WORLD BANK DISCUSSION PAPER NO. 402 Work in progress WDP402 for public discussion August 1999 Trade, Global Policy, and the Environment Per G. Fredriksson, editor Recent World Bank Discussion Papers No. 334 Managing Price Risk in the Pakistan Wheat Market. Rashid Faruqee and Jonathan R. Coleman No. 335 Policy Optionsfor Reform of Chinese State-Owned Enterprises. Edited by Hlarry G. Broadman No. 336 Targeted Credit Programs and Rural Poverty in Bangladesh. Shahidur Khandker and Osman H. Chowdhury No. 337 The Role of Family Planning and Targeted Credit Programs in Demographic Change in Bangladesh. Shahidur R. Khandker and M. Abdul Latif No. 338 Cost Sharing in the Social Sectors of Sub-Saharan Africa: Impact on the Poor. Arvil Van Adams and Teresa Hartnett No. 339 Public and Private Roles in Health: Theory and Financing Patterns. Philip Musgrove No. 340 Developing the Nonfarm Sector in Bangladesh: Lessonsfrom Other Asian Countries. Shahid Yusuf and Praveen Kumar No. 341 Beyond Privatization: The Second Wave of Telecommunications Reforms in Mexico. Bjorn Wellenius and Gregory Staple No. 342 Economic Integration and Trade Liberalization in Southern Africa: Is There a Rolefor South Africa? Merle Holden No. 343 Financing Private Infrastructure in Developing Countries. David Ferreira and Karman Khatami No. 344 Transport and the Village: Findingsfrom African Village-Level Travel and Transport Surveys and Related Studies. Ian Barwell No. 345 On the Road to EU Accession: Financial Sector Development in Central Europe. Michael S. Borish, Wei Ding, and Michel Noel No. 346 Structural Aspects of Manufacturing in Sub-Saharan Africa: Findings fromn a Seven Country Enterprise Survey. Tyler Biggs and Pradeep Srivastava No. 347 Health Reform in Africa: Lessonsfrom Sierra Leone. Bruce Siegel, David Peters, and Sheku Kamara No. 348 Did External Barriers Cause the Marginalization of Sub-Saharan Africa in 'World Trade? Azita Amjadi Ulrich Reincke, and Alexander J. Yeats No. 349 Surveillance of Agricultural Price and Trade Policy in Latin America during Major Policy Reforms. Alberto Valdes No. 350 Who Benefitsfrom Public Education Spending in Malawi: Resultsfrom the Recent Education Reform. Florencia Castro-Leal No. 351 From Universal Food Subsidies to a Self-Targeted Program: A Case Study in Tunisian Reform. Laura Tuck and Kathy Lindert No. 352 China's Urban Transport Development Strategy: Proceedings of a Symposium in Beijing, November 8-10,1995. Edited by Stephen Stares and Liu Zhi No. 353 Telecommunications Policiesfor Sub-Saharan Africa. Mohammad A. Mustafa, Bruce Laidlaw, and Mark Brand No. 354 Saving across the World: Puzzles and Policies. Klaus Schmidt-Hebbel and Luis Serven No. 355 Agriculture and Germtian Reunification. Ulrich E. Koester and Karen M. Brooks No. 356 Evaluating Health Projects: Lessonsfrom the Literature. Susan Stout, Alison Evans, Janet Nassim, and Laura Raney, with substantial contributions from Rudolpho Bulatao, Varun Gauri, and Timothy Johnston No. 357 Innovations and Risk Taking: The Engine of Reform in Local Government in Latin America and the Caribbean. Tim Campbell No. 358 China's Non-Bank Financial Institutions:Trust and Investment Companies. Anjali Kumar, Nicholas Lardy, William Albrecht, Terry Chuppe, Susan Selwyn, Paula Perttunen, and Tao Zhang No. 359 The Demandfor Oil Products in Developing Countries. Dermot Gately and Shane S. Streifel No. 360 Preventing Banking Sector Distress and Crises in Latin America: Proceedings of a Conference held in Washington, D.C., April 15-16,1996. Edited by Suman K. Bery and Valeriano F. Garcia No. 361 China: Power Sector Regulation in a Socialist Market Economy. Edited by Shao Shiwei, Lu Zhengyong, Norreddine Berrah, Bernard Tenenbaum, and Zhao Jianping No. 362 The Regulation of Non-Bank Financial Institutions: The United States, the European Union, and Other Countries. Edited by Anjali Kumar with contributions by Terry Chuppe and Paula Perttunen No. 363 Fostering Sustainable Development: The Sector Investment Program. Nwanze Okidegbe No. 364 Intensified Systems of Farming in the Tropics and Subtropics. J.A. Nicholas Wallis No. 365 Innovations in Health Care Financing: Proceedings of a World Bank Conference, March 10-11, 1997. Edited by George J. Schieber No. 366 Poverty Reduction and Human Development in the Caribbean: A Cross-Country Study. Judy L. Baker No. 367 Easing Barriers to Movement of Plant Varietiesfor Agricultural Development. Edited by David Gisselquist and Jitendra Srivastava (Continued on the inside back cover) WORLD BANK DISCUSSION PAPER NO. 402 Trade, Global Policy, and the Environment Per G. Fredriksson, editor The World Bank Washington, D.C. Copyright © 1999 The International Bank for Reconstruction and Development/THE WORLD BANK 1818 H Street, N.W. Washington, D.C. 20433, U.S.A. All rights reserved Manufactured in the United States of America First printing August 1999 Discussion Papers present results of country analysis or research that are circulated to encourage discussion and comment within the development community. The typescript of this paper therefore has not been prepared in accordance with the procedures appropriate to formal printed texts, and the World Bank accepts no responsibility for errors. Some sources cited in this paper may be informal documents that are not readily available. The findings, interpretations, and conclusions expressed in this paper are entirely those of the author(s) and should not be attributed in any manner to the World Bank, to its affiliated organizations, or to members of its Board of Executive Directors or the countries they represent. The World Bank does not guarantee the accuracy of the data included in this publication and accepts no responsibility for any consequence of their use. The boundaries, colors, denominations, and other information shown on any map in this volume do not imply on the part of the World Bank Group any judgment on the legal status of any territory or the endorsement or acceptance of such boundaries. The material in this publication is copyrighted. The World Bank encourages dissemination of its work and will normally grant permission promptly. Permission to photocopy items for internal or personal use, for the internal or personal use of specific clients, or for educational classroom use, is granted by the World Bank, provided that the appropriate fee is paid directly to Copyright Clearance Center, Inc., 222 Rosewood Drive, Danvers, MA 01923, U.S.A., telephone 978-750-8400, fax 978-750-4470. Please contact the Copyright Clearance Center before photocopying items. For permission to reprint individual articles or chapters, please fax your request with complete information to the Republication Department, Copyright Clearance Center, fax 978-750-4470. All other queries on rights and licenses should be addressed to the World Bank at the address above or faxed to 202-522-2422. ISSN: 0259-21OX Per G. Fredriksson is a consultant at the World Bank Library of Congress Cataloging-in-Publication Data Trade, global policy, and the environment / edited by Per G. Fredriksson. p. cm. - (World Bank discussion paper ; no. 402) Papers presented at a World Bank conference in April 1998. Includes bibliographical references (p. ). ISBN 0-8213-4458-7 1. Free trade-Environmental aspects-Congresses. 2. International trade-Environmental aspects-Congresses. 3. Foreign trade regulation-Environmental aspects-Congresses. 4. Environmental policy-Economic aspects-Congresses. 5. Economic development-Environmental aspects-Congresses. 6. Free trade- Environmental aspects-Developing countries-Congresses. I. Fredriksson, Per. II. World Bank. III. Series: World Bank discussion papers; 402. HF1703.T73 1999 333.7-dc2l 99-10251 CIP Contents Foreword ix Abstract xi List of Contributors xiii Chapter I Trade, Global Policy, and the Environment: New evidence and issues 1 Per G. Fredriksson 1. Introduction 1 2. Background- Scale, Composition, and Technique Effects 1 3. Overviewx of the Volume Papers 3 4. Policy Conclusions and Recommendations 11 Chapter 2 Will Trade Liberalization Harm the Environment?: The case of Indonesia to 2020 13 Anna Strutt and Kym Anderson 1. Introduction 13 2. Adding an Environmental Module to the Projections Model 19 3. Empirical Projections of Environmental Impacts in Indonesia of Structural and Policy Changes to 2020 20 4. Conclusions 29 Appendix 31 Chapter 3 Trade, Environment, and Public Health in Chile: Evidence from an economy-wide model 35 John Beghin, Brad Bowland, Sebastien Dessus, David Roland-Holst, and Dominique van der Mensbrugghe 1. Introduction 35 2. The TEQUILA Model 36 3. A Brief Description of the Santiago Health Model 37 4. Policy Reform Scenarios 38 5. Resultsfrom Policy Reform Simulations 39 6. Conclusions 47 Appendix 49 iii iv Trade, Global Policy, and the Environment Chapter 4 Testing the Impact of Trade Liberalization on the Environment: Theory and evidence 55 Judith M. Dean 1. Introduction 55 2. The Links between Trade and the Environment- Econometric Evidence 56 3. A General Model of Trade and the Environment 57 4. An Alternative Heckscher-Ohlin-Vanek Specification 58 5. An Application to China 59 6. Conclusion 61 Chapter 5 Industrial Pollution in Economic Development: Kuznets revisited 65 Hemamala Hettige, Muthukumara Mani, and David Wheeler 1. Introduction 65 2. Development and Industrial Pollution 66 3. Data 68 4. Econometric Results 69 5. Implications of the Results 75 6. Summary and Conclusions 77 Appendix 79 Chapter 6 Preferential Trading Arrangements between Kenya and the EU: A case study of the environmental effects of the horticulture sector 83 Anil Markandya, Lucy Emerton, and Sam Mwale I. Introduction 83 2. The Horticultural Sector in Kenya 84 3. Tariff Structurefor Horticultural and Related Products 88 4. Environmental Costs of Different Crops in Kenya 88 5. An Economic Analysis of Environmental Impacts 92 6. Estimating the Impact of Preferential Trading on Land Use, the Environment, and other Indicators 94 7. Conclusions 97 Chapter 7 Fuel Prices, Woodlands, and Woodfuel Markets in the Sahel: An integrated economic-ecological model 101 Kenneth M. Chomitz, Charles Griffiths, and Jyotsna Puri 1. Introduction 101 2. Background 102 3. Model 104 4. Results 106 5. Discussion and Conclusions 108 Appendix 110 Chapter 8 In Search of Pollution Havens? Dirty industry in the world economy, 1960-1995 115 Muthukumara Mani and David Wheeler 1. Introduction 115 2. Development, Regulation, and "Pollution Havens" 116 3. Defining Dirty Industries 116 4. Pollution-Intensive Production in the OECD 118 5. Pollution-Intensive Production in Developing Asia and Latin America 122 6. Conclusions and Implications 126 Contents v Chapter 9 The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade 129 Paavo Eliste and Per G. Fredriksson 1. Introduction 129 2. Empirical Analysis 130 3. Empirical Results 132 4. Conclusion 134 Appendix 136 Chapter 10 Pollution and Capital Markets in Developing Countries 141 Susmita Dasgupta, Benoit Laplante, and Nlandu Mamingi 1. Introduction 141 2. Dataset 142 3. Event-Study Methodology 143 4. Empirical Results 146 5. Conclusion 148 Appendix 151 Chapter 11 The Credibility of Trade Sanctions in International Environmental Agreements 161 Scott Barrett 1. Introduction 161 2. Trade Leakage and Trade Linkage 162 3. Free-Rider Deterrence in Linked Games 164 4. The Strategy of Trade Sanctions in a Self-Enforcing International Environmental Agreement 164 5. Conclusions 168 Appendix 170 Chapter 12 The Importance of Trade for the Ratification of the 1992 Climate Change Convention 173 Per G. Fredriksson and Noel Gaston 1. Introduction 173 2. The Data on Legislative Delay 175 3. The Determinants of Legislative Delay 177 4. The Results 180 5. Conclusion 183 Appendix 185 Chapter 13 Achieving Carbon Emission Reductions through Joint Implementation 191 Will Martin 1. Introduction 191 2. Joint Implementation and Technical Change 192 3. Relevant Elasticities of Demandfor Energy and Emissions Intensities 195 4. Some Stylized Experiments 197 5. Supply Side Considerations 198 6. Conclusions and Policy Implications 199 Chapter 14 Carbon Abatement: Lessons from second-best economics 201 Ian W. H. Parry 1. Introduction 201 2. Resultsfrom Analytical Models 202 3. Resultsfrom Numerical Models 207 4. International Implications 209 5. Conclusion 211 vi Trade, Global Policy, and the Environment Figutres A3.1 Production nesting 49 5.1 Manufacturing share in GDP vs. per capita income, 1975-1994 70 5.2 Industrial BOD intensity vs. income per capita 71 5.3 Water pollution intensity vs. income per capita 74 5.4 Industrial pollution and economic development 76 5.5 Counterfactual simulations 76 6.1 Marginal profitability 94 7.1 Actual vs. predicted kerosene share 107 7.2 End-use energy shares as a function of kerosene price 107 7.3 End-use energy shares as a function of charcoal prices 108 8.1 Energy intensity in Japanese manufacturing 117 8.2 Land intensity in Japanese manufacturing 117 8.3 Investment/output ratios in Japanese manufacturing, 1972-1991 117 8.4 Capital/ output ratios in Japanese manufacturing, 1985-1989 117 8.5 Labor intensity in Japanese manufacturing 118 8.6 Japanese production ratio (polluting/non-polluting) 118 8.7 Japan's production of polluting products 118 8.8 Urban commercial land price index (1990 prices) 119 8.9 Japanese Environmental Agency budget 119 8.10 Pollution control investment by big enterprises 119 8.11 Consumption-production ratio of polluting products in Japan 120 8.12 Import-export ratio of polluting products in Japan 120 8.13 Production of polluting products in the United States and Canada 121 8.14 Production of polluting products in Europe 121 8.15 North America's import/export ratio 121 8.16 Europe's import/export ratio 121 8.17 United States-Canada consumption-production ratio 121 8.18 Euorpe's consumption-production ratio 121 8.19 Latin America's production of polluting products 122 8.20 Latin America's import/export ratio 122 8.21 Asia (excluding Japan): Polluting sector production share, 1964-1998 122 8.22 Asia (excluding Japan): Import/export ratio for polluting products, 1964-1994 123 8.23 Openness in Asia 123 8.24 Republic of Korea: Pollution-intensive industry share, 1963-1993 125 8.25 NIEs' production of pollution products 125 8.26 NIEs' import/export ratio 125 8.27 Developing East Asia's production of polluting products 125 8.28 Developing East Asia's import/export ratio 125 8.29 South Asia's production of polluting products 125 8.30 South Asia's production of polluting products 126 8.31 Consumption-production ratio in Asia (excluding Japan) 126 11.1 Provision of the public good 166 12.1 Survival function estimates 175 12.2 Hazard function estimates 176 13.1 Impacts of augmenting technical change on the demand for an input 194 14.1 Pre-existing taxes and the marginal cost of abatement 205 Contents vii Tables 2.1 Assumptions made in the projections: cumulative [and annual] percentage changes in GDP and factor endowments for the period 1992 to 2020 16 2.2 Percentage changes in sectoral output levels and in sectoral shares of GDP due to economic growth, Indonesia, 1992-2010 and 2010-2020 17 2.3 Percentage changes in sectoral output levels in Indonesia following Uruguay Round and APEC trade reform by 2010 and 2020 18 2.4 Recent and projected levels of atmospheric emissions in the base cases, Indonesia, 1992, 2010, and 2020 (kt) 21 2.5 Decomposition of changes in pollution as a consequence of economic growth and structural changes, Indonesia, 1992-2010 and 2010-2020 22 2.6 Recent and projected levels of water use and quality in the base cases, Indonesia, 1992, 2010, and 2020 23 2.7 Decomposition of pollution effects from Uruguay Round trade reform (including China), Indonesia, 2010 25 2.8 Sectoral decomposition of the total change in emissions due to Uruguay Round implementation, Indonesia, 2010 26 2.9 Decomposition of pollution effects in Indonesia under APEC liberalization, 2020 27 2.10 Decomposition of pollution effects in Indonesia under APEC liberalization, with 0.5 percent p.a. extra GDP growth in APEC economies, 2020 28 2.11 Percentage changes in resource-sector output levels in various regions of the world following Uruguary Round trade reform (including China), 2010 29 A2.1 Import tariff rates in Indonesia without and with Uruguay Round liberalization, by sector, 2010 31 3.1 Impact of policy reform on aggregate variables 40 3.2 Impact of policy reforms on national effluent emissions 42 A3.1 Impact of environmental policy reform on health endpoints for Santiago 50 A3.2 Impact of environmental policy reform on mortality and morbidity health damages for Santiago 53 4.1 The impact of openness on growth of emissions 61 5.1 Log (manufacturing share of total output) vs. Log (income per capita), 1975-1994 70 5.2 Log (sector-weighted BOD intensity) vs. Log (income per capita), 1975-94 71 5.3 Intensity equations for pollution and labor (prices and regulation) 72 5.4 Intensity equations for pollution and labor (prices and regulation) 72 5.5 Intensity equations for pollution and labor (in income per capita) 73 5.6 Income and pollution abatement 75 5.7 Industrial pollution and economic development: Simulation experiments 75 5.8 Trends in international emissions -Selected countries, 1977-1989 77 5.9 Estimated industrial BOD emissions-Selected countries, 1977-1989 77 6.1 Indicators for Asian vegetables 84 6.2 Indicators for French beans 85 6.3 Exports of fresh horticultural products: 1985-1993 85 6.4 Returns from horticultural crops 86 6.5 Land potential and environmental sensitivity of selected crops 87 6.6 Impact of removing preferential tariffs on profitability 88 6.7 Techniques for valuing major environmental impacts of agricultural production in Kenya 92 6.8 Summary of net environmental costs of agricultural production 93 6.9 Average profitability, land areas, and implied values of a and I 95 6.10 Impacts of full environmental cost internalization on land use 95 6.11 Impacts of introducing Mediterranean basin tariffs 96 6.12 Impacts of introducing third country tariffs 97 viii Trade, Global Policy, and the Environment Tables (continued) 7.1 Estimated relative fuel demands 106 A7.1 Base run 110 A7.2 Reduced kerosene price run 110 A7.3 Increased transport cost 111 8.1 Ranking of pollution-intensive industries 116 8.2 Openness and economic progress in selected developing countries in Asia and Latin America 124 8.3 Initial national environmental legislation in Asia 124 9.1 Regression results for the stringency of environmental regulations (STRING) 133 9.2 Regression results for government assistance (PSE) 133 A9.1 Variable definition and data sources 136 10.1 Capitalization of the stock market of Argentina, Chile, Mexico, and the Philippines, 1990-1994 142 10.2 Market concentration in the IFC general indexes, end-1994 143 10.3 Number of news, 1990-1994 143 10.4 Description of data set 144 10.5 Positive events 147 10.6 Government actions vs. other positive events 148 10.7 Negative events 149 10.8 Complaints vs. all other negative events 150 A10.1 Complete name of companies in sample set 151 A10.2 Reaction of market to positive news 152 A10.3 Reaction of market to negative news 154 12.1 Descriptive statistics for legislative delay times 176 12.2 Descriptive statistics for covariate analysis 178 12.3 Estimates of proportional hazards model -Effects of trade measures 181 13.1 Long run elasticities of demand in the industrial sector of the G-7 196 13.2 International Energy Agency estimates of energy and emissions data for the industrial sector of ANNEX I parties 197 13.3 Impacts of improvements in fuel use efficiency on total C02emissions 197 14.1 Differences between Pigouvian and second-best taxes 208 14.2 International comparison of tax rates 210 Foreword W A T e live in an increasingly interconnected country-specific. No two countries are identical. With world. Trade flows worldwide are grow- freer trade each country will specialize in (at least ing rapidly and global production pat- somewhat) different sectors than before trade was lib- terns are shifting as countries follow their comparative eralized. Since the impacts are country-specific, effec- advantage in production via trade. At the same time, tive policymaking requires that particular attention however, there is growing concern about potential ad- has to be paid to both the existing policies and future verse environmental impacts from increasing trade. economic structure of a country in order to anticipate Questions such as the following are commonly asked: outcomes and determine appropriate actions. In ad- • How does trade liberalization influence the envi- dition, policy analysis capacity in developing coun- ronment, and how does environmental policy in- tries is often weak. fluence trade flows? This volume attempts to increase the level and * Does increased economic integration induce a scope of our knowledge about the nexus between "race to the bottom" whereby policymakers seek trade liberalization, and local and global environmen- to improve the ability of domestic firms to com- tal quality. Several chapters in this volume address pete with their foreign rivals though the lowering these linkages, including the environmental and of environmental regulations? health effects of trade liberalization and consequent * What are possible ways to avoid or reduce any economic growth, the possible inverted U-shaped re- negative environmental effects from freer trade? lationship between growth and emissions, and the * How do trade measures, such as sanctions, influ- effects of trade distortions and environmental poli- ence a country's decision to participate in global cies on environmental damage from sectors such as environmental policymaking? horticulture and forestry. All of these concerns are To respond to these concerns policymakers are linked to the question of sustainable development. seeking guidance on the links between trade and their How can we ensure that countries grow sufficiently national environment, and appropriate policy re- fast to offer their populations improved economic sponses to address any adverse impacts. This is not conditions, without causing environmental and health an easy task, however. The environmental impacts of effects that are irreversible? the various bilateral, regional, or multilateral trade Another concern often raised in the trade and liberalization programs undertaken are sometimes environment debate is the possible existence of "pol- difficult to anticipate, and thorough analysis often lution havens" in developing countries. Will countries requires large amounts of data. This is particularly with relatively lax environmental regulations end up true because the impact of trade liberalization is very becoming new hosts for polluting firms migrating ix x Trade, Global Policy, and the Environment from richer countries, with supposedly more stringent discusses available policy tools to address global en- controls? Most existing studies have not found strong vironmental problems. support for this hypothesis. A number of chapters in Although this volume does not offer the final this volume add to the debate by studying the impor- solution to all of these issues, it does provide new tance of wages, subsidies, and capital markets in de- perspectives and thinking on these topics. The results termining industrial location. offered here should help policymakers create im- Subsidies, such as those commonly found in the proved local and global environmental and trade pol- agriculture and energy sectors, often have multiple icies. The importance of these issues can only grow in detrimental effects -on the environment, on govern- the future, as the world becomes an increasingly inte- ment finances, on world market prices, and on com- grated place, and trade flows help ensure improved petition. There is great scope for so-called "win-win" living standards for all parties. policies in this area, whereby subsidy reduction or The World Bank has been pleased to be a part of removal is good for both the economy and the envi- the dialogue and I would like to express my thanks to ronment. This is one area where free-trade proponents Per Fredriksson for leading this effort and editing this and nongovernmental organizations appear to have volume, to all of the authors and reviewers of the much in common and could potentially form a strong material included here, to the two keynote speakers coalition against protectionist forces. at the Trade, Global Policy, and the Environment Con- Not all environmental issues are local. The issue ference, Daniel Esty (Yale University) and David Reed of climate change is one challenging example of a glo- (World Wide Fund for Nature), and to Sida, the Swed- bal environmental concern. The costs of emission re- ish International Development Agency, for support- ductions are borne by populations living today, but ing this important work. the benefits will probably not be enjoyed until sever- al decades from now. How can the global community overcome this problem in an efficient and equitable John A. Dixon way? What role does trade policy have in this area? Environmental Economics and Indicators The volume addresses the linkages between trade, The Environment Department trade policies, and international agreements, and also The World Bank Abstract his Discussion Paper contains 13 chapters eral trade liberalization would result in significantly that contribute to knowledge in three broad higher economic gains than the former two, but would areas: the environmental effects of trade have a negative effect on the environment and on ur- liberalization and growth, the "pollution-haven" hy- ban morbidity and mortality. MERCOSUR has a sim- pothesis, and economic instruments for global envi- ilar, but less intense, effect. NAFTA integration would ronmental problems. Earlier versions of these chapters be environmentally benign in terms of several types were presented at a conference held at the World Bank of pollution emissions relative to the other trade inte- in April 1998. An introductory chapter gives a sum- gration scenarios. mary of the findings and policy conclusions. Dean's chapter develops a methodology for es- The chapter by Strutt and Anderson is a case timating the effects of openness on income growth, study simulating how Indonesia's environment would and of income growth on environmental damage. She be affected by two different reforms: the Uruguay uses water-pollution data from Chinese provinces and Round and Asian-Pacific Economic Cooperation finds that trade liberalization has both a direct and an (APEC) most-favored-nation trade liberalization. The indirect effect on emissions growth. Increased trade results suggest that air and water pollution would openness in China directly aggravates environmen- decline as a result of trade-policy reforms. However, tal damage by inducing an expansion of polluting incorporating an 0.5 increase in the rate of income sectors. However, income growth indirectly reduces growth would mean that APEC liberalization would emissions. yield up to 15 percent more air pollution and up to 12 Hettige, Mani, and Wheeler provide further ev- more water pollution by 2020. The authors argue that idence on the relationship between industrial water the economic surplus created could be used for envi- pollution and income. They find that total industrial ronmental protection. water pollution rises rapidly through middle-income The chapter by Beghin, Bowland, Dessus, Ro- status and remains approximately constant thereaf- land-Holst, and van der Mensbrugghe examines links ter, not following an inverted-U. Simulations show between Chilean trade policy, atmospheric pollution, that during the 1980s the Asian developing economies and public health in the Santiago metropolitan area. replaced Organisation of Economic Co-Operation and The authors examine three trade policy scenarios: (i) Development (OECD) countries as the world's larg- Chile's accession to the North American Free Trade est producers of industrial water pollution. Association (NAFTA), (it) Chile's accession to MER- Markandya, Emerton, and Mwale study the en- COSUR (Southern Cone Market), and (iii) complete vironmental effects in Kenya of removing European trade liberalization with the rest of the world. Unilat- Union preferential trading arrangements with Ken- xi xii Trade, Global Policy, and the Environment yan horticultural production, and of other environ- Multilateral environmental agreements are in- mental policy measures. Horticultural production creasingly seen as crucial policy tools with which to turns out to be largely insensitive to such measures. address global environmental problems. The chapter Chomitz, Griffiths, and Puri study the effect of trade by Martin discusses joint implementation, arguing and tax policies on deforestation in the Sahel. The that increases in fuel efficiency resulting from tech- authors' model shows that distortions in kerosene and nology transfers may have important second-round gasoline prices have almost no impact on the rate of effects. Increased efficiency would lower the price of woodland degradation. Only substantial kerosene energy relative to other inputs and lower the price of subsidies induce a shift towards this fuel. Efficiency- energy-intensive products relative to other products. enhancing policy reforms, while desirable on econom- If the fuel whose efficiency is being improved is ini- ic grounds, may sometimes have less environmental tially the most emissions-intensive, the aggregate ef- impact than hoped. fect of price changes may result in an increase in One issue that has attracted much attention is emissions, especially if this fuel continues to be the whether lower trade barriers will result in a special- most polluting variety. ization by developing countries in pollution-intensive The issue of trade sanctions in international en- industries. Mani and Wheeler's chapter argues that vironmental agreements is contentious. Barrett argues such "pollution-haven" effects are insignificant be- that sanctions are more likely to be used when non- cause production is primarily for the domestic mar- signatories could potentially develop a compara- ket, not for export. The increase in the developing tive advantage in pollution-intensive industries. In countries' share of dirty-sector production is attribut- such cases sanctions against nonsignatories increase ed to a highly income-elastic demand for basic indus- the supply of emission reductions and are more like- trial products. As income levels have increased, this ly to be credible. The chapter by Fredriksson and Gas- elasticity has declined, and the stringency of environ- ton addresses the issue of "political drag," or mental regulations has been raised. "regulatory chill;" that is, the hypothesis that because Two chapters argue that there may be other rea- of exposure to international competition countries sons why researchers have not found strong effects may delay the enactment of environmental legislation. of environmental regulations on plant locations and Overall, this argument is not supported by data from trade patterns. First, Eliste and Fredriksson present the 1992 Climate Change Convention. Institutional empirical evidence suggesting that producers may factors and emissions play a more important role in have been compensated by policymakers for their influencing the likelihood of cooperation on global environmental protection costs. This may explain the environmental policymaking, however. minimal effect on trade flows. Second, Dasgupta, Parry provides a survey of the literature on the Laplante, and Mamingi argue that making sufficient "double dividend;" the idea that emission taxes yield information available to capital markets may provide a double return by both reducing emissions and rais- financial and reputational incentives to invest pollu- ing revenues that can be used to reduce distorting tax- tion-reduction efforts in both developed and devel- es elsewhere in the economy. However, pre-existing oping countries. Developing-country capital markets tax distortions may render environmental policies were found to react favorably to announcements of more costly. Parry concludes that if the policy raises positive environmental performance. Multinational tax revenues that are used to reduce other taxes, such investments in pollution-intensive sectors in devel- additional cost can be reduced. oping countries may be affected by concerns over le- gal liability and reputational damage. List of Contributors Keynote Speakers Anna Strutt, University of Waikato Daniel Esty, Yale University David Wheeler, World Bank David Reed, World Wide Fund for Nature Discussants Authors Kym Anderson, University of Adelaide Kym Anderson, University of Adelaide Scott Barrett, London Business School Scott Barrett, London Business School Shantayanan Devarajan, World Bank John Beghin, Iowa State University Ishac Diwan, World Bank Brad Bowland, University of Minnesota William Easterly, World Bank Kenneth M. Chomitz, World Bank Bruce Gardner, University of Maryland Susmita Dasgupta, World Bank Noel Gaston, Bond University Judith M. Dean, Johns Hopkins University Jenny Ligthart, International Monetary Fund Sebastian Dessus, OECD Development Centre Anil Markandya, University of Bath Paavo Eliste, West Virginia University Alex Mourmouras, International Monetary Fund Lucy Emerton, Policy Research Group, Nairobi Mohan Munasinghe, World Bank Per G. Fredriksson, World Bank H&kan Nordstrom, World Trade Organization Noel Gaston, Bond University David Roland-Holst, Mills College Charles Griffiths, U.S. Environmental David Tarr, World Bank Protection Agency Scott Vaughan, World Trade Organization Hemamala Hettige, Asian Development Bank Session Chairs Benoit Laplante, World Bank Maureen Cropper, World Bank Nlandu Mamingi, University of the West Indies John A. Dixon, World Bank Muthukumara Mani, World Bank Hossein Farzin, University of California, Davis Anil Markandya, University of Bath Ricardo Melendez-Ortiz, International Centre for Will Martin, World Bank Trade and Sustainable Development Dominique van der Mensbrugghe, OECD Mohan Munasinghe, World Bank Development Centre Gunnel Nycander, Swedish International Sam Mwale, Policy Research Group, Nairobi Development Cooperation Agency Ian Parry, Resources for the Future John D. Shilling, World Bank Jyotsna Puri, World Bank Kamoji Wachiira, Canadian International David Roland-Holst, Mills College Development Agency xiii Chapter 1 Trade, Global Policy, and the Environment New evidence and issues Per G. Fredriksson 1. Introduction 2. Background -Scale, Composition, and he trade and environment debate intensified Technique Effects in the 1990s, in part because of the attention Although many issues in the trade and environment drawn to the issues by the North American debate are contentious, a consensus appears to be Free Trade Agreement (NAFTA) and the Uruguay emerging on a few matters. Many participants in the Round free trade negotiations. Interest has increased debate now agree that (a) more open trade improves further because many proposed solutions to the cli- growth and economic welfare, and (b) increased trade mate change problem have potential implications for and growth without appropriate environmental poli- the global trading system. Focus is likely to sharpen cies in place may have unwanted effects on the envi- as the next multilateral trade negotiation round ap- ronment. However, in some situations more open pears on the horizon, but the debate is often built on trade may also reduce pressure on the environment little or no empirical evidence. To inform the debate, This ambiguity occurs because trade policy and trade guide policymakers toward solutions, and help set flows have several conflicting effects on both the en- priorities more empirical work is clearly needed. vironment and resource use. It has proven useful to This volume is an attempt to further our under- view the various effects of trade liberalization in three standing of the empirical links between trade and the categories: scale, composition, and technique effects. This environment. The 13 papers included, which were is now a standard way of thinking about the problem presented at a World Bank conference in April 1998, and a helpful tool for analyzing the issues involved. focus on three main themes: Many of the papers in this volume are devoted to 1. Effects of trade liberalization and growth on the deepening our understanding and empirical knowl- environment edge of these effects. Only through a firm understand- 2. The "pollution haven" hypothesis ing of the linkages involved can well-founded policy 3. Economic instruments for resolving global envi- advice be formulated. ronmental problems. The papers address a number of different issues Scale effect within each of the themes, offering new data or new questions and approaches. This introductory chapter The scale effect refers to the fact that more open trade provides an overview of the issues raised and the main creates greater economic activity, thus raising the de- insights and policy conclusions obtained. In the fol- mand for inputs such as raw materials, transporta- lowing section some of the main linkages between tion services, and energy. If output is produced and trade and the environment are discussed to offer the delivered using unchanged technologies, an increase reader a better understanding of the issues. in emissions and resource depletion must follow. 1 2 Trade, Global Policy, and the Environment Composition effect Since trade liberalization generates increased in- The composition effect stems from changes in the rela- come levels, demand for environental quality is also likely to increase. Assuming that this leads to tive size of the economic sectors following a reduc- political pressure for more stringent environmen- tion in trade barriers. Lowering trade barriers changes t poli.cessand forcme the enitolmu- the relative prices between goods produced in differ- tion load will be lower. ent sectors, so that producers and consumers face new * If investment liberalization also takes place, for- trade-offs. Countries tend to specialize production in eign investment may bring modern technologies sectors in which they have a comparative advantage; which are likely to be cleaner than older versions. this tendency becomes more pronounced with freer * As the relative price of intermediate inputs trade. If the difference between abatement costs and changes when tariffs are lowered, the input mix the price of resource extraction is sufficiently large - chosen by firms is adjusted; the new mix may be making environmental regulations more important in more or less pollution-intensive. the determination of comparative advantage- Governments may begin competing for invest- countries with lax regulations are likely to shift away ment and jobs by setting lower environmental from relatively clean sectors and specialize in more standards - a "race to the bottom." However, if for- polluting or resource-dependent sectors, thus dam- eign consumers demand goods produced with aging the environment, cleaner methods, international trade could reduce If, on the other hand, the base for interational pollution intensities, instead stimulating a "race comparative advantage is differences in the supply to the top." of labor and capital or in the efficiency of technolo- Closely related to the previous point, incentives gies, then the impact of changing sector composition for lobby groups to pressure governments for more gies, ~~~~~~~~~~~~favorable environmental legislation may shift as (in response to trade liberalization) on environmen- a result of liberalization. If the sectoral composi- tal quality and resource extraction will be ambigu- t effet (disedlabove iesoa shif ous. More open trade encourages countries to shift tion effect (discussed above) implies a shift into produs.tMore itopsenctors t e rakheay counies of theif more pollution-intensive sectors, both industry production into sectors that make heavy use of their and environmental interests can be expected to relatively abundant factors -sectors in which they intensify their efforts to receive favors from envi- have a comparative advantage. The final effect de- ronmental policymakers - at higher output levels pends on whether the new sector composition is more more is at stake, both in terms of profits and envi- or less polluting than the original one; that is, it de- ronmental degradation. pends on the relative pollution-intensity of the ex- In sum, the technique effect has an ambiguous panding sectors compared to the contracting ones. pandeeing ectoutris trend to thae acomprartivnes. effect on pollution and resource extraction, but is gen- aDvantaping lboutr-intensive setors,vea whchmareagen- erally believed to be positive for environmental qual- eradvycleantger tn capita-intensive sectors.w h at den- ity. In addition, seen from a global perspective, free trade results in a more efficient use of resources; thus veloping countries rich in natural resources may fewer raw materials and inputs are used to produce a experience an expansion of resource extraction follow- ing trade reform. Countries just beginning the indus- the amount of output produced is not constant. trialization process will naturally experience a rapid The three main effects described above often increase in the size of the manufacturing sector. have both local and global environmental implica- tions, and their relative importance differs among Technique effect countries. The final impact of trade liberalization on The technique effect refers to changes in production the environment is therefore ambiguous. A quantifi- methods that follow trade liberalization. Pollution cation of the relative magnitude of these effects, and emissions per unit of output do not necessarily stay their final result, is therefore useful to understand the constant; final intensity depends on a number of sub- range of the relative significance of the three effects components: in different countries. Trade, Global Policy, and the Environment 3 The country-specific effects point to a need to mates of the likely impact of new technology on emis- identify and forecast the effects of existing and future sions; thus incorporating one component of the tech- trade and environmental policies -which puts great nique effect. pressure on policymaking institutions. Sufficient in- The authors' results suggest that, at least with stitutional capacity is not always in place to permit respect to air and water, trade policy reforms would environmental problems to be prevented or handled in many cases improve the environment and reduce as they arise, underscoring the need for a thorough the depletion of natural resources. In the worst cases analysis of future environmental effects as foreign it would add only slightly to environmental degrada- trade continues to open up. The collection of papers tion. According to the authors these results would in this volume should contribute to an improved occur even without strengthening enforcement of ex- analysis of the complex issues involved in such isting environmental regulations or adding new ones, situations. and even if the reforms stimulate a faster rate of eco- nomic growth. This would be the case, they argue, 3. Overview of the Volume Papers because the textile and apparel sectors would expand The three sections below provide an overview of the as a result of liberalization. The damage caused by papers contained in this volume, according to the three trade liberalization is estimated to be only a fraction major themes cited above. At the end of each section of the damage that normal projected economic growth the main empirical findings are reiterated. and structural change would cause by the year 2020 if trade and environmental policies remained un- Effects of trade liberalization and growth changed. However, it should be noted that even with- on the environment out trade liberalization environmental degradation is oprojected to be substantial. Any further damage could The six papers addressing the first broad theme study therefore potentially have large marginal effects. the linkages between trade regimes, growth, and en- the scenario de aboe is basednst The scenario described above is based on a stat- vironmental degradation from different perspectives. ic model that does not include an increase in the eco- The first two studies simulate the environmental and nomic growth rate as a result of trade reform. The health effects of trade liberalization and consequent omission of dynamic effects means that the scale ef- economic growth. The following two contributions fect is likely to be underestimated. For the case of study the inverted-U (or environmental Kuznets APEC liberalization the authors therefore extend the curve) and the role of trade openness in this relation- analysis by performing a simulation in which Indo- ship. The last two papers in this section discuss the nesia'. income growth increases by an extra 0.5 per- effects of trade distortions and environmental poli- cent per year as a result of liberalization. This implies cies on environmental damage from horticultural pro- that by 2020 air pollution would be 12 to 15 percent duction and deforestation. greater, and water pollution 6 to 12 percent greater, Trade, Growth, and Environmental and Health than otherwise would have been the case. The growth Effects. The first two papers in this volume use com- rate is clearly important for the resulting increase in puterized general-equilibrium models to predict the emissions; a 0.5 percent higher growth rate may be environmental impacts of trade liberalization. In chap- conservative. The authors argue that the economic ter 2 ("Will Trade Liberalization Harm the Environ- gains from trade reforms and the surplus (which ment? The Case of Indonesia to 2020"), Anna Strutt would be much greater if GDP growth increased fur- and Kym Anderson present a case study simulating ther) available for adopting well-targeted environ- the ways in which Indonesia's environment in the year mental policies to reduce serious damage would be 2020 will be affected by two different reforms: the sufficiently large that aggregate social welfare would Uruguay Round and most-favored-nation liberaliza- almost certainly be improved substantially by liber- tion by Asian-Pacific Economic Cooperation (APEC) alization. countries. The authors focus on the additional envi- For the policymaker the challenge is to make sure ronmental damage caused by trade liberalization. An that the environmental reforms (in particular in important contribution of the paper is its use of esti- pollution-intensive or resource-extracting sectors) re- 4 Trade, Global Policy, and the Environment quired to make trade reform welfare-improving are The authors argue that both the MERCOSUR and put in place before severe environmental degradation unilateral liberalization scenarios would have a sig- has occurred; it seems particularly important to tar- nificant effect on the environment and on urban mor- get the right sectors. As discussed in chapter 3, the bidity and mortality. Damages due to rising morbidity need for information to ensure that this take places is and mortality are substantial. Unilateral trade liber- great. Failure to take preventive measures may result alization results in damages equal to 13 percent of the in substantially higher clean-up costs once damage income gains arising from free trade. Taxes on air pol- has occurred. lutants, however, produce net welfare gains from re- The paper by John Beghin, and others (chapter duced health damages. The authors find that 3, "Trade, Environment, and Public Health in Chile: emissions of small particulates, SO2 and NO2, have Evidence from an Economywide Model") uses an the strongest impact on local mortality and morbidi- empirical simulation model to examine links between ty. These three pollutants appear to be highly com- Chilean trade policy, atmospheric pollution, and pub- plementary; if the level of one is reduced, the other lic health in the Santiago metropolitan area. The pa- ones will also fall. This implies that if only a few types per recommends a coherent policy that links reform of pollutants are to be targeted (for budget reasons, programs in the three areas. It makes an important for example), the complementary pollutants should contribution by explicitly incorporating links from be targeted first, given their toxicity. Some pollutants trade to the environment to public health indicators, can be reduced at a low cost in relation to GDP. For rather than only measuring pollution incidence or oth- example, reducing SO2 by 25 percent (and therefore er environmental variables. Thus the toxicity of pol- automatically also particulates and NO2 by 23 and 25 lutants is factored into the analysis. percent, respectively) would cost 0.2 percent of GDP Three trade-policy scenarios are examined and through 2010; CO2 reduction is even less costly. compared with a business-as-usual scenario; (1) The authors' results point to the highly differing Chile's accession to NAFTA; (2) Chile's accession to trade and growth effects, and therefore environmen- MERCOSUR (the regional trade agreement between tal and health effects, of different liberalization pro- Argentina, Brazil, Paraguay, and Uruguay); and (3) grams. Just as the economic welfare effects of trade unilateral (complete) trade liberalization by Chile with liberalization programs are highly dependent on the the rest of the world. Moreover, the paper simulates type of reform undertaken, their enviromnental effects environmental tax policy reforms assumed to be suf- depend on what sectors are liberalized and by how ficiently large to reduce a particular pollutant by 25 much, as well as on which economies integrate their percent. Unilateral trade liberalization results in sig- markets. These factors determine the scale-, compo- nificantly higher economic gains (GDP would be 5.6 sition-, and technique effects, and as the Chile case percent higher than under the business-as-usual sce- study shows, they differ substantially according to the nario) than NAFTA or MERCOSUR accession (+1.4 liberalization scenario. The chapter argues that the percent and +0.6 percent, respectively). However, resulting input mix (which depends heavily on the completely free trade also produces a substantial environmental policies in place) is an important fac- worsening of pollution levels and an expansion of tor affecting the environmental outcome of liberaliza- resource-based sectors (resource-based exports in- tion. crease), partly because it facilitates access to cheaper The Environmental Kuznets Curve. A county's trade energy. orientation has important implications for its growth NAFTA integration, however, would be environ- rate. This has led to an intense discussion on the en- mentally benign in terms of several types of pollu- vironmental impacts of economic growth. At the heart tion emissions, relative to the other trade-integration of this debate is the question of whether economic scenarios. It reduces environmental damage because growth is welfare-improving, when all effects are in- it diverts trade flows so that Chile reduces its reliance corporated. The inverted-U hypothesis offers a hope- on cheap energy, unlike the other two scenarios. Join- ful answer to this question. It says that at a certain ing NAFTA, however, would have little impact on GDP level, pollution levels fall with further income Chile's GDP level. growth, so that there is no contradiction between eco- Trade, Global Policy, and the Environment 5 nomic growth and enviromnental quality beyond this ronmental quality increases, environmental regula- point. However, although this hypothesis has been tions become more stringent -a technique effect. As found to hold true for several types of pollutants, in argued by Strutt and Anderson in chapter 2, greater other cases it does not. The hypothesis also implies growth also yields an economic surplus available to that low-income countries seem destined to go be used for environmental protection measures. As through a substantial worsening of the environment income increases a shift may also take place toward (with the consequent health effects) before things get cleaner subsectors within the manufacturing sector. better. Thus trade liberalization also indirectly mitigates en- What are the underlying relationships that de- vironmental damage. Dean's work improves our un- termine the effects of income growth on the environ- derstanding of the relationship between trade, income ment? What role does the trade regime play? If we growth, and emissions. had the answers to these questions, we would be in a The inverted-U relationship differs between vari- better position to prescribe policy recommendations ous pollutants, and empirical explanations of the un- that could help us cut off the top of the inverted-U derlying workings of economic development on and also reduce the income level at which the top of pollution levels have not yet been developed. More the inverted-U is reached. In short the issue is how to work is needed to understand this process, and for facilitate the transition of low-income countries to- which pollutants it does or does not work. For exam- ward improved environmental quality. Other ques- ple, the inverted-U relationship should be more like- tions include: How can high-income countries reduce ly to hold for local forms of pollution than for global the level of pollutants that are now increasing with pollutants, because of free-riding problems. Another income? Are international environmental agreements problem with existing evidence is that no country has necessary to control these pollutants? yet been shown to have actually followed the stipu- In chapter 4, "Testing the Impact of Trade Liber- lated path. alization on the Enviromnent: Theory and Evidence," In chapter 5, "Industrial Pollution in Economic Judith Dean develops a methodology for estimating Development: Kuznets Revisited," Hemamala Het- the relationship between trade liberalization and in- tige, Muthukumara Mani, and David Wheeler pro- dustrial emissions. She takes into account the effects vide further evidence by testing for a Kuznets of openness on income growth and of income growth relationship between industrial water pollution and on environmental damage. The approach consequent- economic growth. They separate out three compo- ly allows for the possibility that trade may have sev- nents of the effects of income growth on water pollu- eral ambiguous effects on the environment. The tion. First, the authors ask what happens to the share literature on both trade and growth and income of manufacturing in total output when income increas- growth and environmental damage (the inverted-U es. This is clearly important for total industrial pollu- hypothesis) are thus incorporated into the model. She tion. Second, they look at the sectoral composition of uses water pollution data from Chinese provinces as relatively clean and dirty sectors within the manufac- an example. turing sector; and third, they examine the intensity of The results support the view that trade liberal- industrial pollution (per unit of output). The authors ization has both a direct and an indirect effect on emis- find that manufacturing's share of output, when iso- sions growth, and that these effects have opposite lated, follows a Kuznets-type trajectory. Next, sectoral impacts. Dean finds that China has a comparative composition gets "cleaner" through the middle- advantage in pollution-intensive goods, so that in- income range and then stabilizes. Finally, water pol- creased trade openness directly aggravates environ- lution intensity declines strongly with income, con- mental damage by inducing a specialization in these firming Dean's finding in the preceding chapter. The sectors - a composition effect. Increased openness to authors attribute this phenomenon in part to stricter trade is also shown to strongly increase income regulation as incomes increase, and in part to a shift growth. However, income growth has an indirect neg- to more modern (and therefore cleaner) production ative effect on emissions growth, improving pollution technology. When they combine the three relation- levels. This is probably because as demand for envi- ships, however, they do not find a Kuznets-type rela- 6 Trade, Global Policy, and the Environment tionship. Instead, total industrial water pollution ris- environmental policies. The analysis shows that hor- es rapidly through middle-income status and remains ticultural production (land use) is largely insensitive approximately constant thereafter. One important to such measures. The environmental impact was also question is the level at which water pollution stabi- negligible; damage to the environment was reduced lizes: Does this happen before or after a serious im- only slig'htly by either action. The authors argue that pact on humans, fish, and aquatic life has occurred? the main reason for this is the relatively high profit- To illustrate the implications of their findings, ability of most horticultural production (the area de- the authors simulate the trends in industrial water voted to this production represents a small share of pollution for four groups of industrial economies be- poor farmers' total production). In Kenya it appears tween 1977 and 1989. They find approximately stable that policies that internalize the social costs of envi- emissions in the OECD and ex-COMECON countries ronmental damage could fairly easily be implement- (note that the ex-COMECON economies experienced ed, and could also generate revenue that could be used constant or declining incomes during the period), by the government for environmental protection. moderate increases in the newly industrialized coun- The authors, however, do not model the impacts tries, and rapidly growing pollution in the develop- on marginal or other vulnerable producers. They ar- ing countries of Asia. Their estimates suggest that gue that if tariff preferences were removed the stron- during the 1980s the latter replaced the OECD econo- gest impact would be on these producers, who are mies as the world's largest generator of industrial barely surviving. Some form of social protection water pollution. Overall, however, the negative feed- would be needed if the preferences were to be altered. back from income growth to pollution intensity was One question that arises from the analysis is what ef- sufficient to keep total world pollution growth to fect the policy changes would have on production around 15 percent during the 12-year sample period. methods in the remaining productive areas. Overall, Horticulture, Forestry, and Environmental Degrada- the paper suggests that EU preferential trade with tioms Agriculture is an important, but often neglect- Kenya has little effect either on land use or on Ken- ed, sector in many countries, and trade liberalization ya's rural environment. could offer new opportunities to agricultural produc- Kenneth Chomitz, Charles Griffiths, and Jyots- ers in developing countries. Their exports are often na Puri study the effect of trade and tax policies on taxed, thus driving down domestic prices. But trade deforestation in the Sahel in chapter 7 ("Fuel Prices, policies favorable to developing-country producers Woodlands and Woodfuel Markets in the Sahel: An can also be cited. For example, in chapter 6, Anil Mar- Integrated Economic-ecological Modelg). There has kandya, Lucy Emerton, and Sam Mwale study the long been concern that overexploitation of woodlands environmental effects of the European Union (EU) for fuels leads to a vicious circle of environmental preferential trading arrangements with Kenyan hor- degradation, rising fuel prices, and increasing pover- ticultural producers ("Preferential Trading Arrange- ty. Urban demand for charcoal and fuelwood leads to ments between Kenya and the EU: A Case Study of increased areas of deforestation around urban centers, the Environmental Effects of the Kenyan Horticulture as open-access forests are used as sources of energy. Industry"). The study is particularly timely because Urban fuel costs increase, since supplies must be the EU and 71 developing countries have begun ne- brought in from increasingly distant harvest areas. gotiations for an arrangement to follow the end of the This process is influenced by the prices of kerosene, fourth Lome Convention, in which preferential access LPG (both substitutes for woodfuels), and gasoline to the EU markets was granted. A large share of or diesel fuel (a major component of the cost of wood- Kenya's horticultural output is produced by small- fuel production). Fuel prices, in turn, are to a large scale farmers, and thus contributes significantly to the extent determined by policy; for example, trade poli- incomes of the poor, according to the authors. cy. Whereas some countries subsidize kerosene with Horticultural activities are environmentally the explicit goal of reducing pressures on forests and damaging, however, and the paper analyzes the im- woodlands, in other countries the price of modern pact on the environment of two policy changes: the fuels is above world market prices as a result of pe- removal of preferential tariffs and the introduction of troleum product trade, manufacture, procurement, Trade, Global Policy, and the Environment 7 distribution, and tax policies. The question is the ex- The pollution haven hypothesis tent to which, for example, trade and tax policies can One issue that has attracted much attention is wheth- affect woodland degradation, rural income, and ur- er reduced trade barriers will result in a specializa- ban fuel consumption. tion by developing countries in pollution-intensive The authors' model shows that distortions in industries; that is, will this group of countries become kerosene and gasoline prices have almost no impact "pollution havens?" For decades commentators have on the rate of woodland degradation. According to predicted that relatively less stringent environmental their model only substantial kerosene subsidies in- regulations in developing countries will lead pollut- duce a shift over to kerosene (due to the demand func- ing firms to migrate there from the OECD economies. tions used), although such a shift may also occur when Economic theory also argues that low-income coun- woodfuels start to be transported from much greater tries may develop a comparative advantage in distances to urban centers. The paper suggests that pollution-intensive goods, even when environmental reductions in policy distortions, while desirable on damage is taxed in an optimal fashion. The task of economic-efficiency grounds, may sometimes have distinguishing the empirical effects of environmental less environmental impact than hoped. Other factors, regulations from other trends is certainly a difficult such as road access, appear to be more important than one. Most existing empirical studies have found ei- trade policy. Reductions in gasoline taxes will have ther no, or weak, effects of environmental regulations distributional effects, shifting rents away from wood on trade flows. Chapters eight, nine, and ten present harvesters to urban consumers. The authors, howev- new evidence related to this issue (chapters four and er, qualify their conclusions by pointing to the sensi- five also touched on the subject). tivity of the results to the assumptions made. The Industrial Delocation. Muthukumara Mani and Day- difference between short- and long-run effects may id Wheeler find a pattern that initially appears con- also be Important. sistent with the pollution havens hypothesis (chapter SUMUARY OF EMPIRICAL RESULTS 8, "In Search of Pollution Havens? Dirty Industry in - Combined trade and environmental reforms offer the World Economy, 1960-1995"). However, they ar- the greatest possibility for positive welfare effects. gue that their evidence shows that pollution-haven Pollution abatement that results in significant, effects have not been of major importance for a num- positive health effects sometimes costs only a frac- ber of reasons. First, consumption and production of tion of the surplus created by trade liberalization. pollution-intensive sector products in the developing Environmental policy reforms that target pollut- world have remained roughly equal throughout the ants that are mutually complementary are more period of study. This indicates that production has effective, because several pollutants are reduced been primarily for the domestic market, and firms are simultaneously. not taking advantage of lax regulations to export else- * The inverted-U relationship between pollution lev- where. The authors argue that trade is not to blame els and income holds for some pollutants only. for any increase in the production of pollution-inten- Some types of industrial water pollution appear sive goods. Nonetheless, production aimed not to fall, even at high income levels. primarily at the home market can still have severe en- * In some developing countries trade liberalization vironmental effects. directly aggravates environmental damage directlyaggravaes envionmenta damageSecond, a significant part of the increase in the through its effect on relative prices and sectors of secod, ctign pare in reasein te specialization, but also indirectly mitigates such dirty-sector production share in developing regions damage because of the effect of liberalized trade seems to be due to a highly income-elastic demand on income growth. for basic industrial products. As income levels have * Different trade liberalization programs have highly risen, this elasticity has declined. Thus further increas- differing effects on environmental quality, human es in income now yield a smaller effect on demand health, and resource use. Some liberalization pro- than at lower GDP levels. Finally, the authors find that grams may have little effect on environmental deg- the stringency of environmental regulations increas- radation. The effects are country- and policy-specific. es along with incomes. They argue that although pol- 8 Trade, Global Policy, and the Environment lution havens may appear to have emerged, in prac- emissions. High pollution intensity of production may tice, like "low-wage havens," they are transient. spark concern among investors about the inefficiency Alternative Explanations. Two papers in this volume of a firm's production process and the risk of future argue that there may be other reasons why plant lo- liabilities; it may spur regulators, environmental cations and trade patterns do not appear to be strongly groups, and a plant's neighbors to increase surveil- affected by enviro.nental regulations. The explana- lance; and it may result in the loss of reputation and tion by Paavo Eliste and Per Fredriksson (chapter 9, goodwill. The announcement of sound environmen- "The Political Economy of Environmental Regula- tal performance or investment in cleaner technologies tions, Government Assistance, and Foreign Trade") has the opposite effect. is based on the theory that firms subject to environ- It appears that certain conditions are required mental regulations may have been compensated by for this process to work, however. The existence of policymakers for the associated abatement costs potential fines or other penalties may play a role in , . , , , .,, ~~~~~~firm and market behavior, if it is the net present val- through various types of transfers such as tariffs, pro- f duction subsidies, or pollution-abatement subsidies. ue of possible future liabilities to which the market They present empirical evidence from the agricultur- reacts. Moreover, not all polluters are listed on the al sector that supports this hypothesis. stock market; for example, publicly owned firms and a Thsetretsupporindicats that farmers have beensuc small firms are generally not motivated by this form The results indicate that farmers have been suc-v ,a .. . . ~~~~~~~~of incentive.2 cessful in the political game of extracting income trans- inene.t fers from other groups in society. The compensation thes resultsomay cnrbthest our undertad provided for costs associated with environmental reg- png of ith pol lutionhvnvhpthens is epected ulation implies that producers are unlikely to be fac- pintabilit an defowofinvestmntrs inpoltin ing the true social cost of the damage caused; that is, itensv sect orseinsdverlopin countr may hep thatthePoluterPay Prncipe des ot hld.It ol- been affected by concerns over legal liability and rep- that the Podutio incmany coes is le tolb utational damage that could influence a multination- al worldwide. This would reduce the incentive to greater and more intense than optimal, affecting en- vironmental~~~~~~~~ qult.eaiey T h xetta relocate to countries with less stringent environmen- vironenta quaity egatvely To he etentthat tal regulations. It becomes more difficult to find sup- these phenomena have taken place in other sectors, tar re tionsut beoen moredifficul st finds the implication is that regulations are less likely to influence trade and investment patterns.1 that consider relocating to pollution havens in order In "Pollution and Capital Markets in Develop- to take advantage of lax regulations are discouraged ing Countries" (chapter 10) Susmita Dasgupta, Benoit from moving. Laplante, and Nlandu Mamingi offer a new perspec- S oF E R tive on environmental regulations and firm location Developing-country production is primarily for in developing countries through their discussion of the domestic market, and firms appear not to have the relationship between firms' environmental per- taken advantage of lax regulations to export else- formance and the behavior of capital markets. Some where. The increase in developing-country, dirty- observers argue that firms in developing countries sector production is, to a significant extent, due to have little incentive to undertake pollution control increasing demand for basic industrial products because of weak monitoring and enforcement of en- at low income levels. vironmental regulations -where regulations exist. * Farmers subject to environmental regulations ap- However, the authors argue that if sufficient infor- pear to have been compensated for their environ- mation is available to capital markets, these markets mental protection costs through various support may provide financial and reputational incentives to programs. This should have mitigated the trade invest in appropriate efforts to reduce pollution. The effects of environmental regulations. chapter shows that capital markets in several devel- * Information on the environmental performance of oping countries react favorably to new media an- firms listed on developing country stock markets nouncements of superior environmental performance have significant effects on share prices, at least in and negatively to citizens' complaints about firm the short run. Trade, Global Policy, and the Environment 9 Global environmental policymaking that it represents a special case and sanctions are less The last set of papers focuses on global environmen- likely to work for many other type of IEAs. For exam- tal policymaking. The first two discuss linkages be- ple, many more producers are involved in the pro- tween trade, trade policies, and international duction of CO2 than of CFCs. agreements; the last two explore policy tools to ad- In chapter 12 ("The Importance of Trade for the dress global environmental problems. Ratification of the 1992 Climate Change Convention") Per Fredriksson and Noel Gaston investigate the re- International Environmental Agreements and Trade lationship between trade flows and delays in ratifica- Policy. One issue that attracts considerable interest is tion of the United Nations Framework Convention on the use of trade sanctions in international environmen- Climate Change (UNFCCC). Their work addresses the tal agreements (IEAs). Clearly, the search for enforce- hypothesis of "political drag," or "regulatory chill" ment mechanisms is important for IEAs since no in environmental policymaking due to trade concerns. supranational enforcer exists. The Montreal Protocol, This hypothesis, closely related to the "race to the which is widely seen as very successful in reducing bottom" argument, argues that countries may delay production of chlorofluorocarbons (CFCs), contains enacting environmental legislation and associated provisions for the use of such sanctions. In "The Cred- abatement efforts for reasons of international compet- ibility of Trade Sanctions in International Agree- itiveness, particularly if they are more open to inter- ments" (chapter 11), Scott Barrett discusses the general national trade. The authors argue that the speed of applicability of trade sanctions as a means to enforce ratification is important to the success of an IEA and an international agreement to provide a global public for countries' bargaining positions in the next negoti- good, such as CO2 reduction. He identifies two re- ation round. Slow ratifiers of the 1992 Climate Change quirements that must be satisfied for sanctions to suc- Convention may have gained an upper hand in the ceed in deterring free-riding. 1997 Kyoto bargaining round. First, given that sanctions are imposed against The evidence on the importance of the openness free-riders, every country must be better off as a sig- to trade, however, is weak. Whereas increased exports natory than as a nonsignatory. If this condition were have tended to increase ratification speed, imports satisfied, free-riding would be deterred completely. haetndtoicasrtfctonpe,imrs sSfeod,hen free-riding wocculd be gdeterred cmpltbe have had the opposite effect. The results suggest that Second, .when free-riding occurs, signatories must be eprigcutishv encnendwt eu betrofb' moigsntosta ycniun exporting countries have been concerned with repu- better off by imposing sanctions than by continuing tation, whereas importers worry about their compet- to trade with nonsignatories. If this condition were itiveness. However, it should be noted that large satisfied, the threat to impose sanctions would be cred- exporters are likely to be large importers. In sum the ible. However, the second condition is less likely to .political drag" argument cannot be supported by this be met, because sanctions hurt the countries impos- data. The authors also discuss the implications of the ing them as well as those being targeted. The issue of results for the use of trade sanctions. "leakage" is important in this context If signatories However, a countrs ratfication of the UNFC- of an IEA reduce their emissions, nonsignatories could CC is found to be more likely, the greater its total CO develop a comparative advantage in pollution-inten- CCissfoundt be morecle, the graerits total 2 sive industries, thus reducing the impact of the IEA emissions and, in particular, civil liberties. Apparent- sivemindusriens, thus reducing the impact of the lEA ly, countries that were heavy polluters came under Barrett arguess that sanctions are more likely to great political pressure, either internally or external- Barrett argues that sanctions are more ly, to ratify the UNFCCC. Political and institutional bkel used - wthti,hat the srobleond condito is morge. factors may be substantially more important than likely to hold -when the problem of leakage iS large. trdarnemtsidtrin atcitonn In that case sanctions against nonsignatories increase geg prip the supply of emission reductions and are credible, global environmental policymaking. Democracies provided that benefits exceed costs. If both conditions seem to view international norms as important, and hold, the threat of trade sanctions would be sufficient to be less likely to shirk their global responsibilities. to deter free-riding, and the sanctions would never Policy Instruments for Global Environmental Poli- be employed. Barrett concludes that the sanctions in cymaking. In chapter 13 ("Achieving Carbon Emis- the Montreal Protocol were effective and credible, but sion Reductions through Joint Implementation") Will 10 Trade, Global Policy, and the Environment Martin takes a closer look at joint implementation, one what the overall level of employment and investment of several ways to coordinate reduction of global CO2 in the economy. Preexisting taxes mean that the lev- emissions. Martin's study has implications for the els of employment and investment are already below future design of joint implementation projects. The levels needed to maximize economic efficiency. Hence large differences between the cost of reducing such further reductions result in additional efficiency loss- emissions in industrialized and developing countries es. However, if the environmental policy raises tax suggest potentially great scope for lowering overall revenues, and the latter are used to reduce other dis- emission reduction costs by focusing primarily on tortionary taxes, then much of the additional economic projects in developing countries. Mostjoint implemen- cost can be offset. To achieve CO2 reduction a tation proposals focus on installing new, more effi- revenue-raising instrument (such as an emissions tax cient equipment to reduce energy use for a given or auctioned emissions quota) is thus preferable to a output level. nonrevenue-raising instrument (a nonauctioned emis- Martin argues that such increases in production sions quota). It is also important to include the poten- efficiency could have some important second-round tial impact of a given policy instrument on private impacts that have thus far been ignored. Increased i t efficiency would lower the price of energy relative to gies ( o Martin' chapterg13). other inputs and lower the price of energy-intensive any deeoing countr h s ed products relative to other products. The direct impact bor markets. Part of the labor force may be employed of a fuel-price reduction is an increase in consump- iormalketor in whi taxes are levedwlet tion of that fuel. In addition substitution effects will in a formal sector while t generally reduce the use of other fuels and associated o the t system.nThemeffector th ibot emissions. If the fuel being made more efficient is al- covered by the tax system. The effect of the tradition- al tax system is to distort the allocation of production repady thaprie leastemisions-indtenve, the cminged by implicitly subsidizing the informal sector, since the imatofapic eucinan terpie-hne latter is not taxed. If energy inputs are taxed to re- is likely to favor the environment, because relatively duce carbon emissions and the revenues are used to more of the relatively cleaner fuel would be consumed. duce formissior and the taxus areduse However, if the fuel being made more efficient is ini- reduce formal taxeor th an is tially the most emission-intensive, the aggregate ef- the effective subsidy to the informal sector. That is, segmented labor markets offer the possibility of a fect of such price changes may result in an increase in emissions, especially if this fuel continues to be heavily double dividend from environmental tax reform. polluting. The final impact of demand shifts on fuel SUZ4RY OF EIRUPCAL RESULTS consumption and energy emissions will also depend The hypothesis that increased trade makes legis- on how the supply side of energy markets react. If latures less willing to enact global environmental energy supply is inelastic, reductions in energy de- policies could not be confirmed. Democratic insti- mand resulting from joint implementation (because tutions appear to positively affect legislative ac- of increased efficiency) will reduce the price of ener- tion on global environmental issues. gy and stimulate energy use, partially offsetting the * Joint implementation programs that are aimed at initial fall in demand. the cleanest fuel are most likely to be favorable to In the final chapter Ian Parry surveys the litera- the environment, because the combined impact of ture on the "double dividend" ("Carbon Abatement: price reductions and other price changes stimu- Lessons from Second-Best Economics"). The "double late use of the relatively cleaner fuel. dividend" refers to the idea that emission taxes yield * Environmental policies can be substantially more a double return by both reducing emissions and rais- costly when their effect on preexisting tax distor- ing revenues that can be used to reduce distorting tax- tions is taken into account. Such policies reduce es elsewhere in the economy. Environmental policies somewhat the overall level of employment and in- can be substantially more costly when their effect on vestment in the economy. If an environmental preexisting tax distortions is taken into account, be- policy raises tax revenues, however, much of the cause environmental policies tend to reduce some- additional cost may be offset. Trade, Global Policy, and the Environment 11 4. Policy Conclusions and Recommendations national cooperation and avoid some distortions could The collection of papers presented here form a fol- be multi-tier harmonization agreements. Such agree- low-up volume to the 1992 World Bank Discussion ments may avoid the free-riding problem that occurs Paper edited by Patrick Low ("International Trade and with transboundary pollution problems. the Environment"). Since the publication of the Low A few policy conclusions can be drawn from the volume, a substantial amount of research has been papers in this volume. For example, the ambiguity of carried out in this area. In particular, much effort has environmental effects of trade liberalization places been devoted to documenting the existence of an in- heavy demands on existing institutions charged with verted-U relationship between income and pollution environmental policy formulation and implementa- and the effects of environmental regulations on trade tion-to prevent potential problems and respond as patterns. negative effects appear. Therefore, local institutional However, our knowledge on most issues in the capacity building on environmental policy is an im- area of trade and environment is still quite limnited. portant priority. This must include the capacity to Some of the reasons for this include serious data lim- identify economic trends and to forecast the environ- itations and the fact that the underlying theories (for mental effects of structural changes. Many environ- example, the inverted-U) are often underdeveloped. mental impacts of trade liberalization could be A major factor behind the uncertainties is also the in- addressed at the national or local level by appropri- herent ambiguity of the impact of trade liberalization ate environmental policies. Coordination of trade and on environmental quality. Forecasting the relative environmental policies is thus important. Appropri- magnitude of the various effects of trade liberaliza- ate environmental policy responses may also reduce tion for a specific country requires a substantial resistance to continued trade reforms, which in turn amount of data and becomes especially complex and would offer protectionist interests less ammunition. uncertain because the final outcome depends on a Actions to improve environmental quality need politically determined policy response at both the not be expensive. Increased dissemination of infor- national and international level. Although the papers mation has been found to induce capital markets to included in the present volume improve our under- police environmental performance. This leads to an standing of these complex matters, much work re- environmental policy argument for privatization, be- mains to be done. For example, as income growth cause nonlisted firms -such as public companies - proceeds in the future, will the relationships discussed are not subject to such scrutiny. Reduced subsidies to change over time, particularly if abatement costs and polluters, such as agriculture and energy producers, pollution levels increase substantially? would have positive impacts on environmental qual- A few issues, however, appear to be approach- ity, world market prices, and government finances. ing settlement. Advances can be cited in understand- The promotion of democratic ideas and practices also ing of the appropriateness of trade measures for has a beneficial effect on environmental policymak- environmental protection. Most observers now agree ing. that environmental policy, in general, is more efficient Policymakers prefer low-cost options, but they and appropriate than trade policy for addressing en- cannot rely only on win-win policies. If the Polluter vironmental problems. Trade policy measures have Pays Principle is to be effective, environmental regu- uncertain effects and create their own distortions. Only lations should be allowed to have economic and trade in special cases, such as international environmental effects. These have been found to be small or negligi- agreements on transboundary pollution issues, is ble, in particular because abatement costs are relative- trade policy likely to play a role. Moreover, uniform ly low. harmonization of environmental policy across coun- tries is inappropriate, since countries have different Notes pollution problems and dissimilar valuations of en- I would like to thank the participants and, in particu- vironmental damages. Thus different countries re- lar, the discussants at the Trade, Global Policy, and the quire different environmental policies; otherwise new Environment conference for many useful suggestions distortions are created. One way to undertake inter- on the papers in this volume. This chapter draws on 12 Trade, Global Policy, and the Environment their insightful comments. I am also grateful to John abatement costs are low relative to labor and capi- Dixon, Ola Larsson, Michael Linddal and Alison Rafa- tal costs, and therefore the cost differentials between el for suggestions on this introduction, and to Jim countries are necessarily low. Moreover, some pol- Cantrell for great desktop work. I am solely responsi- lution-intensive sectors are dependent on access to ble for all remaining errors. In addition to the papers natural resources, which determines their location; presented, David Reed, World Wide Fund for Nature, other industries, such as petroleum refineries, tend and Daniel Esty, Yale University, addressed the con- to locate close to markets. The "Porter Hypothesis" ference as keynote speakers, and their participation argues that environmental regulations stimulate in- is gratefully acknowledged. Support for the confer- novations and improvement of technology, ence and this volume was provided by the Swedish increasing overall productivity in the process. International Development Cooperation Agency 2. There is some recent evidence that publicly held (Sida) through the funding of the "Trade, Macroeco- firms have worse environmental performance than nomic Reform, and the Environment" project, which privately held firms (see R. Hartman, M. Huq, and is gratefully acknowledged. D. Wheeler, "Why Paper Mills Clean Up: Deter- 1. Alternative or complementary explanations for the minants of Pollution Abatement in Four Asian absence of changes in observed trade patterns in- Countries," Policy Research Department Working clude the fact that in most sectors pollution Paper 1710, Washington, D.C.: World Bank, 1977). Chapter 2 Will Trade Liberalization Harm the Environment? The case of Indonesia to 2020 Anna Strutt and Kym Anderson 1. Introduction known as GTAP is used to project the world econo- ost-favoured-nation (MFN) trade liberal- my to 2010 and 2020 without and with those trade izations will always improve global eco- reforms. (This long-run view allows us to abstract s _ nomic welfare even in the presence of en- from the disruptions of the current financial and po- vironmental externalities, provided optimal environ- litical crisis.) As explained in Section 2, an environ- mental policies are in place (Anderson and Blackhurst mental module is attached to the Indonesian part of 1992; Corden 1997). However, in a world in which that global model so as to measure the effects of struc- national environmental standards differ markedly tural and policy-induced changes in economic activi- between countries and interational environmental ty on air and water pollution in Indonesia. The results, spillovers are significant, globally optimal environ- presented in Section 3, identify the proportional con- mental policies will differ from nationally optimal tributions of changes in the aggregate level and com- ones. That, plus the fact that in many (especially de- position of output, and in production techniques, to veloping) countries the enforcement of environmen- changes in environmental indicators. A base case pro- tal policies is often less than optimal even from a jection without trade reform is compared with national viewpoint, raises in some people's midnds (for alternative scenarios involving (a) full global imple- instance, Chichilnisky 1994) the question of whether mentation of Uruguay Round commitments, and (b) liberalizing trade between rich and poor countries is the additional move to MFN free trade by APEC coun- desirable. To reduce the risk that this concern leads tries by 2020. The paper concludes in Section 4 with a to excessive opposition to trade liberalization initia- brief summary of results and suggestions for further tives, and to begin to assess whether the standard research. gains from trade are sufficient to outweigh any loss in wlfar dueto aded nvirnmenal dmage em- Projecting the level and composition of output to in welfare due to added environmental damasre, em- 22 ihu n ihtaeplc eom pirical studies of the resource depletion and environ- mental degradation effects of such reforms are needed. Rapid economic development and on-going policy This paper provides a methodology for doing reforms in Indonesia and other countries of the world that and illustrates it with a case study of Indonesia, will change substantially the level, composition and a large newly industrializing country that is rich in location of production and consumption during the natural resources and committed to taking part in next two decades. In this section we project global major multilateral and regional trade liberalizations economic growth and structural changes for the peri- over the next two decades. Section 1 describes how a ods 1992-2010 and 2010-2020. We also model the modified version of the global economy-wide model Uruguay Round and APEC trade liberalization com- 13 14 Trade, Global Policy, and the Environment mitments over those periods. The Uruguay Round el developed by the Australian Government in the agreements should be fully implemented by 2005, 1980s but has been hugely improved during the 1990s before the end of the first period, and 2010 is the date from its current home at Purdue University in the agreed at Bogor in November 1994 for completion of United States. The model, which is implemented and trade liberalization by APEC industrialised countries. solved using GEMPACK (Harrison and Pearson 1996), The year 2020 was agreed by Indonesia and other is in use by over one hundred researchers in more APEC developing countries to be the date for com- than 30 countries on five continents. Hence space is pleting their move to free trade, and it also happens not used here to describe its myriad features and data to be the end of Indonesia's Second Long Term De- base.1 velopment Plan. The model utilizes a representation of consum- For the present purpose of projecting the world er demands which allows for differences in both the economy to 2020 we use the GTAP database and mod- price and income responsiveness of demand in dif- el of national and international markets for all prod- ferent regions depending upon both the level of ucts and countries/regions of the world (see Hertel development of the region and the particular con- 1997). There are numerous advantages of using such sumption patterns observed in that region. In the sim- a global, economy-wide CGE model even if, as with ulations presented below, many of the East Asian the GTAP model used here, it is comparative static in economies are projected to continue to experience very nature. The economy-wide approach makes explicit rapid economic growth rates (assuming a reasonably the assumed sources of economic growth that expand rapid recovery from the present financial crisis), so the demand for and supply of various products; it that the income elasticities of demand play an impor- ensures countries can import only what they can pay tant role in the model. Non-homothetic preferences for through exporting or borrowing; and it includes are captured through use of a constant difference of in the base scenario the inter-sectoral structural chang- elasticities (CDE) function (Hertel and Tsigas 1997). es that normally accompany economic development. This lies between the commonly used constant elas- The advantage of using a global model rather than a ticities of substitution (CES) function and fully flexi- national one, even though the primary focus of this ble functional forms. Such a demand system enables paper is on results for Indonesia, is that the economic non-homothetic demand to be calibrated to replicate growth and structural and policy changes of other a pre-specified vector of own-price and income elas- countries can be incorporated explicitly. This ensures ticities of demand. that those changes abroad in combination with Indo- On the supply-side, differences in relative rates nesia's changes are used to generate new terms of of factor accumulation interact with different sectoral trade for Indonesia. But it also allows the resource factor intensities to drive changes in the sectoral com- depleting effects of international events on Indonesia position of output. The GTAP production system used to be compared with those effects on other economies. here distinguishes sectors by their intensities in five World Bank GDP, labor force, investment and primary factors of production: land, other natural re- population projections together with the Global Trade sources, unskilled labor, skilled labor, and physical Analysis Project (GTAP) Version 3 data base and capital. Thus in a region where physical capital is ac- model are used to generate market projections to the cumulating rapidly, relative to other factors, we ex- year 2020. The full GTAP model divides the world pect the capital intensive sectors to expand at the economy up into 37 sectors and 30 countries or coun- expense of unskilled labor intensive sectors such as try groups (including the 16 major APEC economies). agriculture in East Asia. Producers are assumed to In order to keep the present analysis and presenta- choose inputs that minimize production costs subject tion of results tractable, however, the data base is ag- to separable, constant returns to scale technologies. gregated up to 23 product groups and to 5 regions in Constant elasticity of substitution (CES) functions addition to Indonesia. describe substitution possibilities between primary The GTAP model is a standard comparative-stat- factors and market clearing conditions equate supply ic multi-region computable general equilibrium model with demand for each factor of production. For inter- of the Johansen type that began as the SALTER mod- mediate inputs, the assumption of a Leontief function Will Trade Liberalization Harm the Environment? 15 implies no substitution between different intermedi- creating supply-side pressures for changes in the com- ates or between them and a composite primary position of output in these economies (Krueger 1977; factor. Leamer 1987). The relatively high rates of accumula- Land and other natural resources (minerals and tion of human capital in developing economies also energy raw materials) are assumed to be sector-spe- are likely to contribute to pressures for structural cific in this study, except that some movement of land change as developing countries upgrade the skill-in- within agricultural sectors and between agriculture tensity of their product mix. Taking all these things and forestry is allowed. It is assumed that 60 percent into account and starting with the 1992 baseline, the of value added by capital in each of the natural re- model generates projections of the world economy source sectors is attributable to the specific factor (fol- assuming no changes to existing trade and other pol- lowing Arndt and others 1997). The single factor icies. That base scenario is then compared with sce- labour in GTAP is split into skilled and unskilled la- narios involving trade policy reforms. bour for this study, whereby the global GTAP data- For Indonesia, the assumed rates of factor and base is adjusted using recent estimates of labour GDP growth are close to government expectations and payments by skill level (Liu and others 1997, p. 17).2 are in line with past trends. Over the 13 years from A composite capital nest is created for human and 1980, for example, the population and labor force physical capital, following Arndt and others (1997). growth rates were a little higher than those being pro- The present paper follows the methodology used jected here for the 18 years to 2010 (1.7 and 2.3 per- in Hertel and others (1996) and Anderson and Pang- cent historically compared with assumed rates of 1.4 estu (1998) but projects the world economy from 1992 and more than 2.0 in table 2.1), while the rates of not just to 2005 but to 2010 before looking at the long- growth of physical capital and real GDP were a little run effects of Uruguay Round trade policy reforms to lower than those projected here (7.1 and 5.8 percent be implemented between now and 2005. It does the historically compared with assumed rates of 7.4 and same from 2010 to 2020, to get a more realistic mea- 6.6 in table 2.1). sure of the long-run effects of APEC reforms. We use The model can be closed with either gross do- a carefully constructed set of Uruguay Round shocks, mestic product (GDP) or total factor productivity to take into account the reality that actual reforms in (TFP) as exogenous targets. Since projections for GDP Indonesia and elsewhere, particularly for farm prod- are available, these are imposed on the model, while ucts, will be much less than was earlier expected, total factor productivity is endogenized. Empirical thanks to 'dirty tariffication' (see Hathaway and Ing- evidence suggests that agriculture has a higher total co 1996). factor productivity growth rate than other sectors (see Table 2.1 reports the assumed rates of growth in Martin and Mitra 1996). Therefore, the assumption factors and real GDP (from which the implied rates of made here is that agricultural productivity increases total factor productivity growth may be derived) in at a rate of 0.7 percent per annum higher than other the reference case for the periods from 1992 to 2010 sectors. and 2010 to 2020. Exogenous projections of each re- With these and myriad other assumptions in- gion's endowments of physical capital, unskilled and cluding those incorporated in the GTAP model (see skilled labor, and population are utilized. These are Hertel 1997), a projection of the world economy to based on combinations of historical data and World 2010 is generated assuming no trade policy changes. Bank projections of the growth in population, labor Then the model is re-run several times: with the Uru- force, real GDP and investment.3 It is clear from these guay Round being fully implemented with China first estimates that the structure of the world economy will excluded but then included in the WTO (the main dif- change in a number of important ways in this base ference being whether China is excluded or included case, with the developing countries constituting a con- in getting expanded access to US and EU textile and siderably larger share of the global economy by 2020. clothing markets - see Anderson and others 1997); and Furthermore, given the particularly high rates of sav- then with APEC liberalization commitments also be- ings and investment in East Asia, the capital-labor ing implemented by 2020. The scenario for 2010 with ratios of these economies are expected to increase, the Uruguay Round fully implemented is the starting 16 Trade, Global Policy, and the Environment Table 2.1 Assumptions made in the projections: cumulative [and annual] percentage changes in GDP and factor endowments for the period 1992 to 2020 (a) 1992-20 10 Physical Unskilled Skilled Region Real GDP capital labor labor Population Indonesia 215 260 44 449 27 [6.6] [7.4] [2.0] [9.9] [1.4] Other APEC developing 202 312 26 167 19 economies [6.3] [8.2] [1.3] [5.6] [1.0] Other developing and 73 61 43 151 40 transition economies [3.1] [2.7] [2.0] [5.3] [1.9] APEC high-income 66 101 16 150 16 economies [2.9] [4.0] [0.8] [5.2] [0.8] Other high-income 55 53 1 394 3 economies [2.5] [2.4] [0.1] [9.3] [0.2] (b) 2010-2020 Physical Unskilled Skilled Region Real GDP capital labor labor Population Indonesia 95 135 17 77 14 [6.9] [8.9] [1.6] [5.9] [1.3] Other APEC developing 72 88 9 51 9 economies [5.6] [6.5] [0.9] [4.2] [0.8] Other developing and 49 46 29 62 18 transition economies [4.1] [3.9] [2.6] [5-0] [1.7] APEC high-income 27 47 3 53 7 economies [2.5] [3.9] [0.3] [4.3] [0.7] Other high-income 28 34 -4 79 0 economies [2.5] [3.0] [-0.4] [6.01 [0.0] Source: Strutt (1998, Ch.4) drawingonAnderson and others (1996), Arndt and others (1997) and, for2010-2020, Bach (1997). point from which to project the world economy to plemented, the real value of global output is project- 2020. This too is done assuming no further trade pol- ed to increase by 65 percent between 1992 and 2010, icy changes as a base case, and that scenario is then and then by a further 35 percent by between 2010 and compared with one in which the remaining trade bar- 2020 after the Uruguay Round is implemented but riers of APEC countries are removed. Indonesia's without any APEC regional liberalization. Develop- nominal rates of import protection for each sector at ing countries are projected to gain enormously in sig- the beginning of each of these reform scenarios are nificance, particularly developing APEC economies shown in Appendix table A2.1. which are projected to more than double their share How do all these changes affect the world econ- of world output, from 6 to 14 percent during the 1992- omy? Even without the Uruguay Round being im- 2020 period, and treble their share of world trade. Will Trade Liberalization Harm the Environment? 17 Indonesia in particular is projected to almost tre- tries grow. The grain sectors' share of GDP is project- ble its contribution to world output (from 0.5 to 1.5 ed to roughly halve by 2010, for example, and to fall percent), to increase its real volume of output and by a further one-third or more in the subsequent trade more than six-fold over the projection period, decade (columns 5 and 6) - even though the absolute and to change the sectoral shares of its GDP substan- level of output keeps rising in these as in all other sec- tially. The latter are summarized in table 2.2. It shows tors (columns 3 and 4). Another example is that while Indonesia's agricultural and other natural resource the depletion of natural resources continues, forestry, based sectors continuing to decline in relative impor- fishing and mining outputs are projected to grow much tance as textiles and other light manufacturing indus- less rapidly than aggregate national output. Table 2.2 Percentage changes in sectoral output levels and in sectoral shares of GDP due to economic growth, Indonesia, 1992-2010 and 2010-2020 Change in Change in Change in Change in sectoral sectoral real value real value share of share of 1992 2010 of output ofoutput GDP GDP output output 1992-2010 2010-2020 1992-2010 2010-2020 Sector (US$b) (US$b) (%) (0/6) (%/0) (%) Paddy rice 7.5 14.1 87 35 -41 -31 Other grains 0.8 1.0 23 1 -61 -48 Non-grain crops 12.4 19.5 58 Is -50 -41 Livestock 3.2 6.9 113 36 -32 -30 Forestry 2.5 5.1 100 43 -36 -26 Fisheries 3.8 7.0 85 23 -41 -37 Coal 0.8 1.8 124 49 -29 -23 Oil 7.4 15.8 114 64 -32 -16 Gas 6.1 12.4 103 59 -36 -18 Other minerals 3.1 7.1 131 82 -27 -7 Food processing 24.0 44.7 87 34 -41 -31 Textiles, clothing, leather 14.1 77.4 449 177 74 42 Wood products 7.2 12.5 73 32 -45 -32 Paper products 2.7 11.7 331 132 37 19 Petroleum & coal products 5.3 18.8 253 121 12 13 Chemicals, rubber & plastics 9.4 35.8 282 120 21 13 Non-metallic mineral products 1.9 6.8 267 125 17 15 Other manufactured products 20.0 95.0 375 201 51 55 Electricity, water & gas 2.8 10.2 268 118 17 12 Construction 22.1 75.4 241 125 8 16 Trade & transport 25.0 101.0 304 120 28 13 Other private services 36.2 142.5 293 114 25 10 Other public services 8.6 46.9 447 61 74 -18 Total, all sectors 227.0 769.5 215 95 Note: The projections for the period to 2010 maintain initial protection data, while those for the period 2010 to 2020 in columns 5 and 7 assume that the Uruguay Round, including China, has been fully implemented by 2010. Source: GTAP V3 database and authors' model results. 18 Trade, Global Policy, and the Environment Against these massive structural changes that Since liberalization is expected to raise GDP growth traditionally accompany economic growth, the mod- rates as well,4 we also simulate the APEC reform as- el's projected changes caused even by very large pol- suming each APEC economy's annual GDP growth icy shocks are relatively modest. Table 2.3 shows, for rate over the 20-year implementation period (2000 to example, how much additional impact by 2010 the 2020) is half a percentage point higher than in the base Uruguay Round's implementation would have on the case, due to faster total factor productivity growth. output of different sectors in Indonesia, both without The impact of these reforms would have to be judged and then with China included, and then how much as rather small in most sectors, relative to the large extra impact the APEC reforms to 2020 would add. changes that normal economic growth is projected to Table 2.3 Percentage changes in sectoral output levels in Indonesia following Uruguay Round and APEC trade reform by 2010 and 2020 APEC Liberalization 2020 (with extra GDP growth of Uruguay Round Uruguay Round APEC 0.5% pa in (without China) (with China) Liberalization APEC 2010 2010 2020 economies) Paddy rice -0.6 -0.3 -1.6 5.9 Other grains 3.2 4.7 14.9 22.6 Non-grain crops -5.1 -4.6 -13.4 -4.5 Livestock -0.2 0.1 3.1 13.2 Forestry -3.4 -1.1 -0.2 9.4 fisheries - 1.1 -0.7 -4.1 5.8 Coal -12.1 -7.1 18.4 31.1 Oil -5.4 -3.3 0.6 11.9 Gas -5.4 -3.4 0.7 11.1 Other minerals -8.1 -5.2 -1.6 8.3 Food processing -0.6 -0.3 -1.7 5.8 Textiles, clothing, leather 61.9 38.5 -2.6 2.9 Wood products -6.9 -2.4 1.2 11.5 Paper products -7.8 -3.7 6.7 17.8 Petroleum & coal products 0.9 0.5 -2.1 7.0 Chemicals, rubber & plastics 1.1 2.5 9.2 20.8 Non-metallic mineral products -7.5 -4.4 23.8 33.6 Other manufactured products -19.6 -12.3 -1.9 7.4 Electricity, water & gas 2.5 1.5 1.1 10.7 Construction 0.5 -0.1 - 1.5 5.9 Trade & transport -2.4 -1.3 4.9 16.3 Other private services -2.0 - 1.4 1.3 12.1 Other public services -0.6 -0.5 - 1.0 9.3 Real GDP growth 1.9 1.4 1.2 10.8 Source: Authors' model results. Will Trade Liberalization Harm the Environment? 19 generate (compare Tables 2 and 3). Nonetheless they types of environmental degradation. The authors use bring substantial increases in Indonesia's economic a case study approach to project anticipated changes welfare as traditionally measured even by compara- in technology to 2020. Twelve case studies generated tive static models such as the one used here: the Uru- data reflecting the views of experts assuming a con- guay Round with China included boosts real GDP for tinuation of current policies. Specialists such as chem- Indonesia by 1.4 percent (or 1.9 percent if China were ical engineers, hydrologists, environmental scientists, to be excluded), and the APEC reform (to 2020) adds energy experts and agricultural scientists were con- another 1.2 percent - even ignoring the likelihood that sulted on the technologies likely to be adopted in com- GDP growth would be accelerated by reform. ing decades.6 For water use there are data on the However, such welfare measures ignore chang- volume of water used and discharged by sector. Four es in resource depletion and the environment as a measures of the water pollution content of the efflu- consequence of the increased level and changed com- ent are provided: biological oxygen demand (BOD), position of Indonesia's output. Many environmental chemical oxygen demand (COD), dissolved solids groups would claim that adverse resource depletion (DS), and suspended solids (SS). The available air pollutant indicators are carbon dioxide and oxides of and environmental degradation effects of trade poli- supur n irgn cy reform will be substantial, but very few empirical sulphur and nitrogen. studies have sought to test that hypothesis. On envi- Based o tata f Ducinean oersi(1993) we assemble a matrix of environmental coefficients ronmental degradation, the following section suggests estimate the to esiaeteenvironmental impact per unit of eco- a way to examine how the changes in the aggregate nomic activity in each sector for 1992, 2010 and 2020 level of output, the composition of that output and in by assuming trends in environental parameters per the inputs and technologies used is likely to impact unit of output are linear over the period 1985-2020. on air and water pollution levels. The paper then pro- The GTAP 1992 benchmark database for Indonesia is vides some empirical results for Indonesia's environ- calibrated to this 1992 matrix of total emissions to ment, followed by a discussion of results on resource derive environmental damage coefficients per unit of depletion. GTAP sectoral output in that base year. The propor- tional changes in these environmental coefficients over 2. Adding an Environmental Module to te time are then multiplied by the GTAP 1992 environ- P!rojections Model mental coefficients to obtain GTAP environmental Accompanying economic growth and market reform coefficients for 2010 and 2020. This approach captures are changes in the scale of output, in tastes, in the rel- the expected change in environmental coefficients in ative size of sectors, and in inputs and production a consistent way that is used to augment the GTAP technologies. These can all affect the level of pollu- analysis. tion. How can we model these interacting forces and Three sources of environmental effects of policy decompose the projected changes in environmental changes are able to be identified: the change in the degradation to determine how they drive environmen- level of aggregate economic activity, the change in the tal change? contribution of each sector to output, and the change The model providing the projections of structural in production technology. This decomposition is use- change and trade liberalization presented above pro- ful for disentangling the causes of changes in envi- vides a starting point, to which needs to be added ronmental damage.7 Define the total change in environmental side modules to analyse the implica- pollution (P) as the sum of the changes in pollution tions of these economic changes for environmental in each sector (P ): degradation.5 In this paper we use side modules to n project environmental outcomes in Indonesia for wa- -L i" ter use, water pollution and air pollution. The data for the side modules are based on a comprehensive The change in pollution in each sector j is the sum of environmental input-output data set prepared by the "aggregate activity" effect (A' ). the "intersectoral Duchin and others (1993) using data collected in In- composition" effect (C, ), and the "technology" effect donesia for 1985 and 2020 by industry for various (7) 20 Trade, Global Policy, and the Environment P3 =AJ+CJ+TC Xj*(xj -g)*E, In the aggregate activity effect, increased eco- Thirdly, there is the "technology" effect, which nomic activity leads to increased demand for all goods is modelled using Duchin and others' (1993) set of en- and services and therefore increased emissions. The vironmental parameters reflecting expert opinion on change in output due to the aggregate activity effect anticipated changes to production methods. Chang- is the proportional change in aggregate real output in es in technology will change the amount of degrada- the economy (g) multiplied by the initial output in tion caused by each unit of output in each sector. Total each sector (X1). This gives the change in the scale of emissions with the new coefficients are compared to output in each sector with all sectors growing at the total emissions with the old environmental coefficients aggregate growth rate of the economy. The change in in place. The first square bracketed term of the fol- the scale of output in each sector is then multiplied lowing equation reflects the new environmental coef- by the initial environmental coefficient for each sec- ficient (E') applied to both the aggregate activity and tor (E,) to give the change in environmental emissions the intersectoral composition components of changes in each sector due to the aggregate activity effect: in output. The second square bracketed part of the equation reflects the idea that the initial output in each A'= X* g* E sector will also be produced using the new technolo- gy and will therefore contribute to a change in The second effect is the intersectoral composi- emissions. tion effect. Because some sectors are more polluting than others, changes in the composition of output will [(AJ - AJ)+ (C,; - C})]+ [XJ * (E; - E°)] change pollution, even if aggregate output were to remain constant. The intersectoral effect is measured where by allowing the composition of output to change while A'= X *g*E' maintaining aggregate output at its initial level. Some and sectors contract and others expand. This has some sim- ilarities with Dean's (1996) composition effect, where = XX* (xj - g)* E, emissions decrease if income growth shifts preferences toward income elastic cleaner goods, but we model However, for policy changes such as trade liberaliza- the general equilibrium-deterrnined intersectoral ef- tion where we start from the appropriate updated fects. Both producers and consumers respond to the database, we assume that the new technology is in changed incentives, given their behavioural functions place and that the trade reform itself does not change and the various constraints on the economy. Demand the environmental damage coefficients. and supply of each commodity in each region of the world respond to changing relative prices, given the 3. Empirical Projections of Environmental elasticities implicit for each sector. The change in sec- Impacts in Indonesia of Structural and toral output due to the intersectoral composition ef- Policy Changes to 2020 fect is found by multiplying the initial output in each sector by the difference between the proportional Projected environmental effects due to growth and change in output in that sector (xj) and the aggre- structural changes gate proportional change in output in the economy This section uses the detailed environmental side (g) to give the change in the relative size of each sec- modules to analyse some of the environmental impli- tor. This change in the contribution of each sector is cations of first the growth and structural changes pro- multiplied by the initial environmental coefficient for jected for Indonesia and then the trade policy changes each sector to give that sector's change in environ- by 2010 and 2020. mental emissions due to the intersectoral composition Table 2.2 shows the 1992 and projected 2010 out- effect, (C,), where put levels for each sector, evaluated at 1992 prices, Will Trade Liberalization Harm the Environment? 21 and the proportional changes in output due to struc- not surprising. Carbon emissions increase by 134 per- tural changes associated with economic growth pro- cent in the first projected period and by 56 percent for jected over that period, assuming no trade policy the decade to 2020. Sulphur oxides increase by 132 changes. Changes over the subsequent decade also and 50 percent and nitrogen oxides increase by 162 are shown. With the large growth in the economy pro- and 65 percent. jected from 1992 to 2010 and 2010 to 2020, all sectors The aggregate output effect increases each sec- exhibit increased output levels in Indonesia but some tor's output, while the technology and intersectoral expand much more than others. We use environmen- composition effects may add to or dampen the im- tal side modules to estimate the effects of these chang- pact of increased aggregate output on emissions. Ta- es in output on air and water pollution. ble 2.5 decomposes these air pollution effects to give a more precise indication of the relative magnitudes Anz POLLUTION of the aggregate activity, the intersectoral composi- tion and the technology effects. The table suggests the Atmospheric emission changes are estimated for car- aggregate activity effects are the main driving force bon dioxide and oxides of sulphur and nitrogen. Ta- behind the increase in projected emissions, but that ble 2.4 lists the initial 1992 level and projected new the intersectoral composition effects of structural levels of emissions for 2010 without the Uruguay change adds to that effect for all air pollutants. This is Round or APEC being implemented, and 2020 after because there is a relatively high increase in the con- the Round's implementation but without APEC trade tribution to output of high air polluting sectors such reform. Large increases are projected for all of these as the electricity, water and gas sector and the trade air pollutants. Since the Indonesian economy is pro- and transport sector. Sectors that are not very high jected to grow by 215 percent between 1992 and 2010 air polluters, such as agricultural sectors, tend to de- and a by further 95 percent by 2020, this finding is cline in relative importance. Table 2.4 Recent and projected levels of atmospheric emissions in the base cases, Indonesia, 1992, 2010, and 2020 (kt) 1992 2010 I 2020 carbon sulphur nitrogen carbon sulphur nitrogen carbon sulphur nitrogen Paddy rice I 0.0 0.1 2 0.0 0.1 3 0.0 0.2 Other grains 16 0.0 0.9 20 0.0 1.1 21 0.0 1.2 Non-grain crops 241 0.3 14.0 378 0.5 22.0 415 0.5 24.2 Livestock 310 0.4 17.9 677 0.8 39.1 931 1.1 53.8 Forestry 246 0.3 14.4 485 0.6 28.3 682 0.8 39.8 Fisheries 531 0.6 31.1 882 1.1 51.6 1,014 1.2 59.3 Coal 853 25.2 5.8 956 28.2 6.5 589 17.4 4.0 Oil 4,463 53.4 31.1 9,187 109.9 64.0 14,244 170.3 99.2 Gas 4,096 0.8 39.2 6,129 1.2 58.7 7,549 1.5 72.3 Other minerals 409 11.1 1.5 650 17.6 2.4 837 22.7 3.1 Food processing 489 13.0 1.8 752 19.9 2.8 890 23.5 3.3 Textiles, clothing, leather 293 7.7 1.1 770 20.1 3.0 1,160 29.9 4.6 Wood products 481 12.9 1.8 880 23.6 3.2 1,167 31.3 4.2 Paper products 217 6.3 1.5 712 20.8 5.0 1,317 38.9 9.7 Petroleum &coal products 1,305 17.4 8.3 4,047 54.4 25.7 8,302 112.4 52.8 Chemicals, rubber & plastics 3,330 35.6 26.6 5,930 65.9 47.3 4,867 60.8 38.6 Non-metallic mineral products 894 25.6 5.4 2,503 73.0 16.8 4,448 131.9 32.5 Other manufactured products 880 23.0 3.4 1,997 52.0 7.8 2,074 53.5 8.2 Electricity, water &gas 7,843 168.2 102.8 18,045 347.1 241.6 26,637 434.2 366.8 Construction 10,547 69.2 37.9 25,007 164.1 89.9 42,587 279.5 153.1 Trade & transport 10,322 129.8 532.5 30,564 384.5 1,578.4 52,865 665.4 2733.0 Other private services 193 1.2 0.8 559 3.4 2.2 943 5.7 3.7 Other public services 709 4.1 2.8 2,882 16.8 11.4 3,724 21.6 14.8 Total, all sectors 48,668 606 882 114,014 1,405 2,309 177,264 2,104 3,782 Note: 2020 levels indude Uruguay Round implementation. Source: Authors' model results. 22 Trade, Global Policy, and the Environment While the aggregate activity effect, and to a much module.8 Manufacturing sectors face two offsetting lesser extent the intersectoral composition effect, in- trends in their use of water. Growth occurs in water- crease air pollution during the period to 2020, many intensive sectors like pulp and paper, but new tech- sectors' emissions of carbon and oxides of sulphur and nologies for conserving water are expected to be nitrogen grow less rapidly than output because of adopted over time. Overall there is a significant in- improvements in energy efficiency. This is shown by crease in water uptake in the textiles, other manufac- the technique effect which is negative for all air pol- turing and pulp and paper sectors. Even by 2010 these lutants in table 2.5, reflecting the improved technolo- more than double their water use, while household gies expected to become available. water use increases by almost 50 percent. However increases in water use are dwarfed by the savings in WATER USE AND POLLUTION water uptake for paddy rice, which is the largest user Table 2.6 presents water use and water pollution re- of water in our model. That comes from the signifi- sults, calculated for the various sectors using GTAP cantly improved efficiencies anticipated in irrigation simulation results and a water use and pollution side delivery systems as well as froxn the changing inter- Table 2.5 Decomposition of changes in pollution as a consequence of economic growth and structural changes, Indonesia, 1992-20 10 and 2010-2020 1992-2010 Total Intersectoral pollution Aggregate composition Technology change' activity effect effect effect Carbon (kt) 65,346 [134] 104,607 10,149 -49,409 Sulphur (kt) 799 [132] 1,302 214 -716 Nitrogen (kt) 1,427 [162] 1,897 392 -862 Water in (bm3)b -12 [-4] 685 -388 -309 Water out (bm3) 0.8 [126] 1.3 0.7 - BOD (kt) 81 [52] 337 176 -433 COD (kt) 341 [64] 1,149 726 -1,534 DS (kt) -17 [-46] 79 -47 -48 SS (kt) 105 [23] 1,002 638 -1,536 2010-2020 Total pollution Aggregate Intersectoral Technology change' activity composition effect Carbon (kt) 63,982 [56] 107,244 16,904 -60,166 Sulphur (kt) 707 [50] 1,323 276 -893 Nitrogen (kt) I ,495 [65] 2,165 366 - I ,035 Water in (bm3)b -109 [-36] 296 -167 -236 Water out (bm3) 0.4 [29] 1.3 1.0 -2 BOD (kt) -13 [-5] 223 146 -382 COD (kt) -2 [-0] 822 587 -1412 DS (kt) -13 [-65] 19 -12 -19.5 SS (kt) -211 [-37] 545 474 -1231 a Percentages changes from base case are shown in square parentheses. b. This does not include the change in household water use. Source: Authors' model results. Will Trade Liberalization Harm the Environment? 23 Table 2.6 Recent and projected levels of water use and quality in the base cases, Indonesia, 1992, 2010, and 2020 Base level for 1992 Water in Water out Dissolved Suspended (millions (millions BOD COD solids solids ofM3) ofm3) (kt) (kt) (kt) (kt) Paddy rice 313,072 0 0 0 0 0 Livestock 8 0 0 0 0 0 Food processing 124 97 21 30 37 49 Textiles, clothing, leather 102 102 1 8 72 0 87 Paper products 217 97 64 217 0 70 Chemicals, rubber, plastics 5 4 0 0 0 0 Other manufactures 307 307 54 216 0 261 Households 10,704 0 0 0 0 0 Total, all sectors 324,538 608 157 534 37 466 2010 Water in Water out Dissolved Suspended (miMlons (millions BOD COD solids solids of m3) ofm3) (kt) (kt) (kt) (kt) Paddy rice 300,439 0 0 0 0 0 Livestock 16 0 0 0 0 0 Food processing 132 104 12 17 20 24 Textiles, clothing, leather 278 278 32 127 0 129 Paper products 519 262 III 402 0 83 Chemicals, rubber, plastics 9 9 0 0 0 0 Other manufactures 720 720 82 329 0 334 Households 15,712 0 0 0 0 0 Total, all sectors 317,825 1,372 238 875 20 571 2020 Water in Water out Dissolved Suspended (millions (millions BOD COD solids solids ofm3) ofm3) (kt) (kt) (kt) (kt Paddy rice 190,557 0 0 0 0 0 Livestock 21 0 0 0 0 0 Food processing 104 82 5 7 7 7 Textiles, clothing, leather 460 460 37 147 0 115 Paper products 645 390 115 449 0 37 Chemicals, rubber, plastics I I 10 0 0 0 0 Other manufactures 822 822 65 263 0 206 Households 18,494 0 0 0 0 0 Total, all sectors 211,114 1,764 223 866 7 365 Note: 2020 levels include Uruguay Round implementation. Source: Authors' model results. sectoral composition of output. As a consequence, Between 1992 and 2010, we project water dis- total water withdrawals fall over the projection peri- charge to increase by 126 percent, with a further 29 ods, by 4 percent to 2010 and by a further 36 percent percent increase by 2020 (column 2 of table 2.5). The by 2020. decomposition in table 2.5 shows that the intersectoral 24 Trade, Global Policy, and the Environment composition effect augments the aggregate activity eralization. Leaving aside the final scenario in which effect a little. The relative increases are in textiles, pulp economic growth is assumed to be boosted by APEC and paper and other manufactures, which are all large liberalization (discussed separately below), some sec- producers of waste water. However, improved tech- tors reduce and other sectors increase their output nologies dampen the effect of increases in water dis- level because of trade reform, in contrast to the mid- charged. dle columns of table 2.2 for structural change projec- The water pollution changes we model are bio- tions where all sectors increase their output. We logical oxygen demand (BOD), chemical oxygen de- therefore can expect the composition effects to be mand (COD), dissolved solids (DS) and suspended much stronger relative to the aggregate activity ef- solids (SS). These emissions are assumed to be deter- fects in these reform cases, in contrast to the growth mined by the quantity of waste water produced. Once and structural change scenarios discussed above. we have calculated the proportional change in water The results in table 2.3, coming from a global output for each sector, we can estimate the change in model, include the effects on output levels in Indone- water pollution by sector. Because of the aggregate sia of changes in protection and relative prices in oth- activity effect, emissions of all water pollutants ex- er regions. The sector that experiences the greatest cept for dissolved solids rise between 1992 and 2010 proportional increase in Indonesia with Uruguay (table 2.5a). However, emissions rise by significantly Round implementation is textiles and clothing, with less than the proportional increase in total output in a 60 percent boost to output anticipated if China is Indonesia. This is mainly due to the improved tech- kept out of the WTO, or just under 40 percent if Chi- nology assumed to be available in 2010. The intersec- na is able to join. With additional APEC liberaliza- toral composition effect for all water pollutants, with tion, the effects on the textile sector are much less the exception of dissolved solids, is positive due to pronounced because MFA quotas are assumed to have the increased relative signiibeen already phased out as part of the Round's im- dtheincrieas. Th latve m sitnieffctnc bothe projcted pe- plementation. The sectors that tend to do well with APEC reform are instead the coal and non-metallic rnods moves production into the sectors we model as minerals - sectors which Indonesia's own policies being the most important producers of water pollut- tend to discriminate against. The corn (coarse grains) ants, particularly textiles, pulp, paper, and other man- ' ' ' ' ~~~~~~~~~~sector also is projected to do well. ufactures. For dissolved solids, the composition effect What do these output changes do to pollution is negative with the reduced significance of the food levels? Again, we consider effects on first air and then processing sector. water, recognising that emissions will increase in some For the period to 2010, the assumed technology sectors and fall in others in response to Uruguay effect offsets over 80 percent of the aggregate activity Round and APEC trade reforms. and intersectoral effects for all water pollutants. And for the period to 2020, the technology effect is suffi- Air pollution ciently strong to overturn the positive aggregate ac- thvity and intersectoral effects to give a net reduction Table 2.7 indicates that a reduction in air pollution is finityand pollution torall waterf ts togiveanetreducfion projected for Indonesia under Uruguay Round liber- alization (including China), rather than the increase Projected environmental effects of Uruguay Round feared by environmentalists. The reduction from 2010 and APEC trade reforns baseline levels is 0.6 percent for carbon and sulphur oxides and 1.0 percent for nitrogen oxides. The de- How much difference will it make to those environ- composition in table 2.7 shows that the aggregate ac- mental effects of economic growth to impose on In- tivity effect adds to air pollution but the change in donesia and others some trade reforms? The first two the intersectoral composition of output reduces air columns of table 2.3 show the proportional change in pollution by more. output due to Uruguay Round liberalization, first When the total change in emissions is examined without and then with the inclusion of China as a by sector (table 2.8), we find that the most significant WTO member. The second pair of columns show the reduction is contributed by the trade and transport projected sectoral changes in output due to APEC lib- sector. The output of textiles rises more than that in Will Trade Liberalization Harm the Environment? 25 Table 2.7 Decomposition of pollution effects from Uruguay Round trade reform (including China), Indonesia, 2010 (percent change from 2010 baseline level shown in curved parentheses, percent of the 1992-2010 absolute change is in brackets) Total Aggregate Intersectoral change Activity composition Carbon (kt) -733 1,585 -2,318 (-0.6) (1.4) (-2.0) [-1.1] [2.4] [-3.5] Sulphur (kt) -8 20 -27 (-0.6) (1.4) (-1.9) [-1.0] [2.4] [-3.4] Nitrogen (kt) -22 32 -54 (-1.0) (1.4) (-2.3) [-1.5] [2.2] [-3.8] Water in -0.8 4 -5 (billion i3) (-0.3) (1.4) (-1.6) [-7] [35] [-42] Water out 0.01 0.02 -0.01 (billion m) (0.6) (1.4) (-0.8) (billion ~~~[I .1] [2.4] [-1I.3] BOD (kt) -2.0 3 -5 (-0.9) (1.4) (-2.3) [-2.5] [4.1] [-6.6] COD (kt) -6.5 12 -19 (-0.7) (1.4) (-2.1) [- 1.9] [3.6] [-5.5] DS (kt) -0.05 0.3 -0.3 (-0.3) (1.4) (-1.7) [-0.3] [1.6] [-2.0] SS (kt) 5.3 8 -3 (0.9) (1.4) (-0.5) [5.0] [7.6] [-2.5] Source: Authors' model results. any other sector, but since it is starting from a rela- Indonesia are possibly more than offset by a reduc- tively low base of air emissions, the increase in air tion in emissions in China following its accession to pollutants from this sector is more than outweighed WTO and thereby its assumed greater access to tex- by reductions occurring in other sectors. If China is tile markets in the United States and the EU.9 not included in the WTO and hence by assumption With additional APEC trade liberalization, air does not liberalize its trade, the reductions in Indone- pollution is projected to increase but, as shown in ta- sia's air pollution almost double relative to the reduc- ble 2.9, the increases are only between 2 and 4 per- tions shown in table 2.7 when China is included. This cent. Moreover, a small number of sectors drive the is primarily because the Indonesian textile and cloth- results. For example, the trade and transport sector ing sector does not grow as much when China is in- contributes over 45 percent of the increase in air pol- cluded and hence that sector does not pull as many lution (unreported further decomposition of results resources away from other more-polluting sectors. in table 2.9). This makes it relatively easy to target However, the greater carbon and other emissions in that pollution with environmental taxes to reduce the Table 2.8 Sectoral decomposition of the total change in emissions due to Uruguay Round implementation, Indonesia, 2010 Carbon Sulphur Nitrogen Water in Water out BOD COD DS SS (kt) (kt) (kt) (bm3) (bm3) (kt) (kt) (kt) (kt) Paddy rice -0.01 0.00 0.00 -0.78 0.00 0.00 0.00 0.00 0.00 Other grains 0.93 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00 Non-grain crops -17.33 -0.02 -1.01 0.00 0.00 0.00 0.00 0.00 0.00 Livestock 0.95 0.00 0.05 0.00 0.00 0.00 0.00 0.00 0.00 Forestry -5.09 -0.01 -0.30 0.00 0.00 0.00 0.00 0.00 0.00 Fisheries -6.35 -0.01 -0.37 0.00 0.00 0.00 0.00 0.00 0.00 Coal -67.89 -2.00 -0.46 0.00 0.00 0.00 0.00 0.00 0.00 Oil -301.33 -3.60 -2.10 0.00 0.00 0.00 0.00 0.00 0.00 Gas -207.77 -0.04 -1.99 0.00 0.00 0.00 0.00 0.00 0.00 Other minerals -33.86 -0.92 -0.13 0.00 0.00 0.00 0.00 0.00 0.00 Food processing -2.03 -0.05 -0.01 0.00 0.00 -0.03 -0.05 -0.05 -0.07 Textiles, clothing, leather 296.17 7.71 1.15 0.11 0.11 12.20 48.81 0.00 49.57 Wood products -21.47 -0.58 -0.08 0.00 0.00 0.00 0.00 0.00 0.00 Paper products -26.13 -0.76 -0.18 -0.02 -0.01 -4.08 -14.74 0.00 -3.05 Petroleum & coal products 21.45 0.29 0.14 0.00 0.00 0.00 0.00 0.00 0.00 Chemicals, rubber & plastics 150.03 1.67 1.20 0.00 0.00 0.00 0.00 0.00 0.00 Non-metallic mineral products -108.88 -3.17 -0.73 0.00 0.00 0.00 0.00 0.00 0.00 Other manufactured products -246.05 -6.40 -0.96 -0.09 -0.09 - 10.14 -40.55 0.00 -41.18 Electricity, water & gas 276.08 5.31 3.70 0.00 0.00 0.00 0.00 0.00 0.00 Construction -27.51 -0.18 -0.10 0.00 0.00 0.00 0.00 0.00 0.00 Trade & transport -385.10 -4.84 -19.89 0.00 0.00 0.00 0.00 0.00 0.00 Other private services -7.71 -0.05 -0.03 0.00 0.00 0.00 0.00 0.00 0.00 Other public services -13.83 -0.08 -0.05 0.00 0.00 0.00 0.00 0.00 0.00 Total, all sectors -732.75 -7.74 -22.09 -0.78 0.01 -2.05 -6.53 -0.05 5.27 Source: Authors' model results. Will Trade Liberalization Harm the Environment? 27 impact of trade reform on emissions, should that small of 0.3 percent with Uruguay Round implementation, increase be considered a problem. while table 2.9 shows that water withdrawals reduce The key point to draw from these results, how- by a further 1.6 percent with APEC liberalization. ever, is that the air pollutive effects of even these major These water use reductions are largely due to a re- trade liberalizations is tiny (at less than 4 percent of duction in paddy output. the base level), and is especially small compared with Most water pollutants also decline with Uruguay the increases that will accompany normal economic Round implementation, as shown in table 2.7. The de- growth and structural changes, as can be seen by the clines are just under 1 percent for BOD, COD and dis- numbers in square parentheses in tables 2.7 and 2.9. solved solids, but there is an increase of just under 1 percent in suspended solids. For APEC liberalization, table 2.9 reports a 2.4 percent increase in BOD and Water use and pollution COD but reductions in solids of between 1 and 2 per- Water withdrawals are reduced by both trade liberal- cent. Thus as with air pollution, these results show izations. Table 2.7 shows a reduction in withdrawals that trade reform will at most add only a very small Table 2.9 Decomposition of pollution effects in Indonesia under APEC liberalization, 2020 (percent change from 2020 baseline level shown in parentheses, percent of the 1992-2020 absolute change in brackets) Total Aggregate Intersectoral change activity composition Carbon (kt) 3,736 2,124 1,612 (2.1) (1.2) (0.9) [2.9] [1.6] [1.3] Sulphur (kt) 72 25 47 (3.4) (1.2) (2.2) [4.8] [1.7] [3.1] Nitrogen (kt) 144 45 99 (3.8) (1.2) (2.6) [4-9] [1.6] [3.4] Water in -3.0 2.3 -5.3 (billion m3) (-1.6) (1.2) (-2.8) [-2.5] [1.9] [-4.4] Water out -0.002 0.02 -0.02 (billion m3) -0. 1) (1.2) (-1.3) [-0.2] [1.8] [- 1.9] BOD (kt) 5.4 2.7 2.7 (2.4) (1.2) (1.2) [7.9] [3.9] [4.0] COD (kt) 21.1 10.4 10.8 (2.4) (1.2) (1.2) [6.2] [3.0] [3.2] DS (kt) -0.13 0.09 -0.21 (-1.8) (1.2) (-3.1) [-0.4] [0.3] [-0.7] SS (kt) -4.5 4.4 -8.9 (-1.2) (1.2) (-2.4) [-4.2] [4.1] [-8.4] Source: Authors' model resuks. 28 Trade, Global Policy, and the Environment amount to water use and pollution, an amount that grow substantially faster (by half a percentage point would not be discernible alongside the increased BOD per year over the 20-year implementation period to and COD pollution associated with the general ex- 2020) through a boost to their total factor productivi- pansion of the economy over time. ty growth. The impact of APEC reform including that faster growth on the pollution results is shown in ta- What if trade reform boosts economic growth? ble 2.10, based on the output effects summarized in The above simulations of trade reform are from a com- the final column of table 2.3. Not surprisingly, that parative static model and so do not include the im- change in assumption raises the effect of liberaliza- pact that trade reform would have in boosting tion on pollution. Even so, the numbers are relatively economic growth. Hence it understates the extent of small: air pollution is 12-15 percent greater and water pollution that might result. To get a feel of how large pollution 6-12 percent greater, than would have been that bias might be, we re-ran the APEC liberalization the case in 2020 (instead of no more than 4 percent as but assumed that APEC economies' GDPs would when we assume no growth effect of APEC reform). Table 2.10 Decomposition of pollution effects in Indonesia under APEC liberalization, with 0.5 percent p.a. extra GDP growth in APEC economies, 2020 (percent change from 2020 baseline level shown in parentheses, percent of the 1992-2020 absolute change in brackets) Total Aggregate lntersectoral change activity composition Carbon (kt) 21,142 19,091 2,051 (12) (I11) (1) [16] [15] [2] Sulphur (kt) 283 227 57 (14) (I11) (3) [19] [15] [4] Nitrogen (kt) 557 407 149 (I15) (I11) (4) [19] [14] [5] Water in I1 21 -9 (billion m3) (6) (I1) (-5) [9] [17] [-8] Water out 0.15 0.19 -0.04 (billion in3) (9) (I1l) (-2) [13] [16] [-3] BOD (kt) 27 24 3 (12) (I11) (1) [39] [35] [4] COD (kt) 104 93 I1 (12) (I11) (1) [31] [28] [3] DS (kt) 0.4 0.8 -0.4 (6) (I11) (_S) [2] [3] [-1] SS (kt) 26 39 -14 (7) (I11) (-4) [24] [37] [-13] Source: Authors' model results. Will Trade Liberalization Harm the Environment? 29 This amount is less than one fifth of the air pollution 4. Conclusions (and a somewhat larger fraction of the water pollu- If present environmental policies remain unchanged, tion) that would result from the normal output ex- projected economic growth and structural changes pansions and structural changes that would take place over the next two decades would, according to the without reform. Moreover, that extra pollution due above simulations, add to environmental degradation to accelerated growth is accompanied by a much and resource depletion in Indonesia. This is not an greater boost to economic welfare as conventionally argument against economic growth of course, but measured than when we assume there is no growth rather for the need to introduce or strengthen the en- effect of trade reform: Indonesia's GDP in 2020 is 10.8 . . . forcement of environmental and resource policies so percent higher in this growth-enhancing case, com- as to internalize some of the externalities associated pared with only 1.2 percent higher in the earlier APEC with output and consumption expansion. When opti- reform case that assumed no growth effect. Clearly mal environmental (and other) policies are in place this compensates generously for the extra pollution and are continually adapted to remain optimal over and provides great scope for spending some of that time, it is necessarily the case that economic growth extra income on pollution abatement. enhances social welfare. That may not preclude a Resource depletion worsening of environmental degradation or further resource depletion, but at least those changes would The impact of trade liberalization on natural resource be optimal from that society's viewpoint, given the depletion can be crudely inferred from changes in pri- actual or opportunity cost of avoidance or abatement. mary production. In the case of the Uruguay Round, Likewise, trade reform can contribute to environmen- the first column of table 2.11 shows that most primary tal damage and resource depletion, but again that will production is reduced by that liberalization. This sug- not be nationally welfare-reducing so long as optimal gests that less rather than more depletion of Indone- environmental (and other) policies are always in place. sia's natural resources will take place because of the A concern of some people, though, is that devel- Uruguay Round reforms. Of course there are some off- oping countries' environmental and resource policies setting changes in other economies, but the final col- may not be optimal even nationally, let alone from a urmn of table 2.11 shows that in aggregate the changes global perspective, and that trade liberalization with to natural resource use from the Round will be tiny. no change in those environmental and resource poli- Table 2.1 1 Percentage changes in resource-sector output levels in various regions of the world following Uruguary Round trade reform (including China), 2010 Other Other APEC developing & APEC high- Other high- developing transition income income Total Indonesia economies economies economies economies world Paddy rice -0.3 2.9 -1.3 -1.0 -3.1 0.48 Non-grain crops -4.6 4.3 -0.4 2.0 -2.9 0.59 Livestock 0.1 -1.4 -1.6 0.9 1.2 -0.06 Forestry -I . -0.7 -0. I -0.0 1.9 -0.03 Fisheries -0.7 -7.4 0.1 -0.4 5.1 -0.21 Coal -7.1 -0.6 0.2 -0.3 1.0 0.03 Oil -3.3 -2.9 0.2 0.1 0.4 -0.04 Gas -3.4 -1.4 0.1 0.5 0.1 0.06 Other minerals -5.2 -5.0 -0.7 -1.4 1.9 -0.39 Source: Authors' model results. 30 Trade, Global Policy, and the Environment cies therefore could be bad for the environment. Hence is often lessened. A case in point is the inclusion of the reason in the present empirical study for looking China in the WTO that (hopefully) allows China great- at trade reform without changing environmental and er access to US and EU markets under the Uruguay resource policies.10 Round Agreement on Textiles and Clothing: the above This case study of Indonesia suggests that trade results show that this would reduce Indonesia's ca- policy reforms slated for the next two decades in most pacity to expand exports of light manufactures and cases would improve the environment (at least with so keep resources in more-polluting activities in In- respect to air and water pollution) and reduce the donesia - but it would mean China moves away from depletion of natural resources in that country and in some of its very pollutive coal-intensive heavy man- the worst cases would add only very slightly to envi- ufacturing, thereby potentially reducing not only lo- ronmental degradation and resource depletion even cal air pollution but also global warming. The latter without toughening the enforcement of existing en- could be quantified by extending the environmental vironmental and resource regulations or adding new side modules developed here for Indonesia to other ones. The economic gains from the trade reforms and countries and regions of the world included in the the scope for adopting well-targeted environmental GTAP model.12 and resource policies to reduce any serious damage We have set up a framework for modelling and are such that social welfare almost certainly is going decomposing the major environmental impacts of to be improved substantially by these liberalizations.11 growth and policy reform in as transparent a way as Furthermore, a related study (Strutt 1998, Ch. 3) possible. The results presented here indicate sectors which focuses on land degradation through soil erosion of particular concern, given available information and and associated off-site damage draws a similar conclu- our choice of model.13 Needless to say, caution should sion. That study incorporates feedback effects of that be used in interpreting the above results, particularly damage on land productivity and thereby is able to val- given the still poor quality of environmental data. ue the loss of production associated with that erosion. There are sectors and types of environmental dam- Again using GTAP to model the effects of implement- age that are not adequately represented here.14 Clear- ing the Uruguay Round agreements, the study finds that ly, this kind of research is in its infancy; there are many the aggregate output expansion and shift in its compo- future directions and areas where improvements can sition does add slightly to soil erosion, but that the cost be made. For example as improved environmental of the damage caused by that increased erosion is minis- data become available for other regions, the environ- cule, amounting to less than 0.2 percent of the national mental effects in other countries could be traced. In economic welfare gain (as traditionally measured) from particular, improved modelling of air pollution across the Uruguay Round liberalization. countries will be possible with the upgraded energy Moreover, this study has focused only on one component of future versions of the GTAP data base.15 country's resources and environment. The natural More-direct inclusion of emissions and abatement resource impact of the Uruguay Round can be seen in activities in the GTAP model may also be desirable, table 2.11 to be positive rather than negative in most rather than having just side modules. Among other other regions too. It is negative mainly in Western things, the model could then be modified to enable Europe ('Other high-income economies'), where re- induced substitution towards less environmentally source policies are well developed and could easily damaging output and the adoption of less-polluting be adapted to cope with any undesired increase in technologies when environmental taxes are imposed exploitation. And it happens that when environmen- or increased. Endogenizing environmental policies to tal damage occurs in Indonesia because of the change income growth,16 trade policy changes and changes in the composition of its output following trade re- in pollution would be another useful extension, pro- form, damage to the environment of other countries viding a rich future research agenda. Will Trade Liberalization Harm the Environment? 31 Appendix Table A2. 1 Import tariff rates in Indonesia without and with Uruguay Round liberalization, by sector, 2010 (percent) 2010 base 2010 after UR Paddy rice 9.0 9.0 Other grains 0.0 0.0 Non-grain crops 54.7 38.3 Livestock 4.8 4.8 Forestry 14.4 14.4 Fisheries 29.8 29.8 Coal 5.0 5.0 Oil 0.0 0.0 Gas 5.0 5.0 Other minerals 4.9 4.9 Food processing 12.3 11.3 Textiles, clothing, leather 28.7 22.5 Wood products 34.4 31.0 Paper products 8.0 8.0 Petroleum & coal products 4.7 4.7 Chemicals, rubber & plastics 6.6 6.6 Non-metallic mineral products 14.1 12.9 Other manufactured products 15.6 15.4 Electricity, water & gas 0.0 0.0 Construction 0.0 0.0 Trade & transport 0.0 0.0 Other private services 0.0 0.0 Other public services 0.0 0.0 Source: GTAP data base and authors' model results. 32 Trade, Global Policy, and the Environment Notes environmental policies here, only the scenario of Thanks are due to Christian Bach, Betina Dima-ranan current trends is used. and Faye Duchin for helpful inputs, and to confer- 7. The decomposition developed here is in some ence participants at the GETS Trade and Environment ways similar to the "scale," "composition," and Workshop, London, June 1998, the World Congress "technique" effects of income growth on the level of Environmental and Resource Economists, Venice, of environmental emissions discussed by Dean June 1998, and the Trade, Global Policy, and the En- (1996, 1998), Beghin and others (1997, 1998) also vironment Conference, particularly Shanta Devarajan, discuss such a three-way decomposition. David Roland-Holst, Tom Rutherford and David Tarr, for insightful comments. The authors are also grate- 8. Increases in household water use are taken from ful to the Australian Centre for International Agricul- estimates in Duchin and others (1993) and entered tural Research (under PN9449) and the World Bank exogenously, assuming Indonesia's population in- for financial assistance. creases to 263 million by 2020. 1. See Hertel (1997, especially chapters 2 and 3) and 9. When China is excluded, the group of "Other APEC McDougall (1997) for detailed descriptions of the developing economies" (which includes China) ex- GTAP model and data base. Updated information pand their output of textiles and dothing by only 8 is available at the following website: http:// percent following Uruguay Round implementation, www.agecon.purdue.edu/gtap. whereas with China included, that sector expands 2. Version 4 of the GTAP data base will include this 25 percent (Strutt 1998, chapter 5). split between skilled and unskilled labour. It will 10. For more on modelling the responses of environ- also break out a natural resource input calibrated mental policies to trade reforms (something which to the target elasticity of supply in resource-con- has not been attempted in the present study), see strained sectors (McDougall 1998). the recent paper on Mexican agriculture by Beghin 3. Growth rates for 1992-2010 are adapted from and others (1997). Anderson and others 1997; and Arndt and others 11. This is not inconsistent with the finding by Lindert 1997, while growth rates for 2010-2020 are adapted (1996) that there is virtually no evidence over many from Bach 1997. decades of net soil degradation in Indonesia. 4. See, for example, the theoretical reasons presented 12. A new paper by Cole, Rayner, and Bates (1998) in Grossman and Helpman (1991), and the rap- applies emission coefficients to another set of Uru- idly growing empirical evidence presented by guay Round output results using GTAP for the Baldwin (1992), Dollar (1992), Edwards (1992), world as a whole and places monetary values on Sachs and Warner (1995), Harrison (1996), and the estimated changes in emissions. While the lat- USITC (1997), and the references therein. ter values are open to question of course, their 5. The approach of augmenting CGE models with en- global results nonetheless are consistent with the vironmental side models has been taken by a above findings for Indonesia in suggesting that any number of researchers. For example, Bandara and increases in pollution from the Uruguay Round Coxhead (1995) look at soil erosion in a single are likely to reduce developing countries' welfare country model. Perroni and Wigle (1997) use an gains from liberalization by much less than 2 per- innovative side model to analyse global externali- cent while raising the welfare gains to some ties and abatement costs with GTAP. There have advanced economies. Another new empirical also been attempts to incorporate environmental study by Unteroberdoerster (1998), looks at APEC equations and parameters more directly into a CGE trade liberalization alone and again finds very model (for example, Xie 1996). small effects on the environment. 6. Other scenarios are also presented where the gov- 13. There are of course more sophisticated methods erinent is assumed to place heavier emphasis on of projecting economic growth, using endogenous environmental protection and resource conserva- growth and incorporating imperfect competition tion. Since we do not explicitly model improved and scale economies. Will Trade Liberalization Harm the Environment? 33 14. For example, the most excessive pollutant in In- Baldwin, R. 1992. "Measurable Dynamic Gains from donesian rivers is faecal coliform which exceeds Trade." Journal of Political Economy 100(1): 162-74. recommended standards by more than a thousand- Bandara, J.S., and I. Coxhead. 1995. "Economic Re- fold in some places (World Bank 1990, p. xxxi). We forms and the Environrnent in Sri Lanka." Agri- have not been able to include this in our present cultural Economics Discussion Paper 27/95, La analysis. Nor have we accounted for the human Trobe University, June. health effects of pollution (as was done for Chile Beghin, J., B. Bowland, S. Dessus, D. Roland-Holst, in Beghin and others 1998). and D. Van der Mennsbrugghe. 1998. "Trade, En- 15. The weakness of the energy data in version 3 of vironment and Public Health in Chile: Evidence GTAP led us to not focus particularly on energy from an Economywide Model." Chapter 3 in this in the current work. Details of the project aiming volume. to collect consistent data on energy quantity flows, Beghin, J., S. Dessus, D. Roland-Holst, and D. Van der prices and taxes to be incorporated into future Mennsbrugghe. 1997. "The Trade and Environ- versions of the GTAP data base are available at ment Nexus in Mexican Agriculture: A General http://www.agecon.purdue.edu/gtap/doe/ Equilibrium Analysis." Agricultural Economics index.htm. 17(2-3): 115-31. 16. The reasons for expecting citizens to seek a Chichilnisky, G. 1994. "North-South Trade and the tightening of environmental standards and regu- Global Environment." American Economic Review lations/ taxes on pollution and resource depletion 84(4): 851-74. as incomes rise, at least after middle-income Cole, M.A., A.J. Rayner, and J.M. Bates. 1998. "Trade status is reached, have been canvassed by, among Liberalisation and the Environment: The Case of others, Selden and Song (1994), Grossman and the Uruguay Round." The World Economy 21(3): Krueger (1995), and Hettige, Mani, and Wheeler 337-47. (1998). Corden, W.M. 1997. Trade Policy and Economic Welfare (second edition). Oxford: Clarendon Press. References Dean, J. 1996. "Testing the Impact of Trade Liberal- Anderson, K. and R. Blackhurst, eds. 1992. The Green- ization on the Environment." CIES Seminar Paper ing of World Trade Issues. Ann Arbor: University 96-11, University of Adelaide, August. of Michigan Press and London: Harvester Wheat- Dean, J. 1999. "Testing the Impact of Trade Liberal- sheaf. ization on the Environment: Theory and Anderson, K., B. Dimaranan, T. Hertel and W. Mar- .iatn ot Environm end tin. 1997. "Economic Growth and Policy Reform Dollar D 1992. "Outward-oriented Developing in the Asia-Pacific: Trade and Welfare Implications Dollar,eD. 1ouwrd-ore Developin by 2005." Asia-Pacific Economic Review 3(1): 1-18. ecnomi Rl LD Grow M or id evi- Anderson, K., and M. Pangestu. 1998. "Structural dence from 95 LDCs,1976-1985." Economic Devel- Changes in a Reforming World Economy: Impli- opment and Cultural Change 40(3): 523-44. cations for Indonesia." Bulletin of Indonesian Eco- Duchin, F., C. Hamilton, and G. Lange. 1993. Environ- nomic Studies 34(3). ment and Development in Indonesia: An Input- Anderson, K., and A. Strutt. 1996. "On Measuring the Output Analysis of Natural Resource Issues. Jakarta: Environmental Effects of Agricultural Trade Lib- BAPPENHAS-Ministry of Forestry for USAID. eralization." In M.E. Bredahl, N. Ballenger, J.C. Edwards, S. 1992. "Trade Orientation, Distortions and Dunmore, and T.L. Roe, eds., Agriculture, Trade and Growth in Developing Countries." Journal of De- the Environment. Boulder and London: Westview velopment Economics 39(1): 31-57. Press. Grossman, G.M., and E. Helpman. 1991. Innovation and Arndt, C., T. Hertel, B. Dimaranan, K. Huff, and R. Growth in the Global Economy. Cambridge: MIT McDougall. 1997. "China in 2005: Implications for Press. the Rest of the World." Journal ofEconomicIntegra- Grossman, G., and A. Krueger. 1995. "Economic tion 12(4): 505-47. Growth and the Environment." Quarterly Journal Bach, C. 1997. Personal communication. of Economics 110(2): 353-77. 34 Trade, Global Policy, and the Environment Hathaway, D.E., and M.D. Ingco. 1996. "Agricultural Martin, W., and D. Mitra. 1996. "Productivity Growth Liberalization and the Uruguay Round." Ch. 2 in in Agriculture and Manufacturing." Mimeo, the W. Martin and L.A. Winters, eds., The Uruguay World Bank, Washington, D.C., September. Round and the Developing Countries. Cambridge and McDougall, R.A., ed. 1997. Global Trade, Assistance, and New York: Cambridge University Press. Protection: The GTAP 3 Data Base. Purdue Univer- Harrison, A. 1996. "Openness and Growth: A Time- sity: Center for Global Trade Analysis. Series, Cross-Country Analysis for Developing , ed. 1998. Global Trade, Assistance, and Protec- Countries." Journal of Development Economics 48(2): tion: The GTAP 4 Data Base, Purdue University: 419-47. Center for Global Trade Analysis. Harrison, J., and K. Pearson. 1996. 'Computing Solu- Perroni, C., and R. Wigle. 1997. "Environmental Policy tions for Large General Equilibrium Models us- Modeling." Chapter 12 in T.W Hertel, ed., Global ing GEMPACK." Computational Economics Trade Analysis: Modeling and Applications. Cam- 9: 83-127. bridge University Press. Hertel, T.W., ed. 1997. Global Trade Analysis: Modeling Sachs, J.D., and A. Warner. 1995. "Economic Reform and Applications, Cambridge and New York: Cam- and the Process of Global Integration." Brookings bridge University Press. Papers on Economic Activity 1: 1-118. Hertel, T.W., W. Martin, K. Yanagishima, and B. Selden, T.M. and D. Song. 1994. "Environmental Qual- Dimaranan. 1996. "Liberalizing Manufactures *ity and Development: Is There a Kuznets Curve Trade in a Changing World Economy." Ch. 7 in for Air Pollution Emissions?" Journal of Environ- W. Martin and L. A. Winters, eds., The Uruguay mental Economics and Management 27(2): 147-62. Round and the Developing Countries, Cambridge and Strutt, A. Forthcoming. Economic Growth, Trade Policy, New York: Cambridge University Press. and the Environment in Indonesia. Draft PhD thesis, Hertel, T.W., and M.E. Tsigas. 1997. "Structure of University of Adelaide. GTAP." Chapter 2 in T.W Hertel, ed., Global Trade Unteroberdoerster, 0. 1998. "Trade Policy and Envi- Analysis: Modeling and Applications. Cambridge ronmental Regulation in the Asia Pacific: A Simu- University Press. lation." Mimeo, Brandeis University, Waltham Hettige, H., M. Mani, and D. Wheeler. 1998. "Indus- MA, March. trial Pollution in Economic Development- USITC. 1997. The Dynamic Effects of Trade Liberaliza- Kuznets Revisited." Chapter 5 in this volume. tion: An Empirical Analysis. Publication 3069, Wash- Krueger, A. 1977. "Growth, Distortions and Patterns ington, D.C.: U.S. International Trade Commission, of Trade Among Many Countries." Princeton Stud- October. ies in International Finance, Princeton, NJ. World Bank. 1990. "Indonesia: Sustainable Develop- Leamer, E. 1987. "Paths of Development in the Three ment of Forests, Land, and Water," Washington, Factor n-good General Equilibrium Model." Jour- D.C.: The World Bank. nal of Political Economy 95(5): 961-99. Xie, J. 1996. Environmental Policy Analysis: A General Lindert, P.H. 1996. "Soil Degradation and Agricultural Equilibrium Approach. Aldershot: Ashgate Publish- Change in China and Indonesia." CIES Seminar ing Limited. Paper 96-14, University of Adelaide, September. Liu, J., T.T. Vo, R. Tyers, and T. Hertel. 1997. "Disag- gregating Labor Payments by Skill Level in GTAP." Draft Technical Paper, Center for Global Trade Analysis, Purdue University. Chapter 3 Trade, Environment, and Public Health in Chile Evidencefrom an economy-wide model John Beghin, Brad Bowland, S6bastien Dessus, David Roland-Holst, and Dominique van der Mensbrugghe 1. Introduction suring pollution incidence or other environmental T n he policy significance of trade and environ- variables. Secondly, this paper is empirical, and in- ment linkages has increased sharply in recent tended to strengthen the basis of evidence for the rap- years, largely because of a higher profile in idly evolving policy debate on trade-environment trade negotiations such as the Uruguay Round and linkages (Beghin and Potier 1997). The present paper the NAFTA. Among academic observers, a consen- gives empirical evidence for Chile, but the methodol- sus has emerged that trade policy is not an adequate ogy can be extended to other countries. Using an ap- tool for environmental protection (Beghin and others plied general equilibrium model, we investigate the 1994), but many other aspects of this linkage remain interactions between trade and environmental poli- contentious issues. Unfortunately, there is still rela- cies, focusing particularly on trade liberalization and tively little empirical evidence to inform this debate, coordinated policies of effluent taxation. We provide and this is the main objective of the present paper. In estimates of emissions for detailed pollution types at particular, we seek to quantify the direct and indirect the national level, identifying patterns of pollution effects of environmental taxes, including their reve- intensity that emerge with greater outward orienta- nue, cost, and output effects, as well as their interac- tion. Although we estimate increased intensities for tion with trade policies and their incidence upon the several pollutants when trade liberalization is under- environment, public health, and elsewhere in the econ- taken without concurrent environmental taxes, none omy. For fast growing developing economies, great- of these appear alarming. er outward-orientation holds great promise in terms A second motivation for the present study is to of growth and efficiency. Pursuing this goal blindly, make more tangible the linkages between economic, however, may jeopardize long-term prosperity be- environmental, and public health indicators, build- cause of the environmental costs of such a strategy. ing upon recent and current work on urban pollution Hence, it is essential to assess the environmental im- and health in Santiago (World Bank 1994; Ostro and pact of trade policy generally and trade liberalization others 1995; O'Ryan 1994). This is an essential step in in particular, and to examine how these might be bet- support of policy formulation that takes more explic- ter coordinated with environmental policies to miti- it account of economy-environment linkages. Past gate environmental degradation. emphasis in this area has been on resource depletion, Our paper makes two contributions. Firstly, we which is appropriate but seriously limited, since it explicitly incorporate links from trade to environment omits more direct and immediate personal costs of to public health indicators, rather than simply mea- environmental degradation. We quantify the incre- 35 36 Trade, Global Policy, and the Environment mental mortality and morbidity associated with com- The infrastructure problem exacerbates air pol- bined economic and environmental polices and their lution in Santiago by contributing to emissions of sus- monetary damages. Because its topology, local cli- pended particulates and other effluents in the air. This mate, and economic concentration make this urban problem combined with unique topological and cli- area comparable to Mexico City and Jakarta, pollu- matic conditions (thermal inversion) put Santiago in tion in Santiago poses a major environmental chal- the league of the most-polluted cities in the world. lenge to Chilean policy makers, now and well into Rising income and health concerns are at odds with the next century. this situation. With the assistance of international or- In this context, we find that abatement of three ganizations, Chile has started addressing these envi- air pollutants (small particulates, SO2, and NO2 (a ronmental problems, especially air and water determinant of ozone)) has the largest impact on mor- pollution in Santiago, and the depletion of forest re- tality and morbidity and far outweighs the health ben- sources (see World Bank 1994). efits which might arise from abatement of other air A critical mass of information has recently been pollutants in Santiago. We also find that Chile's ac- accumulated on urban pollution in Santiago (0' Ryan cession to the NAFTA, compared to unilateral trade 1994; Sanchez 1992; Turner and others 1993; and liberalization, would reduce the emissions of many World Bank 1994); we make use of this information pollutants and have a relatively benign effect on ur- when we link national pollution estimates to pollu- ban public health. Unilateral liberalization, by con- tion concentrations in Santiago. Our study is a useful trast, would appear to induce a significant transfer of contribution to the existing work on Santiago because pollution capacity to Chile from the Rest of the World, it provides estimates of pollution emissions at the na- adversely affecting the environment and public health. tional level and of their variations induced by policy Here the case for coordination with environmental changes. policy is compelling indeed. Until 1975, Chile represented a textbook case of 2. The TEQUILA Model import-substitution, replete with trade distortions, The Trade and Environment EQUILibrium Analysis slow growth, foreign exchange restrictions and result- (TEQUILA) model is a prototype computable gener- ing misallocation of resources. Following a series of al equilibrium model developed at the OECD policy reforms under the structural adjustment of the development Centre for research on sustainable de- 1980s, Chile has become a thriving outward-oriented velopment. The full model is described in details in economy (Papageorgiou and others 1990; World Bank Beghin and others (1996). The TEQUILA model is re- 1999). Growth of output and exports has been spec- cursive dynamic: each period is solved as a static equi- tacular in natural resource-based industries such as librium problem given an allocation of savings and agriculture, fisheries, forestry, and mining sectors in expenditure on current consumption. It is multi-sec- which Chile has traditionally been competitive. These toral (75 sectors for Chile) with careful disaggrega- expansions have fostered rising living standards and tion of natural-resource-based sectors and their concems for the environmental consequences of the forward linkages to manufacturing. Natural resource resource intensity of the growth (World Bank 1994). activities include five agricultural sectors, forestry, In parallel, urbanization is already well ad- fisheries, and five mining/extraction sectors. Their vanced in Chile, where about 85 percent of the popu- linkages to manufacturing are captured by twelve lation live in or within the vicinity of major cities (for agricultural processing sectors, four wood-based sec- example, Santiago Metropolitan Area and Valparai- tors, four oil-based chemical industries, and eight so). The income growth and rapid urbanization have mineral-based ones. outpaced the development of infrastructures such as Output is characterized by CRS technology and paved roads, public transportation equipment and the structure of production consists of a series of nest- sewage treatment systems. Several environmental ed CES functions. Final output is determined from the problems in urban areas are linked to the poor road combination of (non-energy) intermediate inputs and infrastructure and the use of untreated wastewater a composite bundle of energy and value added (la- used in irrigated agriculture (World Bank 1994). bor, and capital (machinery and land)). Non-energy Trade, Environment, and Public Health in Chile 37 intermediate inputs are assumed to be utilized in fixed lutants such as SO2, NO2, CO2. volatile organic com- proportions with respect to total non-energy interme- pounds (VOC), and particulate intensity (PART); and diate demand. The energy-value-added bundle is fur- finally, water pollution measured by biological oxy- ther decomposed into a labor aggregate, and a gen demand (BOD), and total suspended solids (TSS). capital-energy bundle. Labor demand is further de- We calibrate the TEQUILA model using a de- composed into ten occupations. The capital-energy tailed social accounting matrix of Chile for 1992. The bundle is further disaggregated into capital demand model is neoclassical with all markets reaching equi- and demand for an energy aggregate. The energy librium. Trade is modeled assuming goods are bundle is itself decomposed into four base fuel com- differentiated with respect to region of origin and des- ponents. In this production structure, emissions are tination. On the import side, we account for the het- linked to intermediate consumption (inputs) rather erogeneity of imports and domestic goods with the than final output. Figure A3.1 in the appendix shows CES specification attributed to Armington. We assume the nested structure of production. a CET specification for domestic output, in which pro- Most existing CGE models investigating pollu- ducers are assumed to differentiate between the do- tion issues assume fixed proportion between sectoral mestic and export markets. We assume that Chile is a output and emissions associated with that sector. By small country. Trade distortions are expressed as ad contrast, we posit substitution possibilities between valorem tariffs. This assumption is consistent with the value added, energy and non-energy intermediate recent tariffication of most trade distortions in Chile goods, which allow the decrease of pollution associ- following its structural reforms. ated with production if pollution taxes are put in place. This is a major improvement in the incorporation of 3. A Brief Description of the pollution in economywide modeling. Santiago Health Model' We econometrically estimate the pollution efflu- This section briefly describes how we map predicted ents by sector as being function of energy and input pollution emissions from our simulations into health use (Dessus and others 1999). Estimates of these in- effects for residents of Santiago, and then ascribe put-based effluents intensities are obtained by monetary damages to health impacts of pollution. In matching data from a social accounting matrix disag- summary, the model estimates the change in health gsregated at the 4-digit ISIC level to the correspond- status associated with a change in major air pollut- ing IPPS pollution database of The World Bank ants by each of 72 industrial activities in Santiago. (Martin and others 1991). Emissions are generated by Changes in industry emissions used are obtained from both the final consumption and the intermediate use the economywide model. The health effects model of polluting goods. Excise/effluent taxes are used to transforms these emissions data into corresponding achieve pollution abatement. These taxes are mea- changes in health status (such as reduction in PM-10 sured as unit of currency per unit of emissions and related mortality). In so doing, the health effects com- are uniform taxes per unit of effluent for all sectors. ponent is used to estimate the potential health dam- Since every sector has different effluent intensities, the age savings (costs) corresponding to alternative trade pollution tax, expressed per unit ofoutput, varies across and environment policy scenarios analyzed by the sectors. The latter taxes are tacked on to the producer economywide model. price of the polluting commodity. In characterizing emissions, we use baseline in- Pollution by sector is characterized by a vector formation on major air pollutants and emission sourc- of 13 measures of various water, air and soil efflu- es. This step involves collecting data on pollutants ents. Pollution intensity varies by sector and with rel- known to cause significant health problems in Santi- ative prices, since the use of "dirty" inputs is ago, the corresponding emission sources, and influenced by relative price changes induced by poli- baseline average annual emissions and ambient con- cy intervention. The 13 pollution measures include: centration levels. The data are used to estimate the toxic pollutants in water, air and land (TOXAIR, TOX- portion of economywide emissions attributable to WAT, TOXSOL); bio-accumulative toxic metals in air, Santiago, as well as calibrate the health module of the soil, and water (BIOAIR, BIOWAT, BIOSOL); air pol- CGE model to initial conditions. 38 Trade, Global Policy, and the Environment Dispersion modeling maps effluent emissions damages for Chile. Damages due to mortality are into ambient concentration levels, and population- based on the value of a statistical life, which indicates weighted concentration levels are used to determine the aggregate valuation by individuals of reducing exposure rates for health impacts. The next step in- the risk of dying. For Santiago, our estimate is rough- volves calculating the health status response to chang- ly .55 million dollars per life, in 1992 (purchasing pow- es in concentrations of air pollutants. Dose response er parity) US dollars. This estimate corresponds to the functions express the change in incidence of mortali- value of a life reached in 2010 under the reference ty/morbidity induced by changes in pollution con- business-as-usual scenario (Bowland 1997). centrations (Ostro and others 1995). The figures on Because of the scarcity of corresponding morbid- health end-points presented in the results section ity estimates available for industrialized countries, our should be interpreted as increases or decreases in morbidity willingness-to-pay measures are less so- mortality and morbidity with respect to the mortality phisticated. Available estimates from industrialized and morbidity that would have prevailed at a prede- countries were simply scaled down to reflect the per termined safe standard of pollution concentrations. capita income differences between Chile and these We look at various morbidity and mortality industrialized countries, expressed in (PPP) 1992 indicators: US dollars. 1. Premature mortality due to PM-10, SO2, and ozone 2. Premature mortality in males of age 40-59 due to 4. Policy Reform Scenarios lead The time horizon of the simulations is the period 1992- 3. Respiratory hospital admissions (for PM-10, 2010. Every year, savings determine the pool of new ozone) 4. Emergency room visit (for PM-l0) investment resources for the next period and the model 5. Restricted activity days (for PM-10) solves for an equilibrium. This equilibrium determines 6. Lower respiratory illness for children population savigs going to the new investment pool for the sub- of age less than 17 (PM-10) sequent period. Each period, sectoral resource alloca- 7. Asthma symptoms for asthmatic population (for tion adjusts to new prices. Labor moves freely across PM-10, ozone) sectors; existing capital is reallocated across sectors, 8. Respiratory symptoms (for PM-10, ozone) but to a lesser extent due to a partial mobility (vin- 9. Chronic bronchitis in population of age 25 or older tage capital) assumption in the model. The endoge- (for PM-10) nous variables of interest, which adjust at every 10. Minor restricted activity days (for ozone) period, are sectoral inputs, factor use, and output, 11. Respiratory symptoms in children population (for consumption, trade, pollution emissions associated S02) with production and consumption. Aggregate real 12. Chest discomfort in adult population (for SO2) income serves as an approximate gauge of welfare or 13. Respiratory symptoms in adult population (for economic efficiency. We do not attempt to measure NO2) the cost of pollution and characterize "externalities" 14. Eye irritation in adult population (for ozone) only by the level of pollution emissions estimated in 15. Number of headache in adult population (for CO) each scenario. 16. IQ decrement in children population (for lead) We first define a reference trajectory for the econ- 17. Cases of hypertension in adult male population omy based on DRI-McGraw-Hill predictions of GDP (for lead) growth until 2010. Factor and energy productivity 18. Non-fatal heart attacks in male population age 40- changes are endogenously determined such that the 59 (for lead). GDP forecast and the model are consistent with each The last step is to attach a monetary value to the other. All policies are held constant in this reference health impact figures. We follow a willingness-to-pay scenario, called the business-as-usual (BAU) scenar- approach to valuing morbidity and loss of life due to io. For the years 1992 to 2010, the model gives a refer- a change in mortality, relying on the large body of ence trajectory base for output, absorption, trade, and information and data on such measures for industri- pollution emissions, for this BAU scenario. This is the alized economies to econometrically estimate these base or reference trajectory of the economy for our Trade, Environment, and Public Health in Chile 39 analysis. All reported results are expressed in devia- objective is to investigate the implications of coordi- tions (in percent) from this BAU scenario and for 2010, nated trade and environment policies. Analytical re- which is the final year of the simulation exercise. sults (Copeland 1994; and Beghin and others 1997) The first reform scenario imposes taxes on pol- imply that the coordinated piecemeal approach -grad- lutants, one at the time.2 Each tax is such that the ual changes of two instruments to correct for trade emissions of the targeted pollutant progressively de- and environmental distortions- leads to welfare im- crease over time and reach a 25 percent decrease rela- provements. In the context of joint trade and environ- tive to its level in the BAU results by 2010. The phasing mental reforms, efficiency gains are obtained because in of these taxes is set to obtain gradual reductions of trade distortions are reduced and because environ- 10 percent in 1995, 15 percent in 2000, 20 percent in mental degradation can be reduced as well. Recall we 2005, and 25 percent in 2010. The tax rates per unit of want to investigate the effect of such joint reform on effluent are the shadow prices of the quantitative con- sectoral allocation, trade, and pollution abatement. straints on the pollution emissions. Free trade removes border distortions (domestic bor- The second scenario considers a gradual trade der prices are equal to world prices) and the incen- integration, combining unilateral trade liberalization tives to change input mixes to abate pollution in through tariff reductions, with a concurrent but mod- production have been altered, compared to the case est improvement of terms of trade. Terms of trade are of the single environmental reform. The differences parametric for Chile, assumed to be a small country, in the incentive structures lead one to expect contrast- and the terms of trade improvement is introduced as ing results conceming the indirect abatement achieved an exogenous shock. We assume that export prices via complementarity and substitution among emis- increase to simulate this improvement that should sion types, which occurs under the two scenarios. result from the integration of trading countries. This is equivalent to an improvement of the terms of trade. 5. Results from Policy Reform Simulations We decrease the ad-valorem tariffs, progressively to free trade, from their reference levels (1992) as 90 per- Res: foruequence t hh reform sce- cent of original tariffs in 1995, 60 percent in 2000, 30 narios: environmental tax reform, trade integration percent in 2005, and no tariff in 2010. Terms-of-trade (unilateral liberalization, NAFTA, and MERCOSUR), imprvemets ae exresed a an ncrese i ob and then combined trade integration and environmen- served world prices for exports by 2 percent in 1995, tal protection. Results are presented for the final year, 4 percent in 2000, 7 percent in 2005, and 10 percent in 2010, in percent deviations from their BAU values. Table 3.1 summarizes the salient results of the simu- 2010. The terms-of-trade assumption allows us to see how0. The tenvrmsonmtden issuaffecte anlows oustwdori lations in aggregate. Table 3.2 shows the effects of the ented growth strategy. various scenarios on pollution emissions. A longer We consider analogous regional integration and report is available upon request. We first note some liberalization scenarios with NAFTA and MERCOd stylized facts emerging from the Social Accounting SUR countries. Disaggregated data on trade flows al- Matrix on sectors which appear to be pollution hot- low us to consider these alterative trade iberalization spots in Chile. The following sectors exhibit high in- tensities and levels for several effluent types: scenarios. In these two other trade scenarios, we re- move tariffs and increase export prices following a agriculture, sugar refining, mining, chemicals, met- similar progression as in the previous scenario, but als, pottery, electricity, gas, and transportation sec- only with respect to trading partners which are mem- tors. bers of these two regional agreements. Our objective is to impose a sizable trade shock on the Chilean econ- Efluent taxes omy to estimate changes in sectoral composition of Effluent taxes have a small negative impact on growth production and trade. These changes determine the except for the tax on bio-accumulative emissions re- pollution emitted and induced by the outward trade leased in water (BIOWAT), which has a larger impact orientation. (an 8.1 percent decrease in GDP over 18 years with The last group of reform scenarios combines the respect to what it would have been under BAU). The first two types of reforms. For this last scenario, the effects of these taxes on other aggregate measures of Table 3.1 Impact of policy reform on aggregate variables Environmental reform: Aggregate abatement of 25 percent by type of effluent emission Aggregate variables TOXAIR TOXWAT | TOXSOL | BIOAIR | BIOWAT | BIOSOL | SO2 | NO2 I CO2 VOC I PART I BOD I TSS Real GDP -0.7 -0.8 -0.7 -0.3 -8.1 -0.3 -0.2 -0.2 -0.1 -0.4 -0.3 -0.7 0.0 Production 0.4 0.3 0.3 0.3 -8.1 0.4 -2.4 -2.4 -0.8 -3.0 -2.6 0.3 -0.1 Consumption -0.4 -0.5 -0.4 0.0 -1.6 0.0 -1.3 -1.3 -0.2 -1.8 -1.3 -0.4 0.0 Investment -2.1 -2.6 -2.2 -0.8 -23.2 -0.7 -1 .3 -1.3 -0.4 -2.0 - I.5 -2.3 -0.1 Exports -1.6 -1.9 -1.7 -1.0 -10.2 -0.8 -3.1 -3.1 -0.6 -2.1 -3.2 -1.7 0.0 Imports -1.2 -1.4 -1.3 -0.7 -9.6 -0.5 -3.0 -3.0 -0.5 -1.7 -3.1 -1.3 0.0 Labor supply -0.2 -0.3 -0.2 -0.1 -3.1 -0.1 0.2 0.2 0.0 -0.1 0.2 -0.2 0.0 Capital supply -0.9 -1 .1 -0.9 -0.4 -10.4 -0.3 -0.6 -0.6 -0.2 -0.7 -0.7 -1.0 0.0 Real income -0.3 -0.5 -0.4 0.0 -1.3 0.0 -1.2 -1.2 -0.1 -2.5 -1.2 -0.4 0.0 Absorption -0.8 -1.0 -0.8 -0.2 -7.1 -0.2 -1.2 -1.2 -0.2 -1.7 -1.2 -0.8 0.0 Trade policy reforma Combined NAFTA and environmental policy reformb Aggregate variables UNILATERAL | NAFTA |MERCOSUR BIOAIR SO2 T NO2 C CO2 | VOC r PART Real GDP 5.6 1.4 0.6 1.2 1.2 1.2 1.4 0.9 1.1 Production 7.3 1.6 0.6 1.8 -1.1 -1.1 0.7 -1.8 -1.4 Consumption 9.2 2.1 0.9 2.1 0.6 0.6 1.8 -0.1 0.6 Investment 17.7 4.3 1.8 3.5 2.7 2.7 3.7 1.9 2.4 Exports 18.0 3.6 2.7 2.7 0.1 0.1 2.9 1.2 -0.1 Imports 29.1 6.0 3.9 5.3 2.4 2.4 5.3 3.9 2.3 Labor supply 2.0 0.8 0.2 0.7 1.0 1.0 0.8 0.6 1.0 Capital supply 7.2 1.7 0.7 1.4 1.1 1.1 1.5 0.9 0.9 Real income 8.6 2.0 0.8 2.1 0.6 0.6 1.9 -0.9 0.6 Absorption 10.5 2.4 1.0 2.3 1.1 1.1 2.2 0.5 1.0 Trade, Environment, and Public Health in Chile 41 -D N 'IR C~ -. o " 'D economic activity tend to be small as well, with the a2 e( - 0i as i4 0 same exception of the tax on BIOWAT. Trade decreas- es by about 10 percent and investment decreases by O c-m o q n e o . 23 percent. The moderate aggregate output effect of a Cv4 s a -so as othe environmental taxes is a result from aggregation. It dissimulates substantial variations at the disaggre- o. %6 v- 4 '° 4 d 6 oD gated sectoral level and reallocation of resources across sectors. u v o u) r- . o t Lo Next we look at noticeable sectoral output effects, ,C v .6 . vo 2 * -s' 6 v for instance, substantial changes in output occurring in some of the 75 disaggregated sectors included in uw ui '0 ' e C. , r4 0'. E the model. For the first four taxes (all three toxics, BIO- .' _ AIR), fish and seafood output increase significantly E4 0> es e > es a m t > ^ > jy(increases of 60 tol93 percent). For the same effluent @ Z ui ci %6 i = 6 6 c6 0o so a) °dtaxes, mining activities decrease sharply (-17 to -60 -CS >2 -gpercent). The tax on BIOWAT, which induces the larg- 4 O r-. so N-> ri s E E est decrease in aggregate output, has a negative ef- s , = fect on virtually all sectors, and it especially has a strong effect on iron, coal, and basic metals (-30 to -59 3 o + Os _ . uc d as o0 so ^ u w percent). 8 o ui s as r- > 0 * O CD °' 6 W 4) Trade contracts with the effluent taxes. At the 0 e = § disaggregated sectoral levels, trade effects are mixed < .< = (some decreases, some increases) and moderate. Some rN S v cr ,f,os < u) OssO _ 0 3 wexceptions arise: imports and exports of fish increase c., o Lo ->4 ,, 6 o ?L by over 100 percent for the taxes on toxic pollution; E c o" imports of wine and liquors increase by 120 percent .2 d with the tax on VOC. The same VOC tax has a strong .5 U < e _ es o 0' o s o C 0 E negative impact on many pollution-intensive manu- c o . O facturing exports (furniture, chemicals, petroleum o 83 = c)refining, and rubber). c E The simulation results indicate that the impact i : E c6. r r-- 6. e. c O. . j of the taxes on pollution abatement is diverse. Strong L! 4)0 U , complementarities are observed in several subsets of XD 4 o cOthe 13 effluent types, despite the clear possibility of bC o ,< E ffsubstitution among pollution emissions implied by -| 8 i:; vn u- v m. ° . our model since we do not impose any fixedpropor- I.D-a N tions between output and emissions. An increase in 0 Z m the tax on one effluent induces a decrease in another o.o Oa 2 effluent level. All toxics are such a group, so are all ,o g r ' 0. O 0 c , -- . ,, '4 ' E bio-accumulative emissions, and NO2, S2 and PART (PM-10). The larger subset of toxics and bio-accumu- O *O & lative emissions follows such a pattern. More intrigu- 0~~~~~~~~~~~ o 2 m ing is the presence, in the aggregate, of substitution possibilities among effluents. For example, S02 and E ->I>,, NO2 are substitutes for TSS and for bio-accumulative * = *.or wc -c a 8 emissionsin airand soil. -C) n E Xc t t ' - U ID ^ i The tax rates implied by the targeted decrease -a io W c 5 m m ; - ! ° 3 t in emissionsarerealistic: onaveragethepollutiontax " U _ X .J U _1 .A per unit of sectoral output is 4 percent or less for all Table 3.2 Impact of policy reforms on national effluent emissions Aggregate abatement of 25 percent by type_of effluent emission Environmental policy reform F TOXAIR | TOXWAT | TOXSOL | BIOAIR | BIOWAT BIOSOL I SO2 I NO, I CO2 VOC PART BOD TSS Eeffluent emissions TOXAJR -25.0 -27.4 -25.0 -15.7 -11.9 -14.5 0.9 0.7 0.9 -0.1 0.8 -25.2 0.1 TOXWAT -22.7 -25.0 -22.7 -13.8 - 11.5 -12.8 -0.5 -0.6 0.7 -1.2 -0.5 -22.9 0.1 TOXSOIL -25.0 -27.4 -25.0 -15.4 -12.2 -14.3 1.1 1.0 1.0 0.0 1.1 -25.2 0.1 BIOAIR -29.5 -31.5 -29.1 -25.0 -18.7 -19.8 4.4 4.1 -4.5 2.0 0.4 -29.1 -3.0 BIOWAT -2.0 -2.3 -2.0 -1.7 -25.0 -0.7 -0.3 -0.2 -4.1 -1.5 -3.0 -2.1 -2.0 BIOSOL -37.3 -39.9 -36.8 -27.3 -13.7 -25.0 4.2 3.8 1.6 3.7 4.1 -36.9 0.0 SO2 -0.6 -1.4 -0.5 0.4 -5.2 0.2 -25.0 -25.0 -4.1 -10.2 -25.2 -0.5 0.5 NO2 -0.6 -1.4 -0.5 0.4 -5.2 0.2 -25.0 -25.0 -4.1 -10.1 -25.3 -0.6 0.5 CO2 0.4 0.3 0.6 -5.1 -29.8 0.6 -8.0 -7.9 -25.0 -4.7 -23.6 0.6 -11.9 VOC -0.9 -1.5 -0.9 0.1 -4.0 0.1 -3.5 -3.4 -0.6 -25.0 -3.4 -0.9 0.0 PART -0.6 -1.3 -0.5 -0.2 -7.8 0.2 -23.1 -23.1 -6.2 -9.3 -25.0 -0.5 -0.8 BOD -24.7 -27.2 -24.8 -15.2 -12.2 -14.0 1.1 0.9 1.0 -0.1 1.1 -25.0 0.1 TSS 0.6 1.2 0.9 -11.7 -55.5 0.2 9.9 10.0 -47.0 0.8 -21.9 0.9 -25.0 Trade policy reform' Combined NAFTA & environmental policy reformb | UNILATERAL NAFTA | MERCOSUR BIOAIR S 5°2 | NO2 CO2 | VOC | PART Effluent emissions TOXAIR 8.6 - 1.0 3.5 -13.9 -0.3 -0.4 -0.2 - .4 -0.3 TOXWAT 9.5 -0.4 3.3 -11.8 -1.2 -1.3 0.1 -2.1 -1.3 TOXSOL 8.6 -0.8 3.5 -13.5 0.2 0.0 0.1 -1.1 0.1 BIOAIR 8.4 -3.6 8.1 -25.0 1.2 0.9 -8.6 - 1.5 -3.2 BIOWAT 14.8 3.6 1.4 1.9 3.2 3.2 -0.9 1.7 0.2 BIO SOL 4.0 -4.8 4.8 -27.7 -0.4 -0.7 -3.4 -0.8 -0.5 SO2 19.9 3.1 1.6 3.4 -25.0 -25.0 -1.8 -8.4 -25.2 NO2 19.8 3.2 1.6 3.4 -25.0 -25.0 -1.8 -8.3 -25.2 CO2 11.8 2.2 0.3 -2.8 -6.7 -6.7 -25.0 -3.2 -24.0 VOC 13.2 3.6 1.2 3.7 -0.4 -0.4 2.9 -25.0 -0.4 PART 18.9 3.1 1.5 2.8 -23.0 -23.0 -4.2 -7.5 -25.0 BOD 8.8 -0.8 3.5 -13.3 0.1 0.0 0.1 -1.1 0.1 TSS 2.8 1.4 -1.2 -10.0 12.6 12.6 -49.3 2.2 -22.8 Trade, Environment, and Public Health in Chile 43 _ o 13 scenarios. The individual tax rates (per sector and t:2 co 0 c -6 es .- - . by effluent) vary from zero to less than 15 percent for all 13 scenarios, except for the scenario targeting re- a c> 0IR V: 10o o Iq t- vo a, w r-- ID duction in VOC. In the latter scenario the pollution o ,fl, _ uo > _ t a>-_ n tax rate on wine and liquors jumps to 52 percent, and the corresponding tax rate on furniture products is 37 O' v: -. C O° ~ i percent. These high rates are caused by the fact that E = o 0 C, C4 '0 cs these two sectors account for most of the VOC pollu- tion in production. ac ° ° a< ' > Cy Co O 'n 0 The decomposition of abatement into scale (ag- gregate output expansion), composition (composition E2 O q: _ v ^ > o (s S ° CD LoE r E of GDP), and technique (input substitution) effects en C4 a> -0 > .o X = > > > = {f ¢ reveals interesting results.3 First, the composition ef- :5 _ S fect seems overwhelming both in the abatement in o C _ ° ° production and consumption. The effect is more sub- e stantial in production than in consumption, that is, q ° °-t as co CR o. . i , E imports substitute for domestic output in pollution- ~ _4 _0 -2 _ _ _,,, , _ _.° intensive sectors. The technical effect in production is cu _J oEmoderate, and the scale effect is marginal for most e C CD 'O Cr en < n ci C - un. : O pollutants except for the case of the tax on BIOWAT ' (production scale effect of -8.1 percent). Surprisingly, o _ 2 a few simulations exhibit positive scale effects in pro- ae W v tr e. O q t C3 C duction abatement (all toxics, BIOAIR, BIOSOL, and + tE o os D os 0 0 Cs °o C,, BOD). Since the scale effect is the aggregation of sec- c3 n~ c4 ° "toral output effects over all sectors, the latter result c e _ D o may be due to the expansion of activities that are not 'IC 0 o s -; CZ intensive in the pollutants being taxed. This expan- sion, weighted by prices, outweighs the decrease in 4 C _J ...Doutput in polluting sectors. For example, the taxes on E c ` o > co s v r - -o . , all three toxics decrease mining activities as well as 4c I metallic industries, but stimulate fisheries and sea- l Eo _ X tfood, and forestry and wood products. _ m ' This example shows the limitation of tackling 0 ~ ~ ~ ~ ~ 0 o o ooe>s o e . 0= vl. environmental degradation by a single pollution ef- I bo fluent at the time. Abatement of one effluent gives c o _ , .= rise to an increase in resource-intensive activities such C I S( o to o _ o . L , o m 6n , + , 4! as forestry and may induce additional degradation ,E ,_= N 1, Cs - andwelfarelossesifexternalitiesarepresentinthese sectors. This insight reinforces the finding that target- _ g! o ing one specific pollutant can have unintended and _- O~ D damaging consequences on emissions of "substitute" o .'. O pollutants, and calls for an integrated approach to the S eE ~ n design of environmental policies. E In addition, the decomposition of abatement 'l of I_ c ,E,sheds light on the substitutability between effluents. M -o Y F: O U , A variety of patterns emerges. Substitution between n O > ffi 0 < 3 ° U - a two effluent types occurs when all three effects are O 000 oO00 o a.c positive (for example, TSS response to tax on TOX- 44 Trade, Global Policy, and the Environment WAT), or when two or less out of the three effects are bidity and mortality are only applied to Santiago's positive and larger in magnitude than the remaining population. As suggested by the table some taxes such effect(s) (for example CO2 response to BOD tax). as the tax on VOC induce negligible net gains in wel- The impact of the effluent taxes on the concen- fare. tration in Santiago is diverse and to some extent, fol- lows the complementarity/substitution patterns Trade integration observed for emissions. As shown in table 3.2, all three We look at two types of trade integration leading to toxic taxes provide significant decreases in lead (about three scenarios: with the world (unilateral liberaliza- 10 percent), but nothing else, except for a slight in- tion), and regional integration (NAFTA, and MERCO- crease in CO concentration (1.1 percent). The three SUR). Unilateral liberalization induces the largest bio-accumulative pollution taxes decrease lead con- increase in GDP (+5.6 percent), followed by NAFTA centrations as well (by 10 to 20 percent). The tax on (1.4 percent) and MERCOSUR (0.6 percent). These BIOWAT has negative and sometime large effects on gains are small -they represent the relative gains over other concentrations as well -remember it is the tax 18 years. These small changes originate in the out- which has the largest negative scale effects among the ward-orientation Chile has been following; large gains effluent taxes. Air pollution taxes also produce simi- from liberalization have already occurred. Neverthe- lar concentration patterns. Emission taxes on either less these reforms have more significant positive im- NO2, SO2, or PM-10 leads to a substantial decrease in pacts on aggregate trade and aggregate gross the other two (averaging about 19 percent), and some investment. decrease in CO (averaging about 5 percent). Taxes on Moving to disaggregated sectoral output effects, CO and VOC also achieve substantial decreases in the three trade reforms exhibit sharp contrast. The concentration in Santiago. The taxes on water pollu- unilateral trade reform stimulates the output of fruit, tion (BOD, TSS) have marginal impact on most of the forestry, iron, other mining, food processing, wood concentrations. products, paper, and petroleum refining. Converse- As shown in table A3.1, the health endpoints ly, petroleum and gas production, chemicals, glass changes are striking for the taxes on SO2, NO2, and and other manufacturing contract with free trade. PM-10. Premature mortality due to PM-10, SO2 and With NAFTA integration, fruit, agricultural services, ozone decreases by more than 30 percent. With these other mining, food processing, wine and liquor, would three taxes, most endpoints show improvements with expand significantly, whereas copper, iron, and pa- deCreaSeS of morbidity of about 30 percent for seven per would decrease. Hence, NAFTA integration de- oereases mo measures. Thereois a marginal sete- parts significantly from free trade in terms of of the morbidity measures. Tfhere is a marginal dete- ineaoalscaiztn.MROUitgrin international specialization. MERCOSUR integration rioration of morbidity incidence linked to lead (about does not induce any strong effect, except for a major 1 percent). This result is the consequence of the slight increase in transportation material and a decrease in increase in BIOAIR emissions induced by the taxes fish and seafood. On 502 and NO (around 4 percent). onS02 an 2 (an ) The trade effects of these reforms are as follows. Table A3.2 presents the health damages reduc- The unilateral reform induces major increases in vir- tion induced by the environmental taxes. Figures with- tually all sectoral imports and exports, except for im- in parentheses indicate a reduction in damages. The ports of chemicals, glass, and other manufacturing. tax on PM-10 induces a decrease in monetary damag- NAFTA integration has a smaller effect on trade than es equivalent to 0.82 percent of the BAU 2010 GDP; unilateral reform, except for noticeable increases in taxes on SO2 and NO2 reduce damages by an amount imports of agriculture and sugar, and smaller increas- equivalent to 0.65 percent of BAU 2010 GDP. The lat- es for livestock, forestry, fish, mining sectors, sugar, ter taxes induce net gains as approximated by the loss wood products, furniture, paper, and plastic; exports of aggregate income plus the reduction in damages. of fruits, mining (other than copper, coal, and iron), These results show the importance of accounting for dairy, wine and liquor, furniture, and pottery. nonmarket benefits when considering the impact of Finally, the MERCOSUR integration induces in- environmental taxes. The estimated welfare gains are creases in imports of agricultural products, iron, oils, lower-bound estimates because the decreases in mor- sugar, tobacco, petroleum refining, and metals; im- Trade, Environment, and Public Health in Chile 45 ports of fish would decrease. On the export side, sub- marginal increases in the health end-points. Although stantial reductions occur in exports of fish, iron, and NAFTA decreases several types of emissions, these seafood; but food processing, chemicals, plastics, and decreases do not translate into major gains for urban printing expand significantly. health because these improvements are not relevant The pollution implications of these trade reforms for air pollution in Santiago, except for a small im- are next. Unilateral liberalization is pollution inten- provement in lead concentration. The damages asso- sive, for example, NO2, SO2, and PM-10 have an elas- ciated with the health incidences are substantial for ticity of 3.5 with respect to GDP increases induced by the unilateral trade liberalization. As suggested by this unilateral reform. By contrast, MERCOSUR and table A3.2, the damages represent 13 percent of the NAFTA have elasticity values around 2.7 and 2.2 re- aggregate income gains induced by trade liberaliza- spectively, for the same effluents. NAFTA integration tion (damages as percent of gains in GDP). By con- induces decreases in several pollutants (the three tox- trast the damages under the NAFTA scenario are ics, BIOAIR, BIOSOL, and BOD). MERCOSUR induc- moderate due to the small deterioration of the aver- es a decrease in TSS only. The trade diversion of age health status in Santiago. NAFTA integration provides a significant environ- mental benefit in terms of mitigated emissions, rela- Trade integration with environmental protection tive to other two trade liberalization scenarios. This In this last set of reforms, we first combine NAFTA is an overlooked insight on trade diversion in pres- reforms and effluent taxes on a subset of pollutants ence of externalities. The decrease in effluents under (air pollutants). Then, we consider the unilateral trade the NAFTA scenario is achieved through strong com- liberalization coordinated with effluent tax on one position effects in production, outweighing the scale pollutant at the time. The effluent taxes are designed expansion induced by NAFTA. By contrast, the uni- as in the first set of scenarios on environmental re- lateral trade liberalization induces higher intensities forms (incremental and leading to a 25-percent de- in SO2, NO2, and PART (PM-10) via strong technical crease in emissions of the taxed effluent). The tax rates effects towards pollution-intensive input combina- corresponding to these reforms are slightly higher tions. than in the environmental reforms alone. The aver- Still under free trade, we observe marginal in- age tax rates on pollution, expressed in percent of the creases for all toxics, BIOAIR, CO2, VOC, and BOD; producer price per unit of output, do not exceed 5.8 we have marginal decreases for TSS, and BIOSOL. percent. A few individual rates increase sharply. For Finally, we see substantial increases for PM-10, SO2, instance, the tax on TOXWAT emitted by nonmetal- and NO2. These increases are observed after 18 years lic minerals increases to 23.6 percent. As expected, the of expected growth and hence do not represent any- tax rates on VOC for wine and liquor and for furni- thing dramatic. By contrast, NAFTA membership in- ture products increase further to 73 and 53 percent, duces decreases in pollution intensity of GDP or respectively. These increases in tax rates originate in production. This difference between the two trade the output and pollution expansion induced by trade reforms is caused by the cheap energy import occur- liberalization. The pollution expansion requires higher ring under free trade but not under NAFTA. tax rates to be abated back to the level corresponding For the health end-points in Santiago, the uni- to a 25 percent decrease with respect to the BAU lateral trade liberalization scenario has negative con- level. sequences for both mortality and almost all measures The aggregate effect of the combined reforms of morbidity. Premature mortality due to PM-10, (NAFTA cum effluent tax) is small in general, but dif- ozone, and S02 increases by 25 percent and prema- fers according to the pollutant considered. For exam- ture death in males of age 40-59 due to lead increases ple, the effluent tax on CO2 has practically no effect by 9.2 percent, as shown in table A3.1. Morbidity in- on aggregate measures, whereas, the tax on VOC has creases range from 9 percent for cases of hyperten- a negative impact on production, consumption and sion and non-fatal heart attacks to about 26 percent real income. The disaggregated sectoral output vari- for chest discomfort episodes and respiratory symp- ation is more insightful. The iron ore, petro-gas and toms in children. NAFTA and MERCOSUR induce petroleum refining sectors decrease their activity con- 46 Trade, Global Policy, and the Environment siderably for several of the effluent taxes. The VOC Aggregate trade expands less under the coordi- tax drastically reduces the output of wine and liquors, nated reforms than under the simple unilateral trade and of chemicals. Finally, the tax on BIOAIR (lead) liberalization, although some sectoral import induced induces an expansion of fish, seafood and fruit, but a by the latter reform, grow even more under the coor- strong contraction of copper. dinated scenario. For instance, imports of fish are larg- The net trade effects of the combined NAFTA er under the combined scenario than under the free and environmental policy reform are next. Specifical- trade scenario. These exacerbated surges are explained ly, imports of fruits, iron ore, coal, other mining, non- by the almost additive effects of the two policies: free metallic minerals, electricity and transportation trade and environmental protection imply the same increase for most effluent taxes; conversely, imports international specialization. For example, fish imports of petro-gas and petroleum refining decrease. Exports increases significantly with the environmental reforms of fish, iron ore, seafood, food processing, feeds, pa- and with free trade. However, the effect under coor- per, petroleum refining, glass, nonmetallic minerals, dinated policies is lower than the sum of the individ- and transportation decrease. ual one. The pollution abatement figures, including the The inventory of emissions tends to duplicate the multiplier effects of the tax on pollutants that are not patterns reached under the single effluent tax reform directly targeted by the tax, are surprisingly similar since we target the amount of pollution in a similar to the abatement figures for the reforms limited to fashion (-25 percent for each effluent type). Neverthe- environmental reforms alone. The abatement on the less, the substitution between bio-accumulative and targeted emission is of course exactly similar by de- toxic pollutants as a group and the air pollutants (SO2, sign, but the indirect abatement of the other pollut- NO2, VOC, PM-10, and C02) as another group is am- ants does not have to be because relative prices are plified by free trade. This increased substitution is different under the two scenarios. The result is sur- caused by a selective increase in pollution dictated prising because changing border prices affects spe- by the change in relative prices of pollutants when cialization and hence pollution. This result is due to only one type of pollutant is taxed. For instance, the the fact that NAFTA integration has a mitigated im- copper and other-mining sectors decrease their activ- pact on the Chilean environment. ity for the combined scenario targeting toxic and bio- The impact of coordinated reforms -free trade accumulative emissions, but increase their activities cum environmental taxes- appears almost additive on under the four coordinated scenarios targeting SO2, aggregate output, trade and consumption: the aggre- NO2, PM-10 and CO2. VOC emissions increase under gate effect of the coordinated approach is the sum of most scenarios except the one which taxes VOC emis- aggregate effect of the two individual reforms. This sions. is a recurrent result in this type of simulation exer- Finally, as indicated by Table A3.1, the urban cise (Lee and Roland-Holst 1994; Beghin, Roland- health impact of the coordinated reforms reflects these Holst, and van der Mensbrugghe 1995, 1997b). stronger substitutions between broad groups of pol- However, the disaggregated output and trade figures lutants. Mortality due to air pollution increases dra- reveal more interesting, if not surprising, and diverse matically under the combined scenarios involving patterns. For example, iron ore output increases by bio-accumulative and toxic pollution, because the 51 percent with trade liberalization and decreases by emissions of PM-10, SO2, and NO2 are stimulated. 14 percent with the tax on S02. Nonetheless, the com- Similarly, the morbidity induced by SO2, NO2, PM- bined reform (free trade + SO2 tax) induces a margin- 10 and CO increases under the same combined sce- al increase in output of 1.4 percent. This diversity of narios. The VOCs-ozone increases have a negative patterns comes from the difference in relative cost of effect on many morbidity measures: increase in re- abatement by increasing imports (composition) and stricted activity days, in the number of asthma attacks, by changing the input mix (technical effect) under respiratory symptoms, in minor restricted activity different policy regimes. Output of fish, seafood, and days, and in eye irritation cases. As shown in the last wood products increases considerably for several of part of Table A.2, damages reductions under coordi- the free trade cum effluent tax scenarios. nated reforms are less substantial than under the en- Trade, Environment, and Public Health in Chile 47 vironmental tax alone, because of the substitution forc- (efficiency gains from trade, and protected environ- es at work among pollutant types. For example, dam- ment); they are characterized by economic expansion ages caused by lead pollution are substantial in the and decreases in the emissions of the targeted pollut- coordinated scenario involving a tax on PM-10. Nev- ant as well as its polluting "complements". Never- ertheless, the net welfare gains of combined reforms theless, emissions of untaxed substitute pollutants are much higher than for trade liberalization alone. increase considerably. These strong substitutions have For example, the tax on PM-10 combined to free trade a negative impact on urban health, with notable in- induces net welfare gains which are 14 percent high- creases in mortality and morbidity when toxic and bio- er than the net gains under free trade alone. accumulative pollutants are the targets. Further, several natural-resource based sectors expand as well, 6. Conclusions hence increasing the dimensionality of policy coordi- This paper seeks to elucidate linkages between trade, nation (trade policy, effluent taxes, natural resource environment, and public health status in an outward- management). This is a result specific to our investi- oriented economy. From our results, it is apparent that gation of Chile. By contrast, our analysis of trade and such linkages are quite complex, and policy makers environment linkages in Mexico suggests mostly com- relying on intuition alone are unlikely to achieve any- plementarity between effluent types (Beghin and oth- thing close to optimality. Policies in all three areas are ers 1995, 1997b). clearly interdependent, and better coordination could The observed substitutability among pollutant reduce the social and economic costs of economic types and its implications for urban health raises two growth and environmental mitigation. More detailed additional coordination and targeting issues. The first empirical work is needed, however, to support such one is the coordination of environmental programs policies. targeting subgroups of pollutants (for example, tox- Trade liberalization scenarios offer different out- ic, bio-accumulative, air criteria pollutants). Given the comes in terms of growth, international division of substantial substitutability between these groups, an labor and environmental consequences. Integration integrated approach to environmental reform encom- into NAFTA is relatively benign to the environment passing all major groups of pollutants appears and has the smallest pollution elasticity with respect appropriate to avoid unintended environmental to the trade-induced growth (the percentage change degradation or negative health consequence. in pollution with respect to the percentage change in The other issue is the hopeful observation that GDP). Unilateral trade liberalization, with no abate- strong complementarities exist within some groups ment policy, induces higher growth and patterns of of pollutants and that a policy targeting any pollut- specialization more adverse towards the environment, ant within a group would achieve substantial abate- leading to detrimental impacts on public health in ment in most emission types included in the group. Santiago and considerable monetary damages asso- This finding is common to most of our case studies ciated with the negative health impact. MERCOSUR and emerges as an empirical regularity in these link- simulations do not indicate substantial changes in ages. income, pollution or public health, except for in- Another regularity shared by this study and the creased emissions of bio-accumulative pollutants, and other case studies using the same methodology is the small increases in mortality and morbidity linked to relatively low cost of pollution abatement in terms of lead pollution in Santiago. foregone aggregate income. In this specific case of Considering effluent taxes alone, the abatement Chile and Santiago, we establish this result in terms of three pollutants, SO2, NO2, and PM-10 achieves the of welfare. The monetary damages equivalent to the largest decrease in both mortality and morbidity in health impact of air pollution are greatly reduced by Santiago. The health damage reduction exceeds the environmental taxes, especially by the tax on PM-10, foregone aggregate income and corresponds to a net NO2, and SO2, such that the these welfare gains ex- welfare gain to the Chilean economy. ceed the loss of GDP induced by the taxes. A net wel- Coordinated scenarios are well-grounded in eco- fare gain emerges. This statement should be qualified nomic theory and represent the best of both worlds because the resource reallocation implied by the ef- 48 Trade, Global Policy, and the Environment fluent taxes is substantial on a sectoral basis and we vironmental reform encompassing all major groups do abstract from explicit adjustment cost. of pollutants appears appropriate to avoid unintend- The observed substitutability among pollutant ed environmental degradation. types raises two additional coordination and target- The other interesting point is the hopeful obser- ing issues. The first one is the coordination of vation that strong complementarities also exist with- environmental programs targeting subgroups of pol- in some groups of pollutants and that a policy lutants (for example, toxic, bio-accumulative, air cri- targeting any pollutant within a group would achieve teria pollutants). Given the substantial substitutability substantial abatement in most emission types includ- between these groups, an integrated approach to en- ed in that group. Figure A3. I Production nesting Output (XP) Non-energy intermediate demand bundle (ND) Capital-energy-labor bundle (KEL) C0ES (0, 0) ES ( 0) (XAPnf) Labor demand bundle (AL) Capital-energy bundle (KE) IE CES(oK ' Demand by region of origin CES 1) Labor demand by skill type (LI) Energy bundle (Ev) Capital demand (Kv) CES (af, Xe) Energy demandb fel type (XAp7 Demand by region of origin I. Each nest represents a different CES bundle. The first argument in the CES function represents the substitution of elasticity. The elasticity may take the value zero. Because of the putty/ semi-putty specification, the nesting is replicated for each type of capital, that is, old and new. The values of the substitution elasticity will generally differ depending on the capital vintage, with typically lower elasticities for old capital. The second argument in the CES function is an efficiency factor. In the case of the KE bundle, it is only applied on the demand for capital. In the case of the decomposition of labor and energy, it is applied to all components. 2. Intermediate demand, both energy and non-energy, is further decomposed by region of origin according to the Armington specification. However, the Armington function is specified at the border and is not industry specific. 3. The decomposition of the intermediate demand bundle, the labor bundle, and the energy bundle will be specific to the level of aggregation of the model. The diagram represents only schematically the decomposition and is not meant to imply that there are three components in the CES aggregation. Table A3.1 Impact of environmental policy reform on health endpoints for Santiago AGGREGATE ABATEMENT OF 25 PERCENT BY TYPE OF EFFLUENT EMISSION HEALTH ENDPOINTS TOXAIR TOXWAT TOXSQ1 ROAIR RIOWAT PIOS01 SO, N0C' CO7 VOC PART ROn Premature Mortality/Year 0.5 -0.4 0.6 0.9 -10.3 1.1 -30.3 -30.3 -7.7 -13.0 -32.4 0.6 Premature Mortality of males age 40-59Near -12.6 -13.5 -12.2 -15.8 -24.8 -8.9 1.7 1.6 -10.0 -4.8 -4.8 -12.2 RHA/Year 0.3 -0.4 0.4 0.7 -7.7 0.9 -19.5 -19.4 -5.1 -15.2 -20.9 0.4 ERV/Year 0.5 -0.4 0.7 0.5 -11.6 1.1 -29.3 -29.2 -8.7 -12.6 -32.1 0.6 RADNear 0.5 -0.4 0.7 0.5 -11.6 1.1 -29.2 -29.2 -8.7 -12.6 -32.1 0.6 LRINYear (Children < age 17) 0.5 -0.4 0.7 0.5 -11.6 1.1 -29.2 -29.2 -8.7 -12.6 -32.1 0.6 Asthma Attacks/Year (Asthmatics) 0.3 -0.3 0.3 0.7 -5.7 0.8 -14.4 -14.4 -3.3 -16.5 -15.1 0.3 Respiratory Symptoms/Year 0.4 -0.4 0.5 0.6 -9.1 1.0 -23.0 -22.9 -6.4 -14.3 -24.9 0.5 Chronic Bronchitis/Year 0.5 -0.4 0.7 0.5 -11.6 1.1 -29.2 -29.2 -8.7 -12.6 -32.1 0.6 MRAD/Year 0.2 -0.3 0.2 0.8 -4.0 0.7 -10.0 -9.9 -1.6 -17.7 -10.0 0.2 Respiratory Symptoms/Year (Children) 0.5 -0.6 0.6 1.6 -7.4 1.2 -33.2 -33.1 -5.3 -14.2 -33.4 0.6 Chest Discomfort Episodes/Year 0.5 -0.6 0.6 1.6 -7.3 1.2 -33.1 -33.0 -5.3 -14.2 -33.3 0.6 Respiratory SymptomsNear (Adults) 0.4 -0.6 0.6 1.5 -7.1 1.1 -32.6 -32.5 -5.2 -13.8 -32.8 0.5 Eye IrritationsNear (Adults) 0.2 -0.3 0.2 0.8 -4.0 0.7 -10.0 -9.9 -1.6 -17.7 -10.0 0.2 Headaches/Year 1.8 1.7 2.0 -8.0 -52.0 1.5 -8.8 -8.7 -42.9 -6.8 -35.4 2.0 IQ decrements -12.6 -13.5 -12.2 -15.8 -24.8 -8.9 1.7 1.6 -10.0 -4.8 -4.8 -12.2 Cases of Hypertension/i million males age >20 -12.6 -13.5 -12.2 -15.8 -24.8 -8.9 1.7 1.6 -10.0 -4.8 -4.8 -12.2 Non-fatal HeartAttacks/i million males aae 40-59 -12.6 -13.5 -12.2 -15.8 -24.8 -8.9 1.7 1.6 -10.0 -4.8 -4.8 -12.2 Table A3. I Impact of trade reform and combined NAFTA and environmental policy reform on health endpoints for Santiago (continued) TRADE POUCY REFORM* COMBINED NAFrA AND ENVIRONMENTAL POUCY REFORM** HEALTH ENDPOINTS UNI NAFTA MERCOSUR BIOAIR S02 N02 C02 VOC Premature Mortality/Year 24.8 3.2 2.0 3.7 -30.6 -30.6 -5.7 -11.4 Premature Mortality of males age 40-59/Year 9.2 -0.6 8.4 -15.1 1.3 1.2 -11.6 -5.9 RHANear 18.7 3.0 1.5 3.4 -18.8 -18.7 -3.0 -14.2 ERV/Year 24.2 3.2 1.9 3.4 -29.5 -29.4 -6.9 -10.9 RAD/Year 24.2 3.2 1.9 3.4 -29.4 -29.4 -6.8 -10.9 LRINear (Children < age 17) 24.2 3.2 1.9 3.4 -29.4 -29.4 -6.8 -10.9 Asthma AttacksNear (Asthmatics) 15.9 2.9 1.3 3.4 -13.3 -13.3 -1.0 -16.0 Respiratory SymptomsNear 20.7 3.1 1.6 3.4 -22.6 -22.6 -4.4 -13.0 Chronic BronchitisNear 24.2 3.2 1.9 3.4 -29.4 -29.4 -6.8 -10.9 MRAD/Year 13.5 2.7 1.1 3.4 -8.4 -8.4 0.8 -17.5 Respiratory SymptomsNear (Children) 26.4 3.3 2.2 4.5 -33.8 -33.7 -3.1 -12.6 Chest Discomfort EpisodesNear 26.3 3.3 2.2 4.5 -33.6 -33.6 -3.1 -12.6 Respiratory Symptoms/Year (Adults) 25.8 3.3 2.1 4.4 -33.1 -33.1 -3.0 -12.2 Eye IrritationsNear (Adults) 13.4 2.7 1.1 3.4 -8.4 -8.4 0.8 -17.5 HeadachesNear 17.6 3.0 0.2 -5.1 -6.6 -6.5 -43.9 -4.6 10 decrements 9.2 -0.6 8.4 -15.1 1.3 1.2 -11.6 -5.9 Cases of Hypertension/i million males age >20 9.2 -0.6 8.4 -15.1 1.3 1.2 -11.6 -5.9 Non-fatal Heart Attacks/1 million males age 40-59 9.2 -0.6 8.4 -15.1 1.3 1.2 -11.6 -5.9 *Reflects unilateral trade liberalization, NAFTA integration and MERCOSUR integration by 2010 with no explicit environmental policy reforms. **Reflects combined policy reforms of NAFTA integration and aggregate abatement of 25 percent by type of effluent emission. Note: All figures are percentage changes with respect to base trends in 2010. KEY: RHA = respiratory hospital admissions ERV = emergency room visits LRI = lower respiratory illness RAD restricted activity days pphm = parts per hundred million pg/m3 = micrograms per cubic meter MRAD = minor restricted activity days Table A3.1 Impact of coordinated trade and environmental policy reforms on health endpoints for Santiago. Unilateral trade liberalization with aggregate abatement of 25 percent by type of effluent emission (continued) Premature Mortality/I million males age40-59 -5.9 -7.1 -5.2 -14.3 -24.5 0.9 12.4 12.2 -4.7 -1.4 2.5 -5.1 3.5 RHA/Year 18.9 16.8 19.2 20.2 4.9 20.3 -16.5 -16.4 5.8 -7.2 -18.3 19.1 18.2 ERV/Year 24.6 21.9 25.0 25.9 4.0 26.3 -28.5 -28.4 3.4 0.8 -32.3 24.9 22.8 RAD/Year 24.6 21.9 25.0 25.9 4.0 26.2 -28.5 -28.3 3.4 0.8 -32.3 24.9 22.8 LRI/Year (children < age 17) 24.6 21.9 25.0 25.9 4.0 26.2 -28.5 -28.3 3.4 0.8 -32.3 24.9 22.8 Asthma Attacks/Year (Asthmatics) 15.9 14.2 16.2 17.3 5.3 17.2 -10.4 -10.3 7.1 -11.3 -11.1 16.1 15.8 Respiratory Symptoms/Year 20.9 18.6 21.3 22.3 4.6 22.5 -20.8 -20.7 5.0 -4.3 -23.3 21.2 19.9 Chronic Bronchitis/Year 24.6 21.9 25.0 25.9 4.0 26.2 -28.5 -28.3 3.4 0.8 -32.3 24.9 22.8 MRAD/Year 13.3 11.9 13.5 14.7 5.7 14.5 -4.9 -4.8 8.3 -15.0 -4.7 13.4 13.7 Respiratory Symptoms/Year (Children) 26.8 23.7 27.2 29.4 10.8 28.7 -33.3 -33.1 9.3 -0.2 -33.1 27.1 27.4 Chest Discomfort Episodes/Year 26.7 23.6 27.1 29.3 10.8 28.5 -33.2 -33.0 9.3 -0.2 -33.0 27.0 27.3 Respiratory Symptoms/Year (Adults) 26.1 23.1 26.5 28.5 10.7 27.8 -32.8 -32.7 9.0 -0.1 -32.6 26.4 26.7 Eye Irritations/Year (Adults) 13.3 11.9 13.5 14.6 5.7 14.5 -4.9 -4.8 8.2 -15.0 -4.7 13.4 13.7 Headaches/Year 19.0 17.9 19.6 11.0 -49.7 20.4 0.4 0.7 -46.0 4.7 -39.0 19.6 -5.3 IQ decrements -5.9 -7.1 -5.2 -14.3 -24.5 0.9 12.4 12.2 -4.7 -1.4 2.5 -5.1 3.5 Cases of Hypertension/I million males age >20 1 2 1 2 0 12 12 2 1 2 1 Non fatal Heart Attacks/ I million males age 0 1 2 1 2 0 12 12 21 2 1 Trade, Environment, and Public Health in Chile 53 Table A3.2 Impact of environmental policy reform on mortality and morbidity health damages for Santiago (in millions 1992 PPP$) AGGREGATE ABATEMENT OF 25% BY TYPE OF EFFLUENT EMSSON HEALTH DAMAGES TOXAIR TOXWAT TOXSOL BIOAR BIOWAT BtOSOL SO2 NO2 002 VOC PART BOD TSS Mortality (16) (44) (13) (13) (348) 9 (844) (844) (240) (375) (918) (13) (32) Morbidity (422) (467) (406) (530) (1,028) (286) (393) (396) (474) (397) (658) (407) (190) Total (438) (511) (418) (543) (1,376) (276) (1,237) (1,240) (714) (773) (1,576) (419) (222) Total (% Chile BAU (0.23) (0.27) (0.22) (0.28) (0.72) (0.14) (0.65) (0.65) (0.37) (0.40) (0.82) (0.22) (0.12) Note: AU mortality and morbidity figures are valueo at a reference level of GUP[capita income in 2010 under BAU in 1992 PPP$. TRADE POLJCY REFORM' COMBINED NAFTA AND ENVIRONMENTAL POUCY REFORM HEALTH DAMAGES UNI NAFTA MERCOSUR BIOAIR SO2 NO2 CO2 VOC PART Mortality 716 88 76 67 (853) (853) (187) (333) (935) Morbidity 703 34 317 (457) (405) (408) (494) (409) (699) Total 1,419 122 393 (389) (1,258) (1,262) (682) (743) (1,634) Total (% Chile RAU GDP! 0.74 0 06 0.?1 (0.-- (0-6- - (0.66)- (0.36) (0 39'- (tO85) *Reflects unilateral trade liberalization, NAFTA integration and MERCOSUR integration by 2010 with no explicit environmental policy reforms. -Reflects combined policy reforms of NAFTA integration and aggregate abatement of 25% by type of effluent emission. AGGREGATE ABATEMENT OF 26% BY TYPE OF EFFLUENT EMlSSION HEALTH DAMAGES TOXAIR TOXWAT TOXSOL BIOAtR BIOWAT BOSOL SO, NO CO2 VOC PART BOD TSS Mortality 691 607 704 718 109 755 (804) (802) 131 11 (901) 701 683 Morbidity 193 109 224 (71) (764) 452 (3) (6) (90) (79) (397) 225 488 Total 884 716 927 647 (656) 1,207 (806) (808) 41 (68) (1,297) 927 1,171 Tntal (°A .hil PRAIJ )ld fl 7 nAR f;B (fA i Rt f9 (B9) 0n (X On ) A n RA; n OAR nli Notes: Ali mortality and mobidity figures are at a reference level of GOP/captia in 12010 under BAU (measured in 1992 PPP$). Values in parentheses represent welfare gains from policy reform relative to BAU (measured in 1992 PPP$). 54 Trade, Global Policy, and the Environment Notes Bowland, B.J. 1997. "Marginal Benefits of Trade and Environmental Policy: Valuing Air Pollution and Thanks are due Gunnar Eskeland and Raul O'Ryan Health in a Developing Country." Thesis. North for useful discussions. The views expressed in this Calina Seversity, R h. paper are those of the authors and should not be at- Cola B.R. 14Internatyonalanthe tributd to heir ffilited istituions.Copeland, B.R. 1994. 'International and the Environ- tributed to their affiliated institutions. ment: Policy Reform in a Polluted Small Open 1. The health-module for Santiago is formally pre- economy." Journal of Environmental Economics and sented in Bowland (1997), which describes in full Management 20: 44-65. details the different steps and issues involved in Copeland, B.R., and M.S. Taylor. 1994. "North-South the derivation of pollution concentration in Trade and the Environment." Quarterly Journal of Santiago, health impact and monetary damages. Economics 109: 755-87. 2. Taxing all pollutants simultaneously raises diffi- Dessus, S., D. Roland-Holst, and D. van der Mens- culties. First, tracing the effect of any single tax on brugghe. 1994. "Input-based Estimates for Envi- resource allocation becomes impossible. Second ronmental Assessment in Developing Countries." several tax combinations lead to the same decrease OECD Development Centre Technical Papers, in all pollutants, but with different implications Paris. on sectoral allocation, consumption and trade. Grossman, G.M., and A.B. Krueger. 1992. "Environ- mental Impact of a North American Free Trade 3. Our decomposition of pollution follows Copeland Agreement." CEPR Discussion paper 644, London. and Taylor (1994) and Beghin and others (1997), Lee, H., and D. Roland-Holst. 1997. "The Environment which differs from Grossman and Krueger's and Welfare Implications of Trade and Tax Policy." (1992). The major difference resides in the tech- Journal of Development Economics 52(1): 65-82. nique effect. Grossman and Krueger include Low, P., ed. 1992. International Trade and the Environ- technical change in their technique effect; by ment. World Bank Discussion Paper No 159, Wash- contrast we consider movements along an iso-pro- ington D.C. duction surface away from polluting inputs with Martin, P., D. Wheeler, M. Hettige, and R. Stengren. given technology. 1991. "The Industrial Pollution Projection System: Concept, Initial Development, and Critical Assess- References ment." Mimeo, The World Bank. Beghin J., S. Dessus, D. Roland-Holst, and D. van der 0'Ryan, R.E. 1993. "Cost Effective Policies to Improve Mensbrugghe. 1996. "Prototype CGE Model for the Urban Air Quality in Developing Countries: Case Trade and the Envirounment Program. Technical Study for Santiago, Chile." Thesis. University of Specification." OECD Development Centre Techni- California, Berkeley. cal Papers No 116. Ostro, B., and others. 1995. "Air Pollution and Mor- _ 1997. "The Impact of Free Trade and Pollution tality: Results from Santiago, Chile." Policy Re- Taxes on Mexican Agriculture. A General Equilib- search Working Paper (No.n1453). The World Ban, rium Analysis." Agricultural Economics 17: 115-31. Washpe gton D.C. Beghin J., D. Roland-Holst, and D. van der Mens- 1990. Liberalizing Foreign Trade in Developing Coun- brugghe. 1994. "The Trade and Environment tries. The Lessons of Experience. World Bank Publi- Nexus: Global Dimensions." OECD Economic Stud- cations, Washington D.C. ies 23: 167-92. Sanchez, J. M. 1992. "The Costs of Urban Pollution: . 1995. "Trade Liberalization and the Environ- The Case of Santiago." ILADES/Georgetown Uni- ment in the Pacific Basin: Coordinated Approaches versity. to Mexican Trade and Environment Policy." Ameri- Turner, S.H., C.S. Weaver, and M.J. Reale. 1993. "Cost can Journal ofAgricultural Economics 77: 778-785. and Emissions Benefits of Selected Air Pollution . 1997a. "Trade and Pollution Linkages: Piece- Control Measures For Santiago, Chile." Final Re- meal Reform and Optimal Intervention." Canadian port Submitted to The World Bank. Engine, Fuel, Journal of Economics XXX: 442-455. and Emissions Engineering, Inc: Sacramento, CA. Beghin J., and M. Potier. 1997. "Effects of Trade Lib- World Bank. 1994. "Chile: Managing Environmental eralization on the Environment in the Manufac- Problems - Economic Analysis of Selected Issues." turing Sector." The World Economy 20(4): 435-456. Report No. 13061-CH, Washington D.C. Chapter 4 Testing the Impact of Trade Liberalization on the Environment Theory and evidence Judifh M. Dean 1. Introduction es income. Some would argue that openness actually Industrial countries have recently raised concersraises the growth rate of income (Dollar 1992; Harri- over whether or not "dirty industries" migrate. son 1996; and Edwards 1992). If the inverted-U hy- The concern has focussed on a perceived loss of pothesis is correct (Grossman and Krueger 1995; comparative advantage in these industries because of Selden and Song 1994), higher incomes eventually result in lower levels of environmental damage. Since m-ore stringent domestic environmental regulations the static and dynamic effects of trade liberalization compared to developing countries. Developing coun- may work in opposite directions, it is possible that tries, in contrast, are concerned that trade liberaliza- increased openness could lower the growth of emis- tion will promote specialization in dirty industries, sions, and could lower the share of pollution-inten- thus aggravating environmental damage (Dean 1992a, sive goods in output, even for a country with a static b). In a recent study, Copeland and Taylor (1994) show comparative advantage in those goods. that, theoretically, lower income in one country can As a first step in attempting to test the impact of produce a comparative advantage in dirty goods, even trade liberalization on the environment, this paper if externalities are optimally internalized into produc- develops a simultaneous equations model which in- tion costs. corporates both the static and dynamic interrelation- Very little econometric work has been done on ships between trade, environment, and growth. this question (Dean 1992a; Jaffe and others 1995). The Section 2 outlines the theoretical relationship between work by Tobey (1990) and Grossman and Krueger trade and the environment, and reviews three econo- (1993) suggests that trade flows are unaffected by en- metric studies in light of this relationship. Section 3 vironmental abundance (environmental control costs). sets out a simple multi-equation model which incor- Lucas and others (1992) find that the growth rate of porates both the static and dynamic relationships de- toxic intensity of manufacturing output is lower for scribed in section 2. The implications of this for rapidly growing open economies compared to those empirical tests are demonstrated in section 4, using which are closed. Thus openness appears to contrib- the special case of a small country with endogenous ute to cleaner growth. factor supply. In this special case, a two-equation si- But the relationship between trade liberalization multaneous system is required to capture both the and environmental damage is more complex than is direct and indirect effects of trade liberalization on allowed for in these one-equation models. For a coun- growth of emissions. This system is then estimated try with a comparative advantage in dirty goods, the using Chinese provincial data on water pollution for static effect of trade liberalization should worsen en- the period 1987-1995. Section 5 discusses the empiri- vironmental damage. Yet trade liberalization also rais- cal test and presents results. 55 56 Trade, Global Policy, and the Environment Preliminary results show strong support for the "composition effect," therefore, tends to decrease idea that trade liberalization actually has both a di- emissions. The inverted-U hypothesis argues that at rect and an indirect effect on emissions growth, and some income level the scale effect is outweighed by that these effects are opposite in sign. It appears that the composition and technique effects, such that an China has a comparative advantage in pollution-in- increase in income reduces emissions. tensive goods, and thus increased openness directly Since freer trade raises income, it directly con- aggravates environmental damage. At the same time, tributes to increasing pollution levels via the scale ef- increased openness is shown to strongly raise income fect. However, it thereby induces the composition and growth. Income growth itself has a strong negative technique effects of increased income, both of which effect on emissions growth. Thus trade liberalization tend to reduce pollution levels. If a country is on the indirectly mitigates environmental damage. right side of the inverted U, then freer trade will tend to reduce the stock of environment, altering that coun- 2. The Links between Trade and the try's comparative advantage. Thus, even for a coun- Environment-Econometric Evidence try with a relative abundance of environment, freer Theoretically, the impact of trade liberalization on trade may not increase pollution levels. This outcome pollution levels is not clear. Consider the use of the is even more likely if more open economies actually environment (reflected by "emissions") as an input do grow faster, due to access to better technology or into the production process. Using a conventional exposure to global competition. They presumably find Heckscher-Ohlin approach, if a country has a relative themselves moving over the inverted-U faster, rais- abundance of "environment," freer trade leads to in- ing the likelihood that rising incomes will cause pol- creased specialization in "environment-intensive" lutant levels to fall.2 (pollution-intensive) goods. This "composition effect" The few studies which econometrically test the will tend to increase pollution levels. However, fol- relationship between trade liberalization and the en- lowing the Stolper-Samuelson theorem, the price of vironment focus on the composition effect alone, as- environment relative to other inputs will be bid up,1 suming a fixed stock of the environmental factor. causing all industries to switch to less pollution-in- Tobey (1990) and Grossman and Krueger (1993) in- tensive techniques. Thus, the " technique effect" of free vestigate the impact of environmental regulations on trade will lead to a reduction in overall emission lev- trade flows in the context of a Heckscher-Ohlin-Vanek els. The composition effect and the technique effect (HOV) model. The stock of environment is treated as will have the opposite results on pollution levels, if an additional input which, along with labor, capital, the country has a relative scarcity of environment. and natural resources should affect the pattern of The critical point to note here is that the stock of trade. Thus, a relatively large stock of environment environment at any point in time is endogenous, and (compared to one's trading partners) should give a depends upon the income level of the country. Income country a comparative advantage in pollution-inten- growth is thought to have three effects on the exist- sive goods. For such a country, trade liberalization ing amount of pollution emissions. Greater economic should lead to increased specialization in these goods, activity raises demand for all inputs, hence increases hence aggravating environmental damage. emissions (the "scale effect"). However, if people in- Tobey sets up a multi-factor, multi-commodity crease their demand for a clean environment as in- model. Using 1975 data on 23 countries, Tobey pre- come rises (that is, if clean environment is income dicts net exports of 5 different industry groups which elastic), then they will only tolerate higher levels of are classified as pollution-intensive.3 Explanatory pollution if the effluent charge is higher. Since higher variables are the stocks of productive factors, where effluent charges encourage firms to shift toward dean- the stringency of environmental regulations is used er production processes, this "technique effect" tends as a proxy for the stock of environment. The level of to reduce emissions. Finally, if income growth shifts stringency is itself proxied by an index which gives a preferences toward cleaner goods (that is, if clean single ranking to a country. In none of his 5 regres- goods are relatively income elastic), then the share of sions is the environmental stringency variable pollution-intensive goods in output will fall. This significant.4 Testing the Impact of Trade Liberalization on the Environment 57 Grossman and Krueger (1993) explain 1987 U.S. by changing the composition of output towards clean- imports from Mexico (relative to total U.S. shipments) er sectors.9 in 135 industry categories using factor shares as ex- The results from LWH run counter to the popu- planatory variables. In this framework, factor lar notion that developing countries have a compara- shares reflect factor intensity of each industry. tive advantage in pollution-intensive goods (due to Environmental-intensity is proxied by the ratio of lax environmental regulations). There are at least two pollution abatement costs (operating costs) to total potential explanations for this result. One is that oth- value-added in that U.S. industry.5 Grossman and er primary factors, such as unskilled labor, dictate Krueger find that U.S. imports from Mexico are in- comparative advantage. If, as the authors suggest, deed lower in industries which are highly capital- unskilled labor-intensive manufactures are relatively intensive (either physical or human). However, pol- clean, then openness increases specialization in rela- lution abatement costs have no significant impact on tively clean goods. The second possibility is that open- U.S. imports from Mexico. ness increases growth, increasing the likelihood of Given the fact that, in all but a handful of U.S. being on the right side of the inverted U. It is to these industries, pollution abatement costs as a percent of two relationships which we now turn. operating costs are below 1 percent (Low 1992), many authors have argued that these costs are simply too 3. A General Model of Trade and the small in importance to influence trade flows. Gross- Environment'0 man and Krueger's results suggest that with respect A simple simultaneous system is presented in equa- to the United States, Mexico's comparative advantage tions (1)-(4) below. This system attempts to capture lies in unskilled-labor-intensive goods. Thus, we the multiple ways in which trade can affect the envi- would not expect increased openness to cause further ronment (as shown in table 4.1), given that the "stock specialization in dirty goods unless environment and of environent" is actually endogenous. unskilled labor are complements in production. Based on Tobey's estimates, this seems unlikely. He found Y = A(t)h(L, K, E) (1) that a relative abundance of unskilled labor has a pos- E = f (z, Y, Q) (2) itive significant impact on net exports of only one cat- I = g(E, Y) (3) egory of dirty goods-non-ferrous metals.6 U = z(t, Y) (4) Lucas, Wheeler and Hettige (1992) (henceforth LWH) focus directly on the toxic intensity of output Equation (1) shows total output (real income, Y) (rather than on trade flows), where this is measured as a function of the level of trade restrictions (t), the as aggregate emissions per unit of GDP.7 The main stock of conventional factors of production (L, K) and focus of LWH is to test for an inverted-U relationship the level of emissions (E). Increased openness is as- between output growth and the pollution-intensity of sumed to lead to higher total factor productivity output. However, for the period 1960-1988, LWH re- (A'0, andf3>0. Equation (3) is the in- 1980s actually caused it to fall. On the other hand, a verse supply curve for E which is derived from high degree of trade distortion accelerated the growth individuals' utility function. Utility is assumed to be of toxic intensity of output. Thus, it appears that open- increasing in goods, but decreasing in emissions. ness in the trade regime contributes to cleaner growth, Thus, g,>0 and g2>0. 58 Trade, Global Policy, and the Environment We expect the composition of output, Q, to be emissions. That is, the marginal rate of technical sub- negatively related to income (z2<0), as long as clean stitution between K and L is assumed to be indepen- goods are relatively income elastic. However, the com- dent of the level of E. Dirty goods are defined as those position of output will also be related to the restric- which are relatively environment-intensive (pollution- tiveness of the trade regime (t). For a country with a intensive). Thus, production of XD uses a higher ratio comparative advantage in pollution-intensive goods, of environment to conventional factors at any given z <0, since an increase in trade restrictiveness moves factor price ratio than production of X Let F() be an resources toward production of relatively cleaner goods. aggregator of the stock of conventional factors. If we totally differentiate the system in (1)-(4) We assume this country has a relative abundance we can solve for the static effect of trade liberaliza- of environent, and therefore a comparative advan- tion on emissions: tage in dirty goods. The costs of environmental dam- age are internalized via emissions taxes (Xr). Though (dEl dt) A = -hA'(Af g2 + f, + f3z2) - f3z, (5) the country is trading initially, there exists some level where A < 0 is the determinant of the system. The sec- of trade restrictions on imports of Xc. ond term on the right side of (5) is the effect of trade With these simplifying assumptions, we can set out an HOV trade model which captures the relation- liberalization on the demand for environment (as an input) due to a change in the composition of output. ships in (1)-(4), using the approach developed injones If a country's comparative advantage is in pollution- (1965). The unit cost functions for XD and Xc relate intensive goods, z1<°, and thus -f3z1>0. Thus, in- goods prices and factor prices: creased openness would aggravate environmental damage. Counteracting this is the first term which (f- )=(1119I)(PD- (6) captures the fact that trade raises income. The sum in where: are domestic prices of good j, j=D,C; w is the parentheses shows the technique, scale, and compo- P. ' ~~~~~wage paid to conventional factors; ^ indicates pro- sition effects, respectively, of an increase in income prionte chngenina variabl; . isite sre due to trade liberalization. If we are on the "right" pofionpt change in unitbcos fi tpu shand side of the inverted U, this sum is negative. With A'<0, of input i (i=FE) in unit cost of output j, and the first term is negative. Therefore, even if compara- 101 = |0ED - SEC > O- Equation (6) captures changes in tive advantage leads to specialization in pollution- the derived demand for inputs as a function of changes intensive goods, the impact of trade on income could in relative goods prices (assuming fixed stocks of fac- dominate, leading to lower emissions overall. tors of production). Since the country is small, chang- es in the composition of domestic demand do not affect 4. An Alternative Heckscher-Ohlin-Vanek relative outputs nor relative factor demands. Specification Now from (3) we know that the supply of E is To illustrate the importance of accounting for the re- endogenous. Let utility, U =U(CD,CCE) where C. is lationships given by (1)-(4) in empirical testing, con- consumption of good j. Differentiating Uwith respect sider the following special case. Suppose we consider to E one can derive a supply curve for E=E(-c, PD' PC' a small country which produces two types of goods, Y). Let changes in the supply of E be expressed as dirty (XD) and clean (Xc). There is no transborder pol- lution or consumption pollution. Thus all emissions E = (r- ^) )- EY (7) are generated by production. As in Lopez (1994), pro- duction in each sector uses both conventional factors As in Martin and Neary (1980), E, represents a of production (K, L) and E. Rewrite equation (1) as: reduced form supply elasticity with respect to rela- tive factor prices, assuming commodity prices adjust Y = A(t)h[F(L, K), E] (1') to a change in factor prices. If the supply curve does where h(-) is concave in F(-) and in E, and character- not bend backward, E, > 0. The income elasticity of ized by constant returns to scale in K, L, and E. The demand for clean environment, E, > 1. Thus, a rise in production function in (1') assumes weak separabili- income reduces the amount of emissions individuals ty between conventional factors of production and are willing to allow. Testing the Impact of Trade Liberalization on the Environment 59 With constant returns to scale, changes in the (7), and (10) into (8). With some simplification this composition of output can be expressed as yields XD -XC =(/l)(E- F)+a lol(5 -j( ) (8) XD X =S(PD - P) [ Z [(1+FE)F+EylNt)l (12) where: Os is the elasticity of substitution along the where: U=ao+Z- (e,/12A9I) production possibility frontier; A;q is the share of total Z = (I +eyaE) i used in producing j, and |A =|AED - AL > 0. Equation (12) shows that reductions in trade re- strictions will again have two opposing effects on the Let income growth be expressed as relative growth of the pollution-intensive sector. On Y 9+(1Fthe one hand, by increasing the relative price of XD (lEE + A (9) trade liberalization induces more than the usual in- crease in output of XD, since increased demand for where a is the share of input i in total output. As in environment raises the equilibrium stock of environ- Edwards (1992), suppose at any time T, AT = Aer, ment. This is shown by Us, which is the elasticity of where A is a linear function of t, and 6 ' < 0. Then (9) substitution along the variable factor production pos- may be written sibility frontier. If backward bending supply curves at = EE + (2F + ,B(t) (10) are ruled out, 05 > Os. Note that this effect is dimin- y C- +a ,8t 1) ished by a feedback effect due to income growth, cap- tured by 7. On the other hand, trade liberalization Substituting (6) into (7) yields the equilibrium rate of iues income oth hich rede theeqlibri- change in emissions: in~~duces income growth which reduces the equilibri- change in emussions: um stock of the environment. This is seen in the third E(, 1I01)(PD - PC) - (1) term on the right side of (12). Therefore, increased openness could actually lead to a reduction in the share of pollution-intensive goods in output, if the Together, equations (10) and (11) form a simple si- indirect effect of trade liberalization on income multaneous system describing the rates of change in growth, and therefore on the supply of environment, equilibrium emissions and equilibrium real income. eqiiru emsin an .qiirumra noe is stronger than the direct effect on the composition In this system, trade liberalization affects the growth os strgth of emissions in two ways. Recall that changes of output. in changes in the domestic terms of trade (pD-pc)=(p-pc*-i)where * represents world 5. An Application to China prices. Thus, a reduction in trade restrictions will raise Equations (10) and (11) form a system in which in- the relative price of dirty goods (11), leading to in- come growth and the growth of emissions are deter- creased specialization in these goods. This should in- mined simultaneously. The hypotheses to be tested crease the growth of emissions. However, lower levels are as follows. First, trade liberalizahon will affect of trade restrictions will raise factor productivity and emissions growth 'directly,' via its effect on the rela- thereby income (10). This increase in income will re- v duce the growth of emissions since it reduces the will- hve price of pollution-intensive goods. If a country ingness of individuals to supply the environment as has a comparative advantage in those goods, liberal- a factor of production at any given t. Estimating a two- ization will raise the growth of emissions. Second, equation simultaneous model such as this would al- trade liberalization will affect emissions growth "in- low one to sort out these two effects. Since they work directly" via its effect on income growth. If a coun- in opposing directions, a one equation model with try's income is high enough such that it is on the "right trade restrictiveness as an explanatory variable may side" of the inverted U, then liberalization will reduce simply yield an insignificant coefficient. emissions growth. The full impact of trade liberaliza- The implications of trade liberalization for the tion on the environment must account for both these composition of output can be seen by substituting (6), effects. 60 Trade, Global Policy, and the Environment Ideally the system in (10) and (11) should be es- overall trade restrictiveness, specification B removes timated for a country over time. Because of the lack that restriction. Thus, it separates changes in the do- of any lengthy time series data on emissions, we use mestic terms of trade into its two components - chang- pooled Chinese provincial data on industrial water es in the world terms of trade and changes in the black pollution emissions over the period 1987-1995. Data market premium. are obtained from a World Bank dataset compiled by These results appear to validate the hypotheses David Wheeler and Hua Wang. The focus on water outlined above. In Specification A (the first column) pollution emissions (tons of industrial COD discharge) the domestic terms of trade show a strong positive is in part due to data availability, but also due to the relationship with the growth of emissions. An increase fact that Chinese firms do pay levies on water pollu- of 1 percent in the relative price of net exports leads tion. Hence, one can argue that to some extent, envi- to an increase in the growth of emissions of .08 per- ronmental damage from water pollution is indeed cent. This suggests that China may indeed have a static internalized in the costs of the firms. Since emissions comparative advantage in pollution-intensive goods. data are limited to the industrial sector, income is Hence, the direct impact of trade liberalization may measured as the value of industrial output in 1990 indeed be a worsening of the water pollution prob- constant yuan. Data on a broad set of factors of pro- lem. duction are not available at the provincial level. There- At the same time, however, trade liberalization fore, the traditional factors of production included are increases the growth of income (second column). The simply the labor force and physical capital stock. lagged black market premium reflects the overall level No assumption is made as to whether or not of trade restrictiveness at the beginning of period t. A China has a static comparative advantage in "pollution- 1 percent reduction in trade restrictiveness produces intensive" goods. Changes in relative world prices are an increase of .04 percent in the growth rate of income. simply measured as changes in China's net barter Turning to column I again, we see that a .04 percent terms of trade. Assessing trade restrictions at an ag- increase in the growth rate of income causes a decline gregated level is notoriously difficult. For this test, we in the growth of emissions by (.22x.04)=.0088 percent. use the black market premium as a proxy for overall As was argued above, this negative relationship be- trade restrictiveness. We will attempt some sensitivi- tween income growth and emissions growth would ty tests using alternative measures of overall trade reflect the "technique effect." As income rises, people restrictiveness based on actual Chinese tariff and non- increase their demand for a clean environment, hence tariff barrier data. Although China has some restric- imposing higher penalties on water pollution. The tions on trade which differ across provinces, we indirect role of trade liberalization, via its effect on assume that their effect is mitigated by free inter-pro- income growth, is to reduce the water pollution prob- vincial trade. lem. The model in (10) and (11) is estimated using The results for specification B (third and fourth two-stage least squares. Since the average rates of columns) reinforce those in specification A. It appears growth of emissions and income across provinces are that the full effect of changes in the domestic terms of likely to differ based on variation in the types of in- trade on emissions growth can be attributed to chang- dustries concentrated in a province, fixed effects were es in openness. This is not surprising, since the world included. Estimation of the model also required cor- terms of trade for Chinese traded goods changed very rection for both groupwise heteroskedasticity and first little during this period, while the black market pre- order autocorrelation. Table 4.1 reports the results of mium varied widely over time. The effect of open- two specifications of the model. ness on income growth remains the same, while the Specification A uses one variable to capture impact of income growth on emissions growth is changes in the domestic terms of trade (as in equa- slightly more strongly negative. tion (11)). Thus it restricts the coefficient on changes There are certainly a few anomalies in the results in the world terms of trade and changes in the black in table 4.1, which suggest that further work needs to market premium to be the same. Given that the black be done. First, inboth specifications labor force growth market premium is an imperfect proxy for changes in has no significant influence on income growth. This Testing the Impact of Trade Liberalization on the Environment 61 Table 4.1 The impact of openness on growth of emissions Model: (A) (B) Emissions Income Emissions Income Dependent variable: growth ab growth a.b growth ab Growth a,b Emissions growth --- -0.02* -0.01 (-2.08) (-1.56) Income growth -0.22* --- -0.29* --- (-2.03) (-2.21) Domesfic Terms of Trade 0.08** ( percent change) (4.90) World terms of trade --- --- 0.63 --- ( percent change) (0.34) Labor force growth --- 0.59 --- 0.62 (1.13) (1.19) Investment,., --- 0.44** --- 0.42** (3.32) (3.20) Black market premium,., --- -0.04** --- -0.04** (level) (-4.69) (-4.61) Black market premium --- --- -0.08** ( percent change) (-3.64) N= 168f Wald test d__ chi-square 160.43** 34.4 1 1 79.09** 31.73** Notes: t-statistics in parentheses Black market premia are from World Currency Yearbook Net barter terms of trade are from World Bank World Development Indicators. 'Signifcant at the I percent level Significant at the 5 percent level 'AJI variables are measured in annual percent change except black market premium , b Includes fixed effects for provinces. Estimates of the 28 constant terms are not reported. Standard errors corrected forgroupwise heteroskedasticity and first order autocorrelation. ' Data includes 28 provinces over the period 1987-1995. Due to missing data, 1991 values reflect charges between 1989 and 1991. Hainan and Tibet were excluded due to lack of data. d Tests for the significance of the regression as a whole. may be due to the use of provincial employment data industries across provinces. Future tests will attempt to measure the labor force as opposed to provincial to test for the influence of both state ownership and industrial employment data. It may also reflect the pollution intensity of industries. extent to which state owned entreprises dominate in- dustrial production. Second, the growth of emissions 6. Conclusion itself appears to have either no influence or a nega- Recent events such as NAFTA have brought out con- tive influence on income growth. This might be due cems on the part of both industrial and developing the differing concentrations of pollution-intensive countries as to the effects of trade liberalization on 62 Trade, Global Policy, and the Environment the environment. For the latter group, concern has data. Many thanks to Ulrich Kohli and James Riedel, focussed on the idea that less stringent environmen- Kym Anderson, Randy Stringer, Rod Ludema, and tal standards will imply a comparative advantage in Mitch Kaneda for helpful suggestions on the earlier pollution-intensive goods. If so, trade liberalization study. Thanks are also due to the participants in: The will harm the environment. Conference on Trade, Environment, and Growth (Au- Existing empirical work on this question consists gust 1996) in Melbourne, Australia; the CIES seminar of single equation models which focus on the static at the University of Adelaide, Australia; the IEFS ses- relationship between trade and environment. That sion at the ASSA Meetings, January 1997. work found: no relationship between the pattern of 1. This is true to the extent that environmental exter- trade and relative abundance in environment; a neg- nalities are internalized into production costs. ative relationship between growth of toxic intensity 2. As Birdsall and Wheeler (1992) note, if foreign tech- of output and openness. This paper develops an al- nology is cleaner, or exports must be cleaner to ternative simultaneous equations model which allows meet higher foreign standards, then more open for both static effects via comparative advantage, and economies would see cleaner growth. dynamic effects of trade on income growth and there- 3. 'Pollution-intensive industries" are those with pol- fore on the growth of emissions. In this way, the liter- lution abatement and control costs per unit of ature on trade and growth as well as that on income output greater than or equal to 1.85 percent. These growth and environmental damage (the inverted-U 24 SITC industries are then aggregated into 5 in- hypothesis) are incorporated into the model. This dustry groups: mining, paper, chemicals, steel, highlights the fact that trade liberalization indirectly non-ferrous metals. affects relative factor abundance, since the "stock of environment" is endogenous. 4. There are a number of shortcomings in this ap- WVhat emerges is a two equation model which proach. In addition to problems with measuring stringency, using such an index assumes that the simultaneously determines growth of income and regulationsin plc are allymoptial the growth of emissions. Estimation of this model using are not, there is no necessary correlation between Chinese provincial data shows the importance of us- environmental endowment and the index. Aggre- ing a simultaneous model to discern the influence of gation may also be a problem. A test run on one trade liberalization. Results show that there is indeed country across many industries might allow for both a direct and indirect effect of trade liberalization environmental stringency to affect the pattern of on emissions growth, and that these effects are of op- specialization across industries. Finally, import posite sign. Improvements in the domestic terms of restrictions on some of these goods might allow trade lead to increased emissions growth. Hence, the domestic industries to absorb the cost of environ- direct impact of trade liberalization would be to ag- mental regulations without impacting trade flows. gravate environmental damage. However, results also 5. To account for the effects of trade barriers on trade indicate that increased openness significantly raises flows the collected tariff rate for each industry is the growth of income, and that growth of income has also included. As is well known, however, in- a negative and significant effect on emissions growth. creases (decreases) in this rate do not necessarily Thus, the indirect effect of trade liberalization is to correspond to increased (decreased) restrictiveness mitigate environmental damage. in the trade regime. 6. Semi-skilled labor is either unrelated to net exports Notes or significantly negatively related to net exports This paper is a revision and expansion of an earlier for the five industry categories he examined. study, 'Testing the Impadct of Trade Liberalization on 7. They construct a measure of emissions of 320 types the Environment," University of Adelaide Centre for of pollutants from industry i per unit of output of International Economic Studies Seminar Paper 96-10. i. This is then multiplied by the share of industry i I am most grateful to Bill Lahneman for excellent re- in GDP (or alternatively manufacturing output), search assistance, and to Hua Wang, Muthukumara to generate an aggregate measure of emissions per Mani and David Wheeler for their help in obtaining unit of GDP. Testing the Impact of Trade Liberalization on the Environment 63 8. The trade restrictiveness variable is constructed by Harrison, A. 1996. "Openness and Growth: a Time- grouping countries into seven categories, from Series, Cross-Country Analysis for Developing least restrictive to most restrictive, based on Countries." Journal of Development Economics 48: Dollar's (1992) index of trade distortion. 419-447. 9. More recently Lucas (1994) finds considerable evi- Jaffe, A.B., S.R. Peterson, P.R. Portney, and R. Stavins. dence that increased openness reduces the level 1995. "EnvironmentalRegulation and the Com- of toxic intensity of output, when toxic intensity is petitiveness of U.S. Manufacturing." Journal of Eco- measured with respect to individual types of pol- nomic Literature 33(1): 132-163. lutants. Jones, R. 1965. "The Structure of Simnple General Equi- librium Models." Journal of Political Economy 73: 10. The non-technical reader may want to skip sec- 557-572. tions 3 and 4. Lopez, R. 1994. "The Environment as a Factor of Pro- duction: the Effects of Economic Growth and Trade References Liberalization." Journal of Environmental Econom- Copeland, B.R., and M.S.Taylor. 1994. "North-South ics and Management 27: 163-184. Trade and the Environment." Quarterly Journal of Low, P. 1992. "Trade Measures and Environmental Economics August: 755-87. Quality: The Implications for Mexico's Exports." Dean, J. 1992a. "Trade and the Environment: a Sur- In P. Low, ed., International Trade and the Environ- vey of the Literature." In P. Low, ed., International ment. World Bank Discussion Paper No. 159. Trade and theEnvironment. World Bank Discussion Lucas, R.E.B. 1994. "Intemational Environmental In- Paper No. 159. dicators: Trade, Income, and Endowments." Bos- Dean, J. 1992b. "Trade Policy and the Environment: ton University IED Discussion Paper Series No. Developing Country Concerns." Prepared for the 46. Environment Directorate, OECD. Lucas, R.E.B., and others. 1992. "Economic Develop- Dean, J. 1996. "Testing the Impact of Trade Liberal- ment, Environmental Regulation and the Intema- izationon the Environment."Centrefor International tional Migration of Toxic Industrial Pollution: Economic Studies Seminar Paper 96-10, University 1960-88." In P. Low, ed., International Trade and the of Adelaide. Environment. World Bank Discussion Paper No. Dollar, D. 1992. "Outward-Oriented Developing 159. Economies Really Do Grow More Rapidly: Evi- Martin, J., and P. Neary. 1980. "Variable Labour Sup- dence from 95 LDCs, 1976-1985." Economic Devel- ply and the Pure Theory of Intemational Trade." opment and Cultural Change :523-544. Journal of International Economics 10: 549-559. Edwards, S.1992. "Trade Orientation, Distortions, and Selden, T.M., and D. Song. 1994. "Environmental Growth in Developing Countries." Journal of De- Quality and Development: Is There a Kuznets velopment Economics 39: 31-57. Curve for Air Pollution Emissions?" Journal of En- Grossman, G., and A. Krueger. 1993. "Environmental vironmental Economics and Management 27:147-162. Impacts of NAFTA." In P. Garber, ed., The US- Tobey, J. 1990. "The Effects of Domestic Environmen- Mexico Free Trade Agreement. Cambridge, MA: MIT tal Policies on Pattems of World Trade: An Em- Press. pirical Test." Kyklos 43 191-209. .1995. "Economic Growth and the Environment." Quarterly Journal of Economics May: 353-377. Chapter 5 Industrial Pollution in Economic Development Kuznets revisited Hemamala Hettige, Muthukumara Mani, and David Wheeler 1. Introduction not explicitly captured by the theoretical models, A number of recent studies have explored the structural change in the economy and more effective relationship between economic develop- regulation are also potentially-important sources of X &Iment and environmental quality. Theoreti- change in pollution. cal papers by Gruver (1976), John and Pecchenino The empirical results are roughly consistent (1992), and Selden and Song (1995) have derived tran- with a Kuznets curve for conventional air pollutants sition paths for pollution, abatement effort and de- such as suspended particulates and sulphur dioxide, velopment under alternative assumptions about social but the results for water pollution are mixed. In most welfare functions, pollution damage, the cost of abate- cases, however, the implied trajectories are sensitive ment, and the productivity of capital. Empirical stud- to inclusion of higher-order polynomial terms in ies (Hettige and others 1992; Shafik 1994; Selden and income whose significance varies widely. Structural Song 1994; and Grossman and Krueger 1995) have interpretation of the estimates remains ad hoc, since searched for systematic relationships by regressing the existing studies have incorporated almost no cross-country measures of ambient air and water qual- evidence about actual emissions in developing ity on various polynomial specifications of income per countries. capita. This extensive body of work has been moti- This paper attempts to advance the state of the vated by several related questions: Does pollution fol- art, using new data on industrial water emissions in low a "Kuznets" curve, first rising and then falling as developed and developing countries. Our analysis income increases? At what income level does the turn- decomposes total industrial pollution into four around occur? Do all pollutants follow the same tra- proximate determinants: National output; the share jectory? Is pollution reduction in developed economies of industry in national output; the share of polluting due primarily to structural change, or to regulation? sectors in industrial output; and end-of-pipe pollution The theoretical work has shown that a Kuznets, intensities in the polluting sectors. As most of the or inverted-U, relationship can result if a few plausible previously-cited work has noted (without being able conditions are satisfied as income increases: Constant to resolve the issue), declining pollution at higher or falling marginal utility of consumption; rising levels of development must be driven by some marginal disutility of pollution; constant or rising combination of income-related changes in the latter marginal pollution damage; and rising marginal three factors. abatement cost. Of course, actual turnaround points We investigate these changes in three eco- depend on the relative magnitudes of the underlying nometric exercises. Using international panel data, parameters, as well as their signs. Although they are we estimate the effects of economic development on 65 66 Trade, Global Policy, and the Environment industry's share of total output and the industry share The remainder of the paper is organized as of polluting sectors. To study development-related follows. Section 2 develops the models which link our changes in end-of-pipe pollution intensity, we have three pollution factors to economic development. collected factory-level data on industrial water Section 3 introduces the data used for estimation. pollution from national and regional environmental Section 4 discusses the results and their implications, protection agencies (EPAs) in twelve countries: Brazil, while Section 5 provides illustrative estimates of China, Finland, India, Indonesia, Korea, Mexico, recent water pollution trends in a number of Netherlands, Philippines, Sri Lanka, Taiwan (China), developed and developing countries. Section 6 Thailand and the United States. Controlling for concludes the paper. sectoral differences, we use these data to investigate the effects of income per capita, regulatory strictness 2. Development and Industrial Pollution and relative input prices on factory-level pollution The first stages of economic development typically intensity (pollution/output). In a complementary witness the rapid growth of industrial activity and exercise, we add a measure of regulatory strictness to declining environmental quality in densely-populat- a cross-country labor intensity equation to test for the ed urban areas. When new industries are pollution- impact of regulation on the demand for labor. For our intensive, their emissions can increase local ambient international pollution accounting exercise, these pollutant concentrations to harmful levels. To study results provide two inputs: A measure of average this phenomenon, we decompose total industrial water pollution intensity for each industry sector (an emissions in a particular region as follows: input to our study of income-related changes in polluting sectors), and an estimate of the change in P = m(y)Qp(y)17(y), (1) sectoral pollution intensities as income per capita where increases. We combine our econometric results to simulate P Total industrial pollution the total effect of economic development on industrial m = Manufacturing share of total output water pollution. In this case, we do not find an overall Q = Total output inverse U-shaped relationship. The three factors have p = Manufacturing pollution intensity very different relationships with income, and their i = Degree of pollution abatement: 0 < 1 51 joint product with total output is asymptotic, not y = Income per capita parabolic. Industrial water emissions rise until carabourie atdustainrmiddle-inceat uss,and tseunt Equation (1) includes three parameters which we hy- remain approximately constant as they grow richer. pothesize to be functions of economic development: While our results do not support the Kuznets The manufacturing share of total output (m), the pol- hypothiles for indultsti water spollution, they ds lution intensity of manufacturing (p), and the degree revotheal s a oriingdeulariawaty r inpcrosioun,try dn- of pollution abatement by industry (Ti). In this decom- ronmental performance. Our plant-level results position, the effect of economic development on pol- suggestnthat perollutioand. laOr lant-lensiesuith lution depends on the signs and the magnitudes of suggest that pollution and labor intensities with teprmtr oenn h eain ewe ,p respect to output decline continuously, and at almost the parameters governing the relations between m, p, exactly the same rate, as income increases. Thus, TI and y. sectoral pollution/labor ratios remain approximately 2.1 Manufacturing share of total output constant during the development process. This finding provides useful leverage for the analysis of pollution Numerous studies of the relationship between trends across countries and over time. As an industrialization and economic development have illustration, we combine our estimated sectoral suggested an inverted-U relationship between the pollution/labor ratios with panel data on sectoral manufacturing share of output (m) and income per employment to simulate international trends in capita (y).2 During the first phase of economic industrial water pollution during the past two growth, m increases as industry expands more decades. rapidly than agriculture. As the economy begins to Industrial Pollution in Economic Development 67 mature, rapid growth in services becomes the pi,/1+AlogY+ l(iogy,12, dominant factor and m declines. Over the existing (3) range of national incomes per capita, this rela- =s tionship can be approximated with a parabolic function: where j, k, t = country, sector and year respectively, and /j = average pollution intensity for sector j (see logim = azO + or; log y + a2(log y)2 (a, > 0, a2 < 0) (2) Section 4.3). Our empirical analysis uses cross-country evidence 2.3 Pollution abatement for the past two decades to estimate this relationship The marginal cost of abating pollution from industri- and test its intertemporal stability. We focus al sources is a function of the scale of activity, pollut- particularly on changes in 8m/8y as development ant concentration in process influent,4 the degree of proceeds. Controlling for growth in total output, abatement, and local input prices.5 In static partial large movements in m will have a significant impact equilibrium, cost-minimizing firms with flexible on the trajectory followed by industrial pollution. abatement choices will control pollution to the point where their marginal abatement costs equal the 'price' 2.2 Sector-weighted pollution intensity exacted for pollution by affected parties.6 Character- The sectoral composition of industrial activity has an istic production scale and process effluent intensity important effect on its average pollution intensity, or differ significantly by sector, and abatement costs dif- important effect on its average pollution intensity, or fer by location. Differences in the groups affected by pollution per unit of output. Industrial processes dif- pollocanials D if int spatia ariation pollution can also lead to significant spatial variation fer greatly in their production of waste residuals in emissions prices.7 which, in turn, have varying potential for creating environmental damage. Abatement costs also differ POLLUrTON, EMPLOYmENT, REGULATION, AND significantly by industry sector (Dasgupta and others INPUT PRICES 1996; Hartman and others 1997). Even in well-regu- Where the environment is 'cheaper' or abatement is lated economies, these factors cause significant inter- more expensive, the pollution intensity of production sectoral differences in pollution intensity. For in a particular sector should be higher, ceteris pari- example, metals and cement are generally intensive bus. However, data scarcity has made it difficult to in harmful air pollutants; food and paper production test the magnitude of these effects, as well as the im- are disproportionate emitters of organic water pollut- pact of spatial variation in the prices of capital, labor, ants (Hettige and others 1995). energy and materials. At present, we have sufficient Anecdotal evidence suggests that the sectoral data to investigate these relationships in a two- composition of industry follows a 'clean' trend as equation demand system:8 development proceeds. This could reflect domination of early industrialization by InP= R primary industries, which J 0o+XspDspRj +aK Wi-W +aW +KjEW +MWMj j generate heavy pollution (4) loads as they convert bulk raw inJ =O+X;SLDSL +1 materials into primary inputs (for example metals, paper, cement, sugar). During where (for country j) P =Plant-level pollution the development process, primary indus- tries may lose L = Plant-level employment output share to cleaner industries (for example vehicle D = Vector of dummy variables for S sectors and electronics assembly, instruments).3 In this paper, R = An index of regulatory strictness we test the clean trend hypothesis for industrial water WKL,E,M = Prices of capital, labor, energy and pollution by fitting the following equation to an materials international panel dataset: Q = Plant-level output 68 Trade, Global Policy, and the Environment Our data set, described in the following section, latter effect. However, our prior expectation is that combines information from several sources: plant- and the materials-reducing effect should dominate: A sector-level data on emissions and employment from wage increase should reduce water pollution in- national and regional EPA's; sector-level information tensity. on output and employment from national census Energy Price: If labor and energy are gross substitutes bureaus and the World Bank's international database in production, then an increase in the price of en- (BESD); and data from BESD on national income, ergy should increase the labor intensity of produc- population, and a number of other variables. For cross- tion. In the case of pollution, an energy price country consistency, we use summary data by sector.9 increase will reduce energy use for both process- Plant-level relations between scale (Q) and pollution ing and pollution abatement. Abatement activity intensity (P/Q) are not relevant for sectoral ag- should therefore fall, and water pollution inten- gregates, so we impose the assumption of constant sity should rise. returns (aQ Q = = 1) 0 This implies estimation of Capital Price: A capital price increase should also in- the pollution and labor equations in intensity form, crease labor intensity. For pollution, an increase with P/Q and L/Q as the dependent variables. in the interest rate or the price of equipment should Using the BESD database, we estimate sectoral reduce capital and energy use as well as pollution average L/Q ratios for each sample country. We abatement, while increasing the use of labor and construct sectoral average P/L ratios (sectoral materials in processing. Both reduced abatement pollution intensities w.r.t. labor) from the data and increased materials use should lead to more provided by national and regional EPA's. We estimate water pollution. P/Q (pollutionintensity w.r.t. output) by multiplying L /Q and P/L for each sector and country. The results POLLUTION INTENsnlYrN ECONOMIC DEVELOPMENT tQandPLfor estiatsecth f ng equationTheres: lt Regulatory strictness and some input prices (for permit us example, wages) change systematically as per capita p. income increases. To assess the overall impact of in=S. - +YSPDSp wR hRj+aK nW +aL WL +aE Iw economic development, we also estimate our intensity Q' s (5) equations in reduced form: L. inQJ=0+ + SLDSL +8p3InR +JKInWK .+LLhiW L+fEWE. +v In (7J= PJ= rO+Y5 aSDp +PY n y +e£ 6 In some cases, we have clear prior expectations about L . parameter signs: In X = + I YSLDSL +Y In y + v. Regulation: Ceteris paribus, we expect stricter regu- lation to have a negative impact on pollution in- tensity. We have no clear prior about its impact These equations have two specific roles to play in our on labor intensity at the sector level. analysis. First, they provide an estimate of pY, the Labor Price: We naturally expect increasing wages to elasticity of end-of-pipe pollution intensity with reduce the labor intensity of industrial output. The respect to income per capita. We use this to construct effect of wages on pollution intensity is less trans- the index T1 in equation (1). The results also provide parent. Econometric estimates of KLEM (capital, estimates of average sectoral pollution intensities (8,p) labor, energy, materials) models has suggested across countries. We combine these intensities with that (K,E) and (L,M) are complements in produc- our panel data on sector shares by country to construct tion, while the pairs KE and LM are gross substi- estimates of p for use in equation (3). tutes."1 If these relations hold, a wage increase should have the following effects on emissions: 3. Data (1) Materials use and the volume of polluting re- 3.1 Industrial pollution siduals should decline; (2) Labor use should de- crease in both processing and pollution abatement To our knowledge, this is the first comparative inter- activities, with some increase in pollution from the national study of industrial pollution which uses di- Industrial Pollution in Economic Development 69 rect observations on emissions. We have obtained the Hartman and others 1997; Hettige and others 1996; data from environmental protection agencies in Bra- and Wang and Wheeler 1996). Dasgupta and others zil, China, Finland, India, Indonesia, Korea, Mexico, (1995) have advanced the state of the art by developing Netherlands, Philippines, Sri Lanka, Taiwan (China), quantitative indices of regulatory development from Thailand and the United States. Descriptions of the reports filed for the U.N. Conference on Environment data sources are provided in the Appendix. and Development (UNCED -Rio de Janeiro, 1992). We use the pollution information and comple- Their results suggest that international differences in mentary employment data to estimate emissions pollution regulation are well-explained by a model intensities by industry sector in kilograms per day per which incorporates the effects of per capita income, employee. We focus on organic water pollution urbanization, population density, and manufacturing because it provides the most plentiful and reliable share in national output. We have adopted the source of comparable cross-country emissions Dasgupta model to produce a cross-country pollution information. Water pollution data are the most regulation index for this paper. Six of our thirteen plentiful because developing countries have tra- country cases have actually been scored by the ditionally begun industrial pollution control programs Dasgupta exercise. For the remaining seven cases, we with regulation of organic water emissions. They are have calculated the pollution regulation index values relatively reliable because sampling techniques using the Dasgupta equation. for measuring water pollution are more widely understood and much less expensive than those for 3.3 Input prices air pollution. We have computed wages (in $US 1990 per worker) 3.2 Environmental regulation by ISIC sector from UNIDO's reported sectoral totals 3.2 Environmental regulation for employment and payrolls. Our electricity tariff Some comparable measure of regulatory strictness is rates for the OECD and developing countries have necesaryforestiatin ofour ros-coutry been drawn from International Energy Agency data equations. However, credible indices of envi- ronmental regulation are difficult to find. Even in the and the World Bank's Power Sheets database, United States, comparative analyses of state-level respectively. The World Bank's World Development regulatory 'outputs' have generally used input-based Indicators database has provided our national real measures such as expenditures on monitoring and interest rate measures. enforcement, or total employment of inspectors.12 Such measures may have at least some justification 3.4 Employment, income, and output for within-country analyses, since quality- and price- Estimation of equations (2), (3), (5) and (6) requires adjustment problems are not too serious. For cross-country data on total output, industrial output, international comparisons, however, they would be employment, income, population and a number of problematic even if comparable data were available. other variables. We have obtained the relevant panel Most developing countries do not have such data, so data from the World Bank's international database input-based comparisons are notpossible in any case. (BESD). We have used Summers-Heston estimates as A more promising approach has been taken by our measure of income per capita. recent econometric work on the sources of variation in regulatory strictness. This work is helping to 4. Econometric Results identify robust proxies which can be used as 4.1 Manufacturing share in national output instruments in cross-country comparisons. The best Table 5.1 reports panel estimates for equation (2). We instrument is undoubtedly per capita income, which provide comparable results for OLS, fixed- has been shown to affect both formal and informal effects (without time dummies) and random effects regulatory pressure on polluters in the United States models. We prefer the random effects model, but the and Asia (McConnell 1992; Pargal and Wheeler 1996; choice of estimator does not have a major effect on 70 Trade, Global Policy, and the Environment Table 5.1 Log (manufacturing share of total output) vs. Log (income per capita), 1975-1994 Fixed Random Fixed Random Independent variables OLS effects effects OLS effects effects Log income 0.9195 0.5147 0.5726 1.3585 0.7719 0.9402 (2.483) (1.815) (2.076) (7.934) (6.815) (8.323) Log income squared -0.0442 -0.04268 -0.0364 -0.0704 -0.05988 -0.0622 -(1.796) -(2.185) -(1.923) -(6.446) -(8.772) -(8.988) Log income time 0.07527 0.0450 0.0511 --- --- (1.993) (3.126) (3.456) Logincomesquaredtdme -0.0045 -0.0028 -0.0032 --- --- --- -(1.858) -(3.154) -(3.538) Time -0.3194 -0.1614 -0.1911 -0.0120 0.0146 0.0062 -(2.196) -(2.762) -(3.205) -(4.224) (6.612) (3.190) Constant -6.1460 -3.3256 -4.074 -7.9373 -4.2714 -5.3585 -(4.480) -(3.254) -(4.095) -(12.013) -(8.897) -(I 1.415) Numberof 1136 1136 1136 1136 1136 1136 observations Numberoftime 16 16 16 16 16 16 periods Adjusted R-squared 0.299 0.151 0.015 0.297 0.171 0.003 Note: t-statistics in parentheses. the results. They are consistent with an inverted-U position. We have employed panel techniques to esti- model for manufacturing share in national output. mate equation (3) in log-log form, using the log of Our results also suggest some structural change in share-weighted average BOD intensity as the depen- the relationship, since the interactions of time with dent variable. Again, the fixed-effects and random- income and income squared both satisfy classical sig- effects estimates tell the same story: As income per nificance criteria. During the past two decades, the capita increases, overall pollution intensity declines 'inverted-U' appears to have steepened somewhat and because relatively 'clean' sectors grow more quickly. shifted downward. However, our results suggest that the rate of decline To illustrate the implied relationship, we have also decreases. We find no evidence of a structural calculated median manufacturing shares by income change, except for a very slight (but significant) up- class for all 1,717 observations in our sample. The re- shift in compositional pollution intensity. sult (figure 5.1) suggests that man- ufacturing share rises steeply Figure 5.1 Manufacturing share in GDP vs. per capita with income until a country reach- income, 1975-1994 es middle-income status;13 from 25.0 around 10 percent in countries with less than $1000 per capita (Sum- 20.0 mers-Heston income, in $US 1990) c to around 25 percent in countries 0 : 15.0 with incomes of $5,000-$6,000. , s Then the manufacturing share . 9 10.0 slowly declines to around 20 per- cent in countries with $20,000 or Z 5.0 more. 4.2 cangesin setoral0.0 I I I I I II I I I I I I II I I I I l I 4.2 Changes in sectoral . o a 0 0 0 0 0 0 0 0 0 0 o 0 0 0 0 0 0 0 0 0 0 0 composition -° -° °- e u LO LO LO LO LO 0 0 0 0 0 0 0 , \- N ) It~ co co 0 N 'l 0 Table 5.4.2 reports results for our \- _- 2 analysis of changes in sectoral com- Income per capita ($US) Industrial Pollution in Economic Development 71 Table 5.2 Log (sector-weighted BOD intensity) vs. Log (income per capita), 1975-94 Fixed Random Fixed Random Independent variables OLS effects effects OLS effects effects Log income 0.2846 -0.3903 -0.3709 -0.0236 -0.5362 -0.5283 (1.018) -(3.566) -(3.427) -(0.184) -(12.719) -(12.616) Log income squared -0.0234 0.01749 0.0164 -0.0019 0.0269 0.0267 -(1.321) (2.459) (2.344) -(0.249) (11.050) (10.974) Log income time -0.0117 -0.00736 -0.0071 --- --- -(0.416) -(1.214) -(1.171) Lcg income squared tme 0.0009 0.0005 0.0004 --- --- --- (0.578) (1.398) (1.374) Time 0.03203 0.0319 0.0300 0.0034 0.0055 0.0051 (0.278) (1.232) (1.161) (2.166) (6.402) (6.351) Constant 0.6935 3.4369 3.3470 1.7679 3.9912 3.9435 (0.633) (8.168) (8.035) (3.373) (21.143) (20.989) Number of 928 928 928 928 928 928 observations Numberoftime 16 16 16 16 16 16 periods Adjusted R-squared 0.043 0.043 0.043 0.043 0.041 0.041 Note: t-statistics in parentheses. In figure 5.2, we provide an illustration of the Netherlands and Taiwan (China)-have reported relationship between overall pollution intensity and COD (chemical oxygen demand). COD incorporates income during the sample period.14 The figure is the effect of other pollutants on the rate of oxidiza- based on median values of overall intensity for each tion; it is systematically larger than BOD measures. income group in the set of 2,210 observations. It sug- We have controlled for the measurement prob- gests that sector-weighted average water pollution lem by introducing a dummy variable for COD-based intensity declines from nearly 6 Kg to 4Kg per $US1 emissions reports. As expected, the estimated COD million per day, or about 30 percent, as income in- dummy is positive, large and highly significant in all creases to around $5,000 per capita. Then it remains pollution intensity equations. Our sectoral dummy approximately stable over the higher-income range. variable results are also in accord with prior expecta- tions: Food and Paper have the highest average or- 4.3 End-of-pipe pollution intensity ganic water pollution intensities; Metals and Mineral Products have the lowest. In the case of labor intensi- Tables 5.3 to 5.5 report cross-country regression re- sults for equations (5) and (6). We ty, Textiles, Food and Wood Products are highest use dummy variables to control for Figure 5.2 Industrial BOD intensity vs. income per capita sectoral differences in average pol- 5.5 lution intensity; dummy variable 4 controls are also introduced for na- 5 tional differences in reporting pro- ;_ cedures and measures of organic c 45 water pollution. The majority of en- c r= vironmental protection agencies 0 4 (EPAs) have reported emissions of - 3.5 biological oxygen demand (BOD), which is a measure of oxygen re- 3 - I l I I l I I I I l l I I I I I I I moval from water by bacteria which 8 8 8 8 8 8 8 8 are oxidizing organic materials. - cm X to co _ N It r- ° However, three EPAs- for China, Income per capita ($US) 72 Trade, Global Policy, and the Environment Table 5.3 Intensity equations for pollution and labor (in prices and regulation) Dep. Var. - Log. of, Pollution! Pollutionl Labor! Laborl Dep. Var. - Log. O. Output Output Output Output Independent variables Coef t-stat. Coef t-stat. Coef t-stat. Coef t-stat. Log wage - 1.714 -3.055** -0.015 -0.044 -0.711 -8.473** -0.379 -6.380" Log brown index 2.459 0.958 -2.995 -1.601 0.164 0.422 - 1.467 -4.657** Log electricity price 6.123 3.684** 0.620 0.526 -0.098 -0.580 -0.564 -3.354** Log real interest rate 0.455 1.903* 0.029 0.872 Cod 4.308 4.829** 2.406 2.559** Food 5.658 5.044** 4.511 3.940** -0.571 -3.817** -0.813 4.239** Textiles 4.601 4.163** 3.932 3.449** -0.018 -0.125 -0.168 -0.881 Wood products 3.717 2.775** 3.103 2.176** -0.021 -0.140 0.053 0.280 Paper 6.864 6.102** 4.946 4.318** -0.151 - 1.006 -0.231 - 1.205 Chemicals 4.614 3.916** 3.236 2.785** -0.526 -3.320** -0.715 -3.669** Non-metallic minerals 1.290 1.118 1.023 0.879 -0.123 -0.823 -0.242 -11.263 Metals 2.312 1.910* 0.988 0.828 -0.697 -4.383** -0.771 -3.903** Metal products 3.538 3.063** 2.232 1.920* -0.278 - 1.842* -0.502 -2.612** Constant -27.244 -2.466** -0.702 -0.096 -5.253 -3.162** 1.933 1.479 Adjusted R-square 0.63 0.34 0.92 0.81 Number of observations 68 99 80 116 Notes: * Significant at 1% confidence level. ** Significant at 5% confidence level. *Significant at 10% confidence level. Table 5.4 Intensity equations for pollution and labor (in prices and regulation) Pollution! Labor! Pollution! Laborl Dep. Var. - Log of: output output output output Independent Variables Coef t-stat. Coef t-stat. Coef t-stat. Coef t-stat. Log wage - 1.211 -6.153 -0.666 -22.600 Log brown index -4.885 -5.052 -2.872 - 14.642 Log electricity price 5.634 3.565 -0.280 - 1.223 3.765 2.384 - 1.173 -3.871 Log real interest rate 0.370 1.668 0.025 0.807 0.234 0.957 -0.058 -1.27 cod 4.375 4.923 -0.110 -0.852 4.125 4.32 -0.197 -1.071 Food 5.485 4.959 -0.585 -3.979 5.073 4.278 -0.792 -3.767 Textiles 4.586 4.153 -0.018 -0.124 4.556 3.842 -0.016 -0.074 Wood Products 3.654 2.733 -0.028 -0.188 3.429 2.392 -0.131 -0.622 Paper 6.675 6.032 -0.167 - 1.132 6.222 5.247 -0.396 - 1.882 Chemicals 4.246 3.815 -0.558 -3.764 3.364 2.837 -1.023 -4.866 Non-metallic minerals 1.114 0.979 -0.138 -0.941 0.681 0.558 -0.362 - 1.720 Metals 2.014 1.723 -0.722 -4.729 1.243 0.999 - 1.094 -5.045 Metal products 3.323 2.935 -0.296 -2.011 2.784 2.299 -0.560 -2.661 Constant - 16.903 -7.208 -4.403 - 13.734 3.324 0.661 7.522 7.83 Adjusted R-square 0.64 0.93 0.59 0.85 Number of observations 68 80 68 80 Industrial Pollution in Economic Development 73 Table 5.5 Intensity equations for pollution and labor (in income per capita) Dep. Var. - Log of: Pollution/output Labor/output Pollution/labor Independent variables Coef t-stat. Coef t-stat. Coef t-stat. Log income -0.875 -3.26** - 1.003 - 17.041 ** 0.120 0.449 COD 1.908 2.542** 1.930 2.576** Food 4.629 4.096** -0.925 -4.085** 5.492 4.868** Textiles 4.055 3.588** -0.150 -0.662 4.143 3.673** Wood products 3.315 2.350** 0.047 0.206 3.485 2.475** Paper 5.064 4.481 ** -0.350 -1.547 5.353 4.745** Chemical 3.349 2.963** -0.957 -4.225** 4.244 3.762** Mineral 1.151 1.003 -0.361 -1.595 1.414 1.235 Metal 1.119 0.962 -0.964 -4.171 ** 2.038 1.786 Metal products 2.367 2.071 ** -0.635 -2.803** 2.983 2.615** Constant -8.872 -3.61 5** - 1.497 -2.828** -7.246 -2.972** Adjusted R-square 0.35 0.74 0.39 Number of observations 99 116 100 (along with Other Manufacturing, the numeraire sec- Mexico, United States) are in the middle or high tor); Metals and Chemicals are the lowest. income category. Thus, failure to control for the We have also controlled for the possible impact sampling difference might also produce a downward of differences in emissions reporting procedures. In bias in the estimated effect of income or wages on several cases (China, India, Indonesia, Netherlands, pollution intensity. Philippines, Sri Lanka, Taiwan (China), Thailand) the We have introduced a dummy variable to control plant-level information provided by the EPA's in- for this difference, but it is not significant in our cludes employment data. This has enabled us to esti- regressions. In fact, we are not overly surprised by mate sectoral pollution/labor ratios directly from the this result because effective coverage of industrial EPA data. In the other cases (Brazil, Finland, Korea, facilities by both census-takers and regulators is a Mexico and the United States), the EPA's have pro- function of development.15 vided summary pollution data by sector. We have ob- tained summary employment data by sector from THE EFFECTs OF POLLUTION REGULATION AND other national or regional sources, and have used the RELATIVE INPUT PRICES two summaries to calculate sectoral pollution/ labor As expected, the estimated wage-elasticity of labor in- ratios. tensity is large (around -0.70) and highly significant. We recognize the possibility of systematic The wage elasticity of pollution intensity is also neg- differences in the results generated by these two ative, large (-1.71) and highly significant. In the pol- approaches. EPA's in developing countries focus on lution intensity equation, our results are consistent large polluters, so the average pollution intensity of with the hypothesis that labor and pollution are com- these facilities will be reflected in estimates based on plements in production. However, the converse is not plant samples. The situation is potentially quite true. Our index of regulatory strictness is not signifi- different when EPA-reported sectoral emissions are cant in the labor intensity equation. divided by census-reported sectoral employment. While the latter result is not particularly surpris- Plants which ignore pollution regulations (and whose ing, we also find that our regulatory strictness index reported pollution is therefore zero) may nevertheless is not significant in pollution intensity regressions be registered in an employment census. This might which control for wages. Does this imply that market impart a downward bias to summary-based inten- forces alone drive pollution, and that regulation is ir- sities. In addition, all five countries for which we relevant? Although our results are consistent with this employ summary data (Brazil, Finland, Korea, interpretation, we reject it for several reasons. First, 74 Trade, Global Policy, and the Environment our wage and regulation variables are highly collinear ECONOMIC DEVELOPMENT AND POLLUTION INTENSITY because they are both correlated with per capita in- We have also estimated reduced-form intensity equa- come. As table 5.4 shows, each variable is significant tions which control for per capita income, sector and in equations which exclude the other. Second, a large COD reporting. The results are summarized in table body of empirical work suggests that industrial pol- 5.4.5 for three intensities: labor/output, pollution/ out- lution is responsive to pressure from local communi- put and pollution/labor. In all three equations, the ties (Pargal and Wheeler 1996; Hettige and others results for the sectoral dummies replicate the pattern 1997; Hartman and others 1996); as well as formal reg- of results in tables 5.3-5.4. As before, the dummy vari- ulation. Both forms of regulation are strongly affect- able for COD is positive and significant in the pollu- ed by income, reflecting increasing preferences for tion equations. environmental quality and higher valuation of pollu- The results for per capita income suggest a strik- tion damage. We believe that the estimated wage elas- ing regularity across countries. The income elastici- ties of pollution/output and labor/output are both ticity in our pollution intensity regression is capturing negative, and not significantly different from one. In cross-country income effects on formal and informal tet ati on,fweates fereqt of oneu- reguatin, a wel astheeffct o coplemntaity the third equation, we test for the equality of pollu- regulation, as well as the effect of complementarity tion and labor elasticities (w.r.t. income) by regress- with pollution in production. Withcurrently-available ing pollution/labor on the same set of right-hand information, we cannot distinguish clearly between variables (this amounts to differencing the coefficients these two effects. However, their joint effect clearly in the first two equations). The resulting elasticity of shows the impact of rising income on pollution inten- pollution/labor with respect to income per capita is not sig- sity. nificantly differentfrom zero. Of course, we cannot gen- Our results for energy and capital prices are con- eralize from one sample for one pollutant to all siderably weaker. Surprisingly, neither variable is sig- industrial emissions. However,for industrial water pol- nificant in the labor intensity equation when both are lution, our results suggest that sectoral emissions/labor included. In the pollution intensity regression, the ratios are approximately constant across countries at all estimated electricity price elasticity is positive, large income levels. Developing economies generate much and highly significant. The real interest rate elasticity more pollution per unit of output than developed economies, but they also employ much more labor is also positive, and close to significance at the 5 per- . ' ' per unit of output, and in the same proportion. cent level. However, these results are not robust to Figureu5. and table paye estime - chngsinrih-hndvribesorsmpe opoi Figure 5.3 and table 5.6 portray the estimated re- changes in right-hand variables or sample composi- lationship between pollution intensity (per unit of tion. Dropping the real interest rate increases the sam- output) and income per capita. For ease of interpreta- ple size, because we do not have real interest rate data for Mexico, Brazil and Taiwan Figure 5.3. Water pollution intensity vs. income per capita (China). However, with the 100 larger sample the electricity 90 price elasticity loses signifi- . 80 cance in the pollution intensity equation, while becoming - 60 large, negative, and highly 'sig- nificant' in the labor intensity ° 40 equation. We conclude that our H 30 results for capital and energy g 20 prices are highly sensitive to ' 10 outliers, and we see no reason 0 o to draw any clear conclusions 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 20000 from our results. Per capita income ($US) Industrial Pollution in Economic Development 75 Table 5.6 Income and pollution abatement a measure of total output. We simulate the overall Income per ca pita Percent abatement relationship between economic development and industrial pollution by multiplying the four column 1,500 60 entries in each row. The result combines the effects of 3,000 80 changes in total output, manufacturing share, sectoral 7,000 90 composition, and end-of-pipe pollution intensity. I 5,000 95 Column 5 and figure 5.4 portray the total pollution estimate, which has been normalized to an index value tion, we normalize to an intensity value of 100 for the of 100 at the lowest income level. Our result suggests poorest income category ($500 per capita). The cross- that the inverted U-shaped story is only half right for country evidence suggests a sharp drop in pollution industrial water pollution: Total emissions rise intensity with income growth, as manufacturers re- sharply in the range ($500-$7,000), but remain spond to higher wages and regulatory pressures with constant as income increases further. end-of-pipe abatement and process change. From an To assess the contribution of each factor to the emissions index value of 100 at $500 per capita, pol- overall result, we perform three counterfactual lution abatement is about 60 percent at $1,500, 80 per- simulations which are tabulated in columns 6-8 and cent at $3,000, 90 percent at $7,000 and 95 percent at illustrated in figure 5.5. Each simulation allows one $15,000. of columns 24 to vary while holding the other two constant at the lowest-income level. The experiment 5. Implications of the Results in column 6 holds sectoral composition and end-of- 5.1 The Kuznets hypothesis pipe pollution intensity constant, while allowing the Our estimation exercises have suggested three distinct share of manufacturing to vary with income. The patterns of response to economic development. Indus- result is rapid growth of pollution over the whole try's share of national output rises sharply through income range, and an estimated pollution load at middle-income status and then slowly declines. Sec- $20,000 which is eighty times the initial load. To toral composition follows a 'clean' trend for low-in- produce column 7, we allow sectoral composition to come developing countries, but exhibits little or no vary while holding the other two factors constant. trend beyond the middle income range. End-of-pipe Pollution growth is considerably moderated by pollution intensity, by contrast, declines continuous- comparison with column 6, but the projected load at ly with income. $20,000 is still 40 times the initial load. Finally, we We use simulation to project the net result of test the effect of end-of-pipe change in column 8. This changes in these three factors. Our four simulation experiment clearly identifies the most important variables are in columns 1-4 of table 5.7. Column 1 includes a Table 5.7 Industrial pollution and economic development: Simulation broad range of incomes, from experiments $US500 to $US20,000 per capita. Income Manuf BOD EOP Total Variable Variable Variable Columns 2 and 3 replicate the ($Us) share intens. intens. BOD share BOD EOP informatio on manuf500 11.0 5.4 1.00 100 100 100 100 information on manufacturing I .500 13.4 4.8 0.39 128 366 268 118 output shares and average 2,500 16.9 4.6 0.25 167 771 428 127 pollution intensities in figures 3,500 18.5 4.3 0.19 177 1,179 553 133 4.1 and 4.2. Column 4 repro- 4,500 21.0 4.0 0.15 197 1,726 670 138 duces the pollution intensity 6,000 24.3 4.0 0.12 237 2,663 888 144 index in figure 5.3, renor- 8,000 23.5 4.2 0.09 247 3,424 1,230 150 malized to one for the lowest 10,000 23.3 4.1 0.08 253 4,256 1,531 155 income level. 12,000 22.6 4.2 0.07 255 4,953 1,859 160 We assume a unit pop- 14,000 21.2 4.2 0.06 246 5,399 2,188 163 ulation for convenience, so 17,000 20.3 4.3 0.05 248 6,298 2,709 168 income per capita also serves as 20,000 19.5 4.3 0.04 249 7,904 3,559 175 76 Trade, Global Policy, and the Environment Figure 5.4 Industrial pollution and economic development lution per unit of labor (P/L) re- mains approximately constant across the entire range of incomes. § 250 _ This allows us to use commonly- .: ,/ available sectoral labor/output (L/ c 200.. Q) ratios to predict international changes in industrial water pollu- -O 150 btion. As an illustration, we use the World Bank's BESD database to es- timate sectoral L/Q ratios for fifteen 50 III 11111EEE 11111111I countries during the period 1977- o g g g ° 8 ° ° 1989. To estimate BOD loads by sec- t2 in N ) X 8 0 8 8 8 tor, we multiply the L/Q ratios by sectoral P/L coefficients calculated Per capita income ($US)from our regression results for factor: Projected emissions at $20,000 are only 1.8 P/L.16 times the initial load, if manufacturing share and We have chosen the fifteen countries to represent sectoral composition are held constant. large industrial economies in four major groups: We conclude that pollution levels off in the OECD (represented by the United States, Japan, middle income range because end-of-pipe pollution France and Germany (former F.R); the NICs (Mexico, intensity responds to rising wages and stricter Brazil, Taiwan, Korea, South Africa, Turkey); regulation. By comparison, the manufacturing share Asian LDC's (China, India, Indonesia); and the ex- and sectoral composition are minor players. For COMECON countries (Poland, former USSR). The industrial water pollution, the inverted-U pattern does results are tabulated in table 5.9 and summarized in not emerge because declining pollution intensity table 5.8. Taken together, they illustrate the main almost exactly balances output growth, while implications of our empirical analysis. manufacturing share and sectoral composition remain I n of Our empirc analsis. In the OECD, despite modest continued eco- constant beyond the middle income range. nomic growth, estimated BOD emissions remain 5.2 Trends in interational emissions almost constant. In our view, this reflects the countervailing effects of output growth and increases To explore the real-world implications of our results, in wages and regulation; manufacturing shares and we estimate pollution loads for a set of large industri- the 'clean' sector share change very little. The al economies during the period 1977-1989. Powerful COMECON economies are in relative stagnation leverage is provided by our finding that sectoral pol- during the sample period, so there is little movement Figure 5.5 Counterfactual simulations in their estimated emissions. The story for the NIC's is quite 8,000- Vadable: different. Their estimated pollution X 7,000. . Inte:sityincreases by about 25 percent during the X +~~~- Pollution Intensity * 6,000 . -_ Manuf.Share sample period-substantially less than 0 5,000. -* Sector Mix their growth in per capita income. The = 4,000 increase is relatively moderate because X 3,000. rapid output growth is offset by three g 2,000 . factors: the negative impact of increased 1,000. wages and regulation on industrial pol- lution intensity; the first stage of the S 8 8 8 8 8, 8 8o 8 8o 8 gdecline in manufacturing share; and the In 10 LO to LO o o o o NQ o) w co 0 0 NO last stage of the 'clean' trend in sectoral Per capita income ($US) composition. Industrial Pollution in Economic Development 77 Table 5.8 Trends in international emissions-Selected countries, first effects of rising wages and 1977-1989 regulation on pollution intensity. BOD emissions ('000 Kg/Day) To a striking degree, BOD growth Region 1977 1980 /983 1986 1989 in our international sample is OECD 5,776 5,847 5,501 5,403 5,523 due to increased emissions in NIC's 1,565 1,917 1,848 2,197 2,188 developing Asia. Asian LDC'S 4,617 5,030 5,566 6,183 6,883 The overall result of these COMECON 4,127 4,218 4,302 4,228 4,039 changes is a significant shift in Total 16,085 17,012 17,217 18,011 18,633 group shares of total pollution. Table 5.8 shows that the OECD Percent of sample total and COMECON countries drop 1977 1980 1983 1986 1989 significantly in share; the NIC's OECD 36 34 32 30 30 increase marginally; and the Asian NIC's 10 11 11 12 12 LDC's jump from 29 percent to 37 Asian LDC'S 29 30 32 34 37 percent of the total. Equally COMECON 26 25 25 23 22 impressive, however, is the ap- Total 100 100 100 100 100 parently moderate growth in total The Asian LDC experience is also distinctive. BOD emissions during the period Estimated BOD emissions grow by approximately 55 when world concern over environmental damage was percent in these lower-income economies, because reaching a peak. While economic development was rapid output growth and increasing manufacturing sparking greater interest in pollution, is was also share dominate the cleari compositional trend and the setting the stage for real improvements in envi- ronmental performance. From 1977 to 1989, we estimate that total industrial BOD Table 5.9 Estimated industrial BOD emissions-Selected emissions grew by only 16 percent in countries, 1977-1989 ('000 Kg/day) these fifteen major industrial countries. Country 1977 1980 1983 1986 1989 United States 2,652 2,743 2,551 2,454 2,564 6. Summary and Conclusions France 739 716 683 666 652 Germany (former FR) 929 932 800 789 800 In this paper, we have used new inter- Japan 1,456 1,456 1,467 1,493 1,507 national data to investigate the relation- OECD 5,776 5,847 5,501 5,403 5,523 ship between industrial pollution and economic development. To test for a Kuz- Brazil 611 867 771 965 914 nets effect, we measure the effect of Mexico 109 131 130 179 174 income growth on three proximate de- Korea, Republic of 261 282 296 345 377 terminants of pollution: The share of Taiwan, China 208 239 252 296 282p South Africa 226 238 245 245 262 manufacturing in total output; the sec- Turkey I50 160 155 167 179 toral composition of manufacturing; and NlCs , 565 1,917 1,848 2,197 2,188 the intensity (per unit of output) of in- dustrial pollution at the end-of-pipe. We China 3,118 3,358 3,957 4,551 5,023 find thatthe manufacturingsharefollows India 1,309 1 457 1,380 1,277 1,428 a Kuznets-type trajectory, but the other Indonesia 190 214 230 355 433 two determinants do not. Sectoral com- Developing Asia 4,617 5,030 5,566 6,183 6,883 position gets 'cleaner' through middle- income status and then stabilizes. At the Poland 578 581 546 484 459 end-of-pipe, pollution intensity declines U.S.S.R., former 3,549 3,638 3,756 3,744 3,580 strongly with income. We attribute part Ex-COMECON 4,127 4,218 4,302 4,228 4,039 of this to stricter regulation as income in- creases, and partly to pollution-labor Total 16,085 17,012 17,217 18,011 18,633 complementarity in production. 78 Trade, Global Policy, and the Environment Our results suggest that income elasticities of sions in the OECD and ex-COMECON, moderate in- both pollution- and labor-intensity are approximate- creases in the NIC's and rapidly-growing pollution ly minus one. The remarkable implication is that a in the Asian LDC's. During the 1980's, our estimates sector's pollution/labor ratio is constant across countries suggest that the latter group displaced the major at all income levels. Our findings motivate two illustra- OECD economies as the world's largest generator of tive simulation exercises. First, for a set of income organic water pollution. Overall, however, the nega- benchmarks, we simulate total pollution by combin- tive feedback from economic development to pollu- ing representative measures of manufacturing share tion intensity was sufficient to hold total world in output, sectoral composition, and end-of-pipe pol- pollution growth to around 15 percent during a lution abatement. We do not see a Kuznets-type sto- twelve-year sample period. ry in the result, since total pollution rises rapidly In closing, it is worth asking whether these re- through middle-income status and remains approxi- sults are cause for optimism or pessimism. The ap- mately constant thereafter. In three counterfactual ex- propriate answer seems to be 'both.' It is comforting periments, we assess the relative importance of the to see that industrial water emissions level off in rich- three proximate determinants. Our results highlight er economies because pollution intensity has an elas- the dominance of end-of-pipe reductions as wages and tic response to income growth. Unfortunately, unitary regulation increase with development. The combined elasticity implies that total emissions remain constant influence of changes in manufacturing share and sec- unless other factors intervene. Of course, industry toral composition is lower by almost two orders of tends to deconcentrate over time as infrastructure magnitude. improves and prosperity spreads. Constant total emis- Our second simulation uses international panel sions may therefore be consistent with improving data to explore the implications of constant sectoral water quality in at least some areas. However, the pollution/labor ratios. We estimate recent trends in continued existence of many seriously-polluted wa- water pollution for fifteen major industrial nations in terways, even in the most prosperous countries, sug- the OECD, the NICs, Asian LDC's and the ex-COME- gests that economic development remains far short CON economies. We find approximately stable emis- of a Kuznets-style happy ending in the water sector. Industrial Pollution in Economic Development 79 Appendix- Data Sources Mexico: Data for water emissions in the Monterrey Metropolitan Area were provided by the State Brazil: The water pollution data for the Sao Paulo Water Monitoring Authority. The data cover emis- Metropolitan region of Brazil were collected by sions from 7,500 facilities in 1994. Complementary CETESB, the environmental agency for Sao Paulo employment data were provided by Mexico's Cen- State. Our pollution estimates are based on sus Bureau (INEGI). CETESB's 1250-plant database, which includes Netherlands: Water emissions and employment data measures of BOD loads in kg/day. The corre- for approximately 700 regularly-monitored facili- sponding employment data came from the Sao ties in 1990 were provided by the Emissions In- Paulo State Ministry of Labor, which provided 2- ventory System maintained by the Ministry of digit sectoral information from 1991 on nearly Housing, Spatial Planning and the Environment 41,000 plants and 2.15 million workers. (VROM). China: Water pollution data for China were obtained Philippines: Water emissions and employment data from the National Environmental Protection for factories in the Metro Manila Area (MMA) were Agency (NEPA), which maintains a comprehen- provided the Philippines Department of Natural sive database on major sources of industrial pol- Resources (DENR) and the Laguna Lake Devel- lution in China. Our estimates are based on opment Authority. NEPA's 1993 emissions data for 269 factories scat- Taiwan (China): Water emissions and employment tered throughout China. data for 1,800 plants were provided by the Water Finland: The Finnish economic data, aggregated at Quality Protection division of the Taiwan Envi- the 3-digit ISIC level, were provided by the Cen- ronment Protection Agency. tral Statistical Office, covering both white and blue Thailand: Seatec International, a private-sector envi- collar workers for 1989. The pollution data were ronmental consulting firm in Bangkok, provided provided by the Industrial Waste Water Office of plant-level data from two industrial estates in the National Board of Waters and the Environ- Rangsit and Suksawat. The dataset contained in- ment. They cover water emissions in 1992 from formation on water emissions and employment for 193 large water-polluting factories. approximately 450 facilities in 1992. India: The India data are from the state of Tamil Nadu. Sri Lanka: Water pollution and employment data for Plant-level pollution data and employment data Sri Lanka were obtained from a study of waste for 1993-94 were provided by the Tamil Nadu Pol- water treatment options for the Ekala/Ja-ela In- lution Control Board, which monitors air and wa- dustrial Estate, which includes 143 industrial es- ter pollution for all the manufacturing units in the tablishments with 21,000 employees. The data state. were collected by a joint project of the World Indonesia: The Indonesia data came from two differ- Bank's Metropolitan Environment Improvement ent sources. The plant-level emissions data were Program and the Sri Lankan Board of Investment. provided by BAPEDAL, Indonesia's National Pol- Ekala/Ja-Ela industrial estate is one of the two lution Control Agency in the Ministry of Environ- major industrial estates in Sri Lanka. ment. The economic data are from Indonesia's U.S.A.: The information for the United States were Central Statistics Bureau (BPS). drawn from two main sources. The water emis- Korea: Korean pollution data were provided by the sions data have been collected from regional da- National Pollution Control Agency. They cover tabases which monitor industrial water discharges water emissions by 13,504 facilities in 1991. as part of the U.S. Environmental Protection Complementary employment data have been Agency's NPDES system. Employment data are drawn from Korea's National Statistical Yearbooks from the U.S. Census Bureau's Longitudinal Re- and the ILO's International Labor Statistics, 1991. search Database. 80 Trade, Global Policy, and the Environment Notes surveyed in the following section, with a more Our thanks to the many Bank staff members, consult- detailed description in the Appendix. ants, and officials of national environmental protec- 9. See the Appendix for a description of data sources tion institutions who made the data for this study in each country. available to us. We gratefully accept responsibility for 10. Marginal abatement costs decline with treatment any errors of transcription or interpretation. scale for most pollutants, because abatement capi- 1. A partial exception is the work of Selden and Song tal is lumpy. Thus, the estimated output elasticity (1994), whose regressions employ air emissions of emissions in a plant-level equation is generally instead of ambient air quality measures. The lack less than one. For evidence from Asia, see Pargal of monitoring information forces the authors to and Wheeler (1996). At the sectoral level, however, estimate air emissions from secondary sources: the constant-returns assumption seems appropri- National fuel use data and fuel-based pollution ate for cross-country work. It is possible that parameters which are adjusted for conditions in characteristic plant scale is larger in countries with countries at varying income levels. Data scarcity greater sectoral output, but we have no way to test in developing countries is clearly a problem for this proposition with the available data. this exercise. Of thirty countries in the estimation 11. See Christensen and others 1973. sample only four (China, India, Thailand, Turkey) 12. See for example Levinson (1996) and Beede and are LDC's. others (1992). 2. For a discussion of structural change in develop- 13. We have smoothed the series with a three-inter- ment, see Syrquin (1989). val moving average: Each share observation on the 3. See Mani and Wheeler (1997) for further discus- graph is the average for the previous, correspond- sion. ing, and succeeding income groups. 4. "Influent" refers to emissions from industrial pro- 14. We have also used three-interval smoothing for cesses before treatment (or abatement); 'effluent' figure 5.4.2. See footnote 13. refers to emissions to air, water or land after treat- 15. Both approaches may underestimate 'true' sectoral ment; 'concentration' refers to the quantity of pollution intensities in developing countries, be- pollutant per unit volume of the waste stream. cause existing research suggests that medium and 5. For recent empirical evidence, see Hartman and large plants have lower pollution per unit of out- others (1997) and Dasgupta and others (1996). put than smaller facilities (ceteris paribus). Since smaller plants are covered by regulators in devel- 6. These may, according to the circumstances, include oped economies, our econometric result may local administrators, pressure groups, national actually understate the effect of income on pollu- regulators, stockholders, and "green consumers." tion intensity. We accept the plausibility of this Each group is in a position to impose some cost hypothesis, but we have no way to test it at on a firm or plant if its emissions exceed the norms present. adopted by that group. Thus, even where pollu- 16. For this application, we regress log (P/L) on tion charges are in effect, there is no single "price" dummy variables for COD and the industry sec- of pollution. For a detailed discussion, see Afsah dummy viabe for CO and te inursec- and~~ otes(19) tors. Since income is insignificant, we impose a parameter value of zero by dropping it from the 7. Forrecentevidence,seePargalandWheeler(1996), equation. To calculate sectoral P/L ratios, we as- Wang and Wheeler (1996), and Hettige and others sume the BOD case (COD=O), add the constant (1997). term to the estimated parameters for the sector 8. Data-gathering in this context is not a simple task. dummies, and calculate the antilogs of the results. Even assembly of the relatively sparse dataset used References for this exercise has required a massive canvass of World Bank project files, consultants' reports, and Afsah, S., B. Laplante, and D. Wheeler. 1996. "Control- emissions reports from many national environ- ling Industrial Pollution: A New Paradigm." World mental protection institutions. The data are briefly Bank, Policy Research Department Working Paper. Industrial Pollution in Economic Development 81 Beede, D., D. Bloom, and D. Wheeler. 1992. "Measur- Hettige, H., M. Huq, S. Pargal, and D. Wheeler. 1996. ing and Explaining Cross-Establishment Variation "Determinants of Pollution Abatement in Devel- in the Generation and Management of Industrial oping Countries: Evidence from South and South- Waste." World Bank (mimeo.) east Asia." World Development December. Birdsall, N., and D. Wheeler. 1993. "Trade Policy and Hettige, H., M. Singh, S. Pargal, and David Wheeler. Industrial Pollution in Latin America: Where are 1997. "Formal and Informal Regulation of Indus- the Pollution Havens?" Journal of Environment and trial Pollution: Evidence from the U.S. and Indo- Development 2(1): 137-44. nesia." World Bank Economic Review Fall. Christensen, L., D. Jorgensen, and L. Lau. 1973. "Tran- John, A., and R. Pecchenino. 1992. "An Overlapping scendental Logarithmic Production Frontiers." Re- Generations Model of Growth and the Environ- view of Economics and Statistics 55: 2845 ment." Department of Economics, Michigan State Dasgupta, S., and D. Wheeler. 1996. "Citizen Com- University (mimeo). plaints as Environmental Indicators: Evidence Levinson, A. 1996. "Environmental Regulations and from China." World Bank, Policy Research Depart- Manufacturers' Location Choices: Evidence from ment Working Paper, December. the Census of Manufactures." Journal of Public Eco- Dasgupta, S., A. Mody, S. Roy, and D. Wheeler. 1995. nomics 62: 5-29. "Environmental Regulation and Development: A Mani, M.S., and D. Wheeler. 1997. "In Search of Pol- Cross-Country Empirical Analysis." World Bank, lution Havens? Dirty Industry in the World Policy Research Department Working Paper no. Economy, 1960-1995." World Bank, Policy Re- 1448, April. search Department Working Paper. Dasgupta, S., M. Huq, D. Wheeler, and C.H. Zhang. Mani, M.S. 1996. "Environmental Tariffs on Polluting 1996. "Water Pollution Abatement by Chinese In- Imports: An Empirical Study.' Environmental and dustry: Cost Estimates and Policy Implications." Resource Economics 7: 391-411. World Bank, Policy Research Department Work- Mody, A., and D. Wheeler. 1990. Automation and World ing Paper. Competition: New Technologies, Industrial Location, Grossman, G., and A. Krueger. 1995. 'Economic an rd.Lno:McilnPes G,rowth and the Environment." Quarterly Journal anTrd,Ldo:McilnPes GroftEconom May: 353-37nment." QuarterlyJoumal Pargal, S., and D. Wheeler. 1996. "Informal Regula- Gruver, G.W.1976. "Optimal Investment in Pollution tion in Developing Countries: Evidence from In- Contero Captal ma76 NeOptimlaInvessi eth CnPolluiont donesia." Journal of Political Economy (December). JoutrolCapitalofnironmtal NeocoicsandG h Managemnt. Robison, D.H. 1988. "Industrial Pollution Abatement: 5: 165-177. The Impact on the Balance of Trade." Canadian Hartman, R., M. Huq, and D. Wheeler. 1997. "Why Journal of Economics 21: 702-706. Paper Mills Clean Up: Determinants of Pollution Selden, T., and D. Song. 1994. "Environmental Qual- Abatement in Four Asian Countries." World Bank, ity and Development: Is There a Kuznets Curve Policy Research Department Working Paper, for Air Pollution Emissions?" Journal of Environ- Number 1710. mental Economics and Management 27: 147-162. Hartman, R., M. Singh, and D. Wheeler. 1997. "The . 1995. "Neoclassical Growth, the J Curve for Cost of Air Pollution Abatement." Applied Econom- Abatement, and the Inverted U Curve for Pollu- ics 29(6): 759-74. tion." Journal of Environmental Economics and Man- Hettige, H., R.E.B. Lucas, and D. Wheeler. 1992. "The agement 29: 162-168. Toxic Intensity of Industrial Production: Global Shafik, N. 1994. "Economic Development and Envi- Patterns, Trends and Trade Policy." American Eco- ronmental Quality: An Econometric Analysis." nomic Review Papers and Proceedings 82: 478-481. Oxford Economic Papers 46: 757-773. Hettige, H., P. Martin, M. Singh, and David Wheeler. Syrquin, M. 1989. "Patterns of Structural Change." In 1995. "IPPS: The Industrial Pollution Projection H. Chenery and T.N. Srinivasan, eds., Handbook of System." World Bank, Policy Research Department Development Economics Vol.1. Amsterdam: North- Working Paper, February. Holland. 82 Trade, Global Policy, and the Environment Tobey,J.A. 1990. "The Effects of Domestic Environrmen- Wheeler, D., and A. Mody. 1992. "International In- tal Policies on Patterns of World Trade: An Empiri- vestment Location Decisions: The Case of U.S. cal Test." Kyklos (Fasc.): 191-209. Firms." Journal of International Economics 33: 57-76. Wang, H., and D. Wheeler. 1996. "Pricing Industrial Wheeler, D. 1991. "The Economics of Industrial Pol- Pollution in China: An Econometric Analysis of lution Control: An International Perspective." the Levy System." World Bank, Policy Research World Bank, Industry and Energy Department Department Working Paper, No. 1644. Working Paper, No. 60, January. Chapter 6 Preferential Trading Arrangements between Kenya and the EU A case study of the environmental effects of the horticulture sector Anil Markandya, Lucy Emerton, and Sam Mwale 1. Introduction purpose in some cases (Unemo 1993; Tyler 1996). Oth- s part of the effort to promote development, ers have used sectoral and macroeconomic models for developed countries and regions such as the the same purpose (Cruz and Repetto 1993; Markandya X . XEuropean Union (EU) offer preferential ac- and others 1997). Although arguments can be made cess to producers from developing countries. The for and against each type of model, use is often dic- rationale is obvious: preferential access promotes em- tated by data considerations. The present study was ployment through exports by the affected sectors; pro- undertaken in the spirit of these intersectoral studies. vides an impetus to industrialization; and helps to It looks at one sector, horticulture, and estimates the create the conditions for export-led growth. At the environmental costs of growing several different same time, however, the privileged sectors can cause crops. These costs are then compared with those of harm to the environment and forestall advances to- other crops, especially those for which horticulture ward sustainable development. may substitute. The paper then analyzes the impact An evaluation of preferential trading arrange- of preferential trading policies on land use, the envi- ments must examine both the economic and environ- ronment, and overall employment. We do not take a mental consequences of these arrangements. Until full general equilibrium approach (the data would not recently the environmental dimension was largely permit such an analysis), but rather examine the rela- ignored, although attention was drawn to the envi- tionships between major competing agricultural uses. ronmental consequences of liberalizing trade in The "novel" features are: (a) a detailed assessment of Anderson (1992), Grossman and Kruger (1993), Mar- the environmental costs of different crops, and (b) the kandya (1994), Munasinghe and Cruz (1995), and oth- use of a relatively simple "profitability" rule to allo- er works. It is difficult to assess whether a change in cate land between crops when costs are changed ei- trade regime is beneficial if the change has significant ther by internalizing environmental costs, or by giving environmental impacts. The same may be said of pref- different preferential access to certain products. erential trading arrangements, which change the Section two provides background on the horti- structure of production in both countries and thereby cultural sector in the general context of agriculture in alter the use of natural resources and provoke chang- Kenya. Section three reviews preferential trading ar- es in employment, output, and possibly growth. rangements for horticulture. Section four discusses the The study of such impacts is complex, because environmental impacts of different crops, and section intersectoral effects have to be taken into account. five provides monetary estimates of the damages in General equilibrium models have been used for this Ksh (Kenyan shillings) and U.S. dollars per hectare. 83 84 Trade, Global Policy, and the Environment Section six develops a model for estimating changes ing rapidly. The growth of smallholder production is in land use when the prices received for the different largely a result of land pressure arising from rapid crops are altered and applies it to assess the effects of population growth, outgrower schemes by large pro- preferential trading arrangements and the impact of ducers and exporters, and long-term declining returns internalizing environmental costs. Section seven of- from alternative cash crops. As land size decreases, fers a conclusion, noting the limitations of the present households shift to higher-value crops to ensure in- analysis and the implications of its results for prefer- come security. Income from horticulture is used to ential trading policy. purchase food from markets. Horticultural produc- tion by smallholders is most pronounced in areas 2. The Horticultural Sector in Kenya where the commercialization of agriculture has had the strongest impact. The horticultural sector is one of Kenya's most profit- Anot sourcer able nd rpidl expndin expot setors ItsmainAnother source of growth in horticulture is the able and rapidly expanding export sectors. Its main structural adjustment policies (SAPs) in the agricul- products are cut flowers, Asian vegetables (mainly tural sector that the Kenyan government has pursued brinjals, karella, chillies, and okra), and French beans- over the last decade. Such policies have resulted in The growing globalization of markets brings Europe- real increases in input prices (for example, fertilizer an consumers a year-round supply of fresh fruits, and pesticides), while output prices for food commod- vegetables, and flowers, many of which come from ities have been declining in nominal and real terms. African Caribbean Pacific (ACP) countries. Kenya and The result has been declining net profits in tradition- Zimbabwe are two of the most important ACP horti- al cash crops, encouraging the shift to horticulture. cultural producers. Although this sector is relatively Land intensification, application of better husbandry new in Kenya, by 1993 horticulture had displaced practices, and the introduction of higher-yielding va- coffee as the third most important national foreign rieties have also contributed to growth in the horti- exchange earner, after tourism and tea. cultural sector. The amount of land shifted over to The rapid growth of export horticulture can be horticulture has been relatively small at the national attributed to several factors. First, preferential treat- level, but highly significant in particular regions of ment under the Lom6 Convention between ACP and the country (Richardson 1996). the EU provides concessionary access for Kenyan flowers and vegetables to the European market. Sec- Trends in horticulture ond, the sustained demand for horticultural products The main trends in horticultural production and ex- as a result of high-and growing-incomes in Europe ports are given below. Table 6.1 provides data on pro- provides a stable and growing market for Kenyan duction, area underproduction, and exports for Asian producers. Third, Nairobi's location as a hub of air vegetables. Table 6.2 gives the same data for French transport between Europe and the East and Southern beans. Unfortunately data on cut flowers are more Africa region, and Kenya's role as a major tourist des- fragmented, as described below. tination, ensure that there is sufficient northbound air Production volumes for Asian vegetables and cargo to transport exports. Finally, the presence of French beans vary considerably by year for reasons ample local and international investors, particularly that are not yet known. Trends, however, have been in the cut-flower business, provides Kenya with an upward for Asian vegetables and static for French added advantage. beans. Although large-scale producers dominate the Export volumes of Asian vegetables remained horticultural sector, smallholder production is increas- mostly static from 1988 to 1993. The South Asian pop- Table 6.1 Indicators for Asian vegetables Indicator 1989 1990 1991 1992 1993 1994 1995 Production (tons) 21,773 28,557 46,682 80,662 22,695 70,040 72,357 Area under production (ha) 3,102 2,392 3,675 2,618 1,766 5,162 5,365 Exports (tons)* 5,078 7,260 7,240 7,649 5,173 Note: Asian vegetables counted in exports here are dudhi, okra, karella, chillies, and others. There were no data for brinjals. Sources: HCDA, Ministry of Agriculture. Preferential Trading Arrangements between Kenya and the EU 85 ulation in the United Kingdom is the largest consum- horticultural produce in 1991 were the United King- er of these commodities. Prospects for future growth dom (34 percent), the Netherlands (22 percent), France are limited, as that market is declining and competi- (20 percent), and Germany (13 percent), which togeth- tion from other suppliers is increasing. Kenya's French er accounted for 89 percent of all exported fresh pro- beans have been more or less constant in terms of ex- duce. When other EU-member countries are included, ports, but the area under production has increased. the European Union imported over 93 percent of Ken- They continue to hold their own against stiff compe- ya's fresh produce exports. The European continent tition from North African countries such as Morocco, imported over 95 percent of Kenya's horticultural ex- and new entrants such as The Gambia and Burkina ports in 1991, when non EU-member countries such Faso (See table 6.2). as Switzerland are included (Harris 1992). Table 6.2 Indicators for French beans Indicator 1989 1990 1991 1992 1993 1994 1995 Production (tons) 17,832 13,565 24,265 22,265 19,624 18,271 17,400 Area under production (ha) 3,082 3,707 5,939 6,190 5,807 4,792 4,524 Exports (tons) 15,227 16,329 14,852 15,197 14,476 Sources: HCDA, Ministry of Agriculture. Cut flowers have performed well as a cash crop, Table 6.3 Exports of fresh largely as a result of high-quality production by large- horticultural products: 1985-1993 scale, vertically integrated, local and international Volume Change producers with access to new planting materials. Ken- Year (tons) (percent) yan cut flowers have maintained a large market share 1 985 30,000 2 in Dutch flower auctions, which serve as the princi- 1987 36,550 0.9 pal distributors for Europe and the Middle East. De- 1988 59,119 61.8 spite stiff competition Kenya has managed to retain 1989 49,503 - 19.4 its market share over the past four years. 1990 49,147 - 0.7 Kenya's total horticultural exports have grown 1991 49,848 1.4 rapidly since 1972, from just under 8,000 tons to over 1992 57,363 15.1 65,000 tons in 1994, as shown in table 6.3. Acreage 1993 62,129 8.3 has increased correspondingly, and is currently Source: HCDA6 around 1,445 hectares. As discussed above and else- where, the diversity of commodities may indicate that Market participants Kenya's horticultural sector has internal robustness. Yet analysis of the composition of exports, not only Smallholders and large farms produce both French by volume but also by value, shows significant com- beans and Asian vegetables; small-scale producers modity and market concentration. The principal hor- predominate throughout Kenya, except in Kibwezi ticultural exports in 1991 were cut flowers (33 percent), and Embu. Although some producers are landown- French beans (30 percent), and Asian vegetables (19 ers, land is commonly leased to smallholders for two percent), which together constituted 82 percent of all to three years and to largeholders for five to ten years. horticultural exports (Harris 1992). Principal markets Tenure arrangements vary by region. In Loitokitok, include France, Germany, the Netherlands, and the for example, most producers lease land for short pe- United Kingdom, all of which are EU-member riods,andproductionispredominantlybysmallhold- countries. ers. In Kibwezi, in contrast, there is a significant Relatively few importers dominate horticultur- presence of large-scale farmers, some of whom are al markets. As noted earlier, Asian vegetables are pri- tenants on five- to ten-year leases. However the ma- marily exported to the United Kingdom, French beans jority of producers are smallholders, many of whom to France, and cut flowers to Germany and the Neth- have two- to three-year tenancy arrangements. In erlands. The leading importing countries of Kenyan Machakos and Makueni most horticultural produc- 86 Trade, Global Policy, and the Environment ers are landowners, while in Embu and Meru a mix while flower producers tend to market directly to of tenancy and land ownership prevails. Europe. Cut- flower producer-exporters dominate the Land-tenure arrangements have an important market. One Kenyan cut-flower producer-exporter is impact on the environment. In order to minimize risk responsible for more than one-quarter of all horticul- and maximize output, it is rational for landowners to tural produce earnings, and three-quarters of all cut- invest in the sustainable use of land resources to safe- flower exports come from two firms. Much the same guard their resource base. Conversely, evidence sug- situation holds for French beans and Asian vegetable gests that short-term tenants are unwilling to make markets. Three large firms and a handful of medium- investments in land maintenance and improvements scale exporters dominate the market. Despite the ev- over which they have no long-term security or own- idence of market power, the share not donminated by ership. The evidence on this issue is mixed. In Loito- large firms is highly dynamic, comprising many so- kitok landowners are predominantly pastoral Maasai called "briefcase exporters," who opportunistically who have recently turned to agriculture. In the case buy fresh produce from smallholders. of the Maasai, landowners and tenants differ little in regard to soil and water conservation practices. In Inputs and returnsfor horticultural production and other parts of the country, however, landowners ex- alternative cash crops hibit conventional behavior by investing in soil and Horticultural production is generally quite profitable water conservation, such as sprinkler irrigation, which (table 6.4). In 1995 profits per hectare ranged from Ksh has lower wastewater runoff. Tenants tend to use fur- 9,262,400 for roses (largeholder) to Ksh 17,500 for okra row or flood irrigation, a more environmentally dam- (smallholder). This compares favorably with profits aging practice. from coffee of Ksh 250,502 (largeholder) to Ksh 35,775 In the Aberdares and Limuru regions the bulk (smallholder) per hectare and for tea from Ksh 139,748 of cut flowers (carnations and roses) are produced by (largeholder) to Ksh 43,699 (smallholder). The ex- largeholders, although some smallholders produce change rate in 1995 was Ksh56 to US$1. This rate will statice and alstromelia, because of the capital- be used throughout this paper. intensive and more technical nature of production. Horticulture is labor intensive. Alstromelia, for Most rose operations rely heavily on Israeli and Dutch example, requires 803 person days per hectare, com- expatriates for technical assistance and production pared to 442 for pyrethrum, 318 for tea, and 294 for management. In 1994 just over 50 flower operations coffee (Egerton and others 1995). Moreover, capital existed in Kenya, mostly around Nairobi. outlay and costs are high. For vegetables (French Unlike many of the French bean and Asian veg- beans, karella, okra, chillies, and snow peas) the av- etable producers, cut-flower producers tend to be ver- erage capital outlay covering all costs ranges from Ksh tically integrated and to manage their own overseas 204,809 for chillies to Ksh 49,533 for French beans, an marketing. French bean and Asian vegetable produc- average of Ksh 106,535 per hectare. In comparison ers rely mainly on nongrower exporters as middle- total costs in coffee range from Ksh 26,239 to Ksh men between the producer and European markets, 71,215 per hectare. The average for various produc- Table 6.4 Returns from horticultural crops Total Revenue Profits Crop Yield/ha (Ksh) Total Cost (Ksh) (Ksh) Roses 1,800,000 stems 14,400,000 5,137,600 9,262,400 Statice 435,708 stems 435,708 137,169 298,539 Alstromelia 700,245 stems 854,299 123,411 730,888 French beans 1,606 kgs 96,330 49,533 46,797 Karella 2,149 kgs 214,890 137,789 77,101 Okra 711 kgs 85,363 67,858 17,505 Chillies 3,557 kgs 426,816 204,809 222,007 Snow peas 3,522 kgs 316,911 72,685 244,226 Source: Egerton and others (I 995). Preferential Trading Arrangements between Kenya and the EU 87 tion systems across the country is Ksh 48,424 per hect- tural production are Naivasha and parts of Loitoki- are. Tea costs are lower on average than those for tok, where extensive irrigation facilitates production horticulture or coffee, ranging from Ksh 23,284 (small- of horticultural crops. holder) to Ksh 74,935 (largeholder) per hectare (Eger- In terms of susceptibility to environmental deg- ton and others 1995). radation, suitability class one land is least suscepti- Only a limited amount of Kenya's land is suit- ble; class two is moderately susceptible, and class able for tea and coffee production. Horticulture (in- three is the most susceptible. Suitability class one ar- cluding domestic production) is suited to a wider eas are least susceptible to damage from either rain- range of land classes. Land suitable for horticulture is fed or irrigated agriculture. Due to the steepness of classified into three classes, taking all production fac- the land, however, surface runoff and soil erosion are tors into consideration. The amount of land under potential sources of degradation, particularly once cultivation in each class zone is listed in table 6.5. groundcover has been lost. The impact of pesticides Table 6.5 Land potential and environmental sensitivity of selected crops Land in class I Land in class 2 Land in class 3 Land in classes 1 &2 Horticulture 8.6 29.4 62.0 38 Coffee 21.7 27.9 50.4 49.6 Tea 21.4 16.4 62.2 37.8 Maize 3.4 38.5 58.1 41.9 Beans 6.5 46.4 47.1 52.9 Suitability class one is the best land for horticul- and fertilizer depends on production techniques. tural crops. Major alternative crops are tea, dairy, and Suitability class two areas are more susceptible food-crops . This land consists of deep, medium-tex- to environmental degradation, particularly as a result tured soils with moderate natural fertility that is well of the impact of irrigation and rising water tables on drained, slightly saline, and nonsodic. Examples in- the soil. Light soils are easily erodible and heavy soils clude the French bean and snow pea growing areas are easily waterlogged. Rising water tables cause sa- around Mt. Kenya and the Aberdares, as well as west- linity, and impeded drainage increases sodicity. Fer- em Kenya. Most production is rainfed. tilizer and pesticide residues are retained longer in Suitability class two land has medium potential heavy soil, since leaching occurs more slowly in heavy for horticulture. Major alternative crops are coffee, soil. Most smallholder horticultural production takes dairy, and food-crops. Suitability class two horticul- place in these areas, using some form of irrigation. tural land is characterized by heavy or light textured As will be shown later, the environmental impact on soils with low natural fertility that are moderately this class of land is considerable. deep, slightly saline, nonsodic, and well drained. Ex- Suitability class three lands are the most suscep- amples include the cut-flower growing areas around tible to environmental degradation. The enviromnen- Aberdares, Mt. Kenya, and Limuru. These areas are tal impacts are similar to those on class two land, but largely rainfed, but with some irrigation. more are severe, largely as a result of the production Suitability class three land has low potential. systems and climatic conditions. Most vegetables, These lands are characterized by strongly sodic, light particularly Asian vegetables, are grown in this zone, or heavy, moderately well-drained soil, with low nat- largely by smallholders. The combination of arid cli- ural fertility and moderate salinity. Unlike for class mate and water from furrow irrigation makes fungus one or two lands, horticulture has few competing ar- and pest attacks particularly common for Asian veg- able activities. Marginal food-crop production, live- etables in Machakos and Makueni. Increasing pest stock, and wildlife are the major altemative land uses. attacks have led to intensive pesticide use. Yet the Examples include French bean and Asian vegetable greatest volume and value of horticultural produc- production in Loitokitok, Makueni, Machakos, and tion is from this zone, particularly when the Lake Mwea and cut-flower production in the Aberdares. Naivasha region is included. This zone has the great- Other agro-ecological zones with significant horticul- est irrigation potential in the country, and more land 88 Trade, Global Policy, and the Environment is available for short- and long-term leasing in class on- and off-farm. This section presents a general out- three areas than in class one or two zones. line of the environmental hazards of crop production in Kenya's major horticultural areas. The two major 3. Tariff Structure for Horticultural and levels upon which agricultural production may im- Related Products pact on the environment are: * Agricultural land-use practices: environmental im- Kenya currently has tariff-free access to the EU mar- pacts arising from the use of land, vegetation, wa- ket for horticultural products. This puts the country ter, chemicals, and other inputs to generate farm in a preferential position, for example, with respect output. to non-EU Mediterranean basin countries that face * Intermediary goods andfinal products: environmen- tariff rates of 4.2 percent for cut flowers and 5.3 per- tal impacts arising from the use of inputs in cent for vegetables. The tariffs for other countries' producing, processing, storing, using, and trans- (non-Mediterranean basin or ACP) horticultural prod- porting the intermediary goods necessary to sup- ucts entering the EU are 15 percent for cut flowers port agricultural production and final agricultural and 13 percent for vegetables. products. In the analysis that follows we examine the con- sequences of a change in tariff rates that would: (a) Environmental impact of agricultural land-use impose the non-EU tariff on Kenyan products, or (b) practices impose a "tariff" that internalizes environmental costs. The use of land, vegetation, water, and chemicals in The latter need not, of course, be a tariff, but could be agricultural production can have a number of over- an internal tax, imposed by the government to inter- lapping impacts on the environment. Agricultural nalize environmental costs. Estimates of the environ- land-use replaces natural vegetation systems. Crops mental costs are provided in the next section. introduced may be environmentally inferior to the Table 6.6 Impact of removing preferential tariffs on profitability (Ksh/ha) Average private Tariff at Medit. Tariff at Crop profitability Basin level third-country level French beans 46,797 5,105 12,523 Asian vegetable (karella) 77,101 11,389 27,936 Statice (large producers) 298,539 18,300 65,356 Alstromelia (small producers) 730,888 35,881 128,145 Source: Emerton and others (1997). The consequences of the preferential tariff are vegetation they replace, especially when they are demonstrated in table 6.6. If one of the tariffs suggest- poorly managed or grown in unsuitable or vulnera- ed above were imposed, profitability would decline ble areas. The major land-use and vegetation-based as detailed in the table. hazards of crop production are soil erosion and ex- The impacts of the changes in profitability are haustion, watershed change, and loss of biological discussed further in section five. At this point we sim- resources. ply note that the strongest impact would be on mar- Soil erosion, runoff and sedimentation. Most of the ginal producers, many of which will be small areas where horticulture is practised in Kenya lie ei- enterprises. ther in mountainous or semiarid areas. Much cultiva- tion takes place on steep slopes and unsuitable, 4. Environmental Costs of Different Crops marginal lands where the risk of soil erosion is high. in Kenya Pockets of saline soil have been found in a number of Agricultural production affects the environment irrigated horticultural areas, including Mwea (NES through its use of natural, chemical, mechanical, and 1981). The combination of soil salinity, overintensive human inputs to produce, process, store, and trans- land-use, and soil exhaustion has resulted in a loss of port output. These environmental impacts are felt both soil fertility and depth. Preferential Trading Arrangements between Kenya and the EU 89 When ground protection is lost and soil erosion Multiple varieties of indigenous crops exist; re- occurs, local runoff increases and greater volumes of placing these crops with exotic varieties creates the soil are carried off-site. Sediment loads in rivers can risk that indigenous landraces will no longer be plant- increase significantly, and higher levels of silt are ed and thus lost over time. The destruction of landrac- transported downstream. The rivers located down- es is important primarily because it narrows the stream from horticultural areas, most notably the genetic base of agriculture and increases the risk of Tana, Athi, and Ewaso Ngiro, as well as smaller crop failure. Moreover, undisturbed vegetation streams and watercourses, carry extremely high sus- around horticultural areas, especially within the pended sediment loads (UNEP 1987). Clearance of Mount Kenya, Aberdares, and Naivasha regions, pro- natural vegetation for agriculture and poor subse- vides habitat for a range of animals and birds, includ- quent land management are major factors contribut- ing rare and endangered species. In addition many ing to these high sediment loads. households rely on natural vegetation to supply sub- Increased runoff and sedimentation can lead to sistence and income needs including timber, charcoal changed waterflow characteristics in downstream fuelwood, polewood, medicines, grazing, and wild watercourses and exacerbate dry-season water short- foods (Emerton 1994b). age and wet-season flooding. The diversion of sur- Chemical pollution. The increasing use of chemi- face water for irrigation may also alter waterflow. cals is also having significant effects. Kenya imports Local water tables can fall when groundwater re- chemicals, so the environmental costs of their produc- charge levels decrease, as a result of lower soil mois- tion accrue outside the country. Many of the agro- ture, and when pumping groundwater for irrigation chemicals in general circulation in Kenya have been exhausts aquifers. Overuse of local water supplies is banned or designated as restricted-use chemicals in a major concern in the flower-growing areas around the developed world because of adverse health effects Lake Naivasha (Naitore 1995). Changed waterflow or environmental hazards (Matteson and Meltzer results in downstream production losses and in- 1995). About one-half of the pesticides used for vege- creased damage by seasonal water shortages, floods, table production in Kenya fall into this category (Ki- and reductions in groundwater resources. bata 1993). Within the horticultural sector vegetable crops Agrochemicals, especially herbicides and soil generally provide poorer ground cover than flower sterilants, can contaminate surface water and ground- crops, and are more likely to give rise to the above water resources. As a result of chemical pollution effects. Nonhorticultural perennial products such as water may become unfit for human consumption and tea, coffee, and dairy products, however, provide ex- unable to support plant and animal populations. Pol- cellent ground protection and cause little soil erosion; lutants may accumulate in living organisms and, over maize and beans provide poor groundcover and in- time, reach toxic levels. Regions with intense horti- crease rates of soil erosion. The transfer of land from cultural production, especially lake wetlands and perennial to horticultural or nonhorticultural annual watersheds, are of particular concern. Around Lake crops is likely to contribute to increased soil erosion, Naivasha pyrethroid insecticides have been found in loss of soil fertility and depth, and associated down- runoff and streams; soil sterilants have also been de- stream effects. These effects will be especially intense tected (Matteson and Meltzer 1995). The British Broad- on erodible soils in major watersheds, such as the casting Corporation has reported livestock poisoning Mount Kenya and the Aberdares regions. from Aldrin runoff in the area around the lake (BBC Biodiversity. Exotic agricultural crops replace 1995). There is also apprehension about horticultural natural vegetation and indigenous landraces. Clear- pesticide residues found in Lake Nakuru and their ance of vegetation for agriculture creates a loss of di- effect on the lake's flamingo population. Evidence verse plant species and natural habitats. Natural suggests that fish populations are also affected by vegetation also provides a range of goods and servic- agrochemical water pollution. Studies carried out in es, including direct uses, ecological benefits, and op- Lake Naivasha and the Tana and Athi Rivers, all of tion and existence values, which effectively represent which are located downstream from major horticul- the opportunity cost of agricultural land. tural and irrigation areas, have found a wide array of 90 Trade, Global Policy, and the Environment pesticide residues in fish, some of which make them wastewater. In Mwea, a major area of irrigated horti- unfit for human consumption (Matteson and Meltzer culture, the incidence of waterborne diseases among 1995). A number of water bodies downstream from farmers is well above the norm. Thirty one percent of horticultural areas are displaying signs of eutrophi- the residents show signs of bilharzia, 15 percent have cation (NES 1981). ascaris, 15 percent hookworms, and up to one-half On-site human health risks arise from the inges- show signs of malaria (NES 1981). tion of agrochemicals through contaminated contain- Most of the cash crops grown in horticultural ers and other water and airborne transfers that can areas rely on chemical fertilizers. As soil fertility de- cause skin, eye, gastrointestinal, and respiratory ill- clines as a result of intensive production, causing soil nesses. Long-term health effects include pathogene- erosion and exhaustion, increasing quantities of fer- sis of childhood aplastic anaemia, chronic mercury tilizer are applied to maintain soil fertility. This can poisoning, and impotence (Matteson and Meltzer lead to agrochemical pollution, as discussed above. 1995). Commonly applied pesticides that are reason- Heavy fertilizer use also contributes to greenhouse ably safe for human applicators may be acutely toxic gas emissions. Nitrous oxide (N20), a major green- to fish, birds, bees, beneficial insects, and other desir- house gas, is produced from soils through the pro- able nontarget species. cess of denitrification and nitrification, both of which Most on-site human health hazards arise because are associated with the application of agricultural fer- of unsafe chemical application practices. In the Ken- tilizers. Synthetic fertilizers are more harmful because yan small farms sector most pesticides are hand-mixed they are comprised of extra nitrogenous elements and and applied by hand, twig brooms, or backpack spray- lose up to 4 percent of their nitrogen when applied ers (KNFU 1987). Protective measures are rarely tak- (Adger and Brown 1994). Finally, some evidence in- en, and farmers are generally ignorant of pesticide dicates that a range of agricultural pests have become safety practices (Matteson and Meltzer 1995). Surveys resistant to commonly used pesticides, which could in major horticultural areas have found that most lead to a vicious circle of increased pesticide use to farmers experience occasional poisoning symptoms cope with increased resistance. after applying agrochemicals (Partow 1992). It has been estimated that pesticides poison 7 percent of the Environmental impact of intermediary goods and people in Kenya's agricultural sector every year (Mat- final products teson and Meltzer 1995), and that one-quarter of the Agricultural production relies on a range of interme- farmers in major horticultural areas regularly suffer diary goods and services to generate output. Produc- from pesticide-related ill health (KNFU 1987). Orga- ing final products for the market involves processing, nophosphate pesticides constitute the most acute dan- storage, packaging materials, and transportation. ger because of their relatively high toxicity. About 90 These stages of the production process use inputs that percent of the agrochemical poisoning cases treated involve environmental risk between 1987 and 1990 at Kiambu District Hospital, Most horticulture consists of exotic varieties, which is located in a growth area for coffee, vegeta- such as French beans, Asian vegetables, and cut flow- ble, and flower production, involved organophos- ers. These crops are produced from exotic seeds and phates (Mwanthi and Kimani 1990), and one-tenth of seedlings originally imported from overseas. Import- all illnesses treated at Lake Naivasha North ed pests and viruses may accompany the introduc- Clinic involved pesticide poisoning of flower-farm tion of exotic planting materials. Pests from the United workersomainly rashes, sores and gastrointestinal States and Europe may already have been introduced problems (BBC 1995). High organochlorine residues, into central Kenya via imported flower seeds. Once especially DDT and Dieldrin, have been found in hu- pests are imported with seed, they can quickly spread man milk samples taken from mothers in horticultur- within exotic crops, and may be transferred to indig- al and coffee-growing areas (Matteson and Meltzer enous crops and vegetation. 1995; KNFU 1987). Some horticultural crops are propagated using Irrigation heightens the risk of waterborne dis- plastic trays, containers, sheeting, and pots. Most of ease and increases the health risk from contaminated these items are made from plastics produced from Preferential Trading Arrangements between Kenya and the EU 91 petroleum products. They are usually nonrecyclable, rates of soil erosion and runoff caused by inferior land and their production and disposal entail negative en- management practices and irrigation techniques. Veg- vironmental impacts. etable crops provide poor groundcover and soil-hold- ing capacity, and much cultivation takes place on Environmental impact of different crops erodible or marginal soils. The environmental risks The environmental impact of different crops depends of vegetable production are therefore high. Produc- partly on their natural characteristics and suitability tion of exotic vegetables relies on introduced seed, to the region in which they are being grown, as well raising concerns about the introduction of new pests as on their production technology and input use. Cer- and viruses via imported planting materials. Most tain crops are more likely to cause particular types of production takes place from seed, and there is little environmental damage than others. off-site propagation. Tea. Tea is grown without pesticides, except for Cut Flowers. Fifteen percent of all pesticides used herbicides such as paraquat and glyophosphate (Mat- in Kenya are applied in cut-flower production (Mat- teson and Meltzer 1995). Although it is frequently teson and Meltzer 1995). Flowers use soil sterilants, grown on steep slopes, tea provides good groundcov- fungicides, insecticides, and fumigants, creating a high er and soil-holding capacity. Its production therefore risk of applicator poisoning. Cut flowers are normal- entails little environmental risk. ly grown under drip, sprinkler, and furrow irrigation, After harvest green leaves are transported by which contributes to polluted wastewater off-site. road to local factories for processing. Many of these Most flowers are perennial and provide adequate tea roads are in poor condition and lie in ecologically groundcover, although poor land-management prac- fragile areas. Green leaf is processed by means of in- tices can contribute to soil erosion. The environmen- door drying, using a combination of electric fans and tal risks of cut-flower production are high. Because wood-fed dryers, which consume fossil fuel and fuel- production of cut flowers relies on introduced seed, wood from tea estates and the Forest Department's there is concern about the introduction of new pests exotic plantations. and viruses via imported planting materials. Coffee. Coffee producers use high levels of pesti- Maize and Beans. Maize and bean production is cides including insecticides, fungicides, and herbicides rarely irrigated and provides poor groundcover and that cause a high incidence of pesticide poisoning soil-holding capacity. (KNFU 1987). Although it is frequently grown on steep The maity. slopes, much coffee production takes place on bench Diy h anevrnetlhzrsfo terrces. Tche coffee brodushtrovdes relatvely go od bdairy production are chemical-related risks and meth- ane production. Cattle are regularly dipped in, or groundcover and soil-holding capacity. The main en- v agrocemi- sprayed with, acaricides (the most common organo- vironmental risk of coffee production is agrochemi- phosphates) to control tick-borne disease (Matteson cal pollution. After harvest coffee berries are transported by and Meltzer 1995). These solutions are hign y toxic, road to local drying centers, where they are ferment- but more rapidly biodegradable than many other dip- ed and cleaned in water and then sun-dried in the open ping chemicals. Cattle-dip solution is added to the air. The cleaning process leaves husk residues and natural drainage system without any treatment, and wastage, which is rarely disposed of properly and applicators may be at risk if cattle are sprayed. Meth- flows into nearby rivers and streams. ane is produced directly by the enteric fermentation Vegetables. Tomatoes and export vegetables rely in the intestines of ruminants and from methanogen- on high levels of fertilizers and are usually treated esis of stored manure. Dairy cattle in milk produce with fungicides and insecticides, creating a high risk almost one-third more methane than dairy cattle out of applicator poisoning. Vegetables are normally of milk (Adger and Brown 1994). Apart from their grown using flood and furrow irrigation. Wastewa- contribution to global warming, methane emissions ter from such irrigation transports agrochemical pol- are also of concern because the capability of the at- lutants from farms into local rivers and watercourses. mospheric sink and soil sink to absorb increased con- Soil-to-water chemical transfer is exacerbated by high centrations of the gas is reduced over time. 92 Trade, Global Policy, and the Environment 5. An Economic Analysis of Environmental A certain level of soil erosion can be tolerated Impacts without significant effects on productivity. Beyond Having examined the physical and environmental this level yields decline as a result of falling soil fertil- impacts of different crops in the previous section, we ity and nutrient levels. Crop yield losses are related provide here an analysis of the monetary valuation to topsoil loss, soil susceptibility and the level of in- of these impacts. The analysis is carried out primari- put use. Yield losses resulting from soil erosion for ly by estimating the production losses caused by en- different crops and areas have been estimated as part vironmental impacts, including the cost of replacing of this study. lost environmental functions and averting environ- mental damage. The choice of valuation technique depends on the kind of data available. Ideally, the It is possible to cost downstream sedimentation and estimate of changes of environmental values should changed waterflow in terms of the expenditure nec- be made in terms of willingness to pay. However, giv- essary to prevent these circumstances from occurring. en the scarcity of primary studies of environmental An effective way of minimizing soil erosion, loss of values for the agricultural sector in Kenya, we have vegetation cover, and resulting downstream effects had to limit this estimation to the method noted above. is to employ on-site conservation measures. Data are In most cases this means that the estimated cost of available on the soil conservation measures required damages is the lower boundary of the true cost. Never- for different regions and crops (Kassam and others theless, the estimates are useful and important for 1992) and the cost of carrying them out. These pre- analyzing the effects of policy changes, as will be seen ventive costs are used here as a proxy for the envi- below. Techniques for valuing specific environmen- ronmental cost of downstream sedimentation and tal impacts are outlined in table 6.7. changed waterflow. Table 6.7 Techniques for valuing major environmental impacts of agricultural production in Kenya Environmental effect Technique Method of calculation Decline in soil fertility and depth Effect on production Value of yield losses Sedimentation, water pollution Preventive cost Cost of on-site soil conservation Loss of biological resources Effect on production Value of lost resource uses Health hazards Human capital Cost of therapy and lost earmings Decline in soilfertility and depth Loss of biological resources The best-known predictor of surface erosion is the Uni- The cost of losing biological resources can be calculat- versal Soil Loss Equation (USLE). The USLE links soil ed by looking at the values foregone by agricultural loss per unit area with the erosive power of rain, run- land-use. Although data exist regarding the indirect, off, soil erodibility, vegetation cover, and cultivation option, and existence value of biological resources in methods. The USLE is used to estimate on-site losses Kenya (Brown 1989; Emerton 1992, 1994a; Moran 1994; in productivity resulting from declines in soil fertility Norton-Griffiths 1995), most refer to nonagricultural areas (national parks and forest reserves). Information And depthcaused byLE erosi en, developedforKe is available, however, on the direct use values of natu- A modified USLE has been developed for Ken- ral vegetation obtained by local households surround- ya (Kassam and others 1992). The model specifies that ed by areas of montane forest and dry, bushy scrub the annual soil loss A is a function (Emerton 1994b) the main regions where horticulture is carried out. The value of this foregone use forms one A(R,K,LS,C,M,P), (1) part of the cost of biological resource loss. Estimates of such losses have been made as part of this study. where R is the rainfall erodibility factor, K is the soil erodibility factor, LS is the length and steepness of slope Water pollution factor, C is the vegetation cover factor, M is the man- Agrochemicals pollute surface and groundwater agement factor, and P is the physical protection factor. sources. Most of the chemicals used in horticultural Preferential Trading Arrangements between Kenya and the EU 93 areas have limited mobility in soil and are transferred em Kenya (major horticultural areas) have shown that into surface water via soil particles through agricul- 25 percent of farmers have levels of pesticide concen- tural runoff. There is less leaching of agrochemical tration in their blood above those considered safe residues into groundwater in horticultural areas (Ed- (KNFU 1987). Most pesticide poisoning takes place wards 1970; FAO 1972,1979). Although data are avail- after spraying, the frequency of which varies by crop able concerning levels of chemical use in agriculture, and area. Data are available concerning the frequen- their residue and delivery rates, and guidelines for cy and type of pesticide application for different crops acceptable levels in water, it is difficult to link chem- and regions, medical charges for diagnosing and treat- ical pollution in water to given levels of health or ec- ing different types of pesticide poisoning, and for the osystem damage. Most chemicals are transferred from number of work days lost as a result of pesticide-re- soils to surface water. Limiting runoff and soil ero- lated ill health (HDCA). The cost of medical treatment sion through soil conservation measures minimizes and labor lost as a result of pesticide-related ill health downstream water pollution. Data are available on was estimated as part of this study. required soil-conservation measures for different re- gions and crops (Kassam and others 1992), as well as The environmental impacts of crop production the cost of implementing them. These preventive ex- penditures are used as a proxy for the environmental The level of environmental damage arising from ag- cost of agrochemical production. The cost of soil con- ricultural production depends on the natural charac- servation measures for different crops and areas have teristics of different crops, the nature of the areas in been estimated as part of this study. which they are grown, and the method of cultivation. Details of the environmental costs of producing dif- On-site health hazards ferent crops in different areas are available in Emer- ton and others (1997). With other factors held constant, Aside from water pollution, the most negative on-site tn and othe (199). tith othe fact ld constant effect of chemical use is health hazards to agricultur- an increase in the production of a particular agricul- al workers and local residents. The cost of on-site health hazards can be calculated by using a human- environmental cost. Some of these impacts are quan- capital approach, which takes into account days of tifiable. For quantifiable effects, the net environmen- labor lost due to illness and medical expenditure. Al- tal cost (NEC) per hectare of crop x is calculated as: though little quantitative information is available link- NEC(x) = yield losses from soil erosion + cost of soil ing irrigated agriculture and increased incidence of conservation + loss of natural resource use + cost of waterborne diseases, data describing the effects of medical treatment for on-site health hazards. The es- pesticide use on human health are available. timates are summarized in table 6.8. The presence of high levels of pesticide residues These figures have been used in the next section in blood is one indicator of human exposure to un- to estimate the effect of tariffs on land use and total safe levels of chemicals. Surveys in central and east- environmental damage. Table 6.8 Summary of net environmental costs of agricultural production ('000 Ksh/Ha) Maize French Tomatoes Coffee Tea Dairy & beans Statice Aistromeria Karelia beans Average Loitokitok 2.4 3.0 1.7 2.5 2.5 2.4 Mt. Kenya 7.0 2.0 1.4 3.4 6.4 8.5 8.5 7.1 7.1 5.7 Aberares 7.0 2.0 1.4 3.4 6.4 8.5 8.5 7.1 7.1 5.7 Juia 2.4 2.6 3.0 1.7 11.1 11.1 2.5 2.5 4.6 Limuru 2.0 2.1 3.0 1.4 16.2 16.2 2.1 2.1 5.6 Mwea 2.4 2.0 3.0 1.7 3.8 3.8 2.5 2.5 2.7 Machakos 5.4 3.3 4.7 6.9 6.9 5.5 5.5 5.5 Naivasha 2A 3.0 1.7 3.8 3.8 2.5 2.5 2.8 Average 3.8 2.2 1.6 3.2 3.2 8.4 8.4 4.0 4.0 94 Trade, Global Policy, and the Environment 6. Estimating the Impact of Preferential profitability of DP causes marginal profitability to de- Trading on Land Use, the Environment, and cline by that amount. The new equation for profitabil- other Indicators ity is then given by: Data in the previous section include average rates of profitability for each crop and zone in Kenya. Also Ili= xj -h/h,x (7) provided are some data on the amount of land in each zone devoted to each crop. The methodology for as- The new value of the intercept in the linear equation sessing the impact of a change in tariff rates on profit- allows us to compute the new land area allocated to ability is described below, the crop. The shift in profitability is shown in figure 6.1. Methodologyfor estimating changes in land use The above analysis makes a number of assump- tions that could be questioned. First, it assumes that It is assumed that the average profitability for each lands tusd in a uaythatenurses that p poi crop is obtained from a range which is linear, with land IS used m a way that ensures that private profit- the marginal profitability being zero in each case. Hence ability is equalized at the margin; that is, that there is the marginal profitability beig zerop in zneah case. gene no benefit for the marginal farmer in turning from the the marginal profitability for cop i in zone j is given present crop to a different crop. This may be disput- by P.. where ed if farmers are operating under conditions of poor 2 knowledge or are constrained in some way from l1, =ay -/Xijxii (2) changing their use of land. But as a first approxima- tion it is likely to be satisfactory. Second, the assump- aij> 0 fi i> 0 (3) tion of linearity may be questioned, but in the absence of data it is often taken as a first approximation. Third, Profitability varies with the amount of land X.. allo- it is also being assumed that land can be iremoved cated to crop i in zone j aij and /9 ij are coefficients of from, or brought into, use according to the economic the profitability function that need to be estimated. parameters.3 In some areas land is limited and it may This is carried out as shown below. not be possible to expand production. In other cases Given that the actual amount allocated is X ij, it decreasing production may not be feasible; for exam- is further assumed that the marginal profitability for ple, when no other sources of income are available. the last unit is zero. This allows for any returns that Some sensitivity analysis was carried out using the are necessary to recover a return on capital. Hence above model, under the constraint that the total one equation to determine a and A is given by: amount of land area be held constant. Shifts in pro- duction are then based on relative profitability. As - 1= Y X d (4) noted below, however, the broad conclusions of this The second equation to determine a and b is given by analysis remain unchanged even with that constraint. the fact that average profitability from crop i in zone Figure 6.1 Marginal profitability j is ni,. Hence integrating the marginal profitability rl(Ksh/ha) curve gives: x (5) C , n ij x ij (a=-ij- ,, X ij )X, profitability at original tariff or, profitability wi higher tariff rl a=ij - ij Xij/ 2 (6) Equations (4) and (6) allow us to determine the val- Hectares (X) ues of a and fi,,. The impact of changes in profita- bility is assessed by assuming that a change in Preferential Trading Arrangements between Kenya and the EU 95 Estimating the impact of environmental cost for horticultural crops. Data on land areas for the main internalization crops by region are also available for 1995, and are The model described above has been applied to data summarized in table 6.9, along with the implied val- for the following crops: Asian vegetables, French ues of act and /A ii4 Analysis has been carried out for beans, cut flowers, maize, coffee, and tea. Emerton three zones only, Central, Eastern, and Rift, since these and others (1997) report average profitability data for are the only zones where horticultural products are each of these crops; table 6.4 summarizes this data produced in significant quantities. Table 6.9 Average profitability, land areas, and implied values of a, and PI Profit Land used Implied Implied Crop (000 Ksh/ha) (000 ha) value value Central Eastern Rift Total Asian 94.6 0.086 1.320 1.454 2.86 189.2 -66.23 Vegetables French beans 46.8 0.174 3.191 2.335 5.81 154.2 -73.12 Cut flowers 10,291.8 0.040 0.649 0.616 1.33 20,583.6 -15,488.08 Maize 24.8 8.712 534.528 248.644 1,250.0 49.6 -0.04 Coffee 164.2 4.820 54.099 39.584 156.0 328.4 -2.11 Tea 107.3 2.875 28.672 27.215 97.0 214.6 -2.21 Notes: Profitability is the average for all zones. Separate profitability fgures were not available for each zone. Asian vegetables include only okra and karella. - Total land used is taken from Emerton and others (1997) and Kenya Hortkultural Industry Association data. Table 6.10 Impacts of full environmental cost internalization on land use Area at Environmental Area with fijul Reduction in present cost internalization Change in damage Region/crop (OOOha) (OOOKsh/ha) (OOha) area (OOOha) (OOOKsh) CENTRAL Asian veg. 0.086 9.80 0.081 -0.0045 44.35 French beans 0.174 6.90 0.161 -0.0129 89.32 Cut flowers 0.040 3.44 0.040 0.0000 0.15 Coffee 4.820 2.07 4.779 -0.0414 85.79 Tea 2.875 1.50 2.858 -0.0176 26.36 Maize 8.712 3.28 8.065 -0.6464 2,120.27 TOTAL 16.707 15.984 -0.7230 2,366.20 EASTERN Asian veg. 1.320 3.63 1.294 -0.0267 96.84 French beans 3.191 3.30 3.076 -0.1147 378.58 Cut flowers 0.649 3.80 0.648 0.0007 2.76 Coffee 54.099 2.60 53.547 -0.5522 1435.80 Tea 28.672 1.50 28.497 -0.1752 262.82 Maize 534.53 4.70 479.685 -54.8426 257,760.00 TOTAL 622.46 566.747 -55.7120 259,936.00 RIFT Asian veg. 1.454 3.25 1.428 -0.0265 86.27 French beans 2.335 3.05 2.257 -0.0777 237.01 Cut flowers 0.616 3.80 0.615 -0.0007 2.62 Coffee 39.584 2.40 39.204 -0.3800 912.03 Tea 27.215 1.50 27.049 -0.1663 249.47 Maize 248.644 1.75 237.803 - 10.8409 18,971.54 TOTAL 319.848 308.356 -11.4920 20,458.50 GRAND TOTAL 959.015 891.087 -67.927 282,730.70 96 Trade, Global Policy, and the Environment The results of applying the model are shown in Regionally the effects are likely to be most pro- table 6.10. Note that in table 6.8 the environmental nounced in areas where profitability has been most costs were at the district level. In tables 6.10 and 6.11 reduced. Since separate profitability figures were not we have summarized them into three regions, using available for the regions, this impact could not be eval- agricultural and district maps of Kenya. uated fully, although environmental costs appear to Table 6.9 shows that if estimated internal costs be highest in the Eastern region, where production of were internalized, about 68,000 hectares would be- horticultural crops is minimal. come unprofitable, mostly in maize production. Al- The employment effects of internalization are together this amounts to about 7 percent of the land difficult to gauge. Horticulture loses about 265 hect- area in the three regions that cultivate these crops. ares, amounting to about 210,000 person days, or The decrease in production of horticultural crops about 1,000 employees. The biggest impacts are in would be very small-about 500 hectares total- maize, where the employment effects are not quanti- mostly French beans. The production of cut flowers fied. would hardly be affected, as profitability is high enough to absorb most additional costs. Estimating the Impact of Changes in Preferences The impact of changes in environmental dam- Tables 6.11 and 6.12 provide material from which to age is shown in the last column of table 6.10: a reduc- analyze the impact of changes in tariffs. Table 6.11 tion in damages of Ksh. 282 million, or about US$5 imposes tariffs based on those applying to Mediter- million per annum. Most of the reduction would re- ranean Basin countries; table 6.12 shows the impact sult from using less land for maize. Internalizing the of introducing third-country tariffs as indicated in ta- environmental costs of the horticultural sector has a ble 6.6. The impact of such tariffs is not strong. Total small impact in terms of damages avoided. Howev- land cultivated declines by 530 hectares, with the big- er, it is important to note that this analysis does not gest reductions being French beans and Asian vege- reflect the effect of changes in practices on the level of tables. Export revenues are affected to the extent of envirorunental damages. Such effects could be impor- Ksh 48 million (US$0.9 million). At the same time, tant, but to achieve them would require a tax capable however, reduction in environmental damages is also of distinguishing among practices rather than crops. small (about Ksh 2 million, or US$40,000). The main Table 6.1 Impacts of introducing Mediterranean basin tariffs Reduction Area at Change in in Export Reduction Regionl Present Area Revenues in Damages Crop OOOHA. Tariff New Area 0OOha. OOOKsh OOOKsh CENTRAL Asianvegetables 0.09 11.39 0.08 0.01 539.21 51.39 French beans 0.17 5.1 1 0.16 0.01 925.67 66.30 Cut flowers 0.04 27090.50 0.04 0.00 10.25 4.45 Total 0.30 0.28 0.02 1475.13 122.15 EASTERN Asian vegetables 1.32 11.39 1.24 0.08 831 1.27 293.01 French beans 3.19 5.1 3.01 0.18 16975.00 581.52 Cut flowers 0.65 27090.50 0.63 0.02 167.01 80.13 Total 5.16 4.88 0.28 25453.29 954.65 RIFT Asianvegetables 1.45 11.39 1.37 0.09 9153.79 289.33 French beans 2.33 5.11 2.21 0.13 12420.53 393.26 Cut flowers 0.62 27090.50 0.60 0.02 158.52 76.06 Total 4.40 4.17 0.24 21732.85 758.64 GRAND TOTAL 9.86 9.33 0.53 48661.27 1835.45 Preferential Trading Arrangements between Kenya and the EU 97 Table 6.12 Impacts of introducing third country tariffs Reduction Area at Change in in Export Reduction Regionl Present Tariff Area Revenues in Damages Crop OOOHA. OOOKshlha New Area OOOha. OOOKsh OOOKsh CENTRAL Asian vegetables 0.09 27.94 0.07 0.01 1308.59 124.72 French beans 0.17 12.52 0.15 0.02 2253.27 161.40 Cut flowers 0.04 1186.89 0.04 0.00 36.43 15.82 Total 0.30 0.26 0.04 3598.29 301.94 EASTERN Asian vegetables 1.32 27.94 1.12 0.20 20170.29 711.10 French beans 3.19 12.52 2.76 0.43 41320.65 1415.53 Cut flowers 0.65 1186.89 0.57 0.07 593.46 284.73 Total 5.16 4.46 0.70 62084.41 2411.36 RIFT Asianvegetables 1.45 27.94 1.24 0.22 22214.97 702.16 French beans 2.33 12.52 2.02 0.31 30234.14 957.27 Cut flowers 0.62 1186.89 0.54 0.07 563.30 270.26 Total 4.40 3.80 0.60 53012.41 1929.70 Grand Total 9.86 8.52 1.34 118695.11 4643.00 reason is that producers take the cut in profits rather and thus the analysis is partial. The estimates of envi- than reduce output. Consequently, employment ef- ronmental damages are lower bounds because they fects would also be small, around 1,800 persons. are not based on willingness to pay for reduced dam- The impact of a shift to third-country tariffs is ages, but rather on loss of earnings or output. more pronounced, but still not large. The area under The economic analysis of preferential trading horticulture would decline by 1,340ha, and export was carried out on the assumption that land use is revenues would fall by Ksh 118 million (US$2 mil- adjusted so that marginal rates of return are equal- lion). About 4,800 jobs would be affected. ized across crops. Average rates vary since produc- tivity, technology, and other factors are also variable. 7. Conclusions Given the average profitability of different crops, a This paper has analyzed the impacts of horticultural linear function can be estimated for marginal profit- production in Kenya, focusing on how preferential ability. Using this function we can estimate changes trading arrangements affect the environment Horti- in land use following the removal of preferential tar- culture is an important and growing sector of the econ- iffs or the introduction of measures to internalize en- omy, contributing significantly to exports and vironmental costs. The analysis shows horticulture to employment. Although there is some concentration be largely insensitive to such measures. The main of activity in terms of ownership, much production is impact of internalizing environmental costs across all by smallholders, contributing notably to incomes crops would be to reduce maize production. As a among the poor. whole, horticultural production would drop by only At the same time, horticultural activities are en- about 500 hectares; French bean production would de- vironmentally damaging. The paper provides an anal- crease the most. The impact on the environment ysis of the impact on the environment, quantifying it would also be minimal; environmental damages to the maximum extent possible in monetary terms. would be reduced only slightly. The main explana- The figures show that, per hectare, damages to the tion for these phenomena is the high profitability of environment are higher for horticulture than for tea, most horticultural production. Given this situation, coffee, maize, or beans. The damages arise from var- the government could take measures to internalize the ious sources, however, and cannot all be monetized, cost of environmental damages, and thereby gener- 98 Trade, Global Policy, and the Environment ate revenue that could be used for environmental pro- 3. The regional data are constructed from (a) data on tection in agriculture. land classes in each region (I- III), (b) data on how Similar results emerge from the removal of pref- much of each crop is grown on each type of land erential tariffs. If Kenya were to be allotted Mediter- class and (c) the amount of overall production for ranean Basin tariff rates, horticulture would be each crop. Inevitably there are some approxima- affected similarly to the impact of environmental cost tions involved, as we do not have exact data for internalization. French beans and Asian vegetables each crop in each zone, but the figures should pro- would be the most affected. The imposition of third- vide a reasonable approximation for the purposes country tariffs would produce roughly double the of this analysis. impact of Mediterranean Basin tariffs. This analysis has not modeled the impact on References more vulnerable producers. It is almost certain that, should tariff preferences be removed, these groups Adger, N., and K. Brown. 1994. Land Use and the Causes would be the most heavily affected. All the adjust- of Global Warming. London: J. Wiley Press. ments are being made at the margin. Consequently, Anderson, K. 1992. "The Standard Welfare Econom- the impact on producers who are barely surviving ics of Policies Affecting Trade and Environment." would be strongest, and mneasures to protect these In K. Anderson and R. Blackhurst, eds., The Green- producers would be needed if trade preferences were ing of World Trade Issues. London: Harvester Wheatsheaf. to be altered. BBC (British Broadcasting Corporation). 1995. "Flower Notes from Lake Naivasha." Today Programme (Septem- ber 1). London. This paper draws extensively on "EU Trade Policy, Brown, G. 1989. "The Economic Value of Elephants." the Lome Convention, The Horticultural Sector and LEEC Paper 89-12. International Institute for En- the Environment in Kenya." This study was carried vironment and Development and London Envi- out with financial support from the European Com- ronmental Economics Centre. mission, DG8, under contract B-7-5040/94. In addi- Cruz, W., and R. Repetto. 1993. The Environmental Ef- tion to the above, the following have contributed to fects of Stabilization and Structural Adjustment Pro- the earlier study: Cuthbert Idawo, Daniel, Karanja, grams: 7The Pilippine Case. Washington D.C.: World Gem Kodhek, James Nyoroat, Francis Oguya and Resources Institute. Nicholas Dale. Edwards, G. 1970. Persistent Pesticides in the Environ- ment. New York: Butterworths. 1. Asian vegetables counted in exports here are Egerton and others. 1995. "Liberalization of Process- dudhi, okra, karella, chillies and others. There were ing and Marketing of Industrial Crops and Export no data for brinjals. Diversification." Egerton Policy Analysis Matrix 2. In Kenya, there is little slack land in areas with (PAM) Team. Report by Task Force 2 of the Agri- high potential . Increasing the total agricultural cultural Sector Review. Emerton, L. 1992. "An Economic Valuation of the En- land area implies expansion on to virgin land or vrnetlBnft fteMuFrs ope. into drier and more marginal areas. Much of the vimeo a KFnerseas D opent Agency settled agricultural land around Mount Kenya, the K)Forst Dertent Nairobi. Aberdares and Limuru has been excised from pro- ____ ,, F9r4at " envromenta macAsemn .1994a. "An Environmental Impact Assessment tected forest areas, which contain a diverse range and Environmental Valuation of the Proposed of plant and animal species. Further encroach- Aberdares Natural Resources Development ments will continue to erode natural forest areas. Project." Consultancy report. FAO Investment Hence the marginal environmental cost may be Centre, Rome. much higher than the average costs. This is a fac- . 1994b. "Summary of Current National Value tor we have not allowed for. It adds to the fact that of Use of Kenya's Indigenous Forests to Local the environmental costs are underestimated for Households." KIFCON, Overseas Development much of this exercise. Agency (UK), Forests Department, Nairobi. Preferential Trading Arrangements between Kenya and the EU 99 Emerton, L. and others. 1997. "EU Trade Policy, the Koning, A. 1993. "The Impact of European Union Poli- Lome Convention, the Horticultural Sector and the cies on Sub-Saharan Fruit and Vegetable Exports." Environment in Kenya." Report produced for EC Paper presented at Conference on Financial and DGVIII. Metroeconomica Ltd., Bath. Market Strategies for African Exporters of Horti- FAO (U.N. Food and Agriculture Organization). 1972. cultural Products in Harare, Zimbabwe. Effects ofIntensive Fertilizer Use on the Human Envi- Markandya, A. 1994. "Is Free Trade Compatible with ronment. FAO Soils Bulletin 16. FAO Soil Resources Sustainable Development?" UNCTAD Review (9): 22. Development and ConservationService, Land and Markandya, A. and others. 1997. "EU Trade Policy, Water Development Division; with the coopera- The Lome Convention, and The Third World En- Wate Deelomen Diisin; iththecooeravironmnent: A Study of Economy-Environment tion of the Swedish International Development Lirnkes Throug d MEconomy icanmLtd Authority. Linkages Through Trade.t Metroeconomica, Ltd., .1979. Ground Water Pollution. Rome: FAO. Bath. -1992. Fertilizer Handbook. Rome: FAO. Matteson, P., and M. Meltzer. 1995. Environmental and Economic Implications ofAgricultural Trade and Pro- Grossman, G.M., and A.B. Krueger. 1992. "Environ- motion Policies in Kenya: Pest and Pesticide Manage- mental Impacts of a North American Free Trade ment. EPAT Project working paper. Arlington, Va.: Agreement." In P. Gaber, ed., The US-Mexico Free Winrock International Environmental Alliance. Trade Agreement. Cambridge: MIT Press. Moran, D. 1994. "Contingent Valuation and Biodi- Harris, S.S. 1992. "Kenya Horticultural Subsector Sur- versity: Measuring the User Surplus of Kenyan vey." Report by Development Alternatives Inc. for Protected Areas." Working Paper GEC 94-00. USAID Kenya Exports Development Supports CSERGE, University College London and Univer- (KEDS) Project. Nairobi: U.S. Agency for Interna- sity of East Anglia. tional Development. Munansinghe, M., and W. Cruz. 1995. Economywide Jaffee, S. 1991. "Marketing Africa's Horticultural Ex- Policies and the Environment: Lessons from Experi- ports: A Transaction Cost Perspective." Paper pre- ence. World Bank Environmental Paper 10. Wash- sented at Workshop of Fresh Fruit and Vegetables ington, D.C.: World Bank. System, University of California, Santa Cruz. Mwanthi, M., and V. Kimani. n.d. "Agro-chemicals: A Kassam, A., and others. 1992. "Agro-Ecological Land Potential Hazard to Kenyan Smallholder Farmers." Resources Assessment for Agricultural Develop- Paper presented to the IDRC Pesticides Sympo- ment Planning: A Case Study of Kenya." Mimeo. sium. Ottawa, Ontario: International Develop- FAO/IIASA. ment Research Council. Kenya, Horticultural Development Crops Authority Naitore, J. 1995. "Lake is Choking to Death." Sunday (HCDA). n.d. Various production and export data. Standard (uly 23). . Ministry of Agriculture. 1983. Farm Manage- National Environmental Secretariat. 1981. Kenya: Re- ment Handbook. Nairobi: Ministry of Agriculture, port on Environment and Development. Nairobi: GTZ. United Nations Environment Programme. Various years. Economic Survey. Nairobi. Norton-Griffiths, M. 1995. "The Opportunity Cost of Biodiversity Conservation in Kenya." Ecological . National Environmental Secretariat. n.d. Van- Economics 12: 125-39. ous ecological, soil, water, agrochemical, and en- Partow, H. 1992. "Pesticides: Agents of Exploitation, vironmental data. the Case of Kenya." Mimeo. Environmental Liai- Kibata, G. 1993. "Implementation of Integrated Pest son Centre International, Nairobi. Management in Kenya." Background paper for the Reed, D., ed. 1992. Structural Adjustment and the Envi- IPM Working Group East/Central/Southern Af- ronment. Oxford: Westview Press. rica. IPM Implementation Workshop, Harare, . 1996. Structural Adjustment, the Environment, Zimbabwe. and Sustainable Development. London: Earthscan. KNFU (Kenya National Farmers Union). 1987. "Pes- Repetto, R. 1993. "Trade and Environment Policies: ticide Distribution and Use in Kiambu and Achieving Complementarities and Avoiding Con- Nakuru." Report prepared by KAMFOR for flicts." WRI Issues and Ideas. World Resources In- KNFU. Nairobi: KNFU. stitute, Washington, D.C. 100 Trade, Global Policy, and the Environment Tyler, G.J. 1996. "The Contribution of Community UNEP (United Nations Environment Programme). Based Wildlife to Tourism in Botswana." Mimeo. 1987. Kenya: State of the Environment Report. University of Oxford. Nairobi: UNEP. Unemo, L. 1993. "Botswana, Environmental Impact of Government Policies and External Shocks ih Botswana - A CGE Approach." Working Paper 26, Beijer Institute, Stockholm. Chapter 7 Fuel Prices, Woodlands and Woodfuel Markets in the Sahel An integrated economic-ecological model Kenneth M. Chomitz, Charles Griffiths, and Jyotsna Puri 1. Introduction trade, manufacture, procurement, distribution, and In the Sahel as in many parts of the world, there taxation. Trade liberalization could therefore result in has long been concern that overexploitation of substantial price changes. woodlands for woodfuels leads to a vicious circle Other policies could also modulate the relation- of environmental degradation, rising fuel prices, and ship between fuel markets and the environment by increasing poverty (Schreiber and Cleaver 1994). In affecting supply or demand. Niger has encouraged this stylized story, urban demand for charcoal and conversion of open access areas to community man- fuelwood leads to ever-widening circles of periurban agement while taxing woodfuels produced from un- deforestation, as open-access areas are 'mined' for en- managed areas. In many places, attempts have been ergy. Urban fuel costs increase as the supplies must made to promote or subsidize high-efficiency wood- be brought in from ever more distant areas. stoves, and kerosene stoves, as means of reducing In principle, this dynamic is influenced by trade energy demand. Finally, policies affecting the cost and and fiscal policies related to modern fuels. Kerosene competitiveness of transport could have a large effect and LPG can substitute for woodfuels. If cheap on the supply price of woodfuels, and on the distri- enough and reliably distributed, they could act as a bution of benefits from wood production. backstop technology, limiting price rises of woodfu- While these policies all have plausible impacts els and therefore arresting the expansion of defores- on fuel prices and land degradation, the actual mag- tation. On the other hand, higher motor fuel prices nitude of the impacts depends on the interaction of could accelerate the transition away from woodfuels. woodland ecology with the behavior of consumers, Because the urban wholesale price of woodfuels de- producers, and transporters. How quick are consum- pends mostly on transport cost, higher motor fuel ers to substitute kerosene for charcoal? When do pro- prices restrict the exploitable radius around cities, ducers switch from wood to charcoal? How fast does boosting woodfuel prices and increasing the likeli- the frontier of exploitation expand as prices go up? hood of a transition to a backstop fuel. How fast do exploited woodlands regrow? Are wood- Prices of kerosene, gasoline, and other fuels vary lands being cleared for agricultural expansion in any considerably between countries. This variation is part- case? Policy impacts depend on the interaction of a ly due to geography but mostly due to policy. Some complex system of actors, and it is difficult to make a countries subsidize kerosene with the explicit goal of priori impact predictions. reducing pressures on forests and woodlands. In oth- This paper demonstrates a simulation model for ers, though, modern fuels are well above world pric- analyzing the interrelations between urban fuel mar- es as a result of policies affecting petroleum product kets and rural land cover, building on an earlier ef- 101 102 Trade, Global Policy, and the Environment fort (Chomitz and Griffiths 1997) and similar in spirit Studies in the early 1980's, for instance, warned that to Hofstad (1997). Like those models, the one present- woodlands were being rapidly depleted, with serious ed here uses a von Thiinen-style supply function for consequences (de Montalembert and Clement 1983). woodfuels. Unlike those models, the present one links Loss of woodlands would result in high fuel prices in the supply function to a demand function which al- urban areas. It could also increase fuel-gathering time lows substitution among fuels, and between fuel and in rural areas, and reduce soil fertility as rural dwell- other goods. The model presented here consists of an ers turned to dung as a fuel source (Schreiber and urban fuel demand module; a spatially disaggregate Cleaver 1994). Cattle would lose crucial sources of woodfuel supply module employing GIS data on land browse during the long dry season. Loss of wood- cover and biomass; and a module describing wood- land cover, on some types of soils, can lead to biolog- land regrowth in response to extraction. The modules, ical processes which seal the soil and result in perhaps which can be elaborated to any desired degree of de- irreversible desertization (Le Houerou 1989, p. 94). tail, are integrated through a mechanism which solves More recently, a number of authors have ques- for equilibrium prices across the landscape. tioned the degree to which this "conventional wis- The model is used to simulate the ecological and dom" describes the Sahel and other areas (Nichol 1988; economic consequences of alternative fuel pricing sit- Leach and Mearns 1988; Benjaminsen 1993; Leach and uations in a generic Sahelian setting. The principal Fairhead 1994; Dewees 1995; RPTES 1995; Wilton question is the extent to which policy or market-driv- 1996; Foley and others 1997). At issue are such things en changes in the relative prices of modem and tradi- as: tional fuels can affect woodland degradation, rural * The rate at which periurban deforestation is tak- incomes, and urban fuel consumption. The model also ing place serves to illustrate an approach for linking plot-based * The standing stock of biomass land use models with sectoral or macroeconomic mod- * The relative roles of agricultural expansion, els, which may be useful for examining deforestation woodfuel exploitation, and climatic fluctuations and land use change dynamics in a variety of settings. in causing deforestation A benefit of constructing such a model is that * The price-elasticity of demand for fuels the need to specify parameters highlights areas of un- * The degree to which woodland degradation is ir- certainty. Those areas are many, despite years of at- reversible tention to these issues. An underlying theme of the * The degree to which woodfuels are supplied from paper is that our understanding of policy-environ- dead trees, agricultural land clearing, or as a co- ment-poverty linkages depends a great deal on un- product of farming. derstanding those basic parameters. Nonetheless, The newer view suggests that agricultural land sensitivity analysis allows some conclusions to be clearing, rather than woodfuel exploitation, plays a drawn. major role in loss of woodlands, that stocks and The plan of the paper is as follows. The next sec- growth rates might be higher than had been supposed, tion reviews background material on the nature of the and that land-clearing might be slower. This is essen- fuel market-woodland degradation linkage, the rela- tially an empirical debate about the magnitudes of tion between policies and fuel prices, and the impact various processes and their impacts. To a large extent of fuel prices on woodfuel consumption. The third it can be resolved empirically. Large scale time-series section describes the model and data. Results and con- studies of land use change, such as those currently clusions follow. being undertaken in Senegal' and Nigeria2, will help to resolve some of the questions. Simulation models 2. Background such as the one presented here can use these studies' Fuelwood demand and woodland degradation - an findings to better describe the complex interactions environmental and economic problem of uncertain of agricultural, transport, and fuel markets. In partic- dimensions ular, they can disentangle the effects of woodfuel-driv- Concern about the fuelwood-woodland degradation en degradation from other causes of land cover linkage - especially in the Sahel - is long-standing. change. Fuel Prices, Woodlands, and Woodfuel Markets in the Sahel 103 Fuel price policies cy woodstoves. The underlying assumption has been Barnes (1992) documents wide variation in kerosene that energy demand is inelastic, so that decreasing the prices between countries, ranging from about $0.16/ effective price of energy (by providing more cooking gallon in Mexico, Peru, and Burma to more than $2.00 services per kilogram of wood) results in lower, not in Malawi and Burundi (mean data for late 70s and higher, wood consumption. In practice, the stoves early aws). Similarly, gasoline prices vary widely. A have often proved unattractive to consumers, and 1997 survey found that gasoline pump prices in Sub- where adopted yielded at most modest fuel savings4 Saharan Africa ranged from $0.13/liter in Nigeria to (Barnes, Openshaw, Smith and van der Plas 1993; about $0.70 in Niger, Burkina Faso, Chad, and Mali, Wilton 1996). to $1.14 in Uganda.3 On the supply side, taxes on fuelwood and char- Fiscal policy drives some of this variation. Indo-' coal have often been levied both for revenue and in nesia and Nigeria subsidize kerosene, in part with the order to conserve the resource. However, collection rationale of reducing pressure on forest resources. rates have generally been low. More recently, atten- Though shortages at the official rate are common in tion has focused on the dual approach of securing some subsidizing countries, these subsidies can have community tenure over woodlands while levying a an impact on price. ESMAP (1993a) for instance re- tax on woodfuels coming from remaining open-access ports that kerosene prices in northern Nigeria are only areas (see Foley and others 1997, for a detailed de- one-quarter of economic prices. Other countries, such scription). As demonstrated in the Niger Household as Burkina Faso and Mauritania, have taxed kerosene. Energy Project, this involves setting up approved ru- Many countries tax motor fuels as a means of cross- ral markets linked to communities that are granted subsidizing household fuels, raising general revenues, woodland tenure rights, and agree to harvest quotas or conserving foreign exchange (Gately and Streifel and approved harvesting techniques. Management is 1997). Petroleum products in general account for 40 thought to increase the sustainable yield of wood, percent of fiscal revenues in sub-Saharan Africa and though it involves forgoing the profits of 'mining' all absorb 20 percent to 35 percent of foreign exchange salable wood. Purchasers from these markets pay a earnings (Schloss 1993). small levy which supports community development. In many African countries, the prices of petroleum Those who gather woodfuel from unmanaged areas, products are driven up by a variety of policies beside and therefore do not have proof of purchase from an taxation. A comprehensive study of sub-Saharan Afri- approved market, must pay a higher tax, enforced at ca (Cuneo e Associati 1993; Mayorga-Alba 1993) found roadside checkpoints. This tax is Pigouvian in intent, three sources of inefficiency. Procurement practices are correcting for the dynamic costs of woodcutting hampered by high costs, poor bidding practices, and (Wiedenmann 1991)-that is, the open-access wood- distorted pricing policies. Refining costs are high be- cutter's lack of concern for his impact on future har- cause protected local refiners operate small, antiquat- vests. It is also intended to encourage transporters to ed, underutilized equipment. Inland distribution costs buy from rural markets rather than harvest from open- are high because of lack of competition, poor infrastruc- access areas or steal from managed areas. Its effec- ture, and distribution systems which do not take ad- tiveness in doing so will depend on the spatial vantage of opportunities for cross-national economies distribution (and hence locational rents) of managed of scope and scale. On average, rectification of these and unmanaged producers, on their relative produc- inefficiencies could realize savings of 13 percent of the tion levels, and on the demand elasticity for woodfu- cost of petroleum products (Mayorga-Alba 1993). The els (Chomitz and Griffiths 1997). savings would be higher in some countries, with po- tential savings of about 25 percent in Chad and Niger, Price elasticities and subsitution for example (Cuneo e Associati 1992, p. 15). Policies that affect fuel prices will affect the fuel mar- ket-deforestation nexus only if consumers are price- sensitive in their choice of fuels or their overall There has long been interest in reducing woodfuel consumption of energy. Own-price and cross-price demand by promoting the diffusion of high-efficien- elasticities have been difficult to measure in cross-sec- 104 Trade, Global Policy, and the Environment tional household data due to lack of variation (Barnes The model is implemented as a GAUSS program, and others 1998). There are relatively few large-sam- which integrates supply, demand, regrowth, and ac- ple household studies. Kidane (1991) finds relatively counting procedures and uses a grid search to solve inelastic demand for fuelwood in urban Ethiopia for the market-clearing values of poc and pow. The pro- (own-price elasticity = -.37), and no cross-price effects gram is available from the authors on request. with agricultural residues or modem fuels. Howev- er, experience shows that large changes in the price Supply and change in biomass ratio of alternative fuels can bring about rapid, mas- sive shifts between fuels. Barnes and others (1998, Using data in a GIS (geographic information system), p. 46-7) describe how such shifts occurred inrelatively the landscape is gridded into 1 square kilometer (100 short periods in Thailand, due to the imposition of a hectare) cells, and supply is modeled at each cell. Cells ban on logging, and in Zambia, due to a temporary are classified according to four land cover types: grass- interruption of charcoal supply. They suggest that land, bushland, woodland, and other (including agri- cross-national analysis of prices, market availability, culture, scrubland, inundated grasslands, and and consumption patterns can take advantage of con- settlements). The first three types may be in two ten- siderable cross-variation in relative fuel prices and ure categories: unmanaged (open-access) and man- give a more accurate picture of consumer price-re- aged. The land use types differ in maximum biomass sponsiveness and fuel substitution behavior. Barnes and in biomass growth rates. and others report energy choice and conditional en- Each cell has an associated transport cost for ergy demand estimates using a data set with obser- supplying the market. GIS methods were used to cal- vations on each of five income classes for 45 cities. culate the cheapest path to the market, allowing for They find substantial own-price and cross-price fuel relative differences in travel cost on and off roads (see elasticities, as well as strong access effects. figure). The terrain-adjusted distance 1 to market for cell j was scaled by a unit cost per ton-mile of trans- 3. Model port to yield the cost per ton of transporting fuel from A spatial price equilibrum model cell j to the market. The 'farmgate' price pi, of fuel i at The model assumes that all fuel consumption takes cell j is therefore: place at a single urban market. There are four fuels - wood, charcoal, kerosene, and LPG -used primarily -jz = max[O, pm-retail markup - (unit cost for cooking. Kerosene and LPG are imported, and their of transport) . distance) prices are set exogenously. Wood and charcoal are Suppliers in cell j may be resident (if the cell is supplied from surrounding woodland areas. Supply managed), or not (if it is open-access). Supply is of- of these fuels does not compete with the supply of fered from the cell if at least one woodfuel has a pos- biomass to rural consumers, who are assumed to ob- itive 'farmgate' price. Charcoal is offered if pj>pwj, tain energy from crop residues and urumarketable and p9>production costy wood is offered if p,j> p9 and wood sources. p> production cost7,, where production cost includes Market equilibrium is determined annually the costs of cutting and charcoaling. Although char- through aggregate demand and supply equations. coal's price is generally higher (because it has a high- With time subscripts suppressed: er energy content per gram), a kilogram of wood yields only 120 to 200 grams of charcoal. The result is a von QDi = QDi(poepoe,pok0pol) fi=w,c,k,U; Thtinen landscape in which areas near the market QSi QSi(pOepO.) [i=, c]; supply wood, because avoidance of the charcoal con- QSi = QDi [i=w,c,k,lJ, version losses outweighs the higher transport cost per where: unit of useful energy; areas further out supply char- coal; and those beyond the economic frontier supply iindexes the four fuels (wood, charcoal, kerosene, ntigothceramrk. LPG) nothing to the central market. The amount of biomass cut, converted, and sold Pot is the price at the market center, expressed in terms depends on the land cover type and tenure in the cell. of end-use energy units. Managed areas are assumed to be bound by a man- Fuel Prices, Woodlands, and Woodfuel Markets in the Sahel 105 agement plan, and market only the cell's sustainable pressed in joules) allocated to each of the three fuels; yield. Exploiters of open-access areas are arbitrarily and then the total quantity of energy demanded, giv- assumed to cut a fixed fraction of available biomass en the fuel-share-weighted effective price of energy. each year in economically exploitable cells. The shares equations use a multinomial logit-like These assumptions define the cell-specific sup- specification which ensures that shares are bounded ply of woodfuels i: between 0 and 1, and sum to 1; captures the intuition that shares may change rapidly, in a logistic fashion, qsJ1 = qs(p],biomassf land coverf,tenured as two fuels approach the same effective price per joule; and yet allows for asymptotic non-zero shares Biomass in the cell is decremented by the amount for 'expensive' forms of energy, reflecting specialized, supplied. A growth function then increments the bio- less-substitutable uses of each energy form. The shares mass by an amount which depends on the land cover of energy consumed in the forms of wood, charcoal, and tenure, with more rapid regrowth in managed kerosene, and LPG are given as: areas.5 Introducing time subscripts: s1= exp(xj)/1exp(x). (1) biomass,, =biomass, t1 where +growth (land cover type, tenure type) xi=~a1+ 3l/ny + Zj*khrosene i* In(p o0p 0 (2) where e is the charcoaling efficiency (charcoal yield (with the coefficients normalized to zero for i=kerosene) as a proportion of wood input). y is mean income per capita The aggregrate supply is the sum over all cells j: and prices are expressed per unit of effective energy (end-use, allowing for stove efficiency). QSi(p0,pO;unit cost of transport)=5I it In the simulations reported here, total demand The unit transport cost depends on the cost of motor for end-use energy is taken as price-inelastic. How- fuel, which is exogenously given. ever, the model allows total demand for end-use en- Data and parameters for the supply model were ergy QED to be determined as: drawn from information on Chad (ESMAP 1993a; van der Plas and Gutierrez 1996). (There is however no QED(y,p) = kyWo(price index)02, intent to analyze the Chadian situation, since the sce- where narios to be examined deliberately impose simplify- k o0 and 02 are parameters ing assumptions in order to sharpen the results.) The biomass data are based on a land-cover clas- p is the vector of fuel prices sification of 20-meter resolution remote sensing data for an area of approximately 20,000 square kilometers, price index =r Si 'Poi representing the woodfuel catchment area for N'Djamena, the capital of Chad (see Quarmby and oth- The shares function was constructed using esti- ers 1996). The classification distinguishes several types mates based on the 45-city cross-national data de- of woodland and savanna. Applying biomass density scribed in Barnes and others (1998). The source data factors, we aggregated this information into a 1 km grid were derived from surveys of 20,000 households in of biomass density. Information on road and village 12 countries in 1988. We used city-level aggregate locations was used to impute relative travel costs from data. The sample was restricted to cities where all four each cell to the central market, allowing for differential fuels were consumed.6 From (1) and (2), on and off road travel costs. Further details on param- eters are described in Chomitz and Griffiths (1997). In (q/qker0s) Xi where qkene is the ratio of consumption of fuel i Demand to the reference fuel, kerosene. This equation was es- A pragmatic two-stage aggregate demand function timated for each fuel, and measuring quantity in models first the share of end-use energy demand (ex- joules. The coefficients were then used to construct 106 Trade, Global Policy, and the Environment (1), after correcting the constants so that predicted attached to the use of wood caused by for instance, ratios were in terms of end-use energy. The estimat- the emission of smoke and the awkwardness of han- ed equations are shown in table 7.1. dling and storing wood. Figure 7.3 shows how ener- The table shows negative own-price elasticities gy shares react to changes in charcoal price. Holding as expected. For instance, equation 2 shows an elas- other prices constant, there is a strong shift from char- ticity of substitution of charcoal with respect to kero- coal towards wood as charcoal prices increase, but rel- sene of 4, significant at the 1 percent level. The cross atively little substitution into kerosene. price elasticities have to interpreted in light of the sys- tem as a whole and these are best seen graphically. 4. Results Figure 7.1 shows the actual versus predicted share of kerosene in end-use energy (among kerosene, A baseline scenario was run for 10 years. Kerosene LPG, wood, and charcoal), based on equation 1, show- and LPG start with a negligible share; charcoal is the ing that this somewhat convoluted function represents predominant fuel, but wood is still widely used. To the data reasonably well. Figure 7.2 shows the pre- focus attention on the mutual impact of prices and dicted shares of end-use energy implied by equation resource depletion, income is held constant over the 1, as a function of changes in the price of kerosene.7 simulation period, and there is no allowance for con- As kerosene prices decrease, the graph shows a strong version of land to agriculture (which may increase shift into kerosene from wood, and to a lesser extent short term supply but decrease long term supply of from charcoal. The shift takes place somewhat below woodfuels). The model also assumes that all wood- the point where kerosene is price-equivalent (in terms lands are open-access rather than managed resourc- of end-use energy), meaning that some disutility is es. Population grows at 4 percent annually. The Table 7.1 Estimated relative fuel demands -Equation Obs R-sq F P InQ_wk 23 0.8369 23.08369 0 InQ_ck 23 0.7304 12.19028 0.0001 InQ_Ik 23 0.7602 14.26686 0 Equation I In (wood/kerosene) Quantity (oules) Coefficient t-statistic P> It! In (charcoal/kerosene)price -2.36684 -4.381 0 In (Ipg/kerosene) price -2.090107 -2.824 0.011 In (wood/kerosene) price -0.754981 - 1.648 0.117 In (income per capita) 1.875554 4.099 0.001 constant -5.298216 -3.07 0.007 Equation 2 In (charcoal/Kerosene) Quantity Gjoules) Coefficient t-statistic P> t In (charcoal/kerosene)price -4.024249 -4.51 0 In (Ipglkerosene) price -2.561514 -2.096 0.051 In (wood/kerosene) price 1.636667 2.163 0.044 In (income per capita) 3.492979 4.622 0 constant -10.36086 -3.634 0.002 Equation 3 In (LPG/Kerosene) Quantity (oules) Coefficient t-statistic P> It I In (charcoal/kerosene)price -0.626344 -0.807 0.43 In (Ipg/kerosene) price -4.032859 -3.795 0.001 In (wood/kerosene) price 0.4145364 0.63 0.536 In (income per capita) 3.65188 5.558 0 constant -10.93914 -4.414 0 Note: Price and quantity ratios are based on energy content of fuels and arm not adjusted for end-use efficiency. Fuel Prices, Woodlands, and Woodfuel Markets in the Sahel 107 Figure 7.1 Actual vs. predicted kerosene share Actual kerosene share 0 0 0 0 0 .5 - 0 0 0 O cP 0 co o co ~~~ o * ° -8 °b ° 0I o .5 Predicted kerosene share Figure 7.2 End-use energy shares as a function of kerosene price 100 90 80 60 K Kerosene fectly inelastic.5market, moves out to 13 Charcoal 40 30 UWood 20 10 o o) 0 0) 0D 0 0 0) o) o 0 0) 0I 0 0 0D '- C" cf J LO co I- 0) Kerosene price ($/ton) overall demand for end-use energy is taken to be per- wood production, initially 76 kilometers from the fectly inelastic, market, moves out to 130 kilometers. The charcoal The baseline run shows a nearly constant rate of frontier moves from 106 to 178 kilometers. The envi- depletion of biomass (see Appendix table A7.1). Stocks ronmental impact is evident from a comparison of the in the study area are run down from 7.9 million tons panels of the maps at the end of this section, showing to 5.6 million tons over a decade. The frontier of fuel- the virtual elimination of close-in biomass. 108 Trade, Global Policy, and the Environment Figure 7.3 End-use energy shares as a function of charcoal prices 100 90 80 tio t60 [3 IPG u 50 Kerosene . 1* Charcoal IO 40a U Wood L in 30 20 10 0 50 100 150 200 250 300 350 400 450 500 Charcoal price ($Iton) The retail wood price increases from US $45/ton lower. Charcoal and wood prices are little changed to $55, while charcoal increases more modestly, from relative to the base case, as are total production costs. $146/ton to $159.8 These price increases are not Total supply costs, and producer rents, are about the enough to induce much substitution into kerosene or same as in the base case. LPG; while absolute consumption increases faster In the second comparison run, the cost per ton- than population growth, the share of modem fuels mile of transport was doubled from its initial value of remains negligible throughout. Total cost of energy $0.18/ton-km, representing an extremely large in- supply (including production, transportation, and crease in the price of motor fuel. The effect is a jump retail distribution) increases from $10.7 million to in the prices of both wood and charcoal, followed by $17.5 million. Locational rents (producer surplus) in- a more rapid rise than in the base case. However the crease from $0.56 million to $1.09 million. price rises for woodfuels are inadequate to induce a Two comparison runs (Appendix tables A7.2 and significant shift into kerosene. Thus biomass loss is A7.3) examined the effect of substantial changes in similar to the base case. It is slightly greater because the costs of modern fuels. In one, the price of kero- the shift from wood to charcoal results in more deple- sene was reduced by 40 percent, to a level approxi- tion. The main difference with the base case is that mating the current lowest reported retail prices (net the production cost and producer surplus increase of tax) in the world. As figure 7.2 suggests, even this sharply. By the end of the supply period, the total fuel was not sufficient to make a qualitative difference. The supply costs are about 21 percent higher than in the immediate impact was to boost kerosene consump- base case. (The welfare interpretation of this rise de- tion twelve-fold, but by the end of the period kero- pends on whether or not it was due to an increase in sene's share of end-use energy was still less than 6 gasoline taxes.) The final-year locational rents increase percent. Relative to the base case, retained biomass from $1.09 million to $1.97 million, as closer-in har- was just 2 percent higher and the charcoal frontier just vesters benefit from higher prices. This is a transfer 3 kilometers closer-in. Part of the reason for the lack from consumers, who in the final year spend 25 per- of impact, though, is that the shift into kerosene has cent more than in the base case. come from wood, rather than charcoal (which uses more biomass). At the end of the simulation period, 5. Discussion and Conclusions wood consumption is 12 percent lower than the base In sum, gross perturbations of kerosene and gasoline case, while charcoal consumption is only 2 percent prices had almost no impact on the rate of woodland Fuel Prices, Woodlands, and Woodfuel Markets in the Sahel 109 degradation. This is not surprising in light of the de- the impact of agricultural conversion on short and mand functions, which imply that significant transi- long term woodfuel supply, and the rate at which un- tions take place only at substantially subsidized managed woodlands regenerate. We also lack infor- kerosene prices or when woodfuels are brought in mation on the extent to which open-access harvesters from much greater distances than shown here. It sug- secure locational rents, rather than incurring increas- gests that reductions in policy distortions, while de- ing search and harvest costs as woodland densities sirable on many grounds, may often have less decline. envirornental impact than might be hoped. Reduc- However, we have provided a framework into tions in gasoline taxes will have distributional effects, which this information can be incorporated, and which shifting rents away from harvesters -who could be allows sensitivity analysis for parameters which are Biomass at the beginning of simulations Biomass at the end of the base run (tons/km2) (tons/kM2) t 5p | ~~~~~~~200-1000 1 000-max OM Market Roads urban-based transporters, or rural villagers - to urban difficult to gather. The framework also provides an consumers. example of how macro and sectoral economic process- These conclusions need to be qualified with an es can be linked to spatially explicit environmental admission of the tentative nature of many of the un- conditions. The full potential of this framework will derlying parameters. The demand function is con- be realized in applications which require spatial anal- structed from cross-country data and does not yet ysis, including impacts of fuel markets on carbon incorporate any degree of price elasticity for overall emissions, and impacts of alternative geographical energy demand. Other important unknowns include patterns of establishing managed woodlands. 0 Appendix Table A7. I Base run Charcoal Quantity Quantity Quantity Quantity Locational Wood Charcoal Year Total biomass Wood price price wood charcoal kerosene LPG rents Total supply cost frontier frontier Tons W/Ton $/Ton Tons Tons Tons Tons $ $ Kms Kms 1 7921152. 44.80 146.22 88,759 45,711 136 44 559,260 10,214,746 76.18 106.32 2 7652906. 46.32 147.96 89,022 48,713 144 47 609,382 10,843,023 84.67 115.97 3 7386834. 47.78 149.64 89,517 51,755 152 50 659,503 11,490,205 92.73 125.31 4 7122593. 49.01 151.05 90,502 54,751 161 53 691,545 12,149,373 99.57 133.14 5 6862033. 50.16 152.34 91,685 57,810 169 56 731,286 12,817,601 105.97 140.34 6 6601790. 51.18 153.61 93,261 60,867 178 59 777,982 13,495,979 111.65 147.39 7 6341789. 52.12 155.03 95,205 63,933 188 63 843,772 14,189,699 116.89 155.24 8 6082824. 52.99 156.45 97,436 67,045 199 66 918,489 14,902,830 121.71 163.12 9 5824806. 53.79 157.87 99,936 70,216 211 70 1,001,255 15,638,587 126.16 171.02 10 5563478. 54.55 159.30 102,661 73,467 224 75 1,092,535 16,400,328 130.35 178.96 Note: Kerosene price - $432.50/ton; LPG price - $1,250/ton. Appendix Table A7.2 Reduced kerosene price run Charcoal Quantity Quantity Quantity Quantity Locational Wood Charcoal Year Total Biomass Wood price price wood charcoal kerosene LPG rents Total supply cost frontier frontier Tons $/Ton $/Ton Tons Tons Tons Tons $ $ Kms Kms 1 7,932,459 44.32 146.02 79,148 45,548 1,723 64 509,235 10,175,872 73.54 105.18 2 7,672,860 45.84 147.72 79,164 48,442 1,817 68 557,679 10,783,045 81.98 114.66 3 7,417,216 47.24 149.35 79,513 51,330 1,916 72 600,753 11,407,767 89.73 123.71 4 7,162,684 48.47 150.76 80,203 54,225 2,015 76 633,560 12,046,243 96.60 131.53 5 6,911,731 49.57 152.00 81,196 57,142 2,115 80 665,996 12,693,503 102.72 138.45 6 6,662,235 50.54 153.23 82,553 60,050 2,224 84 704,938 13,350,908 108.11 145.24 7 6,413,085 51.45 154.54 84,172 62,980 2,345 89 761,075 14,021,629 113.12 152.52 8 6,165,290 52.26 155.89 86,101 65,924 2,478 94 825,741 14,710,839 117.63 160.02 9 5,917,597 52.99 157.24 88,298 68,905 2,621 99 896,797 15,421,240 121.71 167.54 10 5,669,309 53.68 158.62 90,703 71,948 2,775 105 976,518 16,156,170 125.52 175.18 Note: Kerosene price - $259.50/ton; LPG price - $1250/ton. Appendix Table A7.3 Increased transport cost Total Charcoal Quantity Quantity Quantity Quantity Locational Total supply Wood Charcoal Year Biomass Wood price price wood charcoal kerosene LPG rents cost frontier frontier Tons $/Ton $/Ton Tons Tons Tons Tons $ $ Kms Kms 1 7,908,985 57.03 165.87 70,184 52,258 174 58 1,028,287 11,790,228 72.09 107.75 2 7,629,641 59.75 169.51 69,305 55,651 188 62 1,111,523 12,622,209 79.64 117.84 3 7,353,213 62.44 172.94 68,535 59,129 201 67 1,196,807 13,479,003 87.09 127.37 4 7,078,935 64.83 175.70 68,106 62,618 214 72 1,251,248 14,348,559 93.74 135.06 5 6,807,147 66.89 178.41 68,325 66,005 228 77 1,314,908 15,225,198 99.46 142.56 6 6,536,209 68.81 181.15 68,863 69,412 243 82 1,406,996 16,112,870 104.79 150.17 7 6,265,604 70.56 184.09 69,851 72,795 261 88 1,529,994 17,022,935 109.66 158.36 8 5,995,891 72.08 187.03 71,285 76,162 281 95 1,662,525 17,960,492 113.89 166.51 9 5,726,809 73.54 190.01 72,921 79,608 303 102 1,817,683 18,929,779 117.93 174.79 10 5,456,404 74.83 192.85 74,854 83,107 326 110 1,971,609 19,935,177 121.50 182.70 Note: Kerosene price - $432.50; LPG price - $1250/ton. 112 Trade, Global Policy, and the Environment Notes to the wholesale price, yielding a more rapid run- This work was supported by the project on the Social up in retail prices. and Environmental Consequences of Growth-orient- ed Policies. We are grateful to Douglas Barnes for shar- References ing data on household fuel consumption, Neil Quarmby of IS Ltd for providing the GIS data used in Barnes, Douglas F., Jeffrey Dowd, Liu Quan, Kerry this report and to Robert van der Plas and Luis Guti- Krutilla, and William Hyde. 1998. The Urban En- errez for a variety of parameters and data. ergy Transition: Energy, Poverty, and the Environment in the Developing World. World Bank. Processed. 1. TheUnitedStatesGeologicalSurvey andSenegal's Barnes, Douglas F., Keith Openshaw, Kirk R. Smith Centre de Suivi Ecologique (CSE) are monitoring and Robert van der Plas. 1993. "The Design and long-term changes in land cover at 150 sites in Diffusion of Improved Cooking Stoves." The World Senegal, complemented by the use of aerial Bank Research Observer 8(3): 119-141. videography and remote sensing to monitor land Benjaminsen, Tor A. 1993. "Fuelwood and Desertifi- cover change over wider areas. See G. Tappan, E. cation: Sahel Orthodoxies Discussed on the Basis Wood, A. Hadj, and A. Bodian 1994, "Monitoring of Field Data from the Gourma Region of Mali." Natural Resource Changes in Senegal: Preliminary Geoforum 24(4): 397-409. Evidence from Field Sites," EROS Data Center, Chomitz, Kenneth M., and Charles Griffiths. 1997. An Sioux Falls, SD. Economic Analysis of Woodfuel Management in the 2. "The Assessment of Vegetation and Land Use Sahel: The Case of Chad. World Bank, Policy Re- Changes in Nigeria between 1976/78 and 1993/ search Department: Policy Research Working Pa- 95." Report submitted by Geomatics International per no. 1788. Inc. to The Forestry Management, Evaluation and Cleaver, Kevin M., and Gotz Schreiber. 1994. Revers- Co-ordinating Unit and the Environmental Man- ing the Spiral: the Population, Agriculture, and Envi- agement Projection (EMP) for Nigeria, 1998. ronment Nexus in Sub-Saharan Africa. Washington DC: World Bank. 3. Data from a June 1997 tabulation by TWUTD, Cuneo e Associati. 1992. Petroleum Products Supply and World Bank. Distribution in Sub-Saharan Africa: Report on Nige- 4. Other benefits, such as the health benefits of re- ria and Landlocked Countries. World Bank. Pro- duced smoke, may be substantial, however. cessed. 5. Not utilized in the simulations reported here. Cuneo e Associati. 1993. Petroleum Products Supply and Distribution in Sub-Saharan Afrtica: Executive Sum- 6. The equations were also run excluding cities with mary. World Bank. Processed. low reported availabilities of any of the fuels. The de Montalembert, M.R., and J. Clement. 1983. Fuelwood coefficients were robust save for the own-price Supplies in the Developing Countries. Rome: FAO. elasticity of LPG, which became statistically insig- Forestry Paper no. 42. nificant and close to zero. Since it is possible that Dewees, Peter A. 1995. "Fuelwood responses to tree low availability of LPG reflected high price, we scarcity: The case of woodfuel." In J.E. Michael used the more intuitive results of table 7.1, which Arnold and Peter A. Dewees, eds., Tree Manage- shows a strong substitution between LPG and ment in Farmer Strategies: Responses to Agricultural kerosene. Intensification. Oxford: Oxford University Press. 7. The model was calibrated for the simulations of ESMAP. 1993a Tchad. El6ments de stratigie pour l'inergie part 4. Constant terms for each of the In ratio equa- domestique urbaine: Le cas de N'djamena. Report no. tions were chosen so as to generate the starting 160/94. share values; these in turn were based on estimated Sector. levels for Chad in 1995. Foley and others. 1995. The Niger Household Energy 8. The retail markup is held constant in absolute Project: Promoting Rural Fuelwood Markets and Vil- terms. To the extent that retailing costs include lage Management ofNatural Woodlands. World Bank capital costs, they might increase proportionately Technical Paper no. 362, Energy Series. Fuel Prices, Woodlands, and Woodfuel Markets in the Sahel 113 Gately, Dermot and Shane Streifel. 1997. The Demand Nichol, J.E. 1989. Ecology of fuelwood production in for Oil Products in Developing Countries. World Bank Kano Region, Northern Nigeria. Journal of Arid Discussion paper no. 359. Environments 16: 347-60. Hofstad, Ole. 1997. "Woodland Deforestationby Char- Quarmby, Neil, Andrew Millington and Hugh coal Supply to Dar es Salaam." Journal of Environ- Vaughn Williams. 1996. Domestic Energy Project: mental Economics and Management 33(17): 17-32. Fuelwood Inventory in the region around Jensen, Axel Martin. 1995. Examen des Politiques, Strat- N'Djamena. Final report, January 1996. Manches- egies, et Programmes du secteur des Energies ter, U.K.: I.S. Ltd. Traditionelles. World Bank: AFTPS, January 1995. RPTES (Review of Policies in the Traditional Energy Processed. Sector). 1995. Regional Report: Senegal-Gambia- Kidane, Asmerom. 1991. "Demand for energy in ru- Burkina Faso-Mali-Niger. Maastricht, Netherlands, ral and urban centres of Ethiopia: an econometric May: 15-17. analysis." Energy Economics 13: 130-34. Schloss, Miguel. 1993. "Does Petroleum Procurement Leach, Gerald, and Robin Mearns. 1988. Beyond the and Trade Matter?: The Case of sub-Saharan Af- Woodfuel Crisis: People, Land and Trees in Africa. Lon- rica." Finance and Development 30(March): 44-46. don: Earthscan. Van der Plas, Robert, and Luis Gutierrez. 1996. Per- Leach, Melissa and James Fairhead. 1994. The Forest sonal communication. Islands ofKissidougou: Social dynamics ofenvironmen- Wiedenmann, Ralf. 1991. "Deforestation from the tal change in West Africa'sforest-savanna mosaic. Re- overexploitation of wood resources as a cooking port to ESCOR of the Overseas Development fuel. A comment on the optimal control model of Administration. Hassan and Hertzler." Energy Economics 13(2): 81- Le Hou6rou, Henry N. 1989. The Grazing Land Ecosys- 85. tems of the African Sahel. New York: Springer- Wilton, Max. 1996. "Traditional Energy-the West Verlag. African Experience." World Bank. RPTES Discus- Mayorga-Alba, Eleodoro. 1993. "Rationalization of sion Paper no. 12. imports, refineries and distribution of petroleum in Sub-Saharan Africa." Natural Resources Forum 17(May): 109-16. Chapter 8 In Search of Pollution Havens? Dirty industry in the world economy, 1960-1995 Muthukumara Mani and David Wheeler 1. Introduction means polluters pay more - for pollution control D uring the past three decades, many poor equipment, conversion to cleaner processes, or pen- countries have experienced rapid economic alties for unacceptable emissions. This regulatory gap development after adopting liberal econom- between developed and developing countries could, ic policies. In manufacturing sectors such as apparel in principle, produce "pollution havens" analogous to assembly, a "cascading" pattern of growth has accom- "low wage havens." Pollution-intensive industries panied this global movement toward openness. Pro- (that is, those with low elasticities of substitution be- duction has accelerated in progressively poorer tween use of the environment and other productive countries, as wage increases in rapidly-developing factors) might join labor-intensive industries in the open economies have changed the pattern of compar- migration from the OECD countries to open develop- ative advantage. Downward migration of garment ing economies, if the latter remained unregulated and assembly has reflected its continuing labor-intensity: environmental pricing were a significant determinant Low unit labor costs in poor economies have been of comparative advantage. sufficient to offset the potential for automated pro- Have "pollution havens" in fact emerged? In this duction in higher-wage countries.1 We might there- paper, we examine the record using international in- fore characterize the world garment story as a formation on industrial production, trade and envi- continuous (and self-defeating) search for "low wage ronmental regulation for the period 1960-1995. The havens" by apparel manufacturers. From the perspec- paper is organized as follows. In Section 2, we use tive of development economics, this has been salu- recent results on the relationship between regulation tary: Exports of garments and other light assembly and development to argue that "pollution havens" goods have provided the first rung on the ladder of must be as transient as "low wage havens:" In the rapid income growth and skills development for mil- worst case, we would expect to see pollution-inten- lions of poor workers. sive sectors follow the "cascading" international Although they are critical factors, relative wag- growth pattern of garment production. Section 3 es and labor skills are not the only determinants of draws on several empirical approaches to identify locational advantage. Other long-recognized factors industry sectors which are clearly among the most include the quality and local price of available ener- pollution-intensive. Focusing on these sectors, the rest gy and raw materials, agglomeration economies, and of the paper examines the evidence on shifts in pollu- so forth.2 More recently, attention has turned to the tion-intensive production from the OECD to devel- possible impact of differences in environmental reg- oping economies. Section 4 focuses on the OECD, ulation. In the OECD economies, stricter regulation giving particular attention to the Japanese case. Sec- 115 116 Trade, Global Policy, and the Environment tion 5 considers the experience of developing Asia and the literature has been to identify pollution-intensive Latin America, while Section 6 provides a summary sectors as those which have incurred high levels of and conclusions. abatement expenditure per unit of output in the Unit- ed States and other OECD economies (Robison 1988; 2. Development, Regulation, and Tobey 1990; Mani 1996). By this criterion, five sectors "Pollution Havens" emerge as leading candidates for "dirty industry" sta- "Low wage havens" are transient because incomes tus: Iron and Steel, Non-Ferrous Metals, Industrial and wages increase continuously with development. Chemicals, Pulp and Paper, and Non-Metallic Miner- Similarly, recent evidence suggests that "pollution al Products.4 Another, more direct, approach is to se- havens," if they exist, may be transient because envi- lect sectors which rank high on actual emissions ronmental regulation also increases with develop- intensity (emissions per unit of output). To determine ment. Dasgupta and others (1996) document the high-ranking sectors by this criterion, we have used striking correlation between national income per cap- detailed emissions intensities by medium for U.S. ita and the strictness of environmental regulation. Re- manufacturing at the 3-digit Standard Industrial Clas- cent studies of regional income and regulation in sification (SIC) level, computed by the World Bank in China (Wang and Wheeler 1996; Dasgupta and Wheel- collaboration with the U.S. EPA and the U.S. Census er 1997) and Indonesia (Pargal and Wheeler 1996) find Bureau (Hettige and others 1995). We have comput- similar relationships. ed average sectoral rankings for conventional air pol- Regulation of industrial pollution increases with lutants, water pollutants, and heavy metals, with economic development for two main reasons. First, results displayed in table 8.1. the demand for environmental quality rises with in- Again, five of the six sectors with highest over- come, both for aesthetic reasons and because the val- all ranks are Iron and Steel, Non-Ferrous Metals, In- uation of pollution damage increases. Secondly, more dustrial Chemicals, Pulp and Paper, and Non-Metallic developed economies have (on average) more high- Mineral Products. We have therefore selected them ly-developed public institutions and are more capa- as the "dirty sectors" for this analysis. If there is a sig- ble of enforcing desirable environmental norms.3 If nificant pollution havens story, it should emerge in the income elasticity of regulation is greater than one, their international development history since 1960. then developing countries will not retain a compara- However, since this is a comparative advantage sto- tive advantage in dirty production. However, the ry, we need to be sure that it is not confounded by interplay of relative prices and agglomeration econo- changes in relative prices of factors other than "envi- mies might lead to a garment-style "cascading" pat- ronment" in which dirty industries may also be inten- tern of growth, in which rapid growth of dirty sectors sive. is visible during transitional periods when regulation In fact, it is quite reasonable to suppose that pol- lags behind the growth of output and income. lution-intensive industries are intensive in other in- puts, particularly bulk raw materials, energy and land. 3. Defining Dirty Industries Pollutants are waste residuals-harmful byproducts To test for pollution havens, we need a clearly-defined of industrial processes which are not profitable to re- set of "dirty" industries. A conventional approach in cycle or resell at existing prices (including the price Table 8.1 Ranking of pollution-intensive industries Rank Air Water Metals Overall 1 371 Iron and Steel 371 Iron and Steel 372 Non-Ferrous Metals 371 Iron and Steel 2 372 Non-Ferrous Metals 372 Non-Ferrous Metals 371 Iron and Steel 372 Non-Ferrous Metals 3 369 Non-Metallic Min. Prd. 341 Pulp and Paper 351 Industrial Chemicals 351 Industrial Chemicals 4 354 Misc. Petroleum, Coal Prd. 390 Misc. Manufacturing 323 Leather Products 353 Petroleum Refineries 5 341 Pulp and Paper 351 Industrial Chemicals 361 Pottery 369 Non-Metallic Min Prd. 6 353 Petroleum Refineries 352 Other Chemicals 381 Metal Products 341 Pulp and Paper 7 351 Industrial Chemicals 313 Beverages 355 Rubber Products 352 Other Chemicals 8 352 Other Chemicals 311 Food Products 383 Electrical Products 355 Rubber Products 9 331 Wood Products 355 Rubber Products 382 Machinery 323 Leather Products 10 362 Glass Products 353 Petroleum Refineries 369 Non-Metallic Min. Prd. 381 Metal Products In Search of Pollution Havens? 117 of pollution). The volume of such residuals is, almost data from Japan, we have computed energy, land and tautologically, largest in weight-reducing industries labor intensities for the five cleanest and dirtiest sec- which transform bulk raw materials into primary in- tors and compared the results. puts for industrial production. Such industries should The results for energy intensity, displayed in fig- be land intensive, because some bulk material inven- ure 8.1, are very clear: The five dirty sectors are about tories must be stored on-site. They should also be en- three times more energy intensive than the five clean ergy-intensive, because transformation processes sectors, and there is striking uniformity within the two generally involve the application of high temperature, pressure, and/or mechanical force to raw material tion the result is basically thmome for land intensity inputs. We are agnostic about their capital- and la- tin h euti aial h aefrln nest bor-inputs.Wen but agnosthe facouto ahire cletarly impa- (figure 8.2): It is about three times higher in the dirty torintenfor bu thesompate fadatorsage alsto learly impor- sectors. Capital intensity is also substantially higher tant~~~~~~~~~ fo th.oprtv avnaesoy in the dirty sectors, with an average ratio around 2:1 To test for differential intensities, we have iden- tified the five "cleanest" sectors using the same pollu- for capital/output (figure 8.3) and investment/out- tion-intensity rankings employed for table 8.1. They put (figure 8.4). Labor intensity shows considerable are SIC sectors 321 (Textiles), 382 (Non-Electrical variation within groups, but the clean sectors are Machinery), 383 (Electrical Machinery), 384 (Transport about 40 percent more labor intensive on average (fig- Equipment), and 385 (Instruments). Using available ure 8.5). Figure 8.1 Energy intensity in Japanese Figure 8.3 Investment/output eatios in manufacturing Japanese manufacturing, 1972-1991 8 0.14 7l l l * l °0.12 6 ~~~~~~~~~~~~~~~~~~~~~~~~~~0.1 0.08 2 ~~~~~~~~~~~~~~~~~~~~0.02 0~~~~~~~~~~~~~~~~~~~~~~~~ 0) i_. 11811 1I 83 2 l j j , 0.02 2 l l 1 0.0 0I i I ! I = W g l l l _ } j |341 351 369 371 372 321 382 383 384 385 341 351 369 371 372 321 382 383 384 385 Figure 8.2 Land intensity in Japanese Figure 8.4 Capital/output ratios in Japanese manufacturing manufacturing, 1985-1989 14 2.5 1221 0C 2 . 0 0- 341 351 369 371 372 321 382 383 384 385 _ - Ct - N - N CO fl 10 St e SD 1- N- N C:t CD cc cc CO XO XO 1C> V) C CO CO XO 1C 118 Trade, Global Policy, and the Environment Figure 8.5 Labor intensity in Japanese through" to customers and temporary escalation of manufacturing output value relative to the value of sectors with low- 0.2 er energy intensity. After each break, however, the 0.18k downward trend quickly reasserted itself. What ex- 0.164- o0.4 plains this decline? Obvious candidates are price 0S 0.12. changes for factors in which the pollution-intensive 0. - .-a 0.08- U 0.06- Figure 8.6 Japanese production ratio o 0.04 (polluting/non-polluting) 0.02- 0, 0.7 _ O) . N N.- CN _ LC) CO N N N OD CO Cv) CO C') CO C' C) X ) C') ) 0.6 0.6 To summarize, our evidence suggests that dirty 0 industries are relatively intensive in capital, energy 03 and land; their clean counterparts are relatively in- tensive in labor, although the difference is not as strik- 0.2 ing. Clearly, the pollution havens hypothesis cannot 0.1 be tested using information on environmental regu- 0 I I E I I I I I I i lation alone. Shifts in pollution-intensive production 1963 1965 1967 1969 1971 1973 1975 1977 1979 1981 1983 1985 1987 1989 might also be explained by differences in the relative Figure 8.7 Japan's production of polluting prices of these other inputs. In the following section, products (as a percentage of total manufacturing) we take a careful look at the relative price story for 0.25 Japan.0. 4. Pollution-Intensive Production in the OECD 0.15 The Japanese story 0.1 THE RELATIvE DECLINE OF POLLUt7ON-INTEIVS'VE SECTORS During the past three and one-half decades, Japan has 0.05 gone through major changes in economic and envi- ronmental conditions. As figures 8.6 and 8.7 suggest, o ....... l l l l l I l these changes have had significant consequences for 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 pollution-intensive industry in Japan. Whether com- pared to the five cleanest sectors (figure 8.6) or Japa- sectors are highly intensive: Energy, land and envi- nese manufacturing as a whole (figure 8.7), the share ronment. of our five dirtiest sectors trended strongly downward during the period 1963-1993. As a proportion of clean- ENERGY AND LANI> PRICES sector production, dirty-sector output dropped from Energy prices escalated sharply in 1973, and this in- nearly 70 percent in 1963 to about 30 percent in the crease may well explain part of the decline in the share mid-1990s. As a proportion of total manufacturing of pollution-intensive production in the 1970s. How- production, dirty-sector output dropped from about ever, relative energy prices were not increasing dur- 25 percent in the early 1960s to about 15 percent in ing the 1960s and 1980s-periods when Japan's dirty the mid-1990s. Both series show clear breaks in trend industry share was also declining. during the two periods of rapid energy price escala- As figure 8.8 shows, the past three decades have tion, 1973-74 and 1978-80. Since the dirty sectors have also witnessed a rapid escalation of land prices in Ja- relatively high energy intensity, the short-run re- pan. However, the decline in dirty-sector share be- sponse to the energy price increases was a "pass- ganslackeninginthel980s,atpreciselythetimewhen In Search of Pollution Havens? 119 Figure 8.8 Urban commercial land price Figure 8.10 Pollution control investment by index (1990 prices) big enterprises 120 10000 100 ~ 8000 so0 6000 60 4 4000 400 2000 20 0 ~~~~~~0 I tI I I Il CD CD f. - _C T CO LD OD -0 - ,~ - - - - .~ 0) 0) 0) 0) 0) 0) 0) land prices began escalating sharply. Land price in- UEA) was set up in 1971 and, as figure 8.9 shows, its creases may therefore have played some role, but it activities grew very rapidly during the 1970s. seems unlikely that they were a decisive factor. As JEA's regulatory activity increased, Japanese industry went through a period of rapid adjustment THE TIMNG AND IMpAcT OF to new environmental norms. Figure 8.10 shows that ENVIRONMENTAL REGULATION the mid-1970s witnessed a surge of investment in pol- Does the timing of stricter environmental regulation lution control by Japanese industry. Thus, it is plau- in Japan jibe with the patter of relative decline in sible to suppose that tightened regulation had an impact on the relative fortune of pollution-intensive pOllution-inytensivenprodctidson ies pllutik Tokyo, production in Japan during in the 1970s and early '80s. Oska ad ytoenctdsoe oluio cnto However, tightened regulation cannot explain the measures by the mid-1950s, and Japan's first water del r, tion-inensieg poution in the quliy reevaio lwwa eacedi 158 Hw decline of pollution-intensive production in the 1960s. quality preservation law was enacted in 1958. How- ever, strong opposition from growth- and industry- INCOME GROWTH oriented national ministries hampered the national movement toward stricter regulation until the late Figures 8.6 and 8.7 suggest a consistent pattern of de- 1960s. From 1967 to 1970, regulations covering indus- cline in the share of dirty industry from the sixties to trial air and water emissions were enacted in rapid the nineties, despite large interim changes in the pric- successio an, water Japaness wenvroenactal Agenc rap es of energy, land and environment. During this en- tire period, of course, Japanese income per capita Figure 8.9 Japanese Environmental Agency increased rapidly. Our five dirty sectors are "basic" budget industries, whose domestic demand elasticity falls as 60000 income increases. Thus, income growth alone may 50000 y explain a significant part of their relative decline. c 50 Figures 8.11 and 8.12 tell an intriguing story in 40000/ this context. During the "pre-environmental" period 0 (o 30000 of rapid growth, from 1963 to the mid-1970s, declin- o ing trends in both consumption/production and 20000 import/export ratios are consistent with the income- 10000 elasticity hypothesis. However, in the mid-1970s both 0 I 1 I1 : I1111 I IIIIIIII ratios reversed course: Production of polluting prod- ucts slowed relative to consumption, and imports in- - _t N c cos ( o 0a ) 0) 0 0 CY)creased much more rapidly than exports. Thus, Ir- Irl, Tl- although relative domestic demand for pollution- 120 Trade, Global Policy, and the Environment Figure 8.1 1 Consumption-production ratio duction and import/export ratios since 1976, coupled of polluting products in Japan with rapid growth of the JEA and sharp increases in industrial pollution control investments. A plausible inference is that stricter environmental regulation had 0.95 - a significant impact on Japan's comparative advan- tage in pollution-intensive products. We conclude that 0.9 regulation probably led to both significant abatement 0.85 by pollution-intensive industries in Japan, and dis- placement of some pollution-intensive production to 0.8 Japan's trading partners. 0.75 I I I I I I l l I l I l l l ^ ° > O S X - > O w ^ - > g ^ Dirty production in North America and Western Europe Figure 8.12 Import-export ratio of polluting Although we have used the Japanese case to examine products in Japan the "environmental transition" in detail, regulation 1 was also increasing rapidly in North America and 0.9 Western Europe during the same period. Rapidly-ris- 0.8 ing real wages in the 1960s and energy price hikes in 0.7 the 1970s were common to all three OECD regions. 0.6 As in the case of Japan, the growth of environmental 0.5 regulation imposed substantial abatement costs on 0.4 - industry in North America and Western Europe dur- 0.3 ing the 1970s. However, other conditions were differ- 0.2 ent. Slower income growth after 1970 should have 0.1 dampened the domestic income elasticity effect and 0 ~ ~ ~ ~~~~~~~~~, kept the escalation of land prices well below Japanese 0 a@ X _ rates. In the case of North America, three additional factors were operative: Relative to Japan and West- intensive products undoubtedly continued to decline ern Europe, North America has low settlement densi- with income growth after 1975, the evidence suggests ty (and land prices), a much greater supply of bulk that the pure "income effect" was outweighed by an- raw materials, and substantially lower energy other factor which suddenly retarded the growth of prices - all factors which would enhance comparative domestic production and exports. In light of the re- advantage in pollution-intensive products. North gression results, the most plausible candidate for this America also has a skilled labor force capable of rap- role is increased environmental regulation. id adjustment toward cleaner production processes. It is therefore at least possible that the environmental Dun-Y INDusTRYws "RETREAT" FROM JAPN A SUMZMRY era witnessed displacement of pollution-intensive PERSPECTIVE production from Japan and Western Europe to North Many factors can be invoked to explain the relative America, whose higher abatement expenses would decline of pollution-intensive industry in Japan since have been compensated by a cost advantage in other 1963. Although energy and land prices are plausible factors. culprits, our regression analysis suggests that they Figures 8.13-8.16 tell us what actually happened. have not played dominant roles in the story. Lower In the case of North America, the evidence suggests income elasticities for dirty-sector products have un- that the impact of environmental regulation out- doubtedly played a role. Indeed, the evidence sug- weighed the potentially-displacing effects of tighter gests that they played a dominant role in the 1960s. regulation in Japan and Western Europe. From 1963 However, the most striking part of Japan's dirty-sec- to 1993, the dirty-sector share of industrial produc- tor story is the rapid increase in consumption/pro- tion declined steadily in the United States and, more In Search of Pollution Havens? 121 Figure 8.13 Production of polluting products Figure 8.16 Europe's import/export ratio in the United States and Canada 1.2 0.25 a . L_ 0.8 0.2'~0.6 O.tS- - CANADA 0 0.1 0.4 - 0.05 l g l l ll l l l l l l l l ll l l l l l l l l ll l l g l0 2 0.2 0 . .a .q .r0 _ o N o 0~~~~~~~~~~~~~~~~~~ 8 co CD N LO co - t $- 0 C t co n CD N- N- Nn- co co CD t_ Figure 8.14 Production of polluting products Figure 8.17 United States-Canada in Europe consumption-production ratio 0.25 1.00006 *_ 01.00004 W0.250998- *A .o 0.2 1.00002 w 0.15 0.99998 0.99996 o o.1 0.99994 I 0 ~~~~~~~~~~~~~~~~0.99992- 0 ~~~~~~~~~~~~~~~~~0.9999 co 0.05 to 0.99988 0 CD ~~~~~~~~ ~~~~~~0.99986 Ct CD N C D - - 0 CO ( C Nv cD LO co u 0 ct o CD CD N- VI- N- CD CD Co tn CD co co C - - . ' W C Cn CD t n CD CD CD CD CD CD CD CD tn T] v) TA C4 T tn) tn tn tn tn tn - Figure 8.1 5 North America's import/export Figure 8.18 Euorpe's consumption- ratio production ratio 1.02- 1.2- 0.98 0.8 0.96 0.6 0.94 0.92 0.4 0.9 0.2 0.88 0 IIi rrtirr c i i i i-ii 0.86 C', CD CD Nx UA, CD - > N- 0 CD CD CD CO r C CD CD CD CD C') CD CD 01 LCD c o CD cD cD CD CD CD COC D CD tn Aw tn tn tl) tn _) tn tn tn tra tv tv tv ts b ts~! !~! ~ 0) 0) 02 a modestly, in Canada. North America's dirty-sector port ratio increases steadily. Thus, despite several consumption/production ratio shows no trend and possible countervailing factors, the North American fluctuates within a very narrow range; its import/ex- experience was actually quite similar to that of Japan. 122 Trade, Global Policy, and the Environment Western Europe also displays a declining share Although other interpretations are doubtless of dirty-sector production throughout the period, al- possible, these data are consistent with the following though its dirty-sector import/export ratio has re- argument: During the 1960s, rapid growth in all re- mained approximately constant. Paradoxically, it is gions coincided with relatively weak environmental the Western European, not North American, con- sumption/production ratio which exhibits a down- Figure 8.19 Latin America's production of ward trend well into the 1980s before returning to its polluting products rning ~~~0.25 1960s level in the 1990s. 0.2 Summarizing the OECD experience To summarize briefly, two general patterns are visi- 0.15 ble in dirty-sector production trends for the OECD economies since 1960. In all three regions, the share 0.1 of pollution-intensive industries has significantly de- clined. In two of the regions-Japan and North Amer- 0.05 ica-this decline has been accompanied by net displacement of polluting production to trading part- 0 I i i I I i i i i c _ A iiiiiii co) Lo N. CD -Z co) LO 0) -; co Lo N- a) CD D co CD r, N, N, N, N coC co co co ners, while approximate trade balance has been pre- served in Western Europe. Part of the decline was probably due to low in- Figure 8.20 Latin America's import/export come elasticity of demand for pollution-intensive ratio products. In all three cases, however, another part is 4 most plausibly attributed to stricter environmental 3.5-- regulation and rising abatement costs. The energy 3-- price shock may also have had an impact, but our re- 2.5-- gression results for Japan cast some doubt on this 2 hypothesis. 1.5 5. Pollution-Intensive Production in 1 Developing Asia and Latin America The general story The international impact of regulation, income growth 1963 1969 1975 1981 1987 1993 and (perhaps) energy price changes are strikingly il- lustrated by the juxtaposition of OECD trends with figures 8.19-8.22 for Latin America and Asia.5 For Figure 8.21 Asia (excludingJapan): Polluting these developing regions, the graphs show a steady sector production share, 1964-1998 upward trend in pollution-intensive production 0.19 share-a mirror image of the downward trends in 0.18 North America, Europe and Japan. Superposed on this 0.17 steady increase are pronounced turning points in Latin 0.16 American and Asian import/ export ratios in the mid- 0.15 1970s. Latin America exhibits a rising trend before- 014 hand, a steep fall afterwards, and a leveling in the 13 1980s. The Asian series also exhibits a sudden shift 0.12 I l l l l I l l l l I l I downward in the mid-1970s, but approximate con- stancy otherwise. 1964 1966 1968 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 In Search of Pollution Havens? 123 Figure 8.22 Asia (excludingJapon): Import/ decreasing in both regions. Latin America's import/ export ratio for polluting products, 1964-1994 export ratio stabilized near unity (balanced trade), 4 - while Asia remaining a significant net importer of 3.8; - W _ .pollution-intensive products. 3.32 t \ Asian sub-regions 2.8s < , We turn to Asia for a more detailed analysis of histor- 2' 4ical trends because its recent growth experience has 2.2 included rapid transitions in income and economic structure for a number of countries. These raise the 1964 1968 1972 1976 1980 1984 1988 1992 possibility of rapid change inlocational advantage for pollution-intensive sectors, and in fact the data sug- regulation in the OECD economies and low, stable gest that such transitions actually occurred. For our energy prices worldwide. During this period, domes- analysis, it is convenient to group the Asian econo- tic income elasticity effects were dominant: Relative mies into three categories (see table 8.3):7 the Newly demand for pollution-intensive basic products such Industrializing Economies (NIEs- Hong Kong, Sin- as metals, chemicals, paper and cement fell in the gapore, Korea, Taiwan); Developing East Asia (DEA OECD economies and grew at least as rapidly as do- -Malaysia, Indonesia, Thailand, Philippines and mestic production in the poorer nations of Asia and China); and South Asia (SA -India, Pakistan, Bang- Latin America (apparently more rapidly in the latter). ladesh, Sri Lanka). When the environment/energy shock hit in the Tables 8.2-8.3 and figure 8.23 provide compara- mid-1970s, the sudden shift in relative prices changed tive evidence on economic growth, degree of open- conditions very significantly at the margin. From a ness (by the Summers-Heston measure8) and the position as net importers of pollution-intensive goods timing of environmental regulation during the past (with imports at three times the level of exports), Lat- three decades. The NIEs were relatively open in 1970, in America and Asia experienced a rapid decline in at the beginning of the environmental era and were import/export ratios as weaker regulation and, pos- already experiencing rapid growth; DEA began liber- sibly, lower energy prices altered comparative advan- alizing significantly and growing rapidly in the 1970s. tage in dirty-sector production. During this period, changes in relative costs augmented the income elas- Figure 8.23 Openness in Asia ticity effect: Pollution-intensive production grew faster in the developing regions, and receded more quickly in the industrial economies, than could have been 200 predicted from income change alone. 150 1 By the mid-1980s, three mitigating factors had 100Nl again changed the picture significantly. First, as in- 50 come growth continued, the more industrialized (and 0 polluted) economies of Latin America and Asia prob- ably experienced some decline in the income elastici- ty of demand for dirty-sector production. Second, 80 rising consciousness of environmental problems led *70 East Asia to enactment and enforcement of stricter environmen- 60 tal regulations in both Latin America and Asia.6 Third, 50 the energy price gap narrowed as world petroleum 40 South Asia 30 prices stabilized and developing countries began 20 abandoning energy subsidies. As a result of these 10 three changes, the share of dirty-sector production o Ia i i : n in ii stopped increasing and import/ export ratios stopped 124 Trade, Global Policy, and the Environment Table 8.2 Openness and economic progress in selected developing countries in Asia and Latin America Openness Econornic Progress Average annual Average annual growth of GNP per People living on less growth of Degree of capita than $1 a day exports openness (percent) (percent) (percent) 1990" 1970-1995 1981-95 1980-95 East Asia China 25.4 6.9 29.4 12.5 Indonesia 52.6 4.7 14.5 10.6 Korea, Rep. of 62.5 10.0 -- 11.6 Malaysia 154.20 4.0 5.6 13.6 Philippines 61.5 0.6 27.5 5.3 Thailand 75.8 5.2 0.1 16.7 South Asia Bangladesh 26.9 1.5 -- 10.1 India 18.8 2.4 52.5 6.5 Pakistan 35.0 2.9 11.6 9.1 Sri Lanka 67.4 3.2 4.0 11.5 Latin America Brazil 12.6 -- 28.7 6.3 Chile 65.5 1.8 15.0 7.3 Colombia 35.4 1.9 7.4 8.1 Mexico 32.7 0.9 14.9 13.0 Peru 26.8 1.1 49.4 2.4 Venezuela 59.6 1.1 11.8 1.0 Notes: -- Notavaflable. a. Summers-Heston openness index defined as (exports+imports)/nominal GDP. Source: World Bank 1997. Table 8.3 Initial national environmental legislation in Asia Country Air Water Toxics Japan 1967 1958 1958 NICs Hong Kong Singapore 1978 ----- ----- Korea, Rep. of Taiwan (China) 1975 Deveoping East Asia Malaysia 1977 1977 1979 Indonesia 1988 1988 ----- Thailand 1975 1975 1989 China 1985 1985 1989 Philippines South Asia India 1974 1981 1986 Pakistan 1983 1983 ----- Bangladesh ----- ----- ----- Source: Brandon and Ramankutty (1993). In Search of Pollution Havens? 125 Figure 8.24 Republic of Korea: Pollution- Figure 8.27 Developing East Asia's intensive industry share, 1963-1993 production of polluting products 023 --0.14- 022 - - 0.12 021 o; 0.1 02 0.08 0.19 20.06 0.180- 0.17 00 0.16 0.02 (0 2 C) - c) C)Cl) (0 N ) CO LO 0) . C LO N 0) (C (C (0(0 N N N N N (C (C D C (C (Cl C) 2D N O N N 2D 1D O N (0 (0 co~~~~~~ ~ ~ ~ ~ co O> w (Dr- r- m m a) x a} - OD co OD Figure 8.25 NiEs' production of pollution Figure 8.28 Developing East Asia's import! products export ratio 6 0.2 5 0.15 ~~~~~~~~~~~~~~4 3- 0.1 2 0 - 0-~~~~~~~( C CN N N C(C( C(0C )0 aT a) a) a) a)1 a)t 0 a) 0) a) CD 0D 0C a a) 0 1976 1978 1980 1982 1984 1986 1988 1990 1992 Figure 8.26 NiEs' import/export ratio Figure 8.29 South Asia's production of polluting products 9 < 0.3 8 7 c 0.25 1963 1966 t969 t972 t975 t978 t98t t984 1987 t990 t993 - - - - - - - - 0.2 5 - 4 - ~~~~~~~~~~~~~~~~~0.15 3 - 00.1 2. 1 ~~~~~~~~~~~~~~~~~~0.05 - 0 0 N 0 ~ I N 0)- z N No N N (C- (C 0 t- cm 1963 1966 1969 1972 1975 1978 1981 1984 1987 1990 1993 126 Trade, Global Policy, and the Environment Figure 8.30 South Asia's production of in the NIEs, DEA and SA. However, in each region it polluting products has remained a markedly short story. In the NIEs and 20 DEA, the combined effect of regulation and falling 18 demand elasticity has apparently stabilized the dirty- 16' sector production share, leaving both regions (and SA) 14-r as net importers of pollution-intensive products. 12 The same story is reflected in the historical pat- io tern of consumption/production ratios (figure 8.31). 6 Although they have declined for over two decades in 4 both DEA and SA, they remain above one in all three 2 regions. In fact, all the ratios have been very close to 0 - :Ti,l1i oiiiiiiI lii ciijii one during the entire period. Thus, our "cascading" zo (0 (0 0 N s r 0 cc co co CD co 0 cn CO CD CD N- N N- N- - ( (0 (0 0 ( M O~ dirty-sector production story has been a decidedly marginal affair. Stricter enforcement of environmental regulation be- gan in the 1970s in the NIEs and the 1980s in DEA. 6. Conclusions and Implications SA began more rapid growth in the 1980s, but stricter The last three decades have witnessed rapid econom- environmental regulation was delayed until the 1990s. ic development, particularly in countries which have Across the three regions, the growth experience pursued relatively open economic policies. Rising of pollution-intensive industry seems to reflect these environmental awareness in the 1960s also led to a developments in a "cascading" pattern. Figures 8.24- rapid tightening of pollution regulation in the indus- 8.30 show changes in dirty-sector production shares trial economies. This created an international gap in and import/export ratios. The dirty-sector share grew environmental pricing between industrial and devel- in all three regions during the 1970s; it leveled off in s the NIEs and DEA, but continued rising in SA. The luon ens in , the9resu Acord he pol- NIEs' dirty-sector import/export ratio fell sharply m.ution havens hypothesis, the result should have been during the 1970s. When regulation tightened and the more rapid growth of dirty industries in unregulated NIEs' ratio leveled off in the early 1980s, it began fall- economies which were open to international trade. ing in DEA. As regulation tightened in DEA during In this paper, we argue that a full understand- the late 1980s, the import/export ratio began falling ing of the pollution havens problem requires good in SA. evidence about the interactions linking economic de- To summarize, for dirty-sector production in velopment to regulation, industrial location, sectoral Asia we see an adjustment pattern which looks like a pollution intensity, energy and other input prices. "cascading pollution havens" story. It begins in Ja- Most of the previous studies have focused on the lo- pan in the early 1970s, and continues for two decades cation issue, typically suppressing the otheir factors by lumping countries into simple "developed/devel- Figure 8.31 Consumption-production ratio oping" categories and basing conclusions on sectoral in Asia (excluding Japan) + South Asia changes evaluated at constant (generally U.S.-based) NIE'S ~ abatement expenditures. In addition, the locational 1.02 \\ / \DEAanalyses have typically not considered many of the 1.015 DEA ~~~~~~~~~standard location factors in drawing inferences about 1.01 the impact of differential regulation and abatement 1.00O costs. 1 Although data restrictions have prevented us 0o995 ssfrom incorporating some of the factors mentioned 0.99 here, our(as we have shown for Japan) results do point 0.985 I I , , C , I to some interesting findings and implications for fu- ture research and policy analysis. In Search of Pollution Havens? 127 Our cross-country analysis has found a pattern However, the evidence suggests that something of evidence which does seem consistent with the pol- like a pollution havens effect is real, even if it has been lution havens story: Pollution-intensive output as a transient for many countries. The results, however, percentage of total manufacturing has fallen consis- also suggest there will be some countries that lag be- tently in the OECD and risen steadily in the develop- hind in their efforts to control pollution now and may ing world. Moreover, the periods of rapid increase in even take years to catch-up with the rest of the world. net exports of pollution-intensive products from de- This also rises serious issues about continued exist- veloping countries coincided with periods of rapid ence of polluted waterways and lands left behind by increase in the cost of pollution abatement in the itinerant dirty industries, and their legacy will remain OECD economies. for generations. However, our evidence also shows that pollu- What, if anything, then should the industrial tion haven effects have not had major significance, countries or rest of the world do about this disparity? for several reasons. First, consumption/production Our results cast strong doubt on the wisdom of inter- ratios for dirty-sector products in the developing vention through trade-related measures. The contin- world have remained close to unity throughout the uous, smooth relationship between income growth period; most of the dirty-sector development story is and eniota performanc e s a op- strictly domestic. Second, a significant part of the in- crease in dirty-sector production share in the devel- ing countries are already making social choices which oping regions seems due to a highly income-elastic reflect the calculus of benefits and costs. Countries demand for basic industrial products. With contin- become less polluted as rising incomes makes a clean- ued income growth, this elasticity has declined.9 er environment more desirable and affordable. For- Third, some portion of the international adjustment tunately, there are progressive alternatives to has probably been due to the energy price shock and heavy-handed intervention: The positive alternative the persistence of energy subsidies in many develop- to heav-handed intervention lies in aiding activities ing countries. Thessubidiehaebeto finance pollution control training; the transfer of mg countries. These subsidies have been on the wane for a decade. Finally, environmental regulation in- cost-effective pollution control technologies; and ap- creases continuously with income and seems to have propriate information systems for regulation and pub- played a role in the shift from dirty to cleaner sectors. lic dissemination of environmental information. At Thus, any tendency toward formation of a pol- each level of development, such assistance can help lution havens seems to have been self-limiting, be- developing countries move closer to locally-appropri- cause economic growth brings countervailing ate levels of pollution control. Ultimately, income pressure to bear on polluters through increased regu- growth will be the answer. As developing economies lation, technical expertise, and "clean sector" produc- prosper and tighten their regulations, we are confi- tion. In practice, pollution havens have apparently dent that the shadow of pollution havens will recede been as transient as low-wage havens. to insignificance. In closing, it is worth asking whether these re- sults are a cause for optimism or pessimism about the Notes relationship between economic development and en- Funding for this study has been provided by the vironmental quality. The appropriate answer seems World Bank"s Research Support Budget and the Neth- to be "both." It is conforting to see that industrial pol- erlands Trust Fund for "Social and Environmental lution levels off or declines in richer countries, because Consequences of Growth-Oriented Policies." The find- pollution intensity has an elastic response to income ings, interpretations, and conclusions expressed in this growth. In addition, our results suggest that pollu- paper are entirely those of the authors. They do not tion haven effects have been transient and relatively necessarily reflect the views of the World Bank, its unimportant. It is clear that no feasible trade policy Executive Directors, or the countries they represent. could "neutralize" this effect. Cross-country differenc- For useful comments, we would like to thank Judy es in environmental regulation reflect a broad contin- Dean, Per Fredriksson and other participants in the uum of experience, and the domestic impacts of workshop on Social and Environmental Consequenc- regulation dwarf intemational displacement impacts. es of Growth-Oriented Policies, World Bank, Wash- 128 Trade, Global Policy, and the Environment ington, DC, May 1997 and the Conference on Trade, Dasgupta, S., and D.R. Wheeler. 1997. "Citizen Com- Global Policy and the Environment, World Bank, plaints as Environmental Indicators: Evidence Washington, DC, April 1998. This chapter is published from China." Policy Research Department Work- with permission from Sage Publications, Inc. It has ing Paper 1704, Washington, DC: World Bank. been published in Journal of Environment and Develop- Dasgupta, S., A. Mody, S. Roy, and D.R. Wheeler. ment, Vol. 7, No. 3, September 1998, pp. 215-247. 1995. "Environmental Regulation and Develop- 1. For a detailed discussion, see Mody and Wheeler ment: A Cross-Country Empirical Analysis." (1990). Policy Research Department, Working Paper No. 1448, Washington, DC: World Bank. 2. See Wheeler and Mody (1992). Hettige, H., P. Martin, M. Singh, and D.R. Wheeler. 3. This is confirmed in a recent study by Hettige, 1995. "IPPS: The Industrial Pollution Projection Mani, and Wheeler (1998). System." Policy Research Department Working 4. Petroleum has been excluded because a very few Paper 1431, Washington, DC: World Bank. 4.oPetroleum hasrben actuall ued incause veroduction Hettige, H., M. Mani, and D.R. Wheeler. 1998. "In- countries are actually involved in its production. utilPluini cnmcDvlpet i ^ * ~~~~~dustrial Pollution in Economic Development: 5. We do not have complete data series for all Asian Kuznets Revisited." Policy Research Working Pa- countries for the entire period. Our Asia series in per 1876. Washington, DC: World Bank. figure 16 includes data for Korea, Singapore, Pa- Mani, M. 1996. "Environmental Tariffs on Polluting kistan, Philippines and India. Imports: An Empirical Study." Environmental and 6. For a detailed analysis, see Wheeler and Mody Resource Economics 7: 391-411 (1992). Mody, A., and D.R. Wheeler. 1990. Automation and World Competition: New Technologies, Industrial Lo- 7. Data were not available for the other Asian devel- cation, and Trade. London: Macmillan Press. oping economies. Pargal, S., and D.R. Wheeler. 1996. "Informal Regula- 8. Summers-Heston openness index is defined as tion in Developing Countries: Evidence from In- (exports+imports)/nominal GDP. donesia." Journal of Political Economy 104: 1314-27. Robison, D.H. 1988. "Industrial Pollution Abatement: 9. Dasgupta and others (1995) also find a very strong, h mato h aac fTae"Cnda monoone ncresingrelaionsip btwee na- The Impact on the Balance of Trade.' Canadian mnonotone increasing relationship between na- tional income per capita and the strictness of Journal of Economics 21: 702-706. environmental regulation. Tobey, J.A. 1990. "The Effects of Domestic Environ- mental Policies on Patterns of World Trade: An Empirical Test." Kyklos 43(2): 191-209. References Wang, H., and D.R. Wheeler. 1996. "Pricing Indus- trial Pollution in China: An Econometric Analysis Birdsall, N., and D.R. Wheeler. 1993. "Trade Policy of the Levy System." Policy Research Department and Industrial Pollution in Latin America: Where Working Paper No. 1644. Washington, DC: The Are the Pollution Havens." Journal of Environment World Bank. and Development 2(1): 137-49. Wheeler, D.R., and A. Mody. 1992. "International In- Brandon, C., and R. Ramankutty. 1993. Toward an vestment Location Decisions: The Case of U.S. Environmental Strategy for Asia. World Bank Dis- Firms." Journal of International Economics 33: 57-76. cussion Paper Number 224. Washington, DC: World World Bank. 1997. World Development Indicators 1997. Bank. Washington, DC: World Bank. Chapter 9 The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade Paavo Eliste and Per G. Fredriksson 1. Introduction tions associated with greater government assistance his paper attempts to explain the political de- to agricultural producers? (ii) Do greater government terminants of environmental regulations and support to polluters lead to more stringent environ- assistance programs for agriculture. The inter- mental regulations? To our knowledge this study is national trading system is heavily influenced by a the first attempt at explaining the empirical relation- variety of subsidy schemes in effect in many coun- ship between environmental regulations and transfer tries. We seek to analyze the relationship between such policies. transfers and environmental policies, which, in turn, We also incorporate the impact of farm lobby- should yield new insights into the potential joint ef- ing on environmental regulations and assistance pol- fect of these factors on patterns of international trade.1 icies, the deadweight loss of transfers, and the We argue that producer-support schemes may environmental pressures caused by agriculture. Such be intimately connected with environmental policies. variables address the question of whether the poli- Increased demand for environmental quality (ob- cies selected tend toward efficiency. Moreover, we served in many countries during recent decades) may investigate the effects of freedom of information and stimulate simultaneous increases in both environmen- political freedoms on environmental and assistance tal regulations and transfers. This would occur if pol- policies. luters received compensation, via transfers, for The empirical results support the first hypothe- expenditures incurred as a result of environmental sis: the stringency of environmental policies appear regulations.2 This is a political economy explanation to be positively associated with the size of transfers for why observed increases in the stringency of envi- in the agricultural sector. This, in turn, suggests that ronmental regulations have been found to have small, trade patterns may have been influenced by such com- insignificant, or even reverse effects on trade pat- pensatory transfers, meaning that the transfers may terns.3,4 have mitigated the effects of regulations on trade This study also examines whether government flows. Many forms of support given to producers to assistance serves as a "carrot" in environmental poli- compensate for environmental protection costs imply cymaking. Can distortionary trade and subsidy poli- inefficient environmental protection and trade pat- cies be justified on the basis that they improve welfare terns, since producers are not facing the true social by enabling policymakers to set more stringent envi- cost of the damage caused, that is, the polluter pays ronmental regulations? principle is violated.5 The following two hypotheses are tested empir- The results do not, however, indicate that gov- ically: (i) Are more stringent environmental regula- ernment assistance programs have a significant im- 129 130 Trade, Global Policy, and the Environment pact on the stringency of environmental regulations.6 ports compiled for the 1992 United Nations Confer- Thus in the aggregate, support programs designed to ence on Environment and Development (UNCED) in facilitate the strengthening of environmental protec- Rio de Janeiro. Each country report is based on iden- tion are hardly justifiable. In particular, programs that tical survey questions and was prepared under clearly have a negative effect on environmental quality defined UNCED guidelines, facilitating cross- should be targeted for a quick phaseout (see OECD country comparison (Dasgupta and others 1995). The 1998a). country reports provide sector specific information The results also indicate that the political power about the state of the environment and natural- of the farm lobby is positively related to the relative resource utilization by the agricultural sector, with a economic importance of the agricultural sector, but focus on existing environmental policies, legislation, negatively related to its members' share of the popu- control mechanisms, and enforcement. Usinlg the in- lation. Another finding is that land-use pressures re- formation gathered, a quantitative index of the strin- sult in more stringent environmental policies, gency of environmental regulations was developed indicating that a concern for efficiency is present. Fi- by Dasgupta and others for an initial set of 31 coun- nally, we find no significant impacts of freedom of tries. The answers to each of 25 questions were as- information and freedom of civil rights on environ- signed a score of zero, one, or two for each of four mental regulations and government transfers, respec- different forms of environmental degradation (air, tively. water, land, and living resources or biodiversity) and The next section describes the data and empiri- the scores were added to yield the index.7 cal specification, while section three examines the re- The pool of countries was extended further by sults, and section four offers a conclusion. our own work using the same methodology as Das- gupta and others. We believe this to be the largest 2. Empirical Analysis available cross-country measure of environmental pol- icies. It has the advantage of measuring the stringen- Our research was aimed attesting whether more strin- cy of environmental regulations in one sector gent environmental regulations are positively associ- (agric'ulture) alone, instead of the broad, aggregate ated with government assistance to agricultural measures used in many other studies. The remaining producers, and whether government support to these data sources are provided in the appendix, table A.9.1. producers enables policymakers to set more stringent environmental regulations. The potentially simulta- Endogenous variables neous nature of the relationship between environmen- tal regulations and transfers implies that environmental Thineofnvrmntlegainsdcibdn tal regulations amaythavefanseffectesonhtransfers,mwhich the previous section is used as a measure of the strin- regulatin tur ifee a tions. Th twransfers, are gency of environmental policy (STRING). Compen- may in tumninfluence regulations. The two effects are sation channeled to farmers in return for increased isolated by simultaneously estimating two equations. expenditures on environmental protection is delivered We estimate one equation for the stringency of envi- through a variety of policy intervenhons. Govern- ronmental regulations and another for the level of ments may use production subsidies, pollution abate- government assistance. We provide both ordinary- ment subsidies, or land set-aside schemes, as well as and two-stage least squares estimation results. import restrictions, tax concessions, or waste-manage- ment assistance for this purpose. Data on environmental regulationsmetastncfothsprs. The Producer Subsidy Equivalent (PSE) is used The empirical analysis uses cross-country data from here as a measure of aggregate government assistance 49 countries for the year 1990. Our measure of the in the agricultural sector. The PSE measures the mon- stringency of environmental regulations is a newly etary value of all transfers to the agricultural sector, designed index based on individual country reports including tariffs, nontariff barriers, and various ex- on the status and extent of environmental regulations plicit compensation schemes for environmental reg- in the agricultural sector. The index was initially con- ulations resulting from agricultural regulatory structed by Dasgupta and others (1995) based on re- policies.8 To reduce the effects of random volatility The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade 131 between years (resulting, for example, from large ex- tal pressure. Both can be expected to have a positive change-rate fluctuations), we use the average PSE for impact on STRING. the three years 1988 to 1990. Political attention to environmental problems can be expected to increase as objective information about The equationfor the stringency of environmental environmental damage becomes more publicly avail- regulations able (see Pargal and Wheeler 1996). Thus we expect- ed that variations in the stringency of environmental This section examines the exogenous determinants of re l ationscould he exlingedcy theavailitynof the stringency of environmental policy (STRING). We inforation Moreoerpotted Sloo ieot ta used the PSE to determine the effect of transfers on presenc of ellinormed eleoreplie a the level of environmental regulations. We had no the inflence of lobyoroups.eGa strong prior expectations regarding the effect of the (1987) developed an indexe of fom of.information. PSE on STRING. (97 eeoe nidxo reo fifrain PSE expected STRING. capita income GDPpctohave The original indicator is an integer: 1 represents the We expected per capita income (GDPpc) to have most free information and 5 the least free informa- a positive effect on STRING. If environmental quality tion. Freedom of information may not be measurable is a normal good, demand should increase along with on a nominal scale, as assumed in Gastil's index, how- income, so that per capita income can be viewed as a ever. We therefore defined a dummy variable (FOI- proxy for the political pressure from environmental dummy), following Murdoch and Sandler (1997). Our interests. variable has a value of 1 when Gastil's index is 2 or The total cost of environmental regulations to less (unrestricted diffusion of information) and 0 oth- producers should depend on the relative size of the erwise (restricted diffusion of information). We ex- agricultural sector. Thus the larger the output level pected that FOIdummy would have a positive impact of the agricultural sector relative to the rest of the econ- on STRING. omy, the more influential the farm lobby since more is at stake. In an extensive survey of the empirical in- The producer subsidy equivalent equation terest-group literature, Potters and Sloof (1996) find Considering the many empirical studies that have that the greater a lobby group's stake in the policy found minimal or no effects of environmental regula- outcome, the greater its success. Moreover, the larger tions on trade flows, we anticipated that STRING the relative size of the sector, the stronger the effect of would have a positive effect on the PSE. regulations on social welfare. We therefore hypothe- Transfer levels are partly determined by the eco- sized that the larger the share of value-added of agri- nomic and political incentives of lobby groups (Beck- culture to total GDP (AGDPsh), the lower the STRING. er 1983; Grossman and Helpman 1994; Aidt 1998; A substantial share of soil erosion and loss of Fredriksson 1998). An interest group's political influ- marginal ecosystems, biodiversity, and carbon sinks ence is enhanced by a common, concentrated, and comes from the conversion of land to agricultural pro- sufficiently large economic interest because free-riding duction. This has been labeled an "extensive pressure" is reduced (Olson 1965). We use the share of agricul- in the literature (Goklany 1996). Another form of pres- tural sector labor in the total labor force as a proxy for sure on the environment comes from more intensive the political power of the farm lobby group (AGLA- cultivation or grazing of existing farmland; for exam- BORsh). The expected effect of an increase in AGLA- ple, farm run-off of nutrients and sediments or "in- BpRsh is ambiguous because of the considerations tensive pressure." In sum existing pressures on noted below. environmental resources are important variables de- First, a small change in the PSE has a larger ag- termining the demand for environmental regulations. gregate effect on the farm lobby's welfare as AGLA- We use the share of agricultural land (defined as the BORsh increases, because of the relatively larger sum of arable and pasture land) from total land area number of transfer recipients. This translates into (AGLANDsh) as an indicator of "extensive" pressure greater lobbying pressure and voting power (Potters on the environment, and fertilizer use per hectare and Sloof 1996). Thus pressure for support increases (FERTph) as an indicator of "intensive" environmen- with AGLABORsh. Yet the farm lobby suffers from 132 Trade, Global Policy, and the Environment increased free-riding, so the pressure for assistance use Gastil's Freedom of Civil Rights as a proxy for decreases as the number of farmers increases (Olson the weight on social welfare relative to political con- 1965). Third, as the share of agricultural producers in tributions. Again, Gastil's index uses integers: 1 rep- the total labor force increases, it becomes more diffi- resents the most secure rights, and 6 the least secure cult and costly to transfer income to farmers. This is rights. We define a dummy variable (DEMOCRACY- the case because as farmers increase their share of the dummy), which takes a value of 1 when Gastil's index population, the share of the remaining population is 2 or less (politically free country) and 0 otherwise necessarily falls. Each unit of income extracted yields (not free) (see also Murdoch and Sandler 1997). If the a smaller marginal benefit to each farmer and a great- influence of the farm lobby falls relatively faster as er marginal cost to the remaining population. If the the level of democracy increases, DEMOCRACYdum- first effect is greater (smaller) than the sum of the sec- my will have a negative sign; if the environmental lob- ond and third effects, AGLABORsh has a positive (neg- by's influence falls faster, the sign will be positive. ative) impact on PSE. The discussion above implies that the equations The ability of producers to obtain transfers for the stringency of environmental regulations and through the political process is greater the lower the for government assistance are given by the following deadweight loss associated with the redistribution of two specifications, respectively, where subscript i public funds (Becker 1983; Gardner 1987; Grossman denotes individual country i, i = 1,...N: and Helpman 1994; OECD 1998b). The greater the elasticity of output supply the more costly it becomes STRING.=fil2 + y2PSE. +f12GDPpc. +J3,1AGDPsh, (1.1) to redistribute through production subsidies, because +J3i,AGLANDshi +f 15FOIdummyi + the supply price is more negatively affected, other things being equal. Thus the greater the elasticity of PSE i=J121 + 1STRINGi +fl22AGLABORshi + (1.2) supply, the lower the level of government assistance. Jl23SUPPLYelasticity, + J24 DEMOCRACYdummy, + e Since cross-country supply elasticity estimates are not available, the estimation relies on a proxy variable 3. Empirical Results (SUPPLYelasticity). We hypothesized that the differ- Our empirical work yields a number of results. For ence in the elasticity of supply across countries is re- the STRING equation (1.1) the OLS results in table 9.1 flected by the value-added per hectare of agricultural are discussed since this is our preferred specification, land. High per-hectare value-added should corre- and for the PSE equation (1.2) we discuss the het- spond to low elasticity of supply, because available . . g resources are allocated in a way that maximizes val- ue-added. A further increase in the subsidy should The STRING equation yield a lower marginal output response and a greater positive effect on the producer's net price. Therefore, The coefficient estimate for government assistance SUPPLYelasticity can be expected to have a positive (PSE) in the equation for STRING is highly insignifi- impact on the PSE. cant. This suggests that government-support policies The political economy approach to trade policy may have no effect on the stringency of environmen- pioneered by Grossman and Helpman (1994) views tal regulations in agriculture. One explanation may lobbying activities as more influential the lower the be that increased government assistance may yield priority that a government puts on aggregate social higher political pressures on environmental regula- welfare (that is, the welfare maximizing policy pack- tions from several directions, and the total effect may age) relative to political support. Since this should be be negligible because individual effects neutralize true for both farm and environmental lobby groups each other (see also Fredriksson 1999). This result has (who should push in opposite directions), an increase implications for environmental policymaking. Our in the relative weight of social welfare will bring the finding does not support the argument that assistance PSE closer to zero (from either a negative or positive policies are justified because they serve to increase value), so that the expected sign is ambiguous. We the stringency of environmental regulations. The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade 133 Table 9.1 Regression results for the stringency of environmental regulations (STRING)5 OLS Robust OLS Robust 2SLS Dependent Variable: STRING CONSTANT 79.884 79.884 79.996" (8.746) (12.070) (12.753) PSE 0.020 0.020 0.067 (0.263) (0.325) (0.485) GDPpc 0.004" 0.004o 0.004- (7.248) (6.659) (5.791) AGDPsh -0.562 -0.562 -0.529" (- 1.880) (-2.123) (-2.444) FERTph 0.002 0.002 0.021 (1.261) (1.418) (1.123) AGLANDsh 0.251" 0.251 0.249' (2.131) (2.528) (2.645) FOldummy 4.656 4.656 4.887 (0.655) (0.757) (0.882) Adjusted R2 0.83 0.83 0.83 No. of observations 49 49 49 s* significant at the 10 percent level; significant at the 5 percent level; significant at the I percent level. Table 9.1 Regression results for government assistance (PSE) OLS Robust OLS Robust 2SLS Dependent Variable: PSE CONSTANT -1.937 -1.937 -17.564 (-0.074) (-0.073) (-0.691) STRING 0.236 0.236 0.370 (1.205) (1.280) (1.947) AGLABORsh -0.682" -0.682' -0.596" (-2.351) (-1.918) (-2.050) SUPPLYelasticity 0.004 0.004 0.003' (1.679) (2.261) (2.964) DEMOCRACYdummy 8.707 8.707 3.650 (0.605) (0.665) (0.294) Adjusted R' 0.46 0.46 0.46 No. of observations 49 49 49 i significant at the 10 percent level; * significant at the 5 percent level; f significant at the I percent level. GDPpc, on the other hand, is highly significant with The coefficient for AGLANDsh has the expected the expected positive sign. This finding is consistent with positive sign and is significant. The coefficient for those of other studies, such as Antle and Heidebrink FERTph is insignificant, although it takes the expect- (1995) and Pargal and Wheeler (1996). AGDPsh has the ed positive sign. Using the Wald test to estimate the expected negative sign, and is statistically significant. combined effect of the two environmental pressure Thus an increase in the relative economic importance of variables, we found a (chi-squared) test statistic of 7.18 the agricultural sector translates into greater political indicating that AGLANDsh and FERTph are jointly influence, which has a negative effect on the stringency significant at the 5 percent level. This implies that of environmental regulations.10 political pressure for more stringent regulations in- 134 Trade, Global Policy, and the Environment creases with the degree of environmental degradation The SUPPLYelastIicity variable has the expected posi- (caused by agricultural activities). The policymaker tive sign, and is statistically significant. Finally, DE- sets more stringent regulations when the negative MOCRACYdummy is found to be positive but externalities are greater, suggesting that efficiency insignificant. This may reflect the ambiguity associ- considerations play a role in environmental policy ated with this variable. Increased democracy may determination. have a similarly negative effect on the ability of all Finally, FOldummy has the expected positive interest groups to buy influence. sign, but is insignificant. The insignificant result could be explained by the following factors. First, the non- 4. Conclusion point nature of some forms of pollution from agricul- Economic theory predicts that an increase in the strin- ture (as opposed, for example, to point-source air and gency of environmental regulations will reduce ex- water pollution) makes this form of environmental ports and increase imports. This paper offered an degradation more difficult to detect for the average explanation for the fact that previous empirical re- citizen, and the absence of information reduces polit- search has largely found no, or only weak, effects of ical pressure for action. Second, pollution from agri- environmental regulations on trade patterns (see Jaffe culture has a lagged effect on environmental quality and others 1995). We presented empirical evidence and human health, sometimes requiring years of ac- from the agricultural sector that support the hypoth- cumulation to result in damage. This particular type esis that producers who face more stringent environ- of information may not be readily available in rural mental regulations also receive compensation for the areas in many countries. Third, the result may reflect expenditures associated with environmental protec- the fact that many governments attach a relatively low tion. This implies that future tests of the effects of en- weight to the health of the rural population relative vironmental regulations on trade flows should include to urban dwellers, so that less funds may be directed measures of transfers. to monitoring activities in rural areas. The variable We found no support, however, for the hypoth- may also be imprecisely measured. esis that higher levels of assistance to agriculture in- crease the stringency of environmental regulations in The PSE equation this sector. It appears that, in aggregate, transfers tend We now turn to the equation for the PSE in table 9.1.2, to be ineffective as a "carrot" to encourage polluters where we discuss the 2SLS results.'" First, STRING is to accept more stringent environmental policies. One positively associated with the PSE. The farm lobby explanation may be that assistance programs have group appears to become relatively more powerful implications for the long-run influence of lobby as the stringency of regulations increases. The coeffi- groups. Government support policies may increase cient is significant at the 10 percent level (p-value 0.06); the relative political pressure of the farm lobby, for thus we reject the hypothesis that STRING has no example, because these measures increase the value effect on PSE. This lends some support to the hypoth- of total agricultural output. This, in turn, increases esis that polluters receive compensation for environ- the lobby's incentive to spend further resources on mental protection costs. Although we cannot infer that seeking political influence. the same phenomena has taken place in other sectors, This study also demonstrated the importance of this finding appears to provide initial evidence that other political factors and deadweight loss in the de- support policies may have neutralized the effects of termination of cross-country differences in both envi- environmental regulations on trade flows. ronmental and assistance policies in the agricultural The coefficient for AGLABORsh has a negative sector. We found that the relative economic influence sign and is significant at the 5 percent level. It appears of farmers, free-riding, and supply elasticity play sig- that the effects of free-riding and increased redistri- nificant roles. Finally, we found that greater environ- bution costs outweigh the effects of increased mem- mental pressures result in stricter regulations. bership numbers on political pressure from the farm This paper does not address the effect of the com- lobby as its membership increases. We are, however, bination of environmental policies and assistance not able to separate the former two individual effects. programs on aggregate pollution levels. The environ- The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade 135 mental effects of commodity programs are relatively sity), the results have implications for environmental well-known, see Just and Bockstael (1991). Since agri- policy (see, for example, Lewandrowski and others cultural production is a significant contributor to en- 1997). One possible implication of our results is that vironmental problems such as soil erosion, water the combination of environmental policies and asso- pollution, wetland losses, and reductions of wildlife ciated transfer policies may, in the aggregate, worsen populations (through stimulating excessive use of pes- environmental quality, even compared to a laissez- ticides and fertilizers, encouraging conversion of land faire approach. to agricultural production, and reducing crop diver- 136 Trade, Global Policy, and the Environment Appendix Table A9. I Variable definition and data sources Variable Definition and source STRING Index of stringency of environmental regulations. Source: Dasgupta et al. (1995) and authors' calculations based on UNCED (1992). PSE Producer Subsidy Equivalent as a percentage of the total value of production.a Sources: OECD (1997) for OECD countries; unpublished OECD data files for Czech Republic, Hungary, and Poland; Valdes (1996) for Latin American countries; USDA (1990, 1994a). PSE for Morocco was computed using unpublished USDA data files. GDPpc Gross domestic product per capita (1987 U.S. dollars). Source: World Bank (1992).b AGDPsh Agriculture value-added as a share of GD P (percent). Source: World Bank (1997a). FERTph Fertilizer consumption per hectare of arable land (kilograms) Source: World Bank (I 997a). AGLANDsh Agricultural land as a share of total land (percent). Source: World Bank (1997b). FOldummy Dummy variable for the Freedom of Information. Source: Scully (1992). AGLABORsh Share of labor force in agriculture (percent). Source: World Bank (1997a). SUPPLYelosticity Agricultural value-added per ha of agricultural land (1987 US dollars). Source: USDA (I 994b). DEMOCRACYdummy D ummy variable for the Freedom of Civil Rights. Source: Scully (1992). a. Data for Thailand and Venezuela is from 1987. b. 1987 local currency GDP was converted to 1987 US$ using average exchange rates for 1987. The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade 137 Notes 4. Other explanations include that of Porter and van We thank Torbjorn Becker, Cees van Beers, Peter den Linde (1995) who argue that environmental Christoffersen, Stratford Douglas, Jerry Fletcher, Bruce regulations encourage innovations that enhance firms' long-run competitiveness. Alternatively, Gardner, Leo Maler, Lisa Segnestam, Ronald Steenb- salaaeetcssrltv olbrcss o lik, David Wheeler, the participants at the Trade, Glo- s bal Policy, and the Environment conference and West intercountry differences in regulatory stringency, or the lack of enforcement, could explain the re- Virginia University Departments of Natural Resource tse laffendothers 1995). Economics and Economics, respectively, for helpful , ,. discussions and comments on earlier versions of this 5. Environmental quality is thus likely to be nega- paper, and USDA for making data sources available tively affected (see Just and Bockstael 1991 for a for this project. Funding from the Swedish Interna- discussion of the environmental effects of com- tional Development Cooperation Agency (Sida), with- modity programs). Many subsidy schemes have a in the project "Trade, Macroeconomic Reform, and the positive output effect and cause problems such as Environment" is gratefully acknowledged. soil erosion and water pollution from fertilizer and pesticide leakage. This implies that increased at- 1. Jaffe and others (1995) provide a survey of the lit- tention should be paid to the effects on the erature on the effects of environmental regulations environment of the comprehensive policy pack- age in place. Compensation policies aimed at 2. Eliste and Fredriksson (1998) develop a theoreti- reducing trade effects may have important impli- cal model which shows that an increase in the cations for environmental quality. However, we demand for environmental quality, which results abstract from an explicit analysis of this issue in in a higher equilibrium pollution tax, also affects the present paper. special interest groups' marginal incentive to in- 6. OECD (1995, p. 29) reports that Austria, Finland, fluence the government's subsidy policy. The and Sweden have introduced fertilizer and pesti- subsidy may increase simultaneously with the tax cide taxes and levies, and the revenues have been as a result of changing relative political pressures used to subsidize crops and exports. In the United from environmental and farmer lobby groups, States, the 1985 and 1990 Farm Bills and the Con- given taxpayers' political pressure. The subsidy servation Reserve Program (CRP) are primary enables the government to respond to an increase forms of environmental programs in agriculture. in the demand for environmental quality without Unless farmers refrain from cultivating highly ero- a large drop in producer support. sive areas, all farm program payments are lost by 3. Recent empirical evidence from the manufactur- the farm. Gardner (1996) reports that CRP has idled ing sectors indicates that more stringent envi- 36.5 million acres, approximately 10 percent of U.S. ronmental policies may have counter-intuitive ef- cropland acreage. In the United Kingdom farmers fects on trade patterns (van Beers and van den receive direct compensation for adopting subop- Bergh 1997). Their results suggest that more strin- timal agricultural practices in "nitrate sensitive" gent environmental regulations may have given areas (Blom 1996). OECD (1998a, p. 55) reports that rise to reduced imports, which is opposite to what in Denmark, Germany, and Sweden payments to is predicted by conventional models of compara- prevent farmers from adopting harmful practices tive advantage. Moreover, Grossman and Krueger often have not led to significant changes in farm- (1993) find that tariff-free exports to the United ing practices. See also Bower and others (1981) and States from the Mexican maquiladora sector were Jenkins and Lamech (1994) for industrial subsidy higher in sectors with relatively high abatement policies. costs in the United States. Kalt (1988) and Tobey 7. Our analysis is limited to a single cross-section be- (1990) found insignificant effects of environmen- cause we have not been able to find or construct tal regulations. However, Low and Yeats (1992) an index of environmental regulations for multiple and Hettige and others (1992) found evidence of a time-periods, precluding a more extensive analy- migration of polluting industries from OECD to sis of the dynamic interactions between regu- developing countries. lations and transfers in agriculture. 138 Trade, Global Policy, and the Environment 8. The PSE is calculated by adding indirect transfers Belsley, D.A. 1992. "Assessing the Presence of Harm- from the government (for example, budget and tax ful Collinearity and other Forms of Weak Data expenditures) to implicit transfers from consum- through a Test for Signal-to-Noise." Journal of ers to producers. The sum is measured by mul- Econometrics 20(2): 211-53. tiplying the volume of domestic production by the Blom, J.C. 1996. "Environmental Policies in Europe gap between the domestic producer price and the and the Effect on the Balance of Trade." In M.E. border price. Bredahl and others, eds., Agriculture, Trade, and the Environment: Discovering and Measuring the Criti- 9. We employed the Breusch-Pagan Lagrange Mul- cal Linkages. Boulder, Colo.: Westview Press. tiplier test for heteroscedasticity applied to the Bower, B.D., R. Barre, J. Ktihner, C.S. Russell, and A.J. structural residuals. The test statistic was equal to Price. 1981. Incentives in Water QualityAManagement: 18.91 for equation (1.2) which implies that we re- France and the Ruhr Area. Washington, D.C.: Re- ject the null hypothesis of homoscedasticity at the sources for the Future. 5 percent level of significance (p-value 0.015). The Dasgupta, S., A. Mody, S. Roy, and D. Wheeler. 1995. test did not indicate a heteroscedasticity problem "Environmental Regulation and Development: A in equation (1.1). We also report the White Cross-Country Empirical Analysis." Policy Re- heteroscedasticity-adjusted estimates for this equa- search Working Paper 1448, Policy Research De- partment, World Bank, Washington, D.C. tion (White 1980). Using a Hausman test, we could patet Wol Bak'ahngo,DC notonjet (Whie 1980).usingl a hypohasmano tnest, ou t Eliste, P., and P.G. Fredriksson. 1998. "Environmen- not reject the null hypothesis of exogeneity of the tal Regulations, Transfers, Institutions, and Trade: PSE variable in equation (1.1). Therefore, the 0LS Theory and Evidence." Mimeo, Environment De- results in table 9.1.1 are our preferred estimates. partment, World Bank, Washington, I)C. The Hausman test rejected the exogeneity hypoth- Fredriksson, P.G. 1997. "The Political Economy of Pollu- esis for STRING in equation (1.2), however (the tion Taxes in a Small Open Economy." Journal ofEnvi- chi-square test statistic equals 7.95), and therefore ronmental Economics and Management 33(1): 44-58. the heteroscedasticity-robust 2SLS results are the _ . 1998. "Environmental Policy Choice: Pollution preferred estimates in table 9.1.2. Abatement Subsidies." Resource and Energy Eco- nomics 20(1): 51-63. 10. Lobbying on pollution taxes also appears to be im- ,, 209lTh P i c o al portant in the industrial sector; see for exarnple,199ThPoicaEonmofrdeLbal portant n the inustrialsector; ee for eample, zation and Environmental Policy." Southern Eco- the discussion in OECD (1998b, p. 30) on tax breaks izaJonal 65(3):o513-25. s ~~~~~~nomic journal 65(3): 513-25. for energy-intensive sectors in Sweden. Gardner, B.L. 1987. "Causes of U.S. Farm Commodity 11. Collinearity diagnostics was performed on the re- Programs." Journal of Political Economy 95: 290-310. duced form equations, and our test shows that the - . 1996. "Environmental Regulation and the Com- linear specification does not suffer from multi- petitiveness of U.S. Agriculture." In M.E. Bredahl, index N. Ballanger, J.C. Dunmore, and T.L. Roe, eds., corieart0 teeg and conditionspectively).SeeBelsley (1982Agriculture, Trade, and the Environment: Discover- are 0.03 and 12.78, respectively). See Belsley (1982). in an Mesrn th CrtclLnae.Budr ing and Measuring the Critical Linkages. Boulder, Colo.: Westview Press. References Gastil, R.D. 1987. Freedom in the World. Westport, Conn.: Greenwood Press. Aidt, T.S. 1998. "Political Internalization of Economic Goklany, I.H. 1996. "Factors Affecting Environmen- Externalities and Environmental Policy." Journal tal Impacts: The Effect of Technology on Long-term of Public Economics 69(1): 1-16. Trends in Cropland, Air Pollution, and Water Re- Antle, J.H., and G. Heidebrink. 1995. "Environment G an Ed a . Heipman. 1994." e an Deeomn:Ter'ndItrainlEi Grossman, G.E., and E. Helpman. 1994. "Protection and Development: Theory and International Evi- for Sale." American Economic Review 84(4): 833- dence." Economic Development and Cultural Change 50. 43(1): 603-25. Grossman, G.M. and A.B. Krueger. 1993. "Environ- Becker, G. 1993. "A Theory of Competition Among mental Impacts of North American Free Trade Pressure Groups for Political Influence." Quarterly Agreement." In P. Garber, ed., The U.S.-Mexico Free Journal of Political Economy 98: 371-400. Trade Agreement. Cambridge, Mass: MIT Press. The Political Economy of Environmental Regulations, Government Assistance, and Foreign Trade 139 Hettige, H., R.E.B. Lucas, and D. Wheeler. 1992. "The Pargal, S., and D. Wheeler. 1996. "Informal Regula- Toxic Intensity of Industrial Production: Global tion in Developing Countries: Evidence from In- Patterns, Trends, and Trade Policy." American Eco- donesia." Journal of Political Economy 104:1314-27. nomic Review Papers and Proceedings 82: 478-81. Porter, M., and C. van den Linde. 1995. "Toward a Jaffe, A.B., S.R. Peterson, P.R. Portney, and R. Stavins. New Conception of the Environment-Competitive- 1995. "Environmental Regulation and the Com- ness Relationship." Journal of Economic Perspectives petitiveness of U.S. Manufacturing." Journal of Eco- 9(4): 119-32. nomic Literature 33(1): 132-63. Potters, J., and R. Sloof. 1996. "Interest Groups: A Sur- Jenkins, G.P., and R. Lamech. 1992. "Fiscal Policies to vey of Empirical Models That Try to Assess Their Control Pollution: International Experience." Influence." European Journal of Political Economy 12: Harvard Institute for International Development, 403-42. Development Discussion Paper 439. Cambridge, Scully, G. W. 1992. Constitutional Environments and Eco- Mass: HIID. nomic Growth. Princeton, New Jersey: Princeton Just, R.E., and N. Bockstael, eds. 1991. Commodity and University Press. Resource Policies in Agricultural Systems. UnvrirPes Besourli andiNew Yor: Sringetur-VSysterla. Tobey, J.A. 1990. "The Effects of Domestic Environ- Berlin and New York: Springer-Verlag. mna oiiso atrso ol rd:A Kalt, J.P. 1988. "The Impact of Domestic Environmen- mental Policies on Patterns of World Trade: An tal Regulatory Policies on U.S. International Com- Empirical Test." Kyklos 43: 191-209. petitiveness." In A.M. Spence and H.A. Hazard, eds., International Competitiveness. Cambridge, Geneva: United Nations. Mass: Harper and Row, Ballinger. USDA. 1990. "Estimates of Producer and Consumer Lewandrowski, J., J. Tobey, and Z. Cook. 1997. "The Subsidy Equivalents: Government Intervention in Interface between Agricultural Assistance and the Agriculture 1982-87." Statistical Bulletin 803. Eco- Environment Chemical Fertilizer Consumption and nomic Research Service. Washington, D.C.: U.S. Area Expansion." Land Economics 73(3): 404-27. Department of Agriculture. Low, P., and A. Yeats. 1992. "Do 'Dirty' Industries .1994a. "Government Intervention in Agricul- Migrate?" In P. Low, ed., International Trade and ture 1982-92." Statistical Bulletin 913. Economic Re- the Environment. World Bank Discussion Paper 159. search Service. Washington, D.C.: U.S. Department Washington, D.C. of Agriculture. Murdoch, J. C., and T. Sandler. 1997. "The Voluntary . 1994b. World Agriculture: Trends & Indicators, Provision of a Pure Public Good: The Case of Re- 1961-91. CD-ROM. Departnent of Economics and duced CFC Emissions and the Montreal Protocol." Statistics. Washington, D.C.: U.S. Department of Journal of Public Economics 63: 331-49. Agriculture. OECD. 1995. Environmental Taxes in OECD Countries. Valdes, A. 1996. "Surveillance of Agricultural Price Paris: Organisation for Economic Co-operation and and Trade Policy in Latin America During Major Development. Policy Reforms." World Bank Discussion Paper .1997.ProducerSubsidyEquivalentsandConsumerSub- 349. Washington, D.C.: World Bank. sidy Equivalents Database, 1979-1996. Paris: Organisation van Beers, C., and J.C.J.M. van den Bergh. 1997. "An for Economic Co-operation and Development Empirical Multi-Country Analysis of the Impact -___ 1998a. The Environmental Effects ofReforming Ag- of Environmental Policy on Foreign Trade Flows." ricultural Policies. Paris: Organisation for Economic Co-operation and Development. Kyklos 50(1): 29-46. _. 1998b. Improving the Environment Through Re- White, H. 1980. "A Heteroscedasticity-Consistent Co- ducing Subsidies, Part II. Paris: Organisation for variance Matrix Estimator and a Direct Test Economic Co-operation and Development. for Heteroscedasticity." Econometrica 48: 817-38. Olson, M. 1965. The Logic of Collective Action. Cam- World Bank. 1992. World Tables. Washington, D.C.: bridge: Harvard University Press. World Bank. Opschoor, J.B., and H.B. Vos. 1989. Economic Instru- .1997a. World Development-Indicators 1997. Wash- ments for Environmental Protection. Paris: ington, D.C.: World Bank. Organisation for Economic Co-operation and De- .1997b. Economic and Social Database. Washing- velopment. ton, D.C.: World Bank. Chapter 10 Pollution and Capital Markets in Developing Countries Susmita Dasgupta, Benoit Laplante, and Nlandu Mamingi 1. Introduction It is indeed generally said that firms in develop- hough environmental regulations have now ing countries do not have incentives to invest in pol- been in use for more than 20 years, it is in- lution control effort because of weak monitoring and creasingly recognized that their efficacy in enforcement of the environmental regulations. This controlling pollution emissions has been dampened argument however assumes that the environmental by a lack of appropriate monitoring and enforcement. regulator is the only agent that can penalize the firm Resources devoted by various regulatory agencies to lacking pollution control effort, or reward the firm for the monitoring of emission standards have typically good environmental performance or innovation in been characterized as insufficient.' Moreover, when environmental technologies. It ignores that capital compliance with the standards is found to be lacking, markets may react negatively to the announcement of it is generally acknowledged that fines and penalties adverse environmental incidents (such as violation of are too low (compared to pollution abatement costs) permits, spills, court actions, complaints, and so forth) to act as effective deterrents. In a recent study of en- or positively to the announcement of greater pollution vironmental regulations in East Asian countries, control effort such as the adoption of cleaner technol- O'Connor (1994, p.94) writes: ogies. The impact of firm-specific environmental news In several of the countries studied on market value may work its way through various here,2 the monitoring problem is channels: a high level of pollution intensity may sig- compounded by weak enforcement. In short, when violators of standards nal to investors the inefficiency of the firm's produc- are detected, if penalized at all they tion process; it may invite stricter scrutiny by envi- often face only weak sanctions. (...) ronmental groups and/or facility neighbors; it may polluters are exempted from fines ei- result in the loss of reputation, goodwill, and so forth. ther on grounds of financial hardship On the other hand, the announcement of a good envi- or because the violators wield undue ronmental performance or of the investment in clean- political influence. Perhaps the most er technologies may have the opposite effect: lesser pervasive problem is that, even when scrutiny by regulators and communities (including the fines are levied, they are frequently so low in real terms that they have fmancial community), greater access to international little if any deterrent value. In virtu- markets, and so forth.3 ally all the countries studied, there Hence, the inability of institutions in develop- remains considerable room for im- ing countries to provide incentives for pollution con- provement on the enforcement front. trol effort via the traditional channel of fines and 141 142 Trade, Global Policy, and the Environment penalties may not be as serious an impediment to pol- p. 480) observe that "one possibility is that the expect- lution control as is generally argued. Capital markets, ed profitability of investment in pollution-intensive if properly informed, may provide the appropriate sectors has also been affected by growing concern over reputational and financial incentives. legal liability or reputational damage." Where tradi- A limited number of papers have analyzed the tional tools and actions may have been unable to cre- reaction of capital markets to environmental news in ate incentives for pollution control, our results give Canada and the United States. These studies have some support to this point of view to the extent that generally shown that firms suffer from a decline in capital markets may reward firms with good environ- market values upon announcement of adverse envi- mental performance and penalize firms with poor en- ronmental news.4 In this paper, we assess whether or vironmental performance. not capital markets in Mexico, Chile, Argentina, and In the next section, we describe our dataset. In the Philippines react to the announcement of firm- section 3, we briefly describe the event-study meth- specific environmental news. To our knowledge, the odology used in this analysis to measure the reaction current analysis is the first of this nature performed of capital markets to environmental news (both posi- in developing countries. Even in those countries where tive and negative news). Results are presented in sec- it is generally argued that the environmental regula- tion 4. We briefly conclude in section 5. tions suffer from poor implementation, we show that capital markets react negatively (decrease in firms' 2. Dataset value) to citizens' complaints targeted at specific firms. We also show that markets react positively (increase The countrieshretained pints - argcnti- in firms' market value) to the announcement of re- na, Chile, Mexico, and the Philippines - are countries wards and explicit recognition of superior environ- where stock markets are believed to work reasonably mental performance. An immediate policy implication well, where market capitalization is relatively high from the current analysis is that environmental regu- and increasing over time (table 10.1), and where mar- lators in developing countries may explicitly harness ket concentration is not an impediment to conduct- those market forces by introducing structured pro- ing event-study analyses (table 10.2).7 grams of information release on firms' environmen- For each country, we selected a newspaper tal performance, and empower communities and which has a large circulation and is of particular in- stakeholders through environmental education pro- terest to the business community.8 Environmental grams.5 news were collected in each of the countries over the These results may also shed some new light on period 1990-94 inclusively. Once these news were col- the pollution haven hypothesis. A large number of lected, we identified those articles involving firms studies have examined the potential impact of envi- traded in local capital markets. As shown in table 10.3, ronmental regulations on international competitive- the amount of environmental news (that is, newsclips) ness.6 Many of these have concluded that pollution collected in each country is relatively large (a total of intensive firms have not invested or relocated in 7,354 environmental news were collected over the developing countries to benefit from lower period 1990-94), with Mexico alone representing 47.5 environmental standards and/or poor enforcement of percent of the total number of news. The number of environmental regulations. Hettige and others (1992, environmental news is also relatively constant over Table 10.1 Capitalization of the stock market of Argentina, Chile, Mexico, and the Philippines, 1990-1994 (in millions of U.S. dollars) Market 1990 1991 1992 1993 1994 Argentina 3 268 18 509 18 633 43 967 36 864 Chile 13 645 27 984 29 644 44 622 68 195 Mexico 32 725 98 178 139 061 200 671 130 246 Philippines 5 927 10 197 13 794 40 327 55 519 Source: International Finance Corporation, Emerging stock markets factbook, 1995. Pollution and Capital Markets in Developing Countries 143 Table 10.2 Market concentration in the IFC general indexes, end-1994 IFCG Index share of IO largest stocks' share of Market total market capitalization total market capitalization Argentina 50.9 41.7 Chile 66.1 46.4 Mexico 63.9 33.8 Philippines 54.4 44.3 Source: International Finance Corporation, Emerging stock markets factbook, 1995. the period of analysis. Approximately 20 percent of ed 18 events (10 positive and 8 negative) with 10 firms. the news involve specific firms, traded and non- Finally, the Mexican sample consists of 35 events (of traded. As expected, the number of news involving which only 4 were positive) involving 10 publicly- publicly traded companies is relatively small in all traded firms. Observe that the number of events in countries. However, publicly traded companies rep- table 10.4 is smaller than the number of news (with resent a much larger share of the number of compa- name of publicly traded companies) in table 10.3. This nies cited in environmental news than their relative is the case since a significant number of newsclips is numbers in the economy. This may be explained by simply a repetition or follow-up on an initial event their generally larger size, thus being of greater and does not provide any additional information to scrutiny. what is already known. In most cases, we have in- Environmental news were divided into two cluded in our dataset only the announcement of the groups: positive (for example rewards, investment in initial event. pollution control, and so forth), and negative (spills, complaints, warnings, and so forth). The sample set 3. Event-Study Methodology is described in table 10.4. As can be observed, Chile The event-study methodology is used in this study to registered 53 events (environmental news) involving examine the reaction of investors to positive and neg- 17 publicly traded firms over the period 1990-94; 20 ative news (also called events).9 The methodology is of those events were positive while 33 were negative. based on the assumption that capital markets are suf- Argentina registered 20 events (5 positive and 15 neg- ficiently efficient to evaluate the impact of new infor- ative) involving 11 firms. The Manila Bulletin report- mation (events) on expected future profits of the firms. Table 10.3 Number of news, 1990-1994 1990 1991 1992 1993 1994 Argentina Total number of environmental news 201 189 168 198 170 With name of non-traded companies 28 32 48 33 27 With name of publicly traded companies 0 0 2 13 15 Chile Total number of environmental news 309 285 293 282 272 With name of non-traded companies 29 48 43 22 32 With name of publicly traded companies 4 25 34 36 16 Mexico Total number of environmental news 625 707 759 613 618 With name of non-traded companies 161 143 118 73 88 With name of publicly traded companies 1 4 25 7 10 8 Philippines Total number of environmental news 317 309 334 265 266 With name of non-traded companies 54 47 44 47 55 With name of publicly traded companies 8 8 4 9 12 144 Trade, Global Policy, and the Environment Table 10.4 Description of data set Nature and number of events Country Name offirm Sector of activity Positive | Negative Argentina Astra Oil I I Ipako Oil 1 2 Perez Oil 0 2 YPF Oil 1 4 Celulosa Pulp and paper I 0 Telefonica Telephone 0 1 Colorin Chemical 0 2 Indupa Chemical I 0 Molinos Rio Food 0 1 Sevel Metal 0 1 Siderca Metal 0 1 Total II firms 6 sectors 5 15 Chile Endesa Electric 3 4 Chilgener Electric 4 4 CMPC Pulpand paper 2 1 CAP Metal 3 4 Volcan Building material 0 1 Minera Investment 0 1 Vapores Transportation 0 1 Emos Water 3 1 Puerto Water 0 1 Victoria Fabric 0 1 lansa Food I I Molymet Metal I I Coloso Fishery 0 5 Iquique Fishery 1 5 Lirquien Building material 0 1 Chilectra Electric I I Eperva Fishery I 0 Total 17 firms 10 sectors 20 33 Mexico Cydsasa Pulp and paper, oil 1 3 Grupo Maya (A) Cement 0 6 Grupo Maya (B) Cement 0 4 Tolteca (Tolmex) Cement 0 2 Met-Mex Penoles (A) Mining 1 6 Met-Mex Penoles (B) Mining 0 3 Femsa Food I 0 Grupo Vitro Manufacture I 0 GC3 Cement 0 1 Kimberly y Clark Pulp and paper 0 2 Grupo Bimbo Food 0 2 Telefonos de Mexico Communication 0 2 Total 10 firms 8 sectors 4 31 Philippines Apex Mining Mining 0 1 Atlas C. Mining Mining I 0 Ayala Land, Inc. Property 0 1 Benguet Mining 3 2 Jolibee Food I 0 Lepanto Mining 0 1 Manila Mining Mining I 0 Mondragon Trading 0 1 San Miguel Food 4 1 Robinson Land Property I 0 Total 10 firms s sectors 10 8 Complete names of firms appear in the Appendix. Pollution and Capital Markets in Developing Countries 145 It involves the following steps: (1) identification of the For each individual event, one can estimate the events of interest and definition of the event window;'0 abnormal return and relevant test statistics at each (2) selection of the sample set of firms to include in instant in time within the event window. However, the analysis;"1 (3) prediction of a "normal" return dur- in order to draw overall inference on the abnormal ing the event window in the absence of the event; (4) return observations for the event(s) of interest, one estimation of the abnormal return within the event can also aggregate the abnormal returns. For any giv- window, where the abnormal return is defined as the en subset of N events (or securities), the sampled ag- difference between the actual and predicted returns; gregated abnormal returns (AAR,) at each instant t and (5) testing whether the abnormal return is statis- tically different from zero. Several methods may be used to obtain to estimate abnormal returns: the sin- gle-index model (constant mean return model), the AAR, =- AR N market model and the capital asset price model i= (CAPM) are the most widely used. For large the variance is The market model assumes a linear relationship between the return of any security to the return of the IV market portfolio: VAR(AAR,) = - a' (5) Rith = a = O A R and Var(e ) = 2 (1) To test for the significance of AAR, a Z (or t) test can be derived. where t is the time index, i = 1,2,...,N, stands for secu- In order to test for the persistence of the impact rity, R± and Rare the returns on security i and the of the event during a period (T2 - T,), the abnormal p rt mt return can be added to obtain the cumulated abnor- market portfolio respectively duringperiodtande mal returns (CARi (T T )) for securityi over the period is the error term for security i. Equation (1) is generally estimated over a peri- (T2 - TO: od which runs between 120 and 210 days prior to the T event up to 10 days prior to the event. The event win- CAR, (7; I,T2 ARi, (6) dow is defined as the period from 10 days prior to the event to 10 days after the event. With the estimates of ct, and A from equation (1), one can predict a "nor- where 7; S IHAF(v) for 1 < v < N- B ( = Vl/(a-l) (A.5a) 1; free-riders do better than the countries that supply A (V) O+by the public good. However, eqs. (A.4) are cumbersome ri (v) = (N - v) + bv (A.5b) and further analytical results are hard to obtain. Since p the model is so special anyway, I consider simulations in the main text of the paper. where the superscript B denotes a trade ban. By ap- propriate choice of parameter values, the loss in the A.3 Trade Ban gains from trade will be large relative to the cost of providing the public good. So (A.5b) will exceed (A.5a) If trade between the countries thiat play Abate and when v is "small" and (A.5a) will exceed (A.5b) when Pollute is banned, then each of the former countries v is "large." will consume cAA = ,lv and each of the latter cpp = / (N-v). Upon substituting into (A.1) we get 172 Trade, Global Policy, and the Environment Notes Bhagwati,J., and T.N. Srinivasan. 1996. "Trade and the Environment: Does Environmental Diversity De- I am grateful to Per Fredriksson, David Roland-Holst, trc formethe Cas forF eetad InversBhagDeti and other participants at the World Bank's Confer- tract form the Case for Free Trade?a In J. Bhagwati ence on Trade, Global Policy, and the Environment, ' R April1998, for helpful comments. This paper was also Prerequisites for Free Trade? Volume 1: Economic presented at the NBER-Yale Global Change Snowmass Analysis p. 159-223. Cambridge, MA: MIT Press. Workshop, August 1998, and I am also grateful to Brack. 1996. International Trade and the Montreal Proto- Jagdish Bhagwati and other participants at this work- col. London: Chatham House. shop for their comments. Carraro, C., and D. Siniscalco. 1997. "R&D Coopera- tion and the Stability of International Environmen- 1. This is an issue-specific linkage. Carraro and tal Agreements." In C. Carraro, ed., International Siniscalco (1997) consider linking R&D coopera- Environmental Negotiations p. 71-96. Cheltenham: tion in an international environmental agreement. Edward Elgar. See also Cesar and de Zeeuw (1994) and Folmer, Cesar, H., and A. de Zeeuw. 1994. "Issue Linkage in van Mouche, and Ragland (1993) for alternative Global Environmental Problems." Nota ke Lavoro analyses of linkage. 56.94, Fondazione Eni Enrico Mattei, Milano. 2. I consider here only trade leakage in goods mar- Esty, D. C. 1994. Greening the GATT: Trade, Environ- kets. For an analysis of trade leakage in the market ment, and the Future. Washington, DC: ][nstitute for for traded fuels, see Hoel (1994). International Economics. Fisher, B. S., S. Barrett, P. Bohm, M. Kuroda, J.K.E. 3. I am ignoring here the optimal tariff that would ac- Mubazi, A. Shaw, and R.N. Stavins. 1996. "An Eco- company these adjustments. See Markusen (1975). nomic Assessment of Policy Instruments for Coi- 4. It is often claimed that weaker environmental stan- batting Climate Change." In J.P. Bruce, H. Lee, and dards confer a competititive advantage and that E.F. Haites, eds., Climate Change 1995 p. 397-439. fair trade demands countervailing remedies that Cambridge: Cambridge University Press. level the playing field of international competition. Folmer, H., P.V. van Mouche, and S. Ragland. 1993. In Barrett (1994b), I identify instances in which a "Interconnected Games and International Environ- country may want to weaken its environmental mental Problems. "Environmental and Resource Eco- policy for reasons of competitiveness, but I argue nomics 3: 313-36. that these cases are special and that countervailing GATT Secretariat. 1992. International Trade 90-91. measures are more likely to be welfare-reducing Geneva: GATT. than welfare-enhancing. See also Bhagwati and Hoel, M. 1994. "Efficient Climate Policy in the Pres- Srinivasan (1996), Esty (1994), and Uimonen and ence of Free Riders." Journal of Environmental Eco- Whalley (1997). nomics and Management 27: 259-74. Hoel, M. 1996. "Should a Carbon Tax be Differenti- References ated Across Sectors?" Journal of Public Economics 59: 17-32. Barrett, S. 1990. "The Problem of GlobalEnvironmen- Krugman, P. 1991. "Is Bilateralism Bad?" In E. tal Protection." Oxford Review of Economic Policy Helpman and A. Razin, eds., International Trade and 6: 68-79. Trade Policy p. 9-23. Cambridge, MA: MIT Press. Barrett, S. 1994a. "Self-Enforcing International Envi- Markusen, J.R. 1975. "International Externalities and ronmental Agreements." Oxford Economic Papers Optimal Tax Structures." Journal of International 46: 878-94. Economics 5: 15-29. Barrett, S. 1994b. "Strategic Environmental Policy and Oliveira-Martins, J., J.-M. Burniaux, and J.P. Martin. International Trade." Journal of Public Economics 1992. "Trade and the Effectiveness of Unilateral 54: 325-38. C02 Abatement Policies: Evidence from GREEN." Barrett, S. 1997. "The Strategy of Trade Sanctions in OECD Economic Studies 19: 123-40. International Environmental Agreements." Re- Uimonen, P., and J.Whalley. 1997. Environmental Is- source and Energy Economics 19: 345-61. sues in the New World Trading System. New York: Barrett, S. 1998. "A Theory of International Coopera- St. Martin's Press. tion." mimeo, London Business School. Van Slooten, R. 1994. "The Case of the Montreal Proto- Benedick, R.E. 1991. Ozone Diplomacy. Cambridge, col." In OECD, Trade and the Environment: Processes MA: Harvard University Press. and Production Methods p. 87-91. Paris: OECD. Chapter 12 The Importance of Trade for the Ratification of the 1992 Climate Change Convention Per G. Fredriksson and Noel Gaston 1. Introduction plied with by ratifying countries. Although no nation T he United Nations Framework Convention on can be forced to strictly adhere to the conditions stip- Climate Change (FCCC) was made available ulated in any international environmental agreement, for signature at the Earth Summit held at Rio Carraro and Siniscalco (1992) and Barrett (1994) ar- de Janeiro in June 1992. It entered into force in March gue that due to their voluntary nature, such agree- 1994 after fifty countries had ratified the Convention.1 ments are self-enforcing.3 The Summit was a response to the growing weight of The primary objective of this paper is to study scientific evidence that the Earth may be affected by the determination of the propensity of countries to climate change in the coming decades.2 participate in global environmental policy making by As a framework treaty, the FCCC sets out prin- ratifying the FCCC. In particular, we are interested in ciples as well as general commitments to adopt na- when countries signed the FCCC, rather than whether tional programs for mitigating climate change and they signed. In a departure from the existing related promote the conservation of greenhouse gas "sinks," literature, we focus on the delay in the ratification of such as forests. The FCCC differentiates between de- an environmental treaty, that is, the duration of time veloping countries and former Eastern Bloc countries it takes for countries to formally take action.4 We ar- and OECD countries (the latter two groups are re- gue that countries that speedily ratify an environmen- ferred to as Annex One countries). By ratifying the tal treaty have a more intense preference for the Convention, the latter two groups committed them- provisions it contains. Thus, in our context the date selves to "aim" to reduce emissions individually or of eventual ratification proxies the relative intensity jointly to their 1990 levels by the year 2000. Develop- of preference for the provisions of the FCCC. A short- ing countries have no significant commitments un- p der the FCCC, but are responsible for cooperation and promotion of the diffusion of technologies and prac- provisions in its immediate national self-interest. In tices that restrict emissions, that is, there is a 'com- Barrett's (1990) terminology, signatories (and ratifi- mon but differentiated responsibility" of nations in ers, in our view) act as "abatement leaders" and the confronting the global environmental problem. The others as "abatement followers." Ratification delay OECD countries were obliged to make financial re- data of the type we use in this paper are informative sources and abatement technologies available to de- about the characteristics of "leaders" and "followers" veloping countries in order to meet these obligations. (that is, given that abatement takes place subsequent According to Chayes and Chayes (1991), the majority to ratification of the agreement). A characterization of international environmental agreements are com- of these countries may facilitate an improvement in 173 174 Trade, Global Policy, and the Environment the design of future international environmental was decided upon less than three years subsequent agreements (IEAs). to Rio. The timing of this process, with new rounds of Working with ratification or delay times not only negotiation occurring at such short time intervals, may avoids the problem of choosing an arbitrary cutoff have been an important reason for countries to have date to investigate determinants of treaty ratification, delayed their ratification and for the perceived sub- but also adds to our knowledge of the political pro- sequent failure of the FCCC. cess underlying treaty ratification. Those countries Unfortunately, slow ratification by major emit- that delay their ratification of a treaty do so, either ters and competitors is likely to either dampen the because they perceive the treaty provisions as too cost- enthusiasm of others to ratify the agreement, or alter- ly and severe or, lacking net benefits (or that it may natively, to reduce the interest in adhering to com- be strategically worthwhile to "hold-out," in which mitments made. The FCCC is a prime example of the case they risk non-implementation of the agreement). voluntary provision of a public good. There is scope Global environmental policy carried out by in- for free-riding both in the ratification and the imple- ternational treaties suffer from at least two drawbacks. mentation stages. A better understanding of the un- First, since participation is voluntary for each sover- derlying behavior of countries in this process is eign state, both in the ratification and the compliance important for the appropriate design of global poli- stages, an international environmental agreement is cies. Hopefully, this paper represents a step forward based on a consensus of all participants and must be in our understanding of why ratifying countries de- self-enforcing. According to Sand (1991), agreed upon viated from the FCCC agreement. environmental standards tend to result in the lowest As well as its unique focus on delays in the rati- common denominator. Barrett (1994) and Murdoch fication of an international environmental agreement, and Sandler (1997) argue that the Montreal Protocol the paper contributes to the policy debate and the lit- merely codified CFC reductions that the countries erature in several ways. We are particularly interest- would have undertaken in the absence of any agree- ed in the impact of the degree of openness to ment. Second, ratification of an international environ- international trade on the ratification process. Trade- mental treaty by parliament results in a time lag, related issues arose during the Protocol negotiations delaying the implementation of the provisions of the when it became apparent that some countries consid- treaty. In addition, the stipulation that multilateral ered that future mitigation measures could have agreements be ratified by a certain number of coun- strong trade impacts (see Assuncao 1997). Based on tries further slows implementation.5 While this con- the experience with the Convention on International dition is aimed at avoiding any free-riding by potential Trade in Endangered Species of Wild Flora and Fau- non-ratifiers, it delays the desired outcome to the time na (CITES) and the Montreal Protocol, some countries until the "slowest truck in the convoy" reaches its believed that trade restrictions might be used, both to destination.6 limit emissions and to induce non-compliant parties Sand (1991) sees time lags and treaty ratification to participate in the new Protocol. Profitable trading as one of the most serious drawbacks of the treaty relationships between nations therefore have impor- approach to global environmental policymaking. tant implications for a country's willingness to take Therefore, the question arises as to whether the tradi- up the obligations under the FCCC. tional approach to multilateral agreements constitutes Increased exposure to international markets may an effective way of undertaking international envi- yield a "race to the bottom," or "political drag" in the ronmental policy. Since the target of the FCCC was to enactment of environmental regulations (see Esty stabilize greenhouse gases within eight years, the ex- 1996; Esty and Geradin 1997; Stewart 1993). Conse- hibited ratification speed may have been important quently, countries may delay the ratification of inter- for the failure of many ratifiers to live up to the agree- national environmental agreements and the associated ment.7 Moreover, it is possible that countries behaved abatement efforts, for international "competitiveness" strategically looking ahead to Kyoto.8 The negotia- reasons. tions on global warming issues are clearly an ongo- In the next section, we describe the legislative ing process. The Kyoto meeting held in December 1997 delay time data. To our knowledge, this is one of the The Importance of Trade for the Ratification of the 1992 Climate Change Convention 175 first papers that uses data on the FCCC. Section 3 de- incomplete since some countries have not ratified the scribes our key hypotheses, and section 4 presents FCCC. That is, some of the data are "right-censored." results from a proportional hazards regression mod- This requires the use of special empirical techniques el. Section 5 provides some concluding observations. that have been developed to handle duration data. At A description of the empirical methodology and tech- the commencement date of the Kyoto meeting (De- nical details are left for the Appendix. cember 1,1997) twenty-one countries had still not rat- ified the convention. 2. The Data on Legislative Delay The total sample size consists of 184 countries. The main part of the data are for countries that were The few empirical studies that have preceded ours members of the United Nations (UN) as of June 1992. on the determinants of signing environmental trea- For countries that came into existence after June 1992, ties have investigated the signatories and non-signa- we compute the delay time from the date of that coun- tories at a particular point in time. Since the date of try's admission into the UN.9 ratification is a continuous variable, this introduces Figure 12.1 plots the (non-parametric) estimates an inevitable degree of arbitrariness in the choice of a of the survival distribution function for the delay time cut-off date for an investigation. It also reduces the data. This survivor function specifies the probability political environmental treaty ratification decision to that the delay time will exceed a certain number of a dichotomous choice. In contrast, we focus on delay days. The average delay time for the countries that times to ratification of the FCCC. ratified the FCCC is 810 days, while the median spell The FCCC was made available for ratification on length is 728 days (indicating skewness in the delay June 4th, 1992. The date of ratification by the coun- time distribution). The quartiles of the delay time dis- tries is recorded by the Convention. By definition, tribution are 375,810, and 978 days. That is, more than "left-censoring" is not a problem. That is, we know a year had passed by the time a quarter of the coun- the date when the FCCC was made available for rati- tries had ratified and well over two years had elapsed fication by all countries. However, some spells are before half the countries had done so. Figure 12.1 Survival function estimates SDF 1.0 + A+ S | +A u I ++ r ++ v +I A i | ~~~~A v 0.8 + ++ a ++ A D + i I 4+ t 0.6 + + r + 3. A b + u I+ t + ++ i I + o 0.4 + + n A F ~~~~~~~~~~~~~+4 U ++ n | A+ c I++ t 0.2 + ++ i I +A.. 4 n 0. 0 + 0 200 400 600 80 0 10001200 1400 1600 1800 2000 Ratification Delay (days) 176 Trade, Global Policy, and the Environment Figure 12.2 plots non-parametric estimates of the ified the FCCC, the less likely it is to do so. This latter hazard function. The hazard rate is the rate at which fact points to the importance of early ratification. countries will ratify the FCCC at any particular date Table 12.1 provides a breakdown by broad coun- (that is, conditional on them not having ratified up try "type," that is, OECD and former Eastern Bloc, until that point in time). The single-peakedness indi- and developing country. Despite the concessions cates an initially increasing hazard or positive dura- granted, it is evident that developing countries tend tion dependence and then a decreasing hazard or to have longer delay times.10 This suggests the im- negative duration dependence. (The peak is at 98 portance of a variety of country characteristics as de- countries, above the critical number of 50 countries terminants of the ratification of international required for ratification.) Negative duration depen- environmental agreement. We turn to this issue in the dence indicates that the longer a country has not rat- next section. Figure 12.2 Hazard function estimates Hazard I 0.0023 3 A 0.0020 . +. + * I + + 0 .0018 + + 4 |+ A+ I ~ ~ ~~~~~~~+ +++ | + ~~~~~~~~~~~~~~~+A I ++ *+ + | I A + a I + +. r | ~~~~~~~~~~~~~~+ + d 0.0013 + + + I + + F + A Ia I + + n + + c 0.0010 + + + t I + + I + + I + + fiI 4 0 , o 0a + +A 0 I0008 +++ 44+ | A+ + .+ + I + + 0.0005 + +4+ I+ + |+ A I 4.A+ I+ 0.0003 + 4 A 0.0000 + I 0 200 400 600 800 1000 1200 1400 1600 1800 2000 Ratification Delay (days) Table 12.1 Descriptive statistics for legislative delay times Countries Complete Mean Std Dev Median Incomplete All 163 810 396 728 21 Developing country 130 849 402 914 19 OECD/Former Eastern Bloc 33 656 336 630 2 Source: Data on ratification dates are from the unfccc official web site (http://www.unfccc.de/). The Importance of Trade for the Ratification of the 1992 Climate Change Convention 177 3. The Determinants of Legislative Delay the effect of democracy on trade. International trade The descriptive information is illuminating, but ad- between democracies is greater than between other .. . . . . ~~~~~~~~countries. Trade also has a negative effect on the ditional insights are gained from an analysis ofr level of conflict and a positive effect on bilateral co- abilistic models that estimate the effects of various factors on ratification delay times. To deal with the operation. censored nature of the data, we use the proportional Hoel and Schneider (1997) argue that gover- hazards regression model (Cox 1972). Further details ments may ratify an IEA for fear of exclusion from of the estimation and specification tests are left for future trade agreements. The costs of non-ratification, as well as of deviating after having ratified, are likely the Appendix. to be increasing in the total trade flows and openness We lnk he elaytim daa wth dta n car- of a country and in the number of trading partners. acteristics of the countries that are drawn from a num- ber of different sources. All the data are publicly Hoel and Schneider (1997) argue that low income available. See table 12.2. Missing observations for countries have little to lose by not joining an IEA be- cause they have small non-environmental costs (such some of the key explanatory variables reduce the sam- as a loss of reputation). Whereas exporters are likely pie size to 114 countries. The column containing cor- to be concerned about both their reputation and abate- relations in table 12.2 provides some preliminary mn ot,ipr-optn etr a aepi insights on some of the hypothesized relationships in ment costs, import-competing sectors may care pri- this section. marily about their costs, and thus trade may cause The majr findustries to take different lobbying stances on the The major focus of this paper iS the association FC between trade relations and environmental policy FCCC. making. Esty (1996) and Esty and Geradin (1997) dis- In the area of interational trade negotiations, cuss the issue of "political drag" in environmental and in the political game between superpowers, reci- policy making. Interpreted in our context, they have procity is often invoked as the correct standard of be- the following concern: Do countries delay the ratifi- havior producing cooperation among states (Keohane cation of IEAs and the associated abatement efforts, 1986). Obligations are important and imply that agents for international "competitiveness" reasons? behave in a generally acceptable fashion. The concept Barrett (1997) studies the important relationship appears to apply to IEAs, such as the FCCC, as well between trade and IEAs. He shows that if trade sanc- as to trade agreements. tions are a credible threat, full cooperation on global The sum of these arguments imply that nations environmental policy can be sustained when a mini- which are relatively reliant on trade interact with other mum participation requirement is used. He also ar- nations more frequently and exchange ideas more gues that in the absence of trade sanctions, the freely, and may therefore behave in a different way environment is not protected at the Pareto efficient than autarkies. While the global benefits of ratifica- level. tion are commonly shared, the costs of non-ratification Moreover, analogous to Fernandez' (1997) rea- are likely to vary with each country's reliance on trade. soning concerning regional trade areas, the commit- The effect of openness, per se, is ambiguous. For ex- ment to environmental cooperation may work as a ample, it should also be noted that openness is also signaling device. That is, it signals that a country's likely to affect compliance costs of an IEA as well as industries are competitive, that the investment climate the trade-related costs associated with ratifying an is attractive, and that future political relationships IEA. However, conditional on compliance with IEAs, with other nations will remain healthy. All these fac- we hypothesize that EXPORTS are likely to be posi- tors stimulate investment and are therefore likely to tively correlated with reputation costs and lead to make ratification of the FCCC more attractive. shorter delay times and a higher hazard rate. Howev- Trade has been seen as a civilizing force for cen- er, greater total IMPORTS increase the inflow of for- turies (see Hirschman 1982; and Schiff and Winters eign influences and ideas, as well as the amount of 1997).11 Chan (1984) argues that the level of conflict technology transfers and foreign competition. between any two nations diminishes if they are dem- We also include other controls and variables ocratic. Polachek (1992, 1997) explains this result by thought to be important determinants of the ratifica- Table 12.2 Descriptive statistics for covariate analysis A Variable Label N Mean Std Dev Minm Maxm p Sources DELAY Ratification delay, days after June 4, 1992 104 795.12 385.57 92.00 1959.00 . a ANNEX I OECD/Former Eastern Bloc 114 0.23 0.42 0.00 1.00 -.28- b TROPICS Tropics 114 0.61 0.49 0.00 1.00 .15 c CAPRCORN South of Tropic of Capricorn 114 0.05 0.22 0.00 1.00 .04 c AREA Total area, '000 sq km 114 877.71 1869.07 0.63 9976.14 -.42' CIA COASTLIN Coastline,'000km 114 5.68 23.78 0.00 243.79 -.30'- CIA POPN Population, 1994, mill. 114 43.64 142.87 0.21 1190.43 -.23" CIA LIFEEXP Life expectancy, years 114 64.01 11.82 37.46 79.31 -.18' CIA EXPORTS Exports 1992, bill. ($US) 114 26.00 65.56 0.04 449.00 -.35" CIA IMPORTS Imports 1992, bill. ($US) 114 27.62 70.50 0.13 582.00 -.34- CIA NPCAPITA National product pre capita, 1993 ($US) 114 6091.75 6500.01 500.00 24700.00 -.27*w CIA SACHSIDX Sachs-Warner trade openness index 104 0.37 0.48 0.00 1.00 -.12 SACHS NOTFREE Gastil civil rights index 114 4.26 2.04 1.00 7.00 .37 SCULLY SOCIALIST Gastil economic system index 114 3.79 2.33 1.00 9.00 .15 SCULLY GDPPER GDP growth, 1983-93, % 114 2.48 3.00 -14.90 8.90 -.11 WRI table 7.1 CO2TOTAL Total C02 Emissions, 1992 (mill. tons) 114 143.84 532.68 0.02 4881.35 -.39"- WRI table 14.1 OILRES Crude oil recoverable reserves (bill. tons) 65 2.00 5.44 0.00 35.62 .11 WRI table 12.3 SCOALRES Soft coal recoverable reserves (bill. tons) 37 8.41 24.40 0.00 134.06 -.54*- WRI table 12.3 HCOALRES Hard coal recoverable reserves (bill. tons) 46 8.43 22.41 0.00 106.50 -.40w WRI table 12.3 OPEC OPEC member country 114 0.09 0.28 0.00 1.00 .13 d Notes: a. Data are from UNFCCC official web site (http://wwwwunfccc.de/). Sample consists of 114 countries of which Io are non-ratifiers. b. Dummy variable = I, if country a member of OECD or former Eastern bloc. c. Latitude calculated from geographic center of country. d. OPEC country= 1. CIA. Data from The World Fact Book 1996 (U. S. Central Intelligence Agency). SCULLY. Data from Scully (1992), Appendix table 5.2, p. 132-39. SACHS: Data calculated from Sachs and Warner (1995). Open = 1. WRI: Data from World Resources: A Guide to the Global Environment, 1996-7, Worid Resources Institute, various tables. t Correlation with Log of Delay time. Statistically significant at I percent (*); 5 percent ("), 1o percent (*) level. N = 104. The Importance of Tradefor the Ratification of the 1992 Climate Change Convention 179 tion of IEAs. First, the financial situation of the coun- reasons other than the purported shorter tenure of try is likely to play a role. Countries that are growing dictatorships, Olson (1993) argues that dictators wish relatively quickly, for instance, may find it less costly to maximize tax revenues and thus oppose any poli- on the margin to ratify an IEA than countries that grow cies that would reduce revenue, for example, those more slowly. Slow income growth, which is associat- that result from increased abatement expenditures. ed with a growing unemployment problem, is often Fredriksson (1997) argues that more democratic coun- cited as the reason for resistance to entering into an tries (in the sense that they put a greater emphasis on IEA. Hence, higher percentage growth in income (GD- social welfare relative to lobbying activities) choose PPER) is likely to be associated with shorter delay local environmental policies that are closer to being times and a higher hazard rate. Also, we use GDP per welfare-maximizing.13 Consequently, democratic capita, NPCAPITA, as suggested by Congleton (1992). countries may therefore also be more likely to speed- Murdoch and Sandler (1997) found a linear relation- ily ratify global environmental policies. ship between GNP and emission cutbacks which they If a change in government or particular political interpreted as evidence in favor of the theory of the group in favor has occurred from the time of the ne- voluntary provision of public goods. To the extent that gotiations, the preferences of the legislative body have environmental quality is a normal good, NPCAPITA changed and thus a rapid ratification may be less like- is likely to yield shorter delay times and a higher haz- ly. lida (1996) argues that ratification failures are more ard rate (Bergstrom and others 1986). likely when negotiators are uncertain about their con- Second, countries that emit more greenhouse stituency's preferences. We assert that a greater de- gases and are more reliant on energy-intensive pro- gree of representative democracy may be expected to duction may be more hesitant about ratification of the yield better information about domestic legislative FCCC because their input costs are likely to rise rela- behavior. Furthermore, great internal divisions imply tively more. On the other hand, in a model with iden- that legislatures are less likely to ratify international tical countries, Barrett (1994) finds that the greater is environmental agreements, although this can to some the ratio of the slope of each country's marginal abate- degree also be used by negotiators to extract conces- ment cost curve to the slope of the global marginal sions (see Milner and Rosendorff 1997). benefit function, the fewer are the number of coun- It is also possible that the ratification of IEAs tries that sign an IEA. If higher emissions imply a low- serves as a commitment to the idea of democracy. By er slope of the marginal abatement cost curve, then participation in international agreements, countries ratification of an IEA is more likely. Countries that can show that they fulfill the requirements for partic- emit more pollutants are also likely to be under greater ipation in various forms of cooperation (Fernandez international pressure to ratify an IEA, and the repu- 1997).14 tational effects of non-ratification are likely to be great- Following previous authors (for example, er. A country's reputation may also affect international Congleton 1992), we use Gastil's indices of civil liber- negotiations on other important issues, such as trade ties and political freedoms (NOTFREE) as well as an policy. We use C02TOTAL as an indicator for green- index of the type of economic system that a country house gas emissions. has (SOCIALIST). Higher values of NOTFREE index Third, Congleton (1992) presents a theory which fewer civil liberties; higher values of SOCIALIST in- shows that autocratic countries select less stringent dex a greater degree of government control of prop- environmental regulations. He argues that dictators erty and less market-oriented or more socialistic tend to have shorter time horizons and should be less economic systems (see Scully 1992). likely, or take longer, to ratify an IEA, since the bene- Fourth, the population size (POPN) may influ- fits of doing so are likely to accrue after they have left ence the ratification date because a country with a office, whereas the costs are incurred earlier.12 The small population may stand to gain more from an IEA same argument implies a longer time before the FCCC than a large country. This is because in the negotia- is ratified. In addition, if more undemocratic govern- tion process a small concession by a large country may ments do have a higher turnover, they are more like- imply a substantial gain for a small country (see Chen ly to be out of office at the time of ratification. For 1997). It may be easier for a small country to free-ride, 180 Trade, Global Policy, and the Environment however, since the effect on global pollution is lower. specification. The findings reveal that the hazard rate, Alternatively, Congleton (1992) used POPN to proxy the conditional probability of ratifying the FCCC, is for a country's labor force. Hence, the extent to which positively correlated with EXPORTS, GDPPER, this is correlated with worker concerns about unem- C02TOTAL, AREA, and ANNEX1 and negatively ployment is likely to delay ratification. correlated with IMPORTS and NOTFREE. A country's resource base is positively related The coefficient for EXPORTS indicates that large to personal income and negatively related to the mar- exporting countries were more likely to ratify the ginal cost of environmental regulations (and therefore, FCCC quickly. Exporters appear to take reputational positively correlated with the equilibrium stringency effects into consideration, possibly because of fears of environmental policies). We use land area (AREA) of consumer boycotts or repercussions in other inter- as a proxy for resource base (which is also correlated national forums, such as the World Trade Organisa- with fossil fuels). Available data for recoverable oil tion. Moreover, they are likely to be more exposed and coal reserves are poor, due to the large number and more receptive to ideas in the area of environ- of missing observations. However, we estimate some mental policy and are therefore more open to cooper- models with the reduced sample and also estimate ation as predicted by Polachek (1992, 1997). To the models with an OPEC country dummy. extent that faster ratification mirrors a greater prefer- Next, the time horizon has implications for the ence for increased abatement efforts and stringent reg- costs and benefits derived from the reduction of emis- ulations, the finding provides no support for the sions. Nordhaus (1991) argues that whereas the costs "political drag" or "race to the bottom" effects induced of environmental regulations are incurred immediate- by freer trade. On the other hand, the negative sign ly, the rewards of regulation are repaid only later. We for IMPORTS is suggestive of a "political drag" or use life expectancy at birth (LIFEEXP) as a proxy for "race to the bottom." Large importing countries have the planning horizon. been hesitant to ratify the FCCC, perhaps out of fear Flooding is expected to increase as a result of of loss of 'competitiveness."'5 This indicates that lob- global warming. We use total COASTLINE as a mea- bying by import-competing industries may have been sure of the propensity for floods. Moreover, desertifi- successful in stalling the ratification process. Evident- cation and increased frequency of malaria are ly, the inflow of novel ideas about environmental pol- expected results of increased temperature. We use icies from abroad through imported goods and distance from the equator as measured by two dum- services has been minimal. The findings for trade are mies (TROPICS, CAPRICORN). subjected to further scrutiny below. Finally, developing countries have no significant The finding for C02TOTAL indicates that coun- obligations under the FCCC. This is likely to reduce tries generating greater total pollution had shorter rat- the costs of ratifying the agreement. We capture this ification delay times. This suggests that large polluting effect by the ANNEX1, which takes a value of one for countries were under great political pressure, either OECD and former Eastern Bloc countries. Preliminary internally or externally, to ratify the FCCC-. Reputa- evidence for this hypothesis is not supported by com- tional considerations or marginal abatement costs thus parisons across the rows of table 12.1 and by the neg- appear to be playing an important role. ative correlation in the penultimate column of table The positive coefficient for GDPPER suggests 12.2. This is, however, consistent with Hoel and that fast-growing countries perceive relatively great- Schneider (1997) who argue that the availability of er benefits, or lower costs, from the ratification of the side-payments, such as technology transfers, provid- FCCC than slower-growing ones. The latter group of ed that a country does not commit itself to coopera- countries is more likely to be concerned with high tion, may actually reduce the incentive to participate unemployment and to have fewer funds available for in IEAs. environmental protection. We also studied the effect of per capita income. We estimated several models. 4. The Results All models excluded POPN, as population appears Main findings: Table 12.3 contains the estimation re- in the denominator of NPCAPITA. In many models, sults. Column (1) presents estimates of the main model GDPPER was deleted. In addition, some of the other The Importance of Trade for the Ratification of the 1992 Climate Change Convention 181 Table 12.3 Estimates of proportional hazards model -Effects of trade measures (5) (1) (2) (3) (4) Sachs- Variable Full No trade Scaled trade Openness Warner AREA(*103) 0.152* 0.147 0.139 0.143 0.153 (0.086) (0.091) (0.090) (0.091) (0.095) POPN(* 102) 0.015 0.056 0.050 0.049 0.035 (0.115) (0.101) (0.102) (0.102) (0.104) EXPORTS 0.0I0** . 1.319 (0.005) (1.038) IMPORTS -0.012** . -1.616 (0.006) (1.150) OPENNESS . . -0.165 -0.067 (0.279) (0.285) NOTFREE -0.275*** -0.243*** -0.275*** -0.240*** -0.222*** (0.080) (0.078) (0.082) (0.078) (0.087) SOCIALIST 0.041 0.050 0.061 0.054 0.043 (0.055) (0.055) (0.055) (0.055) (0.057) GDPPER 0.070* 0.069* 0.083* 0.073* 0.050 (0.041) (0.041) (0.043) (0.042) (0.042) C02TOTAL(* 102) 0.106* 0.071 ** 0.069** 0.071 ** 0.067* (0.057) (0.035) (0.035) (0.035) (0.036) ANNEX I 0.776* 0.597 0.618 0.605 0.518 (0.402) (0.391) (0.388) (0.388) (0.392) TROPICS 0.004 0.071 0.029 0.097 -0.060 (0.357) (0.356) (0.359) (0.358) (0.369) CAPRCORN 0.127 0.238 0.216 0.252 0.140 (0.518) (0.518) (0.516) (0.518) (0.542) COASTLIN 0.007 0.009 0.009 0.009 0.009 (0.006) (0.006) (0.006) (0.006) (0.006) LIFEEXP -0.016 -0.013 -0.013 -0.010 -0.010 (0.014) (0.013) (0.015) (0.015) (0.015) N 114 114 114 114 104 -2 LOG L 749.783 755.266 753.227 754.886 683.115 Likelihood ratio 56.961 51.477 53.517 51.858 44.708 Score 111.695 106.381 106.916 106.381 96.695 Wald 62.301 54.664 55.687 54.810 48.397 Notes: i. Standard errors in parentheses. Statistically significant at I %(*"); 5%(**), 10%(*) level. ii. Column (3): Exports and Imports are scaled by domestic consumption (GDP + Imports - Exports); Column (4): Openness index calculated as (Exports + Imports)/GDP; Column (5): Sachs and Warner (1995) Openness index. size-related variables were rotated out of the some of It lends some support to the theory by Hoel and the models. To the extent that total CO2 emissions are Schneider (1997) that the availability of technology positively correlated with a country's wealth, some transfers to developing countries does not necessari- models were also estimated without C02TOTAL. In ly increase the incentive to cooperate. Since develop- all cases, the income per capita variable never achieved ing countries had no binding commitments under the significance (at the 10 percent level). Hence, our dis- FCCC, the ratification decision may best be viewed cussion in this section relates to the preferred model as having a low priority by their legislatures or other specification, which includes GDPPER, reported in decision-making bodies. column (1) of table 12.3. Robustness of thefindings: Unfortunately, there is The group identity variable (ANNEX1) had a sig- some sensitivity in some of the parameter estimates. nificant and positive impact on the ratification speed. Alternative model specifications were also estimated 182 Trade, Global Policy, and the Environment because of a suspected multicollinearity problem be- risk of flooding (COASTLINE) does seem to have tween some of the variables (for example, the correla- some effect, maybe because of pressure from local tion of the size variables is quite high for our insurance companies. This is not a robust finding, sample).16 In some model specifications, many of the however. Finally, the type of economic system (SO- effects identified above, in connection with the results CIALIST), location (TROPICS, CAPRICORN), and reported in column (1), generally have weak statisti- planning horizon (LIFEEXP) have no significant ef- cal significance. Across all the specifications that were fects. estimated, the signs of significant coefficients were A detailed examination of trade variables: One of our generally robust. In particular, NOTFREE-Gastil's key interests is the effects of trade and trade-related civil liberties index -is always associated with a sig- variables on the ratification of the FCCC. The remain- nificantly lower hazard across all model specifications ing columns of table 12.3 present the estimation re- that were estimated, that is, greater civil liberties re- sults for different measures of trade and trade sult in more rapid ratification. Using the point esti- openness. Column (2) reports the results from a mod- mate in column (1), the marginal effect of a unit el specification without any trade-related variables, increase in NOTFREE on the hazard rate is -24 per- which once again highlights the importance of NOT- cent.17 FREE and C02TOTAL. Also, note that omission of Benedick (1991) argues that the experience with total trade flows, which are positively correlated with the CFC negotiations shows the importance of politi- a country's size, does not increase the significance lev- cal leadership and resistance to lobby groups inter- els or change the coefficient signs of the size variables ested in impeding the legislative process. More such as AREA and POPN. Column (3) scales imports democratic countries may also be less likely to use and exports by domestic consumption (that is, GDP tardy ratification as a tool in the next negotiation plus imports minus exports). These measures there- round. The result underscores the importance of po- fore capture some notion of the importance of imports litical and institutional factors for the ratification of and exports in the domestic economies of each coun- IEAs. try by scaling trade flows using market size. The find- Apart from NOTFREE, the only other variable ings now reveal that the hazard rate is unaffected by that consistently retained both its coefficient sign and either import penetration or scaled exports. Thus, as statistical significance across all the model specifica- above, the economic size of a country appears to be tions that were estimated was C02TOTAL. Howev- the overriding consideration.19 er, this variable is likely to be correlated with economic Columns (4) and (5) use two different measures size and wealth. Scaling this variable by a size vari- of countries' outward orientation or trade openness. able such as POPN yields a statistically insignificant We use the ratio of total trade (imports plus exports) coefficient estimate. That is, CO2 emissions per capita scaled by total output, as well as the measure of trade have an insignificant effect on ratification delay. openness developed by Sachs and Warner (1995).20 Therefore, the most conservative conclusion that can In both model specifications, the results indicate that be drawn is that larger, wealthier countries, which only NOTFREE and C02TOTAL are significant. The have greater pollution levels, were more likely to rat- signs of the other covariates are also stable, albeit ify the FCCC at any point in time. insignificant. Variables that capture endowments (such as Overall, while there is some evidence that trade AREA and POPN) or vulnerability to global warm- may well matter for the passage of the FCCC, specif- ing (COASTLINE) generally appear to be less impor- ically total exports and imports, we find no conclu- tant, although AREA and COASTLINE are significant sive evidence for either the "race to the bottom" or at the 10 percent significance level in column (4) with the "race to the top" hypothesis. In addition, since the the expected signs. AREA is a proxy for resource abun- coefficients of the two trade measures are approxi- dance, that is the marginal cost of environmental reg- mately equal in absolute value, multilateral trade lib- ulations (which is falling in the available resources), eralization is unlikely to have a significant impact on and this has some influence on countries' behavior, the speed of ratification. This is evidenced by the in- which was also reported by Congleton (1992)."8 The significant coefficients for openness and Sachs-Warner The Importance of Tradefor the Ratification of the 1992 Climate Change Convention 183 in columns (4) and (5) of table 12.3. The findings can the ratification process. This is an area where much possibly be explained by the offsetting behavior of work remains to be done. import and export industry lobby groups in the legis- lative process. We note that the evidence on the im- 5. Conclusion portance of trade variables is far from conclusive and This paper represents a unique study of the United indicates the need for both further empirical investi- Nations Framework Convention on Climate Change gation and, in our opinion, greater effort directed at (FCCC). In particular, the study looked at the deter- theoretically modeling the precise avenues through minants of the delay in ratification of the FCCC. The which international trade and environmental policy legislative delays in ratifying the agreement were in- interact with one another. strumental in preventing the timely entry of the FCCC There are (at least) two possible interpretations into force. Consequently, this may have had impor- of the estimated effects of trade related variables on tant implications for the overall lack of success of the the ratification of the FCCC. The conservative statis- agreement. The deadline for reductions was 8 years tical conclusion is that the economic size of countries after the Rio Earth Summit. The credibility of this is of overriding importance and that trade, or trade agreement hinged on its early entry into force. By ex- openness, has had no effect on the speed of the ratifi- amining when countries ratified the FCCC, our ap- cation of the FCCC. A more controversial conclusion proach enabled us to identify the intensity of each is that large exporting countries were quicker to rati- country's preference to ratify, or not to ratify, the fy the agreement because of reputational concerns. On FCCC. We consequently identified the characteristics the other hand, large importing countries were guilty of the "leaders" and the "followers" involved in the of dragging their heels with regards to ratification of process of negotiations on the global warming issue. the agreement. This may have been due to competi- Among the more interesting findings were that tiveness pressures and lobbying by import-competing the conditional probability of signing the FCCC was industries. positively related to total CO2 emissions and the pres- Our findings may have implications for the rati- ence of civil liberties. The latter finding is highly sig- fication of future IEAs as follows. If we take the con- nificant and robust and is consistent with earlier servative high road (that is, that trade is irrelevant), research that found that democratic freedoms raised then trade sanctions against non-ratifiers will have no the probability of signing the Montreal Protocol. As- effect on the speed of ratification. In fact, linking trade suming increased levels of democracy over time, ex- sanctions and the ratification of IEAs is likely to be isting forecasts of future emissions may overestimate counterproductive. Such measures will likely result future CO2 levels. Future research on the ratification in social welfare losses and have no great impact on of global international environmental agreements when ratification of IEAs is achieved. should explicitly consider the political and democratic Alternatively, if one takes the view that trade status of countries at any point in time. If countries does matter, then the threat of trade sanctions may are converging in their degree of democracy, ratifica- reinforce the reputational concerns for exporters. In tion delay may also converge. Interestingly, converg- other words, trade sanctions are a credible threat for ing income levels would not yield this effect. The exporters and may expedite the ratification process. finding for CO2 emissions indicates that large, heavi- On the other hand, the large importing countries are ly polluting countries were under great political pres- the laggards in the ratification process and sanctions sure, either internally or externally, to ratify the FCCC. imposed by those countries that are more likely to This also indicates that the speed of ratification is in- ratify (that is, the exporters with reputational con- deed viewed as important. cerns), would not be a credible threat. Even if sanc- Finally, we found mixed evidence that trade tions are credible, then under the interpretation that mattered for the ratification of the FCCC. Whereas international competitiveness concerns slow the rati- total exports has a positive impact on the speed of fication process, the use of measures that sanction ratification, the opposite is true of total imports. Not exports to these countries would further mitigate these surprisingly then, more aggregate measures of trade competitiveness concerns and not serve to accelerate openness indicate no effect of international trade flows 184 Trade, Global Policy, and the Environment on the probability of ratifying the FCCC, that is, we This game merits further analysis in our opinion. We find no evidence of "political drag" in the ratification consider that our findings may be important for pro- process due to international trade. Moreover, the ef- jections of the future increase of greenhouse gases fects of these trade variables disappear when they are under various scenarios. If democracy spreads fur- scaled by variables capturing size or total wealth, such ther among nations, future abatement efforts may as GDP. The economic size of countries also seems to prove to be more successful than are currently pre- be an extremely important determinant of the deci- dicted. The diffusion of democratic ideas are thus an sion about when to ratify an international environmen- important consideration in the continuing attempts tal agreement. to reduce global greenhouse emissions. Naturally, there are strategic issues involved since the CO2 negotiations are an ongoing process. The Importance of Trade for the Ratification of the 1992 Climate Change Convention 185 Appendix i(t) = f(t)/(1 - F(t)) = f(t)/S(t), (A.3) The essential ideas of a survival model are sketched briefly here. See Kiefer (1988) or Lancaster (1992) for more detailed treatments. The conditional hazard rate where S(t) = Pr(T 2 t) is the survivor function, for is modelled as example, the probability of "surviving" at least until t. The density of completed spells can be found from -Z(tl X) (t) expx. (A.1) equation (A.3), for example, f(t) = S(t)A(t). In addi- 2(tIX) = 20 (t) expXfi . (A.1) tion, note that 2(tIX) is the rate at which countries will ratify the d In S(t)/dt = - f(t)/S(t) = -A(t), (A.4) FCCC at any particular date, given that they have not ratified up until that point in time. Equation (A.1) hence, from equations(A.3) and (A.4), the survivor specifies the hazard rate as the product of two com- ponents: a function of spell length (for example, de- function is lay time), 2A (t) or baseline hazard, and a function of r , 1 the observable country characteristics, which are de- S(t) = expt- J02(s)dJ. (A.5) noted by the vector X. The Cox nonparametric esti- mation method enables us to estimate /3 without Thus, the density of completed spells can be comput- having to make a distributional assumption about ed from the integrated hazard rate function. For our , (t). A central objective of this paper is to obtain purposes, note that "survival" means that the coun- unbiased estimates of the vector .1 I try has still not ratified the FCCC and that a "hazard" occurs when the country ratifies the FCCC. Also note For those readers used to thinking in termspo that the time variable "t" is not calendar time but rath- standard regression analysis, it is helpful to interpret er the delay time, which is measured in days. the estimates of the elements of /3 as being negatively To model the effect of covariates on delay times, related to the respective element in the /3 vector esti- the hazard rate can be written as a conditional func- mated by an OLS regression of X on log delay time. tion of country characteristics, which we denote by For example, if a variable X j has a coefficient A, > 0 the vector X. The most widely-used specification of in equation (A.1), this implies that the effect of X j the conditional hazard rate is the proportional hazards on the hazard rate is positive, for example, that the model, country characteristic X j is associated with a higher hazard or greater conditional probability of exiting A(tlX) = (t)0(X). (A.6) (for example, ratifying the FCCC). Alternatively, it implies that the variable X j is associated with a short- Equation(A.6) specifies the hazard rate as the prod- er delay in the ratification of the FCCC. Note that uct of two components: a function of spell length, A, (t) such direct comparisons and interpretations of coef- or baseline hazard, and a function of the observables, ficients are only meaningful if the data are not heavi- ¢(X). Letting ,(X) = exp(X'O), and taking logs, equa- ly censored (see Kiefer 1988, p.665). tion (A.6) gives: More formally, the hazard rate A(t) is the prob- ability of exiting a state in the time interval [t,t + h], In A(tJX) = lnA0(t) + X'A. (A.7) conditional on having arrived at t in that state, and is defined as Note that A(t)h=limPr(t