 Capital flows and monetary policy spillovers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 135 Global Development Horizons v vi Contents Global Development Horizons Conclusion and policy implications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 139 Notes . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 144 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 147 Boxes 1.1 Different terms, different rates: Purchasing-power adjusted investment vs. investment expressed in national currency . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 1.2 Investment booms are not always associated with sustained growth . . . . . . . . . . . . . . . . . . 23 1.3 Investment in human capital can be a source of economic growth . . . . . . . . . . . . . . . . . . . . . 30 1.4 Energy project financing in the Middle East: A multiparty, multimodal approach . . . . . . . . . . 31 1.5 Structural factors can affect the cross-country allocation of investment financing . . . . . . . . . 36 1.6 Tobin’s Q calculations for developing countries suggest positive growth prospects . . . . . . . . 38 1.7 How transient shocks to uncertainty can have a long-lasting impact on investment . . . . . . . 39 1.8 Alternative financing mechanisms have begun to be adopted in financing infrastructure . . . . 51 2.1 A gradual shift in the global distribution of wealth is under way . . . . . . . . . . . . . . . . . . . . . . . 69 2.2 The geographically asynchronous demographic transition is driven by cross-country differences in some key factors . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 75 2.3 Literature on modeling the effect of aging on saving at the global level . . . . . . . . . . . . . . . . . 78 2.4 Micro analysis is a useful tool for determining likely trends in future savings . . . . . . . . . . . . . 83 2.5 A global dataset of household censuses provides a nuanced view of saving at the micro level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97 2.6 The National Transfer Accounts project is a valuable source of data on age-related public transfers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 99 2.7 The impact of aging on government budgets is clarified by breaking it down into its components . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101 3.1 The Third Age of Financial Globalization follows two previous eras . . . . . . . . . . . . . . . . . . . . .116 3.2 Historically, the dominant net providers and recipients of capital have not been fixed . . . . . 122 3.3 Gross capital flow scenarios are based on historical evidence on the relationship between the flows and their determinants . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 129 3.4 Encouraged by the continent’s growth prospects, demographic trends, and improved investment climate, global investors and creditors are increasingly attracted to Sub-Saharan Africa . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 Figures O.1 Future global saving and investment rates will remain fairly stable in the gradual convergence scenario, but this stability belies substantial shifts in the relative shares of developing and high-income countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 O.2 Developing countries will represent more than half of global capital stocks by 2030 in the gradual convergence scenario, compared with about a third in 2010 . . . . . . . . . . . . 4 O.3 Increased earning power will be the greatest driver of saving by Mexican households . . . 4 O.4 Annual infrastructure needs over the next 20 years are likely to be greatest in East and South Asia . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 O.5 By 2030, nearly half or more of gross capital inflows will likely go to developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 7 Global Development Horizons Contents vii O.6 Schematic diagram describing interactions between saving, investment demand, and investment financing . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12 1.1 Gross investment in developing countries has increased in absolute terms (panel a) and as a share of global investment (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 20 B1.1.1 Differentials in investment rates (panel a) and capital-output ratios (panel b) are greater when measured in PPP terms . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 21 1.2 Developing countries’ rising investment rates (panel a) and growing share of global output (panel b) have contributed to their increased share of investment in global output . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 1.3 The rising share of developing countries’ investment in global output is due to more than just changes in China and India . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 22 1.4 Investment rates among Sub-Saharan African countries of different income levels have followed distinct paths . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 B1.2.1 Many countries experience weak growth following an investment boom . . . . . . . . . . . . . 24 1.5 Global manufacturing investment tends to be concentrated in lower-middle-income countries (panel a), with China currently accounting for the vast majority of investment in those countries (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 25 1.6 The public sector share of output is lower in high-income countries than in other country groups (panel a), but the public sector share of investment has converged among country groups (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27 1.7 Private sector commitments to infrastructure have risen over time, both in major developing countries (panel a) and across most infrastructure subsectors (panel b) . . . . . 28 1.8 Infrastructure investment shares of GDP declined rapidly in major Latin American economies in the late 1980s and have remained subdued in subsequent years . . . . . . . . 29 B1.3.1 Investment in human capital relative to physical capital in high-income and developing countries, 1990–99 to 2000–10 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30 1.9 Adjusting capital stocks by efficiency-adjusted units of labor results in different distributions compared to comparing capital stocks alone . . . . . . . . . . . . . . . . . . . . . . . . 33 1.10 Investment rates and financial sector development in developed and developing countries have evolved similarly . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 37 1.11 A positive relationship exists between investment and institutional quality . . . . . . . . . . . 37 B1.6.1 Tobin’s average Q is higher on average in lower-income developing countries, reflecting market expectations of greater growth potential . . . . . . . . . . . . . . . . . . . . . . . . 38 B1.7.1 Investment rate response to increased uncertainty in high-income and developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39 1.12 Total investment (panel a) and investment share of global output (panel b) will increase in developing relative to high-income economies in the gradual convergence scenario . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40 1.13 Investment rates in many developing countries will trend downward in the gradual convergence scenario, but remain fairly stable in high-income countries . . . . . . . . . . . . . 41 1.14 The yield on loanable funds will rise more rapidly in high-income countries than in many developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42 1.15 Developing countries will account for more than half of global capital stocks by 2030 in the gradual convergence scenario (panel b), compared with about a third today (panel a). . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 43 1.16 Total global investment will be higher for developing countries in the rapid convergence scenario (panel a), but this belies substantial heterogeneity in individual developing-country experiences (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44 viii Contents Global Development Horizons 1.17 Annual infrastructure investment needs in developing countries will be substantial for the next two decades (panel a), with the greatest needs arising in East and South Asia (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 49 1.18 Infrastructure investment financing modes have changed over time and will likely favor greater bond financing in the future. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50 2.1 Developing countries have accounted for a growing share of global saving since around 2000 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63 2.2 Both relative size and saving rates have contributed significantly to the increased global importance of developing countries’ saving . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 2.3 Developing countries’ saving rates have increased . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 64 2.4 The paths of saving rates have differed substantially across major developing economies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65 2.5 Saving rates tend to be lower in lower-income countries . . . . . . . . . . . . . . . . . . . . . . . . . 66 2.6 Household saving varies significantly by region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 2.7 Corporate saving in China and India has grown faster than household saving in recent years . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67 2.8 The public portion of national saving has fluctuated within a fixed range . . . . . . . . . . . . . 68 B2.1.1 The global distribution of national wealth per capita has changed over time . . . . . . . . . . . 70 2.9 Saving rates tend to increase with growth (panel a) and decrease with financial development (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 2.10 Saving rates tend to decrease with population aging . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71 2.11 By 2020, growth in world’s working-age population will be exclusively determined by developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73 2.12 Birth and death rates have shown distinct trends across regions . . . . . . . . . . . . . . . . . . . 74 B2.2.1 Developing countries have made significant progress in reducing child mortality, although from different starting points and at different rates . . . . . . . . . . . . . . . . . . . . . . 75 B2.2.2 The gap in fertility rates across regions has narrowed . . . . . . . . . . . . . . . . . . . . . . . . . . . 76 2.13 The size and timing of the demographic “bulge” differs across regions . . . . . . . . . . . . . . 77 2.14 Saving rates will decline more slowly in developing countries, and by 2030 those countries will account for two-thirds of global saving . . . . . . . . . . . . . . . . . . . . . . . . . . . . 79 2.15 China will continue to be dominant in the global saving picture . . . . . . . . . . . . . . . . . . . . . 80 2.16 Differences in the demographic transition will translate into different saving rate trends . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 80 2.17 National saving rates (panel a) and share of global saving (panel b), by income groups, gradual versus rapid convergence scenario, 2010–30 . . . . . . . . . . . . . . . . . . . . . 81 2.18 National saving rates of Thailand, the Russian Federation, Mexico, and Ghana are similar to those of their regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 84 2.19 Projection of age effects on household saving rates in Ghana and the Russian Federation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 86 2.20 Demographic change will drive large changes in saving rates in the decades ahead . . . . 86 2.21 Due to strong growth, the cohort effect will play an important role in the path of Thailand’s saving rate . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 2.22 Growth is an important determinant of saving rates: Two scenarios for the Russian Federation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 87 2.23 Educational attainment in Mexico has been rising over time . . . . . . . . . . . . . . . . . . . . . . . 88 2.24 The saving life cycle differs significantly by educational attainment (panel a), but across all educational levels, young cohorts tend to save more than older cohorts (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 89 Global Development Horizons Contents ix 2.25 Increased earning power will be the greatest driver of saving by Mexican households . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 90 2.26 The bottom 40 percent of income earners in developing regions are not only less likely to save than the top 60 percent but are also less likely to save using formal financial institutions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 92 2.27 The age composition of households varies by age of the household head; more highly educated individuals are less likely to live in a household headed by someone younger . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 93 2.28 Household size varies by age of the household head, and this relationship varies somewhat by educational level . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 95 2.29 Younger cohorts tend to have smaller family sizes, but this relationship varies significantly across regions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 96 B2.5.1 Family size for the first and latest cohort, points and fitted values, República Bolivariana de Venezuela . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 98 2.30 There is wide variation across countries in the extent to which public transfers tend to rise with age . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100 3.1 Changes in yields are more muted under the gradual convergence scenario than under rapid convergence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 121 B3.2.1 Net capital outflows (current account), the United Kingdom and the United States, 1850–1945 (panel a) and evolution of the global pattern of current account balances, 1980–2010 (panel b) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 122 3.2 In the gradual convergence scenario, there will be sizable net capital flows from China to high-income countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 123 3.3 In the rapid convergence scenario, much of the developing world apart from China will be net capital importers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 124 3.4 Gross capital flows have long expanded relatively faster than net flows in high-income economies, and in recent years a similar trend can be observed in developing economies . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 126 3.5 International investment positions still reflect a large disparity in foreign asset holdings between advanced and developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . 127 3.6 The relationship between GDP per capita and external balances across countries is nonlinear . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 128 3.7 By 2030, most gross capital flows may go to developing countries . . . . . . . . . . . . . . . . 129 3.8 Relative to developing countries’ GDP, gross inflows may hold steady or trend upward, depending on the pace of growth and financial development . . . . . . . . . . . . . . 130 3.9 Small and medium developing countries as well as the BRICs will be important sources and destinations of capital flows . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131 B3.4.1 Gross capital inflows to Sub-Saharan Africa will rise as a share of inflows to developing countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 132 B3.4.2 Cross-border bank loans to Sub-Saharan Africa are on the rise . . . . . . . . . . . . . . . . . . . . 134 3.10 Historically, the composition of capital inflows has differed between income groups, with FDI providing stability in flows to developing countries . . . . . . . . . . . . . . . 135 3.11 International assets and foreign exchange trading are mainly denominated in the U.S. dollar and the euro . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 137 3.12 The U.S. dollar is still the main currency in global credit markets . . . . . . . . . . . . . . . . . . 138 3.13 A number of new central bank liquidity swap lines were extended between December 2007 and December 2010 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 140 x Contents Global Development Horizons Maps 1.1 The geographical distribution of capital stocks has gradually shifted to developing countries over the past three decades . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 32 1.2 Capital stocks per worker are now more equally distributed than ever before . . . . . . . . . . . . 34 Tables B1.2.1 Investment booms have occurred in a broad range of developing and high-income countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23 1.1 There is significant heterogeneity in marginal products of capital, at both economywide and sectoral levels, across developing countries . . . . . . . . . . . . . . . . . . . . 26 B1.4.1 Top five project finance deals in the Middle East, 2009. . . . . . . . . . . . . . . . . . . . . . . . . . . 31 1.2 Both developing and high-income countries will see a rise in the share of investment devoted to the services sector in the future . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 46 2.1 Trends in saving rates differ widely across countries . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 85 2.2 Household saving rates differ significantly along the income distribution . . . . . . . . . . . . . 91 2.3 The prevalence of intergenerational households varies by country income level and by region . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 94 B2.5.1 Number of countries with available census data, by region, 1960–2009 . . . . . . . . . . . . . 97 2.4 The path of age-related fiscal pressures will depend strongly on benefit and coverage levels . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 102 2.5 Population aging will put severe pressure on some countries’ public pension systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 104 Foreword V ERY LONGRUN FORECASTING IS global GDP, have led to a spectacular dominance a hazardous activity because the uncer- by developing countries in global investment tainties and imponderables of life have dynamics. Two-thirds of the growth in global plenty of time to intrude, and bend and buck the investment over the last ten years has originated charted path. At the same time, to craft policy in developing countries. that is rooted in reason and reality, we need to To varying degrees, these forces will remain peer into the future with the best information, sta- in play in the medium to long run. Current pro- tistics, and models that we have. That is what we jections indicate that the world population will have tried to do in Global Development Horizons. grow from around 7 billion in 2010 to more We have marshaled some of our best analysts and than 8.5 billion in 2030. The addition of more researchers to undertake this task. Let me thank than 1.4 billion people will occur almost exclu- them for taking on this hazardous and somewhat sively in developing countries. More crucially, thankless job, but also remind them that one there will be a significant shift in the age struc- advantage they have is that, by the time the accu- ture of both developing and advanced countries. racy of their forecasts becomes known, most of Aging in high-income countries, emerging East them will have moved on to other pursuits. Asia, and the transition economies of Europe The aim of this edition of Global Development and Central Asia will affect both saving and Horizons is to build scenarios for the global econ- investment behavior. In Sub-Saharan Africa and omy over, roughly, the next two decades, with South Asia, on the other hand, the population special focus given to saving and investment. will remain relatively young. Concurrent with Developing countries’ share in global investment these demographic trends, developing coun- has grown from less than 20 percent to almost tries will continue to catch up with the higher one-half in the past 15 years. A similar pattern is productivity levels that exist in more advanced observed for saving. Only a minor part of these economies. trends in saving and investment is due to the These trends raise some pressing questions consequences of the 2008 global financial crisis. about the fi nancing of future global patterns of Most of the changes have emerged as a result of investment, and the answers to these are not at all entrenched economic and demographic forces. obvious. For example, will aging societies starve Productivity catch-up, together with increasing the global economy of capital as the elderly draw integration into global markets, sound macro- down their stock of savings? Or will robust eco- economic policies, and improved education and nomic growth boost younger workers’ incomes health, have contributed to an acceleration of and saving? Will investment opportunities growth and large increases in investment and decline rapidly as the growth of the labor force saving rates in many developing countries. A fur- slows across the world? To address these ques- ther boost has been provided by the demographic tions this edition of Global Development Horizons dividend, which has reached peak levels in many develops two scenarios of investment, saving, and countries. The rising investment and saving capital flows through 2030, based on alternative rates within developing countries, together with paths of productivity and fundamental drivers of the expanding share of developing countries in structural change. Global Development Horizons xi xii Foreword Global Development Horizons The forward-looking scenarios explored in this markets. As a result, the global financial crisis report come at a critical juncture in the global illustrated that developing countries are vulner- economy. In the field of saving, developing- able to what happens in dominant financial cen- country governments are pondering strategies for ters in high-income countries. This is unlikely to expanding social services amid fiscal pressures persist in scenarios in which developing coun- emanating from their aging populations. In the tries’ share of global saving and investment con- investment arena, policy makers in developing tinues to rise. countries are looking to spur domestic economic The printed copy of this edition of Global activity as they seek to reorient away from export- Development Horizons will be just one part of led growth while decoupling from the malaise a broader multimedia ensemble. The accom- still hindering the developed world. And in terms panying website, http://www.worldbank.org/ of capital markets, the global financial crisis has CapitalForTheFuture, includes access to a host promoted a rethinking of the benefits and costs of related electronic resources, including data of unbridled capital f lows, coupled with the sets behind the two main scenarios presented in widespread view that there is a pressing need for this report, background papers, technical appen- reform of the international financial system. dixes, interactive widgets that allow the inter- One key prognosis discussed at some length ested reader to explore variations to some of the in the report is that, in spite of pressures from assumptions used in the projections, and related population aging, the world will not face a short- audio and video resources. age of saving in the future, and yields on capi- By setting out the implications of emerg- tal are expected to remain fairly stable through ing trends as they will play out over the next 2030. Th is stability of saving at the global level, two decades, this edition of Global Development however, belies a significant shift in composi- Horizons should help policy makers become bet- tion. By 2030, developing countries will account ter armed with knowledge both in terms of posi- for around two-thirds of every dollar saved and tioning their respective countries in a new, multi- invested (compared with about half today), and polar, global economy, and in terms of adapting around half of the stock of global capital will to potential challenges they will inevitably face, reside in the developing world, compared to less at home and beyond, in the near and more dis- than one-third today. Th is implies a huge chal- tant future. lenge for financial intermediation in develop- ing countries. Currently, many countries with a Kaushik Basu dominant position in trade and production play Senior Vice President and Chief Economist a relatively minor role in international fi nancial The World Bank Acknowledgments T HIS R EPORT WAS PR EPAR ED BY which undoubtedly improved the overall quality the Emerging Global Trends team of the of the product. World Bank’s Development Prospects Ca roline Freund (Of f ice of the Chief Group (DECPG). Mansoor Dailami was the lead Economist, Middle East and North Africa author and manager of the team in the first phase Region), Marcelo Giugale (Poverty Reduction of production of the report; following his retire- and Economic Management, Africa Region), ment from the Bank, this role was assumed by Ivailo Izvorski (Macroeconomics, Povert y Maurizio Bussolo. The report was prepared with Reduction and Economic Management, Europe direction from Hans Timmer and under the and Central Asia Region), Marilou Jane D. Uy general guidance of Kaushik Basu, World Bank (Office of the President, Special Envoy), Ravi Vish Senior Vice President and Chief Economist. (Economics and Sustainability Group), Orazio The Overview was written by Maurizio Pietro Attanasio (University College London), Bussolo, with contributions from other team and Menzie Chinn (University of Wisconsin, members. Chapter 1 was written by Jamus Madison) served as peer reviewers at the report’s Jerome Lim, with written contributions from concept paper stage, while the group of review- Doerte Doemeland on historical infrastructure ers for the Bank-wide review included Sebastien investment. Chapter 2 was written by Maurizio Dessus (Poverty Reduction and Economic Bussolo, with written contributions from Management, Africa Region), Samuel Freije- Jonathon Adams-Kane and Jamus Jerome Lim. Rodriguez (Poverty, Gender, and Equity Group; Israel Osorio-Rodarte and Sebastian Renner Poverty Reduction and Economic Management; provided many inputs to the micro-level analy- Latin America and the Caribbean Region), sis in the chapter, and Miguel Székely provided David Gould (Office of the Chief Economist, expert advice in this area as well. Chapter 3 was South Asia Region), Bert Hofman (Office of written by Mansoor Dailami, Jonathon Adams- the Chief Economist, East Asia and Pacific Kane, and Dana Vorisek, with contributions Region), Orazio Pietro Attanasio (University from Yueqing Jia, Hyung Sik Kim, and Sergio College London), François Bourguignon (Paris Kurlat. Maryla Maliszewska, under the guidance School of Economics), Menzie Chinn (University of Maurizio Bussolo, led the computable general of Wisconsin, Madison), Da le Jorgenson equilibrium modeling exercise using the World (Harvard University), and Nick Stern (London Bank’s Linkage model, the findings of which are School of Economics). Within the Bank, addi- featured in all three chapters of the report. tional comments were provided by Jean-Pierre Many others provided advice, inputs, and Chaffour, Tatiana Didier, Indermit Gill, Elena comments at various stages of the report’s con- Ianchovichina, A lain Ize, Ivailo Izvorski, ceptualization and preparation. Aslı Demirgüç- Norman Loayza, Aart Kraay, Sergio Schmukler, Kunt coordinated the review process within the Augusto de la Torre, and Galina Vincelette. Development Economics Vice Presidency, pro- Outside the Bank, a number of experts provided viding substantial comments and advice. Prior to inputs through meetings, discussions, and pre- his departure from the Bank, Martin Ravallion sentations of the report’s early findings. They also offered significant input at an early stage, include Thorsten Janus (University of Wyoming), Global Development Horizons xiii xiv Acknowledgments Global Development Horizons Ronald Lee (University of California at Berkeley), Kristina Tan Mercado and Rosalie Marie Andrew Mason (University of Hawaii at Manoa), Lourdes Singson Dinglasan provided production Paul Masson (University of Toronto), and partici- assistance to the Emerging Global Trends team. pants at the Midwest Political Science Association Indira Chand and Merrell Tuck-Primdahl man- meetings. aged dissemination activities. Jeffrey Lecksell The Global Development Horizons website was assisted in the preparation of the maps. Book produced by David Horowitz, Jamus Jerome Lim, design, editing, and production were coordinated and Indira Chand. Technical help in the produc- by Patricia M. Katayama, Cindy Fisher, and tion of the website was provided by Roula Yazigi Denise Bergeron of the World Bank’s Office of and Vamsee Krishna Kanchi. Background papers the Publisher. and related research are available on the website: http://www.worldbank.org/CapitalForTheFuture. Abbreviations AIC Akaike information criterion AIDS acquired immune deficiency syndrome BIS Bank for International Settlements BRICs Brazil, the Russian Federation, India, and China BRIIC Brazil, the Russian Federation, India, Indonesia, and China CGE computable general equilibrium (model/analysis) FDI foreign direct investment GARCH generalized autoregressive conditional heteroscedasticity GCC Gulf Cooperation Council GDH Global Development Horizons (World Bank) GDP gross domestic product GEM Global Economic Monitor (World Bank database) GNI gross national income ICT information and communication technology IMF International Monetary Fund IPP independent power project IPUMSI Integrated Public Use Microdata Series, International IWPP independent water and power project kWh kilowatt-hour LIBOR London interbank offered rate LNG liquefied natural gas MPK marginal product of capital NTA National Transfer Accounts (project) OECD Organisation for Economic Co-operation and Development PPP public-private partnership PPP purchasing power parity R&D research and development SAR Special Administrative Region TFP total factor productivity VAR vector autoregressive WDI World Development Indicators (World Bank database) 3D three dimensional All dollar amounts are in U.S. dollars unless otherwise noted. Global Development Horizons xv Overview M AJOR STRUCTUR AL TR ANS formations under way in the global economy will shape the economic fortunes of nations for decades to come. The gradual acceleration of trend growth in develop- two groups (such as fi nancial development and improvements in institutional quality). In the first scenario, this convergence is gradual, while in the second, convergence is more rapid. Convergence between the developed and ing countries, which started during the 1990s, developing worlds was elusive for much of the has increased their contribution not only to 20th century. But upgrading of education and global investment but also to global saving. Th is health, improvements in governance, continued is a significant change from a few decades ago, economic and financial globalization, and the when many considered low national saving rates rapid diffusion of information and communi- and difficulty in attracting foreign capital to be cations technologies increasingly point toward constraints to investment and growth in devel- such convergence as reality (Spence 2011). The oping countries. Collectively, developing coun- two convergence scenarios do not account for tries’ domestic saving stood at 34 percent of every trend or shock that may have an important their gross domestic product (GDP) in 2010, up economic impact over the time horizon consid- from 21 percent in 1970, and their investment ered, such as climate change, commodity shocks, was around 33 percent of their GDP in 2012, or financial crises. Indeed, these scenarios are up from 22 percent. As a result of the upward not intended to be forecasts and should not be trends in these rates and an accelerated economic judged by their likelihood of materializing. Their expansion, developing countries’ share of global usefulness is in illustrating as clearly as possible saving now stands at 46 percent, nearly double how upcoming changes in the global patterns the level of the mid-1960s. of growth, demographic change, and structural What is behind these trends, and will they variables will affect—in a way fundamentally continue in the future or are they a short-term different from the past—saving, investment, and phenomenon? What do they mean for devel- capital flows. opment and poverty reduction? This report In the gradual convergence scenario, the aver- addresses these questions, identifying key eco- age per capita income of the developing world nomic and structural drivers that, together with will rise from about 8 percent of that in high- demographic shifts, will affect saving and invest- income countries in 2010 to about 16 percent by ment decisions over the next two decades—and 2030. In the rapid convergence scenario, the gap thus the global distribution of capital in the between the two country groups is smaller: an future. The report builds two distinct scenarios average citizen of what is now a developing coun- that serve as “economic laboratories” to study the try will, by 2030, earn 19 percent of the income potential consequences of changes in these driv- of an average citizen of a high-income country. ers. The main differences between the two sce- The gradual and rapid scenarios predict average narios are (a) the speed of convergence between world economic growth of 2.6 percent and 3 per- the developed and developing worlds in per capita cent per year, respectively, during the next two income levels (due to productivity catch-up), and decades; the developing world’s growth will aver- (b) the pace of structural transformations in the age an annual rate of 4.8 percent in the gradual Global Development Horizons 1 2 Overview Global Development Horizons convergence scenario and 5.5 percent in the rapid Outlooks under the one. With gradual convergence, the contribution two scenarios of developing countries to global growth will rise from 73 percent around 2015 to 87 percent by The two scenarios presented in this report offer 2030; with rapid convergence, developing coun- a picture of the future of global capital that does tries’ contribution will reach 93 percent by the not parallel recent trends, providing a template for end of the period. In both scenarios, developing thinking strategically about a number of global countries’ employment in services will account financial and economic challenges looming ahead for more than 60 percent of their total employ- and their implications for policy today. The main ment by 2030, their share of total expenditure on findings of the report are summarized below. food items will be halved, and they will account for more than 50 percent of global trade. The world will not run out of saving in the The increasing global weight of investment future, and global investment rates, saving activity in developing countries under these sce- rates, and yields on capital will remain fairly narios will not mechanically materialize but will stable through 2030 require two conditions to be fulfi lled: first, pro- Aging will tend to be less pronounced in develop- ductivity growth and sectoral shifts will have to ing economies than in advanced economies over create enough investment opportunities. Second, the forthcoming two decades. In addition, eco- both domestic and international investors will nomic growth and the speed of financial market have to be willing to fi nance these investments development will be greater in developing coun- and allocate two-thirds of global savings to devel- tries. Under the gradual convergence scenario, oping countries. Fulfi llment of these conditions the saving rate of the developing world will inch depends on major structural changes and contin- down from a peak of 34 percent in 2014 to 32 ued improvements in governance over the next percent in 2030, while the saving rate for high- two decades. income economies will fall from 20 percent to 16 Thus, the scenario analysis considers how percent over the same period (figure O.1, panel structural and institutional factors may change in b). However, the world will not run out of saving the future. In the gradual convergence scenario, because the global saving rate depends not only these structural factors are assumed to evolve in on country-level saving rates but also on econo- a fashion consistent with their historical patterns. mies’ relative sizes. The shift of global economic In the rapid convergence scenario, however, struc- weight toward high-saving developing economies tural transformations are assumed to break away means that, by 2030, developing countries will from historical trends and proceed more quickly. account for 62–64 percent of total world saving, This would occur if, for instance, the techno- depending on which scenario is considered, up logical wave from the third industrial revolution from 45 percent in 2010. Th is shift in economic becomes entrenched globally and fundamentally weight will compensate for the reduction in sav- changes the economic and institutional struc- ing rates at the country level, so that the global tures of developing countries. In such a scenario, saving rate will remain more or less unchanged. financial markets in economies like Brazil, India, With robust economic growth moderating and those of the Middle East will develop consid- the negative impacts of aging and financial mar- erably, with these countries attaining, by 2030, ket development on developing countries’ saving a level of fi nancial development comparable to rates, and with no significant increase in invest- the United States in the early 1980s. Similarly, ment demand as growth slows moderately after the quality of institutions in developing countries 2020 in most countries, it is unlikely that there will tend to improve significantly. Importantly, will be increased tension between global saving the rate of progress in these variables will differ supply and capital demand. Accordingly, there among countries: for economies starting further will not be significant upward pressure on yields, away from the frontier, the rate of growth will be even in the rapid convergence scenario, under relatively faster. which the average global yield is expected to rise Global Development Horizons Overview 3 FIGURE O.1 Future global saving and investment rates will remain fairly stable in the gradual convergence scenario, but this stability belies substantial shifts in the relative shares of developing and high-income countries a. Income groups’ shares of global investment, b. Income groups’ saving rates and the world investment rate 90 30 World investment as share of GDP, % 40 High income Developing Share of global investment, % 80 35 Saving as a share of GNI, % World 25 70 (right axis) 30 60 20 25 High income 50 20 15 40 15 30 10 10 20 5 High income/ Developing 5 developing differential  10 0 0 0 –5 25 30 75 80 85 90 95 00 05 10 15 20 25 30 70 65 70 75 80 85 90 95 00 05 10 15 20 20 20 19 19 19 19 19 20 20 20 20 20 20 20 19 19 19 19 19 19 19 19 20 20 20 20 20 Sources: World Bank projections, supplemented with calculations using data in the World Bank Global Economic Monitor and World Development Indicators databases. by roughly half a percentage point between 2014 that such redistributive shifts will be replicated and 2030. within countries. Indeed, saving (and wealth) Global investment will expand at a rate in line in developing countries is highly concentrated with overall output, amounting to $26.7 trillion among upper-income households. Data at the (in 2010 dollars, or more than a quarter of the household level show that the least educated global output of $105 trillion) by 2030 under the groups tend to have the lowest lifetime income gradual convergence scenario (figure O.1, panel a). and low or no savings, suggesting an inability to improve their earning capacity and, for the poor- The distribution of the global stock of capital est, to escape poverty traps. If the distribution of will shift toward the developing world, but education among workers of future generations wealth may remain concentrated among high- were to remain as unequal as it is today, this would income households in developing economies perpetuate inequality of earning capacity, saving, The increased significance of developing coun- and wealth in the future. Leveling the playing tries as suppliers and demanders of global saving field in terms of educational opportunities could will lead to a redistribution of the world stock of thus be supported not just in terms of fairness but capital and wealth. In the gradual convergence also—given the positive effect on private saving— scenario, by 2030, half of the global stock of capi- in terms of efficiency. In Mexico, for instance, tal—$158 trillion (in 2010 dollars)—will reside augmenting demographic projections with educa- in the developing world (fi gure O.2). Th is will tional attainment yields a household saving rate mean an acceleration of the shift in the global in 2050 that is almost 5 percentage points above distribution of national wealth. The bimodal a scenario that does not take into account the distribution of national wealth per capita, which impact of changes in education (figure O.3). currently displays a large “hump” of middle- income countries and a smaller peak of a number Sub-Saharan Africa will be the only region of high-income countries, may well become more not experiencing a decline in its saving rate, balanced as more middle-income countries build while East Asia and Eastern Europe and their per capita national wealth. Central Asia will record the largest reductions The more-equal distribution of capital and Differences in the timing of demographic transi- wealth between countries does not imply, however, tions imply that saving, factor prices, and income 4 Overview Global Development Horizons FIGURE O.2 Developing countries will represent more than half of global capital stocks by 2030 in the gradual convergence scenario, compared with about a third in 2010 a. Capital stocks, 2010 b. Capital stocks, 2030a $3.8t $4.6t $4.6t $1.7t $8.8t $11.0t $17.5t $89.1t $23.3t $14.5t $21.5t $115.8t $8.0t $158.2t High-income L. Amer. & the Mid. East S. Asia Sub-Saharan E. Asia & E. Eur. & countries Caribbean & N. Africa Africa Pacific Cent. Asia Source: World Bank projections. Note: Capital stocks are calculated using a perpetual inventory method with a constant depreciation rate of 5 percent. a. Panel b projections assume a gradual convergence scenario of annual world economic growth of 2.6 percent over the next two decades, of which developing (high-income) growth will average 5.0 (1.0) percent annually. distribution will all be affected differently across older than the average person, while those ben- countries. In aging economies, labor becomes efiting from higher wages will tend to be younger. scarcer with respect to capital, thus raising real Set against these heterogeneous demographic wages relative to the return to capital. This diver- pressures is the role of economic growth. Increases gence in the path of factor prices has intergenera- in productivity, resulting from technological tional distribution effects: asset holders adversely advancement as well as institutional improve- affected by lower capital returns will tend to be ments, can raise the amount of output produced with a given amount of input. For some countries and regions, accelerated growth associated with FIGURE O.3 Increased earning power will be the greatest productivity advancements will increase national driver of saving by Mexican households saving, with younger cohorts of the population receiving larger lifetime per capita incomes such 18 that they increase their saving more than older 17 cohorts reduce theirs. Household saving rate, % 16 For these reasons, the stability of investment and saving rates at the global level masks sub- 15 stantial heterogeneity in country experiences. 14 Demographic + education effect Indonesia, Japan, and the Russian Federation, 13 for example, will experience sharp reductions in 12 the working-age share of their populations and Demographic effect thus will see their saving rates fall significantly 11 through 2030. In contrast, economies such as 10 Brazil and Mexico will experience a more gradual 2010 2015 2020 2025 2030 2035 2040 2045 2050 decline. Sub-Saharan Africa, with its relatively Source: World Bank projections using household survey data. young and rapidly growing population and robust Global Development Horizons Overview 5 rate of growth due to productivity catch-up, will account for almost half of all global manufactur- be the only region not experiencing a significant ing investment. decline in its saving rate over the time horizon The heterogeneity in investment patterns considered here. In absolute terms, however, sav- across countries will be more pronounced in the ing will continue to be dominated by Asia and the rapid convergence scenario. With financial devel- Middle East: in the gradual convergence scenario, opment and institutional quality advancing at a China will save far more, in absolute terms, than faster pace under this scenario, developing econo- any other country in 2030—$9 trillion in 2010 mies in general will attract an even greater share dollars—with India a distant second with $1.7 of global saving. Some economies, such as China trillion, having surpassed the levels of Japan and and India, will have investment rates of around 1 the United States in the 2020s. percent higher in the rapid convergence scenario The aging of the population will bring than in the gradual convergence scenario by increases in age-related expenditures such as pen- 2030. Economies such as Brazil and Indonesia, sions and health services, putting pressure on on the other hand, will see slightly lower rates public finances. In China, for example, total age- of investment in the rapid convergence scenario related public expenditures would increase by 5.4 (although their investment levels will be unam- percent of GDP between 2010 and 2030 if the biguously higher) because their output will grow generosity and coverage of the pension, health- even faster than investment. care, and education systems were to remain con- The future patterns of investment will entail stant; if China were to converge with U.S. gen- not only geographic shifts but also sectoral ones. erosity and coverage levels, age-related public As countries become richer, demand shifts toward expenditures would rise by 18.6 percent of GDP. services. However, whether investment in the ser- vices sector rises concomitantly depends on the China and India will account for the majority capital intensities of services relative to the other of global investment, and investment will sectors. In the gradual convergence scenario, ser- shift toward the services sector, especially in vices as a share of total investment in developing infrastructure countries will grow from 57 percent to 61 per- Strong economic growth will underpin China’s cent. Th is shift will likely be realized alongside leadership in global gross capital formation over demographic changes that will increase demand the next two decades. In the gradual conver- for educational, health, and infrastructural ser- gence scenario, China will account for 30 per- vices, as well as a global move toward greater cent of global investment by 2030. Elsewhere trade in services and a larger share of services in the developing world, robust growth will be being embedded in tradable goods. associated with high investment rates as well. A strong flow of investment into services will Brazil, India, and Russia, together, will account be evident in infrastructure. Measured in terms for more than 13 percent of global investment of value, infrastructural needs will be largest in in 2030, more than the United States. Among developing East and South Asia (figure O.4) in high-income countries, relatively strong institu- light of population growth pressures and rapid tions and continued technological advantage will increases in per capita income. Even exempting mean that investment rates remain fairly stable, at maintenance costs and replacement investments, around 17 percent of output. In terms of volumes, fully meeting anticipated population coverage investment in the developing world will reach targets will result in a global infrastructure bill for $15 trillion (in 2010 dollars) versus $10 trillion developing countries that amounts to about $866 in high-income economies. As is the case for sav- billion annually in 2030 (in 2010 dollars), consti- ing, China and India will be the largest investors tuting about 10 percent of all services investment among developing countries—the two countries in developing countries. In Sub-Saharan Africa, combined will represent 38 percent of the global where infrastructure needs are especially acute, gross investment in 2030, versus 40 percent in it will amount to as much as 23 percent of total all high-income countries combined—and will investment. 6 Overview Global Development Horizons FIGURE O.4 Annual infrastructure needs over the next 20 because faster financial development attracts years are likely to be greatest in East and South Asia capital and at the same time moderates private saving. These net capital inflows will not come 350 primarily from the North but from newly indus- trialized East Asian countries, most notably 300 China. The relevance of capital f lows, however, 250 extends well beyond the balance of a coun- try’s borrowing or lending as captured by net inflows or outflows. From a policy perspective, it 200 $ billions is equally important to also examine the future path of gross fl ows. As developing countries’ vol- 150 umes of gross inflows and outflows expand in the future, the potential benefits to these countries 100 are significant: diversification of idiosyncratic national risks, the imposition of greater market discipline on policy making, and the opportunity 50 to supplement domestic saving in ramping up fi xed investment and growth. Gross inflows and 0 outflows can be thought of as trade in fi nancial Sub- E. Asia E. Eur. L. Amer. Mid. S. Asia assets. In this sense, the future path of a country’s Saharan & Pacific & Cent. & the East & Africa Asia Caribbean N. Africa gross capital inflows and outflows depends on (a) the strength of demand for foreign assets; (b) the Water/sanitation Telecoms country’s capacity to supply assets with return Transportation Power and risk characteristics appealing to interna- Source: World Bank projections. tional investors and creditors; and (c) the degree Note: Details of the methodology underlying the projections are given in the Global of frictions—for example, problems of asym- Development Horizons online annex 1.5. metric information—that potentially inhibit this exchange. Developing countries will account for a Institutional improvement and financial greater share of gross capital inflows and market development in the developing world, outflows in the future combined with an environment of rising per- The shifting geographic patterns of saving and ceived risk in high-income economies, appear investment activity will also manifest in pat- set to remove advanced countries’ monopoly terns of global capital f lows. In the gradual on supplying high-quality assets. Encouraged convergence scenario, much of the developing by improvements in the business environment, world will run moderate, gradually attenuat- solid economic growth, and demographic trends ing deficits. This pattern will predominate supportive of growing consumer demand, inves- among countries that are relatively early in their tors have shown greater interest in developing demographic transitions and have significant countries far beyond large emerging markets, scope for fi nancial market development, which as demonstrated perhaps most clearly in recent tends to moderate saving and at the same time years by the growing level of capital inflows to boost investment. For example, India and, on Sub-Saharan African countries. In the future, aggregate, Sub-Saharan Africa fit this descrip- improvements in institutional factors will co- tion well; they will run current account defi- evolve with ongoing regional and global integra- cits averaging 2.4 percent and 3.2 percent of tion of developing countries’ fi nancial markets, GDP, respectively, over 2010–30 in the gradual rendering developing countries much more sig- convergence scenario. Under the rapid conver- nificant sources, destinations, and potentially gence scenario, this pattern will be even starker also intermediaries of global capital flows. Global Development Horizons Overview 7 The scenario analysis estimates that develop- grow to become significant at the global level. ing countries will account for 47–60 percent of Developing countries’ financial markets will global capital inflows in 2030, up from 23 per- play a much greater role in intermediating capi- cent in 2010 (figure O.5). Relative to develop- tal flows than they do today, both in their role ing countries’ GDP, however, the increase is as intermediaries of South-South flows and in less dramatic: inflows are foreseen to represent their intermediation of capital flows to and from 6–11 percent of their GDP in 2030, versus 7 the North. Reserve accumulation will likely percent in 2010. There will also be clear changes decline as a share of developing countries’ out- in the regional distribution of capital f lows. flows, displaced by private flows as financial mar- Disaggregating the projected increase in capi- kets develop and as exchange rate regimes move tal inflows to the developing world by country toward greater flexibility. shows that China will be an important part of the story, but by no means will it be a China story— Complex policy challenges will arise from nor a BRICs (Brazil, Russia, India, and China) changing patterns of saving, investment, and story—alone. By 2030, no single country will capital flows attract as great a share of global inflows as the The patterns of saving, investment, and capital United States or the Euro Area does today. flows through 2030 will affect economic con- Moreover, small and medium-size developing ditions from the household level to the global countries will collectively matter much more in macroeconomic level, with implications not only the global economy than they do today, particu- for national policy makers but also for interna- larly in terms of their role in global financial mar- tional institutions and policy coordination. The kets and in driving capital flows. rapid growth in cross-border capital flows in the With developing countries accounting for decades ahead will make the world’s economies nearly half or more of both global gross inflows more integrated than at any time in history. Eight and outflows under both scenarios, the South- main policy messages stem from the three chap- South component of gross capital f lows will ters in this report: FIGURE O.5 By 2030, nearly half or more of gross capital inflows will likely go to developing countries a. Gradual convergence scenario b. Rapid convergence scenario 25 25 Gross capital inflows, $ trillions Gross capital inflows, $ trillions 20 20 15 15 10 10 5 5 0 0 2000 2005 2010 2015 2020 2025 2030 2000 2005 2010 2015 2020 2025 2030 Advanced countries Developing countries Sources: World Bank projections, supplemented with calculations using data in the IMF International Financial Statistics database. Note: Infl ows are depicted in nominal dollars, assuming a constant 3.5 percent world infl ation rate, where infl ation is based on the 2003–07 fi ve- year average from the World Bank Global Economic Monitor database. Gradual and rapid convergence refer to projected scenarios concerning the relative pace of convergence between developed and developing economies in terms of income and structural changes. 8 Overview Global Development Horizons National policy makers seeking to support invest- pose a major challenge. Infrastructure needs will ment activity in their economies should concentrate have to be financed in a challenging environment their eff orts on establishing a favorable investment in which demographic pressures will exacerbate climate. The importance of financial development public sector funding difficulties. Infrastructure and institutional quality in attracting investment financing is particularly challenging given the from the global pool of saving, amply demon- long-term nature (and associated risks) of these strated by the difference in investment patterns investments. Moreover, as developing countries in the rapid versus gradual convergence scenarios, constitute a larger and larger share of the world underscores the importance of structural and economy, they may outgrow traditional sources institutional change for countries to effectively of infrastructure fi nancing, such as lending by compete in the future world economy. Taking international financial institutions. To meet their steps to improve the development of the financial future infrastructure financing needs, policy sector and adjusting policies to raise the overall makers in developing countries will need to lever- quality of governance can be effective for coun- age private sector fi nancing through public-pri- tries seeking to sustain high rates of private sector vate partnerships as well as tap structured financ- investment. Th is is not to say that public invest- ing from global capital markets. ment is irrelevant, since there may be a role for Government will have to sustainably manage intervention where market failures are clear and public finances with an eye toward the forthcoming where social returns are especially large. If policy demographic changes. A large increase in the share makers do decide to pursue the interventionist of population past working age, in combination route with regard to investment policy, attention with increased demand for health-care services should be paid to the institutional design. High in the later years of life that will come with ris- levels of accountability should be maintained, ing incomes, will strain budgets. The shift in the and interventions should have clear sunset clauses composition of public expenditure toward health to better align private sector incentives in an envi- care and pensions will be offset only margin- ronment of active private sector competition. ally by the expected impact of aging on educa- Policy makers will have to recognize the forth- tion spending. Furthermore, in most developing coming increasing demand for services and facili- countries, the scope for decreases in non-age- tate the needed accompanying investment. In many related expenditures is quite limited. Complex developing countries, governments currently policy challenges will arise from efforts to keep engage in policies that, directly or indirectly, the public burden of health care and pensions favor investment in agriculture and manufactur- under control while limiting the decline in ben- ing while maintaining a protectionist stance in efits and services. In the face of unprecedented the services sector. Many economies, for exam- aging, for many developing countries, the public ple, explicitly prohibit foreign direct investment pension and health-care models pursued in the (FDI) in certain “sensitive” or “strategic” services. past will no longer be viable options. Looking Such policies will become an increasing burden forward, part of the policy challenge will be to on efficiency in a global economy where services transition to systems with greater participation by account for a larger share of output. Furthermore, private markets. if targeted interventions in investment are Demographic shifts due to changes in household deemed desirable, the services sector is likely to structure will increase the importance of financial offer the most social returns in areas such as edu- markets in providing for income support during old cation, health care, and infrastructure. Even in age. As incomes rise, household size tends to fall manufacturing and agriculture, there are subin- as workers are more geographically mobile and dustries where positive externalities may poten- older individuals are more able to live indepen- tially be large; these include areas such as the dently on their accumulated savings. Alongside production of “green” technology components or this reduction in household size tends to come a tropically adapted seeds. profound transformation of the old-age support For developing countries over the next 20 years, structure away from an informal, multigenera- the issue of financing for infrastructural projects will tional household system to more formal private Global Development Horizons Overview 9 pension or public social protection systems. amounts of capital are transferred among devel- Greater reliance on privately fi nanced pensions oping countries, South-South monetary policy rather than household savings and transfers dur- coordination will become more critical in pro- ing old age has the potential to improve welfare, moting stable financial and macroeconomic con- assuming that the elderly and their children all ditions in these countries. In particular, there prefer to live separately. Another benefit is that is capacity for greater regional monetary policy using private fi nancial markets to intermediate spillover from large emerging economies such as pension savings can increase financial depth and Brazil and Russia. At the global level, greater use contribute to development. However, shifting of the renminbi could considerably strengthen from dependence on the income of family mem- the impact of China’s monetary policy on the rest bers to dependence on financial institutions also of the world, partially eroding the dominance of underlines the importance of strong regulation U.S. and Euro Area monetary policy. For small to limit fraud and excessive risk taking among and medium-size developing countries, a world financial intermediaries. with a multipolar currency hierarchy would Policy makers in developing countries have a cen- mean that developing countries will become less tral role to play in boosting private saving through affected by monetary policy spillovers from any policies to raise educational attainment, especially one country. This could be stabilizing on a global for the poor. This is likely to reduce the saving level because liquidity shocks will be more diver- concentration observed across household groups sified, but it could also become more difficult to within countries. Indeed, such concentration also assess the timing and extent of monetary policy has negative implications for economic mobil- spillovers, requiring greater monetary policy ity and thus for the political and social cohesion coordination. Regardless of the currency compo- essential for growth. Not only do high-income sition of capital flows in the future, the increasing households tend to save a greater proportion share of global flows going to and from devel- of their incomes than low-income households, oping countries indicates that these countries but they also account for the bulk of saving in should have a larger role in management of capi- countries at various stages of development and tal flows at the international level, within both demographic transition, although to different bilateral and multilateral organizations. degrees. In countries with high economic mobil- Policy makers will need to prepare for a greater ity, the relationship between low savings and role of capital markets in international financial low income could reflect efforts to smooth con- intermediation and promote the development of sumption by households experiencing temporary domestic capital markets. Looking forward, as income losses. Unfortunately, a similar correla- gross capital inflows and outflows grow in scale, tion is observed across households grouped by their composition in terms of portfolio, bank educational attainment, a proxy for permanent lending, and FDI will also grow in importance. income and thus a more stable condition than the This is because these different types of flows have position in the income distribution at a point in different implications for macro stability. It is time. Consistently, the least educated group in a likely that, globally, capital markets will inter- country has low or no savings, suggesting that mediate a growing share of flows in the future those individuals have an inability to improve and that banks will account for less. Bank lend- their earning capacity and, among the poorest, to ing tends to be highly procyclical and gener- escape a poverty trap. ally less supportive of risk sharing than FDI or The course of global monetary and financial pol- equity portfolio investment. However, total (debt icy making will need to be adjusted as developing and equity) portfolio investment has historically countries become responsible for an expected half or been even more volatile, in relative terms (that is, more of the world’s capital outflows. Both the grad- adjusting for the smaller magnitude of this com- ual and rapid convergence scenarios envisage that ponent of developing countries’ inflows), than developing countries will account for a substan- bank lending. Moreover, as households and firms tially increasing share of the world’s gross capital in developing countries increasingly demand inflows and outflows through 2030. As growing not only greater access to credit but also greater 10 Overview Global Development Horizons choice and variety in financial assets and services, of investment and saving—and the fact that domestic financial markets will have to compete income, saving, and investment affect each globally in terms of both their structure and their other—a structural model is needed. This is depth. Although the many efforts under way to why, in this report, a global computable general improve regulation of the international bank- equilibrium (CGE) dynamic model is adopted ing sector will remain highly relevant, policies as the main analytical tool applied across all the should also be designed to accommodate—and chapters. In addition, other tools—such as panel in some cases actively promote—the development data econometrics—are used (a) to estimate key of domestic capital markets. At the same time, parameters for the global CGE model, and (b) to authorities should monitor the composition of complement the results obtained from it. capital flows; more broadly, they should develop Given the focus of the methodological regulatory institutions to be forward-looking approach on capturing the impact of the main and ready to adapt to potentially destabilizing determinants, the potential effects of some other changes in the composition of capital flows. long-term trends and persistent economic shocks are omitted from the analysis. For example, as dis- cussed in chapter 1, crises (whether commodity- based, financial, or environmental in nature) may Modeling the global dynamics engender increased uncertainty that results in lon- of investment, saving, and ger-term effects on investment. Other elements— capital flows for example, changes in habits arising from shifts The main objective of this report is to identify in cultural factors behind saving behavior, or emerging trends of investment and saving. This changes in the global pattern of migration and is achieved not in terms of a set of forecasts but remittances—also remain unmodeled. rather by asking a series of “what if ” questions It should also be noted that there are numer- and by building scenarios. For example, the report ous potential outcomes of future changes in pro- addresses the question of what are the conse- ductivity, ensuing growth patterns, and accom- quences for global saving if aging will hasten and panying policy changes. Although there is broad will continue to exert negative pressure. Or what agreement on demographic projections, there is will happen to the demand for capital if produc- no consensus on the exogenous values governing tivity catch-up accelerates in developing countries. productivity changes or on the correct param- A simple approach would just rely on corre- eterization of saving functions or equations of lations and, by extrapolating the trends of key demand and supply of capital goods. Thus, even determinants such as aging and productivity with sophisticated models, growth, investment, growth, infer their impact on saving and invest- and saving rates for any specific country or region ment. Thus, in the first “what if” question above, are subject to a large margin of error. With this in this simple approach would suggest that global mind, the main advantage of a model-based sce- saving will decline. However, aging does not hap- nario analysis is that it provides an opportunity pen in isolation, and other factors counterbal- to explore the interaction among broad trends ance its negative impact. Likewise, faster rates rather than providing exact forecasts. of productivity growth will—for a given popula- tion growth rate—translate into faster economic A global dynamic general equilibrium growth and higher per capita incomes; higher framework incomes, in turn, will affect the pace of fi nan- The global CGE model—which is a modified cial development and institutional improvement. version of the World Bank’s Linkage model, a These outcomes reinforce the positive impact that dynamic model—comprises 17 country-regions, faster economic growth alone has on attracting 7 sectors (encompassing agriculture, manufac- investment financing. turing, and services), and 3 factors of production To account for the direct and indirect effects (capital, and skilled and unskilled labor). The of multiple factors on the emerging trends version of the model used here relies on the most Global Development Horizons Overview 11 recent Global Trade Analysis Project (GTAP) countries is somewhat wider: 0.7–3.5 percent dataset, whose base year is 2007. Scenarios are until 2015 and constant thereafter. There is sig- developed by solving for a new equilibrium in nificant variation in TFP growth across develop- each subsequent year through 2030. ing countries, ranging from above 3.5 percent in At its core, Linkage is a neoclassical model China (in line with Bosworth and Collins [2008] with aggregate growth, saving, and investment estimates) to around 1.5 percent in Sub-Saharan endogenously determined and predicated on Africa. assumptions regarding key exogenous determi- The growth in the labor force is derived from nants such as productivity changes, demographic the United Nations age-specific population pro- shifts, fi nancial market development, and insti- jections assuming no changes in participation tutional improvement. Unlike simpler growth rates. According to these population projections, models, however, Linkage has considerably more the demographic transition is asynchronous structure: First, it is multisectoral. This allows for across countries in several respects: the timing more complex productivity dynamics, including of its start, the speed with which the transition differentiating productivity growth between agri- unfolds, and the relative sizes of bulge cohorts. culture, manufacturing, and services and picking In the current version of the model, there is no up the changing structure of demand (and there- differentiation between the growth of skilled and fore output) as growth in incomes leads to a rela- unskilled workers. tive shift into manufactures and services. Second, The other crucial dynamics of the model deal it is linked multiregionally, allowing for the influ- with the accumulation of physical capital, which ence of openness—via trade and finance—on results from the interaction of global saving sup- domestic variables such as output and wages. ply, domestic investment demand, and interna- Third, Linkage includes a set of equations for tional capital allocation (figure O.6), as follows: capturing saving and investment behavior. A full description of the Linkage model is • In each country, saving behavior, in accor- available in Van der Mensbrugghe (2011), and dance with a standard life-cycle approach, this overview briefly describes the equations gov- depends on demography and per capita erning the dynamics of investment and saving as income growth; additional determinants well as the main assumptions concerning the pro- include fi nancial development and social jected paths of exogenous determinants. protection systems. Starting with the latter, the key exogenous • Investment demand is obtained from capi- determinants are productivity and demograph- tal demand, which, in turn, is derived from ics. Productivity change is derived from a com- sector-specific production functions in a bination of estimates and is also subjectively setting of perfectly competitive profit-max- fine-tuned. Agricultural productivity is assumed imizing firms. Investment demand thus to be factor-neutral and exogenous and is set to relates positively to output and negatively estimates from empirical studies (for example, to rental rates. Martin and Mitra 1999). Productivity in man- • The global pool of saving is allocated across ufacturing and services is labor-augmenting countries following a function represent- (Harrod-neutral technical change); it is skill-neu- ing the global financing of investment. This tral but sector-biased, with productivity growth responds to relative rental rates but also higher in manufacturing than in services. Th is to country-specific economic growth and gives rise to a long-term rate of total factor pro- structural factors such as financial develop- ductivity (TFP) growth in the range of 0.1–0.2 ment and the quality of institutions. The percent for the high-income countries in the global financing of investment also cap- gradual convergence scenario, which lies toward tures home bias by including a lagged term. the low end of the Bosworth and Collins (2003) estimates but is consistent with the trends in the For each country, net capital flows make up early and mid-2000s. The range for developing the difference between the level of financing and 12 Overview Global Development Horizons FIGURE O.6 Schematic diagram describing interactions between saving, investment demand, and investment financing Economy A Economy B Economy Q Sector-specific Sector-specific … Sector-specific profit-maximizing profit-maximizing profit-maximizing factor demands factor demands factor demands N a t Investment Investment … Investment i demand demand demand Economy-specific o rental rate n equalizing investment demand a and financing l Investment Investment Investment … financing financing financing G l Growth, structural variables o Capital b flows Global pool of saving a l N … Saving Saving Saving a t i Per capita growth, o demography, n structural variables a l Source: World Bank. domestic saving. Thus the model allows for global will be bid up. Th is will attract capital flows to capital mobility, but only imperfectly because the the country, mitigating the upward pressure on fi nancing of investment responds to rental rate rental rates but not eliminating it fully. differentials with a fi nite elasticity and depends Summing up, in the Linkage model, sav- on other factors as well. In this setup, net capi- ing, investment, output, and income as well as tal flows between countries remain within limits relative factor and good prices are simultaneously that are reasonable given their historical range, determined. However, for any specific country or and there is no guarantee that rental rates will region, income growth rates, investment and sav- equalize across countries. For example, a country ing rates, and net capital flows generated by the benefiting from productivity catch-up will expe- model are subject to a margin of error because the rience stronger capital demand, and rental rates resulting trends in these variables depend on: Global Development Horizons Overview 13 (a) Assumptions on the path of exogenous vari- accounting for the possible endogeneity of eco- ables, specifically on productivity, demogra- nomic growth, faster growth is positively related phy, financial sector development, and qual- with both saving and investment financing. ity of institutions The coefficients on the structural variables (b) Parameterization of the equations and, more also accord with a body of empirical and theo- explicitly, the elasticity of the saving and retical work. For instance, financial development investment rates with respect to aged depen- is positively associated with the investment rate dency (only for saving), income growth, (more sophisticated financial markets are able to fi nancial sector development, and quality of lend more readily to firms for investment pur- institutions poses) (Benhabib and Spiegel 2000) but nega- tively related to the saving rate (households with Some of the key assumptions concerning easier access to consumer credit need to save less point (a) have already been described above. By for consumption smoothing) (Loayza, Schmidt- considering two scenarios, the report takes into Hebbel, and Servén 2000). account uncertainty in the future trends. In the rapid convergence scenario, productivity is exog- Complementary analytical approaches: enously raised by a factor of 50 percent for all enriching the CGE results developing countries, and the rate of growth of The CGE analysis requires a number of simpli- structural variables is boosted so that they reduce fying assumptions and cannot generate some of by 25 percent the distance between their cur- the results that are of interest to policy makers rent levels and that of the United States in 2030 and development practitioners. For example, the (assumed to be the frontier). With regard to point population age structure of a country (or groups (b), a key assumption made is that the coefficients of countries) is approximated by a simple old-age for the economic and structural variables in the dependency ratio; and, given the fact that the saving and investment financing equations are model works with a representative household, it indeed stable. cannot provide any insights on different saving behavior for different categories of households. Complementary analytical approaches: Therefore, complementary analytical tools are estimating parameters used to address these issues. The parameterization of the investment demand In the case of saving, a complementary function draws on the large, and by now stan- approach studies the future of saving by consid- dard, empirical literature that estimates the ering how demography and income growth will parameters of the production function. The lit- affect saving at the household level. Examining erature is, however, somewhat more circumspect the issue from a microeconomic perspective pro- with regard to parameter estimates of the saving vides a more realistic and nuanced view of the and investment financing functions. Accordingly, likely evolution of saving, although at the price of this report performs estimates for the elastici- considering a limited set of countries because of ties corresponding to the described determinants data availability. While the household-level anal- using econometric techniques designed to limit ysis confirms the broad conclusions of the CGE the impact of endogeneity in the regressors as analysis, the micro-level data also expose the well as measurement error (details are given in complexities of the interaction among aging, sav- online annex 1.5).1 ing, and growth. The microanalytical approach The signs of these estimated coefficients are is described in more technical detail in online consistent with underlying theory. For example, annex 2.1. the coefficient on demography in the saving equa- In the case of capital flows, the CGE model- tion—measured by the aged dependency ratio— generated saving and investment scenarios have is negative and statistically significant, which direct implications for net capital flows (essen- implies that economies with older populations tially countries’ saving-investment differentials), tend to save less. As another example, even after but a separate model is required to complement 14 Overview Global Development Horizons these findings with scenarios for countries’ gross ———. 2008. “Accounting for Growth: Comparing volumes of capital inf lows and outf lows. An China and India,” Journal of Economic Perspectives econometric model is specified to estimate gross 22 (1): 45–66. inflows (outflows can be backed out from these IMF International Financial Statistics database, Inter- national Monetary Fund, Washington, DC. http:// and net flows), drawing on the literature to iden- www.imf.org/external/data.htm. tify key determinants, and also controlling for Loayza, Norman V., Klaus Schmidt-Hebbel, and Luis any country-level effects that do not vary over Servén. 2000. “What Drives Private Saving Across time and for global shocks across time. Two sce- the World?” Review of Economics and Statistics 82 narios for gross flows are generated by fitting pro- (2): 165–81. jected paths of the independent variables to the Martin, Will, and Devashish Mitra. 1999. “Productiv- estimated equation, which correspond to the two ity Growth and Convergence in Agriculture and CGE model-generated scenarios (details are given Manufacturing.” Policy Research Working Paper in online annexes 3.3 and 3.4). 2171, World Bank, Washington, DC. Spence, A. Michael. 2011. The Next Convergence: The Future of Economic Growth in a Multispeed World . New York: Farrar, Straus, and Giroux. Note Van der Mensbrugghe, Dominique. 2011. “Linkage 1. The online annexes are available at http://www.world Technical Reference Document, version 7.1.” World bank.org/CapitalForTheFuture. Bank, Washington, DC. World Bank Global Economic Monitor database, World Bank, Washington, DC. http://data.world References bank.org/data-catalog/global-economic-monitor. World Bank World Development Indicators database, Benhabib, Jess, and Mark M. Spiegel. 2000. “The Role World Bank, Washington, DC. http://data.world of Financial Development in Growth and Invest- ba n k .org /d at a-c at a log /world-de velopment- ment.” Journal of Economic Growth 5 (4): 341–60. indicators. Bosworth, Barry P., and Susan M. Collins. 2003. “The Empirics of Growth: An Update.” Brookings Papers on Economic Activity 34 (2): 113–206. 1 The Emerging Pattern of Global Investment A SNAPSHOT IN 2030 WILL REVEAL expected to increase gradually over time. As labor a world in which more than two-thirds productivity in developing countries catches up of all global investment and half of the with the higher levels that exist in high-income global accumulated capital stock will be in devel- countries, the global distribution of capital per oping economies. Th is is in sharp contrast with capita will become more equal. recent history. Currently, almost 70 percent of With much higher productivity growth in the world’s capital stock resides in the developed developing countries, their investment rate has world, and in 2000 developing countries’ share to be substantially higher than in high-income in global investments was only about 20 percent. countries for the capital stock to keep pace with Projections presented in this chapter indicate potential output. Today, developing countries that, by 2030, global aggregate investment activ- collectively invest slightly above 30 percent of ity will far and away reside in China—with 30 GDP, a rate almost twice that of high-income percent—while India and Brazil (7 percent and 3 countries, which currently stands at about 17 per- percent, respectively) will account for shares com- cent. The high investment rates—together with parable to the United States and Japan (11 per- the rising share of developing countries in global cent and 5 percent, respectively). Developing Asia GDP—will increase developing countries’ share will collectively hold capital stocks exceeding 55 in global investment to two-thirds before 2030. percent of the entire developed world’s. However, this outcome will not mechanically Th is shift in investment activity toward the materialize, but will require two conditions to be Global South coincides with rapid catch-up fulfi lled: First, productivity growth and sectoral growth that began during the 1990s as devel- shifts must create enough investment opportu- oping countries integrated into global markets, nities. Second, both domestic and international underwent structural transformations, and investors must be willing to finance these invest- improved their institutions. Assuming a continu- ments, allocating two-thirds of every dollar of ation of needed reforms, this catch-up process is global savings to developing countries. These expected to carry on over the course of the next two conditions are analyzed in this chapter. The two decades. In fact, the shift in allocation of analysis of the first one relies heavily on standard the global capital stock corresponds closely to factor demand equations in a global general equi- a similar rise in developing countries’ share of librium model. The analysis of the second con- global gross domestic product (GDP). Currently, dition brings in econometric analysis regarding roughly 70 percent of global GDP is produced the historical behavior of investors, who react not in high-income countries, and that share will only to growth opportunities and yields, but also decline to around 50 percent by 2030. Indeed, to institutions and other factors, often encapsu- the capital-to-GDP ratios of developing coun- lated by the term “investment climate.” tries as a group and high-income countries as a That the discussion has shifted to how much group differ very little, even as the ratio can vary developing countries will contribute to, and substantially across individual countries: in both reshape, global investment trends is a testimony groups, the size of the capital stock is about 2.5 to the great strides that the developing world has times the size of annual GDP, and this ratio is made over the past half century. Estimates by a Global Development Horizons 17 18 The Emerging Pattern of Global Investment Global Development Horizons group of experts convened by the United Nations of global investment since 2000 has been due placed the total capital requirements of develop- in large part to robust growth in those coun- ing countries in 1950–60 at $19 billion, of which tries. This observation is consistent with they believed only a fraction would be met by optimal factor demand calculations in a domestic savings, leaving an “investment gap” of computable general equilibrium (CGE) $14 billion (UN 1951). Pioneers in development model. It is also confirmed by econometric economics argued over whether structural rigidi- analysis of how investors allocate resources ties in developing countries could ever be over- over countries. However, the historical cor- come such that investment could be productively relation between growth and investment translated into output, engendering an industrial ratios found in the latter analysis indicates takeoff (Meier and Seers 1984). that growth alone will be insufficient to The main messages of this chapter are the fully satisfy optimal factor demand. Th at following: means improvements in other driving forces are needed to attract sufficient capi- • Since 2000, there have been three notable tal. One of those forces is fi nancial devel- changes in the pattern of global invest- opment, or the maturity of the financial ment activity: a shift in global investment sector. A 10 percent increase in a standard toward the developing world, a shift toward measure of fi nancial intermediation—the greater manufacturing investment glob- ratio of private credit to GDP—is associ- ally, and a slow, but definitive, redistribu- ated with a 1–5 percent increase in invest- tion of capital stocks toward the develop- ment. If such improvements are insuffi- ing world . Although a nontrivial share cient, yields relative to those abroad will of these phenomena can be attributed to have to rise in fast-growing economies to large economies such as China and India, attract sufficient capital. significant increases in investment rates • One major area of concern in developing have occurred elsewhere in the developing countries regards infrastructure needs. Even world as well. exempting maintenance costs and replace- • The future structure of production will imply ment investments, fully meeting antici- also a global shift toward investment in ser- pated population coverage targets will vices. This shift is the natural consequence result in a global infrastructure investment of (a) increased shares of services in econo- bill for developing countries that amounts mies with higher per capita incomes, (b) to about $866 billion annually in 2030 demographic changes that will increase (measured in 2010 dollars). Although this demand for educational and health services, amounts to only around 3 percent of all and (c) a larger share of services embedded developing-world investment in 2030, its in tradable goods. Between 2010 and 2030, financing is particularly challenging given services investment will rise from 57 per- the long-term nature (and associated risks) cent to 61 percent of the total investment of these investments. profi le in developing countries, and from • Despite many common changes in investment 75 percent to 78 percent in high-income patterns, the patterns themselves can signifi- economies. Th is rise is expected to occur cantly diff er across countries. There are pre- despite ongoing rapid growth in manu- dictable patterns; for example, low-income facturing investment, especially in several countries invest primarily in agriculture, lower-middle-income countries and regions and middle-income countries invest heavily (such as India and the lower-income econo- in manufacturing and increasingly in ser- mies of Southeast Asia). vices. But there are notable exceptions. For • The overall relationship between economic example, Vietnam (which until recently growth and investment is strong and signifi- was a low-income country) invested a dis- cant, and developing countries’ growing share proportionately large share of all fixed Global Development Horizons The Emerging Pattern of Global Investment 19 investment in manufacturing. Countries any given country will depend, however, have also varied in their overall capital- on the increase in its investment relative to to-GDP ratio, with the natural resource- its output; in China and India, for exam- rich economies of Brazil and the Russian ple, investment rates will rise on the order Federation requiring less capital per unit of about 1 percent, while economies such of output than China and India. Finally, as Brazil, Mexico, and Indonesia will see although the private sector has tended to small declines in their investment rates, of account for a larger proportion of invest- 0.4–1.1 percent. ment than the public sector in most upper- • Policy makers seeking to support investment middle-income economies, the economies activity in their economies should concentrate of China and the Middle East have been their eff orts on establishing a favorable invest- notable exceptions. ment climate that supports private sector • Despite the rising share of developing coun- investment activity, including policies aimed tries, the global investment rate is expected at improving financial sector development to remain relatively stable. If investment and policies to raise the overall quality of rates in individual countries would remain governance, especially concerning the rule constant, the global investment rate would of law. To the extent that direct interven- rise because of the growing share of the tion is warranted, it should be in areas developing world. However, investment where market or government failures are rates in virtually all individual countries clear and where social returns are especially are expected to fall. That drop is largely large. If policy makers do decide to pursue associated with slowing growth in produc- the interventionist route with regard to tion potential, reflecting demographics and investment policy, their efforts may be best the shift toward services (where productiv- placed supporting the upcoming expansion ity growth is lower). The slowdown will in service sector investment, especially in not be uniform across the world. China’s areas such as education, health care, and investment will fall more sharply than else- infrastructure, where spillovers from posi- where in the developing world, while Sub- tive externalities can be especially high. Saharan Africa will actually experience a Insofar as infrastructure is concerned, the rise relative to its recent historical average. trend toward public-private partnerships Th is fall of the investment rate in individ- can offer both additional sources of capi- ual countries compensates for the composi- tal and a disciplinary mechanism to limit tion effect of a larger share of developing potential public sector inefficiencies. countries in the world, leading to a stable global investment rate. • In a scenario where convergence between developing and high-income economies occurs Changing patterns of investment more rapidly, total investment at the global worldwide level will be 7 percent higher than in a sce- There has been a global shift nario where convergence follows recent his- toward greater investment in the torical trends and is far more gradual . In developing world this rapid convergence scenario, aggregate investment in developing countries will Since the turn of the 20th century, the pat- also rise by $2.7 trillion (in 2010 dollars) tern of global investment—measured as gross compared with the gradual convergence capital formation1—has changed significantly. scenario, while investment in high-income Between 1965 and 1999, investment in devel- countries will fall by about $1 trillion rela- oping countries held a relatively constant global tive to the gradual convergence outcome. share—averaging 18.5 percent—but this share Whether investment rates rise or fall in increased dramatically in the first decade of the 20 The Emerging Pattern of Global Investment Global Development Horizons FIGURE 1.1 Gross investment in developing countries has increased in absolute terms (panel a) and as a share of global investment (panel b) a. Total gross investment b. Share of global investment 10 100 Share of global investment, % 8 80 High income High income $ trillions 6 60 4 40 Developing 2 Developing 20 0 0 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 Source: World Bank calculations, using data in the World Bank World Development Indicators database. Note: Shaded area corresponds to the period from 2000 onward, where a break in the series occurred (a simple linear regression on time in the 1965–99 period for devel- oping countries is I = –0.011t + 21.360, where I is total gross investment and t is the year, while that for the 2000–10 period is I = 0.024t – 49.312; the Chow test F = 52.06 is significant at the 1 percent level). 21st century, reaching about 40 percent by 2009 Further, although the investment share of output (fi gure 1.1, panel b); the change is even starker in developing countries has remained strong after when one adjusts for differences in purchasing the crisis, it contracted sharply in the developed power (as further discussed in box 1.1). Moreover, world. Consequently, there is reason to believe this convergence in global investment shares that future saving and investment flows—and has been hastened by the global fi nancial crisis. the associated costs of capital—may in fact sta- Since 2007, investment has fallen more sharply bilize at levels demonstrably different from those in high-income countries than in the developing observed today. world (figure 1.1, panel a). The growing influ- By virtue of sheer size, much of the sharp rise ence of developing countries in the global invest- in developing countries’ share of global invest- ment picture has clearly paralleled the emergence ment can be attributed to China and, to a lesser of developing countries on the world economic extent, India. Moreover, with so much of the stage. change in developing-country investment occur- This convergence in investment shares ring in China and India, neglecting to examine between developing and high-income countries their experiences risks obscuring the important owes as much to increases in the developing dynamics that are taking place there. But even world’s investment rates (figure 1.2, panel a) as in the absence of these two developing-country it does to their larger size in the global economy giants, the share of global investment accounted (figure 1.2, panel b). Although these trends are for by developing countries has risen steadily unlikely to persist indefinitely—the process of since 2000, and currently stands at the high- deleveraging in Europe and the United States fol- est level since the mid-1960s (figure 1.3). Across lowing the 2007–09 crisis will eventually end, developing countries, investment rates have risen and saving and investment rates will ultimately significantly from lows in the 1990s—including reach upper limits in the developing world— countries as diverse as Ghana (20 percent in the investment in developing countries appears likely 1990s to 24 percent in the 2000s), Guatemala (16 to play a greater global role in the future. Indeed, percent to 19 percent), and Mongolia (31 percent the gap between developing and high-income to 35 percent). Looking at the world as a whole, countries’ investment shares of output has never an unmistakable message emerges: the conver- been greater for any period since the mid-1960s. gence of investment performance between the Global Development Horizons The Emerging Pattern of Global Investment 21 BOX 1.1 Different terms, different rates: Purchasing-power adjusted investment vs. investment expressed in national currency In this report, investment ratios are expressed in mar- within developing countries, investment goods are ket prices, as are the underlying optimal capital-output relatively expensive, and investment rates (along with ratios. Investment rates do not depend on the choice of capital-output ratios) turn out to be significantly smaller a specific unit of measurement of the output or invest- when expressed in international PPP units, even if these ment volumes because market prices will refl ect the ratios calculated in market prices are similar. chosen volume units. Nevertheless, it might also be As developing countries catch up, the difference interesting to express investment rates in specific vol- between market prices and PPP units will disappear, ume units—units expressed in purchasing power parity and capital-output ratios in PPP measures will con- (PPP) terms that are the same across countries. verge toward those in high-income countries. In other That PPP-adjusted investment rates tend to be lower words, over time the volume of the capital in develop- in developing countries than rates measured in national ing countries will rise not only as overall productivity market prices has been known for some time (Summers increases but also because the relative price of output and Heston 1991) (figure B1.1.1, panel a). Recent will rise. This important trend can be explained only if research confirms that the difference is primarily attrib- the analysis is conducted in market prices, and using utable to differences in output prices across countries PPP measures of investment may introduce potentially (Hsieh and Klenow 2007), a reflection of the Balassa- significant distortions to the data (Knowles 2001). For Samuelson observation that at low levels of develop- example, the small capital-output differential between ment, lower productivity levels in the tradable sectors high-income and developing countries becomes more relative to nontradables translates into lower price levels pronounced after PPP corrections (figure B1.1.1, panel overall. In contrast, because investment goods are, by b), even though this differential has no bearing on the and large, internationally traded, their price differences investment decisions faced by economic agents within also tend to be small across countries. Consequently, any given country. FIGURE B1.1.1 Differentials in investment rates (panel a) and capital-output ratios (panel b) are greater when measured in PPP terms a. Investment rate differential,1980–2009 b. Capital-output ratio, 2000 6 260 NC Differential Investment rate differential, % (international 250 4 Capital-output ratio, % dollars) PPP 240 2 230 0 220 Differential –2 (national currency) 210 –4 200 1980 1985 1990 1995 2000 2005 High Developing High Developing income income Source: World Bank calculations, using data in the World Bank World Development Indicators database and Penn World Tables (https://pwt.sas.upenn.edu/). Note: PPP = purchasing power parity. The investment rate differential is calculated as the difference between (unweighted) average investment rates for high-income and developing countries, measured in either national currency (NC) or PPP-adjusted international dollars. Capital stocks were calculated using a perpetual inventory method with an assumed constant depreciation rate of 5 percent. Calculations with weighted averages yield similar results. 22 The Emerging Pattern of Global Investment Global Development Horizons FIGURE 1.2 Developing countries’ rising investment rates (panel a) and growing share of global output (panel b) have contributed to their increased share of investment in global output a. Investment rate b. Relative output share 35 15 100 90 Gap in investment rates, % Share of global output, % 80 Investment rate, % 30 10 70 60 25 5 50 40 20 0 30 20 10 15 –5 0 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 High income Gap between high-income Developing and developing countries High income Developing (right axis) Source: World Bank calculations, using data in the World Bank World Development Indicators database. Note: The gap between developing and high-income investment is computed as the difference between the two groups’ investment shares of gross domestic product. FIGURE 1.3 The rising share of developing countries’ developing countries continue to struggle with investment in global output is due to more than just sustaining high levels of investment over time. changes in China and India The upper-middle-income economies of Sub- Saharan Africa, for example, saw an investment Share of investment in global output, % 9 takeoff between 1965 and 1980—peaking at 8 30 percent of GDP in 1978—before falling to annual averages of 15–20 percent (figure 1.4). 7 Similarly, investment growth in major emerging 6 economies such as Brazil, Mexico, and Turkey has been more modest and gradual. 5 Globally, investment booms—character- 4 ized by a fairly sharp increase in an economy’s investment rate over just several years—have 3 been somewhat rare phenomena, although some 2 historical episodes did persist over an extended 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 period. An examination of countries that have Developing Developing Developing without experienced such booms suggests that such spurts without China China and India of investment have not, in general, been associ- Source: World Bank calculations, using data in the World Bank World Development Indicators ated with sustained subsequent economic growth database. (box 1.2). Note: Shaded area corresponds to the period from 2000 onward, where a break in the series occurred (a simple linear regression on time in the 1965–99 period for the full developing Some have questioned China’s heavy reliance country subsample is I = –0.011t + 21.360, where I is total gross investment and t is the year, on investment as an engine of growth, espe- while that for the 2000–10 period is I = 0.024 – 49.312; the Chow test F = 52.06 is significant at the 1 percent level. cially when examined in terms of the relatively low-value-added nature of Chinese manufactur- developed and developing worlds has definitively ing (Bardhan 2010). Consequently, if China— begun and is likely to consolidate in the future. along with other high-investment economies—is The clear upswing in investment in develop- to sustain a “soft landing” for its current high ing countries as a whole is not to deny that many investment rate, observers have argued that its Global Development Horizons The Emerging Pattern of Global Investment 23 investment-centric economic growth model must FIGURE 1.4 Investment rates among Sub-Saharan be ultimately accompanied by structural changes African countries of different income levels have followed to the economy (World Bank 2012a). Such struc- distinct paths tural changes include not only sectoral shifts in 30 investment and production but also more fun- Investment rate, % damental changes in terms of fi nancial market 25 Upper-middle development, its economic and sociopolitical income, Africa institutions, and its financial integration into the 20 Africa average global economy. In sum, it is clear that, since 2000, there has 15 Low income, been an unmistakable shift in the pattern of Africa 10 gross capital formation at the worldwide level, Lower-middle with developing countries becoming increasingly income, Africa 5 important in the global investment picture. This 1969 1974 1979 1984 1989 1994 1999 2004 2009 shift appears to have been led by robust economic growth in the developing world relative to the Source: World Bank calculations, using data in the World Bank World Development Indicators database. developed world and by greater convergence in Note: Investment rates are the unweighted shares for each income group. GDP-weighted investment rates between the two. results are qualitatively similar. BOX 1.2 Investment booms are not always associated with sustained growth Investment booms have historically occurred in a Investment-boom economies also include many broad range of developing and high-income coun- economies that have remained low- or lower-middle- tries (table B1.2.1). However, such booms have not income. Interestingly, China—which has maintained necessarily been accompanied by sustained eco - very high levels of investment since reforms in the nomic growth in the aftermath of the boom. Indeed, 1980s—has not sustained the sorts of investment many economies that have experienced investment surges that would justify the label “boom”: rather, the booms—defined as any period of steady growth in Chinese experience has been one of steady and con- investment rates lasting for at least three consecutive sistent ratcheting up of investment, with occasional years—have experienced subsequent weak growth pauses. Altogether, there have been at least 38 invest- even after the initial postboom period (figure B1.2.1). ment booms over the past half century. TABLE B1.2.1 Investment booms have occurred in a broad range of developing and high- income countries Number of booms Economies 1 Albania; Algeria; Azerbaijan; Bahamas, The; Belarus; Botswana; Chad; Côte d’Ivoire; Cameroon; Eritrea; Iceland; Iran, Islamic Rep.; Kazakhstan; Lebanon; Macao SAR, China; Madagascar; Malawi; Malta; Morocco; Pakistan; Papua New Guinea; Philippines; Saudi Arabia; Serbia; Sri Lanka; Thailand; Trinidad and Tobago; Zambia 2 Jordan; Malaysia; Venezuela, RB 4 Lesotho Source: World Bank calculations, using data in the World Bank World Development Indicators database. Note: Investment booms are defined as any three-year episode where the investment share of gross domestic product increased by at least a total of two percentage points, conditional on an initial investment share of at least 10 percent. (continued) 24 The Emerging Pattern of Global Investment Global Development Horizons BOX 1.2 (continued) FIGURE B1.2.1 Many countries experience weak growth following an investment boom 10 9 8 Average GDP growth, % 7 6 5 4 3 2 1 0 –1 –2 BHS MWI VEN LSO MAR LKA ALB CMR JOR PHL THA MYS DZA CIV MLT BWA IRN ERI Preboom Postboom Source: World Bank calculations, using data in the World Bank World Development Indicators database. Note: Data are for 1960–2010 preboom (postboom) periods, defined as the GDP growth rate in the four-year period just prior to the boom (the four- year period after the end of the boom). For countries experiencing more than one boom, the reported numbers average the rates across all booms. Investment booms are defined as any three-year episode of at least two percentage points’ growth in investment share of GDP, conditional on an initial investment share of at least 10 percent Some of these investment booms have lasted for large part due to natural resources, while some booms fairly long periods. The longest episode was the recent coincided with significant appreciation in real estate Belarussian boom of 2003–10. As is to be expected, (Belarus and Lebanon) or financial asset markets (Macao several fast-growing economies have also experienced SAR, China). Other economies have had booms led by investment booms: these include República Bolivariana the public sector, such as that of Côte d’Ivoire in the de Venezuela in the 1970s and Malaysia and Thailand mid-1970s. Finally, some booms were driven by foreign in the 1990s. Nevertheless, the conditions underlying capital inflows, which may give rise to the boom-bust investment booms differ across economies: Kazakhstan cycles commonly observed in middle-income countries and Saudi Arabia, for example, experienced booms in (Tornell and Westermann 2002). A shift in sectoral distribution shift can be expected to occur in investment pat- toward manufacturing and terns. Indeed, since 2000, capital formation has services is under way gradually moved away from agriculture, and man- ufacturing investment—as a share of global fi xed As an economy’s income level rises, its production investment—has doubled, rising from 11 percent structure tends to move away from the predomi- in 2000 to 22 percent in 2007, even as global nance of agriculture, toward a larger manufactur- investment in agriculture has remained relatively ing base, and eventually to a service-oriented econ- stagnant (of about 9 percent of global fixed invest- omy. As average income levels rise, an analogous ment, roughly half its share in the 1980s). Global Development Horizons The Emerging Pattern of Global Investment 25 FIGURE 1.5 Global manufacturing investment tends to be concentrated in lower-middle-income countries (panel a), with China currently accounting for the vast majority of investment in those countries (panel b) a. Sectoral shares of investment, 1990–2007a b. Global manufacturing investment, 2005–07 HIC UMC LMC LIC LIC 100 UMC LMC Agriculture 75 Percent 50 HIC LMC Manufacturing 25 Services 0 China India 1990–99 2000–07 1990–99 2000–07 1990–99 2000–07 1990–99 2000–07 Vietnam Other Sources: World Bank calculations, using data in the UNIDO Industrial Statistics database; FAO FAOSTAT database; GTAP database; and World Bank World Development Indicators database. Note: HIC = high-income countries, UMC = upper-middle-income countries, LMC = lower-middle-income countries, LIC = low-income countries. a. Shares are the unweighted averages of annual data indicated for each income group. Investment-weighted results are qualitatively similar but overstate the 2000–07 shares for the LMC category because of the large weight of China. Global manufacturing data for 2000–07 averages for LMC include Indian data for only 2005 (the only year data were available). Global agricultural data for 1990–99 averages for LIC include Tajikistan data for only 1995 because of anomalous shares for the later years. This increase in the overall importance of These changes in the sectoral distribution of manufacturing investment notwithstanding, the investment activity will likely be accompanied by global distribution of manufacturing investment gradual improvements in the efficiency of invest- has remained concentrated in lower-middle- ment. Aggregate and sectoral investment figures income countries, a feature that has been rela- often obscure the extent to which investment is tively stable over the past two decades (figure 1.5, actually translated into productive capital, and panel a). Thus, although China has accounted for as economies develop, such allocative efficiency the vast majority of lower-middle-income invest- will probably rise. Estimates of the efficiency of ment in the sector, its share is likely to decline in capital use indicate wide variations, both between the future as lower-middle-income countries— and within countries (table 1.1).2 In China, the such as India and Vietnam (figure 1.5, panel b)— economywide ratio of changes in output for each gradually ramp up their manufacturing bases additional unit of capital deployed (the marginal with added investment. To the extent that future product of capital, or MPK) averaged 22 percent growth in Sub-Saharan Africa remains robust, annually. This is comparable to the MPKs in low-income countries in the continent may also India, Indonesia, and Mexico but is significantly see increases in their manufacturing investment lower than Turkey’s MPK (and higher than shares. Russia’s)—suggesting that, at the economywide As income levels rise in the developing world, level, investment is more efficiently deployed in global investment activity is on the cusp of yet Turkey (and less efficiently in Russia). another major shift in sectoral investment pat- Taking these estimates at face value, the terns. This transition will likely see services fairly low MPK of around 10 percent observed investment rise in both the developed and devel- for China’s and Indonesia’s manufacturing sec- oping worlds, completing the global move away tors (both relative to other economies and within from the primary and secondary sectors and these respective economies) indicates that invest- toward a postindustrial investment landscape ment is less efficiently deployed in manufactur- with the services sector dominating global invest- ing than in the agriculture and services sec- ment activity. tors in these countries. For China, one possible 26 The Emerging Pattern of Global Investment Global Development Horizons TABLE 1.1 There is significant heterogeneity in marginal products of capital, at both economywide and sectoral levels, across developing countries Russian Brazil China India Indonesia Mexico Federation Turkey Economywide 0.17 0.22 0.24 0.26 0.27 0.03 0.48 Agriculture 0.28 0.29 0.19 0.14 0.10 0.03 0.34 Manufacturing — 0.08 0.35 0.11 0.27 — 1.17 Services — 8.78 a 0.19 — 0.24 — 0.19 Sources: World Bank calculations, using data in the UNIDO Industrial Statistics dabatase; FAO FAOSTAT database; GTAP database; and World Bank World Development Indicators database. Note: MPK = marginal product of capital, — = not available. Unless otherwise indicated, data are for 1991–2007. MPKs are computed as the capital output elasticity-weighted ratio of output to capital stock, computed for a sector j (or economywide) in year t from the formula MPKjt = αjYjt /Kjt , where Y and K are gross domestic product (GDP) and the capital stock formation, respectively, and α is computed from the share of compensation to physical capital in that sector (or economywide), excluding natural capital. Sectoral GDPs are computed from sectoral value-added shares of total GDP. Capital stocks are calculated from sectoral investment using a perpetual inventory method with an assumed constant depreciation rate of 5 percent. In most cases, annual data for manufacturing and services are not available for the full period, and reported MPKs are computed from available data. The computation excludes the years 1998–99 (financial crisis) for Brazil; years prior to 1996 (transition) and 1999–2000 (financial crisis) for Russia; 1997–98 (financial crisis) for Indonesia; 1994 (financial crisis) for Mexico; and 1999–2000 (financial crisis) for Turkey. a. Because no data are directly available for investment in services, the fi gures for service sector investment were recovered as the residual of total investment and investment in the other two sectors. Consequently, very large service MPKs (as for China) may be the result of underestimating service sector investment rather than actual high levels of efficiency in the sector. explanation for this finding is that the high levels among developing-country firms. The upshot of capital formation required for the construction of this fi nding is that many potential efficiency of its world-class assembly plants are ultimately gains remain for developing countries to exploit rewarded with relatively less by way of value- as they grow. added output (Koopman, Wang, and Wei 2012). One potential reason for the historical ineffi- In Indonesia, governance problems have under- ciency of investment in many developing coun- mined the efficiency of investment, especially in tries may be the preponderance of public sources projects tied to the public sector (Fisman 2001; as the origin of investment activity. Public sector Olken 2007). India—in spite of its reputation in investment rates—averaging about 9 percent of globally tradable services (Kapur and Ramamurti total output—are often significantly higher in 2001)—possesses an overall service-sector MPK developing economies relative to high-income that is not higher than that of its other sectors. countries (figure 1.6, panel a). In low-income In contrast, Brazil’s high MPK in agriculture countries, in part because of greater infrastruc- appears consistent with the overall evidence in ture needs, this share may be even larger; public favor of the sector’s global competitiveness (Rada investment rates in low-income countries aver- and Buccola 2012). age about 2 percentage points higher than in To the extent that cross-country variations middle-income ones. Moreover, for low-income in MPKs are indicative of the efficiency of capi- economies that are commodity exporters, the tal deployment—with the resulting resource government often invests directly in the extrac- misallocation ref lected in aggregate total fac- tive resource industry. tor productivity (TFP) differentials—the mes- Yet public investment often fails to produce sage of low overall efficiency of investment in economically valuable capital, and even in cases developing countries is a theme that has found where public capital can be economically benefi- resonance in recent academic research (Caselli cial, public investment spending often suffers from and Feyrer 2007). Additional estimates for severe misallocation problems (Pritchett 2000). Africa (Kalemi-Ozcan and Sørensen 2012), However, the role of the state in investment has Eastern Europe (Bartelsman, Haltiwanger, and diminished in most developing countries since Scarpetta 2009), and China and India (Hsieh the 1970s (especially among low-income African and Klenow 2009) all point to the presence of countries exiting a “Big Push” model)3 (figure 1.6, significant inefficiencies in capital allocation panel b). Moreover, future trends may support Global Development Horizons The Emerging Pattern of Global Investment 27 FIGURE 1.6 The public sector share of output is lower in high-income countries than in other country groups (panel a), but the public sector share of investment has converged among country groups (panel b) a. Public share of output, 1972–2008a b. Public share of total investment, 1970–2010b 16 100 LIC 14 80 12 LMC LIC 10 60 Percent Percent 8 LMC 6 UMC 40 4 20 UMC 2 HIC HIC 0 0 1972 1976 1980 1984 1988 1992 1996 2000 2004 2008 1970 1975 1980 1985 1990 1995 2000 2005 2010 Source: World Bank calculations, using data in the World Bank World Development Indicators database . Note: HIC = high-income countries, UMC = upper-middle-income countries, LMC = lower-middle-income countries, LIC = low-income countries. a. (Unweighted) shares of total output are computed as a fi ve-year moving average, given the volatility of annual investment relative output. b. (Unweighted) shares of public investment are computed from public fi xed capital formation as a share of gross fi xed capital formation, with statistically implausible outli- ers (shares greater than unity) dropped. even greater rollback: the increased openness of of public capital to output in low-income coun- economies to foreign direct investment (FDI) tries (Gupta et al. 2011). More generally, the in funding large infrastructure projects, and the gains to output growth from increases in infra- movement of more countries from low-income to structure investment could potentially be large: lower-middle-income status. On this final point, improving infrastructure quality from that of a the development of economies worldwide would median economy to the top quartile translates exert further downward pressure on overall pub- into as much as a 7.7 percent increase in output lic sector investment. Working against these fac- per worker (Caldéron, Moral-Benito, and Servén tors is the growing perception that, in a multi- 2011). Since the 1990s, PPP infrastructure proj- polar world economy with influential sovereign ects have become increasingly common among wealth funds, the state can play a more major role developing countries (figure 1.7, panel a). in economic activity overall (Mostrous, Gue, and In addition to the long-term increase in pri- Dittman 2010), which could signal more active vate sector involvement in infrastructure, aid involvement of governments in future investment flows have become more important as the public activities. sector retrenched in the wake of the global finan- Another factor that may support lower levels cial crisis. In 2008, for example, more than one- of public investment in the future is the increased third of infrastructure spending in low-income involvement of the private sector in the provision countries was aid-financed (compared with only of infrastructure—namely, through public-pri- 4 percent in middle-income countries). With vate partnerships (PPPs).4 These partnerships can government budgets increasingly strained in the be a crucial step toward improving discipline in developed world, however, and fiscal outlooks rel- the allocation of public funds for infrastructure atively grim for the next decade or two, support and enhancing the efficiency of infrastructure for infrastructure from traditional donor sources services. (Such discipline has typically been lack- is likely to slow significantly. Th is leaves nontra- ing in the public sector, eroding the contribution ditional bilateral donors—such as Brazil, China, of public investment to capital accumulation and the Gulf Cooperation Council (GCC) countries,5 growth.) Enhancing public investment efficiency and India—to pick up the slack as major fi nan- is especially important given the contribution ciers of infrastructure projects in Africa. 28 The Emerging Pattern of Global Investment Global Development Horizons FIGURE 1.7 Private sector commitments to infrastructure have risen over time, both in major developing countries (panel a) and across most infrastructure subsectors (panel b) a. PPP infrastructure investment, selected b. Global PPP infrastructure projects, developing countries, 2000–10 by sector, 1984–2010a 180 70 170 160 150 60 130 50 2005 $ billions Countries, no. 120 110 $ billions 40 90 80 30 70 50 20 40 30 10 10 0 0 –10 2000 2002 2004 2006 2008 2010 1984 1987 1990 1993 1996 1999 2002 2005 2008 Turkey Russian Federation India Water Transport China Brazil Other developing Telecommunications Energy Countries with new projects (right axis) Source: World Bank calculations, using data in the World Bank Private Participation in Infrastructure database. Note: PPP = public-private partnership. a. Data for total global distribution for 2000–10 (panel b) have been adjusted for U.S. Consumer Price Index infl ation and are given in real 2005 dollars. For all their benefits, PPP projects remain, at American economies, infrastructure investment this time, limited in their ability to meet com- as a share of GDP collapsed in the late 1980s and prehensive infrastructure needs, with most early 1990s and has remained subdued in the projects limited to the telecommunications and years since, the exception being Chile (figure 1.8). energy sectors. One reason for this is that private This could suggest that private and public investment commitments have been heavily con- investments in infrastructure are complements centrated; the energy and telecommunications rather than substitutes (Calderón and Servén sectors have clearly benefited the most from the 2010). The limited global success of private pro- increase in private investment (figure 1.7, panel vision of important public goods is a cautionary b). In low-income countries, three-quarters of tale against a rush into PPPs in other areas where PPP commitments have been in telecommuni- the state has historically played a central investing cations, as opposed to a shade less than half for role, such as human capital (box 1.3). middle-income countries. This concentration can Perhaps the biggest challenge facing PPP- also be seen geographically, with the major devel- based infrastructure financing is how, historically, oping countries often accounting for more than such financing has been an enormously complex half of all PPP projects in a typical year. India process. Financing large-scale infrastructure proj- has been the top recipient of private sector flows ects typically involves multiple financing sources, in infrastructure since 2006, followed by Brazil, numerous public and private contracting parties, China, Russia, and Turkey.6 intricate legal documentation governed by both Moreover, on net, private investment in infra- national and international laws, and difficult con- structure often has not off set declines in public tracting and institutional design problems. sector investment, leading to cases, such as in The complexity of the multiparty, multimodal Pakistan and South Africa, where infrastructure approach to infrastructure financing is evi- investment in terms of GDP plummeted as pub- dent, for example, in the largest energy projects lic sector investment declined (Commission on recently undertaken in the Middle East (box 1.4). Growth and Development 2008). In major Latin The contours of infrastructure financing have Global Development Horizons The Emerging Pattern of Global Investment 29 FIGURE 1.8 Infrastructure investment shares of GDP declined rapidly in major Latin American economies in the late 1980s and have remained subdued in subsequent years 100 5 Infrastructure investment Share of infrastructure 80 4 as share of GDP, % investment, % 60 3 40 2 20 1 0 0 1980 1983 1986 1989 1992 1995 1998 2001 2004 Public share Private share Infrastructure investment (right axis) Source: World Bank calculations, using data from Calderón and Servén 2010. Note: The six Latin American economies covered in the figure are Argentina, Brazil, Chile, Colombia, Mexico, and Peru. also changed gradually over the past decade, with capital (measured using the perpetual inventory bond issuance replacing traditional bank loans, method),8 was largely concentrated in the devel- and banks based in emerging economies now oped world, especially in the United States and becoming major intermediaries in the infrastruc- Western Europe (map 1.1). By 1995, this situation ture sector. The challenges facing infrastructure had begun to change, with several large develop- fi nancing in the future will be revisited later in ing economies, such as Brazil and China, accumu- this chapter. lating stocks comparable to those of high-income In sum, sectoral investment patterns also countries. The latest data suggest that this shift has imply changing patterns since 2000, with a further consolidated. Today, Brazil, China, India, shift in global investment away from agriculture and Russia together account for about 18 percent toward manufacturing, and a gradual reduction of the global share of capital, more than twice the of the role of public sector investment in devel- share of Germany and near that of the United oping countries. The world appears to be on the States. cusp of yet another shift in investment patterns, The process is far from complete: a normalized this time toward an increased emphasis on the Herfindahl index of capital stocks has fallen only services sector. The expanded role of the services modestly—from a high of 0.08 in the late 1980s sector will likely be manifested in a host of invest- to a low of 0.07 of 2010—compared with a larger ment activities, ranging from services related to fall in the analogous normalized Herfi ndahl of greater human capital production to an expan- production (a decline from 0.13 to 0.10). This sion of infrastructure in the developing world. slow evolution of the Herfindahl points to the high concentration of capital and wealth that continues to reside in the industrialized high- A gradual redistribution of income countries, a reflection of the legacy of global capital stocks has occurred capital accumulation that occurred over the 19th in recent decades and 20th centuries. Nevertheless, the process of a The ongoing shift in the distribution of global gradually less unequal global distribution of capi- investment shares has meant that the exist- tal should continue in the future as developing ing stock of productive physical capital has economies grow in size and absolute investment also been changing over time.7 In 1980, global flows into their domestic capital stocks rise. 30 The Emerging Pattern of Global Investment Global Development Horizons BOX 1.3 Investment in human capital can be a source of economic growth Although the focus of this report is on investment in their total investment billa on human capital, whereas physical capital, it is important to recognize that accu- developing countries spent about half as much over mulation of human capital may stimulate economic the same period (figure B1.3.1). In fact, the most recent growth and development. Though it is possible that decade saw only a slight increase in human capital human capital may be subject to diminishing returns in investment by developing nations, to 19.6 percent the same way as physical capital (Mankiw, Romer, and (from 17.3 percent in the decade before). Given the Weil 1992), some scholars have argued that endog- growing importance of human capital in the deploy- enous growth via increasing returns could be real- ment of technology, even in capital-intensive sectors ized through spillovers from knowledge embedded in such as manufacturing, developing countries will likely human capital (Romer 1990). need to adjust their capital allocation decisions in the Over the past two decades, high-income coun- future. To the extent that they do, this increased invest- tries have outspent developing countries in developing ment in human capital will be yet another element giv- human capital, not only on an absolute basis but also in ing rise to higher levels of investment in the services relative terms. Between 1990 and 2009, high-income sector, since education and research and development countries invested, on average, about 35 percent of (R&D) are typically classified as services. a. Current national income accounting practice classifies expenditure in these sectors as consumption, not investment. For the purposes of box 1.3, the total investment bill is calculated to include these expenditures as forms of “investment.” FIGURE B1.3.1 Investment in human capital relative to physical capital in high- income and developing countries, 1990–99 to 2000–10 12 2000–10 10 Human capital 8 1990–99 investment $ trillions 6 Physical capital 4 2000–10 investment 2 1990–99 0 Developing High income Investment in education (private) Investment in education (public) Investment in R&D Inventory Investment Fixed investment Sources: World Bank calculations, using data in the World Bank World Development Indicators database and World Bank EdStats database. Note: Investment in private and public education and research and development (R&D) are measured as expenditures in each of these sectors. Although the distribution of stocks of capital diminished, the greater abundance of capital in provides an important sense of the cumulative high-income countries need not signify that capi- investment that would be required before the tal has been inefficiently distributed. Capital must, global inequality in the distribution of capital is ultimately, be paired with labor (and other inputs) Global Development Horizons The Emerging Pattern of Global Investment 31 BOX 1.4 Energy project financing in the Middle East: A multiparty, multimodal approach Infrastructure fi nancing related to the highest-value The general resilience of project financing in the energy projects undertaken in the Middle East illus- Middle East in 2009 demonstrates that sound, credit- trates the complexity of large-scale infrastructure worthy projects continue to be able to raise long-term fi nancing operations. Projects typically involve multi- financing in a global environment of credit-constrained ple local and regional investors and financing based on markets, especially when projects employ innovative a capital structure that combines direct equity stakes, approaches to syndicate formation. The Shuweihat 2 debt issuance, and syndicated lending, together with independent water and power project, for example, some elements of export credit (both direct and cov- secured a key loan component of its financing from the ered) and Islamic financing (table B1.4.1). Pricing struc- Japan Bank for International Cooperation after the par- tures are also often complex; most projects, for exam- ticipation of Japan’s Marubeni among the deal’s equity ple, peg pricing to the London interbank offered rate sponsors. The RasGas project is the second time the (LIBOR), with varying markups for tranches according project has sought financing, the first being a 70-30 debt- to the stage of the project. equity offering in 1996 (Dailami and Hauswald 2007). TABLE B1.4.1 Top five project finance deals in the Middle East, 2009 Value Financing breakdown Sponsors Sector Country ($ billions) ($ billions) (country, stake) Dolphin Gas Pipeline Oil and gas United Arab 4.1 Bond (1.3); syndicated Mubadala (United Arab Emirates loan (1.4); covered loan Emirates, 51%); Total (0.2); shareholder loan (France, 25%); Occidental (1.2) Petrol (United States, 25%) Shuweihat 2 IWPP Power United Arab 2.6 Syndicated loan (2.1, Abu Dhabi Water and Elec Emirates 2 tranches); sponsor (United Arab Emirates, equity (0.5) 60%); Suez (France, 20%); Marubeni (Japan, 20%) Rabigh IPP Oil and gas, Saudi Arabia 2.5 Syndicated loan (1.9); ACWA Power (Saudi Arabia, power sponsor equity (0.5) 40%); Kepco (United States, 40%); Saudi Electricity (Saudi Arabia, 20%) Ras Laffan LNG Oil and gas Qatar 2.2 Bond (0.5 3-yr, 1.1 5-yr, n.a. Project 0.6 10-yr) Al Dur IWPP Power and Bahrain 2.2 Syndicated loan (1.3); Suez (France, 40%); GIC desalination senior public bond (0.1), (Singapore 20%); SIO sponsor equity (0.6) (Bahrain, 10%); Cap Mgmt Hse (Bahrain, 25%); Instrata (Bahrain, 5%) Source: World Bank calculations, using data in the PFI database. Note: IPP = independent power project, IWPP = independent water and power project, LNG = liquefied natural gas, n.a. = not applicable. for the purposes of production; a given country message than a focus on stocks alone. The East would only be “capital-starved” if there existed a Asia and Pacific region, for instance, has the significant imbalance between available capital and largest stock of capital, mainly due to the size labor (adjusted for human capital differentials). of China. Once the capital stock is adjusted for Thus, the distribution of such efficiency- efficient units of labor, however, the region falls adjusted capital stocks may offer a different into the lower half of the distribution in terms of 32 The Emerging Pattern of Global Investment Global Development Horizons MAP 1.1 The geographical distribution of capital stocks has gradually shifted to developing countries over the past three decades 1980 1995 2010 Capital stock (constant 2000 $ billions) > $2,500 $1,000–$2,500 $500–$1,000 $100–$500 $50–$100 $20–$50 $10–$20 < $10 no data Sources: World Bank calculations, using data in the World Bank World Development Indicators database and World Bank Global Economic Monitor database. Note: Capital stocks are calculated using a perpetual inventory method with assumed constant depreciation rate of 5 percent. Countries with insuf- ficient data in the constant investment series are backcasted using a regression of the investment defl ator on the GDP defl ator and available invest- ment data. Data for the 2010 chart are either for 2010 or the latest year since 2006. Global Development Horizons The Emerging Pattern of Global Investment 33 capital per effective worker (figure 1.9). The oppo- FIGURE 1.9 Adjusting capital stocks by efficiency- site phenomenon can be seen, most dramatically, adjusted units of labor results in different distributions in the Middle East and North Africa. Finally, the compared to comparing capital stocks alone stock of capital in Sub-Saharan Africa, while sig- 16,000 12 nificantly less than that of other regions, is higher than that of South Asia in per-worker units. 14,000 10 These differences in capital stock per efficiency- 12,000 adjusted labor unit point to the importance of 8 10,000 $ trillions Dollars considering the evolution of complementary fac- 8,000 6 tors when assessing the future demand for capi- 6,000 tal. For example, Latin America may experience 4 weaker demand for investment in the future, 4,000 2 especially because impending population dynam- 2,000 ics (explored in greater detail in chapter 2) may 0 0 mean a slower expansion of the complementary Sub- E. Asia E. Eur. L. Amer. Mid. S. Asia labor force. Conversely, capital demand may be Saharan & & Cent. & the East & greater in regions such as Africa and South Asia, Africa Pacific Asia Caribbean N. Africa given the anticipated expansion of the labor force Capital per effective worker Capital stock in these countries. There also appears to be room (left axis) (right axis) for more capital accumulation in a high-capital- Sources: World Bank calculations, using data in the World Bank World Development Indicators stock region like East Asia. database and World Bank Global Economic Monitor database. Note: Capital stocks are calculated for 2010 (or latest), using a perpetual inventory method, Historical increases in capital per worker also with assumed constant depreciation rate of 5 percent, for the region collectively. Efficiency carry significant implications for overall worker units of labor are calculated as the share of the working-age population participating in the labor market, adjusted by returns to the existing stock of human capital (measured as the aver- welfare. The average worker in 1950 in the United age years of schooling of the population above 15 years of age). Capital stocks per efficiency- States—then, as now, the technologically lead- adjusted units of labor are calculated as simple averages for each region. ing nation—was paired with the equivalent of $72,400 worth of capital (measured in 2005 increases in capital stocks, thereby placing a PPP-adjusted international dollars) in the form natural brake on the convergence of per-worker of machinery, plant, and equipment, a level that stocks of capital. countries such as Chile and Mexico attained by 2008. The catch-up in worker welfare is even more stark when compared to other high-income countries. Today, the average worker in Malaysia Long-run drivers of investment is paired with only 20 percent less capital than a around the world French worker had in 1975, and that does not take With its central role as a signal of anticipated into account the vast improvements in the over- future prosperity, the study of investment has pre- all quality of capital goods. Even in a low-income occupied macroeconomists since the discipline’s economy such as Bangladesh, the average worker emergence. Keynes (1936, 161) termed the spon- works with one-tenth the capital an average British taneous and volatile forces behind investment worker did in the 1980s. “animal spirits.” But the driving forces behind Although absolute capital stocks will con- economic actors’ long-run decisions to invest tinue to (slowly) converge over the next two are not entirely a black box. Traditional models decades, differentials in the per-worker stock of of investment activity appeal to macroeconomic capital between developing and high-income determinants, such as the growth rate and the real countries—which is already substantially more cost of capital (Jorgenson 1963), the shadow cost equally distributed today than it has ever been of capital (Tobin 1969), uncertainty (Lucas and (map 1.2)—may not necessarily fall further. Prescott 1971), and nominal variables such as Th is is because population growth in develop- asset prices ( Dixit and Pindyck 1994) and the ing countries will likely off set their fairly mild money supply (Holmström and Tirole 1997). 34 The Emerging Pattern of Global Investment Global Development Horizons MAP 1.2 Capital stocks per worker are now more equally distributed than ever before Capital per worker (constant 2005 international $ ‘000) > $200 $100–$200 $60–$100 $40–$60 $25–$40 $10–$25 $5–$10 < $5 no data Source: World Bank calculations, using data in the Penn World Tables (https://pwt.sas.upenn.edu/). Note: Data are for 2009. Capital stocks are calculated using a perpetual inventory method with assumed constant depreciation rate of 5 percent, and presented in thousands of 2005 (Geary-Khamis) international dollars. At the global level, investment is ultimately more investment from abroad. Although increases constrained by saving. This accounting fact in investment need not automatically trans- implies that independent changes in global saving late into economic growth (Easterly and Levine patterns will directly affect the extent of global 2001),9 greater investment is naturally associated investment activity. Were there to be reductions with heightened economic activity because firms in global saving in the years ahead, for instance, respond to anticipated future profit opportunities this would imply reductions in world investment by raising investment today. Productivity gains as well. Nevertheless, a single-minded focus on drive heightened growth and investment activ- the global saving supply would leave open the cru- ity, inciting lasting structural transformations in cial question of how, given this supply of saving, the broader economy. The relationship between global investment is subsequently allocated across growth and investment can be substantial: econo- countries. Th is allocation into global portfolios metric estimates suggest that, all else equal, each of investments is subject to two countervailing additional percentage point of growth is associ- forces that may suppress investment activity in ated with a 0.2 percentage point increase in the any given economy: disincentives for additional investment rate. domestic investment, arising from diminishing returns (Kraay et al. 2006); and frictions imped- Structural factors are important ing mobility of capital across national borders, drivers of changes in the both of which give rise to differences in expected allocation of investment financing returns across countries. Understanding what drives such cross-country differences in expected However, given the heterogeneity of developing returns is key to predicting the ability of econo- countries’ experience with investment, even for mies to attract investment financing. economies growing at similar rates, it is clear that Differences in macroeconomic growth, in par- economic growth alone does not fully explain ticular, are a central part of the story. Economies all aggregate investment activity. The opportu- that grow faster tend to invest more and attract nity cost of undertaking any given investment Global Development Horizons The Emerging Pattern of Global Investment 35 activity lies, ultimately, in the attractiveness of percent increase in financial development, invest- other available investment opportunities world- ment can rise as much as 3.3 percent. wide, and this attractiveness is embodied in these Investment is also positively associated with alternative investments’ expected returns to capi- the overall quality of the institutional environ- tal. Such expected return differentials are attrib- ment (figure 1.11). On this account, developing utable not just to differences in countries’ eco- countries have made significant strides over the nomic growth. Of central importance is also the past decade: between 1996 and 2010, developing yield on capital an economy offers relative to oth- countries’ control of corruption and rule of law ers.10 Finally, cross-country differences in struc- advanced by about 4 percent, while that in the tural factors—in particular their levels of finan- developed world slid by about 1 percent, accord- cial development and the overall quality of their ing to the World Bank’s Worldwide Governance institutional environments—can also affect the Indicators.12 Although governance challenges expected returns from investing in a given coun- remain an issue for developing countries, these try because such structural factors can also dis- improvements could well make a difference tinguish the developmental fortunes of nations.11 to their likely investment paths in the future. Two structural factors are especially important Econometric analyses indicate that real invest- in negatively influencing investment fi nancing: ment activity, even after controlling for a host low levels of financial development and inadequa- of other intervening factors, remains statistically cies in institutional quality. Together, these fac- significant (figure 1.11) and can rise by as much tors give rise to (real or perceived) shortages of as 1.6 percent for each 10 percent increase in safe, investible assets (Caballero 2008). Th is is a institutional quality. chronic problem in emerging economies and can Finally, the openness of a country’s borders to engender instabilities that go beyond the country’s foreign capital inflows can also increase domes- borders: the pursuit of safe investment opportu- tic investment. As individual economies become nities can suppress long-term real interest rates, more integrated into the global economy, future decrease the efficiency of global saving allocation, investment— especially in the smaller and ignite speculative bubbles in safe asset-generating poorer economies—may be increasingly led countries, and contribute to unsustainable global by such FDI. Cross-border capital movements imbalances (Caballero, Farhi, and Gourinchas (still) remain a small fraction of what would be 2008). Overall, the empirical evidence supports implied in a world with more frictionless global the notion that structural factors are central to capital markets (Lucas 1990), but if financial determining the amount of investment financ- globalization were to proceed at the pace of ing a country can attract, and it offers a poten- the past two decades, patterns of capital flows tial explanation of why cross-country patterns of could be dramatically aff ected, as discussed in investment differ as much as they do (box 1.5). chapter 3. The importance of these two factors on actual investment activity is evident from the cross- Investor sentiment is a key country data. As countries grow, they also tend driver shaping future investment to yield more mature and sophisticated finan- dynamics cial systems; this increased level of financial development, in turn, appears to move together One way of interpreting Keynes’s “animal spir- with changes in investment rates. Th is relation- its” behind investment is that such spirits may ship is especially striking for developing coun- also embody expectations surrounding future tries, although changes in credit provision have investment dynamics. Such investor sentiment historically followed investment rates closely for is ultimately based on expectations formed with high-income countries as well, with the excep- regard to both future growth prospects and tion of the period after 2000, when credit provi- uncertainty surrounding those prospects. These sion grew sharply in the developed world (figure expectations are not easily captured in econo- 1.10). Econometric analyses suggest that, for a 10 metric analyses of observable variables, but the 36 The Emerging Pattern of Global Investment Global Development Horizons BOX 1.5 Structural factors can affect the cross-country allocation of investment financing Traditional models of investment typically rely on stan- for additional factors affecting expected returns (but not dard macroeconomic determinants—such as output, fully reflected in interest rates), the model should be economic growth, and the real cost of capital—to supplemented by economic and structural variables that explain investment activity. However, macroeconomic may differ systematically between countries. Third, the variables are subject to cyclical shocks of a short-term approach should account for the significant heterogene- nature, and such models may end up capturing fac- ity that may occur in unobservable variables, along with tors whose effects may dissipate over the longer run. the potential endogeneity in the explanatory variables. Just as important, these older models do not capture Although no single model can simultaneously address the potential effects of informational asymmetries these distinct considerations, this report relies on sev- and incomplete markets, a shortcoming that has been eral econometric approaches to specify a range of plau- remedied by more recent models of investment. sible coefficients for the main variables of interest. The important role of uncertainty and market imper- The formal analyses (reported in the online Global fections highlighted by these more recent studies Development Horizons [GDH ] annex 1.3) suggest that hint at the need to consider how structural variables although not all structural variables exert a systematic can affect aggregate investment at the cross-country impact on the cross-country allocation of investment level. After all, such frictions often fi nd root in struc- financing, some clearly do. One important factor is the tural changes such as differences in the ability of the role of financial development: as the level of financial financial sector to intermediate saving (Dailami 1992), development increases, investment tends to rise as the legal frameworks that govern contract enforcement well. This is consistent with the literature (Benhabib and (Cooley, Marimon, and Quadrini 2004), and property Spiegel 2000; Levine 2005; World Bank 2012b), which rights more generally (Besley 1995), along with the finds that financial development has a causal influence broader business climate and political-institutional envi- on, among other things, capital formation. The overall ronment (Acemoglu, Johnson, and Robinson 2005). institutional environment appears to be important for Economic openness, so crucial to the medium-term driving investment as well. For example, Mauro (1995) current account, may also affect long-run investment has demonstrated that widespread corruption causes activity (Chinn and Prasad 2003). lower levels of investment at the cross-country level, Differences in such structural factors can potentially while Besley (1995), using the case of Ghana, has alter the flow of investment financing received by any shown that the same applies to the strength of the given country. Three major considerations must be property rights regime. More generally, strong, posi- taken into account in any successful model that seeks tive relationships tend to exist between institutions and to explain such cross-country investment patterns. investment growth (Acemoglu, Johnson, and Robinson First, the approach should incorporate, at its core, cross- 2005), especially those that are supportive of economic country differentials in interest rates. Second, to account freedom (Dawson 1998). role of investor sentiment can be teased out in investment.14 Values of Q significantly above other ways. Consequently, although the Linkage unity are suggestive of optimism regarding future model used in this report for scenario analyses investment opportunities, and vice versa for values does not explicitly model the effect of investor significantly below one. Computations of Tobin’s sentiment, several other approaches offer hints Q for developing countries suggest that investors about the role that sentiment can play in shaping anticipate continued positive growth prospects for future investment in developing countries. many developing economies (box 1.6). One such approach is to examine the implicit Another alternative way of attempting to cap- (or shadow) cost of capital, as measured by Tobin’s ture the importance of investor sentiment is to Q.13 Estimates of Tobin’s Q point to financial examine the effect that even temporary shocks market participants’ conditional forecasts of to uncertainty can have on future investment. Global Development Horizons The Emerging Pattern of Global Investment 37 FIGURE 1.10 Investment rates and financial sector development in developed and developing countries have evolved similarly 30 8 Fixed investment share of GDP, %. Investment rate Investment rate 7 (developing) 25 (high income) Credit share of GDP, % 6 5 20 4 Financial 15 development 3 (high income, 2 10 right axis) 1 0 5 Financial development –1 (developing, right axis) 0 –2 1963 1968 1973 1978 1983 1988 1993 1998 2003 2008 Sources: World Bank calculations, using data in the World Bank World Development Indicators database and Financial Development and Structure database. Note: Investment rate is defined as the fi xed investment share of GDP, while change in financial development is defined as the three-year moving average of the annual change in the ratio of domestic credit provision to the private sector as a share of GDP; both ratios are unweighted averages for the two country groups. Th is is especially pertinent given how the global FIGURE 1.11 A positive relationship exists between fi nancial crisis of 2007–09 is often associated investment and institutional quality with decreased investor confidence. The logic 30 behind this mechanism is well understood: capi- tal investment is essentially a decision on when to exercise an option on an irreversible project, contingent on uncertainty about its future prof- Real fixed investment 25 itability (Dixit and Pindyck 1994; McDonald and Siegel 1986).15 Therefore, increased uncer- tainty about future economic prospects may dampen aggregate investment as firms retain 20 their option value by postponing capital forma- tion decisions.16 Large variations in uncertainty alone may be sufficient to freeze investment activity and generate sharp recessions, indepen- dent of productivity shocks (Bloom 2009), and 15 may have an effect even when realized real inter- 0 0.5 1.0 1.5 2.0 2.5 est rates themselves remain constant (Fernández- Institutional quality Villaverde et al. 2011). By and large, empiri- Sources: World Bank calculations, using data in the World Bank World Development Indicators cal studies have found that uncertainty does database and Financial Development and Structure database. Note: Data are pooled for all countries over the period 1980–2010. The figure is a bivariate plot have a corrosive effect on investment (Carruth, of real fixed investment and institutional quality, with linear regression line. Investment rate is Dickerson, and Henley 2000). For developing defined as fixed investment in 2000 dollars, and institutional quality is the simple average of (subjective) measures of corruption and rule of law. Both variables are in logarithm form, are countries, the effect of uncertainty is likely to be unweighted, and are significant in both bivariate and multivariate regressions at standard levels. especially pronounced for macroeconomic vola- tility (Aizenman and Marion 1999), uncertainty The deleterious effects of uncertainty are regarding the real exchange rate (Servén 2003), not limited to the short run. Simulations of and commodity price f luctuations (Claessens the response of investment to a transient shock and Duncan 1993). arising from economic uncertainty show that 38 The Emerging Pattern of Global Investment Global Development Horizons BOX 1.6 Tobin’s Q calculations for developing countries suggest positive growth prospects To the extent that Tobin’s Q serves as a reasonable which could reflect long-lasting negative spillover conditional (medium-term) forecast, cross-country effects of the European financial and debt crisis. measures of Tobin’s Q for developing economies reveal Some caution is warranted in placing too much a mixed picture with regard to prospects for investment weight on a single, raw measure of countries’ likely in the developing world. On one hand, it is evident that, future investment. Developing countries—almost by on average, lower-income countries possess higher definition—will tend to possess financial markets that levels of average Q s (figure B1.6.1), a result consis- are less complete and more ineffi cient, which would tent with the notion that expectations of future invest- reduce the predictive value of the market-based valu- ment are greater in lower-income countries. (Although ations on which Tobin’s Q computations rely. Higher not reported, Tobin’s Q calculations for developed valuations for developing-country firms may also reflect countries typically yield values that fl uctuate around intangibles—such as political connections or monopoly the equilibrium-consistent value of unity.) Moreover, rents—rather than bright growth prospects. Indeed, measured Q s in several fast-growing developing coun- while the relationship between investment rates and tries—such as Bangladesh (3.55), Brazil (2.60), China Tobin’s Q is reasonably strong for high-income econo- (2.08), India (2.14), and Indonesia (1.87)—are higher mies, it is far weaker for developing ones, which does than the developing-country average (of 1.53), which limit the confi dence one can place on the predictive accords with the notion that anticipated investment ability of the measure. Nevertheless, with 90 percent prospects in these countries are generally favorable. of Q s measured at levels above unity, the overall mes- On the other hand, Tobin’s Q values for several Eastern sage of these calculations is that investors anticipate European countries indicate that market expectations a continued positive growth path for many developing for investment in the region remain more guarded, economies. FIGURE B1.6.1 Tobin’s average Q is higher on average in lower-income developing countries, reflecting market expectations of greater growth potential 4.0 3.55 Higher income Lower income 3.5 3.0 2.5 Tobin’s Q 2.0 1.53 1.5 1.0 0.57 0.5 0 L U A A S R EX RG M S S F L er R JA e M A CH R N H KR M N R A L G D N A VN A M K N D Y CH ZA CO ag PH LV RU TU Y U Av PE JO A PA EG BR TH BI LK IN H G BG LT ID KE M M U A Source: World Bank calculations, using data in the Bloomberg database. Note: Data are for 2010. Country-level Tobin’s average Q s are computed as the average of the ratio of aggregate market value to book value of assets held by public firms domiciled in the country, with outliers greater than two standard deviations removed. Countries are sorted in order of descending per capita income, measured in 2010 dollars. Country coverage is limited by data availability. The developing-country average is com- puted as a simple average of available data. The positive slope of the linear fit line indicates that Tobin’s Q is higher for economies with lower levels of per capita income. Global Development Horizons The Emerging Pattern of Global Investment 39 although such uncertainty seems to have a more cross-country differences in growth rates and rel- pronounced effect on investment in high-income ative returns to capital (when such relative returns countries, the negative effects of such shocks are understood to embody not just observable may also be acutely felt in developing countries differences in countries’ relative yields but also (box 1.7).17 structural factors), along with unobservable dif- In sum, the existing literature on invest- ferences arising from investor sentiment. On ment suggests that the cross-country variations balance, therefore, the overall pattern of global in investment financing observed in recent and country-level investment through 2030 will times—especially in terms of the shift in invest- depend on the global supply of saving, its alloca- ment activity toward developing countries and tion to investment across countries, and country- the move toward greater investment in second- specific demands for capital. It is to this issue that ary sectors—can be explained by examining the next section turns. BOX 1.7 How transient shocks to uncertainty can have a long-lasting impact on investment The deleterious effects of uncertainty are not lim- after a lag of about eight years in developing countries ited to the short run. Simulations of the response of (fi gure B1.7.1, panel b). For both sets of countries, an investment to a transient shock arising from economic uncertainty shock can be quite persistent, and the uncertainty show that although such uncertainty effect may not disappear completely long after the seems to have a more pronounced effect on invest- shock has ended. Moreover, the sharp results sug- ment in high-income countries (figure B1.7.1, panel gest that shifts in uncertainty of a more fundamental a), developing countries are certainly not exempt; the nature may have even more negative and detrimental negative effects of an uncertainty shock may appear effects on investment. FIGURE B1.7.1 Investment rate response to increased uncertainty in high-income and developing counties a. High-income countries b. Developing countries 1.2 1.2 0.8 0.8 Investment rate, % Investment rate, % 0.4 0.4 0.0 0.0 –0.4 –0.4 –0.8 –0.8 –1.2 –1.2 2 4 6 8 10 12 14 16 18 20 22 24 2 4 6 8 10 12 14 16 18 20 22 24 Years Years Source: World Bank. Note: Investment rate is defined as the fi xed investment share of output. The increase in uncertainty is a one standard-deviation shock to uncertainty, measured as the contribution-weighted principal components of generalized autoregressive conditional heteroscedasticity (GARCH) (1,1) conditional variances of infl ation; real exchange rate; and agricultural, metals, and energy commodity price indexes. The figures show impulse response functions for a one standard-deviation innovation (solid blue) with +/– 2 standard error bands (dotted red), for 25 periods (years) into the future, generated from a level vector autoregressive (VAR) specification with two Akaike information criterion (AIC)–selected lags. Details of the estimation are provided in the Global Development Horizons online annex 1.4. 40 The Emerging Pattern of Global Investment Global Development Horizons The changing face of future The shift in distribution of investment as a investment opportunities share of global output between developing and high-income countries can be understood by The gradual convergence considering the two components underlying scenario: A fairly stable path for this change: changes in each respective group’s the global investment rate, with investment rates and changes in their sizes rela- heterogeneity at the country level tive to each other. The bulk of the increase in After the dip due to the global crisis of 2007–09, the share held by developing countries will actu- the scenario analyses performed for this chapter ally be due to increases in their relative size. indicate that global investment will recover and Collectively, investment in high-income coun- trend steadily upward, growing at a rate in line tries as a proportion of their own output will with overall output (figure 1.12, panel a). By decline slightly, by 0.7 percentage point (from 2030, global investment will amount to $25 tril- 17.5 percent to 16.8 percent), with an even larger lion, measured in 2010 dollars, with developing corresponding decline in the developing world countries accounting for about two-thirds ($15 (a fall from 32.2 percent to 30.1 percent). Thus, trillion) of the total. In absolute terms, invest- developing countries’ increasing share of invest- ment in developing countries will overtake that ment as a percentage of global output will be in high-income countries by the middle of the due to their catch-up economic growth, which 2010s. The global investment rate will remain results in their relative size rising from about more or less stable, consistent with the global sav- one-third of the global economy in 2010 to ing rate of about 22 percent (figure 1.12, panel almost half by 2030. b, and chapter 2). The share of global investment Just as China’s contribution led the surge in as a percentage of world output accounted for developing-world investment, the future decline by high-income countries is expected to fall sig- in investment rates within developing countries nificantly—from 12.0 percent to 9.1 percent of will also be led by China, which will see its very world output—with the rise in the investment high investment rate of 48 percent (as of 2010) share by developing countries more than making edge down to about 45 percent in 2016 (as its up the difference. shorter-term postcrisis stimulus fades) before FIGURE 1.12 Total investment (panel a) and investment share of global output (panel b) will increase in developing relative to high-income economies in the gradual convergence scenario a. Total investment b. Investment share of world output Investment share of world output, % 30 120 30 25 100 25 World GDP, $ trillions World GDP World 20 80 20 (right axis) investment $ trillions 15 World Developing 60 15 Developing investment 10 40 10 High income High income 5 20 5 0 0 0 2005 2010 2015 2020 2025 2030 2005 2010 2015 2020 2025 2030 Sources: World Bank projections, supplemented with calculations using data in the World Bank Global Economic Monitor database and World Bank World Development Indicators database. Note: Figure projections assume a gradual convergence scenario of annual world economic growth of 2.6 percent over the next two decades, of which developing (high- income) growth will average 5.0 (1.0) percent annually. Global Development Horizons The Emerging Pattern of Global Investment 41 FIGURE 1.13 Investment rates in many developing countries will trend downward in the gradual convergence scenario, but remain fairly stable in high-income countries a. Large economies b. Other major developing economies 45 Indonesia China 40 India 50 35 Investment rate, % Investment rate, % 40 30 Russian Federation 25 30 Europe 20 Japan Brazil 20 15 Sub-Saharan 10 Africa 10 United States 5 0 0 2005 2010 2015 2020 2025 2030 2005 2010 2015 2020 2025 2030 Source: World Bank projections, supplemented with calculations using data in the World Bank Global Economic Monitor database. Note: Investment rates for the short run (2011–13) follow estimates and projections from the World Bank’s Global Economic Prospects, while medium- and long-run pro- jections (2014–30) are from the Bank’s global L INKAGE model. Actual realizations for post-2014 are likely to be noisier. Figure projections assume a gradual convergence scenario of annual world economic growth of 2.6 percent over the next two decades, of which developing (high-income) growth will average 5.0 (1.0) percent annually. declining gradually to 37 percent by 2030 (fig- (figure 1.13, panel b). At the same time, not all ure 1.13, panel a). Already, there are signals in developing countries will see continuous declines China—for example, underuse of office build- in investment rates: Sub-Saharan Africa, for ings in major cities and suspension or partial instance, will see its investment rate rise to aver- suspension of many high-speed rail projects— age 24 percent between 2010 and 2030—a rate that investment opportunities are being gradu- higher than its 2005–09 average of 21 percent— ally exhausted. Th is declining investment trend and Russia’s investment rate will remain above 27 in China will occur alongside a slowdown in percent over most of the projection period, signif- the country’s rate of economic growth resulting icantly higher than its historical 2005–09 aver- from a rising old-age dependency ratio, which age. By way of contrast, the investment rate in the will reduce both capital supply (as saving falls) largest high-income countries will remain more and labor supply (from a shrinking working-age or less stable, a testament to the fact that mild population). Of course, the decline of China’s (but nevertheless positive) rates of productivity investment rate to a still-very-high 37 percent is growth will mean that demand for investment not entirely surprising, given the general consen- in these capital-intensive economies will not dip sus by academic, market, and policy economists precipitously. In the United States, for example, that investment rates consistently exceeding 40 investment rates will close 2030 at around 13 per- percent are unlikely, even taking into account cent, which is its average rate over the course of China’s relatively high rate of expected growth the two decades. (Bardhan 2010; Dobbs et al. 2010; World Bank In the gradual convergence scenario, future 2012a). declines in investment rates will be largely driven Although the sharp drop in the investment by a slowing of either productivity growth or the rate in China will not be matched by any other labor supply because these factors determine the developing country, several countries will expe- demand for investment. However, investment rience significant declines relative to anticipated rates can also fall because of insufficient financial peaks in the medium term. These include Brazil resources available for any given country (the sup- (from 23 percent in 2014 to 19 percent in 2030) ply of investment financing). Changes in yields, and Indonesia (from 41 percent to 33 percent) relative to average yields in the world, would 42 The Emerging Pattern of Global Investment Global Development Horizons FIGURE 1.14 The yield on loanable funds will rise more underlying mechanism inducing this change is rapidly in high-income countries than in many developing the converse of Sub-Saharan Africa: shrinking countries labor forces and a deceleration from current fast productivity growth will mean that investment 0.90 demand softens over the course of the next two 0.80 2010 2030 decades. Then—conditional on trends in domes- 0.70 tic saving—the drop in investment rates and the 0.60 lower yields will also be accompanied by a slow- ing of net capital inf lows (discussed in detail 0.50 Yield in chapter 3) as investors seek relatively higher- 0.40 return opportunities elsewhere. 0.30 High levels of investment by developing 0.20 countries in their domestic economies over the 0.10 next two decades will go some way toward cor- recting the inequality in the global distribution 0.00 of capital stocks between the developing world il o and high-income countries. By 2030, half of all pe n es a a a a n a az ic in di si ic ic pa tio at ro ex ne fr fr Br Ch In Ja ra St .A A Eu M capital stocks will reside in developing countries, do de n d N ra In te Fe t& with countries in East Asia and Latin America ha ni n U as Sa ia accounting for the largest shares of these stocks ss .E b- Ru Su id M (56 percent and 15 percent of the developing Source: World Bank projections. world total, respectively) (figure 1.15). Sub- Note: The leftmost (rightmost) bar for each country corresponds to 2010 (2030). Sub-Saharan Saharan Africa will see the third-fastest growth in Africa in these graphs excludes South Africa. Figure projections assume a gradual conver- gence scenario of annual world economic growth of 2.6 percent over the next two decades, of its capital stock among developing regions (after which developing (high-income) growth will average 5.0 (1.0) percent annually. Yields are nor- East and South Asia), albeit from a low base; by malized to unity for 2007, and so reported levels are not directly comparable across countries. 2030, its capital stock will more than double to $4.6 trillion (in 2010 dollars). As a group, devel- oping countries’ capital stocks will grow by 212 indicate whether forces of demand or supply are percent by 2030, compared with a 37 percent responsible for the decrease in investment rates. If increase in high-income countries. (weak) demand is the driving force, relative yields will decline, and investors that are eager to supply The rapid convergence scenario: their funds will mitigate the drop in the invest- A similarly stable path for the ment rate by encouraging more capital-intensive global investment rate, although production. If, instead, (weak) supply is the driv- with slightly higher investment ing force, the relative yield will increase, enticing rates among developing countries still some additional financial capital while reduc- ing demand by lowering the capital-intensity of As an alternative to the gradual convergence sce- production. nario, it is interesting to consider a rapid con- The relatively high yields in Sub-Saharan vergence scenario where convergence between Africa, for instance, signal a difficulty in attract- developing economies and high-income ones pro- ing ample financing from global sources to meet ceeds more rapidly. Such a scenario could arise, their large investment demand and many new for instance, if the tepid post-financial-crisis investment opportunities (figure 1.14). For a recovery in high-income countries plays out over similar reason, yields are also expected to be rela- an extended period and if more rapid catch-up tively high in Russia, Mexico, or high-income growth in the developing world is accompanied countries. The relative decline of yields in China, by significant structural and institutional trans- India, and Indonesia, in contrast, indicate a drop formation. For example, fi nancial development in investment demand from current levels. The in Brazil, India, and the Middle East would Global Development Horizons The Emerging Pattern of Global Investment 43 FIGURE 1.15 Developing countries will account for more than half of global capital stocks by 2030 in the gradual convergence scenario (panel b), compared with about a third today (panel a) a. Capital stocks, 2010 b. Capital stocks, 2030a $3.8t $4.6t $4.6t $1.7t $8.8t $11.0t $17.5t $89.1t $23.3t $14.5t $21.5t $115.8t $8.0t $158.2t High-income L. Amer. & the Mid. East S. Asia Sub-Saharan E. Asia & E. Eur. & countries Caribbean & N. Africa Africa Pacific Cent. Asia Source: World Bank projections. Note: Capital stocks are calculated using a perpetual inventory method with a constant depreciation rate of 5 percent. a. Panel b projections assume a gradual convergence scenario of annual world economic growth of 2.6 percent over the next two decades, of which developing (high-income) growth will average 5.0 (1.0) percent annually. advance to attain, by 2030, a level of fi nancial Thus, as a group, the developing world will see development comparable to the United States in its investment rate exceed that projected under the early 1980s. Similarly, the quality of insti- the gradual convergence scenario by an average tutions in China, Indonesia, and Russia would of about one percentage point throughout the reach, by 2030, a level of the rule of law that is 20-year projection horizon. approximately two-thirds that of Scandinavian For high-income countries, however, the economies such as Norway or Sweden today. investment rate under this rapid convergence Importantly, the rate of progress in these variables scenario will be marginally lower—by about one is assumed to differ between countries: for econo- percentage point—with a widening disparity mies starting further away from the frontier, the between the two scenarios as 2030 approaches. rate of growth will be relatively faster (the specific Overall, the global investment rate rises slightly assumptions for this scenario are documented in (because of the relatively larger size of develop- detail in the GDH online annex 1.6). ing economies), although it remains around 23 The results of this rapid convergence scenario percent. This uptick, nevertheless, will be smaller indicate that global investment will be higher than the decline in global saving rates under this by 2030—led by substantially higher levels of rapid convergence scenario (to be discussed in investment in the developing world than under chapter 2), which results in some upward pressure the gradual convergence scenario—to a total of on yields worldwide (discussed in chapter 3). $26.7 trillion (figure 1.16, panel a). The increases Even though the investment rate for develop- in investment in developing countries are under- ing countries collectively will rise slightly, there pinned by higher productivity, with some rein- will be significant heterogeneity in their invest- forcing effect from changes in the structural vari- ment rates in the rapid convergence scenario rela- ables, especially from financial development.18 tive to the gradual convergence scenario (figure 44 The Emerging Pattern of Global Investment Global Development Horizons FIGURE 1.16 Total global investment will be higher for developing countries in the rapid convergence scenario (panel a), but this belies substantial heterogeneity in individual developing-country experiences (panel b) a. Total investment, by income group b. Investment rate, selected developing economies 30 40 35 25 30 20 25 $ trillions Percent 15 20 15 10 10 5 5 0 0 2005 2010 2015 2020 2025 2030 Brazil China India Indonesia Mexico Russian Sub- Federation Saharan Africa High income (rapid) High income (gradual) Developing (rapid) Developing (gradual) Gradual convergence World (rapid) World (gradual) Rapid covergence Sources: World Bank projections, supplemented with calculations using data in the World Bank Global Economic Monitor database and World Bank World Development Indicators database. Note: Figure projections assume two scenarios: a gradual convergence scenario of annual world economic growth of 2.6 percent over the next two decades, of which the developing (high-income) growth will average 5.0 (1.0) percent annually; and a rapid convergence scenario of annual world economic growth of 3.0 percent over the next two decades, of which the developing (high-income) growth will average 5.9 (1.0) percent annually. 1.16, panel b). For example, China and India The opposite is true of Brazil, Indonesia, and will experience slightly higher rates of invest- Mexico. ment, on the order of around one percentage point. At the other end of the spectrum, econo- mies such as Brazil, Indonesia, and Mexico will Transformations in sectoral experience small declines in their investment rates, of 0.4–1.1 percent. Th is variation between investment in developing countries is attributable to two mutually off set- countries ting factors: (a) a positive contribution from high A shift toward investment growth rates, rapidly growing financial maturity, in services can be expected and fast institutional advancement, all of which attract greater investment fi nancing and hence The transformations in the global picture of raises investment; set against the fact that (b) investment will not occur independently of dra- output may grow even faster, so that as a share of matic shifts in the sectoral distribution of invest- this much higher GDP level , the investment rate ment within developing countries. One major falls (in terms of investment levels, of course, trend in sectoral shifts concerns the anticipated there are absolute increases in investment flows). increase in the proportion of investment in the For China and India, therefore, the growth in all services sectors.19 In the gradual convergence three of these drivers of investment financing—in scenario, 20 by 2030, investment demand by the conjunction with very little change in global sav- services sector as a proportion of total invest- ing—do not lead to output growth outstripping ment demand will rise by about four percentage investment growth, and so investment rates are points in high-income countries, reaching a high higher than in the gradual convergence scenario. of 84 percent in Japan and the United States. In Global Development Horizons The Emerging Pattern of Global Investment 45 developing countries, the increase will be even the future, and services in general will be more larger: the services sector proportion of invest- tradable. This increasing tradability is already ment demand will increase by close to five per- evident in the information and communication centage points (to 60 percent) overall, with the technology (ICT) portion of the services export largest increases in India (from 60 percent to 70 sector in some developing countries, such as vir- percent) and Russia (66 percent to 73 percent).21 tual medical diagnostics in India and Pakistan, There will also be significant upticks in certain human resource processing in the Gulf coun- countries that currently hold relatively small ser- tries, and call centers in African countries such vices shares, such as Indonesia, where the share as Kenya and South Africa (Anand, Gable, and can be expected to rise by around six percentage Mishra 2011). points (table 1.2). The developing economies of The final trend pertains to increased infra- East Asia (excluding China and Indonesia) and structure needs as economies develop (in emerg- Sub-Saharan Africa (excluding South Africa) will ing economies), coupled with the need to upgrade also experience significant increases in their ser- aging infrastructure (in advanced economies), vice sector shares of roughly 6 percentage points. both of which will channel more investment into The increase in the representation of services the services sector. in total investment will be supported by four dis- Developing countries will continue to be the tinct trends: world’s manufacturing workshop. Collectively, developing countries will invest almost twice the • Increasing per capita incomes in the largest amount that their high-income counterparts will developing countries in manufacturing by 2030 (compared with four- • Demographic change fifths of that of high-income countries in 2010). • A rising share of services components along In the currently lower-middle-income economies, the supply chain manufacturing investment will grow steadily over • Infrastructure investment needs (infra- the next two decades: Indonesia and India, for structure being defined in this report, as it instance, will have average annual manufactur- is conventionally, as a service) ing investment growth of around 7 percent and 5 percent, respectively, and in Sub-Saharan Africa The first effect is well known: economies as a whole investment will grow at an annual tend to experience higher services value-added rate of somewhat less than 5 percent (table 1.2). shares as their per capita incomes rise. India, for Other smaller economies in East Asia, such as the instance, had a services share of 38 percent of Philippines and Vietnam, could also contribute output in 1960, but as of 2010 this had risen to to the global manufacturing picture, especially 55 percent. Similar trends have occurred in devel- given their physical proximity to the East Asian oping economies as diverse as South Africa (from production network. Manufacturing in the rest 51 percent to 67 percent over the same period) of developing Asia will grow by an average of and Argentina (from 39 percent to 59 percent about 4 percent per year. between 1965 and 2010). In some high-income countries, there will be The second effect will stem from rising health- a significant falloff in manufacturing investment: care needs faced by aging populations in many Japan, for example, will see its manufacturing regions (this is discussed more thoroughly in investment steadily shrink each year, to reach 15 chapter 2), alongside smaller youth populations percent of its total investment in 2030. Perhaps in search of higher-quality educational services. somewhat surprisingly, Europe and the United These demographic pressures will result in a States will experience small, but nevertheless greater demand for investment in services. positive, annual growth rates in manufacturing Th ird, with perhaps the notable exception of investment, although the share of manufacturing infrastructure—which is classified as a service in in their investment bills will decline (by two per- any case—agricultural and manufactured goods centage points each; it will decline by twice that will embed a far greater services component in in Japan). 46 The Emerging Pattern of Global Investment Global Development Horizons TABLE 1.2 Both developing and high-income countries will see a rise in the share of investment devoted to the services sector in the future Agriculture Manufacturing Services Share (%) Growth (%) Share (%) Growth (%) Share (%) Growth (%) Developing countries East Asia and Pacific 11 6.4 34 7.1 56 8.0 China 10 7.4 35 7.7 56 8.6 Indonesia 22 4.5 24 5.1 54 6.2 Eastern Europe and Central Asia 12 1.8 18 2.8 70 3.8 Russian Federation 17 1.7 11 3.1 73 3.9 Latin America and the Caribbean 10 3.0 19 3.0 71 4.1 Brazil 11 3.4 16 2.9 74 4.3 Mexico 4 2.8 20 2.1 76 3.4 Middle East and North Africa 39 3.1 16 3.9 46 5.0 South Asia 14 5.1 17 6.5 69 8.1 India 13 5.4 17 7.0 70 8.6 Sub-Saharan Africa 26 5.4 17 4.6 57 5.8 South Africa 4 1.6 21 2.1 75 3.3 High-income countries Europe 3 −0.9 16 0.9 82 1.8 Japan 1 −1.2 15 −0.1 84 1.1 United States 3 −0.8 13 1.1 84 2.3 Source: World Bank projections. Note: Investment share is computed as the sectoral share of total investment for each respective sector in 2030. Investment growth is computed as the compound annual percentage growth in sectoral investment for each respective sector between 2010 and 2030, measured in constant prices. Agriculture is defined to include natural resources, manufacturing is defined to include capital goods, and services are defined to include infrastructure and construction. Table projections assume a gradual conver- gence scenario of annual world economic growth of 2.6 percent over the next two decades, of which developing (high-income) growth will average 5.0 (1.0) percent annually. In addition to the growing share of global man- to lower the price of certain product lines, such as ufacturing investment represented by developing single-use medical devices, that would otherwise countries in the coming decades, the nature of be unaffordable or inaccessible, especially in low- manufactured goods will change. Manufactured income countries when they are produced by the goods are likely to embed greater services com- traditional manufacturing industry. 3D-printing ponents in the future and will be more techno- techniques could also defeat component obsoles- logically sophisticated in ways beyond quality cence (supporting poorer farmers and craftsmen improvements, while also becoming increasingly who cannot afford equipment and regular prod- transportable and transferable. Technological uct upgrade cycles) and potentially encourage sophistication could be realized as enhanced the development of products for which the end interdevice interconnectivity (such as home appli- users are predominantly located in low-income ances that seamlessly interact with either a central countries. computer or a mobile phone application) and as For most countries, a larger share of investment cost-competitive, environmentally friendly prod- activity within the services sector will be mirrored ucts (such as high-efficiency vehicles and air con- by a shrinking share held by the agriculture sec- ditioners operating on solid-oxide fuel cells, bio- tor. In developing countries as a whole, the share fuels, or electrofuels). Other advances in additive of investment in agriculture will shrink from manufacturing, such as three-dimensional (3D) 16 percent to 13 percent, although the variance printing, will allow more on-demand, physically across countries will be fairly large. Brazil and complex, and highly customizable production. Sub-Saharan Africa will emerge as the (relative) Such manufacturing processes have the potential leaders in this regard: their shares of agricultural Global Development Horizons The Emerging Pattern of Global Investment 47 investment, as a share of their total gross invest- of the most pressing challenges for developing ment, will be mostly maintained through 2030, countries, especially when considered alongside growing at annual rates of 3 percent and 5 per- rapid advances in per capita income and popula- cent, respectively. This growth would ref lect tion growth pressures in the poorest parts of the Brazil’s comparative advantage in agricultural world. innovation, along with the adaptation of agricul- To better understand what such infrastructure tural technologies for the next frontier of mass investment needs are, it is necessary to go beyond agricultural cultivation in Africa, and is likely to likely investment paths and toward anticipating be realized in the form of locally adapted tech- the investment expenditures that would result if nologies for tropical-weather farming and yield desired targets for infrastructure were fully satis- enhancement. Almost everywhere else in the fied. Indeed, even excluding maintenance costs, world, however, agricultural investment will con- such needs will substantially exceed likely levels of tract more sharply: in high-income countries— actual infrastructure investment.22 Furthermore, excluding investment in natural resources by the given the central role that the public sector typi- high-income countries of the GCC—agricultural cally plays in infrastructural services, meeting investment will shrink, on average, to just 2 per- such anticipated infrastructure needs is a fi rst- cent of the investment bill, and even in the devel- order concern for developing-country policy oping world, agriculture will amount to only 13 makers.23 percent of investment expenditures. In general, Moreover, increased investment in sound declines in pure agriculture (excluding natural infrastructure can serve a broader catalytic role resources) around the world will be somewhat in enabling sectoral shifts in investment as econo- offset by increases in natural resource investment; mies develop. Low-income countries seeking to such developments are likely to reflect the resur- break into global manufacturing markets will gence of shale gas extraction through horizontal need reliable electricity generation, improved drilling and hydraulic fracturing. road and rail networks, and expanded port facili- The bottom line is that there will always be ties. Lower-middle-income economies will need new technologies that absorb investment, and higher-quality roads, more sophisticated water reallocations of investment between sectors ulti- and drainage systems, and improved Internet and mately are led by changes in the relative returns cellular phone coverage as their middle classes to investment in each sector (which in turn is develop and as demand rises for these infrastruc- a function of relative output and factor prices tural services. And upper-middle-income nations between sectors). In short, the world is not about will need modern, high-speed ICT networks and to “run out” of investible opportunities. high-capacity power generation to effectively compete in high-tech manufacturing and glob- ally tradable services while upgrading the overall Developing-country infrastructure quality of their service sectors. investment needs will be A central question that developing-country substantially higher in 2030 policy makers will need to confront as they seek One investment priority consistent with the to scale up their infrastructure investments is how forthcoming shift toward services investment technological progress has changed, and will con- is the pressure in most developing countries for tinue to change, the mechanisms and possibilities expanded infrastructural services. As discussed surrounding infrastructure service delivery and in the previous subsection, services investment operation. The meaning of the term “infrastruc- in developing countries will rise substantially ture” is no longer self-evident: four decades ago, from current levels, amounting to almost $10 fixed-line telecommunications cables appeared trillion by 2030. Of this amount, $626 billion to be a natural monopoly benefiting from pub- (slightly more than 6 percent) will be spent on lic sector provision; starting in the 1990s, a infrastructure. Yet in spite of major advances over wave of privatizations around the world took the past decade, infrastructural needs remain one hold, expanding penetration and coverage and 48 The Emerging Pattern of Global Investment Global Development Horizons enhancing efficiency (Li and Xu 2004). Today, Infrastructure investment needs are likely to be fixed telephone lines are almost anachronistic, most concentrated in East Asian and South Asian with many developing countries leapfrogging countries: the amount in just those two regions fi xed lines and installing cellular transmitter tow- will total $501 billion in 2030, reflecting contin- ers instead. Finally, improved efficiencies due to ued strong demand for infrastructure by the very technological improvements in grid storage and large Chinese and Indian populations (figure 1.17, digital power conversion could mean that future panel b). In China, power needs will dominate, infrastructure spending in power will shift from but this will be coupled with fi xed-line telecom- investment in production toward storage. munications demand as more and more Chinese Although the specific composition and nature consumers and businesses seek high-speed connec- of infrastructural capital will adapt according to tivity. In India, investment will need to be directed changes in the world economy, the nature of eco- toward improving the country’s very poor road nomic services that characterize infrastructure system, although mobile connectivity will draw nevertheless remain the same: high capital inten- infrastructure investment as well. sity, elements of natural monopoly, and location- Yet even the substantial sum of $501 billion specific investment.24 Accordingly, the discussion may underestimate the tremendous infrastruc- in this report will retain the traditional categories ture needs of Asian countries when considering of power, transportation, telecommunications, international infrastructure links (those involv- and water and sanitation, while acknowledging ing improving access between two or more that the specific technologies implied by these cat- countries). Using this more expansive definition egories require some flexibility in interpretation. bumps up infrastructure needs for Asia to an Though estimates are not consistent, analysis estimated $750 billion per year for the decade of demand for infrastructure services generally between 2010 and 2020 (ADB 2010)—about 26 reveals that although demand increases with per percent more than estimated expenditures when capita income, the pace of growth in demand is only intracountry needs are considered. faster at lower per capita income levels. Currently, As for Sub-Saharan Africa, the region’s infra- per capita electricity consumption in developing structure investment needs do not appear sig- countries ranges from an average of 452 kilowatt- nificantly greater than that of other developing hours (kWh) for low-income countries to 3,073 regions in absolute terms. Were they to be met kWh in upper-middle-income countries; by com- in full, however, the amount would constitute a parison, the average in developed countries is significant share of total investment in the region: 10,408 kWh. The sheer number of people who $97 billion, or 15 percent of the total $662 bil- lack access to other forms of infrastructure pro- lion. Moreover, this estimate excludes investment vides another hint of how much investment will needs arising from the need to rehabilitate the be required: 1.1 billion people in the world cur- estimated 30 percent of existing infrastructure rently lack access to safe drinking water, and 2.4 that has been poorly managed (AfDB 2010) or billion face inadequate sanitation services. the additional investment required to enhance In 2030, infrastructure investment needs in regional interconnectivity (PIDA 2011). developing countries will amount to an estimated Importantly, infrastructure investment needs $864 billion, as measured in 2010 dollars (figure at the regional level will not evolve monotoni- 1.17, panel a).25 These needs will be concentrated cally. Given the easy availability of cellular mobile in the power ($243 billion) and transportation phones and their widespread adoption, demand ($254 billion) subsectors, which will together for cellular telecommunications infrastructure, in account for more than half of all infrastructure particular, is likely to be greatest over the next five needs. Fixed-line telecommunications infrastruc- years. This front-loaded demand is likely to reach ture—presumably in the form of digital and opti- saturation in a significant number of developing cal cables—will also be a major area of invest- countries by 2017 or 2018.26 As these needs taper ment, amounting to $172 billion (or 20 percent off, fixed-line infrastructure will increase in impor- of total infrastructure investment needs). tance. Power and transportation needs are likely to Global Development Horizons The Emerging Pattern of Global Investment 49 FIGURE 1.17 Annual infrastructure investment needs in developing countries will be substantial for the next two decades (panel a), with the greatest needs arising in East and South Asia (panel b) a. Total infrastructure needs, 2010–30 b. Regional infrastructure needs, 2030 Water 350 800 Sanitation 300 Fixed 250 600 Mobile 200 $ billions $ billions 400 150 Transport 100 200 50 Power 0 0 2011 2014 2017 2020 2023 2026 2029 Sub- E. Asia E. Eur. L. Amer. Mid. S. Asia Saharan & Pacific & Cent. & the East & Africa Asia Caribbean N. Africa Water/sanitation Telecoms Transportation Power Source: World Bank projections. Note: Details of the methodology underlying the projections are given in the Global Development Horizons online annex 1.5. grow more gradually, a reflection of more gradual sheer size of infrastructure investment needs changes in population demands for these services. (which does not distinguish between public or private investment) suggests that adequately meeting these needs will call for a rethinking Enormous resource mobilization of existing institutional arrangements of service will be required to meet delivery, increased internationalization of fund- infrastructure financing needs ing sources, and innovative fi nancing strategies in developing countries that are better designed to meet the high levels of Successfully meeting developing countries’ cumu- capital intensity. Concurrent with meeting these lative infrastructure financing needs of $14.6 tril- purely financial goals is the necessity of being lion between 2012 and 2030 will require an enor- keenly aware of the broader public policy issues mous mobilization of financing—one akin to of efficiency, access by the poor, and quality of frontier expansion by European countries in the services. New World in the 19th century, which generated More than other sectors, infrastructure massive investment needs in the form of railways, financing has witnessed a distinct shift in terms real estate, and large-scale agricultural projects. of government policy, public attitude, and intel- Although infrastructure investment has tradi- lectual treatment. Since the mid-20th century, tionally been the purview of the public sector, as these trends first encouraged a model of pub- discussed earlier, this trend is changing, toward lic ownership and control, vertically integrated greater public-private partnership. Indeed, the industrial organization, and public fi nancing of 50 The Emerging Pattern of Global Investment Global Development Horizons investment. Since the 1990s, the common model international banks) to provide the necessary has shifted toward market liberalization, unbun- capital for large-scale infrastructure projects is dling of power and telecom utilities, privatization questionable, given weaknesses in their postcrisis of public assets, and fi nancing through private balance sheets (an issue also discussed in chapter markets. More recently, the model has evolved 3). Considered together, these factors will limit toward public-private partnerships in financing, the options available to policy makers seeking to especially in middle-income countries. finance infrastructure. The evolution toward greater private sector One respite offered by the long-term dura- involvement in infrastructure financing could not tion of infrastructure assets is that, in spite of have come at a more opportune time. The financ- global rollbacks in overall syndicated lending, the ing of future infrastructure investment needs will nature of the investment allows for more inno- occur in a global environment where the public vative fi nancing options through, for instance, sector will almost certainly face funding difficul- increased securitization of future cash flows. In ties. As will be further discussed in chapter 2, fact, there is evidence that this shift toward bond fiscal balances worldwide were compromised as a financing has already occurred (figure 1.18) and result of the fi nancial crisis. Th is was especially will continue to expand in the future, perhaps the case in advanced economies but also in the accounting for as much as half of all infrastruc- developing world. Furthermore, questions about ture investment financing. the sustainability of sovereign debt, as detailed The current trend toward globalization of in chapter 3, are likely to negatively affect the corporate finance is also likely to intensify in the global outlook for public sector access to global future, and this will further support infrastructure capital markets. Even the ability of traditional financing sourced from global capital markets. private sector financial intermediaries (such as Such bond market maturation will ultimately lead to lower credit spreads and hence the cost of capi- tal. There is evidence that this is already under way: FIGURE 1.18 Infrastructure investment financing modes emerging market corporate bonds carried spreads have changed over time and will likely favor greater bond over comparable U.S. Treasury securities of about financing in the future 450 basis points in the early 1990s, but the average spread narrowed to 392 basis points in 2011. Of course, such financing structures are feasible only when the underlying financial, regulatory, legal, Bonds and institutional frameworks are sound, which in ($434.3b) Loans turn is premised on the developing world’s prog- Bonds ($54.5b) (projected) ress in developing their financial institutions. In ($3,807b) situations where such frameworks are not properly 2001 1990– in place or are in a state of transition, investors are –11 2000 likely to increase their hurdle rates. This is typically 2012–22 Bonds passed on as higher tariffs for infrastructure ser- (projected) (projected) ($5,033b) Loans ($246.2b) vices, which, as experience indicates, leads to such services being prohibitively expensive, especially Loans ($884.2b) for the poor. But innovative financing need not be limited to securitization. Innovations can also occur in the manner by which funding is raised, especially when aided by modern ICT (box 1.8). Source: World Bank projections, supplemented with calculations using data from Dealogic Bondware and Loanware. Note: Infrastructure financing is calculated as a share of cumulative investment financing needs during the respective time periods. Total values, in billions of current dollars, are given Conclusion and policy directions in parentheses. Projections for 2012–22 are based on the assumption of similar growth rates between the second two periods as between the first two periods and zero infl ation. The The future face of global investment activ- cumulative infrastructure investment bill for 2012–22 is $8.8 trillion. ity will see developing countries in far greater Global Development Horizons The Emerging Pattern of Global Investment 51 BOX 1.8 Alternative financing mechanisms have begun to be adopted in financing infrastructure The idea behind crowdsourced capital has a relatively of this kind are possible pathways into a future where long, and reasonably venerable, history in developing- infrastructure financing breaks free from traditional world fi nance. Distilled to its essence, crowd fi nanc- modalities and embraces the power and connectivity ing seeks to minimize the role of intermediaries in made available by modern information and communica- matching savers with investors. On the saving front, tions technologies. rotating savings and credit associations as well as Another alternative mechanism that has begun to the more recent innovation, community finance asso- take hold in infrastructure financing is Islamic finance, ciations, can be regarded as forms of crowd fi nanc- particularly for large-scale projects in the Middle East. ing (Besley, Coate, and Loury 1993). As for investing, Although the broader applicability of Islamic finance some microfinancing organizations, such as Kiva, have may continue to face economic, legal, and systemic been known to employ crowd funding as their pri- challenges (Hesse, Jobst, and Solé 2008), there is mary financing mechanism (Cull, Demirgüç-Kunt, and reason to believe that Islamic fi nance could be cred- Morduch 2009). ibly scaled up as an alternative medium for cross- Information technology, as well, has expanded the border infrastructure fi nancing. The tenets of Islamic potential to draw on diverse sources to fi nance infra- fi nance—which focus on direct participation in asset structure in developing countries. One such initiative, performance—are well suited for the long-term nature Sokoni, has established a digital marketplace that con- of infrastructure finance, where the rate of return could nects donors and investors directly to project develop- be difficult to predetermine and may be realized years ers and governments. Innovative financing mechanisms after an initial investment. prominence, investing more than high-income that when large economies engage in profligate countries in absolute value and as a share of global public policies or unrestrained private borrowing, investment. Since 2000, developing countries’ there could be material contractions in the global investment as a proportion of their own output supply of saving (to the extent that such actions are has been substantially higher than that of high- not fully offset by Ricardian equivalence). This can income countries. These trends are not expected choke off investment in small, vulnerable develop- to continue indefinitely into the future, but their ing countries, which typically have less fiscal space effects are already palpable, and their full ramifi- to offset negative shocks from abroad. cations will be realized over the next two decades. The anticipated fundamental forces that Although this chapter focuses on real invest- will shape the global economy in the years to ment activity—in particular, gross capital forma- come—asynchronous demographic transforma- tion—across countries, the current account iden- tion, asymmetric productivity growth, structural tity implies that investment should be understood change, and financial globalization—will there- alongside saving and capital flows.27 For many fore alter not just investment needs and opportu- small developing economies, shocks to saving in nities but also saving decisions and choices along major economies, such as the United States, will with the cross-border flow of financial capital and inevitably affect their domestic patterns of invest- international investment positions. Accordingly, ment (Feyrer and Shambaugh 2012). This result is the following chapters will address the issue of a double-edged sword. On one hand, it suggests global patterns in saving (chapter 2), along with that developing countries can leverage increases in the saving-investment imbalance that constitutes national saving in large, high-income countries to international capital flows (chapter 3), tying the relax domestic saving constraints and boost invest- notion of investment needs with opportunities ment at home. On the other hand, it also means together at the global level. 52 The Emerging Pattern of Global Investment Global Development Horizons Policy makers have an important desirable projects and activities. Moreover, role to play in supporting a attracting investment from abroad often requires favorable investment climate the presence of a professional, knowledgeable, and investor-focused investment facilitation Because investment is primarily a private sector- agency, equipped with the capacity to ease cross- led activity, government policy action in most border informational asymmetries (World Bank countries exerts only an indirect effect on chang- 2012c). ing the level of investment in a country. Rather, government’s role is often to establish an envi- For policy makers interested in boosting ronment—the investment climate—that is sup- investment, a central element of the invest- portive of robust and sustainable investment ment climate is the level of development of activity. This is not to say that public investment the fi nancial sector. As discussed in this chap- is irrelevant, since there may be a role for inter- ter, an increasingly mature financial system is vention where market failures are clear and where associated with rising levels of investment. More social returns are especially large. For instance, important, well-functioning financial systems governments in developing countries often have ensure a superior mobilization and corresponding a well-defined role in the provision of infrastruc- allocation of saving toward the most productive ture, which serves as an important foundation investment opportunities (Levine 2005) while for private investment activity. But the policy providing investors diversification and risk man- framework for investment is ultimately an indi- agement benefits. A lengthy discussion of policies rect one. that can stimulate financial sector development is With this overriding framework in mind, the beyond the scope of this report, but recent work importance of changes to structural factors in by the World Bank offers a menu of ideas for attracting investment from the global pool of sav- jump-starting such development, especially as it ing—amply demonstrated by the difference in pertains to the role of the state, and for guard- investment patterns in the rapid versus gradual ing against circumstances of state capture and convergence scenarios discussed earlier—suggests government failure (World Bank 2012b). For the that countries seeking to compete in the future purposes of relaxing credit-related constraints, world economy will need to pay attention to fos- especially for small and medium-size enter- tering a favorable investment climate. In general, prises, the establishment of a credit information a favorable investment climate is characterized agency can be especially helpful. It is important not only by traditional policy areas that can foster to point out, however, that any policies designed private sector investment, such as a stable macro- to encourage financial sector development—a economic and regulatory regime, but also by (a) long-run goal—should not be interpreted simply the broader institutional environment in which as the pure expansion of credit; such a narrow firms operate (which includes secure property view may give rise to the pursuit of unsustainable rights and stable rule of law); (b) the governance short-term policies that are ultimately detrimen- framework, such as adequate control of corrup- tal to economic and financial stability. tion and a limited burden from—and simplified Major developing economies have not stood administration of—direct and indirect taxation idly by in seeking to foster enhanced fi nancing (which includes the risk of outright appropria- facilities for South-South investment. The BRICS tion); and (c) the level of maturity of the financial countries (Brazil, Russia, India, China, and sector. In this sense, getting the investment cli- South Africa) touted a South-South development mate right shifts the policy dialogue away from bank in early 2012, for example, as a mechanism a laundry list of specific actions and policies for to help recycle surplus saving for their developing reform, and toward a discussion of what consti- countries’ investment needs, while regional mul- tutes the appropriate set of rules, governance, and tilateral development banks have also expanded incentives that would stimulate the private sec- their operations dramatically in recent years (a tor to invest in economically efficient and socially topic that will be revisited in chapter 3). Global Development Horizons The Emerging Pattern of Global Investment 53 Improvements in the overall quality of gover- maintained; and when interventions have a clear nance in developing countries, especially with sunset clause. regard to the rule of law, can also have a major impact on the investment climate and serve to If policy makers do ultimately decide to pursue attract fi nancing for investment from global the targeted intervention route, their eff orts sources. This is not surprising given that respect may be best placed in supporting the upcom- for the rule of law is essential to ensuring that ing expansion in service sector investment (as investors can realize returns from their invest- discussed earlier), especially in areas (such as ment activities. However, successfully address- education and health care) where spillovers from ing governance challenges as they pertain to the positive externalities could be especially high. investment climate calls for going beyond de jure Even in manufacturing and agriculture, there are corrections to the letter of the law or salutary certainly subindustries (like “green” technology actions such as the establishment of independent components or tropically adapted seeds) where investment and regulatory commissions. Actions social returns appear to be sufficiently large to will need to ensure that the elements of rule of justify the risks of direct intervention. law are maintained de facto by bureaucrats and Although the institutional design that best politicians and that formal institutional bodies supports such successful interventions is likely to can enforce violations of rule of law in an uncom- be idiosyncratic, a wholesale rejection of a public promised fashion. role in investment is probably unwarranted. Just Such fundamental reform, of course, is likely as important, governments that currently engage to encounter resistance by entrenched interests, in policies that, either directly or indirectly, favor and so reform efforts must be undertaken with investment in agriculture and manufacturing (for a mind toward building coalitions for change example, many governments explicitly prohibit through expanding the scope of stakeholders and FDI in certain “sensitive” or “strategic” services) encouraging partnership models of stakeholder will place a burden on their economies’ efficiency interactions (World Bank 2008). Even seemingly in a world where services account for a larger innocuous reform efforts, such as routine, unbi- share of output. ased implementation of investment-related com- petition law, can face staunch political opposition Indeed, the area where government policy if they would weaken existing monopolistic or is likely to offer some of the greatest social oligopolistic positions of incumbent businesses. returns is in the provision of infrastructure. As discussed earlier, this is also a sector of the econ- Successful investment policy, of course, seeks omy that will likely face increased demand for to not only correct areas of government failure future investment. Yet even in this area, which as described above but also to suggest direct has been the traditional bulwark of public sec- intervention in cases where market failures are tor investment, government roles are evolving. sufficiently evident (and remain uncorrected) PPPs have helped move along projects that cash- or where social returns are suffi ciently high. constrained public sectors sometimes have had Recent research (Aghion et al. 2012; Bloom and trouble completing on their own. For example, van Reenen 2007; Rodrik 2009) has revived the by offering guarantees on minimum concession idea that targeted interventions can better align charges for toll-road operations, private investors private sector incentives and foster private sector in Chile, India, and Mexico have funded trans- competition (rather than rent seeking)—espe- portation infrastructure in many remote areas cially when such policies are not so much targeted that were previously underserved. In Jordan and at a specific sector (where the risk of subsequent Russia, new airports have been built and operated capture can be high) but at particular constraints by foreign investors, which in turn contribute (such as inadequate knowledge of good man- to positive network externalities for future busi- agement practices or difficulties in obtaining ness opportunities. Even in areas where citizen- credit); when high levels of accountability are state confl icts have arisen in the past—such as 54 The Emerging Pattern of Global Investment Global Development Horizons water utilities—a willingness to employ profes- in terms of both economic payoffs and political sional management in state-owned facilities has institutions, that would ensure that large invest- helped improve efficiency and service quality in ment expenditures are not misallocated. Colombia, Gabon, and the Philippines. 4. PPPs are typically formed through the establish- International financial institutions such ment of a single long-term contract between a government and a private investor that bundles as the World Bank, International Financial infrastructure investment and service provision. Corporation, and the Multilateral Investment 5. The Gulf Cooperation Council (GCC) was estab- Guarantee Agency can have an especially impor- lished in 1981 through an agreement between tant role to play in this regard. The World Bank Bahrain, Kuwait, Oman, Qatar, Saudi Arabia, has traditionally been active in infrastructure and the United Arab Emirates. financing, and it can play an important interme- 6. Indeed, in recent years there has been a major diary function in offering low-cost infrastructure shift as investors from advanced economies have financing through its participation in PPP lend- retreated and developing-country investors have ing syndicates or by designing structured financ- emerged as a major source of investment finance ing options (such as infrastructure bonds). But for infrastructure projects with private participa- the institution’s role goes beyond financing. The tion (Ettinger et al. 2005). 7. As discussed in online annex 1.1 (on the Global World Bank is now a repository for knowledge Development Horizons [GDH ] website), the defini- on PPPs in infrastructure, and as such can pro- tion of “productive physical capital” intentionally vide technical assistance and other advisory ser- excludes broader measures of capital, including vices that draw on the diversity of cross-country that of human and natural capital. Such correc- experiences in infrastructure investment projects tions may be appropriate for the study of certain (World Bank 2012d). phenomena—such as the optimal rate of natural resource extraction (World Bank 2006, 2011)— but a considered treatment is beyond the scope of this report. Moreover, as discussed elsewhere in Notes this report, there are potential problems in regard- 1. The definition of “investment” used in this chap- ing all forms of investment (private, public, or ter is, in most cases, gross capital formation, investment in nonreproducible capital) as equiva- including purchases of physical structures, plants, lent to productive capital formation. With these machinery, and equipment, together with inven- caveats in mind, the goal of this subsection is to tory accumulation, but gross of depreciation. provide a better sense of the distribution of global Further information on this and other definitions fi xed capital stocks, and caution is warranted in of investment may be found in online annex 1.1 relying on the capital stock computations pro- (available, with the rest of the annexes for this and vided here in secondary analyses. the other chapters, at http://www.worldbank.org/ 8. Although the capital stock calculations are based CapitalForTheFuture). on constant 2000 dollars and the perpetual inven- 2. It is important to be aware that measurement dif- tory method assuming a constant depreciation rate, ficulties mean that the reported sectoral marginal alternative calculations, using either PPP-adjusted products of capital (MPKs) should be regarded as investment or hyperbolic discounting of the capital indicative, rather than definitive. It is also impor- stock, deliver a similar qualitative message. tant to keep in mind that calculated MPKs cap- 9. The relationship between economic growth and ture the mean of a distribution of firm efficiencies, investment is clearly endogenous. Nevertheless, and so the most productive firms in a country there is a strong correlation between the two or sector may still be highly efficient. Analogous because sustained economic growth is the result of computations using a more familiar metric, the improvements in total factor productivity (TFP) incremental capital-output ratio, are reported in and not factor accumulation alone. TFP, in turn, online annex 1.2, with similar qualitative results. derives from a far richer set of factors over and 3. Although there is still considerable debate over the above high levels of investment. veracity of such models in stimulating growth, it is 10. This relative yield ∼ r is defined in the context of the well recognized today that any such approach can- investment financing equation in online annex 1.6 not occur in the absence of absorptive capacity, (on the GDH website). Global Development Horizons The Emerging Pattern of Global Investment 55 11. Undoubtedly, interest rates should reflect some investment projects, the importance of timing, of the differences that arise from these structural and, crucially, the inherent uncertainty involved in factors. However, interest rates may not fully assessing investment opportunities. capture the impact of structural differences due 16. The theoretical prediction of the negative effect to financial market imperfections. For example, of uncertainty on investment is more ambiguous. financial repression—common in many develop- Caballero (1991), for instance, has noted that the ing countries—may result in a distortion so that response of investment to uncertainty depends a disconnect arises between reported interest rates on assumptions regarding decreasing marginal and expected returns. returns to capital, while Lee and Shin (2001) show 12. The changes are measured by the simple average of that the share of labor in the production function the control of corruption and rule of law measures can affect the balance between the positive and in the Worldwide Governance Indicators for the negative effects of uncertainty. two years (1996 and 2010). For more information 17. As can be seen from box 1.7, figure B1.7.1, there is about the Worldwide Governance Indicators proj- significant uncertainty about the true impact of the ect, see http://info.worldbank.org/governance/wgi shock; the error bands for both impulse responses /index.asp. include zero. Nevertheless, as Jordà (2009) has 13. Theory posits that the investment rate is equal to argued, the inclusion of zero within the confidence the ratio of firms’ shadow value of capital per unit bands for the impulse response need not disqual- price of investment goods, a relationship known as ify the variable from having an effect, especially marginal Q. However, because such shadow prices if the function appears to trend smoothly in one are not observable, economists have operational- direction or another during the propagation of ized the concept by taking the ratio of the firms’ the shock, rather than vacillate unstably around (observable) market valuations to their replace- zero. It should also be noted that other studies (for ment costs of capital, or average Q. Under certain example, Carrière-Swallow and Céspedes, forth- additional assumptions—which include perfect coming) have found that the effect of uncertainty competition and perfect capital markets—these shocks on investment is actually greater for emerg- two ratios are in fact equivalent (Hayashi 1982). ing economies. Recent research has even suggested that future 18. Because labor supply remains unchanged in this investment opportunities are better captured scenario, higher levels of financial development by average, rather than marginal, Q. Empirical attract greater amounts of investment financ- Tobin’s Q studies seeking to explain investment by ing, which increases efficiency-adjusted capital Q alone, however, have had a disappointing track stocks and serves as an additional boost to growth record, which may be attributable to the diver- over and above the productivity increase. When gence between these two measures as a result of decomposed into the contribution of growth ver- financial frictions (Lorenzoni and Walentin 2007). sus structural change (by separately running the 14. A common alternative interpretation of Tobin’s Q scenario with only changes in either), the effect is that it is a proxy for adjustment costs for chang- of growth is generally twice as large as that of ing capital stocks. Absent such costs, the coef- structural changes. Although not reported, these ficient on Q would be unity. A small coefficient results are available on request. estimate for Q is thus interpreted as an indica- 19. Although one may reasonably expect that service tion of large (convex) adjustment costs (although sector investment would rise as countries become recent research by Abel and Eberly [2011] dispute richer (given the trend that output in the service this traditional interpretation). In the reported sector does rise with per capita incomes), this is sample, the coefficient for Q in the full sample not immediately obvious because the capital is 0.01, which is remarkably close to estimates intensity of the service sector differs from that of typically obtained in the literature, using far more the manufacturing sector. In the model used in sophisticated estimation techniques and alterna- this report, the capital intensity of the service sec- tive specifications. tor is indeed, on average, significantly lower than 15. Treating investment decisions as real options that of the manufacturing sector across develop- simultaneously accounts for three features of ing countries. investment that are not well captured by the tra- 20. For space reasons, this section will rely on esti- ditional approach of equating user cost to mar- mates consistent with the gradual convergence ginal product: the irreversible nature of many scenario. Although the numbers differ, the main 56 The Emerging Pattern of Global Investment Global Development Horizons messages that emerge using numbers from the estimate of $626 billion in the gradual conver- rapid convergence scenario are similar. gence scenario. These estimates also suggest that 21. Globally, the services investment share will actu- infrastructure investment grows at an annual rate ally remain stable, at 69 percent. This apparent of only 0.6 percent for 2011–30, a rate one order anomaly can be understood by the fact that any of magnitude slower than the annual growth rate change in global investment can be decomposed of services in developing countries (of 6.2 per- into two effects: change in respective investment cent). As can be seen in figure 1.17, however, this rates and change in relative sizes. Between 2010 slow rate is due mainly to a significant drop-off in and 2030, the investment rate in services will mobile infrastructure spending. Annual growth in increase in both developing and high-income infrastructure investment excluding this category regions. However, as developing countries increase grows by 2.5 percent, which remains slower than their weight in the global economy, the fact that the overall rate of services growth, but not sub- services investment was, on average, much lower stantially so. in 2010 in these economies means that the total 26. This saturation point is assumed to be at two global increase was much more modest. cellphones per user. As of 2010, only Hong Kong 22. Accordingly, in contrast to the general equilib- SAR, China, has coverage (slightly) in excess of rium model adopted for the earlier sections of this this level. chapter, this subsection supplements the earlier 27. The current account identity states that the dif- analysis with a partial equilibrium approach. This ference between national saving and investment methodology provides better estimates of invest- is equivalent to a country’s net capital flows. The ment needs because it abstracts from likely substi- surprisingly high correlation between saving and tution effects that may arise because of changes investment was first observed by Feldstein and in relative prices, technological advances that Horioka (1980), who regarded the finding as may improve energy efficiency units, and chang- evidence of home bias in saving and investment. ing preferences for infrastructure services. Con- Subsequent work has punctured this notion that sequently, the estimates provided may be higher high saving-investment correlations necessarily than actual infrastructure investment. However, imply poor capital mobility (Kraay and Ventura because estimates of investment needs provided 2000), although the more general point that the here do not account for maintenance, includ- three variables are closely linked should not be ing such maintenance needs into infrastructure understated. investment may result in estimates higher than even the figures presented here. 23. Although not the focus of this report, anticipated infrastructure needs are also a priority for many References high-income economies, which are experiencing Abel, Andrew B., and Janice C. Eberly. 2011. “How the gradual erosion of their domestic infrastruc- Q and Cash Flow Affect Investment Without Fric- ture and the need to upgrade legacy systems. tions: An Analytic Explanation.” Review of Eco- Indeed, the anticipated high relative rental rates nomic Studies 78 (4): 1179–1200. in high-income countries could be a major factor Acemoglu, K. Daron, Simon Johnson, and James A. supporting future foreign investment in infra- Robinson. 2005. “Institutions as a Fundamen- structure in high-income economies. tal Cause of Long-Run Economic Growth.” In 24. Such a definition will, for instance, exclude fi xed Handbook of Economic Growth, vol. 1, ed. Philippe capital for education, health, and sports and rec- Aghion and Steven N. Durlauf, 385–472. Amster- reation because these facilities typically do not dam: Elsevier. reflect the continuously diminishing marginal ADB (Asian Development Bank). 2010. Infrastructure costs that are characteristic of natural monopolies. for a Seamless Asia . Tokyo: Asian Development Likewise, cellular transmitter towers, which are Bank Institute. generally well provided by private sector, would AfDB (African Development Bank). 2010. African not constitute infrastructure, while fiber-optic Development Report 2010: Ports, Logistics, and Trade and digital cable deployment would be considered in Africa. Oxford: Oxford University Press. infrastructure. Aghion, Philippe, Mathias F. Dewatripont, Luosha Du, 25. For reasons documented in endnote 24, this Ann E. Harrison, and Patrick Legros. 2012. “Indus- estimate is around 18 percent higher than the trial Policy and Competition.” Working Paper Global Development Horizons The Emerging Pattern of Global Investment 57 18048, National Bureau of Economic Research, Calderón, César A., and Luis Servén. 2010. “Infra- Cambridge, MA. structure in Latin America.” Policy Research Work- Aizenman, Joshua, and Nancy P. Marion. 1999. ing Paper 5317, World Bank, Washington, DC. “Volatility and Investment: Interpreting Evidence Carrière-Swallow, Yan, and Luis Felipe Céspedes. from Developing Countries.” Economica 66 (262): Forthcoming. “The Impact of Uncertainty Shocks 157–79. in Emerging Economies.” Journal of International Anand, Rahul, Susanna Lundstrom Gable, and Sau- Economics. rabh Mishra. 2011. “Service Export Sophistication Carruth, Alan A., Andrew P. Dickerson, and Andrew and Economic Growth.” Policy Research Working G. Henley. 2000. “What Do We Know About Paper 5606, World Bank, Washington, DC. Investment Under Uncertainty?” Journal of Eco- Bardhan, Pranab K. 2010. Awakening Giants, Feet of nomic Surveys 14 (2): 119–53. Clay: Assessing the Economic Rise of China and India. Caselli, Francesco, and James D. Feyrer. 2007. “The Princeton, NJ: Princeton University Press. Marginal Product of Capital.” Quarterly Journal of Bartelsman, Eric J., John C. Haltiwanger, and Ste- Economics 122 (2): 535–68. fano Scarpetta. 2009. “Measuring and Analyzing Chinn, Menzie D., and Eswar S. Prasad. 2003. Cross-Country Differences in Firm Dynamics.” In “Medium-Term Determinants of Current Accounts Producer Dynamics: New Evidence from Micro Data, in Industrial and Developing Countries: An Empir- ed. Timothy Dunne, J. Bradford Jensen, and Mark ical Exploration.” Journal of International Economics J. Roberts, 15–76. Chicago: University of Chicago 59 (1): 47–76. Press. Claessens, Stijn, and Ronald C. Duncan, eds. 1993. Benhabib, Jess, and Mark M. Spiegel. 2000. “The Role Managing Commodity Price Risk in Developing Coun- of Financial Development in Growth and Invest- tries. Baltimore: Johns Hopkins University Press. ment.” Journal of Economic Growth 5 (4): 341–60. Commission on Growth and Development. 2008. Besley, Timothy J. 1995. “Property Rights and Invest- The Growth Report: Strategies for Sustained Growth ment Incentives: Theory and Evidence from Ghana.” and Inclusive Development. Washington, DC: The Journal of Political Economy 103 (5): 903–37. World Bank. Besley, Timothy J., Stephen Coate, and Glenn C. Cooley, Thomas F., Ramon Marimon, and Vicenzo Loury. 1993. “The Economics of Rotating Sav- Quadrini. 2004. “Aggregate Consequences of Lim- ings and Credit Associations.” American Economic ited Contract Enforceability.” Journal of Political Review 83 (4): 792–810. Economy 112 (4): 817–47. Bloom, Nicholas. 2009. “The Impact of Uncertainty Cull, Robert J., Asli Demirgüç-Kunt, and Jonathan J. Shocks.” Econometrica 77 (3): 623–85. Morduch. 2009. “Microfinance Meets the Market.” Bloom, Nicholas, and John M. van Reenen. 2007. Journal of Economic Perspectives 23 (1): 167–92. “Measuring and Explaining Management Practices Dailami, Mansoor. 1992. “Optimal Corporate Debt Across Firms and Countries.” Quarterly Journal of Financing and Real Investment Decisions under Economics 122 (4): 1351–1408. Controlled Banking Systems.” In Business Finance Caballero, Ricardo J. 1991. “On the Sign of the Invest- in Less Developed Capital Markets, ed. Klaus P. ment-Uncertainty Relationship.” American Eco- Fischer and George J. Papaioannou, 229–49. West- nomic Review 81 (1): 279–88. port, CT: Greenwood Press. ———. 2008. “On the Macroeconomics of Asset Dailami, Mansoor, and Robert Hauswald. 2007. Shortages.” In The Role of Money: Money and Mon- “Credit-Spread Determinants and Interlocking etary Policy in the Twenty-First Century, ed. Andreas Contracts: A Study of the Ras Gas Project.” Journal Beyer and Lucrezia Reichlin, 272–83. Frankfurt: of Financial Economics 86 (1): 248–78. European Central Bank. Dawson, John W. 1998. “Institutions, Investment, and Caballero, Ricardo J., Emmanuel Farhi, and Pierre- Growth: New Cross-Country and Panel Data Evi- Olivier Gourinchas. 2008. “An Equilibrium Model dence.” Economic Inquiry 36 (4): 603–19. of ‘Global Imbalances’ and Low Interest Rates.” Dixit, Avinash K., and Robert S. Pindyck. 1994. Invest- American Economic Review 98 (1): 358–93. ment Under Uncertainty. Princeton, NJ: Princeton Calderón, César A., Enrique Moral-Benito, and Luis University Press. Servén. 2011. “Is Infrastructure Capital Produc- Dobbs, Richard, Susan Lund, Charles Roxburgh, tive? A Dynamic Heterogeneous Approach.” Policy James Manyika, Alex Kim, Andreas Schreiner, Ric- Research Working Paper 5682, World Bank, Wash- cardo Boin, Rohit Chopra, Sebastian Jauch, Hyun ington, DC. Kim, Megan McDonald, and John Piotrowski. 58 The Emerging Pattern of Global Investment Global Development Horizons 2010. Farewell to Cheap Capital? The Implications of ———. 2009. “Misallocation and Manufacturing TFP Long-Term Shifts in Global Investment and Saving. in China and India.” Quarterly Journal of Economics Washington, DC: McKinsey Global Institute. 124 (4): 1403–48. Easterly, William R., and Ross Levine. 2001. “It’s Not Jordà, Òscar. 2009. “Simultaneous Confidence Factor Accumulation: Stylized Facts and Growth Regions for Impulse Responses.” Review of Econom- Models.” World Bank Economic Review 15 (2): ics and Statistics 91 (3): 629–47. 177–219. Jorgenson, Dale W. 1963. “Capital Theory and Invest- Ettinger, Stephen, Michael Schur, Stephan von Klaudy, ment Behavior.” American Economic Review 53 (2): Georgina Dellacha, and Shelly Hahn. 2005. 47–56. “Developing Country Investors and Operators in Kalemi-Ozcan, Sebnem, and Bent E. Sørensen. 2012. Infrastructure.” Trends and Policy Options Series “Misallocation, Property Rights, and Access to 3, Public-Private Infrastructure Advisory Facility, Finance: Evidence from Within and Across Africa.” Washington, DC. Working Paper 18030, National Bureau of Eco- FAO (Food and Agriculture Organization of the nomic Research, Cambridge, MA. United Nations). FAOSTAT database. FAO, Rome. Kapur, Devesh, and Ravi Ramamurti. 2001. “India’s http://faostat.fao.org. Emerging Competitive Advantage in Services.” Feldstein, Martin S., and Charles Y. Horioka. 1980. Academy of Management Executive 15 (2): 20–33. “Domestic Saving and International Capital Flows.” Keynes, John Maynard. 1936. The General Theory of Economic Journal 90 (358): 314–29. Employment, Interest and Money. New York: Har- Fernández-Villaverde, Jesús, Pablo Guerrón-Quintana, court, Brace & World. Juan F. Rubio-Ramírez, and Matin Uribe. 2011. Knowles, Stephen. 2001. “Are the Penn World Tables “Risk Matters: The Real Effects of Volatility Shocks.” Data on Government Consumption and Investment American Economic Review 101 (4): 2530–61. Being Misused?” Economics Letters 71 (2): 293–98. Feyrer, James D., and Jay C. Shambaugh. 2012. “Global Koopman, Robert B., Zhi Wang, and Shang-Jin Wei. Savings and Global Investment: The Transmission 2012. “Estimating Domestic Content in Exports of Identified Fiscal Shocks.” American Economic When Processing Trade Is Pervasive.” Journal of Journal: Economic Policy 4 (2): 95–144. Development Economics 99 (1): 178–189. Fisman, Raymond J. 2001. “Estimating the Value of Kraay, Aart C., Norman V. Loayza, Luis Servén, and Political Connections.” American Economic Review Jaume Ventura. 2006. “Country Portfolios.” Journal 91 (4): 1095–102. of the European Economic Association 3 (4): 914–45. GTAP (Global Trade Analysis Project) database. Kraay, Aart C., and Jaume Ventura. 2000. “Current Center for Global Trade Analysis, Department of Accounts in Debtor and Creditor Countries.” Agricultural Economics, Purdue University, West Quarterly Journal of Economics 115 (4): 1137–66. Lafayette, IN. https://www.gtap.agecon.purdue Lee, Jaewoo, and Kwanho Shin. 2001. “The Role of a .edu/#2. Variable Input in the Relationship Between Invest- Gupta, Sanjeev, Alvar Kangur, Chris Papageorgiou, and ment and Uncertainty.” American Economic Review Abdoul Wane. 2011. “Efficiency-Adjusted Public 90 (3): 667–80. Capital and Growth.” Working Paper WP/11/217, Levine, Ross. 2005. “Finance and Growth: Theory International Monetary Fund, Washington, DC. and Evidence.” In Handbook of Economic Growth, Hayashi, Fumio. 1982. “Tobin’s Marginal q and Aver- vol. 1, ed. Philippe Aghion and Steven N. Durlauf, age q: A Neoclassical Interpretation.” Econometrica 866–934. Amsterdam: Elsevier. 50 (1): 213–24. Li, Wei, and Lixin Colin Xu. 2004. “The Impact of Hesse, Heiko, Andreas A. Jobst, and Juan Solé. 2008. Privatization and Competition in the Telecommu- “Trends and Challenges in Islamic Finance.” World nications Sector around the World.” Journal of Law Economics 9 (2): 175–93. and Economics 47 (2): 395–430. Holmström, Bengt R., and Jean Tirole. 1997. “Finan- Lorenzoni, Guido, and Karl Walentin. 2007. “Finan- cial Intermediation, Loanable Funds, and the Real cial Frictions, Investment and Tobin’s q.” Work- Sector.” Quarterly Journal of Economics 112 (3): ing Paper 13092, National Bureau of Economic 663–91. Research, Cambridge, MA. Hsieh, Chang-Tai, and Peter J. Klenow. 2007. “Relative Lucas, Robert E., Jr. 1990. “Why Doesn’t Capital Flow Prices and Relative Prosperity.” American Economic from Rich to Poor Countries?” American Economic Review 97 (3): 562–85. Review 80 (2): 92–6. Global Development Horizons The Emerging Pattern of Global Investment 59 Lucas, Robert E., Jr., and Edward C. Prescott. 1971. Tornell, Aaron, and Frank Westermann. 2002. “Boom- “Investment Under Uncertainty.” Econometrica 39 Bust Cycles in Middle Income Countries: Facts and (5): 659–81. Explanation.” International Monetary Fund Staff Mankiw, N. Gregory, David H. Romer, and David Papers 49 (Special Issue): 111–55. N. Weil. 1992. “A Contribution to the Empirics of UN (United Nations). 1951. “Measures for the Eco- Economic Growth.” Quarterly Journal of Economics nomic Development of Underdeveloped Coun- 107 (2): 407–37. tries.” Report by a Group of Experts appointed by Mauro, Paolo. 1995. “Corruption and Growth.” Quar- the Secretary-General of the United Nations, UN terly Journal of Economics 110 (3): 681–712. Department of Economic Affairs, New York. McDonald, Robert L., and Daniel Siegel. 1986. “The UNIDO (United Nations Industrial Development Orga- Value of Waiting to Invest.” Quarterly Journal of nization) Industrial Statistics database. UNIDO, Economics 101 (4): 707–27. Vienna, Austria. http://www.unido.org/statistics Meier, Gerald M., and Dudley Seers, eds. 1984. Pioneers .html. in Development. New York: Oxford University Press. World Bank. 2006. Where Is the Wealth of Nations? Mostrous, Yiannis G., Elliott H. Gue, and David F. Measuring Capital for the 21st Century. Washington, Dittman. 2010. The Rise of the State: Profitable DC: World Bank. Investing and Geopolitics in the 21st Century. Upper ———. 2008. The Political Economy of Policy Reform: Saddle River, NJ: FT Press. Issues and Implications for Policy Dialog and Devel- Olken, Benjamin A. 2007. “Monitoring Corruption: opment Operations. Washington, DC: World Bank. Evidence from a Field Experiment in Indonesia.” ———. 2011. The Changing Wealth of Nations: Measur- Journal of Political Economy 115 (2): 200–49. ing Sustainable Development in the New Millennium. PFI (Project Finance International). Newsletter data- Washington, DC: World Bank. base and website, Thomson Reuters, London. ———. 2012a. China 2030: Building a Modern, http://www.pfie.com. Harmonious, and Creative High-Income Society. Pritchett, Lant. 2000. “The Tyranny of Concepts: Washington, DC: World Bank and Development CUDIE (Cumulated, Depreciated, Investment Research Center of the State Council, the People’s Effort) Is Not Capital.” Journal of Economic Growth Republic of China. 5 (4): 361–84. ———. 2012b. Global Financial Development Report PIDA (Programme for Infrastructure Development 2013: Rethinking the Role of the State in Finance. in Africa). 2011. Interconnecting, Integrating, and Washington, DC: World Bank. Transforming a Continent. Tunis, Tunisia: African ———. 2012c. Global Investment Promotion Best Prac- Development Bank. tices 2012. Washington, DC: World Bank. Rada, Nicholas E., and Steven T. Buccola. 2012. “Agri- ———. 2012d. World Bank Group Infrastructure Strat- cultural Policy and Productivity: Evidence from egy Update FY2012–2015: Transformation Through Brazilian Censuses.” Agricultural Economics 43 (4): Infrastructure. Washington, DC: The World Bank. 355–67. World Bank EdStats database. Education statistics Rodrik, Dani. 2009. “Industrial Policy: Don’t Ask website, World Bank, Washington, DC. http:// Why, Ask How.” Middle East Development Journal www.world bank.org/education/edstats. 1 (1): 1–29. World Bank Global Economic Monitor database. World Romer, Paul M. 1990. “Endogenous Technological Bank, Washington, DC. http://data.worldbank.org/ Change.” Journal of Political Economy 98 (5/2): data-catalog/global-economic-monitor S71–S102. World Bank Private Participation in Infrastructure Servén, Luis. 2003. “Real-Exchange-Rate Uncertainty database. World Bank, Washington, DC. http:// and Private Investment in LDCs.” Review of Eco- ppi.worldbank.org/. nomics and Statistics 85 (1): 212–18. World Bank World Development Indicators database. Summers, Robert, and Alan W. Heston. 1991. “Penn World Bank, Washington, DC. http://data.world World Table (Mark 5): An Expanded Set of Interna- ba n k .org /d at a-c at a log /world-de velopment- tional Comparisons, 1950–1988.” Quarterly Jour- indicators. nal of Economics 106 (2): 327–68. Tobin, James. 1969. “A General Equilibrium Approach to Monetary Theory.” Journal of Money, Credit and Banking 1 (1): 15–29. 2 Global Saving in 2030 W HETHER SAVING IS A SERIOUS developing world will put downward pres- impediment to growth in developing sure on household saving rates. However, countries has long been an issue of aging will also make (young) labor scarcer concern. In the late 1960s, developing countries with respect to capital, raising real wages saved only 18 percent of their income, and low- relative to the return to capital. Th is shift income economies only 13 percent, compared in relative factor prices will entail a redistri- with 21 percent in high-income economies. bution of income from asset holders, who Given developing countries’ limited ability to tend to be older and thus have lower saving attract foreign saving, it was felt that inadequate rates, to workers who tend to be younger domestic saving constrained investment, and and thus have higher saving rates, contrib- thus growth, necessitating large transfers from uting to mitigate downward pressure. high-saving rich countries. Whether saving was The higher economic growth rates in the primary constraint on development can still developing countries will also encour- be debated, but the situation has changed dra- age saving. Indeed, a scenario of gradual matically. Developing countries’ domestic sav- convergence between the developed and ing averaged 32 percent of their income in 2009, developing worlds in terms of income and while the rate in developed countries had fallen structural change assumes that, for a large to 17 percent.1 This sharp increase in saving rates group of developing countries, growth rates among developing countries has not laid to rest will continue to increase up to around 2020, the concern that inadequate saving may limit when they will stabilize and then slowly their growth prospects, however, because aging decelerate. For some countries, notably in populations may put downward pressure on their Sub-Saharan Africa, growth rates are pro- saving rates in the coming decades. jected to follow an increasing trend through A conclusion of this chapter is that despite 2030. These accelerations of growth are even the tendency of population aging to lower sav- stronger in a scenario where the productivity ing rates, saving is unlikely to be a binding con- gap between high-income countries and the straint on global growth. Although this conclu- developing world closes more rapidly. These sion applies at the global level, there will be clear positive changes in the pace of growth boost disparities in saving rates across countries and saving and partially offset the drag from the regions and some of them may face difficulties in increasing share of old people. the future. Other growth-related mechanisms will The main messages of the chapter are as also play a role. Longer life expectancy follows: may have the effect of increasing saving to finance additional years of education • Changes in saving rates in developing coun- and retirement. In turn, the increased tries over time will be driven by three main number of better-educated workers with factors: aging populations, economic growth steeper age-earnings profi les will also tend rates, and the deepening of financial markets. to increase saving rates. Moreover, higher The aging of populations in much of the incomes are associated with fewer children Global Development Horizons 61 62 Global Saving in 2030 Global Development Horizons and greater female labor force participa- • The rise of developing countries will support tion, increasing the potential for saving global saving rates. Developing countries during workers’ most productive years. will experience rapid growth in output, Financial deepening in developing coun- rapid population growth in regions with tries is found to put downward pressure on relatively young populations, and relatively saving rates. Specifically, financial develop- high saving rates between 2010 and 2030, ment can relax liquidity constraints and while high-income economies will expe- thus reduce the need for saving by indi- rience slower growth and declining sav- viduals and firms with high-return invest- ing rates. By 2030, developing countries ments. Financial deepening reflected in the will contribute 62 of every 100 dollars of greater availability and use of insurance and total world saving, up from 45 in 2010. pension arrangements can also reduce the The shift in economic weight to relatively need for precautionary saving. But financial high-saving developing economies will deepening is also associated with a more counterbalance the contraction of saving in efficient allocation of capital, and thus with high-income countries. Developing coun- higher growth in incomes, which would tries’ rapidly growing share of global sav- tend to support saving. This complex inter- ing implies also that their share of global action among aging, growth, and financial wealth will also rise, albeit more slowly. deepening can be expected to result in a • Saving in developing countries appears to be slight decline in developing countries’ aver- concentrated among high-income households. age saving rate, from a peak of 34 percent in Although the concentration of saving (and 2014 to 32 percent in 2030. of wealth) within a small segment of a • The evolution of saving will differ signifi- country’s population has the possibility to cantly by region. In two regions—East support growth by encouraging financial Asia as well as Eastern Europe and Central deepening, concentration also has somber Asia—downward pressure on saving rates implications for economic mobility and thus due to rapid aging will be at least partially for the political and social consensus essen- offset by rapid growth, so that saving rates tial for growth. Not only do high-income will not fall by as much as in high-income households tend to save a greater proportion countries. In Indonesia, for example, the of their incomes than low-income house- old-age dependency ratio will increase at holds, but they also account for the bulk as rapid a pace as in high-income coun- of saving in countries at various stages of tries between 2010 and 2030, but its sav- development and demographic transition. ing rate will fall only from 32 percent to In countries with high economic mobility, 31 percent (under the gradual scenario), in the relationship between low saving and contrast to the expected decline in high- low income could reflect efforts to smooth income countries’ saving rate from 18 per- consumption by households experiencing cent to 16 percent. Some of this difference temporary income losses. Unfortunately, a reflects changes in structural variables such similar correlation is observed across house- as financial sector depth and the size of holds grouped by educational attainment— social security systems, but mostly it is due a proxy for permanent income and thus a to Indonesia’s higher growth rate. At the more stable condition than the position in other end of the demographic spectrum, the income distribution at a point in time. Sub-Saharan Africa’s relatively young and Consistently, the least educated groups in a rapidly growing population, coupled with country have low or no saving, suggesting its increasing growth due to productiv- an inability to improve their earning capac- ity catch-up, will make it the only region ity and, for the poorest, to escape a poverty where the saving rate does not fall during trap. Policy makers in developing countries the time horizon considered here. have a central role to play in boosting private Global Development Horizons Global Saving in 2030 63 saving through policies that raise human Saving at the aggregate level: capital, especially for the poor. Past, present, and future • Changes in household structure will increase the importance of financial markets in pro- Saving has increased in viding for income support during old age. As developing countries incomes rise, household size tends to fall A large portion of global saving now originates in because workers are more geographically the developing world. This is quite a recent obser- mobile and older individuals are more able vation, however. For the two decades from 1980 to live independently on their accumulated to 2000, developing countries’ saving amounted savings. Alongside this reduction in house- to a fairly stable ratio of around 4 percent of hold size tends to come a profound trans- world income (fi gure 2.1). Th at ratio increased formation of the old-age support struc- sharply since 2000, however, to more than 9 per- ture—from an informal, multigenerational cent by 2009, while total saving by high-income household system to more formal private countries fell from 18 percent to 12.5 percent pension or public social protection systems. of global income over the same years. The rapid Reliance on privately financed pensions rise in developing countries’ saving has sustained rather than household saving and wealth global saving rates in the face of declines in high- during old age has the potential to improve income countries. A notable follow-on effect of welfare if the elderly and their children all the increase in saving in developing countries is prefer to live separately. Another benefit that they account for a growing share of global is that using private financial markets to wealth (discussed in box 2.1). intermediate pension savings can increase The growth in developing countries’ contribu- financial depth and contribute to devel- tion to the global pool of saving can be decom- opment. However, shifting from depen- posed into two parts: (a) the size of their econo- dence on the income of family members to mies relative to the global economy and (b) their dependence on fi nancial institutions also saving rates. Looking at changes in developing underlines the importance of strong regula- tion to limit fraud and excessive risk taking by financial intermediaries. FIGURE 2.1 Developing countries have accounted for a • Demographic change will challenge the sus- growing share of global saving since around 2000 tainability of public finances. In developing 30 Saving as a share of global income, % countries experiencing a large increase in the share of their populations past working 25 age, budgets will be strained by mounting World pension and health-care costs. The shift 20 in the composition of public expenditure High income toward old-age-related items, though, will 15 be offset only marginally by the reduc- tion in education spending expected in the 10 face of aging populations. Furthermore, Developing 5 the scope for decreases in non-age-related expenditures is limited. Complex policy 0 challenges will arise from efforts to reduce 1980 1985 1990 1995 2000 2005 2009 the burden of health care and pensions while limiting the decline in benefits and Source: World Bank calculations using data in the World Bank World Development Indicators database. services. Working through these challenges Note: Shaded area corresponds to the period from 2000 onward, where a regime shift in will be necessary to ensure the sustainabil- developing countries’ total saving appears to have occurred (a simple linear regression of developing countries’ saving rates on time in the 1965–99 period is S = 0.000t – 1.464, while ity of public finances without imposing that for the 2000–09 period is S = 0.002t – 47.369, where S and t correspond to developing severe hardships on the old. countries’ total saving and the year, respectively). 64 Global Saving in 2030 Global Development Horizons FIGURE 2.2 Both relative size and saving rates have have increased total saving in developing coun- contributed significantly to the increased global tries by 1.5 percent of global income, while importance of developing countries’ saving another increase of 1.9 percent was directly due 3.5 to changes in economic size. Saving in developing countries has risen 3.0 Change in saving as a share sharply from the levels of some 50 years ago. 2.5 Developing countries as a group saved 18 percent of global income, % 2.0 of their income in the 1960s, and low-income 1.5 countries only 13 percent, compared with 21 per- 1.0 cent in high-income countries.2 These low saving rates were viewed as a major constraint on invest- 0.5 ment and thus on future growth. Other factors 0.0 such as uncertain policy environments, poor –5.0 infrastructure, and undeveloped financial sys- –1.0 tems also hindered investment and at the same 1980–84 1985–89 1990–94 1995–99 2000–04 2005–09 time reduced incentives to save. Contribution of relative size Contribution of saving rate But much has changed since the 1960s. The past three decades in particular have seen impres- Source: World Bank calculations using data in the World Bank World Development Indicators database. sive policy improvements and large increases in Note: The decomposition of the investment share takes total derivatives from the formula growth, saving, and investment. Saving had risen (s/Y ) = (s/y ) × (y/Y ), where s , y, and Y correspond to developing countries’ saving, developing countries’ income, and global income, respectively. to almost a third of developing countries’ income by 2009, despite some decline during the global financial crisis (figure 2.3). Conversely, aging FIGURE 2.3 Developing countries’ saving rates have populations have put significant downward pres- increased sure on advanced economies’ saving rates. The 36 ratio of developing-country saving rates to those 34 in high-income countries rose steadily from 0.9 in 1980 to 2.0 in 2009. Saving as a share of GNI, % 32 As in the case of investment, much of the 30 Without increase in saving by developing countries is 28 China attributable to China, though other large emerg- 26 Developing ing economies, such as India and Indonesia, have 24 made significant contributions as well (figure 22 Without 2.4). The path of saving rates has differed sub- BRIICs 20 stantially across major developing economies: 18 India’s saving rate has increased steadily; Brazil 16 showed a slowly increasing trend in saving until 1980 1985 1990 1995 2000 2005 2009 the peak in 1989 and a distinct downward trend thereafter; and the Russian Federation’s saving Source: World Bank calculations using data in the World Bank World Development Indicators rate has been erratic since the dissolution of the database. Notes: Shaded area corresponds to the period from 2000 onward, where a regime shift in Soviet Union. developing countries’ total saving appears to have occurred (a simple linear regression on time in the 1965–99 period is S = 0.000t – 1.464, while that for the 2000–09 period is S = 0.002t – 47.369, where S and t correspond to developing countries’ total saving and the year, respec- Saving trends in China, in particular, have tively) . BRIICs = Brazil, the Russian Federation, India, Indonesia, and China. received considerable attention in both policy circles and academic literature. China is nota- countries’ saving for five-year periods since 1980 ble not only because of the large size of its saving shows that both relative size and saving rates pool and its high saving rate but also because the have, on average, contributed (fi gure 2.2). For composition of saving is broad: China’s house- the years 2000 to 2009, changes in saving rates hold, corporate, and government saving have all Global Development Horizons Global Saving in 2030 65 seen large increases. The literature points to sev- FIGURE 2.4 The paths of saving rates have differed eral factors in the increasing saving rate, includ- substantially across major developing economies ing the following: 60 • Chamon and Prasad (2010) identify the China 50 Saving as a share of GNI, % rising private burden of housing, educa- tion, and health care expenditures as well 40 Russian as increased precautionary saving as impor- Federation tant drivers of a seven percentage point 30 Indonesia increase in the urban household saving rate India between 1995 and 2005. 20 • Ma and Yi (2010) focus on the rapid Brazil increase in the working-age share of the 10 population, the structural shift from agri- culture and corresponding urbanization, 0 tough corporate restructuring, a shrinking 1980 1985 1990 1995 2000 2005 2010 social safety net, and transition to a par- Source: World Bank calculations using data in the World Bank World Development Indicators tially funded pension system. database. • Wei and Zhang (2011) emphasize competi- tive saving by parents with sons to make them more attractive for marriage given a has been concentrated in high-income coun- skewed gender ratio in the population. tries, but middle-income countries have seen sig- • Curtis, Lugauer, and Mark (2011) show nificant deepening since the early 1990s as well that China’s changing age distribution (Beck, Demirgüç-Kunt, and Levine 2010).3 appears to be an important factor when a bequest motive for saving is taken into Regional saving trends among developing account. countries indicate that the recent rise in saving has been mainly due to increases among coun- Another key factor is the relatively unbal- tries in East Asia and the Pacifi c, the Middle anced pattern of fi nancial market development East and North Africa, and South Asia. Th is in China; households and small and medium-size rise in saving has been concomitant with signifi- enterprises often face tight credit constraints, so cant external surpluses (owing to robust exports they must save to self-finance, while state-owned of manufactured goods and energy), although, enterprises typically have easy access to credit. notably, in South Asia the rise has occurred even But perhaps the most important factor behind in the presence of small deficits. China’s high level of saving in recent decades has Within regions, too, there is significant diver- been its high rate of economic growth, which sity in the level and trend in domestic saving. tends to push up saving rates. Typically, saving rates tend to be lower in low- Although the forces at work in driving up income countries (fi gure 2.5). In East Asia, for saving rates in China apply to many developing example, the region’s relatively high overall saving countries, nowhere is the impact of the forces as rate—averaging 24 percent between 1980 and acute as it is in China; in fact, in much of the 2010—masks much lower rates of saving among developing world, one or more of the character- low-income nations, although saving rates among istics clearly observable in China is absent. For countries in income groups in East Asia have example, Sub-Saharan Africa is at an earlier point risen over time (figure 2.5, panel a). In times of in the demographic transition than other devel- crisis, such as during the Asian financial crisis of oping regions, and the transition is proceeding 1997–98 and the global crisis of 2007–09, saving more slowly than it did in other regions. Also, rates actually dropped because incomes fell and financial systems have deepened worldwide; this debt was incurred to smooth consumption. 66 Global Saving in 2030 Global Development Horizons FIGURE 2.5 Saving rates tend to be lower in lower-income countries a. East Asia and Pacific b. Eastern Europe and Central Asia 50 30 Upper- 25 Saving as a share of GNI, % Saving as a share of GNI, % Upper-middle income middle 40 20 income 15 30 Lower-middle 10 Lower- income 5 middle 20 income 0 10 Low income –5 Low income –10 0 –15 1980 1985 1990 1995 2000 2005 2010 1990 1995 2000 2005 2010 Source: World Bank calculations using data in the World Bank World Development Indicators database. Notes: Saving rates were the unweighted shares for each income group. GDP-weighted results are qualitatively similar, with the exception of shares for the lower-middle- income category for East Asia, due to the large weight of China. The private and public components Private saving in several major developing of saving have varied significantly countries (notably China and India) is atypically across countries and across time high and has actually grown relatively faster than household saving in recent years (figure 2.7). When national saving is broken down into its pri- Theory suggests that corporate saving is negatively vate and public components, it is clear that both related to cash flow (Riddick and Whited 2009), are important. which in turn suggests that external financing constraints—which could result from financial Private saving is typically the greater of the development not keeping pace with growth—may two, and historically private saving has tended be responsible for the recent rise in corporate sav- to fluctuate more than public saving. For many ing in these economies. As the financial sectors in countries, household saving alone accounts for China and India mature, however, this tendency most of national saving, although once country for high corporate (and hence private) saving will data are aggregated to a regional level the extent likely diminish. of household saving varies substantially (fi gure 2.6). Households’ share of total saving has ranged Public saving does not show such an obvious from around 25 percent in Latin America to an long-term trend. Although it has fluctuated to average of 68 percent in India since the 1950s. some extent, the public portion of national sav- One reason for relatively low saving by Latin ing has remained roughly in the range of 10–20 American households is probably that corporate percent of national saving in developing countries saving is high there, and because fi rms are ulti- over the 30-year period 1981–2011 (figure 2.8). mately owned mainly by households, corporate The notable exception was during the years lead- saving substitutes for household saving. Because ing up to the global recession in the late 2000s, household saving is sensitive to both economic when public saving exceeded 25 percent of growth and demographic change, differences in national saving from 2006 through 2008 before the timing of the demographic transition and in quickly returning to historical levels in 2009. growth over the coming decades have the poten- The trend in public saving for the entire world tial to affect global patterns of saving through has followed a path roughly the same as that of their impact on household saving decisions. developing countries over the past three decades, Global Development Horizons Global Saving in 2030 67 FIGURE 2.6 Household saving varies significantly by region a. East Asia and Pacific b. Latin America and the Caribbean c. Middle East and North Africa 100 100 100 Household saving, Household saving, Household saving, % of total saving % of total saving % of total saving 50 50 50 0 0 0 1980 1990 2000 2010 1980 1990 2000 2010 1980 1990 2000 2010 d. South Asia e. Sub−Saharan Africa 100 100 Household saving, Household saving, % of total saving % of total saving 50 50 0 0 1980 1990 2000 2010 1980 1990 2000 2010 Source: World Bank calculations using data in the UN National Accounts Main Aggregate database. Note: Shares are the unweighted average of household saving as a share of national saving across countries in the region, reported only for years in which data are available for enough countries to account for at least 40 percent of the region’s gross domestic product. FIGURE 2.7 Corporate saving in China and India has grown faster than household saving in recent years a. Chinaa, c b. Indiab, c 60 40 35 50 30 Saving rate, % Saving rate, % 40 25 20 30 15 20 10 5 10 0 0 –5 19 2 19 3 19 4 95 19 6 97 19 8 99 20 0 01 20 2 03 20 4 05 20 6 07 20 8 09 19 2 19 3 19 4 19 5 96 19 7 98 20 9 00 20 1 02 20 3 04 20 5 06 20 7 08 9 9 9 9 9 0 0 0 0 0 9 9 9 9 9 9 0 0 0 0 19 19 19 20 20 20 20 20 19 19 19 20 20 20 20 Government Enterprises (corporations) Households Sources: World Bank calculations using data in the ACMR All China Data Center database (China) and CSO National Accounts database (India). a. For China, 2009 household and enterprise saving are imputed from their respective 2008 shares of 2009 total private saving. b. For India, 2007 data are provisional and 2008 are estimates, and household saving is defined as the sum of household financial saving and household physical saving. c. The use of the terms “enterprises” and “corporations” are retained from the original data sources (“enterprises” for China, and “corporations” for India). 68 Global Saving in 2030 Global Development Horizons FIGURE 2.8 The public portion of national saving has international investment positions (for more fluctuated within a fixed range details on the recent evolution of the global dis- tribution of wealth, see box 2.1). 30 Public saving as a share of 25 Growth, demography, and financial development will drive national saving, % 20 changes in the global pattern of 15 savings World A key objective of this report is to provide a pre- 10 Developing view of world saving for the next two decades. A countries 5 definitive theory of saving unambiguously iden- tifying its determinants is not yet available (for 0 a recent survey, see Attanasio and Weber 2010). However, there is broad consensus that economic 81 85 89 93 97 01 05 09 11 19 19 19 19 19 20 20 20 20 growth, demographic transition, and certain Source: World Bank calculations using data in the IMF World Economic Outlook database. structural variables such as financial sector devel- Note: Gross national saving is defined according to the national income accounting standard: opment affect saving. Theoretical models as well the difference between gross national income and public and private consumption, plus net current transfers. Gross public saving is current revenues minus current expenditures, for all as reduced-form empirical equations show the rel- levels of government. Current expenditures consist of government consumption expenditures, evance of these variables. current transfer payments, interest payments, and subsidies; public investment is not included in current expenditures, so it essentially counts toward public saving. For years 1992–2011, the For example, the importance of income graph represents the unweighted average of the ratio of gross public saving to gross national growth and changes in the age structure due saving for all countries for which data exist for all years (omitting Kuwait, which experienced large swings in its saving around the time of the First Gulf War). This sample includes 40 to population growth are central in explain- advanced countries and 102 developing countries. For years prior to 1992, the sample includes ing saving behavior in the life cycle (Modigliani all countries for which data were available in a given year (again omitting Kuwait). and Brumberg 1954) and permanent income (Friedman 1957) models. These models remain although public saving in developing countries common benchmarks against which saving typically accounted for a greater share of total behavior is evaluated. More recent models con- saving than in high-income countries until the sider how the stock of wealth can serve as a pre- late 1990s, and public saving fell more sharply for cautionary buffer against unexpected shocks developing countries than for high-income coun- (Carroll 1997) and how uncertainty about future tries during the early-2000s global recession. income may influence saving at the household It remains an open question whether the global (Weil 1993) and firm (Riddick and Whited financial crisis interrupted what would have been 2009) levels, especially absent complete markets a sustained upward shift in governments’ role for risk insurance (Aiyagari 1994).4 Increased in saving and capital formation (perhaps due to longevity can potentially raise saving rates at greater infrastructure investment in developing all ages (Bloom, Canning, and Graham 2003; countries) or whether the ramp-up was a short- Kinugasa and Mason 2007). Thus, if increased run phenomenon. Regardless of which descrip- growth and improvements in longevity happen tion of the precrisis years is more accurate, it is together, this effect would reinforce the positive clear that any upward trend in public saving in impact of growth on saving. the coming decades is likely to be offset by signif- Structural changes in developing countries can icant pressure from aging populations on public alleviate uncertainty and informational asymme- health care and pension expenditures. tries and contribute to moving toward more com- Although stocks of wealth evolve much more plete markets. For example, financial development slowly than output, saving, investment, and capi- can relax liquidity constraints and hence affect tal flows, developing countries’ rapidly growing saving activity (Deaton 1991; Jappelli and Pagano shares of these f lows will gradually accumu- 1994). At the same time, presence (or absence) of late into a growing share of global capital and social insurance mechanisms can also condition Global Development Horizons Global Saving in 2030 69 how much households in a country choose to save saving equation whose determinants are per capita (Hubbard, Skinner, and Zeldes 1995). income growth, demography (aged dependency), The scenarios for the future of world saving are financial development, social protection, and past constructed using a global dynamic computable saving rates. Following standard practice, this general equilibrium (CGE) model that includes a equation is calibrated using observed data and BOX 2.1 A gradual shift in the global distribution of wealth is under way The rise in savings rates in developing countries the peak on the right, consisting of moderately wealthy relative to high-income countries has resulted in an high-income countries, fell as rapid growth in per capita increase in developing countries’ share of global wealth (in economies such as Hong Kong SAR, China; wealth (the accumulation of national saving over time). Luxembourg; Norway; and Singapore) meant that they The growing importance of developing countries is moved from the right peak to the right tail of the dis- most evident in terms of external wealth holdings: tribution. There was thus an increasing gap between the emerging world saw rapid reductions in negative the wealthiest high-income countries and those around external positions in the 1990s and early 2000s (Lane the right peak—for example, Gulf States and southern and Milesi-Ferretti 2007), which have since become European countries, respectively. positive and continue to grow. a Despite the fact that Most of the world’s stock of wealth remains distributions change at a glacial pace, there was a entrenched in high-income countries. Recent analy- noticeable shift in the global distribution of national sis of household wealth by Davies et al. (2011), for wealth in the 15 years between 1992 and 2007, in example, shows that North America, Europe, and high- both absolute and per capita terms. income Asia-Pacifi c countries account for 90 percent Figure B2.1.1 shows, in panel a, the global intercoun- of global wealth. Among developing countries, wealth try distributionb of wealth per capita in 1992 and 2007, tends to be held in nonfinancial assets, such as hous- along with, in panel b, the difference between the two ing and land (Shorrocks, Davies, and Lluberas 2010). when the 1992 mean is preserved to show the change The continued concentration of wealth in high-income in the shape of the distribution independent of the shift countries, however, has been accompanied by marginal in the global average. That the distribution shifted to progress toward more equal distribution of national the right means that the world became wealthier in wealth per capita among countries of different income real terms. The bimodal pattern of the distribution in levels over the past 15 years. When countries are panel a is indicative of the presence of two clubs of weighted by their populations,c the Gini coefficient fell countries: the change in the peak on the left, which from 0.77 in 1992 to 0.74 in 2007. This moderate con- became slightly higher in 2007, refl ects the fact that vergence in national per capita wealth can be expected a number of previously poor countries (such as China, to continue, albeit gradually, in the future as developing the Dominican Republic, and Namibia) amassed assets countries account for a growing share of global output, and became middle-income countries. the change in saving, and investment. a. Net international investment positions are available from the International Monetary Fund’s International Financial Statistics (http://www.imf.org/ external/data.htm). It should be noted that China’s position constitutes a large part of the developing-country aggregate. b. Milanovic (2005) discusses three broad concepts for measuring income inequality across countries. In principle, these can also be applied to wealth inequality. Intercountry inequality is based on giving each country’s per capita wealth equal weight in the world distribution. International inequality takes into account the relative sizes of countries, weighting each by its population. Global inequality takes within-country inequality into account, measuring the distribution across individuals regardless of their nationality (see, for example, Bourguignon and Morrisson 2002). The global inequality approach is plagued by data limitations. The international inequality approach gives a good picture of what is happening at the global level, but the results can be dominated by one or two large countries (in the case of developing countries, China and India appear as two spikes that dominate the distribution). c. The population-weighted Gini measure of international inequality was calculated by assigning each country’s average per capita wealth to each of its residents and taking the Gini coefficient of the resulting global distribution. Over the same period, the unweighted (intercountry) Gini rose from 0.66 to 0.68, indicating that the result on international inequality depends in large part on changes in populous developing countries such as China and India. (continued) 70 Global Saving in 2030 Global Development Horizons BOX 2.1 (continued) FIGURE B2.1.1 The global distribution of national wealth per capita has changed over time a. Intercountry distribution, 1992 (solid) b. Change in distribution mean preserving, and 2007 (dashed) 2007 versus 1992 0.20 0.02 Differences in densities 0.15 0.01 Density 0.10 0 0.05 –0.01 0 –0.02 10 30 100 1,000 10,000 100,000 10 30 100 1,000 10,000 100,000 Log, national wealth per capita, $ Log, national wealth per capita, $ Sources: World Bank calculations using data in the World Bank World Development Indicators and Global Economic Monitor databases and Lane and Milesi-Ferretti 2007. Note: Panel a shows the probability distribution of the logarithm of total national wealth, computed for the 100 countries for which data are available for both years, as Gaussian kernel density estimates. National wealth is defined as the sum of the capital stock and net foreign assets, defl ated to 2000 dollars. Capital stocks are calculated using a perpetual inventory method with an assumed constant depreciation rate of 5 percent, and are defl ated with the country’s respective investment defl ator. Net foreign assets, by Lane and Milesi-Ferretti (2007), are calculated to account for changes in stock, valuation, and exchange rate, and are defl ated with the respective country’s GDP defl ator. The mean-preserving difference in the per capita wealth dis- tributions for the two years (1992 and 2007), shown in panel b, is computed by subtracting the increase in the mean of 2007 relative to 1992 from the observed countries’ values of 2007 prior to calculating differences. elasticities obtained from econometric estima- income growth is found to be positive. For each tions. Measuring global saving, accounting for the 10 percent increase in the rate of growth of per general equilibrium effects of growth patterns and capita income, the national saving rate rises 0.5– policy changes, and predicting the future are all 1.0 percent. Th is fairly strong relationship goes difficult things to do in economics, and they can some way toward explaining why—over the past all be easily criticized. However, the main advan- decade during which developing countries have tage of a model-based analysis is not in providing enjoyed robust per capita income growth—sav- precise forecasts but in having a framework that ing rates in many developing countries have also is consistent with economic theory and that can moved upward. be used to test and explain the other-things-equal This positive link between saving and growth effects of changes in key determinants. is moderated by two factors: changes in the The elasticities used in the CGE model are aged dependency ratio and changes to the level obtained from a formal multivariate econometric of fi nancial development, both of which have a analysis, and graphical representations of its main negative relationship with the saving rate. As results are offered in figures 2.9 and 2.10.5 the financial sector develops, there is generally a As depicted in figure 2.9 (panel a), the condi- downward pressure on saving: with credit more tional6 relationship of saving rate and per capita readily available in more sophisticated fi nancial Global Development Horizons Global Saving in 2030 71 FIGURE 2.9 Saving rates tend to increase with growth (panel a) and decrease with financial development (panel b) a. Income growth and saving b. Financial development and saving 0.05 0.05 Residuals of saving rate Residuals of saving rate 0 0 –0.05 –0.05 –0.2 –0.1 0 0.1 0.2 –0.4 –0.2 0 0.2 0.4 Residuals of per capita income growth Residuals of financial development Sources: World Bank calculations using data in the World Bank World Development Indicators database and Financial Structure and Development database. Note: The figures depict added variable plots of the residuals of the saving rate against per capita income growth (panel a) and saving rate against financial development (panel b), where saving rate is defined as domestic saving share of gross domestic product (GDP), measured in local currency; financial development is defined as domes- tic credit provision to the private sector as a share of GDP; and per capita income growth is calculated from real per capita income, measured in 2000 dollars. The relation- ship between the x-axis and y-axis variables is conditional in the sense that the relationship takes into account the effects of other determinants whose contributions are held constant. The graphs above are constructed from the results of specification S4 in table 1A.4 of online annex 1.6. All variables are in logarithm form, are unweighted, and are significant in multivariate regressions at standard levels. Plots for the developing country subsample show very similar patterns. markets, households are more likely to draw on FIGURE 2.10 Saving rates tend to decrease with consumer credit to address their financing needs. population aging Similarly, firms face a weaker incentive to retain earnings when it is easier to access capital markets to meet their working capital needs. 0.05 The negative relationship between aged depen- dency and saving is illustrated in figure 2.10, and Residuals of saving rate it confirms a standard result of the life-cycle theory. 0 Forthcoming demographic shifts will have a strong impact on global labor supply and saving Momentous demographic change, already under way, will be a key factor in shaping the future of global saving. Th ree major demographic forces –0.05 must be taken into consideration: (a) an increase in the size of the population of developing coun- –0.04 –0.02 0 0.02 0.04 0.06 tries relative to the world population, (b) change Residuals of aged dependency ratio in the age structure of the world population, and Sources: World Bank calculations using data in the World Bank World Development Indicators (c) the asynchronicity of changes in age structure database and Financial Structure and Development database. 72 Global Saving in 2030 Global Development Horizons across regions. Together with high economic household, which may be quite different, on aver- growth rates, these forces will mean that global age, in developing versus advanced economies. In saving will be increasingly determined by occur- a typical household, however, the life cycle leads rences in developing countries. to a hump-shaped saving profile over time: saving The world population will grow considerably rises from low or negative levels in the first phase, in future decades, from around 7 billion in 2010 reach a maximum level in the second phase, and to 9 billion in 2050 (United Nations Population begin to decline in the third phase. This third Division 2011).7 Although the rate of popula- phase thus implies that as the share of elderly tion growth over this period will be far less people in a country’s population rises, national than during the previous four decades, the addi- saving may fall. tion of more than 2 billion people to the world Indeed, considerable change is under way in through 2050 (an addition roughly equivalent the structure of the world’s population. By 2021, to the current population of China, Indonesia, less than a decade into the future, growth in the and the United States combined) is not incon- world’s working-age population will be exclu- sequential. Importantly, this increase in popu- sively determined by developing-country natives lation will be almost exclusively due to growth (that is, the blue line in figure 2.11 will meet the within developing countries. Already, developing horizontal line at 100 percent). By 2050, this sce- countries represent the main source of growth of nario will be quite different: each 100 working- the world population, and their contribution has age individuals added to the global population steadily increased. In 1950, 80 out of 100 new will be the result of 110 additional working-age people were natives of a developing country; the people among developing country natives and share increased to almost 95 percent by 2010; 10 fewer working-age people in high-income and from 2050 onward, developing countries countries. Assuming that working-age people will likely be the only ones contributing to the have higher saving rates than elderly people, this growth of the world population. As developing growing representation of natives of developing countries represent a growing proportion of the countries in the world’s working-age population world economy, their saving patterns will play a implies that developing countries will account for larger role in shaping the overall picture of global a growing share of global saving. saving. Large disparities in birth, death, and migra- The evolving age structure of both developing- tion rates between developing and high-income and advanced-country populations will also have countries help explain both the developing profound implications for saving rates. Typically, world’s growing population and the change in the households go through three broad phases during world’s age structure. For any given country, the the life cycle: (a) a first phase, when a household gap between death and birth rates (assuming no is raising young children and when its income net migration) is the rate of population growth, is growing, but so are its expenditures and its while the sequencing in the reduction of these investment in the education of the offspring; (b) two rates—the demographic transition—deter- a second phase, during which the household head mines a period of rapid population growth. and possibly the spouse reach their maximum A recent phase8 of the demographic transition earning potential and purchasing power; and (c) is depicted in figure 2.12 for the six developing a final phase, when the main breadwinners cease regions. Several aspects of population dynam- participating in economically productive activi- ics in recent decades are emphasized by this fig- ties and rely on accumulated assets and pensions ure: First, for all the regions, birth rates are still or other transfers to fund consumption. During higher than death rates. Second, death rates are these three phases, income and consumption— still falling, but they seem to be leveling off at key determinants of the welfare of the household around 10 deaths per 1,000. Finally, the differ- members—are strongly influenced by life-cycle ence between death and birth rates and the speed factors such as the age and composition of the of its closing are different across developing Global Development Horizons Global Saving in 2030 73 FIGURE 2.11 By 2020, growth in world’s working-age population will be exclusively determined by developing countries 115 Share of world’s working-age population growth from developing countries, % 110 105 100 95 90 85 80 75 70 50 55 60 65 70 75 80 85 90 95 00 05 10 15 20 25 30 35 40 45 50 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 20 20 20 20 Sources: World Bank projections, supplemented with calculations using data in the World Bank World Development Indicators database and United Nations Population Division 2011. Note: The green line shows the linear time trend in the data presented by the blue line. Where the share exceeds 100 percent, it refl ects loss of working-age people in high-income countries. For example, by 2050, each 100 additional working-age individuals in the global population will be the result of 110 additional working-age people in developing country natives and 10 fewer working-age people in high-income countries. regions’ trends (the determinants of which are asynchronous—with high-income countries explored in box 2.2). and some parts of the developing world much A somewhat complex set of mechanisms lies more advanced in their demographic transition. behind the timing of changes in the gap between The asynchronicity is shown by the considerable a country’s birth rate and death rate. When a diversity of the height of the peak of the ratio of country experiences an initial fall in child mor- working-age to nonworking-age population, the tality, this generates a cohort of young people that timing of this peak, and the slope of the curve is much larger than that of earlier generations. At around the peak. a certain point, fertility rates begin to slow down In high-income countries, the ratio of work- and the increase in the very young population ing-age to nonworking-age persons increased ends. But, as shown in box 2.2, figure B2.2.1, the slowly from 1950 to 2000, and now is declin- closing of the gap proceeds at a different pace in ing.9 As a consequence of the fairly slow demo- different regions. The larger the initial gap and graphic transition, this group has experienced the shorter the time period to close this gap, the mild age structure effects. By contrast, in the East bigger will be the “bulge” in the age structure. As Asia and Pacific region, the reduction in fertil- time goes by, the “bulge” cohort enters the labor ity has been rapid (see box 2.2) and the increase market, increasing the ratio of working-age to in the ratio of working-age population steep. nonworking-age population. The opposite effect South Asia, the Middle East and North Africa, occurs when the bulge cohort reaches retirement and Latin America and the Caribbean follow a age, with an increase in the ratio of nonworking- somewhat similar path: their fertility rates have age people to working-age people. been decreasing, but their demographic dividend A graphical representation of this “bulge” con- has not yet reached its maximum. Sub-Saharan cept can be seen for various regions in figure 2.13. Africa is once again an exception, showing a gla- One of the most interesting features highlighted cial pace in its demographic transition. Fertility by this figure is that aging is geographically rates have remained high (as shown in figure 74 Global Saving in 2030 Global Development Horizons FIGURE 2.12 Birth and death rates have shown distinct trends across regions a. East Asia and Pacific b. Middle East and North Africa 50 50 Number per 1,000 people Number per 1,000 people 40 40 Birth rate, crude 30 30 Birth rate, crude Death rate, crude 20 20 Death rate, crude 10 10 0 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 c. Eastern Europe and Central Asia d. South Asia 50 50 Number per 1,000 people Number per 1,000 people 40 40 Birth rate, crude 30 30 Death rate, crude 20 Birth rate, crude 20 Death rate, crude 10 10 0 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 e. Latin America and the Caribbean f. Sub-Saharan Africa 50 50 Birth rate, crude Number per 1,000 people Number per 1,000 people 40 40 30 Birth rate, crude 30 Death rate, crude 20 20 Death rate, crude 10 10 0 0 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 Source: World Bank calculations using data in the World Bank World Development Indicators database. Note: The crude death rate is the number of deaths among the population of a given geographical area during a given year, per 1,000 midyear total population of the given geographical area during the same year. Crude birth rate indicates the number of live births occurring during the year, per 1,000 population estimated at midyear. Subtracting the crude death rate from the crude birth rate provides the rate of natural increase, which is equal to the rate of population change in the absence of migration. Global Development Horizons Global Saving in 2030 75 BOX 2.2 The geographically asynchronous demographic transition is driven by cross-country differences in some key factors For any given country, declines in two major demo- Meanwhile, the rate in the Middle East and North graphic measures, mortality rates and fertility rates, Africa as of 2010 was roughly equivalent to the rate contribute to the aging of a population. Consider first in high-income countries in the 1960s. Under-5 mor- the mortality rate. Reductions in infant and child mortal- tality remains comparatively high in South Asia and ity due to the expansion of access to clean water and Sub-Saharan Africa, however, with rates 10 and 19 sanitation as well as to preventive health care (such as times those of high-income countries, respectively, as vaccination against infectious diseases) are the initial of 2010. For high-income countries, by comparison, drivers of population growth and explain the increase recent reductions in the overall mortality rate are attrib- in life expectancy experienced in the recent past. High- utable to the reductions in old-age mortality, which income countries have long had an advantage in terms result from more expensive health treatments and, of having the financial resources to address these con- more generally, higher incomes. cerns, as refl ected in their dramatically lower under-5 The aging of populations in recent decades also mortality rates in 1960 than in any developing region reflects sharp reductions in fertility rates. In the 1960s, (figure B2.2.1). In the five decades since, however, developing regions (excluding Eastern Europe and developing countries have made significant progress in Central Asia) had fertility rates of around 6.2 births per reducing child mortality, although from different start- woman, or more than twice the rate of high-income ing points and at different rates.a countries. That gap has now become much smaller As of 2010, under-5 mortality rates in the East (fi gure B2.2.2). As of 2010, East Asia and the Pacifi c Asia and Pacific, Eastern Europe and Central Asia, and Eastern Europe and Central Asia had fertility rates and Latin America and the Caribbean regions were equal to that of high-income countries and approxi- similar to those in high-income countries in the 1970s. mately equal to the so-called replacement rate of 2 FIGURE B2.2.1 Developing countries have made significant progress in reducing child mortality, although from different starting points and at different rates 250 200 Under-5 mortality rate 150 100 50 0 E. Asia & E. Eur. & L. Amer & Mid. East & S. Asia Sub-Saharan High-income Pacific Cent. Asia the Caribbean N. Africa Africa countries 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 Source: World Bank calculations using data in the World Bank World Development Indicators database. Note: The under-5 mortality rate is the probability per 1,000 that a newborn baby will die before reaching age fi ve if subject to current age-specifi c mor- tality rates. (continued) 76 Global Saving in 2030 Global Development Horizons BOX 2.2 (continued) births per woman. The fertility rate in Latin America fertility rate of 5 in 2010. Two main reasons have been and the Caribbean is now only slightly higher than advanced to explain this rapid reduction in fertility: fall- the replacement rate, and, at 2.7 and 2.8 births per ing infant mortality rates and increased female edu- woman, respectively, the Middle East and North Africa cation, the latter of which improves the labor market and South Asia are not terribly far above that level. opportunities of women and increases the opportunity Sub-Saharan Africa remains the exception, with a cost of childraising. a. Infant mortality rates, although lower than the under-5 mortality rates shown in figure B2.2.1, show similar regional dispersion and reduction patterns. FIGURE B2.2.2 The gap in fertility rates across regions has narrowed 7 6 5 Total fertility rate 4 3 2 1 0 E. Asia & E. Eur. & L. Amer & Mid. East & S. Asia Sub-Saharan High-income World Pacific Cent. Asia the Caribbean N. Africa Africa countries 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2009 Source: World Bank calculations using data in the World Bank World Development Indicators database. Note: The total fertility rate represents the number of children that would be born to a woman if she were to live to the end of her childbearing years and bear children in accordance with current age-specific fertility rates. B2.2.2), and as long as they continue to do so, versus wages); and pension, health care, and other the average age of the population and its share old-age support needs. These, in turn, will affect of working-age individuals will remain low com- international capital flows (see box 2.3). pared with other regions. The high incidence of acquired immune deficiency syndrome (AIDS) World saving patterns are set to among young adults has added to the reduction shift in the decades ahead in the ratio of workers to young and old. Given this heterogeneity in demographic This report’s view of the future evolution of global conditions, regions will differ in terms of their saving is based on scenarios built with a multi- saving dynamics; relative factor returns (capital country dynamic general equilibrium model. The Global Development Horizons Global Saving in 2030 77 FIGURE 2.13 The size and timing of the demographic “bulge” differs across regions 2.6 E. Asia & Pacific nonworking-age population 2.4 Ratio of working-age to 2.2 S. Asia 2.0 High-income countries 1.8 1.6 L. Amer. & E. Eur. & Cent. Asia the Caribbean 1.4 Mid. East 1.2 & N. Africa Sub-Saharan Africa 1.0 1950 1955 1960 1965 1970 1975 1980 1985 1990 1995 2000 2005 2010 2015 2020 2025 2030 2035 2040 2045 2050 Sources: World Bank projections, supplemented with calculations using data in the World Bank World Development Indicators database and United Nations Population Division 2011. model is used to assess the effects on saving of change in both private and public saving. Given several major forces: the forthcoming changes in its long-term view, the model makes the sim- demographics, the fact that developing countries plifying assumption that governments run bal- will continue to grow faster than high-income anced budgets: any increase in public expendi- countries, and shifts of other structural determi- tures is passed on to households via an increase nants such as the level and coverage of financial in direct taxation. The impact of demographics development and social protection. The approach on both private and public age-related spending is broadly similar to that taken in the second is ref lected in a change in private sector con- group of studies described in box 2.3. sumption and saving and, thus, national saving. Two scenarios are run with the model. The The mechanisms by which demographic change main difference between the two scenarios is affects the public component of national saving the pace of convergence between developing are not explicitly modeled in the CGE framework and high-income countries in terms of economic but are nevertheless captured. A later section of growth and fi nancial development. The gradual the chapter presents a more detailed, but partial convergence scenario assumes a certain degree of equilibrium, analysis of the forthcoming expan- economic growth convergence, accompanied by sion of age-related public expenditures. endogenous increases in financial deepening. The rapid convergence scenario imposes faster Under a gradual convergence growth convergence, together with the assump- scenario, global saving will almost tion that developing countries will be able to double from its 2010 level reduce the gap between their financial devel- opment and that of the United States in 2030 Developing countries will represent a very large by about 25 percent. All of the other exogenous share—almost two-thirds—of all global saving variables (notably, demographic shifts in the age by 2030 (figure 2.14, panel b), up from slightly composition and labor supplies) as well as other less than half in 2010. China and other Asian policy variables (such as tax rates, tariff s, and economies (both developing and high-income) government expenditures) are held at the same will continue to be major savers, with China levels in both scenarios. (The assumptions and accounting for a larger share of global savings parameters of the model are detailed in online than any other individual country. Together, annex 1.6). high-income European countries also will con- The scenarios’ results reported are for national tinue to contribute a significant share of global saving—that is, they represent the combined savings. 78 Global Saving in 2030 Global Development Horizons BOX 2.3 Literature on modeling the effect of aging on saving at the global levela A recent wave of studies uses multicountry or multi- These effects are intensified by a rapid demographic region general equilibrium models to account for the transition and a quickly closing gap between an initial impact of cross- country demographic changes on mortality reduction and subsequent fertility reduc- saving—some considering the historical experience, tion. These two aspects of the demographic transition and others the outlook for future decades. It is pos- strengthen the changes in aggregate saving and labor sible to identify at least two different strands of this supply, thus exacerbating the changes in relative factor literature. supplies and the divergence in relative factor prices. Auerbach and Kotlikoff (1987) provide an early For an open economy, geographically asynchronous example of an approach that uses large-scale over- aging across other countries will affect saving, and thus lapping generations models to analyze the implica- the relative factor prices, through international capital tions of policy reform. A large number of papers fol- flows.c A country’s current account deficit and its capital low this approach to study social security reform (for inflows are determined by the excess of its investment example, Conesa and Krueger 1999) and basic tax over its saving (and vice versa for a surplus and capital reform (for example, Altig et al. 2001; Conesa and outflows). Because the supply of saving is related, via Krueger 2006), among other issues. Within this fi rst the life cycle, to the age structure of its population, the group of studies, a subset of papers focuses on the current account balance is affected, all things equal, by economic consequences of population aging in closed demographic change. In a world with free capital move- economies. A specific question addresses the adjust- ment, capital should flow from fast-aging regions, where ments required in the social security system because capital is relatively abundant, to regions with fast-grow- of demographic shifts. Important examples include ing populations, where labor is abundant. Huang, I ˙mrohorog ˘ lu, and Sargent (1997); De Nardi, A second strand of studies extends the previous ˙ Imrohorog ˘ lu, and Sargent (1999); and, with respect to approaches by explicitly considering these open-econ- asset prices, Abel (2003). omy issues. By taking into account international links In a closed-economy setting, geographically asyn- (trade and capital flows) and using multicountry or mul- chronous aging does not matter, and the saving effects tiregion general equilibrium models, this third strand of the demographic transition—particularly in the latter of literature reassesses the impact of cross-country phases, when the society is aging—are determined by demographic changes on various economic issues. the speed of change in the age structure and size of the Several studies use demographic projections to quan- bulge cohort. More specifically, in an economy closed tify, for example, the effects of pension reforms on the to international capital fl ows, changes in labor supply viability of social security systems (Aglietta et al. 2007; and aggregate saving affect the relative returns of labor Attanasio, Kitao, and Violante 2006, 2007; Börsch- and capital and the growth of the economy. In an aging Supan, Ludwig, and Winter 2006; Fehr, Jokisch, and economy, young labor becomes scarcer with respect Kotlikoff 2003). Krueger and Ludwig (2007) study the to capital, thus the return to labor rises relative to the impact of demographic forecasts on the distribution return to capital. of wealth and welfare in Organisation for Economic This change in factors’ relative prices also has inter- Co-operation and Development countries. Other stud- generational distribution effects. The people who lose ies adopt an ex post approach. For example, using from lower capital returns will tend to be asset hold- calibrated life-cycle models, Domeij and Flodén (2006), ers who are older than the average person, while those Feroli (2003), and Henriksen (2002) find that changes in who gain from higher wages will tend to be younger demographics explain a large part of historical current than the average.b accounts. a. This box draws heavily on Marchiori (2011) and Krueger and Ludwig (2007). b. The intergenerational distribution effects are further complicated when the funding of pension schemes comes from payroll taxes. c. International migration will also be affected, and the mechanisms will be similar to those described for international capital fl ows. Global Development Horizons Global Saving in 2030 79 Compared with the start of the 21st century, the global saving rate will remain constant at saving rates will have fallen by 2030, a reflection around 24 percent. of increased demographic pressures from aging China will continue to be dominant in the populations in much of the world and slowing global savings picture. The continuous increase growth rates. In both the developing and devel- in the country’s saving over the scenario horizon oped worlds, saving rates will decline from their will mean that by 2030, its absolute saving will respective peaks (figure 2.14, panel a). For devel- be far and away the largest among all economies oping countries collectively, saving is anticipated worldwide, at around $9 trillion (measured in to reach a peak of 34 percent of their income in 2010 dollars) (figure 2.15). Th is large amount of 2014 and steadily creep downward to 32 percent saving is a function of China’s large population in 2030; for high-income countries, the corre- and high propensity to save, as often noted, but sponding reduction will be from 20 percent to those strong upward pressures on saving will be 16 percent. The significantly slower decline in to some degree off set by a rising old-age depen- the developing world’s saving rate, coupled with dency ratio (China’s ratio will surpass 2010 U.S. the increasing size of developing countries in the levels in the latter half of the 2020s) and decreas- global economy, will contribute to a widening ing economic growth. Thus, although China’s difference between the global saving contribu- saving will undoubtedly remain large, it will tions of the two groups. nevertheless be smaller than it would be in the Another implication of the increased relative absence of changes in population structure and size of developing countries is that the global the slowdown of economic expansion. saving rate will not fall in spite of decreases in The effects of a large population on national the saving rates of both groups. The larger eco- saving will also be seen in India, which will over- nomic size of the developing world, alongside take both Japan and the United States in terms the slower rate of decline in its relatively higher of its absolute level of national saving sometime saving rate, will off set the decline of saving in in the 2020s. A major factor supporting the rela- high-income countries. In summary, although tively rapid rate of growth of saving in India is the global flow of saving will rise dramatically its favorable demographics: India’s old-age depen- in absolute terms between 2010 and 2030, dency ratio will rise far more slowly than in East FIGURE 2.14 Saving rates will decline more slowly in developing countries, and by 2030 those countries will account for two-thirds of global saving a. Saving rates, 2005–30 b. Shares of global savings, 2030 40 Saving as a share of GNI, % 30 20 10 Developing China 0 Europe Russian Federation 2005 2010 2015 2020 2025 2030 United States India Japan Brazil High income Developing Other high income Other developing Source: World Bank projections. 80 Global Saving in 2030 Global Development Horizons FIGURE 2.15 China will continue to be dominant in the evident when comparing the saving rate paths global saving picture among developing countries (figure 2.16). In economies that will experience a simultane- 2.5 China 10 ous sharp reduction in the working-age share of (right axis) Saving in China, $ trillions their population and an increase in the old-age 2.0 8 United States share of their population, such as Indonesia and Saving, $ trillions 1.5 6 Russia, the saving rate will fall significantly. The more moderate decline in saving rates in Mexico 1.0 Japan 4 will be due in part to its relatively more positive India demographic evolution. 0.5 2 A remarkable exception to these general trends Middle East will be Sub-Saharan Africa, where the saving rate 0.0 0 will not fall thanks to a stable old-age depen- 2010 2015 2020 2025 2030 dency ratio and healthy rates of economic growth (the region will average about 5 percent annual Source: World Bank projections. Note: Saving levels are measured in 2010 U.S. dollars. growth in the gradual convergence scenario). Indeed, Sub-Saharan Africa stands as a remark- able outlier to the global pattern of declining sav- FIGURE 2.16 Differences in the demographic transition ing rates worldwide, largely because of the favor- will translate into different saving rate trends able combination of productivity growth and little demographic pressure. 40 35 India Under the rapid convergence scenario, the absolute level of Saving as a share of GNI, % Indonesia global saving will be slightly 30 Russian Federation larger 25 Global saving will be $27 trillion under the rapid Mexico scenario, whereas under the gradual conver- 20 gence scenario, it will be $25 trillion. This larger Sub-Saharan Africa amount of global saving is mainly due to the Brazil assumption of a higher growth rate in the global 15 economy. The overall global saving rate under 10 this new scenario will, however, still fall slightly: 2010 2015 2020 2025 2030 this is mainly because, even though developing countries save more in absolute terms, the nega- Source: World Bank projections. tive impact of higher levels of financial develop- ment on saving rates will dominate—households and firms need to set aside less saving in the pres- Asia or high-income nations. Still, as discussed ence of better financial intermediation—and the earlier in the chapter, India’s demographic divi- net effect will be a reduction in developing coun- dend will gradually peter out as it approaches tries’ saving rates, to 30 percent, versus 32 percent 2030, and this will moderate its increase in sav- in the gradual convergence scenario (figure 2.17, ing. Regardless, the saving rate in India will panel a).10 remain higher than that of many other develop- In the rapid scenario, faster fi nancial devel- ing countries, in no small part because of growth opment in developing countries tends to bring in per capita incomes. down their saving rates. The impact on the global The expected effects of asynchronous demo- saving rate, however, is offset by the fact that the graphic changes on saving activity are most global weight of developing countries (which still Global Development Horizons Global Saving in 2030 81 tend to have higher saving rates than advanced development and thus a greater negative impact countries) is greater relative to the gradual sce- on its saving rates. nario. Thus, the path of the global saving rate is The small reduction in global saving implies roughly the same across the two scenarios. an increase in the rental rate of capital worldwide, Relative to the first scenario, developing coun- however, because lower global saving reduces the tries will capture a slightly larger share of global supply of capital, while higher overall growth saving under the rapid convergence scenario (fig- increases the demand for capital (as discussed ure 2.17, panel b). Th is small change obscures a more thoroughly in chapter 1). Interest rates turn significant degree of heterogeneity in the paths out to be significantly higher in the rapid con- of different countries, however. This heterogene- vergence scenario than in the gradual scenario. ity, in turn, results from the difference in impact However, stronger price signals, and better finan- of fi nancial development on saving in this sec- cial intermediation in terms of coverage and effi- ond scenario: for countries starting at low ini- ciency, mean that savings will be channeled more tial levels of financial development, reducing by effectively to its most productive uses. one-fourth their distance from the United States means a larger increase of financial development. China, for example, will experience a decline in A more nuanced view of the its saving rate of 7 percentage points (from 52 future of saving: Distribution, percent to 45 percent), whereas Indonesia will see a sharp drop in its saving rate of more than public finance, old-age support, 7 percentage points (from 32 percent to 25 per- and other impacts cent). Both countries will follow healthy growth The scenarios presented in the previous sec- paths, but Indonesia, which in 2010 had a level tion explored the implications for global saving of financial development only a fifth of the high- of forecasts for productivity growth, demogra- income-country average (compared with China, phy, and financial development. The principal with a level close to two-thirds in the same year) conclusions were that higher growth tends to will experience a larger change in its fi nancial increase saving, a rise in the elderly dependency FIGURE 2.17 National saving rates (panel a) and share of global saving (panel b), by income groups, gradual versus rapid convergence scenario, 2010–30 a. Saving rate b. Share of global saving 40 70 Saving as a share of GNI, % 35 60 Share of global saving, % 30 50 25 40 20 30 15 20 10 5 10 0 0 2010 2015 2020 2025 2030 2010 2015 2020 2025 2030 High-income High-income Developing Developing World World (gradual) (rapid) (gradual) (rapid) (gradual) (rapid) Source: World Bank projections . Note: The gradual and rapid convergence scenarios refer to numerical simulations based on two sets of assumptions on productivity growth and structural changes (see details in the main text). 82 Global Saving in 2030 Global Development Horizons ratio tends to reduce saving, and higher financial Th is second part of the chapter complements development may lower saving rates but improves the general equilibrium-generated scenarios of the efficiency of investment allocation and thus global saving discussed above. The more disaggre- increases growth. Overall, average global saving gated analyses of the household and government rates would remain roughly constant through saving in this part are based on different tools 2030, in part because of the rising global gross and datasets, and therefore some of the mecha- domestic product (GDP) share of the relatively nisms operating in these settings—for example, higher-saving developing countries. Th is macro- the impact of education on the life-cycle saving economic analysis requires a number of simpli- behavior and the effects of demographic changes fying assumptions. For example, the population on government expenditure—were not integrated age structure of a country (or groups of countries) in the CGE modeling. is approximated by a simple old-age dependency ratio, growth is assumed to benefit all households Saving at the household level: in the same way, and saving is not disaggregated Who actually saves? into households, government, and firms. Other approaches to studying the future of The evolution of household saving and economic global saving are to consider how demography development are strongly interrelated. A glance and growth will affect saving at the household back at figure 2.6, for instance, shows that house- level, and how aging will affect public fi nances. hold saving represents a large share of national Examining saving from a microeconomic per- saving of an economy and, therefore, its capacity spective can provide a more realistic and nuanced to fi nance investment using domestic resources. view of the likely evolution of saving, although at Indeed, fundamental economic decisions con- the price of considering a limited set of countries cerning saving are made at the household level: given the data availability. Although the house- how time is allocated between leisure and work; hold-level analysis confirms the broad conclusions how income is earned and shared; how much of of the macroeconomic analysis, the micro-level income is consumed (and for which goods and data also expose the complexities of the interac- services) versus saved; and what proportion of tion among aging, saving, and growth. The anal- resources is devoted to investment in the educa- ysis fi nds that (a) aging tends to reduce saving; tion of the young.12 Analysis at the micro level (b) the size and composition of households simul- can thus provide substantial insights, also because taneously affect saving, labor supply, and other it considers distributional aspects ignored by rep- key economic decisions; and (c) all three of these resentative-agent approaches. factors have feedback effects on demography. It The microeconomic analysis undertaken here also finds that saving tends to be highly concen- considers the household-level data of a group of trated, which may initially help boost develop- developing economies in which saving behavior ment of financial systems but also has potentially is driven by distinct characteristics: a middle- negative implications for economic mobility and income economy with modest economic growth inclusion.11 rates (Mexico), a middle-income country with These fi ndings have significant public policy robust growth (Th ailand), a transition economy implications, not least of which is that the decline with an aging population (Russia), and a coun- in the extended family support system in many try with a large demographic dividend (Ghana). developing countries will require greater reliance The analytical approach adopted here is a cohort on formal fi nancial institutions, which in turn (or pseudo-panel) approach that uses a series of will affect both financial markets and government repeated cross-sectional data on both income and finances. Many countries will find it challenging consumption. This approach allows not only the to provide health care and retirement benefits for measurement of saving rates of individuals at dif- their aging populations while ensuring financial ferent ages, but also determines whether these sustainability and encouraging healthy rates of individuals belong to different cohorts (that is, if economic growth. they were born in different years). In other words, Global Development Horizons Global Saving in 2030 83 the differences in incomes and saving of a younger online annex 2.1 provide additional methodologi- individual relative to an older one can be decom- cal details. posed into a portion attributable to their different Following cohorts over time is appealing ages (the age effect) and a portion attributable to because it helps shed light on how the two key the fact that the younger individual belongs to drivers of saving—demography and growth— a cohort that, on average, has different income operate at a more disaggregated level. Cohort and saving (the cohort effect). In addition, saving analysis can help determine whether saving has rates change from one year to another (the time been rising because the share of the population at effect) as agents respond to shocks. Box 2.4 and high-saving ages has increased or because younger BOX 2.4 Micro analysis is a useful tool for determining likely trends in future savings Separating the age, cohort, and time effects in the esti- Most of the findings that follow from this analysis mation of the age-saving profile is not straightforward are illustrated by graphing variables of interest along and requires several identification restrictions. The micro the life cycle. Thus, it is useful to describe how these analysis presented in this report follows an approach graphs are constructed. In each country and for each that was pioneered by Heckman and Robb (1987) household survey, approximately 14 five-year birth and subsequently used by Deaton and Paxson (1994); cohorts are constructed. Cohort 1 is the youngest MaCurdy and Mroz (1995); Gokhale, Kotlikoff, and cohort; for example, in Mexico, cohort 1 is represented Sabelhaus (1996); Paxson (1996); Attanasio (1998); and by all household heads born 1985–89, so that, in the more recently, Chamon and Prasad (2010). The saving most recent surveys of 2009 –10, they are 20 to 25 data, consisting of data points for each cohort and each years old. Cohort 2 represents household heads who year, are the dependent variables; these are regressed were born 1980 – 84 and were repeatedly observed on dummies for age, cohort, and time (where time is in the surveys of 2004 at the “average” age of 22, in the year of the survey). The restrictions are equivalent the survey of 2005 at the age of 23, and so on, until to assuming that “all deterministic trends in the savings the most recent survey of 2010 at the age of 28. The rate data originate from a combination of cohort and age oldest cohort, cohort 14, represents household heads effects” (Attanasio and Szekely 2003, 15).a who were born 1920–24; in the first Mexican survey of For the four countries examined here, household 1984, these heads were 60–64 years old. The graphs saving is estimated from high-quality household sur- plot the cohort’s average for the variable of interest veys. These surveys are all nationally representative against the age of the head, and so the age profile and were conducted using a broadly comparable meth- of the cohort is obtained by connecting the different odology over a fairly long period covering the mid- points in time when the cohort is observed. Given that 1980s to the late 2000s. The saving variable estimated the surveys cover a period of about 20 years, cohorts from the surveys is obtained as simply the difference can be followed over a large portion of their life cycles. between total household disposable income (which Furthermore, this construction allows different cohorts includes imputed amounts—for example, imputed to be observed at the same age. Figures 2.23, 2.27, rents from owning a house, income in kind, and con- 2.28, and 2.29 are presented in this manner. sumption of automobiles) and total expenditures.b, c a. For more details on the estimation procedure and its relationship with the life-cycle theory, see online annex 2.1. b. Other definitions of saving have been used, but the main results do not vary significantly. Additional definitions include (a) excluding investment com- ponents—such as purchases of durable goods and spending on education and health—from the total expenditures; and (b) netting transfers, such as pensions, from the income aggregate. Note that for the case of Russia, the estimation of consumption is not reliable, but the questionnaire included a specific question on saving. Therefore, for the case of Russia, we used this self-reported saving variable. c. These micro-based estimations of household saving are not strictly comparable with domestic saving in the national accounts (NA): for these coun- tries, NA’s domestic saving includes corporate saving, and the definitions of income and consumption are not consistent across the NA and the house- hold surveys. 84 Global Saving in 2030 Global Development Horizons FIGURE 2.18 National saving rates of Thailand, the Russian Federation, Mexico, and Ghana are similar to those of their regions a. Thailand b. Russian Federation 40 40 THA (3-yr MA) Saving as a share of GNI, % Saving as a share of GNI, % 35 35 RUS (3-yr MA) E. Eur. & Cen. Asia 30 30 25 25 20 20 15 East Asia & Pacific 15 (without China) 10 10 5 5 0 0 1980 1984 1988 1992 1996 2000 2004 2008 2011 1980 1984 1988 1992 1996 2000 2004 2008 2011 c. Mexico d. Ghana 40 40 Saving as a share of GNI, % Saving as a share of GNI, % 35 35 30 MEX (3-yr MA) 30 25 25 20 20 15 15 L. Amer. & 10 the Caribbean 10 Sub-Saharan Africa 5 5 GHA (3-yr MA) 0 0 1980 1984 1988 1992 1996 2000 2004 2008 2011 1980 1984 1988 1992 1996 2000 2004 2008 2011 Source: World Bank calculations using data in the UN National Accounts Main Aggregate database. Note: Series represent three-year moving averages (MA). cohorts have been saving more, possibly because analyses can be used to make qualified general- of stronger economic growth. Discerning the izations regarding broader groups of countries. difference between these two cases is important As is the case for national saving rates, trends because it may suggest whether the expansion of in household saving rates in the four countries aggregate saving is temporary (because a large of interest have varied greatly in recent decades cohort is currently in the working-age stage of (table 2.1). In Ghana and Thailand, there has its life cycle) or more permanent. The estimated been an upward trend in household saving rates age and cohort profiles of saving can also be used since the late 1980s. Th ailand’s household sav- to make exploratory inferences about how saving ing rate was also much higher than in the other will evolve in the future. three countries over the same period. Mexico’s Before considering the results of the analysis household saving rate shows some sign of increas- in detail, it is useful to examine some descrip- ing toward the mid-late 2000s, but it is reduced tive statistics on saving. Figure 2.18 shows that again if 2010 is included, presumably an effect of the levels and movements of national saving rates the global financial crisis. The household data for of Thailand, Russia, Mexico, and Ghana are Russia show not only a much lower saving rate similar to those of their regions, thus supporting relative to the other three countries but also little the notion that the fi ndings of specific country variation over time. The low rate of household Global Development Horizons Global Saving in 2030 85 saving in Russia is perhaps not surprising because TABLE 2.1 Trends in saving rates differ widely across household saving in the Eastern Europe and countries percent Central Asia region is lower than in other devel- oping regions. Russian These micro data confirm one of the main Year Mexico Federation Thailand Ghana findings from the macro trends described ear- 1984 7.2 —  —  —  lier in the chapter: apart from Russia, develop- 1987 —  —  —  4.9 ing countries have increased their saving rates, 1988 — —  8.6 6.4 and micro data show that some of this increase 1989 11.1 —  —  —  is attributable to increases in households’ saving 1990 —  —  17.1 —  rates. 1991 —  —  —  14.2 1992 12.0 —  20.9 —  1994 14.2 —  21.2 —  How do demographic differences 1996 9.5 —  27.8 —  drive differences in savings? 1997 —  4.7 —  —  Ghana versus Russia 1998 13.0 3.8 31.6 13.2 1999 —  3.4 33.5 —  Together with the assumptions of the life-cycle 2000 11.0 5.2 31.2 —  hypothesis, the age-saving profiles identified 2001 —  5.5 —  —  with the pseudo panel approach can be used to 2002 11.0 5.4 29.9 —  make some simple inferences on the outlook for 2003 —  4.6 —  —  household saving. Theory suggests that the bulk 2004 10.3 3.3 27.6 —  of an individual’s lifetime saving occurs dur- 2005 14.2 4.4 —  —  ing the middle of his or her life cycle—that is, 2006 10.8 5.9 20.6 11.3 during the individual’s working years. Thus, an 2007 —  5.3 22.8 —  2008 20.3 4.0 —  —  increase in the share of a population of work- 2009 —  —  24.1 —  ing age should increase aggregate savings. Th is 2010 11.1 —  —  —  inference rests on strong assumptions, however: Source: World Bank calculations using household surveys. it assumes no behavioral change and no price, Note: For all countries except Russia, household saving rates are estimated as the ratio of wage, rental rate, or other general equilibrium the difference between the household income and household consumption over house- hold income. For Russia, the household saving rate is self-reported by survey participants. effects. In fact, it is mainly an accounting projec- — = not available. tion, and as such it should be taken with a good bit of caution. However, studying the relationship between the age structure and saving is still rel- fraction is the income weight of cohort c for evant and interesting in that it can give an indica- the year t. These weights are also predicted tion of how differences in countries’ progression for time t using the age and cohort effects for along the demographic transition will affect their the income levels (Y ˆc,t ) and the future popula- saving profiles in the coming decades. The case of tion levels (Nc,t ). Ghana versus Russia is considered first. The results of this exercise are shown in figure The projected saving rate (as shown in figure 2.20, which depicts an upward trend in projected 2.19) is estimated using the following equation: aggregate household saving rates in Ghana and a downward trend in Russia. Yˆ ,t N c,t ˆ = Stot,t ∑ c Sˆc,t ∑ cY ˆ N These opposite trends in household saving c c, t c ,t (2.1) are due to the different shapes of the age effects for saving (as reported in figure 2.19) and to the where Sˆ tot,t is the projected aggregate sav- different patterns of demographic transition in ing rate for year t , Sˆ c,t is the saving rate of Ghana and Russia. Figure 2.20 (panel a) shows cohort “c” predicted for year t (using the age that the ratio of working-age population to and cohort profi les identified by the regres- nonworking-age population will rise from about sions described in online annex 1.5), and the 1.4 to 1.8 in Ghana (and, in fact, for the entire 86 Global Saving in 2030 Global Development Horizons FIGURE 2.19 Projection of age effects on household saving rates in Ghana and the Russian Federation a. Ghana b. Russian Federation 45 8 40 7 35 6 Saving rate, % Saving rate, % 30 5 25 4 20 3 15 10 2 5 1 0 0 27 32 37 42 47 52 57 62 67 72 23 28 33 38 43 48 53 58 63 68 73 78 Age of head of household Age of head of household Source: World Bank calculations using household survey data. FIGURE 2.20 Demographic change will drive large changes in saving rates in the decades ahead a. Ratio of working- to nonworking-age population b. Household saving rates 2.4 18 5 2.2 17 Ghana 4 16 2.0 Russian Federation 15 4 Percent Percent 1.8 Ratio 14 1.6 3 13 1.4 Ghana 12 Russian Federation (right axis) 3 1.2 11 1.0 10 2 2010 2015 2020 2025 2030 2035 2040 2045 2050 2010 2015 2020 2025 2030 2035 2040 2045 2050 Source: World Bank calculations using household survey data. Sub-Saharan region, as shown previously in fig- can be explained by considering more closely ure 2.13) between 2010 and 2050. In Russia, how the forecast is generated. First, note that however, the ratio will fall considerably over the the saving rate among people of retirement age same decades, from about 2.3 to less than 1.4. in Thailand will fall only slightly (figure 2.21, panel a). Second, the cohort effect will play an important role in Thailand (figure 2.21, panel How do demography and growth b). In the forecast, the future (still-unborn) interact to affect saving? Thailand cohorts are assigned the cohort effect of the most versus Russia recent cohort (that is, those near the vertical In Thailand, the saving rate is forecast to rise axis)—which, in the case of Thailand, reflects a between 2010 and 2050 despite the expected fairly high saving rate (as younger cohorts’ rapid decline in the share of the working-age popu- growth in incomes results in higher saving rates lation over the projection period. This result than for older cohorts). Thus the aggregate saving Global Development Horizons Global Saving in 2030 87 FIGURE 2.21 Due to strong growth, the cohort effect will play an important role in the path of Thailand’s saving rate a. Saving rate (age effect) b. Saving rate (cohort effect) 35 45 40 30 Household saving rate, % Household saving rate, % 35 25 30 20 25 15 20 15 10 10 5 5 0 0 22 27 32 37 42 47 52 57 62 67 72 1 2 3 4 5 6 7 8 9 10 11 Age of household head Cohort (lower number = younger cohort) Source: World Bank calculations using household survey data. Note: Cohorts are defined in terms of fi ve-year age ranges of birth year. For Thailand, cohort 1, the youngest cohort, is represented by all household heads born 1985–89, so that, in the most recent surveys of 2009–10, they are roughly 20 to 25 years old. Cohort 2 represents household heads who were born 1980–84, and so on. rate rises over time as younger (higher-saving) FIGURE 2.22 Growth is an important determinant of cohorts replace older (lower-saving) cohorts. This saving rates: Two scenarios for the Russian Federation effect is reinforced because these younger cohorts 5.0 are larger in size and move along an income-age profile that is above that of older cohorts, so their Household saving rate, % weight in aggregate saving rises over time. 4.5 Another illustration of how higher income growth increases saving can be seen by comparing 4.0 Alternative the baseline forecast for Russia with an alternative case. In the alternative case, higher growth is sim- 3.5 Baseline ulated by increasing the cohort effect on incomes of the new (still-unborn) cohorts.13 As will be the 3.0 case for Thailand, these new cohorts will have higher saving rates and higher incomes than older 2.5 cohorts. By 2030 in the alternative, increased- 2010 2015 2020 2025 2030 2035 2040 2045 2050 growth scenario, the share of total income earned Source: World Bank calculations using household survey data. by the younger cohorts (considered those born Note: The figure shows the difference in projected saving rates under two income growth sce- between 1990 and 2009) would reach 37 percent, narios for Russia that approximate the gradual (baseline) and rapid (alternative) convergence scenarios discussed in the first part of the chapter. versus 27 percent in the baseline scenario. The gap between the two scenarios is even larger, with incomes of the younger cohorts (now considered those born between 2010 and 2029) accounting the saving rate will be about half a percentage for 51 percent of the total, versus 26 percent in point higher by 2050 in the alternative scenario the baseline scenario. than in the baseline scenario (figure 2.22). Th is As the share of total income accounted for by conclusion—that rising incomes are associated cohorts with higher saving rates rises, so will the with higher saving rates—reinforces the finding aggregate saving rate. Thus, in the case of Russia, from the previous macro modeling section that 88 Global Saving in 2030 Global Development Horizons growth generates the means to sustain itself (it is The saving profiles of Mexican households also consistent with a basic prediction of the stan- differ significantly by education level. The age- dard life-cycle model).14 saving profi les tend to be quite flat for the least- educated group and higher and steeper for the more-educated groups (see figure 2.24, panel a). Increases in educational There are some mild declines in saving rates for attainment will support growth household heads who are above age 65, especially and saving: The case of Mexico in the case of the group with tertiary education. In Mexico, as in other developing countries, the Higher saving by more-educated heads of house- entry of larger numbers of skilled workers into the holds reflect their anticipated rising income lev- labor market in the coming decades will support els, while heads with lower education face a much growth in both incomes and saving rates, even flatter age-earning profi le. Because education is after accounting for some erosion in the wage typically a good proxy for permanent income,15 premium due to the increased supply. This effect this analysis helps explain why low-income (and can be seen by grouping households according to typically not well educated) households tend to educational attainment of the household head. As save less than high-income (and typically better- shown in figure 2.23, panel a, household heads of educated) households. young cohorts are much more likely than those Another observation that can be gleaned from of older cohorts to have a secondary or higher figure 2.24 (panel b) is that young cohorts (those education: only about 10 percent of household closest to the vertical axis) have higher saving heads born in the 1920s have a secondary educa- rates than old cohorts: saving rates for the young- tion, but that figure has steadily risen over time to est cohorts are 10 percentage points higher than about 50 percent for household heads born in the for the oldest cohorts, a significant change. Th is mid-1980s. Similar progress can be observed for indicates that the long-term trend (at least for the share of heads with higher education (figure the most recent decades observed in these sur- 2.23, panel b). veys) has been upward-sloped for saving.16 The FIGURE 2.23 Educational attainment in Mexico has been rising over time a. Household heads with secondary schooling b. Household heads with higher education 50 40 40 Household heads, % Household heads, % 30 20 20 10 0 0 30 40 50 60 70 80 20 40 60 80 Age of household head Age of household head Source: World Bank calculations using household survey data. Note: Box 2.4 provides details of how the cohorts shown in these figures were constructed Global Development Horizons Global Saving in 2030 89 FIGURE 2.24 The saving life cycle differs significantly by educational attainment (panel a), but across all educational levels, young cohorts tend to save more than older cohorts (panel b) a. Saving rates by education and age b. Saving rates by education and age cohort 0 1 0 1 40 20 10 Household saving rate, % Household saving rate, % 20 0 0 –10 –20 –20 2 3 2 3 40 20 10 20 0 0 –10 –20 –20 20 40 60 80 20 40 60 80 0 5 10 15 0 5 10 15 Age Cohort Source: World Bank calculations using household survey data. Note: 0 indicates whole population, 1 population with no education or just primary education, 2 secondary education, and 3 tertiary education. Cohorts are defined in terms of fi ve-year age ranges of birth year. steepness of the upward trend, however, differs three groups by its share of total income. These across education groups. The data for Mexico education-augmented demographic projections show that the positive cohort effect is strongest (figure 2.25, panel a) show that by 2050, the for the most-educated group. least-educated group will account for the small- To estimate the path of aggregate saving in est share of nonworking-age people in Mexico. Mexico over coming decades, we begin with a Because the least-educated group’s earnings and baseline forecast in which the impact of educa- saving rates—in terms of both age and cohort tion on saving rates is not taken into account. effects—can be expected to continue to be lower The results show that the aggregate saving rate than those of other groups, the economywide sav- will rise until about 2025 because the share of ing rate will be almost 5 percentage points above working-age people in the total population will the baseline scenario by 2050. increase over the same period, thus allowing aggregate saving rates to rise because more of the Poor households save little, population will be earning income. The forecast limiting their ability to escape aggregate saving rate levels off in 2025, however, poverty: Mexico and Thailand as the share of the nonworking-age population starts to decrease after reaching a peak. In most countries, saving tends to be concen- An alternative scenario, which ref lects the trated in households that are relatively high- impact of rising education levels on savings, pre- income compared with the country average. In dicts considerably higher saving rates over the Mexico and Th ailand, saving rates are actually forecast period. To generate projections that take negative up to the 25th percentile of the income into account the different saving and income pro- distribution. Even if this is partly due to measure- fi les, equation (2.1) can be applied separately to ment error or temporary fluctuations,17 the poor- three groups of households: those with low or no est quarter of households seems to accumulate education, those with secondary education, and much less saving, while saving rates of households those with tertiary education. An average saving beyond the 75th percentile of the income distri- rate is then calculated by weighting each of the bution are high: saving rates among the richest 1 90 Global Saving in 2030 Global Development Horizons FIGURE 2.25 Increased earning power will be the greatest driver of saving by Mexican households a. Ratio of working- to nonworking-age population, b. Household saving rate, with and by education level of household head without education effect 2.5 18 nonworking-age population 17 Household saving rate, % 2.0 Ratio of working- to 16 1.5 15 14 1.0 13 12 0.5 11 0 10 2010 2015 2020 2025 2030 2035 2040 2045 2050 2010 2015 2020 2025 2030 2035 2040 2045 2050 Primary or Secondary Tertiary Demographic Demographic + no education education education effect education effect Source: World Bank calculations using household survey data. percent of households are consistently above 60 income. Grouping the saving data by the level of percent in Mexico and 70 percent in Thailand education of the household head provides a good (table 2.2). Furthermore, this gap between rich proxy for permanent income because heads sel- and poor households’ saving rates does not appear dom change their educational attainment after to have diminished in recent decades. forming a household. Therefore, tracking house- Admittedly, there are drawbacks to comparing holds grouped by education over time minimizes the saving behavior of households across income the potential bias introduced by composition percentiles to ascertain whether there are perma- effects. nent vulnerabilities for certain groups. Low or In both Mexico and Thailand, households negative saving rates may result from consump- where the head has low educational attainment tion smoothing by households facing a temporary tend to also have low saving. According to the shock. Similarly, the highest-income households most recent household survey data—2010 for are disproportionately in the high-earning years of Mexico and 2009 for Th ailand—the least-edu- their life cycles, and their earnings likely include cated group accounted for 17 percent and 34 per- greater than average non-age-related transitory cent of saving, respectively. By contrast, in Mexico components as well. Total transitory components about 60 percent of total saving was provided by of income can be large and, because of economic the most-educated group of heads of household mobility, saving by those households that make versus 47 percent in Thailand. Moreover, saving up an income percentile at a given point in time tends to be even more concentrated than income: may be determined either by long-term saving shares of total income accounted for by the most- behavior or by composition effects. Th at is, low educated groups were 50 percent and 37 percent saving among the lowest percentiles may reflect for Mexico and Thailand, respectively, according income smoothing by households that are tem- to the most recent survey data. The fact that these porarily poor, while high saving among the high- income shares are much lower than the saving est percentiles may reflect smoothing by house- shares means that more-educated people not only holds that are temporarily flush. Alternatively, save more in absolute terms but also save, on aver- low saving may reflect a permanent incapacity of age, a higher proportion of their incomes. some households to accumulate assets for their This concentration of saving among household retirement or to cope with unexpected drops in heads with greater education, and thus greater Global Development Horizons Global Saving in 2030 91 TABLE 2.2 Household saving rates differ significantly along the income distribution   Percentile 1st 5th 10th 25th 75th 90th 95th 99th Year Mexico 1984 −177.4 −69.7 −38.2 −12.9 22.7 39.2 48.0 64.1 1989 −215.9 −79.1 −49.1 −14.9 25.6 41.9 52.5 70.5 1992 −140.0 −60.6 −40.3 −13.0 21.7 36.4 46.1 65.4 1994 −120.4 −47.5 −29.5 −8.6 22.8 38.4 47.4 64.6 1996 −135.9 −55.2 −34.2 −12.7 17.2 32.0 41.8 60.4 1998 −178.9 −66.8 −39.7 −13.4 21.9 38.8 47.1 64.9 2000 −142.8 −54.9 −33.5 −10.3 22.2 36.8 45.4 62.2 2002 −131.5 −53.3 −32.3 −8.7 23.6 39.0 47.6 61.7 2004 −158.2 −72.9 −45.2 −14.7 24.6 40.2 49.0 62.7 2005 −178.8 −66.8 −39.2 −11.2 27.9 43.2 51.5 67.5 2006 −154.5 −66.7 −41.7 −12.4 24.2 39.1 47.7 63.8 2008 −151.6 −76.9 −49.7 −16.8 31.5 49.0 58.5 73.8 2010 −237.2 −91.8 −55.0 −17.5 26.8 42.9 52.1 68.0   Thailand 1990 −387.2 −141.6 −81.8 −28.4 26.7 45.2 55.3 71.6 1992 −302.2 −127.2 −76.3 −23.1 28.0 46.3 55.8 73.8 1994 −332.7 −130.5 −80.0 −26.3 31.8 49.6 58.9 74.6 1996 −193.7 −108.3 −64.1 −14.5 36.9 54.6 63.1 76.5 1998 −254.5 −97.9 −55.8 −11.7 37.2 54.3 63.0 78.3 1999 −264.3 −95.4 −51.5 −9.3 36.2 52.9 62.3 75.6 2000 −226.2 −88.5 −47.7 −8.1 37.5 54.6 63.5 77.7 2002 −239.9 −85.0 −46.5 −8.0 37.9 53.8 62.0 76.1 2004 −231.3 −83.3 −46.4 −8.7 37.3 53.7 62.5 76.7 2006 −278.6 −98.2 −54.9 −16.1 24.9 43.3 53.9 72.7 2007 −214.4 −79.6 −44.5 −10.6 26.3 44.8 55.0 72.9 2009 −166.9 −60.3 −32.2 −6.9 26.2 43.5 54.3 73.4 Source: World Bank calculations using household survey data. permanent earning capacity, has increased over Mexico, reductions in the skill premium may the most recent two decades for Thailand but have moderated the increase in the share of the not for Mexico. In Th ailand, the share of total highly educated in total saving.19 saving by households with secondary education Low saving among the poorly educated likely has basically been stable, while there has been reflects a long-term incapacity to save because an increase among the group with higher edu- they face much f latter age-earning profiles. cation, counterbalanced by a reduction for the Negative saving may ref lect these households’ group with no or very little education. This trend long-term incapacity to accumulate resources toward increased concentration of savings par- because they face unemployment or other bar- tially reflects the increased weight of educated riers to income generation, while consumption individuals in the total population. The share exceeds income because of either public or pri- of households with a highly educated head has vate transfers. Lack of saving makes it difficult for increased quite significantly as well, not just for these households to achieve sustainable increases the two countries considered here but also in the in income by investing in the education of their developing world more generally.18 However, for members or to cope with adverse shocks. The 92 Global Saving in 2030 Global Development Horizons concentration of saving thus exacerbates the dif- poor households’ limited participation in formal ficulties involved in reducing poverty, raising the financial markets further limits the financial prospect of a permanent underclass with little inclusion of other poor households, potentially potential for upward mobility. contributing to poverty traps. The concentration of saving among high earners also has important implications for Saving behavior depends fi nancial inclusion in developing countries. On on household formation and one hand, concentration of fi nancial services, composition over the life cycle in terms of either geography or income levels, can make it less costly for financial institu- There are several reasons why household data for tions to enter and operate in developing-coun- developing countries may not reveal the expected try markets. On the other hand, the resulting inverted-U shape for the relationship between concentration of financial institutions among aging and saving (saving rises during a person’s the high-income segment of a population, and working years, then levels off and declines during in the areas where they live and work, has the retirement): potential to perpetuate lack of access to financial services for the poor. Analysis of recent survey • Because consumption is measured at the data on saving vehicles in developing countries household level, it is often difficult to pre- (Demirgüç-Kunt and K lapper 2012) shows cisely attribute consumption amounts to that the bottom 40 percent of income earners individual members of the household. As in developing regions are not only less likely to pointed out by Deaton and Paxson (2000), save than the top 60 percent but are also less many individuals may live in multigen- likely to save using formal fi nancial institutions erational households, such as a 45-year-old (fi gure 2.26). Th is suggests that the concentra- household head living with his 20-year- tion of saving associated with income inequality old son and 70-year-old father. At the may be exacerbated by a spillover effect in which household level, the positive saving of the 45-year-old may be cancelled out by the negative saving of the young and old mem- bers of his household. Thus, the age pro- FIGURE 2.26 The bottom 40 percent of income earners in file of household saving may be biased and developing regions are not only less likely to save than the quite different from the age profile of indi- top 60 percent but are also less likely to save using formal financial institutions vidual saving. • Elderly individuals who are relatively poor 100 may be more likely to live with their chil- 90 dren than richer ones, thus creating a selec- 80 tion bias. 70 • Different mortality rates may affect the 60 composition of the cohorts because longev- Percent 50 ity and wealth may be positively correlated, 40 as Attanasio and Hoynes (2000) point out. 30 As a cohort ages, richer individuals survive longer and become the larger subgroup 20 within the cohort. If they continue saving 10 to leave bequests, the inverted U age-saving 0 Sub- E. Asia & E. Eur. & L. Amer. Mid. East S. Asia profile may disappear mainly because of Saharan Pacific Cent. Asia & the & N. Africa changes in the composition of the cohorts. Africa Caribbean Saved with a Saved only Did not save The potential biases introduced by these financial institution by other means observations help explain why the hump-shaped Source: World Bank calculations using data in the World Bank Global Financial Inclusion data- relationship between aging and saving predicted base and Demirgüç-Kunt and Klapper 2012. Note: The left bar for each region represents the bottom 40 percent of the population by by theory is not always observed in the cases income; the right represents the top 60 percent. examined here. Global Development Horizons Global Saving in 2030 93 Figure 2.27 shows, for Mexico, Thailand, and FIGURE 2.27 The age composition of households varies Russia, the average age of the household head by age of the household head; more highly educated against the age of the individuals who live with individuals are less likely to live in a household headed by that head. Four panels are displayed for each coun- someone younger try: one for the whole population, and three for the a. Mexico different education levels. The 45-degree line indi- 0 1 cates the cohort profile that would be observed if 80 all individuals were heads or were living with heads 60 Age of household head of their same age. However, as shown in the figure, 40 there are clear deviations from this case, especially 20 at the beginning and end of the life cycle. 2 3 At the beginning of the life cycle, household 80 heads tend to be older (above the 45-degree line) 60 than other members of their households. Th is is because there are fewer heads at fairly young ages, 40 and some young individuals still live with their 20 parents. The opposite effect occurs toward the 20 40 60 80 20 40 60 80 end of the life cycle, when the elderly cease being Age of other individual(s) in household household heads when they return to live with their children, thus making household heads b. Thailand younger (below the 45-degree line) than other 0 1 80 members of their households. The deviations from the 45-degree line dis- 60 Age of household head play a common pattern across all countries: the 40 higher the education level, the less likely it is 20 that old people live in a household headed by a 2 3 person younger than themselves. Th is suggests 80 that more-educated individuals may have better 60 means by which to support themselves (accumu- 40 lated saving) during old age. Conversely, families 20 play an important role of smoothing consump- tion for old people with lower education and with 20 40 60 80 20 40 60 80 lower earning and saving capacity. Age of other individual(s) in household Apart from this commonality, the three coun- c. Russian Federation tries differ significantly in the relevance of this 0 1 intergenerational (or extended) household setup. 80 Specifically, this phenomenon seems less preva- 60 lent in Mexico than in Russia or Th ailand. Th is Age of household head 40 finding can also be observed at the regional level, where a larger proportion of old people maintain 20 2 3 head-of-household status in Latin America than 80 in East Asia or in Eastern Europe and Central 60 Asia, as shown in table 2.3. The table also illus- trates that the proportion of elderly people who 40 are not household heads, and for whom the sup- 20 port of extended families is important, decreases 20 40 60 80 20 40 60 80 as income levels increase: the share declines from Age of other individual(s) in household about 30 percent in low-income countries to 10 Source: World Bank calculations using household survey data. percent in high-income countries. The reduction Note: In each set of four panels, 0 indicates whole population, 1 population with no education is even more striking for females. or just primary, 2 secondary education, and 3 tertiary education. Box 2.4 provides details of how the cohorts shown in the panels were constructed. The 45-degree line indicates the Notice that figure 2.27 also provides a visual cohort profile that would be observed if all individuals were heads or were living with heads of corroboration of the potential bias due to the their same age. 94 Global Saving in 2030 Global Development Horizons TABLE 2.3 The prevalence of intergenerational households varies by country income level and by region percentage who maintain head-of-household status Population above 70 Population above 65     Female Male Female Male Low-income economies 23 66 24 71 Lower-middle-income economies 30 76 28 80 Upper-middle-income economies 47 81 45 82 High-income OECD members 56 89 53 90 East Asia and Pacific 32 79 32 82 Eastern Europe and Central Asia 50 76 48 77 Latin America and the Caribbean 45 85 44 87 Middle East and North Africa 5 24 5 29 South Asia 14 72 14 77 Sub-Saharan Africa 42 88 43 89 Source: World Bank calculations using household surveys of the Global Income Distribution Dynamics, data for 2005. Note: OECD = Organisation for Economic Co-operation and Development. change in household composition discussed in Eastern Europe and Central Asia, South Asia, above. Even if the group of household heads born and Sub-Saharan Africa after reaching a peak, in the same year is followed through time, the age indicating once again that multigenerational composition of their households is changing, and households are common in these regions of the this may affect saving decisions beyond the pure world. (individual head’s) age effect. Changes in household size during the life Evolving systems of old-age cycle also affect household behavior. Figure 2.28 income support plots household size against the age of the house- hold head for the four countries across different In most developing countries, extended fami- cohorts and education levels. As postulated by lies have traditionally cared for the elderly who Becker (1993) in his economic theory of family can no longer work. However, this is changing formation, there is a trade-off between quantity for the educated population in the four countries and quality of offspring. Households with low analyzed here: the higher the level of education, levels of income tend to opt for more children and thus the higher the level of accumulated sav- because they are unable to invest in their human ing, the less likely it is that elderly persons join capital. Th is is clearly visible across all the four younger households. Thus, rising incomes in countries, where household size is larger for lower developing countries can be thought of as shift- education groups. ing the support system for the elderly from infor- Additional observations that can be gathered mal provision by relatives to formal provision from figure 2.28 are that younger cohorts tend through pension systems. The shift from infor- to have smaller family sizes than older ones, and mal to formal provision of support for the elderly, that family sizes differ quite a lot across differ- which is likely to accelerate as incomes rise in ent countries.20 Once again, these findings can be the developing world, changes the nature of the validated by a broader regional analysis, as shown risks the elderly face in sustaining their levels of in figure 2.29 (additional background informa- consumption after retirement. It will also require tion on how this figure was generated is discussed changes in the regulation of developing countries’ in box 2.5). financial systems. In the regional breakdown, it is also possible The existence of large pools of saving, and to see that household size declines quite slowly the need to maintain the real value of this Global Development Horizons Global Saving in 2030 95 FIGURE 2.28 Household size varies by age of the household head, and this relationship varies somewhat by educational level a. Mexico b. Thailand 0 1 0 1 2.0 1.5 1.5 1.0 1.0 Log household size Log household size 0.5 0.5 0 2 3 2 3 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0 20 40 60 80 20 40 60 80 20 40 60 80 20 40 60 80 Age of household head Age of household head c. Russian Federation d. Ghana 0 1 0 1 2.0 1.5 1.5 1.0 1.0 Log household size Log household size 0.5 0.5 0 0 2 3 2 3 2.0 1.5 1.5 1.0 1.0 0.5 0.5 0 0 20 40 60 80 20 40 60 80 20 40 60 80 20 40 60 80 Age of household head Age of household head Source: World Bank calculations using household survey data. Note: In each set of four panels, 0 indicates whole population, 1 population with no education or just primary, 2 secondary education, and 3 tertiary education. Box 2.4 pro- vides details of how the cohorts shown in the panels were constructed. saving over several decades, will encourage the pension plans incur substantial investment losses growth of fi nancial intermediaries such as pen- or are inadequately regulated. Poorly designed sion funds, investment firms, and life insurance public pension systems present similar invest- companies. The growth of such intermediar- ment risks for the elderly in addition to the ies has the potential to improve the efficiency potential loss of benefits that may accompany a of investment by increasing the involvement by redefi nition of benefits or age-related eligibility professionals who specialize in evaluating the levels. Changes in the taxation of pensions, or profitability and soundness of diff erent invest- the presence of high infl ation that reduces the ment vehicles. On the other hand, the growth real value of fi xed-income annuities, can nega- of formal pension systems can also present the tively affect both private and public pensions. elderly with new risks, including a reduction in Of course, informal support systems also present their expected standard of living if their private significant risks for the elderly, principally in the 96 Global Saving in 2030 Global Development Horizons FIGURE 2.29 Younger cohorts tend to have smaller family sizes, but this relationship varies significantly across regions a. East Asia and Pacific b. Eastern Europe and c. Latin America and Central Asia the Caribbean 8 8 8 6 Family size Family size Family size 6 6 4 4 4 2 2 2 15 25 35 45 55 65 75 85 95 15 25 35 45 55 65 75 85 95 15 25 35 45 55 65 75 85 95 Age of household head Age of household head Age of household head d. Middle East and e. South Asia f. Sub-Saharan Africa North Africa 8 8 8 6 Family size Family size Family size 6 6 4 4 2 4 2 2 15 25 35 45 55 65 75 85 95 15 25 35 45 55 65 75 85 95 15 25 35 45 55 65 75 85 95 Age of household head Age of household head Age of household head Source: World Bank calculations using census data from the Integrated Public Use Microdata Series, International (IPUMSI). Note: Values in each panel are simple averages. See box 2.5. form of unanticipated declines in their family care and pension expenditures are likely to grow members’ incomes. The important point is that even more rapidly than incomes. Moreover, tax the risks will evolve in the decades ahead. revenue is heavily dependent on the size of the working-age population, which will shrink in some developing regions in the years ahead. The Aging will challenge the rising burden of expenditures on the elderly will sustainability of fiscal policy in not be off set by the declines in expenditures on both developing and advanced education that one might expect to result from a countries decreasing share of young people because rising Two of the largest components of government incomes will also lead to increased demand for expenditures in many countries are health care education per child. Thus, aging and growth can and pensions. Both of these components require be expected to increase the fiscal burden of age- more resources as the average age of the popula- related expenditures. tion rises. At the same time, as incomes increase, Th is strain on public fi nances has important the share of these expenditures in income tends to implications for national saving. The public com- rise. Thus, as populations age and incomes con- ponent of national saving could contract because tinue to increase in developing countries, health of fiscal deficits or reduced public investment. In Global Development Horizons Global Saving in 2030 97 addition, various policy changes to counteract countries. Pension systems are also potentially demographic pressures—such as tax increases, highly sensitive to demographic change, so pro- reduction in coverage or generosity of age-related jections of public pension costs are critical for benefits, or cuts to non-age-related spending— thinking about the magnitude and timing of may affect household saving behavior. These demographic pressures on overall public finances. important saving dynamics are not explicitly 21 Most projections suggest that the increased pub- accounted for in the general equilibrium model- lic spending on health care will be as important generated scenarios; therefore, this section com- as pensions, accounting for roughly half of the plements the previous analysis, although in a par- increase in age-related liabilities over the next few tial equilibrium setting. decades.22 Already, pensions are a significant component The data available on age-related public expen- of the government budget in many developing ditures by age of recipient (from the National BOX 2.5 A global dataset of household censuses provides a nuanced view of saving at the micro level The data behind figure 2.29 come from censuses birth year of the heads. For example, household heads that have been carried out in 43 developing countries born between 1870 and 1874 belong to cohort 1, while since the 1960s. These censuses are collected and those who were born between 1875 and 1879, to standardized by the Integrated Public Use Microdata cohort 2, and so on. These cohorts are then followed Series, International (IPUMSI, http://international. through time for all the dates for which the censuses ipums.org). Table B2.5.1 shows the number of coun- were repeated. For most countries, there are at least tries by region with available census data. four or fi ve repeated censuses. The data for the vari- Using these data, we calculated the year of birth for ous countries were then averaged to obtain information each household head in the censuses and thus iden- at the regional level. The averages were calculated as tified all household heads between 15 and 90 years simple averages or as weighted averages where the old. Their birth years range from 1870 to 1994. Five- weights are the population sizes of the countries. year cohorts were then constructed according to the TABLE B2.5.1 Number of countries with available census data, by region, 1960–2009 WB Region 1960–64 1968–72 1973–77 1978–82 1983–87 1988–92 1992–97 1998–02 2003–07 2008–09 East Asia and Pacific 2 3 6 1 6 2 Eastern Europe and 1 1 4 Central Asia Latin America and 7 5 4 7 2 9 3 10 3 the Caribbean Middle East and 2 3 North Africa South Asia 1 1 2 1 3 1 Sub-Saharan Africa 3 5 2 9 2 2 Total 7 7 6 11 7 21 9 32 9 4 Source: World Bank compilation. (continued) 98 Global Saving in 2030 Global Development Horizons BOX 2.5 (continued) Family size represented in fi gure 2.29 is a variable this polynomial method fi ts two set of points relating directly coded in the IPUMSI censuses. Smoothed lines family size and household heads’ age for two separate connecting the cohort data were plotted by adopting a cohorts (the youngest and the oldest) in an example for polynomial approximation. Figure B2.5.1 shows how República Bolivariana de Venezuela. FIGURE B2.5.1 Family size for the first and latest cohort, points and fitted values, República Bolivariana de Venezuela 7 6 Family size 5 4 3 20 40 60 80 100 Age of household head Source: World Bank calculations using República Bolivariana de Venezuela census data. Transfer Accounts [NTA] project, discussed in ages. Finally, education expenditures tend to be box 2.6) show wide variation across countries in a larger share of GDP in high-income countries the extent to which public transfers rise with age than in developing countries. Th is implies that (figure 2.30). Health care transfers to the elderly even as aging reduces the share of developing tend to be much greater in advanced countries, countries’ school-age populations, higher enroll- not only relative to GDP per capita but also ment rates and more rapid growth in per-student relative to health care spending at other ages (in expenditures than in incomes will limit any per capita terms). This is consistent with find- decline in total education expenditures relative to ings that, at the micro level, health care at very GDP. advanced ages is a luxury good (see, for example, Although the burden of age-related expen- De Nardi, French, and Jones 2010). In the case ditures will be significant in many countries in of public pensions, transfers tend to begin at sub- the years ahead, age-related expenditures will be stantially younger ages in developing countries highly sensitive to the generosity and coverage of than in advanced countries; thus, there may be pension and health care systems (the model gen- significant room for reforming developing coun- erating these scenarios is described in box 2.7). tries’ public pension systems by raising retirement Absent changes in benefits and coverage, four of Global Development Horizons Global Saving in 2030 99 the six developing countries considered in table countries; the resulting increase in the ratio of 2.4—Brazil, Chile, China, and Costa Rica—will public transfers to GDP from 2010 to 2050 in experience a substantial increase in the fiscal bur- the six developing countries ranges from 14 per- den of their old-age expenditures as their popula- centage points in India to 47 percentage points in tions age in future decades. But this calculation China. Even adopting U.S. coverage and benefits may understate the potential burden on public levels would increase the public transfers-to-GDP expenditures because rising incomes could result ratio significantly from current levels, ranging in demands for higher benefits and more wide- from 9 percentage points in Brazil to 19 percent- spread coverage in countries where the public sec- age points in China. tor currently provides only limited pension and These scenarios include projections of public health-care services. health care expenditures. It should be born in Some further insight can be gained by com- mind that speculating on the future of public paring (a) projections of public age-related expen- health care expenditures is subject to more error ditures in individual countries assuming an than predicting pension expenditures because unchanged policy scenario, with (b) projections the costs of public health care depend to a large assuming convergence with the 2003 benefits extent on factors beyond the control of policy and coverage levels offered by a European coun- makers. Prices of medical services and medi- try with a deep benefit system (Sweden) and (c) cines, for instance, are determined at least in part convergence with the somewhat lower benefits by market conditions—international as well as provided in the United States, which has a more domestic—and technological innovations not modest public pension program even than most only affect prices but also make altogether new developing countries but high public health care treatments available. Thus, any projections on expenditures on the elderly. A path of conver- health care spending have a high degree of uncer- gence with Swedish levels of benefits and cov- tainty, even assuming an unchanged policy envi- erage would be untenable for most developing ronment.23 The changing nature of demand for BOX 2.6 The National Transfer Accounts project is a valuable source of data on age-related public transfers Data on age -related public expenditures are typi- Data are currently available for only one year for each cally not disaggregated by age group. However, the country, and the year varies by country. National Transfer Accounts (NTA) Project a aims to Some countries have undergone significant reforms measure public transfer infl ows and outfl ows by one- since the year for which its accounts are available—for year age cohort, and data for at least some series have example, Brazil, in the case of public pension reform— been prepared for 21 countries so far. Most of these which have undoubtedly altered age-related expendi- countries have series on public education, health care, tures since then, both in terms of magnitude and age and pension transfer inflows by age, and among these distribution. Bearing these caveats in mind, the NTA are six developing countries with data on all three: data are nonetheless useful estimates of age-related Brazil, Chile, China, Costa Rica, India, and Mexico. public transfers disaggregated by age. The NTA meth- The project is still under way; more countries will be odology is harmonized so that age distributions of added, and some existing data may yet be revised. transfers are directly comparable across countries. a. The NTA Project is a collaborative effort to measure, analyze, and interpret macroeconomic aspects of age and population aging around the world (http://www.ntaccounts.org/web/nta/show/). Funded by an array of public and private sources, the lead institutions for the project are the Center for the Economics and Demography of Aging at the University of California, Berkeley, and the Population and Health Studies Program of the East-West Center in Honolulu. For details on the project, see Lee and Mason (2011) and http://www.ntaccounts.org. 100 Global Saving in 2030 Global Development Horizons FIGURE 2.30 There is wide variation across countries in the extent to which public transfers tend to rise with age a. Brazil, 1996 b. Sweden, 2003 120 120 Avg. public transfers per person, Avg. public transfers per person, 100 100 % per capita GDP % per capita GDP 80 80 60 60 40 40 20 20 0 0 0 10 20 30 40 50 60 70 80 90+ 0 10 20 30 40 50 60 70 80 90+ Age Age c. China, 2002 d. Japan, 2004 20 120 Avg. public transfers per person, Avg. public transfers per person, 100 15 % per capita GDP % per capita GDP 80 10 60 40 5 20 0 0 0 10 20 30 40 50 60 70 80 90+ 0 10 20 30 40 50 60 70 80 90+ Age Age e. Mexico, 2004 f. United States, 2003 20 120 Avg. public transfers per person, Avg. public transfers per person, 100 15 % per capita GDP % per capita GDP 80 10 60 40 5 20 0 0 0 10 20 30 40 50 60 70 80 90+ 0 10 20 30 40 50 60 70 80 90+ Age Age Education Health care Pensions Source: World Bank calculations using data in the National Transfer Accounts (NTA) project (http://www.ntaccounts.org/web/nta/show/). Note: NTA data are at the one-year age cohort level, in terms of average transfers to a person at that age. They are shown here as percentages of the country’s GDP per capita, in fi ve-cohort rolling averages, for the year for which each country has NTA data available. The vertical axes of the China and Mexico graphs are on a different scale than the graphs of the other four countries. Global Development Horizons Global Saving in 2030 101 BOX 2.7 The impact of aging on government budgets is clarified by breaking it down into its components Fiscal balances are affected by changes in the age per person of each age into average expenditures per distribution of a population mainly because public participant in the program Eage /Page , and the share of health care and pension expenditures are greater for the population of the cohort covered by the program older people, education expenditures are greater for Page /Cage . This highlights that average expenditures younger people, and tax revenue comes mainly from per person in the age group can change because of people in their working years. A simple decomposition either an increase or a decrease in the generosity of of expenditures (equation B2.7.1) common in the liter- the system, or because of diminished or expanded ature (see, for example, Miller, Mason, and Holz 2011), coverage. shows aggregate public expenditures E, on a particu- In principle, all of the above components can be lar program or category of programs (for example, expected to change over time. For example, coun- public pensions), as a fraction of GDP Y, expressed as tries that experience a fall in the ratio of the school- the sum across ages of expenditures relative to GDP age population to the working-age population are often on people of each age. For each age cohort, this is experiencing rising incomes at the same time (which broken down into average expenditures per person is typically associated with having fewer children but Eage /Cage (where Cage is the population of the cohort), also with rising school enrollment rates) as well as relative to output per working-age person Y/W (where increased education spending per student (Becker and W is the working-age population, for example, aged Lewis 1973). Thus, the first two ratios on the right- 20– 64 years), and the ratio of the population of each hand side of the identity rise while the third falls, and age cohort to the working-age population Cage /W. For the sign of the net change in education expenditures as a given Eage /Cage for each age, total expenditures E/Y a fraction of GDP depends on the relative magnitudes increase relative to Y/W as Cage /W rises for ages with of the effects. However, holding expenditures per per- higher than average costs per person Eage /Cage , that is, son constant relative to average income (or relative to as more expensive cohorts grow in size relative to the income per working-age adult) can be a useful thought working-age population. experiment, and in any case there is not always an a The above identity can be further disaggregated priori reason to expect program generosity or coverage (equation B2.7.2), separating the average expenditures to change in one direction or another. ⎡ ⎛ E0 ⎞ ⎤ ⎡ ⎛ E1 ⎞ ⎤ ⎡ ⎛ E 90 + ⎞ ⎤ ⎢⎜ ⎟ ⎥ ⎢⎜ ⎟ ⎥ ⎢⎜ ⎟⎥ E ⎢ ⎝ C 0 ⎠ ⎥ ⎡ C 0 ⎤ ⎢ ⎝ C1 ⎠ ⎥ ⎡ C1 ⎤ ⎢ ⎝ C 90 + ⎠ ⎥ × ⎡C 90 + ⎤ , =⎢ × + × + ... + (B2.7.1) Y ⎛Y ⎞⎥ ⎢ ⎥ ⎣W ⎦ ⎢ ⎛ Y ⎞ ⎥ ⎣W ⎦ ⎢ ⎥ ⎢ ⎛Y ⎞ ⎥ ⎢ ⎣W ⎦ ⎥ ⎢⎜ ⎟ ⎥ ⎢⎜ ⎟ ⎥ ⎢ ⎜ ⎟ ⎥ ⎢ ⎝ W ⎠ ⎥ ⎢ ⎝ W ⎠ ⎥ ⎢ ⎝ W ⎠ ⎥ ⎣ ⎦ ⎣ ⎦ ⎣ ⎦ Avg. expenditures on Zero-year-old each zero-year-old, rel- population, relative ative to avg. output per to working-age working-age person population ⎡ ⎛ E age ⎞ ⎤ ⎢⎜ ⎟⎥ E ⎢ ⎝ Page ⎠ ⎥ ⎡ Page ⎤ ⎡C age ⎤ 90 + (B2.7.2) = ∑ ⎢ ×⎢ ⎥×⎢ ⎥ Y age = 0 ⎢ ⎛ Y ⎞ ⎥ ⎥ ⎢ ⎦ ⎣W ⎦ ⎣C age ⎥ ⎜ ⎟ ⎢ ⎝W ⎠ ⎥ ⎣ ⎦ Generosity: Avg. expenditures Coverage: Share Cohort popula- on each program participant in of the cohort tion, relative to age cohort, relative to avg. out- participating in working-age put per working-age person the program population 102 Global Saving in 2030 Global Development Horizons health services must also be taken into account in are useful—first, for illustrating that U.S. lev- assembling any estimate of future costs. At pres- els are not extraordinarily high for an advanced ent, health care spending for persons near the end country (and therefore that convergence with of their lives is much higher in advanced coun- those levels is not unthinkable for developing tries than in developing countries. If elderly pop- countries over a time horizon of 40 years); and, ulations in developing countries begin to demand second, for illustrating just how greatly outcomes costlier medical services in their final years, may vary if per capita expenditures are driven up the impact of aging on public fi nances in those by rising health care costs because the fiscal effect countries will be much greater than it would be of a large increase in costs would be similar to assuming a static level of services. that of a large increase in generosity. Indeed, the Table 2.4 illustrates that if the generosity of greatest task for governments in preparing for the public health care programs were to gradually impact of aging on public health care liabilities converge with current U.S. levels, for a number may not be to tweak coverage and generosity of of countries the fiscal burden of health care costs conventional systems but to make deep institu- would be roughly double what it would be from tional reforms to health care systems so that ben- aging alone. Sweden, with one of the most lav- eficiaries have stronger incentives to seek low-cost ish public health care systems in the world for its heath care options, reducing contingent liabilities elderly, is an outlier. However, the projections in stemming from unforeseeable shocks to health which generosity converges with Swedish levels care costs. TABLE 2.4 The path of age-related fiscal pressures will depend strongly on benefit and coverage levels Total ( ) Eage Pensions Health care Education change Cage (% GDP) (% GDP) (% GDP) (% GDP) Country (W Y )  2010 2030 2050 2010 2030 2050 2010 2030 2050  2010–50 Brazil constant 9.1 14.0 20.9 3.0 3.5 4.5 2.6 1.9 1.7 12.4 Brazil →Sweden 9.1 10.9 12.8 3.0 5.9 13.2 2.6 4.9 6.9 18.2 Brazil →United States 9.1 14.2 11.0 3.0 4.6 8.8 2.6 3.0 3.6 8.7 Chile constant 5.5 8.8 11.7 2.2 2.6 2.9 2.2 1.7 1.5 6.3 Chile →Sweden 5.5 11.1 18.2 2.2 6.3 15.0 2.2 4.8 7.2 30.5 Chile →United States 5.5 9.5 11.8 2.2 4.7 9.8 2.2 3.0 3.8 15.5 China constant 3.4 5.9 8.1 1.7 2.2 3.0 2.1 1.6 1.5 5.4 China →Sweden 3.4 13.6 33.7 1.7 5.4 14.3 2.1 4.0 6.4 47.2 China →United States 3.4 7.0 12.8 1.7 4.2 9.7 2.1 2.5 3.3 18.6 Costa Rica constant 4.3 7.1 11.0 4.7 5.4 6.8 4.2 3.0 2.6 7.2 Costa Rica →Sweden 4.3 9.8 20.3 4.7 6.8 13.4 4.2 5.4 6.7 27.2 Costa Rica →United States 4.3 7.3 11.1 4.7 5.6 8.9 4.2 3.6 3.5 10.4 India constant 1.0 1.2 1.4 2.3 2.4 2.6 1.7 1.3 1.0 0.1 India →Sweden 1.0 1.6 2.9 2.3 4.4 8.1 1.7 5.7 8.2 14.3 India →United States 1.0 3.6 6.9 2.3 3.4 5.7 1.7 3.3 4.4 12.0 Mexico constant 1.5 2.3 3.2 2.1 2.2 2.6 3.5 2.5 2.0 0.8 Mexico →Sweden 1.5 8.0 22.0 2.1 5.0 11.9 3.5 5.9 7.6 34.4 Mexico →United States 1.5 4.1 9.8 2.1 3.8 8.0 3.5 3.7 4.0 14.8 Sources: World Bank projections, supplemented with calculations using data in the National Transfer Accounts (http://www.ntaccounts.org/web/nta/show/), United Nations Population Division 2011, IMF 2011, and World Bank World Development Indicators database. Note: Figures in the table are estimated using age distribution data on public transfers (from the National Transfer Accounts [NTA]) and projected demographic change. The 2010 pension expenditures are calibrated to public pension expenditure estimates given in IMF (2011). The NTA data are combined with population projections by five-year ( )/( ) age cohort from the United Nations. The methodology underlying the projections is given in online annex 2.2. Eage Y denotes average expenditure on each person of a given age, relative to average output per working-age person; additional details are given in box 2.7. Cage W Global Development Horizons Global Saving in 2030 103 Overall, these simple scenarios suggest not although, from the 2030s on, the latter will be only that the fi scal challenge of rising old-age- a significant factor in China and also in Brazil related costs in the coming decades will be sig- given the enormous scale of Brazil’s current pub- nificant but also that gradual changes in benefits lic pension system and thus the sensitivity of its or coverage can have an impact as great as that fiscal balance to even modest demographic shifts. of demographic change. These scenarios should Some East Asian countries, as well as India, be viewed as illustrating the potential impact of Mexico, Pakistan, and South Africa, do not face aging on fiscal balances, not as forecasts of expen- enormous pressure because they do not have large diture levels. Additionally, it is important to public pension systems. recognize that the evolution of government sav- Increases in pension expenditures as a share of ing patterns in the future will depend on other GDP can imply significant increases in dollar val- changes that have the potential to moderate ues. For example, if Turkey’s GDP growth aver- the fiscal impact of aging: for example, pension ages 4.2 percent from 2010 through 2030 (in line reform, tax reform, labor market reform, changes with a range of projections), then an increase in in labor market participation, and changes in pension expenditures from 6.3 percent of GDP in non-age-related public expenditures. 2010 to 9.6 percent in 2030 would imply annual growth of 6.4 percent—more than tripling expenditures in constant dollars. Countries fac- Aging will increase the cost of ing the prospect of this kind of sustained growth public pensions in pension liabilities will need to consider means When considering a broader sample of countries of reducing coverage and/or benefits or make than that of the National Transfer Accounts proj- large cuts in other areas of expenditures. ect, it is clear that pension expenditures are set to The two most dramatic projections in table 2.5 rise substantially relative to GDP across the devel- are for Brazil and Ukraine, where pension costs oping world (table 2.5).24 Th is scenario assumes would exceed 20 percent of GDP by 2050. Brazil no change in pension coverage or policies, so that made significant reforms in 1999 and 2003, but expenditures per person at each age are held con- more will be necessary soon because the aged stant relative to GDP per working-age person, dependency ratio is set to nearly double by 2030. and changes in expenditures as a share of GDP In 2011, Ukraine enacted overdue but politically are driven purely by the changing age distribution unpopular reforms to overhaul its already quite of the population. Lacking data on the distribu- fiscally burdensome, Soviet-era public pension tion of expenditures across age cohorts for most system, in part by raising retirement ages (this of these countries, an approximation is made by change is not reflected in the scenario). assuming that average pension payments are the For countries with unsustainable pension sys- same for all ages 50 and older, and zero for ages tems, several avenues of reform are available.25 It under 50. Under this assumption, pension expen- may be necessary to reduce payments per recipi- ditures as a share of GDP change proportionally ent, relative to contributions. Policy makers can to a modified dependency ratio (the methodology generally adjust the generosity and coverage of is given in online annex 2.3). pension systems, subject to political constraints; Developing countries vary widely in their in most cases, the promise of future pension pay- exposure to rising pension costs. Countries in ments is implicit—an intergenerational agree- Latin America, Eastern Europe, and the Middle ment that is not legally binding—and is subject East and North Africa (to the extent that a to renegotiation. But significant headway can also generalization can be made based on the Arab be made by the relatively straightforward route of Republic of Egypt and Jordan) generally face sig- raising retirement ages. Th is change would have nificant increases in their pension liabilities in the the added benefit of expanding the labor force, decades ahead. These increases are due mainly to thus adding to economic growth and tax rev- ballooning retirement-age populations and much enues. There is considerable scope for later retire- less to declining working-age population shares, ment in some countries, and pension systems are 104 Global Saving in 2030 Global Development Horizons TABLE 2.5 Population aging will put severe pressure on some countries’ public pension systems Public pension costs Age 65+ Age 20–64 (% of GDP) (% of population) (% of population) 2010 2030 2050 2010 2030 2050 2010 2030 2050 Latin America and the Caribbean (62.8% of population covered) Argentina 7.4 8.8 11.2 10.6 13.6 19.1 56.1 58.2 56.9 Brazil 9.1 13.8 20.1 7.0 13.7 22.5 59.1 61.7 57.5 Chile 5.5 8.1 10.6 9.3 17.3 23.5 59.9 58.8 56.0 Mexico 1.5 2.3 3.2 6.3 11.7 19.9 54.9 59.6 57.7 East Asia and Pacific (90.2% of population covered) China 3.4 5.5 7.7 8.2 16.5 25.6 64.5 63.3 56.3 Indonesia 0.7 1.1 1.6 5.6 10.5 19.2 58.4 62.1 58.4 Malaysia 3.0 4.3 5.5 4.8 10.3 15.0 55.7 57.9 58.2 Philippines 1.7 2.2 3.0 3.6 6.7 10.8 50.5 55.5 58.2 Thailand 1.0 1.6 2.1 8.9 17.6 25.1 63.0 61.6 55.5 Eastern Europe and Central Asia (72.9% of population covered) Bulgaria 7.2 9.2 11.0 17.5 22.3 28.7 63.4 58.1 51.0 Latvia 6.1 7.6 8.9 17.8 21.3 25.7 62.4 57.2 53.6 Lithuania 7.6 9.6 11.3 16.0 20.4 24.2 62.0 57.5 55.2 Romania 7.5 10.0 12.9 14.9 19.2 27.9 64.3 61.1 52.7 Russian Federation 8.1 10.7 12.6 12.8 19.1 23.1 66.4 58.9 54.7 Turkey 6.3 9.6 13.8 6.0 11.4 19.6 58.7 61.8 58.6 Ukraine 17.7 21.8 26.2 15.5 19.5 24.1 64.2 59.2 54.6 South Asia (85.7% of population covered) India 1.0 1.3 1.8 4.9 8.3 13.5 54.7 59.7 60.8 Pakistan 0.6 0.7 1.1 4.3 6.0 10.4 49.1 57.5 61.7 Middle East and North Africa (26.7% of population covered) Egypt, Arab Rep. 4.0 5.2 7.3 5.0 8.7 14.2 53.8 57.9 59.2 Jordan 4.1 5.8 9.3 3.9 5.6 12.8 47.6 59.5 60.7 Sub-Saharan Africa (5.9% of population covered) South Africa 1.9 2.3 2.9 4.6 7.8 10.1 55.3 58.2 61.2 Sources: World Bank projections, supplemented with calculations using data in the IMF 2011 and United Nations Population Division 2011. Note: These projections assume that the ratio of average expenditures per person of retirement age (50+) to GDP per person of working age (20–64) remains constant. Given this simplifying assumption, these projections are less precise than those given in table 2.4, and, for this reason, the paths for countries that are in both sets of projections (Brazil, China, Chile, India, and Mexico) differ somewhat between the two tables. Also, any pension reforms that had not come into effect by 2010 are not accounted for. Details are given in online annex 2.3. not always designed well to encourage labor force for women. With adult mortality rates and fertil- participation by those in their mid-50s to 60s who ity both falling, the average length of retirement are still able to work productively. For example, and the ratio of retirees to the labor force is set even after the last round of reforms, the average to rise dramatically in China, and policies aimed retirement age of those who retire under Brazil’s at later retirement have been among the main Length of Contribution rule26 is 54 for men and reforms being discussed to address this.27 51 for women (Gragnolati et al. 2011). In China, A more fundamental reform is to transition the statutory retirement age is 60 for men and 55 from a pay-as-you-go to a funded system. A Global Development Horizons Global Saving in 2030 105 number of countries are making the change, par- debt, making the fiscal situation still worse in ticularly in Latin America and Eastern Europe. the future. One long-term option is to raise taxes, The transition itself is costly, but the switch to which may carry with it a political cost but will a funded system is a long-term fi x, relieving the probably be a component of any balanced policy government of the imbalance between inflows package to adapt to aging. Second, unnecessary and outf lows that results from demographic and wasteful spending will have to be cut in non- change. It is also noteworthy that the change can age-related areas; the areas and scope for cuts be expected to have a positive impact on total vary by country, but most developing countries national saving, provided that the funding of the have space for non-age-related expenditure cuts system is financed by taxation rather than public that can go part of the way toward freeing fiscal borrowing.28 space for age-related liabilities.30 Th ird, reform- ing the scale and institutional design of public pension and health care systems will be essential. The fiscal burden of aging will Th is process is already under way in a number affect the composition and level of of countries facing the most imminent pressure national saving from aging, but it will need to be accelerated in As shown above, public saving has not shown many countries. any strong trend historically at the global level, The path of age-related public expenditures either up or down. It did rise noticeably for a few has important implications for household saving years leading up to the 2007–09 global financial as well as public saving. Increased generosity and crisis but quickly returned to its normal histori- coverage of public health care and social security cal range of 10–20 percent of global GDP when programs is associated with significant reductions the crisis hit. Even if public saving does resume in household saving,31 all things equal; not only an upward trend in the next few years, the sce- is there less need to smooth consumption over the narios just discussed suggest that, in the medium life cycle, but precautionary saving is also reduced to long run, most countries will face pressure on because public programs play an insurance role, public finances—in some cases, quite severe pres- pooling the risk associated with the uncertainty sure—from the aging of their populations, and of lifespan and health care needs. The effect can this can be expected to constrain public saving. work in reverse as well: China experienced this Furthermore, developing countries may face sig- during the 1990s and 2000s, with urban house- nificantly greater demand for infrastructure as hold saving rates increasing as social safety nets economic growth proceeds (see chapter 1); and if were scaled back.32 public saving levels need to be kept up to support Over the coming decades, developing coun- rising infrastructure investment without running tries will experience some of the typical increases sustained fiscal deficits,29 demographic consider- in demand for public services that have, histori- ations will be still more pressing. cally, accompanied rising incomes. But at the Looking forward, gross public saving will same time, governments’ ability to ramp up likely stay in the range that it has been in for the pension and health care programs will be con- past several decades, constituting roughly 10–20 strained by aging to a much greater extent than percent of gross national saving. This is required high-income countries experienced when they for the public sector to play a significant role in were going through a similar stage of economic any sustained investment in infrastructure. To development. If governments do not follow the maintain this level as age-related expenditures path that high-income countries did in the past, increase, policy makers will have a few options. and instead reduce the generosity or coverage of In the short to medium term, they may run fis- age-related programs in the face of a swelling cal deficits, but this solution is not sustainable number of beneficiaries, then private saving can and only delays inevitable costly adjustments to be expected to rise because a greater proportion of long-term demographic shifts; indeed, running retirement and health care costs will be paid for deficits will only burden the government with out of pocket. Increased generosity and coverage 106 Global Saving in 2030 Global Development Horizons of education systems should be feasible because choices includes lower standards of living for the fiscal impact will be offset by the decreasing retirees, longer working lives, higher saving dur- number of children as a share of the population. ing the working life, and larger transfers from Pensions and health care are a different story, workers to retirees. and there will likely be an important role here Most developed countries have opted for for markets, and private saving, alongside public (or are considering) a mix of a delayed retire- programs. Higher tax rates will also affect house- ment age, increased private saving for retirement hold saving, so even if public saving does not fall, (through favorable tax regimes), and reformed private saving can be expected to fall if tax hikes public pension systems—while most have constitute a large part of the policy response, and, rejected the option of declining living standards in this case, total national saving will fall. for the elderly (see Pensions Commission 2005 for the United Kingdom). Depending on its position in the demographic Conclusion and policy directions transition, a specific developing country will The downward pressure on saving rates associated more likely be successful in applying one or a with the aging of the global population through mix of these options. For example, countries at 2030 is unlikely to become a severe constraint early stages of their demographic transition, such on growth overall. However, some crucial policy as the Sub-Saharan countries, could probably challenges lie ahead, especially for those coun- set up tax incentives to encourage saving. A key tries affected by more rapid aging and with less issue is assessing whether tax-deferred individual potential for productivity catch-up. Th is chap- retirement saving crowds out other forms of sav- ter outlines two of the main challenges: first, ing. The literature on the United States finds little the potential pressure of old-age-related public evidence that 401(k) contributions substitute for expenditures on government finance, and sec- other forms of personal saving (Poterba, Venti, ond, the concentration of saving, in terms of both and Wise 1996; Benjamin 2003; Gelber 2011). amounts and rates, within a restricted segment of For other regions, longer working lives and the population. larger transfers may be a better solution. As Looming policy challenges with respect to described in the chapter, there are some strong government fi nances include uncertainty about rationales for transfers to elderly households: (a) the rate at which old-age-related expenditure will redistribution (productivity growth will raise increase and the lack of policy frameworks to deal incomes of future cohorts, and public programs with this long-term issue. As Heller (2003) clearly are replacing family transfers); and (b) correction put it: “While budgets increasingly encompass a of market failures such as limited insurance mar- medium-term framework of three to five years, kets against the risk of long-term care needs. few countries provide long-term scenarios. . . . For developing countries with an increasingly Most countries’ sustainability analyses focus on large share of old people, an often overlooked their ability to service current debt and antici- policy challenge consists of recognizing the het- pated future deficits. . . . Projections of broader erogeneity among the elderly. This is not a homo- fiscal aggregates lack credibility, because they are geneous group, but it varies in terms of financial based on unchanged policy assumptions, current disparities, health status, capacity to work, and life laws, or constant shares of revenues and expendi- expectancy. For example, recent research on labor ture in output. Budget processes also lack mecha- market conditions at older age (Gruber and Wise nisms to foster debate on policy commitments 2007) has found that higher tax on earnings from or guarantees whose fiscal consequences emerge ages 55 to 69 is positively correlated with unused only over the long term.” (emphasis added) labor capacity (workers aged 55 to 65). A fundamental policy choice with respect Regarding the second issue—the concen- to the allocation of the burden of demographic tration of saving within restricted population change is about its distribution between future segments—this chapter highlighted that the retirees and future workers. The menu of policy within-countries high concentration of saving Global Development Horizons Global Saving in 2030 107 and wealth may create inequality traps that have 4. There are, however, some empirical findings negative implications for economic mobility and that are inconsistent with the life-cycle model: thus for the political and social consensus essen- for example, preference for default saving plan tial for growth. In fact, the recent literature has options, demand for commitment devices that tie emphasized other, more direct channels through one’s hands in making saving decisions, and evi- dence of habit formation in saving behavior (Atta- which inequality may affect productivity and nasio and Weber 2010). growth. A main lesson from this body of work 5. Detailed results are reported in online annex 1.5, (see World Bank 2005) is that in the highly and for a technical description of how the econo- imperfect markets of poor countries, individu- metric estimates are used in the CGE model, see als without enough wealth or social status would online annex 1.6; all annexes are available at tend to underinvest or use their resources for http://www.worldbank.org/CapitalForTheFuture. some less productive purpose, reducing economy- 6. “Conditional” in the sense that the relationship wide productivity. And the related broad policy between saving rate and per capita income growth advice has been to reduce the unequal access takes into account the effects from other deter- to finance. A simple initial step could include minants—in this specific case: lagged saving, per improving financial literacy: many households capita GDP, financial development, aged depen- lack adequate knowledge and decision-making dency ratio, social protection (and other controls as shown in specification S4 in online annex tools when it comes to credit markets. An inter- 2.5)—whose contributions are held constant. esting example is offered by a policy intervention 7. The UN’s high scenario estimates that the world in Mexico consisting of an information cam- population will reach 36.4 billion by 2300, and paign on how to use the banking system to chan- the low scenario estimates 2.3 billion. nel remittances toward investment rather than 8. The global demographic transition is assumed to immediate consumption. last three to four centuries. It started in Europe The chapter has also shown that consistently around 1700, and it will probably be complete for the least-educated groups have low or no savings, the whole world in 2100. Figure 2.12 thus depicts suggesting an inability to improve their earning only a “snapshot” of about 60 years of this three- capacity and, for the poorest, to escape a poverty or four-century process. trap. Leveling the playing field in terms of edu- 9. Note that the heterogeneity highlighted in figure cation opportunities would thus have the added 2.13 still masks the full heterogeneity because countries within each group differ in their demo- benefit of boosting private saving. graphic transition. In the case of high-income countries, there are sharp differences between Europe and the United States. In the former, fer- Notes tility rates are quite below the replacement rate 1. Note that for maximizing country coverage for and net migration has been slow. comparison across time, the figures in this para- 10. The dominance of the financial development graph represent ratios of saving to gross domestic effect can also be verified by separately running product (GDP). Following standard practice, the the scenario with only the change in productivity rest of the chapter uses ratios of saving to gross or only the change in financial development (not national income (GNI). reported, but available on request). Performing 2. Undoubtedly, the composition of countries is not such a decomposition makes it clear that the effect static over time, and the data reported generally of a reduction in the saving rate in developing use the maximum available sample coverage. How- countries is, indeed, due primarily to more rapid ever, these variations in the sample over time do financial development. not alter the main qualitative messages in the text. 11. Economists have been interested in the relation- 3. As discussed more extensively in relation to the ship between economic inequality and saving since scenarios of saving in the coming two decades, the original observation of Kaldor (1957) that the financial deepening is associated with lower rates rich should have a higher propensity to save than of saving, which may also help explain why saving the poor, and thus higher inequality should have rates have tended to stay low in Latin American a positive effect on saving. In fact, various post- countries. Keynesian models (Lewis 1954; Pasinetti 1962) 108 Global Saving in 2030 Global Development Horizons stress the link between the functional distribution earnings in the economy and no economywide of income and saving. These models conclude that net savings. However, if the population is expand- redistribution (for example, via differential tax ing, the total amount of saving by the young will rates) toward profits from wages should increase be larger than the total amount of dissaving by saving because capitalists tend to have higher the elderly. Similarly, even if the population shares propensities to save than workers. More recent of savers and dissavers are constant, growth in research has focused on the personal distribution income (per capita) implies that the working-age of income, but empirical analyses have not yet population has a higher lifetime income and that found a significant relationship between income the economy will accumulate more wealth than inequality and saving (see Leigh and Posso 2009 that held by the elderly. Thus, the saving rate rises for a recent example). as population or income (or both) grow. 12. These micro decisions can sometimes be aggre- 15. The fact that the lines representing these educa- gated to create a representative household within tion proportions for the various cohorts tend to an economy. In these cases, an aggregate “macro” be quite flat means that as cohorts proceed along approach and a “micro” approach provide the same their life cycles, their composition in terms of edu- type of results and could be used interchangeably. cation does not change much. But this is not always possible, and, for the many 16. Note that this long-term trend is captured by the cases where there is an “aggregation problem,” cohort effect given the estimation restrictions these two approaches provide different and com- imposed on the time dummies, as these sum to plementary insights. This problem exists whenever zero and are orthogonal to a time trend and thus the aggregate agents’ behavior, such as aggregate they pick up deviations from a long stable trend. private demand, cannot be “treated as if it were 17. Income is more likely to be underestimated for the outcome of the decision of a single maximizing poor households because of greater informal and consumer” (Deaton and Muellbauer 1980, 148). in-kind components, thus biasing imputed saving The conditions for aggregation are quite severe and downward for these households. Underestimation break down easily in the presence of discontinui- could also arise from temporary fluctuations in ties and nonlinearities. Consider, for example, the income; temporary low-income levels in a cross- case of participation in the labor market. Theory section would correspond to negative saving rates. says that participation should increase with the 18. For Ghana, Mexico, Russia, and Thailand as a wage rate. At the micro level, this can be obtained group, the share of households with a head with as increasing hours for an already employed indi- tertiary education has risen from 6–10 percent of vidual or as the entrance in the market of some- the total in the 1980s or early 1990s to 20–25 per- one who was not working. The effects, in terms cent in the late 2000s. of income distribution, may be very different in 19. This is consistent with findings in the literature. these two cases. However, at the macro level, it is For example, López-Calva and Lustig (2010) show impossible to model these two alternatives with an that a large part of the recent reduction of inequal- aggregate representative worker (even if the num- ity in Latin America is due to the entry of a greater ber of representative workers is greatly increased). number of educated people into the labor market And comparable cases, where aggregation condi- and the related reduction of skill premia in wages. tions do not hold, arise in almost all markets with 20. The vertical distance between cohorts—in this case, imperfection and asymmetric information: goods, the difference in household sizes for two cohorts at credit, education, and others. the same age—cannot always be interpreted as a 13. Note that this simulated growth rate approximates pure cohort effect. Because cohorts are observed at the growth rate of the rapid convergence scenario the same age but in different time periods, the verti- of the global general equilibrium model described cal distance may be due to a year effect. However, in the macroeconomic projection section. it may be difficult to justify that year effects are 14. In the simplest version of the life-cycle model, important for a variable such as household size (or consumption-smoothing individuals save during number of children). Therefore, in this specific case, their working years to purchase assets that they the vertical distances can be assumed to represent will sell later in life to finance consumption (in the cohort effect. other words, they dissave later in life). Without 21. The general equilibrium model includes govern- income or population growth, the model predicts ment saving; however, consistent with its long-run that there will be a constant ratio of wealth to nature, it assumes that each government’s deficit Global Development Horizons Global Saving in 2030 109 (or surplus) remains small and fixed throughout 30. For an example of methodology for estimating the projection period. Any change in public expen- the scope for cutting non-age-related expenditures diture is compensated by an immediate counter- (applied to European countries), see Heller and balancing change in direct taxes. In other words, Hauner (2006). households’ disposable incomes are affected directly 31. Baldacci et al. (2010) finds that for OECD coun- by changes in public expenditures, and this affects tries from 1990 to 2008, an increase in social their saving decisions. security spending of 1 percent of GDP reduced 22. For a recent set of pension projections for both household saving by 0.22 to 0.29 percent of GDP, developing and advanced economies, see IMF and an increase in public health spending of 1 per- (2011). For a detailed survey of official projections cent of GDP reduced household saving by 0.70 to of age-related expenditures for Europe, see EC 0.78 percent of GDP (evaluated at the respective (2006). For a concise survey of projections and sample means). Chou, Liu, and Hammitt (2006) literature on the health care component of future find a qualitatively similar impact of public health public expenditures, see Gray (2005). care provision in Taiwan, exploiting the natural 23. Heller and Hauner (2006) discuss reform of fiscal experiment of an expansion of public health insur- institutions to address the problem of uncertainty ance coverage. of long-term paths of public liabilities. 32. Although there are many factors behind high sav- 24. In the case of pensions, paths of public expendi- ing in China, Chamon and Prasad (2010) identify tures can be reasonably approximated without the rising private burden of education and health detailed cohort-level data because transfers come care expenditures, and increased precautionary almost entirely from one age group and go almost saving, as important drivers of a 7 percentage entirely to another. Thus, the projections given in point increase in the urban household saving rate table 2.5 are not limited to countries with NTA between 1995 and 2005. data. Details are given in online annex 2.3. 25. For a primer on public pension reform, see Jousten (2007). References 26. Under Brazil’s Length of Contribution rule, men Abel, A. 2003. “The Effects of a Baby Boom on Stock are eligible for a pension after making contribu- Prices and Capital Accumulation in the Presence tions for 35 years (30 for teachers), and women are of Social Security.” Econometrica 71 (2): 551–578. eligible after making contributions for 30 years ACMR All China Data Center database, All China (25 for teachers). Marketing Research Co., University of Michigan, 27. It is worth noting that an increase in retirement Ann Arbor. http://chinadataonline.org/. age tends to reduce household saving because life- Aglietta, M., J. Chateau, J. Fayolle, M. Juillard, J. L. time income rises and expected length of retire- Cacheux, G. L. Garrec, and V. Touzé. 2007. “Pen- ment falls (Barrell, Hurst, and Kirby 2009). Also, sion Reforms in Europe: An Investigation with a the increased expected future size of the labor Computable OLG World Model.” Economic Model- force should tend to raise investment demand. ling 24 (3): 481–505. 28. Samwick (2000) estimates that a country with a Aiyagari, S. Rao. 1994. “Uninsured Idiosyncratic Risk provident fund or fully funded system tends to and Aggregate Saving.” Quarterly Journal of Eco- have an average saving rate that is 3.68 percentage nomics 109 (3): 659–84. points greater than a country with a pay-as-you-go Altig, David, Alan J. Auerbach, Laurence J. Kotlikoff, system, all things equal. Part of the explanation is Kent A. Smetters, and Jan Walliser. 2001. “Simulat- that intergenerational Ricardian equivalence does ing Fundamental Tax Reform in the United States.” not hold; Loayza, Schmidt-Hebbel, and Servén American Economic Review 91 (3): 574–95. (2000) find that Ricardian equivalence does not Attanasio, Orazio P. 1998. “Cohort Analysis of Saving hold empirically, even in the long run. Behavior by U.S. Households.” Journal of Human 29. Gross public saving is linked with public invest- Resources 33 (3): 575–609. ment; because gross public saving is defined as Attanasio, Orazio P., and Hilary Williamson Hoynes. current revenues minus current expenditures, 2000. “Differential Mortality and Wealth Accu- and public investment is not included in current mulation.” The Journal of Human Resources 35 (1): expenditures, for a given fiscal balance (total rev- 1–29.] enues minus total expenditures) gross public sav- Attanasio, O., S. Kitao, and G. Violante. 2006. “Quan- ing must rise 1-for-1 with public investment. tifying the Effects of the Demographic Transition 110 Global Saving in 2030 Global Development Horizons in Developing Economies.” The B.E. Journal of Carroll, Christopher D. 1997. “Buffer-Stock Saving Macroeconomics advances.6 (1): 1–44. and the Life Cycle/Permanent Income Hypothesis.” ———. 2007. “Global Demographic Trends and Social Quarterly Journal of Economics 112 (1): 1–55. Security Reform.” Journal of Monetary Economics 54 Chamon, Marcos D., and Eswar S. Prasad. 2010. “Why (1): 144–98. Are Saving Rates of Urban Households in China Attanasio, Orazio P., and Miguel Szekely. 2003. The Rising?” American Economic Journal: Macroeconom- Family in Flux: Household Decision-Making in Latin ics 2 (1): 93–130. America. Washington, DC: Inter-American Devel- Chou, Shin-Yi, Jin-Tan Liu, and James Hammitt. 2006. opment Bank. “Households’ Precautionary Behaviors: The Effects Attanasio, Orazio P., and Guglielmo Weber. 2010. of the Introduction of National Health Insurance “Consumption and Saving: Models of Intertem- in Taiwan.” Review of Economics of the Household 4 poral Allocation and Their Implications for Pub- (4): 395–421. lic Policy.” Journal of Economic Literature 48 (3): Conesa, Juan C., and Dirk Krueger. 1999. “Social 693–751. Security Reform with Heterogeneous Agents.” Auerbach, Alan J., and Laurence J. Kotlikoff. 1987. Review of Economic Dynamics 2 (4): 757–95. “Evaluating Fiscal Policy with a Dynamic Simu- ———. 2006. “On the Optimal Progressivity of the lation Model.” American Economic Review 77 (2): Income Tax Code.” Journal of Monetary Economics 49–55. 53 (7): 1425–50. Baldacci, Emanuele, David Coady, Giovanni Callegari, CSO National Accounts database, Central Statisti- Ding Ding, Jaejoon Woo, Pietro Tommasino, and cal Organisation, Ministry of Statistics and Pro- Manmohan S. Kumar. 2010. “Public Expenditures gramme Implementation, New Delhi. http://mospi. on Social Programs and Household Consumption nic.in/Mospi_New/site/home.aspx. in China.” Working Paper WP/10/69, International Curtis, Chadwick C., Steven Lugauer, and Nelson C. Monetary Fund, Washington, DC. Mark. 2011. “Demographic Patterns and House- Barrell, Ray, Ian Hurst, and Simon Kirby. 2009. “How hold Saving in China.” Working Paper 16828, to Pay for the Crisis or Macroeconomic Impli- National Bureau of Economic Research, Cam- cations of Pension Reform.” Discussion Paper bridge, MA. 333, National Institute of Economic and Social Davies, James B., Susanna Sandström, Anthony Shor- Research, London. rocks, and Edward N. Wolff. 2011. “The Level and Beck, Thorsten, Asli Demirgüç-Kunt, and Ross Levine. Distribution of Household Wealth.” Economic Jour- 2010. “Financial Institutions and Markets across nal 121 (551): 223–54. Countries and over Time: The Updated Financial Deaton, Angus S. 1991. “Saving and Liquidity Con- Development and Structure Database.” World Bank straints.” Econometrica 59 (5): 1221–48. Economic Review 24 (1): 77–92. Deaton, Angus, and John Muellbauer. 1980. Econom- Becker, G. S. (1981) 1993. A Treatise on the Family. ics and Consumer Behavior. Cambridge: Cambridge Cambridge, MA: Harvard University Press. University Press. Becker, Gary, and H. Gregg Lewis. 1973. “On the Deaton, Angus, and Christina Paxson. 1994. “Inter- Interaction between the Quantity and Quality of temporal Choice and Inequality.” Journal of Politi- Children.” The Journal of Political Economy 81 (2): cal Economy 102 (3): 437–67. S279–88. ———. 2000. “Growth and Savings among Individu- Benjamin, Daniel J. 2003. “Does 401 (k) Eligibility als and Households.” The Review of Economics and Increase Saving? Evidence from Propensity Score Statistics 82 (2): 212–25. Subclassification.” Journal of Public Economics 87 Demirgüç-Kunt, Asli, and Leora F. Klapper. 2012. (5 –6): 1259–90. “Measuring Financial Inclusion: The Global Findex Bloom, David E., David Canning, and Bryan Graham. Database.” Policy Research Working Paper 6025, 2003. “Longevity and Life-Cycle Savings.” Scandi- World Bank, Washington, DC. navian Journal of Economics 105 (3): 319–38. De Nardi, Mariacristina, Eric French, and John B. Börsch-Supan, A., A. Ludwig, and J. Winter. 2006. Jones. 2010. “Why Do the Elderly Save? The Role of “Ageing, Pension Reform, and Capital Flows: A Medical Expenses.” The Journal of Political Economy Multi-Country Simulation Model.” Economica 73 118 (1): 39–75. (292): 625–58. De Nardi, Mariacristina, Selahattin İmrohoroğlu, and Bourguignon, François, and Christian Morrisson. Thomas J. Sargent. 1999. “Projected U.S. Demo- 2002. “Inequality Among World Citizens: 1820– graphics and Social Security.” Review of Economic 1992.” American Economic Review 92 (4): 727–44. Dynamics 2 (3): 575–615. Global Development Horizons Global Saving in 2030 111 Domeij, David, and Martin Flodén. 2006. “Population Heller, Peter, S. 2003. “Who Will Pay?” Finance & Aging and International Capital Flows.” Interna- Development, September. tional Economic Review 47 (3): 1013–32. Heller, Peter S., and David Hauner. 2006. “Fiscal Policy EC (European Commission). 2006. “European Econ- in the Face of Long-Term Expenditure Uncertain- omy: The 2005 Projections of Age-Related Expen- ties.” International Tax Public Finance 13 (4): 325–50. diture (2004–50) for the EU-25 Member States: Henriksen, Espen R. 2002. “A Demographic Explana- Underlying Assumptions and Projection Meth- tion of U.S. and Japanese Current Account Behav- odologies.” Special Report No. 4, Directorate- ior.” Unpublished manuscript, Carnegie Mellon General for Economic and Financial Affairs, EC, University, Pittsburgh. Luxembourg. Huang, He, Selahattin İmrohoroğlu, and Thomas J. Fehr, H., S. Jokisch, and L. Kotlikoff. 2003. “The Sargent. 1997. “Two Computations to Fund Social Developed World’s Demographic Transition: The Security.” Macroeconomic Dynamics 1 (1): 7–44. Roles of Capital Flows, Immigration, and Policy.” Hubbard, R. Glenn, Jonathan S. Skinner, and Stephen Working Paper 10096, National Bureau of Eco- P. Zeldes. 1995. “Precautionary Saving and Social nomic Research, Cambridge, MA. Insurance.” Journal of Political Economy 103 (2): Feroli, Michael. 2003. “Capital Flows Among the G-7 360–99. Nations: A Demographic Perspective.” Finance IMF (International Monetary Fund). 2011. “The Chal- and Economics Discussion Paper 2003-54, Board lenge of Public Pension Reform in Advanced and of Governors of the [U.S.] Federal Reserve System, Emerging Economies.” Fiscal Aff airs Department Washington, DC. report, IMF, Washington, DC. Friedman, Milton. 1957. A Theory of the Consump- IMF World Economic Outlook database, International tion Function. Princeton, NJ: Princeton University Monetary Fund, Washington, DC. http://www.imf Press. .org/external/data.htm. Gelber, Alexander M. 2011. “How Do 401(k)s Affect Integrated Public Use Microdata Series (IPUMSI) Saving? Evidence from Changes in 401(k) Eligibil- database, University of Minnesota, Minneapolis. ity.” American Economic Journal: Economic Policy 3 https://international.ipums.org/international/. (4): 103–22. Jappelli, Tullio, and Marco Pagano. 1994. “Saving, Gokhale, Jagadeesh, Laurence J. Kotlikoff, and John Growth, and Liquidity Constraints.” Quarterly Sabelhaus. 1996. “Understanding the Postwar Journal of Economics 109 (1): 83–109. Decline in United States Saving: A Cohort Analy- Jousten, Alain. 2007. “Public Pension Reform: A sis.” The Brookings Papers on Economic Activity, Primer.” Working Paper WP/07/28, International The Brookings Institution, Washington, DC. Monetary Fund, Washington, DC. Gragnolati, Michele, Ole Hagen Jorgensen, Romera Kaldor, N. 1957. “A Model of Economic Growth.” Eco- Rocha, and Anna Fruttero. 2011. Growing Old in nomic Journal 67 (268): 591–624. an Older Brazil: Implications of Population Aging on Kinugasa, Tomoko, and Andrew Mason. 2007. “Why Growth, Poverty, Public Finance, and Service Deliv- Countries Become Wealthy: The Effects of Adult ery. Directions in Development series. Washington, Longevity on Saving.” World Development 35 (1): DC: World Bank. 1–23. Gray, Alistair. 2005. “Population Ageing and Health Krueger, Dirk, and Alexander Ludwig. 2007. “On the Care Expenditure.” Ageing Horizons 2005 (2): Consequences of Demographic Change for Rates of 15–20. Returns to Capital, and the Distribution of Wealth Gruber, Jonathan, and David A. Wise. 2007. “Intro- and Welfare.” Journal of Monetary Economics 54 (1): duction to Social Security Programs and Retirement 49–87. around the World: Fiscal Implications of Reform.” Lane, Philip R., and Gian Maria Milesi-Ferretti. 2007. In Social Security Programs and Retirement around “The External Wealth of Nations Mark II: Revised the World: Fiscal Implications of Reform: National and Extended Estimates of Foreign Assets and Lia- Bureau of Economic Research Conference Report , bilities, 1970–2004.” Journal of International Eco- 1–42. Chicago: University of Chicago Press. nomics 73 (2): 223–50. Heckman, James J., and Richard Robb. 1987. “Using Lee, Ronald, and Andrew Mason, eds. 2011. Popula- Longitudinal Data to Estimate Age, Period, and tion Aging and the Generational Economy: A Global Cohort Effects in Earnings Equations.” In Cohort Perspective. Cheltenham, UK: Edward Elgar. Analysis in Social Research: Beyond the Identification Leigh, Andrew, and Alberto Posso, 2009. “Top Incomes Problem, ed. William M. Mason and Stephen E. and National Savings.” Review of Income and Wealth Fienberg, 137–50. New York: Springer-Verlag. 55 (1): 57–74. 112 Global Saving in 2030 Global Development Horizons Lewis, W. A. 1954. “Economic Development with Poterba, James M., Steven F. Venti, and David A. Wise. Unlimited Supplies of Labour.”Manchester School of 1996. “The Effects of Special Saving Programs on Economic and Social Studies 22: 139–191. Saving and Wealth.” In The Economic Eff ects of Aging Loayza, Norman V., Klaus Schmidt-Hebbel, and Luis in the United States and Japan, 217–40. Cambridge, Servén. 2000. “What Drives Private Saving Across MA: National Bureau of Economic Research. the World?” Review of Economics and Statistics 82 Riddick, Leigh A., and Toni M. Whited. 2009. “The (2): 165–81. Corporate Propensity to Save.” Journal of Finance López-Calva, Luis F., and Nora Lustig, eds. 2010. The 64 (4): 1729–66. Decline in Inequality in Latin America: A Decade of Samwick, Andrew A. 2000. “Is Pension Reform Con- Progress? Baltimore: Brookings Institution Press and ducive to Higher Saving?” The Review of Economics the United Nations Development Programme. and Statistics 82 (2): 264–72. Ma, Guonan, and Wang Yi. 2010. “China’s High Shorrocks, Anthony, James B. Davies, and Rodrigo Saving Rate: Myth and Reality.” Working Paper Lluberas. 2010. Credit Suisse Global Wealth Report. 312, Bank for International Settlements, Basel, Zurich: Credit Suisse Research Institute. Switzerland. UN National Accounts Main Aggregates database, MaCurdy, Thomas E., and Thomas Mroz. 1995. United Nations, New York. http://unstats.un.org/ “Measuring Macroeconomic Shifts in Wages from unsd/snaama/introduction.asp. Cohort Specifications.” Unpublished manuscript, United Nations Population Division. 2011. World Pop- Stanford University, Stanford, CA; and University ulation Prospects: The 2010 Revision. New York: UN. of North Carolina, Chapel Hill. Wei, Shang-Jin, and Xiaobo Zhang. 2011. “The Com- Marchiori, Luca. 2011. “Demographic Trends and petitive Saving Motive: Evidence from Rising Sex International Capital Flows in an Integrated Ratios and Savings Rates in China.” Journal of World.” Economic Modelling 28 (5): 2100–120. Political Economy 119 (3): 511–64. Milanovic, Branko. 2005. Worlds Apart: Measuring Weil, Philippe. 1993. “Precautionary Savings and the Global and International Inequality. Princeton, NJ: Permanent Income Hypothesis.” Review of Eco- Princeton University Press. nomic Studies 60 (2): 367–83. Miller, Tim, Carl Mason, and Mauricio Holz. 2011. World Bank. 2005. World Development Report 2006: “The Fiscal Impact of Demographic Change in Ten Equity and Development. World Bank, Washington Latin American Countries: Education, Health, and DC. Pensions.” In Population Aging: Is Latin America World Bank Financial Inclusion database, World Bank, Ready? ed. Daniel Cotlear, 233–72. Washington, Washington, DC. DC: World Bank. World Bank Financial Structure and Development Modigliani, Franco, and Richard Brumberg. 1954. database, World Bank, Washington, DC. “Utility Analysis and the Consumption Function: World Bank Global Economic Monitor database, An Interpretation of Cross-Section Data.” In Post- World Bank, Washington, DC. Keynesian Economics, ed. Kenneth K. Kurihara, World Bank Global Income Distribution Dynamics 383–436. New Brunswick, NJ: Rutgers University database, World Bank, Washington, DC. http:// Press. econ.worldbank.org/ W BSITE/E XTER NA L/ Pasinetti, Luigi L. 1962. “Rate of Profit and Income EXTDEC/EXTDECPROSPECTS/0,,contentMD Distribution in Relation to the Rate of Economic K:21909753~menuPK:5963193~pagePK:64165401 Growth.” Review of Economic Studies 29 (4): 267–79. ~piPK:64165026~theSitePK:476883,00.html. Paxson, Christina. 1996. “Saving and Growth: Evidence World Bank World Development Indicators database, from Micro Data.” European Economic Review 40 (2): World Bank, Washington, DC. http://data.world 255–88. ba n k .org /d at a-c at a log /world-de velopment- Pensions Commission. 2005. A New Pension Settlement indicators. for the Twenty-First Century. London: The Statio- nery Office. 3 Capital Flows in the Third Age of Financial Globalization C OMING CH A NGES IN THE developing economies will become much more patterns of investment and saving, as deeply integrated into the fabric of global finance described in the first two chapters of this and will account for a growing share of the world’s report, raise several imperative questions. Should a gross capital inflows and outflows. Th is age fol- shortage of saving be anticipated at the global level, lows two previous eras: the first era was domi- which would push up yields on investment? Or, on nated by Great Britain through the 1930s, and the the other hand, will the future bring a global excess second, which began after World War II and now of saving and a lack of investment opportunities appears to be concluding, has been dominated by that can return current yields? Similar questions the United States (box 3.1). can be asked at the country level, but the analysis Relative to their contribution to global output is more complicated. In a specific country, future and trade, emerging and developing economies domestic saving might be sufficient to finance all historically have accounted for a much smaller investment opportunities at current yields, but this share of activity in global finance. In recent years, domestic saving might also be invested abroad, and developing countries have accounted for 30 per- foreign saving may flow into the country as well. cent of world production and trade and more What are the resulting net capital flows, and what than half of global growth but only 22 percent are the pressures on yields in individual countries? of the world’s capital market capitalization of This chapter not only addresses all of these perti- $49.5 trillion, merely 8 percent of gross capital nent questions, but also goes beyond them to cover flows, and 9.6 percent of global external assets. what may be the most dominant trend in interna- Of the 25 banks that set the London interbank tional finance in the coming decades: the increase offered rate (LIBOR) based on their costs of bor- in gross capital flows and the role of developing rowing in one or more of the 10 currencies cov- countries in that process. ered by the measure, not a single bank is based What the analysis in this chapter finds is that in a developing country. Even after the surge developing countries will become more impor- of cross-border capital flows in the 1990s and tant as both sources and destinations of capital 2000s, global investors’ exposure to developing through 2030, and for the first time in history countries remains limited relative to their expo- they seem poised to play a significant role in global sure to developed countries. No emerging-market intermediation of capital. In tracing the potential currency is used internationally to a significant path of capital flows across borders through 2030, extent. But all these statistics may change consid- this chapter focuses on the interplay of three erably through 2030. key long-term structural forces: economic con- The defining aspect of the Third Age of vergence, demographic transition, and fi nancial Financial Globalization will be a change in the globalization. Policy choices at the national and dynamics of global capital f lows. Because the international levels, of course, will affect the speed relevance of capital flows extends well beyond and intensity of these forces on global capital the balance of a country’s borrowing or lend- flows, chief among them the management of what ing captured by net inf lows or outf lows, the Global Development Horizons (GDH ) identifies as approach taken in this chapter is to examine the the Th ird Age of Financial Globalization, when future paths of both gross flows and net flows. Global Development Horizons 115 116 Capital Flows in the Third Age of Financial Globalization Global Development Horizons As developing countries’ volumes of gross inflows world. But the potential risks involved—par- and outflows expand in the future, the potential ticularly with developing countries’ absorption benefits to these countries are significant: the of large amounts of capital inflows—cannot be diversification of idiosyncratic national risks, the overlooked. Exchange rate appreciation (and thus imposition of greater market discipline on policy decreased trade competitiveness), the formation of making, the opportunity to supplement domes- asset bubbles, and economic overheating have all tic saving in ramping up fixed investment and been problematic in the past. A potential reversal growth, and the promotion of financial sector in capital inflows can also have detrimental effects. competitiveness. Designing policy frameworks at the national and Realization of these outcomes could be partic- international levels that take into consideration all ularly valuable for developing countries with rela- of these potential benefits and risks will become tively young populations and significant potential more urgent as the amount of capital moving for economic convergence with the developed across borders rises in the decades ahead. BOX 3.1 The Third Age of Financial Globalization follows two previous eras During the First Age of Financial Globalization, starting Eichengreen 2003). Obligations under the Organisation in the second half of the 19th century, large amounts for Economic Co-operation and Development’s Code of capital were directed from European countries to of Liberalization were broadened to include virtually all the New World, mostly for investment in railways, real capital movements, including short-term transactions estate, and large-scale agricultural projects (Obstfeld by enterprises and individuals. Rapid globalization in the and Taylor 2003). By the start of World War I in 1914, financial industry in the 1990s and 2000s brought even more than one-quarter of British wealth was invested more dramatic change to the landscape of the global outside of Great Britain, mainly in foreign government financial system, not only encouraging steep increases securities and railroads (Gilpin 1987). In 1913, almost in cross-border capital fl ows as money market instru- half of Argentine and one-fifth of Australian capi- ments, forwards, swaps, and other derivatives were tal stock was owned by foreign investors in Europe created, but also allowing developing countries to be (Taylor 1992). This age wound down as European integrated into the global financial system in earnest. countries dramatically reversed their nondefense At present, the world appears to be in a transition capital outfl ows during World War I, eventually giving into a Third Age of Financial Globalization. The begin- way to the Bretton Woods arrangements shortly after nings of this shift would likely have occurred in the World War II. early 2000s as developing countries became more As noted by several observers of international mon- integrated into the global financial system and capital etary affairs, the Bretton Woods Agreement gave pri- infl ows to them became signifi cant in absolute terms ority to multilateral trade expansion and investment at for the fi rst time. The trend became more noticeable the expense of financial liberalization. To the extent that during the global fi nancial crisis, when gross infl ows private capital moved across borders in the post–World of capital to developing countries declined much War II era, it was confined to the foreign currency busi- less than infl ows to advanced countries, after having ness of major banks, largely in the form of trade financ- increased more rapidly on an annual basis over 2003– ing and payment services. 07. Scenarios for gross capital flows produced for this Progress toward full capital market liberalization chapter indicate that developing countries will likely among developed countries took a large step forward in account for a steadily increasing share of inflows in the the post–Bretton Woods period, which may be regarded future—a continuation of the trend that began in the as the Second Age of Financial Globalization (Bordo and precrisis years. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 117 Several key messages emerge from the chapter: from the perspective of financial market development and integration—and, cru- • Investment demand and saving are likely to cially, global financial stability—gross remain more or less in balance at the global flows are a much more relevant metric than level, causing no substantial changes in global net flows. To complement the computable rates of return over the coming decades. At general equilibrium (CGE)-based scenar- the country level, however, trends in invest- ios for saving, investment, and net capital ment and saving may result in significant flows, this chapter builds a picture of gross tensions and current account imbalances. At capital f lows corresponding to the same the global level, not only will returns stay gradual and rapid convergence scenarios. relatively constant under the two scenarios By 2030, gross capital inflows to develop- considered, but global saving and invest- ing countries will reach $6.2 trillion under ment relative to global gross domestic prod- the gradual convergence scenario (47 per- uct (GDP) is also expected to be surpris- cent of the global total) and more than ingly stable: the effect of falling rates within $13 trillion under the rapid convergence countries will be offset by the rising share in scenario (60 percent), rising from $1.3 the world economy of developing countries trillion (23 percent of the global total) in with relatively high saving and investment 2010.1 The increase will be driven by more rates. At the country level, tensions exist for rapid economic growth and slower popu- specific countries in terms of their excess lation aging in developing countries than demand for capital, relative to supply, at ini- in advanced countries, as well as by devel- tial rates of return. These tensions will put oping countries’ relatively greater scope pressure on rates of return, attracting capital for increasing openness and strengthening inflows and affecting current account posi- financial sector institutions. The seemingly tions. Such tensions are more pronounced large increase in the share of gross flows in the rapid convergence scenario. destined for developing countries is not as Countries and regions with relatively dramatic when considered relative to out- young populations and those with the put, however: the scenario analysis fore- greatest scope for financial market devel- sees gross inflows to developing countries opment—both of which tend to moderate of 6–11 percent of their GDP as of 2030. saving and at the same time boost invest- The previous all-time peak was 9 percent, ment—have the greatest potential to expe- in 2007. rience increases in investment relative to Institutional improvement in developing domestic saving and net inflows of capital countries, combined with greater perceived from abroad. For example, India and, on risk in high-income economies, is set to aggregate, Sub-Saharan Africa will run cur- remove advanced countries’ monopoly on rent account deficits averaging 2.4 percent supplying high-quality liquid assets in the and 3.2 percent of GDP, respectively, over decades ahead. Encouraged by positive 2010–30 under the gradual convergence business environment changes, solid eco- scenario. The corresponding net capital nomic growth, and demographic trends inflows will not come primarily from the supportive of growing consumer demand North but from newly industrialized East in the long term, investors are showing Asian countries, most notably China. increasing interest in developing countries • Developing countries are poised to account for far beyond the largest emerging markets. a growing share of global gross capital fl ows • As developing countries become more deeply in the decades ahead. Although establishing integrated into the global financial system and a picture of net capital flows in the future account for a larger share of the world’s gross offers important information about the capital flows, they will also have a greater path of balance-of-payments imbalances, impact on financial and monetary policy 118 Capital Flows in the Third Age of Financial Globalization Global Development Horizons making. Because the use of international • Policy makers will need to prepare for a greater currencies in financial transactions extends role of capital markets in international finan- beyond their home country borders, the cial intermediation and promote the develop- monetary policies of the Euro Area and ment of domestic capital markets. Looking for- the United States currently have a dispro- ward, as gross capital inflows and outflows portionate effect on global monetary con- grow in scale, the composition of these flows ditions and developing countries’ access to will become more important because dif- capital. Looking ahead, this suggests that ferent channels have different capacities for developing countries may become increas- creating stabilizing or destabilizing condi- ingly exposed to spillovers of Euro Area tions in recipient countries. Globally, capital and U.S. monetary policy shocks as their markets will likely intermediate an increas- international balance sheets grow. This out- ing share of gross flows in the future, and come is uncertain, however, considering bank loans will account for less. Whether that because developing countries account this will be good news for countries that for a growing share of global capital flows, currently receive most of their inflows in the their own monetary policies might have a form of bank loans remains to be seen; bank greater impact on the world. This is a pos- lending tends to be highly procyclical and sibility for China, in particular. There is generally less supportive of risk sharing than also the potential for greater regional mon- foreign direct investment (FDI) or equity etary policy spillover from large emerging portfolio investment (Brunnermeier et al. economies such as Brazil and the Russian 2012). In middle-income countries, port- Federation. folio investment has historically been even For small and medium-size developing more volatile than bank lending, in rela- countries, a world in which China, the tive terms (that is, adjusting for the smaller Euro Area, and the United States all have magnitude of this component of develop- key international currencies could be sta- ing countries’ inflows) (World Bank 2012). bilizing in that they will be less affected Moreover, as households and firms in devel- by monetary policy spillovers from any oping countries increasingly demand not one source. At the same time, as growing only greater access to credit but also greater amounts of capital are transferred among choice and variety in financial assets and developing countries, monetary policy services, domestic financial markets will coordination between developing countries have to compete globally in terms of their will become more critical in promoting structure and their depth. Although the stable fi nancial and macroeconomic con- many efforts under way to improve regula- ditions among them. An increasing share tion of the international banking sector will of global flows going to and from develop- remain highly relevant, policies should also ing countries indicates that these countries be prepared to accommodate—and in some should have a larger role in management cases actively promote—the development of capital flows within multilateral orga- of domestic capital markets. At the same nizations as well as bilaterally. But, given time, regulatory authorities in developing that monetary policy coordination has countries should monitor the composition never been perfectly feasible even among a of both capital flows and domestic financial small number of major players—and could intermediation, and, more broadly, should become even more difficult in an increas- develop regulatory institutions to be more ingly multipolar world—second-best forward-looking—ready to adapt to poten- policy solutions at the national level may tially destabilizing changes in the composi- become increasingly crucial to stability as tion of balance sheets and financial market well. innovations. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 119 Rates of return around the world Th is report estimates each country’s yield to It has long been argued that large differences in suppliers of capital as the marginal product of output per worker across countries imply simi- reproducible capital, after adjusting for the rela- larly large differences in capital-labor ratios and tive price of capital to consumption goods and thus higher returns to capital in developing coun- depreciation.5 By this measure, the average yield tries than in high-income countries. Th is result in developing countries for 2007 was 12.4 per- came to be viewed as something of a puzzle, with cent, compared with 5.3 percent in high-income a number of hypotheses competing to explain countries.6 Of course, the true extent of differ- why, given high rates of return to capital, invest- ences in yields across countries remains an open ment is not much greater in the developing world question, and an important one. Steps toward or, alternatively, why returns are not in fact higher resolving this question can shed light on a wide in developing countries despite lower capital range of policy questions, such as the potential stocks. Th is question was articulated most nota- for official aid flows to stimulate growth, effects bly by Lucas (1990), who argued that accounting of political instability on investment, and the role for disparate endowments of complementary fac- of policies designed to spur human capital accu- tors, in particular human capital, goes a long way mulation and knowledge spillovers in driving toward resolving the puzzle.2 Some other poten- investment and growth. tially relevant factors are political risk and barri- ers to capital flows. There will not be much pressure Empirically, the evidence on cross-country on global rates of return to differences in returns is mixed. Observed interest investment rates tend to be significantly higher in develop- ing countries, but these rates depend on default Saving and investment are set to remain more or risk and, in some cases, on fi nancial repression, less in balance at the global level, so there will not both of which are not straightforward to quan- be significant pressure on equilibrium yields. The tify but almost certainly tend to have a greater global saving rate can be expected to hold more impact on interest rates in developing countries or less steady over the next two decades. Th is is than in advanced ones.3 The underlying return to because, at least up to 2020, acceleration of eco- capital is not directly observable, and alternative nomic growth in developing countries will mod- assumptions upon which its measurement can be erate the negative impacts of aging and financial based produce widely varying estimates. Gollin market development on developing countries’ (2002) made an important advance in attempting saving rates, so that they tend to fall only mod- to account for self-employment income, which is erately. More important, the effect of falling sav- often incorrectly counted as capital income and is ing rates across countries on the global aggregate important for cross-country comparisons because saving rate will be offset by the growing weight of self-employment typically constitutes a greater relatively high-saving developing countries in the share of economic activity in developing coun- world economy. tries than in developed ones. Caselli and Feyrer To compare trends in demand for capital to (2007) extended this line of reasoning to land trends in its supply, a measure of investment and natural resources and used estimates of the demand must be constructed that is indepen- stocks of these other factors to measure the return dent of yields. This measure—“notional” demand to reproducible capital distinctly from the income for investment—reflects the level of investment that should be attributed to them. Th is greatly that producers desire, at a given fi xed yield.7 Of reduces the variation in returns across countries course, at the global level the quantity of invest- because land and natural resources tend to play a ment demanded is ultimately determined by the larger role in production in developing countries supply of saving, with yields adjusting so that the than in developed countries.4 two are equal. 120 Capital Flows in the Third Age of Financial Globalization Global Development Horizons In the scenario of gradual economic con- demand for investment exceeds notional supply, vergence between the developed and develop- therefore signaling a tension in favor of higher ing worlds, global notional investment demand yields. However, the excess demand is modest remains fairly close to global saving supply, and decreasing over time, hence yields in India which demonstrates that investment demand between 2010 and 2030 remain fairly stable, at is not constrained by the availability of saving.8 around 10 percent. The current account deficit Consequently, the global average yield stays for India—and hence capital inflows—likewise essentially constant in this scenario. In the rapid remains fairly stable, at around 2 percent of GDP. convergence scenario, however, rapid growth and The excess demand for investment is slightly financial development in the developing world higher for the case of Sub-Saharan Africa, and results in faster global capital demand growth trending upward; this leads to a small increase in relative to saving supply, which translates to some yields (by close to one percentage point over two moderate upward pressure on yields.9 Even in this decades), to around 13 percent. Were saving to scenario, however, the deceleration of labor force remain unchanged, this would imply a worsening growth (and even shrinking labor forces in some current account deficit over time; however, ris- parts of the world) will mean that global invest- ing African saving actually offsets the increasing ment will not be constrained by the supply of sav- excess demand, resulting in an improvement in ing much more severely than it is today.10 the current account (which nevertheless remains in slight deficit for the entire projection horizon). The evolution of returns to It is also illuminating to contrast these coun- capital at the country level has tries’ experiences in the rapid convergence sce- implications for net capital flows nario, where brisk growth has a disproportion- ate effect on the demand for, relative to supply At the country level, to better understand the of, capital. In this scenario, for the case of India, forces that are shaping the direction of yields to excess demand remains high throughout the capital, it is useful to conduct a thought experi- projection period. Th is induces a more substan- ment where not only the ex ante notional demand tial increase in yields, by almost two percentage is estimated but also the ex ante notional supply. points. This increase attracts more capital inflows This will allow one to identify whether the demand and, as a consequence, current account deficits in for capital—at a fi xed yield—is met by a supply India rise from −2 percent to −6 percent of GDP. of investment financing at the same fi xed rate. If Sub-Saharan Africa, in the rapid convergence notional demand exceeds supply, for instance, scenario, offers the starkest example of how ten- there will be pressure for the rate of return to rise; sions between notional investment demand and the observed ex post rate of return is a reflection, in supply can affect the evolution of yields and equilibrium, of the easing of such tensions, if any, resultant capital flows. Here, excess demand for between notional demand and supply of capital. capital rises substantially throughout the 20-year In the gradual convergence scenario, invest- period—the ratio of notional demand to supply ment demand in most economies remains fairly rises by five percentage points—and the region’s muted, which translates to a fairly stable path yields reach 14 percent by 2030 (from slightly less for yields between 2014 and 2030 (figure 3.1, than 15 percent). As a result, Sub-Saharan Africa panel a). There is greater heterogeneity in invest- as a whole eventually runs current account defi- ment demand in the rapid convergence scenario, cits of almost 7 percent of GDP. with certain economies experiencing signifi- cant increases in capital demand; this results in sharply rising yields over the next 15 years in some countries (figure 3.1, panel b), in contrast to Changing patterns in global average yields. net capital flows Consider, for example, the case of India in Capital flows, on a net basis, reflect the difference the gradual convergence scenario. Here, notional between a country’s saving and investment; as Global Development Horizons Capital Flows in the Third Age of Financial Globalization 121 FIGURE 3.1 Changes in yields are more muted under the gradual convergence scenario than under rapid convergence a. Gradual convergence b. Rapid convergence 14 14 12 12 10 10 8 8 Yield, % Yield, % 6 6 4 4 2 2 0 0 pe n es il do na a Sa de ia n a pe n es il a a a n a az si ic az in si di ic pa ra tio pa tio d at at ro i ro ne fr ne fr Ch n Ch In Br Br Ja Ja ha ra ra St St I A A Eu Eu do de n n d d ra In In te te e Fe F ha ni ni Su ian n U U Sa ia ss ss b- b- Ru Ru Su 2014 2030 2014 2030 Source: World Bank calculations. Note: This report estimates each country’s yield to suppliers of capital as the marginal product of reproducible capital, after adjusting for the rela- tive price of capital to consumption goods and depreciation. Yields are reported for 2014 (the first year where the model attains potential GDP) and 2030 (the end of the projection period). Gradual and rapid convergence refer to the relative pace of projected convergence between developed and developing economies in terms of productivity growth and structural changes. such, the disparities in levels of saving and invest- fi nancial market development has the potential ment within individual economies through 2030 to reduce many developing countries’ current described in chapters 1 and 2 naturally imply the account surpluses (or raise their deficits). China continued movement of capital across borders is a special case, facing relatively severe aging but over that period (as detailed in online annex 3.2). also having less scope for a rapid expansion in In the decades ahead, financial markets in devel- credit, so that investment will likely fall at least oping countries will become considerably deeper, as much as saving in the years ahead. To put the more sophisticated, and more integrated region- scenarios in context, box 3.2 provides a brief look ally and globally; these countries will tend to face at the historical record of shifts in the global pat- better growth prospects than advanced econo- tern of net capital flows. mies; and the timing of demographic change will vary widely across countries. These will be The evolution of net capital flows the key drivers of changes in the global pattern depends on the paths of growth of net capital flows and net international asset and financial market development positions.11 Investment may be spurred by lower transac- The configurations of which countries are net tion costs in credit markets—for example, from importers or exporters of capital will continue more efficient collateralization—and at the same to shift in the future.12 Under the gradual con- time, greater access to credit will tend to reduce vergence scenario, developing countries’ sav- saving. Taken together, these effects mean that ing rates will tend to fall somewhat because of a 122 Capital Flows in the Third Age of Financial Globalization Global Development Horizons BOX 3.2 Historically, the dominant net providers and recipients of capital have not been fixed Between 1850 and 1929, Great Britain was a net pro- China since around 2000), together with major oil vider of capital to the rest of the world, on the order exporters, have become significant net providers of of about 3.5 percent of its GDP annually and as much capital to the rest of the world, with the United States as 8–9 percent in some years (figure B3.2.1, panel a), and a diverse group of Latin American, African, and fi rst by running a trade surplus as a major exporter peripheral European economies serving as the pri- of goods such as textiles, and later because repatri- mary recipients of that capital (figure B3.2.1, panel ated earnings from overseas investments more than b). There has been much debate over the causes, compensated for a declining trade balance. Following consequences, and sustainability of large balance-of- World War I, the United States gradually took over payments positions (see, for example, Blanchard and that mantle, running an annual balance-of-payments Milesi-Ferretti 2010; Lin and Dailami 2010; Caballero surplus of 1.9 percent in the interwar period and again and Krishnamurthy 2009; Chinn and Ito 2007; and through most of the post–World War II period until the Obstfeld and Rogoff 2007). But to the extent that such mid-1970s. Japan and Germany became net lenders positions are believed to be destabilizing to the global of significance in the 1980s. economy, any sustainable long-run rebalancing must More recently, the global pattern of balances of address underlying cross-country variations in incen- payments has undergone yet another transformation. tives for investment and saving behavior. Newly industrialized East Asian economies (particularly FIGURE B3.2.1 Net capital outflows (current account), the United Kingdom and the United States, 1850–1945 (panel a) and evolution of the global pattern of current account balances, 1980–2010 (panel b) a. Net capital outflows, 1850–1945 b. Current account balances, 1980–2010 15 2,000 1,500 Capital outflow, % of GDP 10 1,000 5 500 $ billions 0 0 –500 –5 –1,000 –10 –1,500 –15 –2,000 50 60 70 80 90 00 10 20 30 40 19 0 19 2 19 4 86 19 8 19 0 19 2 1994 19 6 20 8 20 0 02 20 4 20 6 08 10 8 8 8 8 9 9 9 9 0 0 0 18 18 18 18 18 19 19 19 19 19 19 19 20 20 United States China Other developing United Kingdom United States Other advanced Major oil exportersa Sources: World Bank calculations using data in Jones and Obstfeld 2001 and the IMF International Financial Statistics database. a. Major oil exporting countries, as defined by the IMF (International Monetary Fund), are Algeria, Angola, Bahrain, Canada, Colombia, Ecuador, the Islamic Republic of Iran, Iraq, Kuwait, Libya, Mexico, Nigeria, Norway, Qatar, the Russian Federation, Saudi Arabia, the United Arab Emirates, and the República Bolivariana de Venezuela. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 123 FIGURE 3.2 In the gradual convergence scenario, there will be sizable net capital flows from China to high-income countries a. High-income countries and the b. Selected developing countries developing world and regions Current account balance, % of GDP Current account balance, % of GDP 12 12 10 10 8 8 6 6 4 4 2 2 0 0 –2 –2 –4 –4 –6 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 20 20 2 20 4 20 6 20 8 30 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 30 0 0 0 0 1 1 1 1 1 2 2 2 2 0 0 0 0 1 1 1 1 1 2 2 2 2 2 20 20 20 High income China China Russian Federation Developing without China India Brazil Sub-Saharan Africa Source: World Bank projections, supplemented with calculations using data in the World Bank Global Economic Monitor database. Note: Net capital fl ows are reported as the current account position, with surpluses indicative of capital outfl ows. The current account positions for the two groups do not exactly offset because they are computed as a share of each respective group’s output. It is also well known that historical data on current account balances do not generally sum to zero at the global level, as required by theory, because of measurement problems. combination of aging and moderate but steady As discussed in chapter 2, under this scenario, financial market development. At the same aging and financial development lead to a fall time, investment will slow because of a moder- in China’s saving rate of roughly 7 percentage ate slowdown in growth. For most developing points, from 52 percent of GDP in 2010 to 45 countries, these two effects will roughly can- percent in 2030. However, China’s investment cel out and have fairly little impact on current rate will decline even further, falling from 48 accounts. Excluding China, the developing world percent to 37 percent over the same period. The will, on aggregate, maintain a moderate, gradu- scenario analysis implicitly accounts for a transi- ally attenuating deficit between 2015 and 2030 tion from heavy state involvement in investment (figure 3.2, panel a). For example, India and, on decisions to a more market-driven structure of aggregate, Sub-Saharan Africa will follow this investment in China.13 At the same time, steadily pattern, being fairly early in their demographic rising wages will result in a steep decline in its transitions and having significant scope for finan- rental-wage ratio, rendering returns in China cial market development. Brazil will fit this pat- relatively less competitive, and this manifests in tern for most of the projection period as well, but, the investment rate falling from its very high his- facing more severe aging, its deficit will level off torical levels, although it remains high by inter- and then begin to rise slightly in the mid- to late national standards. The net result will be that 2020s. Net capital flows from China to advanced China maintains its role as a significant exporter countries, on the other hand, will increase sub- of capital to the rest of the world. stantially, leveling off after reaching 6–7 percent Under the rapid convergence scenario, because of China’s GDP (figure 3.2, panel b); the flows of steady credit growth, saving rates in developing will represent a rising share of high-income coun- countries will be lower in the future. Conversely, tries’ output because these economies will expand their investment rates will fall only margin- relatively more slowly than China. ally, sustained by institutional development and 124 Capital Flows in the Third Age of Financial Globalization Global Development Horizons FIGURE 3.3 In the rapid convergence scenario, much of the developing world apart from China will be net capital importers a. High-income countries and the b. Selected developing countries developing world and regions Current account balance, % of GDP Current account balance, % of GDP 12 15 10 8 10 6 5 4 2 0 0 –2 –5 –4 –6 –10 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 30 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 20 0 20 2 20 4 20 6 20 8 30 0 0 0 0 1 1 1 1 1 2 2 2 2 2 0 0 0 0 1 1 1 1 1 2 2 2 2 2 20 20 High income China China Russian Federation Developing without China India Brazil Sub-Saharan Africa Source: World Bank projections, supplemented with calculations using data in the World Bank Global Economic Monitor database. Note: Net capital fl ows are reported as the current account position, with surpluses indicative of net capital outfl ows. robust productivity growth. The resulting out- policy community: how to manage China’s cur- come is a channeling of much of the world’s sav- rent account surplus. Some reduction in China’s ing to the developing world. However, there will surplus should be desirable for the country, at be a widening imbalance between China and least to the extent that surpluses are the result of other developing countries, with most developing domestic distortions such as lack of social protec- countries running gradually increasing deficits tion and weak corporate governance.16 Benefits to (figure 3.3).14 Relative to the gradual scenario, China include regaining monetary independence China’s current account surplus will evolve more through reducing reserve accumulation, smooth- gradually, with smaller net flows to high-income ing international relations, and improving global countries due to relatively less attractive invest- fi nancial stability. At the same time, it must be ment opportunities there, in a world in which acknowledged that a rapid reduction in China’s global growth is driven almost entirely by the reserve accumulation has the potential to destabi- developing world.15 lize the global financial system; thus, the timing and speed of adjustment are important. But this result also highlights a caveat that Scenarios of China’s current must be made to the approach taken to gener- account balance highlight looming ate these two scenarios. The determinants of policy challenges investment and saving in the model were selected Under the two scenarios explored in this report, because they tend to be important across coun- China will run large current account surpluses tries, and the sensitivity of saving and investment over the next two decades. This will finance invest- to these factors is assumed to be the same across ment mainly in high-income countries in the countries (online annexes 1.6 and 2.6). This gradual convergence scenario and mainly in devel- approach may miss some idiosyncratic factors oping countries in the rapid convergence scenario. that influence saving and investment behavior, These results highlight a major policy challenge and thus net capital flows, in any particular coun- faced by China and, indeed, by the international try. Given the institutional and policy challenges Global Development Horizons Capital Flows in the Third Age of Financial Globalization 125 faced by China in the context of its atypical eco- and social safety net measures such as pension nomic system, such idiosyncratic factors might be and public health reforms have the potential to especially important there, which deserves special significantly reduce Chinese savings and exter- mention because of the country’s size and global nal imbalances, and these policy choices are only economic influence. crudely captured by the scenario analysis.17 Of In the context of investment rates moderat- course, policy makers in deficit countries, par- ing to more sustainable levels, current account ticularly the United States, have a role to play in imbalances can only be resolved by implementing resolving global imbalances as well.18 policies to bring saving down more quickly than it would fall from demographic change alone. One policy avenue is institutional reform to pro- Expanding volumes of gross mote broad-based financial market development. capital flows to and from Most developing countries will experience a faster decline in saving rates than investment rates in developing countries the rapid convergence scenario because of credit Although the scenarios for the path of net capital growth, as discussed earlier, but China will sus- flows offer important information on the extent tain its surplus under this scenario. of balance-of-payments imbalances in the decades This result for China is observed chief ly ahead, assessing net capital f lows alone over- because the country starts out with a high initial looks important monetary and fi nancial stabil- ratio of credit to GDP and thus has less room for ity aspects of transferring capital across national convergence. Domestic credit to the private sector borders. From the perspective of assessing these as a share of GDP, a standard aggregate measure issues, measuring gross capital flows—along with of financial sector development, was 130 percent the composition of those flows—is more informa- of GDP in China in 2010, well above the aver- tive than looking at net balances. age of 73 percent for middle-income countries Furthermore, the impact of fi nancial market (World Bank World Development Indicators development on developing countries’ volumes of database) and implying that China already has a gross inflows and outflows is quite distinct from highly developed financial sector. In Tunisia and its impact on the net balance between the two. Uruguay, also middle-income countries, domes- Although greater access to finance has the poten- tic credit extended to the private sector was 69 tial to both stimulate investment and reduce sav- percent and 22 percent of GDP, respectively. The ing, thus narrowing current account balances, situation in China is considerably more nuanced development and integration of developing than this one metric indicates, however. On an countries’ fi nancial markets can be expected to index of household access to financial services expand gross flows significantly and to potentially developed by Honohan (2008), for example, alter the composition of capital flows as well. An China scored 42 out of a possible 100, the same expansion of developing countries’ gross capital as Tunisia and Uruguay. Financial underdevel- flows will also be driven by relatively favorable opment has been identified as one of the factors demographic conditions for attracting foreign behind increases in China’s saving, as the need investment as well as by healthy productivity for credit-constrained households to save for big- improvements and growth. A country’s pace of ticket items and precautionary motives increased growth and path of financial market development as social safety nets shrank (Chamon and Prasad and integration depend on institutional change 2010). On the corporate saving side, distortions and policy choices—a range of possibilities high- such as barriers to competition, corporate gov- lighted by considering alternative scenarios of ernance weaknesses, undertaxation, and advan- how these variables may evolve in the future and tageous access to fi nance for state-owned enter- the corresponding paths of gross capital flows. prises have led to high retained earnings (Lin International financial integration is reflected and Dailami 2010). Institutional reforms of the by growth in both inflows and outflows of capi- corporate sector, financial market liberalization, tal as investors at home and abroad diversify and 126 Capital Flows in the Third Age of Financial Globalization Global Development Horizons their countries share risk. Greater gross f lows capital inflows and outflows has vastly exceeded can mean more access to finance for some types the increase in their net counterparts (figure 3.4). of investment projects because the composi- Initially, this was a high-income-country phe- tion of inflows and outflows tends to differ.19 nomenon, visible in the takeoff in gross inflows to Thus, investment projects can go ahead despite and outflows from those countries. The upward residents’ relatively low tolerance for risk and trend continued for two decades, so that from without the country necessarily borrowing on 1990 to the peak in 2007, the sum of high-income net. Large amounts of gross flows can be desta- countries’ gross capital inf lows and outf lows bilizing, though; cross-border banking flows, in grew tenfold, compared with a less than fourfold particular, are volatile and highly procyclical, increase in trade flows (imports plus exports) and potentially magnifying risk instead of spreading a less than two-and-a-half-fold increase in nomi- it (Brunnermeier et al. 2012). Given that volumes nal GDP over that period.20 of gross capital flows have been growing much Developing countries’ integration into the more rapidly than net balances globally, and that financial system during the 1990s remained the path of gross flows is both an essential under- quite modest relative to their economic size, per- pinning of investment and a potentially destabi- ceived investment opportunities, and engage- lizing force with important policy implications, ment in global trade. Evidence to this effect can thinking about how gross flows will evolve in the be observed in patterns of global capital flows, in future is a valuable exercise. the role of developing-country banks and other fi nancial institutions in the process of interna- tional financial intermediation, and in developing Developing countries’ inflows and countries’ capacity to influence the rules of global outflows of capital remain small finance. It was only during the expansionary relative to those of high-income years just prior to the global financial crisis that economies, but they have begun developing countries came to play a greater role to take off in international financial intermediation and that Since the mid-1990s, financial globalization has gross inflows to and outflows from developing meant that the significant expansion in total gross countries began to take off.21 By 2007, 11 percent FIGURE 3.4 Gross capital flows have long expanded relatively faster than net flows in high- income economies, and in recent years a similar trend can be observed in developing economies a. High-income economies b. Developing economies 25 25 Capital flows, % of GDP Capital flows, % of GDP 15 15 5 5 0 0 –5 –5 –15 –15 –25 –25 19 0 19 2 1984 1986 1988 1990 1992 1994 19 6 2098 20 0 20 2 2004 2006 2008 10 19 0 19 2 1984 1986 1988 1990 1992 1994 19 6 2098 20 0 20 2 2004 2006 2008 10 8 8 9 0 0 8 8 9 0 0 19 19 Current account surplus Current account deficit Gross capital outflows Gross capital inflows Source: World Bank calculations using data in the IMF International Financial Statistics database. Note: Current account surplus (deficit) is equal to the sum of current accounts across surplus (defi cit) countries in each group, divided by the group’s GDP. Gross capital fl ows are summed across all countries in each group, divided by the group’s GDP. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 127 of gross global capital inflows went to developing FIGURE 3.5 International investment positions still countries, versus 4 percent in 2000. reflect a large disparity in foreign asset holdings between Although the financial crisis brought on a dra- advanced and developing countries matic downturn in the volume of global capital 120 200 flows, capital flows to and from developing coun- External assets, % of world GDP tries were much less affected during the crisis 100 External assets, $ trillions than flows in and out of advanced countries.22 160 Before the crisis, not only did high-income coun- 80 tries account for the great majority of both global 120 gross inflows and outflows, but current account 60 balances also constituted a much smaller share 80 of gross flows in high-income countries than in 40 developing countries. Even as high-income coun- 40 20 tries’ flows have recovered in the immediate post- crisis years, developing countries’ share of global 0 0 flows is significantly greater than a decade ago, 80 82 84 86 88 90 92 94 96 98 00 02 04 06 08 10 although still not commensurate with their share 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 of global output, growth, and trade. International Developing countries Advanced countries % of world GDP investment positions reflect a large disparity in (right axis) foreign asset holdings between advanced and Source: World Bank calculations using data in the IMF International Financial Statistics developing countries as well (figure 3.5). database. Several ongoing trends will provides structural support for cross-border capi- support a rapid expansion of tal movement in that capital tends to flow out of capital flows to developing countries with high saving rates. countries in the future One promising line of research views gross Rapidly growing economies attract increased capital f lows from the perspective of interna- capital flows because that growth implies sizable tional trade in fi nancial assets; as such, a coun- investment opportunities and, often, improv- try’s engagement in global finance depends on ing creditworthiness and more opportunity for its residents’ demand for foreign assets as well as investment diversification. Stronger, more devel- its capacity to supply assets with return and risk oped financial institutions, in particular, tend to characteristics sufficiently appealing to foreign allow developing countries to attract larger capi- investors (Broner et al. 2013). Thus, gross capital tal flows. In terms of intermediation, although flows into and out of a country can be expected to stock markets would need to be of some threshold be related not only to the country’s generation of size to attract institutional investors, the interna- profitable investment opportunities but also to its tionalization of the banking industry has been an level of financial market sophistication and global important vehicle of information transmission integration.23 for global investors (Portes and Rey 2005), in It has been a longstanding stylized fact turn allowing a growing amount of capital to be that international portfolio diversification has directed to developing countries. occurred much less than implied by models of Demographic conditions also matter: coun- risk sharing unless they incorporate frictions such tries with a large share of working-age people as asymmetric information or other sources of tend to attract more capital inflows because of transaction costs.24 A reduction in these frictions their relatively large labor pools and potential would imply a large increase in gross flows into for high rates of productivity growth (Lane and and out of a given country, as domestic agents Milesi-Ferretti 2002; Domeij and Flodén 2006, diversify by sending capital abroad and foreign among others). The large diversity in the demo- agents diversify by sending capital in. Advanced graphic conditions across the world, as well, countries seem to have been following this 128 Capital Flows in the Third Age of Financial Globalization Global Development Horizons FIGURE 3.6 The relationship between GDP per capita lagged behind their shares of global output and and external balances across countries is nonlinear trade. This line of reasoning is also consistent with the prospect that developing countries’ vol- 6 umes of capital flows are set to accelerate. Total external assets, multiples of GDP GBR BEL ISL NLD The expansion of developing 4 countries’ gross capital flows will depend on the pace of growth and AUT financial development FRA NOR FIN SWE DNK The econometric work underlying this report’s 2 y = 9.6 – 2.6x + 0.18x 2 PRT DEU forecast of gross capital flows is supportive of pre- BWA MYS HUN ESP ITA USA vious literature in that it finds that higher levels SWZ ISR JPN KWT CHL LSO DJI BOL EGY IRQ JOR NAM BGR ZAF ARG URY RUS VEN LVA EST SVN GRC NZL AUS CAN of capital inflows—encompassing FDI, portfolio KHM CHN CZE UKR THA KAZ LTU SLE MOZ UGA TGO BFAMLI KGZ BEN CIV MDA NGA YEM SLB SENNIC IND BTN PHL SYR HND PRY MAR CPV AGO GTM ECUALB IDN GEO ARMSLVTUN BIH MKD AZE PER JAM CRI ROM DOMCOL TUR POL HRV SVK KOR f lows, and bank lending—are associated with TZA NER RWA GNB HTI PAK MDV BRA MEX 0 BGD BLR increasing fi nancial development and openness LIC LMC UMC HIC (measured by domestic credit to the private sec- Log GDP per capita tor as a percentage of GDP and the Chinn and Sources: World Bank calculations using data in the IMF International Financial Statistics and Ito [2006] index of capital account openness, World Bank World Development Indicators databases. respectively) as well as with higher GDP growth, Note: The external balance sheet is measured as total external assets plus total external a lower dependency ratio, and global monetary liabilities, divided by two. Countries identified as offshore financial centers are excluded. On the horizontal axis, mid-points (of the log) of the 2011 World Bank income classification ranges conditions. Th is work provides the basis for two are shown, where LIC denotes low-income, LMC lower-middle income, UMC upper-middle scenarios of projected gross inflows to 2030, cor- income, and HIC high income. responding to the gradual and rapid convergence scenarios presented in the previous two chapters (box 3.3 and online annex 3.4). pattern at least until the onset of the global finan- Under the gradual convergence scenario, cial crisis, as their fi nancial markets integrated global gross capital inflows will reach $13 tril- and gross flows expanded rapidly. lion in 2030 (figure 3.7, panel a), with the share The relationship between gross flows and over- destined for developing countries rising to 47 all development, proxied by per capita GDP, is percent ($6.2 trillion). In the rapid convergence also evident when the accumulation of gross flows scenario, developing countries’ comparatively into stocks of international assets and liabilities stronger growth and convergence with developed is examined. As presented in figure 3.6, the rela- countries in terms of financial sector develop- tionship between GDP per capita and external ment and globalization will contribute to a faster balances across countries is nonlinear: a 1 per- increase in the growth of capital flows than in the cent difference between countries in GDP per first scenario. Global gross capital inflows will capita corresponds to an increasingly large differ- reach nearly $22.6 trillion by 2030 in the rapid ence in external assets as a percentage of GDP as convergence scenario, with a larger share of the countries at higher levels of GDP per capita are total (60 percent, or $13.5 trillion) expected to go compared. to developing countries (figure 3.7, panel b). Were countries to follow a similar curve over One factor behind the increase in develop- time (which is admittedly a strong assumption ing countries’ share of global gross flows is sim- but makes for an interesting thought experiment), ply the change in their relative size. Under the then developing countries would first move along gradual scenario, gross inf lows to developing the downward-sloping and relatively flat portions countries will change little relative to their output of the curve. This is consistent with the observa- (figure 3.8), so the increase in their share is due tion that these countries’ shares of gross capital largely to their growing relative economic size, flows and international investment positions have coupled with a moderate decline in high-income Global Development Horizons Capital Flows in the Third Age of Financial Globalization 129 BOX 3.3 Gross capital flow scenarios are based on historical evidence on the relationship between the flows and their determinants The construction of gross capital inflow projections developing countries as distinct groups. The paths of begins with the estimation of four specifications of the key explanatory variables correspond to the paths a model capturing the determinants of gross capital of these variables as assumed, or endogenously gen- inflows along five broad categories of explanatory vari- erated by the CGE model, in the two global scenarios ables. These categories include country-level demo- (gradual convergence and rapid convergence) devel- graphic factors, country-level macroeconomic factors, oped in chapters 1 and 2. The only exception is the country-level structural and institutional factors, global measure of capital account openness, which is not a economic factors, and controls for global shocks over variable in the CGE model; scenario paths for capital time and country and income group effects. The model account openness are built specifi cally for the gross is estimated for a large number of developing and devel- capital flows projections. oped countries using data for the period 1980–2010. Finally, estimates of gross capital outflows are Using the results of the preferred specification inferred from the difference between gross inflows and of the model (column 4 in table 3A.1 of online annex net flows (the current account). A full explanation of the 3.3), forecasts of gross capital inflows to 172 countries variables, the modeling process, and the projection pro- through 2030 are constructed, treating developed and cess is presented in online annexes 3.3 and 3.4 . FIGURE 3.7 By 2030, most gross capital flows may go to developing countries a. Gradual convergence scenario b. Rapid convergence scenario 25 25 Gross capital inflows, $ trillions Gross capital inflows, $ trillions 20 20 15 15 10 10 5 5 0 0 2000 2005 2010 2015 2020 2025 2030 2000 2005 2010 2015 2020 2025 2030 Advanced countries Developing countries Source: World Bank projections, supplemented with calculations using data in the IMF International Financial Statistics database. Note: Infl ows are depicted in nominal U.S. dollars, assuming a constant 3.5 percent world infl ation rate that is based on the 2003–07 fi ve-year aver- age from the World Bank Global Economic Monitor database. Gradual and rapid convergence refer to projected scenarios concerning the relative pace of convergence between developed and developing economies in terms of productivity growth and structural changes. countries’ inflows as a share of output due to a capital inflows to developing countries is illus- slowdown in growth. trated by the rapid convergence scenario. Here, The potential contribution of financial devel- the rise in gross capital inf lows to developing opment and integration to an expansion in gross countries will outstrip economic growth, so that 130 Capital Flows in the Third Age of Financial Globalization Global Development Horizons FIGURE 3.8 Relative to developing countries’ GDP, gross Developing economies’ growing inflows may hold steady or trend upward, depending on importance as sources and the pace of growth and financial development destinations of capital flows will 16 not be a BRICs story alone Gross capital inflows, % of GDP 14 Disaggregating the scenario results by country 12 shows that China will be an important part of 11% Linear (historical) the story, but by no means will changing future 10 9% patterns of capital flows be a China story alone 8 Rapid convergence Historical scenario (figure 3.9), nor will it be a BRICs (Brazil, the 6% 6 Russian Federation, India, and China) story Gradual convergence scenario alone. Financial development and integration 4 2 have the potential to boost gross capital flows across the developing world. Sub-Saharan Africa, 0 for example, can be expected to not only receive 30 a growing volume of capital flows, but also to 00 02 04 06 08 10 12 14 16 18 20 22 24 26 28 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 attract an increasing share of total capital flows to Source: World Bank projections, supplemented with data in the IMF International Financial Statistics database. developing countries (box 3.4). Under the gradual Note: Gradual and rapid convergence refer to projected scenarios concerning the relative convergence scenario, developing countries other pace of convergence between developed and developing economies in terms of productivity than the BRICs will collectively account for 17 growth and structural change. percent of global gross inflows in 2030 and 15 percent of outflows. In this multipolar world, no single country will attract as great a share of global inflows reach 11 percent of their GDP in 2030, inflows as the Euro Area or the United States do exceeding the level in 2007 (9 percent) when today, not even China; small and medium-size capital flows peaked in the lead-up to the global developing countries will collectively matter much financial crisis. This works together with their more in the global economy than they do today, growing relative size to raise their share of global particularly in terms of their role in the global capital flows substantially under this scenario, financial markets and capital flows. although it is still somewhat conservative relative On the other hand, by 2030, China may be to the growth rate of capital flows to developing as important a source of capital flows as Europe countries between 2000 and 2011 (represented by and the United States are today. Under both the the linear trend line in figure 3.8). gradual and rapid convergence scenarios, China is The likelihood of one scenario or the other expected to account for over 40 percent of global playing out depends not only on several factors outflows in 2030.25 This will be driven mainly by that are out of the hands of developing-country its robust growth, greater integration with global policy makers—such as global productivity financial markets, and significant current account advancements and demand for exports—but also surpluses stemming from its investment rate (cur- on factors that are very much influenced by policy rently very high) falling somewhat faster than its at the national level. These include (a) education saving rate as wages rise (some details are pro- reform; (b) efforts to reduce corruption, barriers vided in chapters 1 and 2). to competition, and distortions in factor markets that limit the adoption of technologies and thus The composition of global constrain economic growth; (c) policies to pro- capital flows in terms of financial mote fi nancial market development, along with instruments may be poised for complementary reform of regulatory institutions change to limit volatility and risk and enhance corporate governance and transparency; and (d) trade and Financial market development and integration financial market integration. can be expected to affect the composition, as Global Development Horizons Capital Flows in the Third Age of Financial Globalization 131 FIGURE 3.9 Small and medium developing countries as well as the BRICs will be impor- tant sources and destinations of capital flows a. Global gross capital inflows b. Global gross capital outflows 100 100 90 90 80 80 70 70 60 60 Percent Percent 50 50 40 40 30 30 20 20 10 10 0 0 2010 2030 (gradual) 2030 (rapid) 2010 2030 (gradual) 2030 (rapid) High-income countries China Brazil, India, and the Russian Federation Other developing countries Source: World Bank projections for 2030; World Bank calculations using data in the IMF International Financial Statistics database for 2010. Note: Gross capital fl ows include portfolio, foreign direct investment, and banking transactions. Gradual and rapid refer to projected scenarios concerning the relative pace of convergence between developed and developing economies in terms of productivity growth and structural change. BRICs = Brazil, the Russian Federation, India, and China. well as the overall levels, of capital flows. A brief changes in the composition of inflows thus may examination of trends in recent decades shows affect an economy’s exposure to exchange rate long-standing differences in the composition risk and monetary policy spillovers.26 of capital inf lows to developed countries ver- There is widespread speculation that a larger sus developing countries. Developing countries percentage of the world’s capital flows will be receive mainly equity inflows, while their out- intermediated by capital markets in the future, flows are predominantly into relatively safe secu- implying that banking and FDI inflows will repre- rities, meaning that a disproportionate amount of sent relatively less of the total. This is an uncertain risk is borne abroad (compensated for by higher outcome, however, considering that portfolio flows returns on these foreign liabilities than on foreign currently make up a relatively small share of capital assets). Indeed, most capital inflows to develop- flows into and out of developing countries and that ing countries were in the form of FDI during the these countries’ flows can be expected to consti- 1990s and early 2000s (figure 3.10). Inflows to tute a growing share of global aggregate flows in advanced countries were more diverse, with the the future. But even if the predicted increase in lion’s share in the form of loans and portfolio the use of capital markets as a channel of finan- investment. The pattern of inflows to developing cial intermediation is not borne out globally, capi- countries changed somewhat in the 2000s, how- tal markets are likely to be increasingly important ever, with loans representing a rising share. in developing countries as their financial markets The relative shares of FDI, bank lending, and mature, with the long-term effect of developing- portfolio investment in gross inflows have impor- economy convergence with advanced economies in tant implications for macroeconomic stability terms of the composition of their capital inflows. and policy making; banking and portfolio flows Ultimately, the extent to which capital mar- tend to be more procyclical and volatile than kets play a greater role in funding investment FDI. Currency denomination varies by compo- in these countries depends to some degree on nent as well. Because bonds are typically issued policy choices. Historically, policies deliberately in only a handful of international currencies, designed to promote FDI, such as tax incentives, 132 Capital Flows in the Third Age of Financial Globalization Global Development Horizons BOX 3.4 Encouraged by the continent’s growth prospects, demographic trends, and improved investment climate, global investors and creditors are increasingly attracted to Sub-Saharan Africa Sub -Saharan Africa is poised to receive steadily large developing economies such as Brazil, China, and increasing amounts of capital infl ows over the next India. Notwithstanding the decline in the level of FDI two decades. In the gradual convergence scenario infl ows to Sub-Saharan Africa in the aftermath of the presented in this chapter, gross capital inflows to global financial crisis, the long-term expansion of the the region will reach $25 4 billion in 2030 (figure global economy and rich natural resource endowments B3.4.1), compared with $62 billion in 2012. As a of countries in the region are likely to keep Sub-Saharan share of gross infl ows to developing countries, Sub- Africa’s FDI inflows at robust levels in the years ahead. Saharan Africa will rise from 5.4 percent in 2012 to Guinea alone contains 26 percent of the world’s known 8.2 percent in 2030 under the gradual convergence bauxite reserves; the Democratic Republic of Congo scenario. Although the share of developing-country accounts for 45 percent of cobalt reserves; Botswana, infl ows going to Sub-Saharan Africa is not expected the Democratic Republic of Congo, and South Africa to reach the peaks observed during periods of severe together represent 58 percent of industrial diamond economic tension in recent years, it will remain well reserves; and South Africa accounts for more than 95 above the level observed during most of the 2000s. percent of platinum-group metals (USGS 2012). Strong The largest single stream of gross capital infl ows global demand for energy indicates that international to Sub-Saharan Africa is in the form of FDI, and much capital will continue to flow to oil development projects, of this is directed to the natural resources sector—oil even in challenging governance environments such as in Angola, Nigeria, and other West and Central African that in South Sudan. countries; copper in the Democratic Republic of Congo Although extractive industries are set to be an and Zambia; bauxite in Guinea; and natural gas in a important target of capital inflows to Sub-Saharan number of countries. In recent years, the level of FDI Africa in the future, there are also signs of increasing inflows has been driven in part by the rapid expansion of investor interest in financing the infrastructure, retail, FIGURE B3.4.1 Gross capital inflows to Sub-Saharan Africa will rise as a share of inflows to developing countries 300 10 9 250 8 7 200 2010 $ billions 6 Percent 150 5 4 100 3 2 50 1 0 0 02 04 06 08 10 12 14 16 18 20 22 24 26 28 30 20 20 20 20 20 20 20 20 20 20 20 20 20 20 20 Volume of inflows to Inflows to Sub-Saharan Africa as a share of Sub-Saharan Africa (left axis) total inflows to developing countries (right axis) Sources: World Bank projections, supplemented with calculations using data in the IMF International Financial Statistics database. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 133 BOX 3.4 (continued) and financial services sectors. Infrastructure financ- Asia) in other developing regions. The large proportion ing is especially important because insufficient power of young people in Africa means that the region is at an generation, lack of basic water and sanitation systems, advantage in terms of accumulating savings over the overcrowded urban roads, and ineffi cient ports pose long term and in terms of having a large proportion of major constraints to productivity and international trade working-age population for decades to come. in the region and have a significant impact on liveli- One clear signal of the world’s growing assurance hoods. Reversing these defi ciencies will require sig- of the power of the African consumer was Walmart’s nificant amounts of capital from domestic and foreign $2.4 billion purchase of the South African retail giant sources, but at the same time infrastructure improve- Massmart in 2011. Although intraregional investment ments promise to make doing business in Africa signifi - currently accounts for a small proportion of total FDI cantly easier. infl ows—only 4 percent (in terms of value) of cross- Estimates produced for chapter 1 indicate that border mergers and acquisitions and greenfield FDI aggregate water and sanitation sector investment projects undertaken in Sub-Saharan African countries needs among Sub-Saharan African countries will rise between 2003 and 2010 were funded by other coun- from $2.2 billion in 2012 to $3.8 billion in 2030. This fig- tries in the region—such projects can help convince ure will be dwarfed by investment needs in the power companies elsewhere in the world considering invest- sector, expected to nearly triple, from $4.9 billion to ing in Africa to move forward. Good examples include $13.8 billion, over the same years. Roads will require an high-profile investments such as Mauritius-based estimated $11.2 billion of investment in 2030. Nigeria, SEACOM’s $600 million investment in an underwater with the region’s largest (and rapidly growing) popula- fiber-optic cable that linked South Africa, Mozambique, tion, will require power sector investment of almost Tanzania, and Kenya to Europe and South Asia in 2009. three-and-a-half times its 2012 level in 2030. Relatively Ecobank, for another example, a bank operating in rapid population growth in Sub-Saharan Africa means some 30 Sub-Saharan Africa countries, was among the that infrastructure financing needs will grow faster top FDI investors in the region during 2003–11 (Ernst & than in other developing regions. At the same time, the Young 2012). growing proportion of global economic growth attribut- Looking ahead, there may be an adjustment in the able to developing countries suggests that other devel- type of capital going to Sub-Saharan Africa. A long- oping countries will be a key source of capital flows to standing trend is that most capital inflows to the region Sub-Saharan Africa in the decades ahead. China and have been in the form of FDI. But in recent years, India are already active investors in infrastructure proj- access to bank lending and international bond markets ects and telecommunications projects in the region. has improved. Figure B3.4.2 tracks capital inflows in the Another driver behind the growth in capital inflows form of cross-border bank loans, clearly depicting a ris- to Sub-Saharan Africa over the next two decades will ing trend. In 2007, the peak year for cross-border capi- be the region’s rising per capita incomes. The expec- tal fl ows, nine international syndicated loans of more tation that Africa’s emerging middle class will support than $1 billion went to borrowers in Angola, Mauritius, a range of consumer goods industries—processed Nigeria, and South Africa in the retail, telecommuni- food and beverages, ready-to-wear clothing, cosmet- cations, construction, oil, and mining industries. The ics, household appliances, and home improvement—is lenders in these transactions are broadly based geo- increasingly becoming reality for multinational compa- graphically. Despite dropping in 2008 in the face of nies and retail banks, their confi dence was bolstered the global fi nancial turmoil, the volume of bank lend- by the very visible success of the mobile phone indus- ing has increased in each subsequent year. Here, too, try on the continent. Just as important are key demo- there is evidence of growing intraregional involvement: graphic statistics: 42 percent of Sub-Saharan Africans Standard Bank group, the largest commercial bank in are under the age of 15, versus a range of 17 percent Sub-Saharan Africa, was among the top three arrangers (Eastern Europe and Central Asia) to 32 percent (South of syndicated loans in 2011 (Bloomberg 2012). (continued) 134 Capital Flows in the Third Age of Financial Globalization Global Development Horizons BOX 3.4 (continued) FIGURE B3.4.2 Cross-border bank loans to Sub-Saharan Africa are on the rise 35 Cross-border bank loans, $ billions 30 25 20 15 10 5 0 1995 1997 1999 2001 2003 2005 2007 2009 2011 Syndicated Bilateral Source: World Bank calculations using data in the Dealogic Loan Analytics database. Note: The figure includes intra-African cross-border loans. Raising capital on international bond markets has For all the positivity, Sub-Saharan Africa faces unde- also become a possibility for some African countries. niable challenges in creating an environment suitable The governments of Gabon, Ghana, Namibia, Senegal, for large capital inflows, not least of which is the rela- and the Seychelles raised a total of $3.18 billion on tive underdevelopment of its domestic banking sector. international bond markets between 2006 and 2011. Only 24 percent of people age 15 and above have an Though few corporations based in the region have account at a formal financial institution, compared with access to markets outside their home countries, the 33 percent in South Asia and 55 percent in East Asia increased ability of sovereigns to issue debt on inter- and the Pacific (World Bank Global Financial Inclusion national markets is likely to pave the way for enhanced database). a Domestic credit to the private sector as a corporate access in the future. A similar narrative can percentage of GDP is less than 30 percent for most be found in access to international capital markets: Sub-Saharan African countries, and in several countries only 37 Sub-Saharan African companies are listed on that rate is in the single digits. Investors still shy away equity markets outside the region—17 on European from some countries because of entrenched corruption exchanges and 11 on U.S. exchanges (Bloomberg data- and other risks. In the immediate future, increased con- base). That said, international investors have demon- sumer spending by the middle class will remain con- strated increasing interest in African capital markets, centrated in commodity-exporting and other relatively albeit from a low starting point. wealthy countries. a. For all three of the regions mentioned, the figure does not include countries classified as high-income by the World Bank. Thus, for Sub-Saharan Africa, Equatorial Guinea is excluded. have been one factor behind the predominance and transaction costs. 27 These constraints can of FDI in inflows to the developing world, along generally be expected to relax as income levels with trade barriers, shallow financial markets, rise, but these changes do involve shifts in policy. barriers to competition in the economy, and, at Policy makers should recognize that initiatives the broadest level, informational asymmetries designed to deepen and better integrate financial Global Development Horizons Capital Flows in the Third Age of Financial Globalization 135 markets may also promote portfolio inflows, and FIGURE 3.10 Historically, the composition of capital that regulatory institutions should be strength- inflows has differed between income groups, with FDI ened in tandem with liberalization, given the providing stability in flows to developing countries volatility of private capital flows (portfolio flows a. Advanced countries in particular). In addition to changes in the composition of 12 capital flows that are likely to accompany the 10 increasing share of inflows and outflows attribut- Gross capital inflows, $ trillions able to developing countries, the two scenarios 8 of gross capital f lows in this chapter indicate that the coming decades are likely to see much 6 greater South-South capital flows (as illustrated in online annex 3.5). GDH 2011 documented the 4 rise of South-South FDI, which is already under way and will become an increasingly important 2 component of investment in developing countries 0 in the future (World Bank 2011). As developing countries’ banks grow in size and become more –2 internationally active, as their stock and bond markets continue to develop, and as their finan- –4 cial markets become more regionally integrated, 02 9 1 4 7 0 3 6 05 08 11 9 8 8 8 9 9 9 20 20 19 19 19 19 19 19 19 20 20 they will have an opportunity to intermediate a greater share of South-South flows and, broadly, Loans Portfolio FDI to play a much larger role in intermediating global capital flows than they do today. At present, more than half of capital outflows b. Developing countries from developing countries are in the form of 1.4 government reserve accumulation, while exter- nal investment activity in advanced countries 1.2 Gross capital inflows, $ trillions is mainly accounted for by private sector enti- ties. But reserve accumulation among developing 1.0 countries can be expected to decelerate, perhaps dramatically, as more countries move toward float- 0.8 ing exchange rate regimes and as capital markets 0.6 are allowed to become more open, meaning that developing countries will become more similar to 0.4 advanced countries in terms of the composition of their capital outflows. China will be a major 0.2 determinant of the speed and extent of developing countries’ changing capital outflow composition. 0.0 –0.2 02 9 Capital flows and monetary 1 4 7 0 3 6 05 08 11 9 8 8 8 9 9 9 20 20 19 19 19 19 19 19 19 20 20 policy spillovers Loans Portfolio FDI As discussed above, in the years ahead the vol- Source: IMF International Financial Statistics database. umes of gross capital flows to and from develop- Note: Each component depicted is the annual net change in liabilities to nonresidents. Loans consist mainly of lending by banks and nonbank financial institutions, but also include any ing countries are likely to expand dramatically. other type of liabilities to nonresidents not counted as FDI or portfolio investment. Portfolio The developing world’s financial markets will investment includes equity and debt portfolio investment. 136 Capital Flows in the Third Age of Financial Globalization Global Development Horizons become deeper, more sophisticated, and more foreign exchange markets. The first three of these globalized. These economies’ capital flows and sources, at least, will be in flux over the coming stocks of international assets and liabilities will two decades as developing countries come to not only grow in size but also change in terms account for a much greater share of global capital of their composition, rates of return, and risk flows than they do today, as their domestic finan- profi les. South-South investment will increase. cial markets mature and become more sophisti- Developing countries’ banks and capital markets cated, and as South-South flows increase. will intermediate a growing share of global capi- As of the end of 2010, the bulk of capital flows tal flows, both between developing countries and were still conducted in U.S. dollars and euros; between developing and advanced economies. 83 percent of international bonds outstanding An implication of these developments is that and 80 percent of international banking assets developing countries may become more exposed were denominated in one of these two currencies than ever to spillovers of monetary policy shocks (figure 3.11). Despite the increasing importance through their international balance sheets. of the euro, the U.S. dollar remains the domi- However, the dominant position of today’s global nant international reserve currency. The share of fi nancial centers will be eroded to some extent. reserves held in dollars is more than double the Thus, the configuration of the sources and chan- share of reserves held in euros.28 nels of international shocks may become more In the current multipolar international mon- dispersed than today, reducing the exposure of etary system—in which the U.S. dollar and euro the world economy to shocks originating in any serve as the principal reserve currencies, most one economy and offsetting the potential for emerging economies operate under some vari- greater instability that developing countries face ant of a managed floating exchange rate regime, as their financial markets expand and integrate. and there is high capital mobility across national borders—advanced countries with international currency status have a disproportionate inf lu- The international monetary ence on global monetary conditions. 29 Should system is anchored to the very low interest rates (such as those currently in U.S. dollar and the euro place in key currency countries) be maintained Use of the national currencies of a few major over an extended period, they could potentially advanced countries as the world’s dominant inter- lead to excessive risk-taking behavior on the national currencies has been a prominent feature part of fi nancial market participants, including of the post–World War II global financial system. banks, and stimulate portfolio and banking flows Currently, the international monetary system is to developing countries as investors and fi nan- anchored to the U.S. dollar and, to a lesser extent, cial intermediaries “reach for yield” and seek to the euro (Dailami and Masson 2011). Th is situ- maintain returns. Under these conditions, policy ation has long had strong relevance for develop- makers in developing countries face a difficult ing countries, which rely more than advanced challenge in deciphering whether capital inflows countries on credit denominated in foreign cur- are the result of improvements in their domestic rencies. Monetary policy may be neutral in the fundamentals or rather are a transitory effect trig- long run, but at any given time prevailing mon- gered by low yields in advanced countries, subject etary conditions have real effects, and the insta- to reversal when world interest rates rise. bility that can result from an inability to control One gauge of the extent of the international monetary conditions is an important issue facing domain of the U.S. dollar and euro in interna- developing-country policy makers. A long-term tional monetary affairs is the volume of interna- shift in which currencies are most prominent at tional bank claims denominated in those curren- the global level, as well, has policy implications. cies, which is particularly relevant for the large Demand for international currency originates number of developing countries where interna- from four sources: accumulation of official for- tional banks account for the bulk of capital flows. eign exchange reserves, international bank credit, As the international banking industry has grown, international securities issues, and turnover in the U.S. dollar has persistently played a major Global Development Horizons Capital Flows in the Third Age of Financial Globalization 137 role as the preferred currency for global banking FIGURE 3.11 International assets and foreign exchange transactions, although there are signs that this trading are mainly denominated in the U.S. dollar and the may change in the future as the dollar gradually euro gives way to other currencies (World Bank 2011). 100 It is interesting to note that loans by non-U.S. 90 banks, mainly European banks, account for a sig- 80 nificant part of dollar-denominated loans. 70 Global credit denominated in both dollars and 60 euros grew rapidly in the decade leading to the Percent 50 2007–09 global fi nancial crisis, although it has 40 been fairly flat from 2008 to 2011 (figure 3.12). U.S. dollar credit stood at $31.4 trillion at the end 30 of 2011, up from $12.6 trillion at the end of 1999. 20 Global credit in euros was €23.9 trillion ($31.4 10 trillion) at the end of 2011, up from €11 trillion 0 ($11.1 trillion) at the end of 1999. Although the International International Reserve Foreign banking bond market allocation exchange amounts of credit outstanding globally that are tradinga denominated in dollars and euros are roughly equal, dollar-denominated credit makes up a U.S. dollar Euro Pound sterling Swiss franc Yen Others substantially greater share of credit outside of the Sources: World Bank calculations using data in the BIS Locational Banking Statistics and IMF United States than the euro does outside the Euro COFER databases. Area—$11.8 trillion compared with $4.4 trillion. Note: Data are for 2010. Denomination of assets is calculated in terms of shares of the global stocks of banking assets and bonds for which the denomination was reported to the BIS, One implication is that because the U.S. dollar and shares of reserves for which the denomination was reported to the IMF (in 2010, these is more internationalized, the global economy is accounted for 56 percent of total reported reserve holdings). more sensitive to U.S. monetary policy than it is a. Because each foreign exchange transaction involves two currencies, the shares of transac- tion volume in which each currency was involved sum to 200 percent of the total volume. to Euro Area monetary policy, in part through the international banking system. For emerging-market economies, the ability it is likely that some developing countries will to insulate domestic monetary policy depends on become major players in international fi nancial their willingness to fully accept the exchange rate intermediation. Regarding how monetary policy movements that would result from changes in the spillovers will evolve, these trends in capital flows monetary policy stance of international currency have two main, countervailing implications: First, countries. Because many countries fear the result- were the present composition of capital flows (in ing volatility, they intervene to limit the exchange terms of type and currency denomination) to rate changes through accumulation of foreign remain roughly the same through 2030, as devel- exchange reserves. Though countries may be suc- oping countries become more deeply integrated in cessful in the short run in sterilizing the effects global financial markets, the impact of U.S. and on domestic liquidity, empirical evidence shows euro monetary policy on these countries would that there are limits to sterilization. In the cur- increase. Second, developing countries’ financial rent environment, for example, this means that markets are likely to become deeper and more some of the U.S. monetary expansion is likely to sophisticated in the years ahead, meaning that appear domestically in emerging-market coun- their capital markets and banks have the poten- tries as well.30 tial to intermediate a growing share of the world’s capital flows, and key developing countries’ mon- etary policy may become more influential on a The international monetary global scale. Monetary spillovers from key devel- system will be in flux in the future oping countries to other developing countries Looking forward, developing countries as a whole are particularly plausible given that South-South will account for a much greater share of gross cap- capital flows are expected to increase in the com- ital inflows and outflows than they do today, and ing years. Th is change, on its own, would mean 138 Capital Flows in the Third Age of Financial Globalization Global Development Horizons FIGURE 3.12 The U.S. dollar is still the main currency in global credit markets a. Dollars 2011 total: $31.4 trillion 35 30 Credit to non-U.S. residents: 25 $11.8 trillion $ trillions (2011) 20 15 Credit to U.S. 10 residents: $19.5 trillion 5 (2011) 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Claims on non-U.S. residents Cross-border claims on U.S. residents U.S. domestic credit by non-U.S. banks U.S. domestic credit by U.S. banks b. Euros 2011 total: $30.9 trillion 35 Credit to 30 non-Euro Area residents: 25 $4.4 trillion $ trillions 20 (2011) Credit to Euro Area 15 residents: 10 $26.4 trillion (2011) 5 0 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 2011 Claims on non-Euro Area residents Cross-border claims on Euro Area residents Euro Area domestic credit by non-Euro Area banks Euro Area domestic credit by Euro Area banks Sources: World Bank calculations using data in the BIS Consolidated Banking Statistics and IMF International Financial Statistics databases. Note: U.S. (Euro Area) domestic credit by non-U.S. (non-Euro Area) banks is measured as local currency claims on local residents of the United States (Euro Area) using BIS Consolidated Banking Statistics data. Domestic credit provided by home banks is estimated by subtracting local currency claims on local residents from total domestic credit using IMF International Financial Statistics data. Cross-border claims on residents are from the BIS Locational Banking Statistics. Claims on nonresidents are approxi- mated by adding cross-border claims on nonresidents and local claims on nonresidents from the BIS Locational Banking Statistics; this underestimates this component slightly because it omits credit to nonresidents by their countries’ domestic banks denominated in the currency in question; data on this subcomponent are not readily avail- able, but it is likely to be quite small relative to foreign banks’ claims in the currency. that the share of developing countries’ capital is uncertain. One can even imagine a scenario in flows denominated in advanced-country curren- which greater liberalization and integration of cies would likely shrink as a growing share are developing countries’ financial markets proceeds intermediated by developing country–based capi- hand in hand with a trend toward greater con- tal markets and banks. solidation and concentration of global fi nancial Thus, the evolution of developing countries’ intermediation, with only a few countries in the exposure to advanced countries’ monetary policy world specializing in fi nancial services. In such Global Development Horizons Capital Flows in the Third Age of Financial Globalization 139 a scenario, developing countries might not play The costly string of banking and sovereign a significant role in international intermedia- debt crises the world has experienced in recent tion despite becoming increasingly important as years has left policy makers keenly aware of the a source and destination of international capital trade-off between the instability that can accom- flows. The 2007–09 global financial crisis appears pany a highly complex and integrated global to have interrupted a trend toward greater con- financial system and the benefits of deepen- centration. Whether a reversion to this trend ing global integration via capital flows. Capital occurs depends, in part, on whether policies are inf lows can be key in supporting economic pursued at the national and regional levels to pro- growth, and open cross-border financial channels mote the deepening and strengthening of local are essential in channeling capital from coun- financial markets. tries with excess savings to productive projects It is likely, in any case, that developing coun- in countries with insufficient savings. Over the tries will have a more central place in the global years, the steady integration of developing coun- financial system in the years ahead than they do tries into the international financial system by today. China, in particular, has the capacity to way of international capital flows has provided have a much larger impact on global interest rates investors with an important means of diversifi- and other monetary conditions if the renminbi cation. But it is the risks of an increasingly inte- follows a path toward becoming an international grated international financial system featuring currency.31 And as developing countries account the movement of capital across countries on a for a greater share of global gross capital inflows massive scale, and how to best manage the system and outflows, the composition of those flows may to avoid future crises, that continue to dominate very well change as well. Reserve accumulation the minds of fi nancial sector policy makers and will likely diminish relative to private flows as regulators at the moment. developing countries move toward more flexible Several major concerns lie at the forefront exchange rate regimes. This will likely mean that of policy discussions related to the stability of fewer foreign assets will be denominated in dol- the international fi nancial system: (a) the long- lars and euros;32 even more important, returns on standing underpricing of sovereign risk in some foreign assets will rise as reserves are displaced advanced economies, (b) the global impact of by private investments intermediated by more unprecedented easy monetary policy in coun- mature financial markets. tries containing major fi nancial centers, and (c) the ability of the international financial system to fulfi ll its primary task of intermediating capital across national borders in a manner that promotes Conclusion and policy productive investment and sustained growth implications rather than fueling asset bubbles and instabil- The rapid expansion in the volume of global capi- ity. Regardless of the short-term versus long-term tal flows expected during the next two decades nature of each of these issues, the magnitude of in the two scenarios presented in this chapter the expected expansion in the volume of interna- will render the world’s economies more finan- tional capital flows through 2030, in both abso- cially interconnected than at any time in his- lute amounts and as a percentage of GDP, means tory. Interpreting how this increased integration that having an appropriate policy framework in will affect the global economy in the long term, place to manage international capital flows will or how policy makers should react to it, is by no become even more crucial in the future. means a straightforward task. History can cer- There is broad agreement that central banks tainly provide guidance, but there is also value and other economic planning authorities must be in estimating the scope of international finan- well equipped to respond quickly and resolutely cial links in the future, and additionally in envi- to mitigate contagion across countries during sioning potential policy responses that might be times of fi nancial distress, and must be vigilant appropriate. in monitoring and acting upon conditions that 140 Capital Flows in the Third Age of Financial Globalization Global Development Horizons contribute to instability. Expanding the capacity around the world set up a large network of liquid- of economic policy makers to act in a coordinated ity swap lines starting in late 2007 (figure 3.13). manner at the international level in monitoring Under these arrangements, the home central and responding to episodes of instability is, by bank of currencies in short supply globally at the many accounts, also important. Although inter- time could provide its currency to commercial national coordination efforts are complicated by a banks operating in other countries, with the cen- host of political economy and other issues, crises tral banks of the countries in which the commer- offer strong incentive to overcome such obstacles. cial banks were located operating as intermediar- ies (Allen and Moessner 2010). The People’s Bank of China, for instance, agreed to provide a total Financial crises can provide of RMB 650 billion under swap lines with the strong incentive for international central banks of Argentina; Belarus; Hong Kong monetary policy and financial SAR, China; Indonesia; the Republic of Korea; stability coordination and Malaysia.33 After implementing several prior An obvious place to look for evidence on these rounds of swap agreements with other major policy matters is the global financial crisis of sev- central banks, the U.S. Federal Reserve agreed eral years ago, which saw unprecedented coor- to swap lines of unlimited size with the Bank dination of monetary policy among national of England, Bank of Japan, European Central monetary authorities. For example, central banks Bank, and Swiss National Bank for a period of FIGURE 3.13 A number of new central bank liquidity swap lines were extended between December 2007 and December 2010 Poland Switzerland ECB Hungary United Norway Kingdom Canada Iceland Sweden Estonia United States Mexico Denmark Latvia India Brazil Singapore Philippines New Australia Zealand Japan Malaysia Indonesia Korea, Rep. Belarus Thailand China United Arab Argentina Emirates Source: Based on Allen and Moessner 2010, with updates. Note: Color of the arrows indicates the swap network, generally according to the central bank at the center of the network: blue is Switzerland, green is the ECB (European Central Bank), orange is China, red is Japan, and black is the United States; aqua arrows show a multilateral swap net- work between the central banks of China, Japan, and the Republic of Korea. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 141 six months starting in October 2008 (the height New modes of cross-country of the crisis). cooperation may be beneficial Global and regional organizations responded as the international financial to the crisis by increasing their volume of lend- system becomes ever more ing to client countries. Among multilateral devel- interconnected opment banks, significant capital infusions also allowed more resources to be devoted to devel- Moving forward, a potentially fundamental opment initiatives in their member countries.34 question is whether the current decentralized International negotiations surrounding postcrisis model—in which national (or regional) central measures to reduce the probability of future finan- banks have control over their own monetary pol- cial crises have been led by organizations such as icy and financial regulation is crafted at the coun- the International Monetary Fund (IMF) and the try level—is adequate for the task of managing an Group of 20 Finance Ministers and Central Bank increasingly globalized financial system. The real- Governors (G-20). The IMF’s revised Flexible ity is that any model must fit into the constructs Credit Line, for example, together with its newly of a world made up of sovereign states. In such a established Precautionary and Liquidity Line, are world, perhaps the best way to arrive at outcomes intended to improve the institution’s capacity to that serve the needs of the most countries possible assist countries heavily affected by the crisis and is to encourage cooperation at both the bilateral to prevent future crises. In 2010, the IMF also and multilateral levels, while acknowledging that launched three new concessional financing facili- a high degree of coordination is not always politi- ties specifically for low-income countries. The cally feasible and that national policies will con- IMF’s proposed Global Stabilization Mechanism tinue to be key in shaping the world’s monetary would provide liquidity in conjunction with and financial policy framework. bilateral and regional arrangements with the spe- The scope for cooperative efforts is broad and cific goal of arresting contagion across countries could take a number of forms. One form is partic- during financial crises. ipation in multilateral initiatives—such as those International commitment to preventing epi- on monetary and financial stability and bank- sodes of fi nancial contagion is currently being ing supervision led by the Bank for International put to the test by the ongoing sovereign debt Settlements—or public commitments to time- crisis in the Euro Area. In one sense, the world lines for policy objectives within the G-20. One is signaling its recognition of the gravity of the potential contribution of the G-20 process would situation by allocating additional resources to be to emphasize the spillover and mutual depen- safety nets. As of the June 2012 meeting of the dence aspects of key currency countries’ financial G-20, 37 IMF member countries had contrib- sector and monetary policy, and to ensure that uted a total of $456 billion to the institution’s key currency countries move away from a nar- fi rewall fund, which the institution intends to row policy focus on just their own economies. disburse only as a secondary line of defense after Synthesizing the efforts of multilateral institu- resources available via its standard quota arrange- tions might also be called for. ments and the New Arrangements to Borrow are Another form of cooperation is at the bilateral significantly drawn down. Developing countries level, such as strengthening and institutionalizing have been remarkably active in contributing to the swap lines that provide liquidity to non-key this fund, with the BRICS, along with Mexico currency countries during periods of instabil- and Turkey, pledging a combined $95 billion of ity or sharing knowledge and experience across the total. This level of involvement by developing countries. Recent crises in advanced economies countries in international crisis prevention and hold valuable lessons for developing countries resolution mechanisms will become increasingly whose financial markets will become increas- important as developing countries account for a ingly sophisticated and globally integrated in growing share of gross global capital flows in the the years ahead. Regulatory institutions must years to come. be more forward-looking, with systems in place 142 Capital Flows in the Third Age of Financial Globalization Global Development Horizons to monitor new innovations in financial mar- authorities in emerging economies—facing kets in real time—for example, by making this strong capital inflows as a result of low interest an explicit part of the job description for bank rates in major economies—may feel it is neces- supervisors—and to quickly evaluate and adapt sary to ease interest rates to discourage inflows to innovations before they cause systemic risk. even if domestic macroeconomic conditions do The bottom line is that cooperation in moni- not otherwise call for such a move, although toring and managing the international financial the inflationary pressure of lowering rates must system cannot be overemphasized. In a highly also be considered. Macroprudential responses interconnected system, international cooperation might also be considered, perhaps targeting a spe- in managing the system is key for its stability. cific sector, such as real estate, that is subject to Multilateral institutions will continue to have a overheating. role in this management in the future, particu- Faced with strong demand for their assets, larly in helping countries establish the conditions even economies with f loating exchange rates supporting the development and functioning of may resort to capital controls, such as Brazil and financial markets and assisting in coping with Taiwan, China, did in 2009, or as Switzerland periods of crisis or near-crisis. But initiatives implicitly began to do in 2011 when it commit- within international forums must be considered ted to maintaining an exchange rate floor of the in light of efforts by authorities in individual Swiss franc against the euro. In the longer term, countries. Regarding the response to the global countries reliant on capital inflows representing fi nancial crisis, the unprecedented provision of a large share of their GDP risk fiscal distress and liquidity by major central banks to help resolve recession should there be a sudden turnaround the crisis has fueled heated debate about the scale in those flows, at which point monetary authori- and nature of risks being taken in the process ties must step in to assist. It could be argued of unwinding their positions. At the same time, that there is potential danger for policy makers there is concern that extremely low interest rates to err on the side of overly strict capital controls, in advanced economies could result in an over- choking off access to finance, when what is really abundance of capital inflows to emerging and needed is careful reform of regulatory institutions developing countries as investors search for yield and greater international regulatory coordination. elsewhere, with potentially destabilizing effects in Understanding the nature of capital inflows to those countries in the future. emerging and developing economies—whether they are driven by good economic fundamentals within the country or by more transitory forces As developing countries come such as low yields in advanced economies or to account for half or more of heavy use of carry trades—can help policy mak- the world’s capital outflows, ers determine the best response. Deciphering their policies regarding capital whether fundamental or transitory forces are flows are likely to become more more powerful in driving capital inf lows to a influential at the global level country is no doubt a difficult task, but one that For a given country, the challenges of managing has clear implications for policy. In the case of the surges of capital inflows are by now well under- former, it has been argued that allowing exchange stood and continue to be relevant in understand- rate appreciation is the most appropriate response; ing how emerging and developing countries may in the latter, monetary, fiscal, and financial poli- react to accommodative monetary policy con- cies should be employed (Sidaoui 2011). ditions that persist in advanced economies. For Recognizing the potential impacts of prevail- emerging economies with fi xed exchange rates, ing conditions in advanced economies on the excessive capital inflows can present a difficult rest of the world is no doubt essential to craft- choice between allowing currency appreciation ing policy responses to those conditions. But and intervening in foreign exchange markets to how might policy responses need to change in hold down the value of the currency. Monetary the long term? The two scenarios presented in Global Development Horizons Capital Flows in the Third Age of Financial Globalization 143 this chapter anticipate sharp transitions in the currencies implies that developing countries will regional patterns of capital flows through 2030. become less affected by monetary policy spill- Under both scenarios, the share of global capital overs from any one country. This could be stabi- inflows going to Europe and the United States is lizing on a global level because liquidity shocks projected to decline substantially, while the share will be more diversified, but it could also be going to China will increase considerably. A simi- destabilizing as it becomes more difficult to assess lar trend is expected in terms of the share of capi- the timing and extent of monetary policy spill- tal outflows. Several future policy implications overs. Regardless of the currency composition of follow from the scenarios. capital flows in the future, an increasing share of Broadly, the course of global monetary and global flows going to and from developing coun- financial policy making will need to be adjusted tries indicates that these countries should have a at national and international levels to account for larger role in management of capital flows at the the implication that developing countries will international level, within both bilateral and mul- become responsible for an expected half or more tilateral organizations. of the world’s capital outflows. The two scenarios There are also exchange rate regime implica- presented in this chapter envisage that develop- tions of a potential shift in the currency composi- ing countries will account for 49–58 percent of tion of global capital flows. With more diversifi- the world’s global capital outflows by 2030, rising cation in the currency makeup of capital flows, from 28 percent in 2010. Th is shift implies that countries with fi xed exchange rate regimes may developing countries’ monetary and financial sec- fi nd that it is more sustainable to peg to a bas- tor policies could be more influential at the global ket of currencies (for example, the dollar, the level—although, as noted previously, this notion euro, and the renminbi) than to any one of those must be balanced against the currency composi- currencies. tion of global capital flows. The rise in the volume of global capital flows Among developing countries, China is unique in the decades ahead will increase the influence in its potential to position its currency as a truly of foreign investors on capital accumulation in global currency during the coming years. Should many countries. For some countries, it is conceiv- the increase in the share of global capital flows able that most of their total investment will be attributable to China be accompanied by greater determined by overseas investors and creditors in use of the renminbi as the currency denominat- the future, thus bringing an even greater inter- ing these flows (internationalization of the ren- national dimension to fi nancial policy-making minbi is a stated policy goal in China), then efforts. The efforts of the G-20, IMF, and BIS China’s monetary policy stance can be expected will be key in developing a financial supervisory to exert considerably stronger influence on the architecture that enhances global financial mar- international financial system than it does at ket stability through, by other measures, avoiding present. Under such a scenario, the world would excessive volatile speculative capital transactions. also experience a decrease, to some extent, in Policy makers will need to monitor the chan- the dominance of U.S. and Euro Area mon- nels of capital inf lows (bank loans, portfolio etary policy on the global stage. One conceivable investment, and FDI) because these channels arrangement is that the U.S. dollar, the euro, and have different capacities for stabilizing or desta- the renminbi will all be globally significant cur- bilizing conditions in recipient countries. There rencies by 2030 (World Bank 2011). Of course, is wide speculation that capital markets will alternate scenarios consider that either the dollar intermediate a growing amount of global capital remains preeminent in global finance during the flows in the future and that banks will account next two decades, or there is no clearly preemi- for less. Whether this will be good news for coun- nent currency by 2030. tries that currently receive most of their inflows For small and medium-size developing coun- in the form of bank loans remains to be seen; tries, a world in which the United States, the bank lending tends to be highly procyclical and Euro Area, and China all have key international generally less supportive of risk sharing than FDI 144 Capital Flows in the Third Age of Financial Globalization Global Development Horizons or equity portfolio investment (Brunnermeier Estimates for the informal agricultural sector in et al. 2012). But, in middle-income countries, Ghana, for example, place nominal rates there at portfolio investment has historically been even as much as 350 percent (Udry and Anagol 2006), more volatile, in relative terms (that is, adjusting and real returns to microenterprises in Sri Lanka have been placed at several percentage points for the smaller magnitude of this component of higher than prevailing market interest rates (De developing countries’ inflows), than bank lending Mel, McKenzie, and Woodruff 2008). (Broner et al. 2013).35 Though the share of bank 4. Some approaches effectively lump land and natural loans in global capital flows may well recede in resources together with capital by estimating the the long term, it must also be acknowledged that income share of capital as one minus the income loans will continue to represent a significant por- share of labor, in which case the income share of tion of global capital flows for years to come. The capital is overestimated. Caselli and Feyrer (2007) many efforts under way to improve regulation of estimate that only about half of this amount, on the international banking sector thus will remain average, should actually be attributed to reproduc- highly relevant. ible capital, but this follows from estimates of the Finally, as growing amounts of capital are contribution of land and natural resources to pro- duction that may be overly aggressive. transferred among developing countries, there 5. Details are reported in online annex 3.1, available at is even capacity for greater regional monetary http://www.worldbank.org/CapitalForTheFuture. policy spillover from large emerging economies 6. Averages for each income group were computed by such as Brazil and Russia. These countries are not weighting yields for each country by their relative expected to have the global reach that China may, share of the group’s capital stock. but it is conceivable that their influence could 7. The notional investment demand equation is cal- become increasingly apparent at the regional level. culated across time at the country level, holding Certainly, in a multipolar world, South-South yields constant at 2014 levels, and this is summed monetary policy coordination will become more across countries to estimate a path of global critical in promoting stable financial and macro- notional investment demand. This straightfor- economic conditions in developing countries. ward measure of notional demand abstracts from the endogeneity of variables such as output on the right-hand side of the factor demand equation, as these variables follow the path generated by Notes the model, which is codetermined with the real- 1. As in other recent scholarship, gross capital ized path of investment as opposed to the path of inflows in this report are measured as purchases notional investment demand. In the special case of domestic assets by foreigners net of sales of when, globally, yields tend to change little over domestic assets by foreigners, and gross outflows time, the distortion from not accounting for endo- are measured as purchases of foreign assets by geneity is quite minor because notional demand domestic residents net of sales of foreign assets by barely differs from realized demand in this case. domestic residents. Net capital flows are defined as 8. More precisely, 2030 global notional investment gross inflows minus gross outflows. demand is within 0.04 percent of the supply of 2. Lucas (1990) noted that if positive spillovers from saving. Note that, at the global level, the supply of human capital accumulation are sufficiently great capital will be equal to the pool of saving, and this and sufficiently local (spilling over more within represents the notional global supply of saving for countries than across borders), then differences investment because—by the assumptions of the in human capital might eliminate differences model—saving is not affected by yields. between countries in the return to capital entirely. 9. More precisely, 2030 global notional demand However, not enough progress has been made in exceeds global supply by 2.6 percent; this pressure quantifying the scope of external benefits from corresponds to a cumulative increase in global knowledge spillovers or the degree to which these average yields (weighted by capital stocks) of benefits spill across borders to resolve this question roughly half a percentage point between 2014 and empirically. 2030. 3. Interest rates can also vary as widely between sec- 10. It is also important to recognize that a scenario with tors within countries as they do between countries. rapid growth alone (unaccompanied by changes in Global Development Horizons Capital Flows in the Third Age of Financial Globalization 145 structural factors) does not result in global saving is greater, which raises country-specific relative and investment deviating significantly from the yields and hence capital inflows to finance realized gradual convergence scenario. This underscores investment; second, realized investment is also the importance of allowing structural variables greater because developing countries can attract to experience substantial change—as is consid- more investment financing as their financial ered in the rapid convergence scenario—to grasp development and institutional quality improve the full extent of potential transformations in the substantially. developing world for global saving and investment 15. Under the rapid convergence scenario, high- outcomes. income countries’ investment falls from 17.3 per- 11. Higgins (1998) and Lane and Milesi-Ferretti cent of the group’s GDP in 2010 to 14.9 percent in (2002) point to potential effects that changes in 2030, compared with 16.4 percent in 2030 under the age structure of populations have on current the gradual convergence scenario. The group’s account balances, but the theoretical direction of saving rate in 2030 does not differ significantly the effect is ambiguous because aging can affect between the two scenarios, so the difference in the both saving and investment negatively (and poten- group’s aggregate current account balance is due tially nonlinearly). In the scenario analysis in this mainly to this difference in investment rates. report, the impact of demographic change on net 16. Blanchard and Milesi-Ferretti (2010) distinguish capital flows is not directly identified, but the between “good imbalances” and “bad imbal- model generating the scenario results does incor- ances,” and provide a summary of how domestic porate projected changes in the age structure. Fur- and systemic distortions can give rise to poten- ther details are given in online annex 1.6. tially destabilizing imbalances. The main role of 12. Details of how the two scenarios are constructed national-level policy makers is to address domestic can be found in online annex 1.6. Details of the distortions of incentives to save and invest in their current account balance projection calculations economies, whereas there is also a role for interna- are in online annex 3.2. tional policy coordination in addressing systemic 13. In the CGE model, the parameters in the invest- distortions—for example, by improving global ment equation are calibrated uniformly across liquidity provision. countries based on results from a panel regression, 17. In the Linkage model, national saving does with an implicit assumption that the future path depend negatively on the degree of social security of investment will be determined similarly across protection. Details are given in online annexes 1.5 countries. Details are given in online annexes and 1.6. 1.5 and 1.6. For China, investment has histori- 18. Advanced economies might reduce current account cally been higher than would be predicted by the deficits by increasing public saving and encourag- model, so its projected path reflects an adjustment ing private saving by avoiding negative real inter- down to what would be expected for a typical est rates over extended periods (see, for example, country, given the paths of the right-hand-side Blanchard and Milesi-Ferretti 2010). variables. 19. Lane and Milesi-Ferretti (2001, 2007), for exam- 14. The relatively large differential in high-income ple, document a transition in the composition of versus developing-country yields described earlier emerging markets’ international balance sheets to raises the Lucas (1990) question of why capital a large FDI and portfolio equity component of can flow “uphill” from the South to the North liabilities, with heavy reserve accumulation on the in the model. The reason for this is that the key asset side. endogenous variables in the World Bank’s Link- 20. At the peak in 2007, global gross capital flows age model are calibrated to observed 2007 data, amounted to $11 trillion—an amount equiva- allowing surpluses (deficits) to coexist with posi- lent to 19 percent of world GDP—up from $0.4 tive (negative) yield differentials. Consequently, trillion in 1980 and approximately $3.7 trillion only subsequent movements of yields, along with in 2000. At the nadir in 2009, total global gross the other determinants of investment financing, capital flows had dropped to $1.8 trillion. have the potential to alter the path of net capi- 21. In 2010, for example, among majority foreign- tal flows. The mitigation of the Lucas paradox owned banks in developing countries, 34.6 per- relative to levels today in the rapid convergence cent (350 banks) had another developing country scenario thus has two interrelated reasons: first, as the main nationality of ownership, up from investment demand in many developing countries 29.3 percent (212 banks) in 2000 (authors’ 146 Capital Flows in the Third Age of Financial Globalization Global Development Horizons calculations based on bank shareholder data on could lead to more inflows in the form of bonds banks in Bankscope). denominated in domestic currency. 22. Didier, Hevia, and Schmukler (2011) show, how- 27. Hooper and Kim (2007) estimate the association ever, that the crisis affected economic performance of various measures of transparency with the three in high-income and developing countries simi- main components of capital inflows to devel- larly when differences in precrisis growth rates are oping countries and find that the relationships accounted for, and also that there was wide varia- vary widely, both by dimension of transparency tion among developing countries, with Eastern and by type of inflow, so that, broadly speak- European countries among the most affected and ing, an increase in transparency over time can low-income countries relatively isolated from the be expected to affect the composition of capital global crisis. flows. For example, an index of competitiveness 23. For example, Martin and Rey (2004) build a is found to have a much larger positive association theoretical model with transaction costs in which with portfolio flows than with FDI, and a negative financial assets are imperfect substitutes and the association with bank lending; stock market depth number of assets is endogenous, which gives rise similarly has a positive association with portfolio to a link between market size and integration, flows and a negative association with bank lending and asset returns. Another formal model of how (and, in this case, an association with FDI that is greater integration of financial markets can drive not statistically significant). an expansion of capital flows is developed by 28. These shares are based only on the portion of Evans and Hnatkovska (2005); the model implies reserves voluntarily identified by central banks to a relationship between integration and changes in the International Monetary Fund to be denomi- the composition and volatility of capital flows as nated in a specific currency. In 2010, the currency of well. 56 percent of total reserve holdings was identified. 24. Lucas (1990) compares international diversifica- 29. These effects may not be additive because liquidity tion to that predicted by a portfolio model and in each of the currencies may not be easily substi- reviews a range of possible sources of home bias. tutable, particularly during times of financial dis- Kraay et al. (2005) incorporates sovereign risk in tress, as experienced during the recent crisis when such a model to reconcile the predictions of the acute dollar shortage in international markets model to the data. A growing body of literature pushed up the costs of borrowing in foreign cur- supports the idea that informational asymmetries rencies and swapping the proceeds into dollars. In are important in explaining capital flows; see, particular, the intensive funding pressures expe- for example, Kang and Stulz (1997); Ahearne, rienced by European banks increased the euro- Griever, and Warnock (2004); and Leuz, Lins, dollar implied swap basis spread. and Warnock (2009). 30. A second source of monetary policy spillover is 25. The scenarios for outflows are calculated as a resid- that, given the importance of the United States ual of the gross inflow and current account pro- in the world economy, spillovers from the United jections, rather than being explicitly modeled (as States may influence economic conditions in other in online annex 3.4), and thus are admittedly less countries, such that business cycles become syn- theoretically grounded. Some factors that drive chronized and corresponding fiscal and monetary residents’ capital outflows from a country are not policy stances become similar across countries. taken into account, and these may differ signifi- For instance, a demand shock in the United States cantly from the factors that drive foreigners’ capi- may increase both U.S. economic activity and tal flows into the country and the determinants U.S. imports, inducing a rise in interest rates in of current accounts. To the extent that income the United States and abroad. growth and the development and global integra- 31. This outcome depends on political commitment in tion of financial markets are likely drivers of gross China to move toward full currency convertibil- outflows, however, these effects are accounted for ity and institutional transparency. An expanded by the modeling strategy via their intermediary discussion of the conditions necessary for the ren- impact on gross inflows. minbi to become an international currency is in 26. In addition, the currency makeup of the compo- World Bank (2011). nents themselves may evolve with financial market 32. In the context of the present uncertainty sur- development. For example, in some larger develop- rounding the institutional arrangement of the ing countries, domestic bond market development Euro Area, it should be noted that the extent to Global Development Horizons Capital Flows in the Third Age of Financial Globalization 147 which the euro will be used as an international 310, Bank for International Settlements, Basel, currency over the next one or two decades depends Switzerland. to some extent on the resolution of inconsisten- BIS Consolidated Banking Statistics, Bank for Interna- cies between the Euro Area’s monetary and fiscal tional Settlements, Basel, Switzerland. institutions. BIS Locational Banking Statistics, Bank for Interna- 33. East Asian countries had a network of swap lines tional Settlements, Basel, Switzerland. already in place, part of the legacy of the 1997–98 Blanchard, Olivier J., and Gian Maria Milesi-Ferretti. Asian financial crisis, although the People’s Bank 2010. “Global Imbalances: In Mid-Stream?” Dis- of China created new swap lines and expanded cussion Paper 7693, Center for Economic and Pol- existing ones (Allen and Moessner 2010). icy Research, Washington, DC. 34. The International Monetary Fund, tradition- Bloomberg. 2012. “Banks Battling European Debt ally considered the global lender of last resort for Crisis Lose on African Deals.” May 8. http://www countries facing financial crises, tripled its lending .bloomberg.com/news/2012-05-07/banks-battling- capacity in response to the crisis. Multilateral devel- european-debt-crisis-lose-on-african-deals.html. opment banks (MDBs), as well, expanded their vol- Bloomberg database, New York. http://www.bloom ume of lending to client countries. The increase in lending by MDBs was possible following significant berg.com. capital increases within many of the institutions. In Bordo, Michael D., and Barry J. Eichengreen. 2003. March 2011, the International Bank for Recon- “Crises Now and Then: What Lessons from the Last struction and Development’s board of governors Era of Financial Globalization.” In Monetary His- approved an $86.2 billion (31 percent) increase in tory, Exchange Rates and Financial Markets: Essays in subscribed capital, including $5.1 billion over the Honor of Charles Goodhart, vol. 2, ed. Paul Mizen, next five years. Shareholders of the Asian Devel- 52–91. London: Edward Elgar. opment Bank agreed in April 2009 to triple the Broner, Fernando A., Tatiana Didier, Aitor Erce, and institution’s capital base, from $55 billion to $165 Sergio Schmukler. 2013. “Gross Capital Flows: billion, the largest general capital increase among Dynamics and Crises.” Journal of Monetary Eco- the major MDBs, including a 4 percent increase of nomics 60 (1): 113–33. paid-in capital over the next five years. In 2010, the Brunnermeier, Markus, José de Gregorio, Barry Inter-American Development Bank approved a $70 Eichengreen, Mohamed El-Erian, Arminio Fraga, billion (70 percent) increase in general capital, and Takatoshi Ito, Philip R. Lane, Jean Pisani-Ferry, the African Development Bank approved a $63 bil- Eswar Prasad, Raghuram Rajan, Maria Ramos, lion (200 percent) increase. Hélène Rey, Dani Rodrik, Kenneth Rogoff, Hyun 35. At the same time, rapid adjustment of asset prices Song Shin, Andrés Velasco, Beatrice Weder di means that foreign holders of equity absorb much Mauro, and Yongding Yu. 2012. “Banks and Cross- of a negative shock before they can exit, and thus Border Capital Flows: Policy Challenges and Reg- may be less likely to exit in response to the shock. ulatory Responses.” Report of the Committee on However, investors face high costs of monitoring a International Economic Policy and Reform, Brook- fund manager’s medium- to long-run performance, ings Institution, Washington, DC. and this can tend to induce a focus on short-term Caballero, Ricardo J., and Arvind Krishnamurthy. returns and herding behavior (De la Torre, Ize, and 2009. “Global Imbalances and Financial Fragility.” Schmukler 2011; World Bank 2012). American Economic Review 99 (2): 584–88. Caselli, Francesco, and James D. Feyrer. 2007. “The Marginal Product of Capital.” The Quarterly Journal References of Economics 122 (2): 535–68. Ahearne, Alan G., William L. Griever, and Francis E. Chamon, Marcos D., and Eswar S. Prasad. 2010. “Why Warnock. 2004. “Information Costs and Home Are Saving Rates of Urban Households in China Bias: An Analysis of U.S. Holdings of Foreign Rising?” American Economic Journal: Macroeconom- Equities.” Journal of International Economics 62 (2): ics 2 (1): 93–130. 313–36. Chinn, Menzie D., and Hiro Ito. 2006. “What Matters Allen, William, and Richhild Moessner. 2010. “Cen- for Financial Development? Capital Controls, Insti- tral Bank Co-operation and International Liquidity tutions, and Interactions.” Journal of Development in the Financial Crisis of 2008–9.” Working Paper Economics 81 (1): 163–92. 148 Capital Flows in the Third Age of Financial Globalization Global Development Horizons ———. 2007. “Current Account Balances, Financial Monetary Fund, Washington, DC. http://www Development and Institutions: Assaying the World .imf.org/external/np/sta/cofer/eng/index.htm. ‘Saving Glut’.” Journal of International Money and Jones, Matthew T., and Maurice Obstfeld. 2001. “Sav- Finance 26 (4): 546–69. ing, Investment, and Gold: A Reassessment of His- Dailami, Mansoor, and Paul Masson. 2011. “Prospects torical Current Account Data.” In Money, Capital for a Multipolar International Monetary System.” Mobility, and Trade: Essays in Honor of Robert A. Report 2011:13, Dansk Institut for Internationale Mundell , ed. Guillermo A. Calvo, Rudiger Dorn- Studier (DIIS), Copenhagen. busch, and Maurice Obstfeld, 303–63. Cambridge, De la Torre, Augusto, Allain Ize, and Sergio Schmuk- MA: MIT Press. ler. 2011. Financial Development in Latin America Kang, Jun-Koo, and Rene M. Stulz. 1997. “Why Is and the Caribbean: The Road Ahead . Washington, There a Home Bias? An Analysis of Foreign Portfo- DC: World Bank. lio Equity Ownership in Japan.” Journal of Finan- De Mel, Suresh, David McKenzie, and Christopher cial Economics 46 (1): 3–28. Woodruff. 2008. “Returns to Capital in Micro- Kraay, Aart, Norman Loayza, Luis Servén, and Jaume enterprises: Evidence from a Field Experiment.” The Ventura. 2005. “Country Portfolios.” Journal of the Quarterly Journal of Economics 123 (4): 1329–72. European Economic Association 3 (4): 914–45. Dealogic Loan Analytics database, Dealogic. Lane, Philip R., and Gian Maria Milesi-Ferretti. 2001. Didier, Tatiana, Constantino Hevia, and Sergio L. “The External Wealth of Nations: Measures of Schmukler. 2011. “How Resilient and Counter- Foreign Assets and Liabilities for Industrial and cyclical Were Emerging Economies to the Global Developing Countries.” Journal of International Financial Crisis?” Policy Research Working Paper Economics 55 (2): 263–94. 5637, World Bank, Washington, DC. ———. 2002. “Long-Term Capital Movements.” In Domeij, David, and Martin Flodén. 2006. “Popula- NBER Macroeconomics Annual 2001, vol. 16 , ed. tion Aging and International Capital Flows.” Inter- Ben S. Bernanke and Kenneth Rogoff, 73–136. national Economic Review 47 (3): 1013–32. Cambridge, MA: MIT Press. Ernst & Young. 2012. “Building Bridges: Ernst & ———. 2007. “The External Wealth of Nations Mark Young’s 2012 Attractiveness Survey: Africa.” Sur- II: Revised and Extended Estimates of Foreign vey report, Ernst & Young. Assets and Liabilities, 1970–2004.” Journal of Inter- Evans, Martin D. D., and Viktoria Hnatkovska. 2005. national Economics 73 (2): 223–50. “International Capital Flows, Returns and World Leuz, Christian, Karl V. Lins, and Francis E. Warnock. Financial Integration.” Working Paper 11701, 2009. “Do Foreigners Invest Less in Poorly Gov- National Bureau of Economic Research, Cam- bridge, MA. erned Firms?” Review of Financial Studies 22 (8): Gilpin, Robert. 1987. The Political Economy of Interna- 3245–85. tional Relations. Princeton, NJ: Princeton Univer- Lin, Justin Yifu, and Mansoor Dailami. 2010. “Com- sity Press. ments on Dealing with Global Imbalances.” In Gollin, Douglas. 2002. “Getting Income Shares Reconstructing the World Economy, ed. Olivier Right.” Journal of Political Economy 110 (2): 458–74. Blanchard and Il SaKong, 111–20. Washington, Higgins, Matthew. 1998. “Demography, National DC: IMF and Korea Development Institute. Savings, and International Capital Flows.” Interna- Lucas, Robert E., Jr. 1990. “Why Doesn’t Capital Flow tional Economic Review 39 (2): 343–69. from Rich to Poor Countries?” American Economic Honohan, Patrick. 2008. “Cross-Country Variation in Review 80 (2): 92–96. Household Access to Financial Services.” Journal of Martin, Philippe, and Hélène Rey. 2004. “Financial Banking and Finance 32 (11): 2493–2500. Super-Markets: Size Matters for Asset Trade.” Jour- Hooper, Vince, and Suk-Joong Kim. 2007. “The Deter- nal of International Economics 64 (2): 335 –61. minants of Capital Inflows: Does Opacity of Recip- Obstfeld, Maurice, and Kenneth Rogoff. 2007. “The ient Country Explain the Flows?” Economic Systems Unsustainable U.S. Current Account Position 31 (1): 35–48. Revisited.” In G7 Current Account Imbalances: Sus- IMF International Financial Statistics database, Inter- tainability and Adjustment, ed. Richard H. Clarida, national Monetary Fund, Washingtion, DC. http:// 339–76. Chicago: University of Chicago Press. www.imf.org/external/data.htm. Obstfeld, Maurice, and Alan M. Taylor. 2003. “Glo- IMF COFER (Currency Composition of Official For- balization and Capital Markets.” In Globalization in eign Exchange Reserves) database, International Historical Perspective, ed. Michael D. Bordo, Alan M. Global Development Horizons Capital Flows in the Third Age of Financial Globalization 149 Taylor, and Jeffrey G. Williamson, 121–90. Chicago: World Bank. 2011. Global Development Horizons 2011: University of Chicago Press. Multipolarity: The New Global Economy. Washing- Portes, Richard, and Hélène Rey. 2005. “The Deter- ton, DC: World Bank. minants of Cross-Border Equity Flows.” Journal of ———. 2012. Latin America Copes with Volatility: International Economics 65 (2): 269–96. The Dark Side of Globalization. Washington, DC: Sidaoui, José. 2011. “Challenges for Emerging Market World Bank. Economies.” Paper No. 68, Bank for International World Bank Global Economic Monitor database, Settlements, Basel, Switzerland. World Bank, Washington, DC. http://data.world Taylor, Alan M. 1992. “External Dependence, Demo- bank.org/data-catalog/global-economic-monitor. graphic Burdens, and Argentine Economic Decline World Bank Global Financial Inclusion (Global Fin- After the Belle Époque.” Journal of Economic History dex) Database, World Bank, Washington, DC. 52 (4): 907–36. http://go.worldbank.org/1F2V9ZK8C0. Udry, Christopher, and Santosh Anagol. 2006. “The World Bank World Development Indicators data- Return to Capital in Ghana.” Working Paper 932, base, World Bank, Washington, DC. http://data. Economic Growth Center, Yale University, New worldbank.org/data-catalog/world-development- Haven, CT. indicators. USGS (U.S. Geological Survey). 2012. “Mineral Com- modity Summaries 2012.” Annual report, U.S. Geological Survey, Reston, VA. E C O -AU DI T Environmental Benefits Statement The World Bank is committed to preserving Saved: endangered forests and natural resources. • 7 trees The Office of the Publisher has chosen to • 3 million Btu of total print Capital for the Future on recycled energy paper with 50 percent post-consumer waste, • 582 lb. of net greenhouse in accordance with the recommended stan- gases dards for paper usage set by the Green Press • 3,155 gal. of water Initiative, a nonprofit program supporting • 211 lb. of solid waste publishers in using fiber that is not sourced from endangered forests. For more informa- tion, visit www.greenpressinitiative.org.