The Impacts of Trade Liberalization and Macroeconomic Instability on the Brazilian Economy

Home , Liberalization, Linder hypothesis

THE IMPACTS OF TRADE LIBERALIZATION AND MACROECONOMIC INSTABILITY ON THE BRAZILIAN ECONOMY

DISSERTATION

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy in the Graduate School of The Ohio State University

Mauricio Vaz Lobo Bittencourt, M.S., M.A.

* * * * *

The Ohio State University 2004

Dissertation Committee: Approved by Dr. Donald W. Larson, Adviser

Dr. David S. Kraybill ______Adviser Dr. Stanley R. Thompson Graduate Program in Agricultural, Environmental, and Development Economics

ABSTRACT

For decades, Latin America, and particularly Brazil, adopted traditional protectionist policies that created an economic structure based on high import tariffs and prohibitions that generated a severe anti-export bias that discouraged both the growth and diversification of exports. However, the large number of trade agreements worldwide was also implemented in Latin America in the late 1980s, reducing substantially the level of protection in these countries. Brazil was one of the last closed Latin American countries to open its economy to the foreign market in the beginning of the 1990s, with the creation of the Mercosur, together with Argentina, Paraguay and Uruguay. After this, Brazil trade with its Mercosur partners increased largely, and new free trade agreements began to be debated between Mercosur and other countries. Mercosur is still negotiating two other main agreements. The first involves Mercosur and the European Union, and their main issues have been the agricultural products. This issue also seems to be one of the obstacles of the second main agreement, the Free Trade Area of Americas (FTAA), which was initially planned to be implemented in January 2005. The FTAA, if successfully implemented, will include all countries in the North, Central, and South

Americas, except Cuba, and it will be the largest free trade area in the world.

ii The main goal of the Brazilian trade liberalization program is to reverse the negative effects of protectionist policies adopted in the past. Traditional trade theory predicts that trade liberalization reallocates resources according to comparative advantage, reduces waste, and lowers the price of imported goods in a more transparent economic regime, with less lobbying activities, and exports not only grow rapidly, but also become more diversified. Most economists also share that open countries fare better in the long run than do closed ones, but the short run impacts from trade liberalization can harm the poor. Since Brazil is one of the countries with larger inequality in the distribution of income, with high levels of poverty and regional differences, this study takes these concerns seriously by assessing the economic impacts of a reduction in import tariffs on poverty and distribution of income, identifying a combined policy that can reduce possible negative impacts from trade reform on the poor, through a single-country multi-regional computable general equilibrium model (CGE) applied to Brazil. The main findings show that sectoral reduction on import tariffs can bring better results than an overall reduction on such tariffs. Poverty and regional income inequality can be reduced through combined trade and tax policies.

Because of the ongoing Mercosur trade agreement and also the negotiations of the proposed FTAA, the role of macroeconomic policies in the involved countries in the process of opening a country’s economy is very relevant. In recent years, countries like

Argentina and Brazil have experienced many different economic crises due to their own domestic instabilities, which have contributed to delayed market opening in these countries, and have threatened the evolution of new trade agreements, such as the FTAA.

This study also emphasizes the lack of macroeconomic policy coordination between

iii Mercosur and FTAA countries, notably the exchange rate policy through the impact of real bilateral exchange rate volatility on trade. Excessive price and exchange rate fluctuations caused by uncoordinated macroeconomic policies among trade partners can affect trade and resource allocation among members of a free trade area. Therefore, a sectoral gravity model is estimated to evaluate not only the role played by the lack of macroeconomic policy coordination, but also to better evaluate the patterns of trade in the

Mercosur and in the proposed FTAA. The overall results show that, at the same time that the reduction in the level of exchange rate volatility can increase bilateral trade, gradual reduction in the level of tariffs and increase in countries’ income are also important pro- trade variables.

Dedicated to my wife Marcia and my daughter Bruna

v ACKNOWLEDGMENTS

This dissertation would not be possible without contributions and support from many people. I will try to thank all of them in the lines below. Please forgive me if I forgot someone.

First of all, I wish to mainly thank my loved wife and daughter, Marcia Adriane and Bruna, for their enormous love, support, patience, and for being my main source of inspiration during the graduate studies.

I also would like to thank Professor Donald Larson, my adviser, for his continuous support, guidance, and patience throughout these 4 years. Prof. Larson was not only my advisor during this period, but he also became a great friend of mine. Thanks for everything, Professor Larson! I am also thankful to his wife, Mrs. Karen Larson, for the grammar suggestions and for her patience in reading the whole dissertation.

I am grateful to professors David Kraybill and Stanley Thompson for their intellectually challenging and stimulating discussions and comments.

I wish to thank my parents, Joao Alfredo and Scheila, for all their support and for believing in me.

Thanks to my father-in-law and his wife, Antonio Carlos and Maria de Lourdes, for their willingness to help and for trying to be with us most of the time.

vi I am grateful to many people who helped me in one way or another to be here.

Among them, I am particularly thankful to Armando Vaz Sampaio, Jose Gabriel Porcile,

Mauricio Serra, Nilson de Paula, Ramon Fernandez, and all professors from the department of Economics (Federal University of Parana) for their unconditional support during these 4 years. Special thanks to my friend, Armando Vaz Sampaio, who was always ready to help and to assist us in whatever we needed.

Thanks to professors Judas Tadeu G. Mendes and Ricardo Shirota for being so supportive and also for their optimism about my accomplishments in the graduate studies.

I would like to thank professor Joaquim Bento Ferreira for his encouragement and technical support with chapter 2 of this dissertation. I also would like to thank Dr. Hans

Lofgren, from International Food Policy Research Institute (IFPRI), for supplying the basic Brazilian social accounting matrix used in chapter 2.

I also wish to thank Mr. Samuel Munyaneza from the United Nations Conference on Trade and Development (UNCTAD) for allowing me to access their trade database in

November/2003, which I used in chapter 3 of this dissertation. I am also thankful to the

Brazilian Embassy for having me in Washington, D.C. to download this data set.

I am grateful to my friends and colleagues, Eric Rangel, Marcos Hasegawa,

Ratapol Teratanavat, and Tufan Ekici, for listening and encouraging me throughout these

4 years.

I must especially acknowledge the CAPES Foundation and the Federal University of Parana for the financial support during the whole 4-year period in which I was on leave in the doctorate program at the Ohio State University.

vii VITA

May 6, 1970 ………………………… Born – Curitiba, Parana, Brazil

1992 …………………………………... B.Sc. Agronomy, Federal University of Parana

1995 …………………………………... M.S. Agricultural Economics, Escola Superior de Agricultura “Luiz de Queiroz” (ESALQ), University of Sao Paulo, Brazil

1995 - 1997...... Substitute Professor, Department of Agricultural Economics, Federal University of Parana, Brazil

1995 - 1997...... Agric. Economics Analyst and Consultant, Agromarket Socio-Economic Consultant Ltd., Brazil

1997 – present ...... Assistant Professor, Graduate Studies in Development Economics, Department of Economics, Federal University of Parana, Brazil

2000 – 2004 ...... Fellowship Recipient, Fundacao CAPES, Brazil

2004 ...... M.A. Economics, The Ohio State University

2004 - present ………………………… Graduate Teaching Associate, The Ohio State University

viii PUBLICATIONS

Journal Articles

1. Oliveira, A., Larson, D., Bittencourt, M., and Graham, D. "The Potential for Savings Mobilization in Rural Mozambican Households". Savings and Development XXVIII (2), (2004).

2. Barros, G.S.C., and Bittencourt, M.V.L. “Price Formation Under Oligopsony: The Poultry Market in Sao Paulo”. Brazilian Economic Review (Revista Brasileira de Economia) 51 (2): 181–199, (1997).

3. Bittencourt, M.V.L. (1996). “Price Formation of Soybean in Parana and the Infuence of External Market”. Agrarian Sciences Review (Revista do Setor de Ciencias Agrarias)15 (2): 07–13, (1996).

4. Bittencourt, M.V.L., and Barros, G.S.C. “Prices Ratios of Poultry in the Brazilian South and Southeast Regions”. Brazilian Society of Agricultural Economics and Rural Sociology Journal (Revista Brasileira de Economia e Sociologia Rural) 34 (3/4): 147–172, (1996).

Book Chapters

1. Porcile-Meireles, J.G., Bittencourt, M.V.L., and L. Bertola. “The Thirlwall Law Revisited: A VAR Application to the Brazilian Economy in the Post-War”. In Dynamics Macroeconomics: Growth, Cycles, Development, and Economic Policy (Macroeconomia Dinamica: Crescimento, Ciclos, Desenvolvimento, e Politica Economica). M.A. Dias (Editor). Maringa State University, (2002).

FIELDS OF STUDY

Major Fields: International Trade and Development Economics

ix TABLE OF CONTENTS

Page Abstract …………………………………………………………………………….. ii

Dedication ………………………………………………………………………….. v

Acknowledgments ………………………………………………………………….. vi

Vita …………………………………………………………………………………. viii

List of Tables ………………………………………………………………………. xii

List of Figures ……………………………………………………………………… xviii

Chapter 1: Introduction ….……………………………………………………………...... 1

Chapter 2: Short to Medium Run Regional Effects of Trade Liberalization: a Computable General Equilibrium Model For Brazil ...... 8 2.1 Introduction ……………………………………………………………..... 8 2.2 The issue …………………………………………………….…………..... 11 2.3 Objectives of the study …...... ….………………………………….………. 19 2.4 Literature review …………………………..……………………………… 21 2.4.1 CGE and Walrasian Models…………………..……………..…….... 21 2.4.2 CGE Models to Evaluate Changes in Trade Policy.………………… 24 2.4.3 CGE Studies About Trade in Brazil…….…..……………………..... 29 2.5 Social accounting matrix (SAM) …………………………………………. 33 2.5.1 Regional Sectoral Disaggregation……..……………………...... 37 2.5.2 The Balance Procedure ………………..……………………...... 38 2.6 The standard CGE model ………………..………………………………... 44 2.6.1 Prices, Activities, Production, and Factor Markets……..…………... 45

x 2.6.2 Institutions…………………………………………..…..…………... 48 2.6.3 Commodity Markets…………………………………….…………... 49 2.6.4 Macroeconomic Closures…………………………..…..………….... 50 2.6.5 Inequality and Welfare Measures……………………….…………... 53 2.7 Trade policy simulations ...………………………………………………... 62 2.8 Results and Discussion …………………………..…………….………….. 71 2.8.1 Regional Disaggregated SAM…….…………………….…………... 71 2.8.2 Overall Trade Liberalization (Scenario 1)…...………….………...... 76 2.8.3 Sectoral Trade Liberalization (Scenario 2)…….…………………..... 87 2.8.4 Equity-Efficiency Trade Liberalization (Scenario 3)…...………...... 93 2.9 Conclusions …………………………………..…………….……………... 105

Chapter 3: An Examination of Exchange Rate Volatility in the Mercosur and in the Proposed Free Trade Area of the Americas: Sectoral Trade Impacts in Brazil ………………. 110 3.1 Introduction ...…………………………………………………………...... 110 3.2 Specification of the problem ...... 114 3.3 Literature review …...…………………..…………………………………. 121 3.3.1 Gravity models ……………………………………………………… 121 3.3.2 The proposed FTAA ………………………………………………... 128 3.3.3 Effects of exchange rate volatility on different sectors ……………... 130 3.4 Data and issues …...………...…………………………………………...... 134 3.5 The gravity model ….....……………..………………………………...... 146 3.6 Results and discussion ……………………………………………………. 154 3.6.1 The Mercosur analysis ……………………………………………… 155 3.6.2 The FTAA analysis …………………………………………………. 166 3.7 Conclusions and implications ……………………………………………. 176

Appendix A: Brazilian Social Accounting Matrix (SAM) ………...………………. 183 Appendix B: Cross-Entropy Equations …………………….……...…………...... 185 Appendix C: The Standard CGE Model ………………………...…………………. 187 Appendix D: Main Disaggregated SAM Components ………...…………..………. 198 Appendix E: Additional results and discussion from chapter 2 ……………………. 204 E.1 Disaggregated regional SAM 205 E.2 Overall Trade Liberalization (Scenario 1) 209 E.3 Sectoral Trade Liberalization (Scenario 2) 227 E.4 Equity-Efficiency Trade Liberalization (Scenario 3) 235 Appendix F: Product Disaggregation Across Sectors …………………………...... 241

List of References ...... 246

xi LIST OF TABLES

Table Page

2.1 Gini index for regional monthly labor income for people 10 years old or older in Brazil, 1999 …………………………………………………….. 17

2.2 General description of a SAM structure used in the standard CGE model ...... 35

2.3 Summary of activities, commodities, and factors included in the 1995 Brazilian SAM …………………………………………………………... 36

2.4 Main assumptions and macroeconomic closure of the Brazilian standard CGE model ……………………………………………………………… 53

2.5 Description of the main sets of simulation for the Brazilian trade reform……………………………………………………………………. 63

2.6 Average nominal import tariff by sectors and goods in Brazil, 1995 …... 65

2.7 Maximum income tax rates for selected countries, in percent ………….. 68

2.8 Aggregated national accounts …………………………………………... 73

2.9 Participation of commodities in value added, production, employment, exports, and imports shares ……………………………………………... 74

2.10 Proportion of Brazil’s total factors employed in each region …………... 76

2.11 Simulations results for overall import tariffs reduction (scenario 1), % change from benchmark values …………………………………………. 77

2.12 Regional impacts from an overall elimination of the import tariffs in household’s labor income (% change from benchmark values) ………... 84

2.13 Regional income inequality measures before and after an overall elimination of the import tariffs ………………………………………… 85

xii

2.14 Contribution of the four decompositions to overall labor income inequality before and after simulation …………………………………... 86

2.15 Regional contribution to overall labor income inequality before and after simulation ………………………………………………………….. 86

2.16 Simulation results for sectoral elimination of the import tariffs (scenario 2), % change from benchmark values …………………………………... 89

2.17 Regional income inequality measures before and after elimination of the import tariffs in agriculture ……………………………………………... 91

2.18 Regional income inequality measures before and after elimination of the import tariffs in industry ………………………………………………... 92

2.19 Regional income inequality measures before and after elimination of the import tariffs in a combination of agriculture and industry …………….. 93

2.20 Simulation results for overall import tariffs reduction combined with 20 % in direct tax (scenario 3), % change from benchmark values ………... 97

2.21 Main changes in consumption expenditures by households for scenarios 1 and 3 …………………………………………………………………... 99

2.22 Overall regional impacts from an elimination of the import tariffs combined with an increase in the rate of direct tax on household’s labor income (% change from benchmark values) ……………………………. 101

2.23 Overall regional impacts from an elimination of the import tariffs combined with an increase in the rate of direct tax on capital and land incomes (% change from benchmark values) …………………………... 102

2.24 Regional income inequality measures before and after an overall elimination of the import tariffs combined with an increase in the rate of direct tax ………………………………………………………………… 103

2.25 Contribution of the four decompositions to overall capital income inequality before and after simulation …………………………………... 104

3.1 Average annual growth rate of trade in Argentina and Brazil for the period 1991-2000 ……………………………………………………….. 115

xiii

3.2 Fixed and random effects estimations for trade in the agricultural sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure ……………………………………………………….. 157

3.3 Fixed effects estimations for trade in the livestock sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure ………………………………………………………………….. 158

3.4 Random and fixed effects estimations for trade in the chemicals sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure ……………………………………………………….. 160

3.5 Fixed effects estimations for trade in the manufactured sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure ………………………………………………………………….. 161

3.6 Random and fixed effects estimations for trade in the mining and oil sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure …………………………………………………... 162

3.7 Fixed effects estimations for total trade between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure …………….. 163

3.8 Summary of the statistically significant coefficients for the sectoral trade between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure …………………………………………………... 165

3.9 Fixed effects estimations for trade in the agricultural sector between Brazil and 17 Potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ……………………………………………………….. 168

3.10 Fixed effects estimations for trade in the livestock sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ……………………………………………………….. 169

3.11 Fixed effects estimations for trade in the chemicals sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ……………………………………………………….. 170

3.12 Random effects estimations for trade in the manufactured sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ………………………………………... 172

xiv

3.13 Fixed effects estimations for trade in the mining and oil sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ……………………………………………………….. 173

3.14 Random effects estimations for total trade between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ………………………………………………………………….. 174

3.15 Summary of the statistically significant coefficients for the sectoral trade between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure ………………………………………... 176

A.1 Brazilian social accounting matrix (SAM) (Cattaneo, 1999), 1995-96 aggregated version (1995 bi R$)………………………………………… 184

D.1 Description of the main activities in the disaggregated SAM …………... 199

D.2 Description of the main types of urban labor in the disaggregated SAM …………………………………………………………………….. 201

D.3 Description of the main types of rural labor in the disaggregated SAM …………………………………………………………………….. 202

D.4 Description of the main types of land in the disaggregated SAM ……… 202

D.5 Description of the main types of capital in the disaggregated SAM ……. 203

E.1 Quantity of factors employed by each sector and region ……………….. 206

E.2 Regional distribution of factor endowments for each type of household ……………………………………………………………….. 207

E.3 Budget share for commodities by households …………………………... 208

E.4 Simulation results for the Region North after an overall 100 % reduction in the import tariffs (% change from benchmark values) ……………….. 211

E.5 Factor prices by each activity in the Region North after an overall 100 % reduction in the import tariffs (% change from benchmark values) …. 212

E.6 Household’s labor income in the Region North after an overall 100 % reduction in the import tariffs (% change from benchmark values) ……. 213

E.7 Simulation results for the Region Northeast after an overall 100 % reduction in the import tariffs (% change from benchmark values) ……. 215

E.8 Factor prices by each activity in the Region Northeast after an overall 100 % reduction in the import tariffs (% change from benchmark values) …………………………………………………………………... 216

E.9 Household’s labor income in the Region Northeast after an overall 100 % reduction in the import tariffs (% change from benchmark values) …. 218

E.10 Simulation results for the Region Center-West after an overall 100 % reduction in the import tariffs (% change from benchmark values) ……. 220

E.11 Factor prices by each activity in the Region Center-West after an overall 100 % reduction in the import tariffs (% change from benchmark values) …………………………………………………………………... 221

E.12 Household’s labor income in the Region Center-West after an overall 100 % reduction in the import tariffs (% change from benchmark values) …………………………………………………………………... 222

E.13 Simulation results for the Region South/Southeast after an overall 100 % reduction in the import tariffs (% change from benchmark values) …. 224

E.14 Factor prices by each activity in the Region South/Southeast after an overall 100 % reduction in the import tariffs (% change from benchmark values) …………………………………………………………………... 225

E.15 Household’s labor income in the Region South/Southeast after an overall 100 % reduction in the import tariffs (% change from benchmark values) ………………………………………………………………… 226

E.16 Simulation results for 50 % sectoral import tariffs reduction (scenario 2), % change from benchmark values …………………………………... 227

E.17 Household’s labor income after elimination of the import tariffs in agriculture (% change from benchmark values) ………………………... 229

E.18 Regional changes in household’s labor income after elimination of the import tariffs in agriculture (% change from benchmark values)…………………...... 230

E.19 Household’s labor income after elimination of the import tariffs in industry (% change from benchmark values) …………………………… 231

xvi

E.20 Regional changes in household’s labor income after elimination of the import tariffs in industry (% change from benchmark values) …………. 232

E.21 Household’s labor income after elimination of the import tariffs in a combination of agriculture and industry (% change from benchmark values)…………………………………………………………………… 234

E.22 Regional changes in household’s labor income after elimination of the import tariffs in a combination of agriculture and industry (% change from benchmark values) ………………………………………………… 235

E.23 Factor prices by each activity in the Region North after combining trade/tax reform (% change from benchmark values) …………………... 237

E.24 Factor prices by each activity in the Region Northeast after combining trade/tax reform (% change from benchmark values) …………………... 238

E.25 Factor prices by each activity in the Region Center-West after combining trade/tax reform (% change from benchmark values) ………. 239

E.26 Factor prices by each activity in the Region South/Southeast after combining trade/tax reform (% change from benchmark values) ………. 240

F.1 Sectoral participation in trade between Brazil and Mercosur partners, 2001 ...... 242

F.2 Main countries considered in the proposed FTAA analysis ……………. 242

F.3 Main products included in the livestock sector for the Mercosur and the FTAA analysis …………………………………………………………... 243

F.4 Main products included in the agricultural sector the Mercosur and the FTAA analysis …………………………………………………………... 243

F.5 Main products included in the chemical sector the Mercosur and the FTAA analysis …………………………………………………………... 244

F.6 Main products included in the manufactured sector the Mercosur and the FTAA analysis ………………………………………………………. 244

F.7 Main products included in the mining and oil sector the Mercosur and the FTAA analysis ………………………………………………………. 245

xvii LIST OF FIGURES

Figure Page

2.1 Gini coefficient for the income distribution in Brazil, 1992-2001 …… 15

2.2 The poverty problem in Brazil, 1977 to 1999 ………………………… 18

2.3 Regional production technology in the standard CGE model for Brazil ………………………………………………………………….. 46

2.4 Flows of regional marketed commodities in the standard CGE model ………………………………………………………………….. 52

2.5 Hicksian equivalent variation (EV) …………………………………... 62

2.6 Transmission of trade shocks in the domestic market of a good ……... 70

2.7 Direct tax rates at the base year and for the simulation in scenario 3 (%) …………………………………………………………………….. 95

2.8 The main effects of different simulations on household’s welfare changes from base (%) ………………………………………………... 98

3.1 Index of Brazilian real exchange rate (Brazil/USA), Jan/99 = 100, period January 1999 to April 2003 ………………………………...... 116

3.2 Bilateral real exchange rate volatility (moving standard deviation measure) in Mercosur, 1989 – 2002 ………………………………….. 138

3.3 Bilateral real exchange rate volatility (Peree and Steinherr measure) in Mercosur, 1989 – 2002 ……………………………………………….. 139

3.4 Correlation between bilateral real exchange rate and economic fundamentals between Brazil and other countries, 1989-2002 ……….. 146

xviii CHAPTER 1

INTRODUCTION

Most economists agree that the Brazilian economic crises in the last decades are linked closely to the chronic public finance imbalance that has become an important obstacle to macroeconomic stabilization, which can be considered an important factor for long-run sustainable growth. The public sector has played a crucial role in Brazil’s development, adopting a model of industrialization based on import substitution for many decades, distorting the economy with protectionist tariffs, credit, subsidies, fiscal incentives, taking large amounts of foreign loans that would contribute to economic crises, and implementing unsuccessful economic measures to control the inflation, worsening the socio-economic problems of the country as a whole. The industrial policy adopted, including all governmental measures that can affect the allocation of resources across the different sectors of the economy, contributed to a higher rate of structural change in Brazil during the last five decades.

The late fifties witnessed the implementation of new capital-intensive industries, as a consequence of the import substitution policy (ISP) adopted, led by the metal- mechanical (especially vehicles) and the chemical industries (the second phase of

1 import-substitution, ISI-2). The design and implementation of industrial policy was carried out in very different political and institutional conditions in Brazil and this had an impact on industrial development.

As Fishlow (1990) pointed out, the import substitution policy in Brazil was compatible with accelerated industrialization and high rates of aggregate growth. Its share of regional income increased from 43 to 54 % during the 1953 to 1973 period.

Industrial deepening was carried out in the framework of Kubitschek's Plano de

Metas (Targets Plan) that provided consistent support for industrial development, including subsidies and closed markets for new industries1 during five years. The domestic political environment was always favourable to the "developmentalist" project, which was pushed forward even when mounting disequilibria in the domestic and external front became evident. There was a broad consensus in Brazil as to the need for rapid industrial growth, which sustained the "developmentalist" coalition2.

In 1974, Brazil adopted a specially ambitious program of industrial development, the II PND (Plano Nacional de Desenvolvimento), aimed at implementing a new set of capital (and technology) intensive industries, mainly in the intermediate and capital goods

1 The implementation of the Targets Plan was in charge of the so-called "Executive Groups", ad hoc bodies that managed specific areas in development planning, like vehicles, agricultural machinery and equipment, naval construction, heavy machinery, transportation and railways. These Executive Groups operated with considerable autonomy and were quite effective in overcoming bureaucratic resistance, as they were formed by representatives from various governmental agencies. An especially important role was played by the GEIA (Executive Group of the Automobile Industry), which offered significant benefits (exchange rate and tariff exemptions for imports of inputs and machinery, tax rebates and subsidized official credits by the Bank of Brazil and the National Development Bank) in exchange of a certain level of "nationalization" of the components of the car. The National Development Bank (BNDES), in turn, was an important player in the coordination of the investment efforts in the public and private sector (Leopoldi, 1991).

2 On the political conditions of the Targets Plan see Benevides (1976).

2 sectors3. This move was prompted by the 1973 oil crisis and sought to "complete" the industrial matrix through a new wave of import-substituting industrialization. In addition,

Brazil attempted to diversify its export structure by increasing manufactured exports, especially to other developing countries. As a result, the import coefficient of the economy was further reduced, while the export coefficient increased.

In order to achieve this objective, a comprehensive array of policy measures was adopted, which included financial subsidies for the new industries, more rigorous import restrictions (based largely on non-tariff barriers, managed by the CACEX4) and subsidies to manufactured exports, combined with an active diplomacy towards developing countries in Africa, the Middle-East and Latin America5. The abundance of foreign capital was then instrumental in broadening the degree of autonomy that Brazil needed to finance the new industrial projects. As mentioned before, this industrialization drive of

Brazil succeeded in promoting the convergence of its industrial structure with respect to that of the industrialized countries (Bertola et al., 1998, 1999; Porcile at al., 2000).

The contrasting experience in industrial transformation of the Brazilian economy ended with the 1982 debt crisis. Brazil had followed policies that compromised (for different reasons) competitiveness and external equilibrium, and the array of subsidies provided by the ISI-2 represented an additional source of tension as the government faced a growing fiscal deficit. Moreover, the policies had been sustained on the basis of the

3 See Barros de Castro e Souza (1985). 4 Carteira do Comercio Exterior (Foreign Commerce Department of the Banco do Brasil).

5 However, trade relations with Argentina were restrained as a result of an enduring diplomatic conflict related to geopolitical rivalry and the construction of the Itaipu dam. In addition, Brazil strengthened its diplomatic and economic links with Europe, especially with Germany, in order to set forward its nuclear project. 3 external debt. The increase of the international interest rates in the early eighties launched a financial crisis, put an end to the policies of the seventies and opened up "the lost decade", which was characterized by large resource transfers, high real interest rates, large deficits financed by internal debts, accelerating inflation and economic stagnation.

During the whole “lost decade” and the beginning of the 1990’s, the Brazilian economy experienced much macroeconomic instability, with chronic inflation rates, increasing poverty and income inequality, and low growth rates. In the period 1985 to

1993, there were five different economic plans, trying to control the inflation and promote growth. The stability and inflation control were achieved with the Real Plan in

1993-94, but it had a price with an increasing public debt6, contributing to important negative economic effects from Asian, Russian, Argentinean, and Brazilian crises in the following years.

In the beginning of the 1990’s, Brazil abandoned the import-substitution policy and started a policy with emphasis in free market and international market liberalization.

In 1990 the free trade area of the Southern Cone (Mercosur), with Argentina, Brazil,

Paraguay and Uruguay, was created, which was responsible for an overall fall in the main import and export tariffs of tradable goods among member countries, increasing substantially the trade among these countries.

In the last 50 years, the Brazilian economy passed through important transformations in all sectors, but it is possible to identify two distinct periods: the first is

6 The actual public debt is about 54 % of GDP, in recent information from the Brazilian central bank in September 2003. See www.bcb.gov.br for more information. 4 the period before 1994, where the economy was basically unstable and overprotected, and the second, after 1994, with economic stability and a trade-oriented economy, although some degree of protection is still in place mainly in the industry.

The Brazilian economy has showed important improvement in recent years, but there are some crucial problems that the government needs to account for, such as the increasing poverty and income inequality. Based on gains from trade without market distortion, it is believed that these problems would be relieved with an increase in

Brazilian trade and a fall in the remaining high effective import tariffs that “protect” the main industrial sectors, which suggests that the strengthening of the Mercosur, and the creation of the Free Trade Area of Americas (FTAA), may have an important role to play to reduce such important problems.

This dissertation is about international trade of the Brazilian economy, showing how trade can affect Brazil’s economic performance, how it can help to reduce poverty and income inequality, and also how important the macroeconomic imbalances are in explaining the trade flows between Brazil and other countries in the formation of new free trade areas. The dissertation is divided into three more chapters following this introduction. In chapter two a computable general equilibrium (CGE) model is used for the Brazilian economy, which analyzes the effects of overall and sectoral reductions in import tariffs on poverty and income inequality. The third chapter estimates a gravity trade model for Brazil in two different scenarios, Mercosur and Free Trade Area of

Americas (FTAA), in order to capture the main determinants of trade flows and the role that the volatility of exchange rates has in trade between Brazil and other trade partners.

Chapter four lists the main references cited throughout the dissertation.

5 Chapter two uses a CGE model to investigate the impact of tariff reduction on the

Brazilian economy, trying to identify a combined policy that can reduce possible negative impacts from trade reform on the poor. Brazil has not only high levels of poverty and inequality in the distribution of income, but also large regional disparities, which contribute to increased income concentration and poverty. Therefore, it is essential to capture and understand the main effects of tariff reduction in the Brazilian economy, since a large fall on tariffs is expected due to the Mercosur and the Free Trade of

Americas (FTAA) trade agreements. This study uses a single-country multiregional CGE model for Brazil, and performs different levels of reduction in import tariffs at overall and sectoral levels, in order to identify the trade policy that can bring gains for the poor and for the income distribution, and to design an alternative policy that can be combined with the trade reform in order to reduce poverty and improve the distribution of income.

Chapter three evaluates the lack of coordination of macroeconomic policies in the

Mercosur and in the proposed FTAA, which is expected to be implemented in 2005.

Different economic stabilization plans adopted at different times, and implemented by different countries in the Mercosur and in the proposed FTAA, can be responsible for most of the medium to long term real exchange rate volatility. Long swings in real exchange rate, caused by country-specific economic stabilization plans, can increase the level of uncertainty among domestic and foreign (trade partners) economic agents, bringing unexpected outcomes for trade. Therefore, this chapter investigates Brazil’s main trade determinants in the Mercosur and in the proposed FTAA, accounting for the possibility that the lack of stable macroeconomic policies might hurt Mercosur trade and it would be a problem to the implementation of the FTAA as well. The main focus is the

6 different effects of medium to long run exchange rate volatility on different sectors, to be captured through estimation of a gravity trade flow model based on a panel data bilateral trade between Brazil and 17 other countries, under different proxies of exchange rate uncertainty.

7 CHAPTER 2

SHORT TO MEDIUM RUN REGIONAL EFFECTS OF TRADE LIBERALIZATION: A COMPUTABLE GENERAL EQUILIBRIUM MODEL FOR BRAZIL

2.1. Introduction

A wave of trade liberalization policies started for many developing countries after

the Mexican crisis in the late 1980s. The main belief about such trade policies was that

free trade would bring welfare gains and growth for these countries. Brazil was one of the last countries to adopt such liberal trade policies. At the end of the 1980s, Brazil’s trade policy still had features of an import substitution regime. In general, import substitution policy aimed at helping in structural changes necessary to improve the Brazilian economy was not successfully implemented. The policy created externalities and distortions that resulted in frustrated attempts to reduce its consequences. The agricultural sector, for example, was penalized in order to finance the manufacturing activities. We can say that the main consequences of the import substitution policy were the use of import quotas, exchange rate controls, high import tariffs, overvalued exchange rates that contributed to unemployment and underutilization of capital, and the penalization of the exports. In

1988 the “New Industrial Policy” was launched, partially removing some non-tariff barriers (NTBs), and simplifying the taxation on imports and exports. But the changes

8 were not effective, since the system of import licensing still remained in place, including

NTBs, such as “the law of similar”, which prevented competitive imports of goods similarly produced in Brazil. In the early 1990s, under the Asuncion Treaty, Brazil established a trade partnership called Mercosur, with Argentina, Paraguay and Uruguay.

This partnership had the same purpose as NAFTA (the North American Free Trade

Agreement) and the European Union, which was the creation of a common market among countries to facilitate trade, to coordinate trade policy, and to distribute proportionately, increases in revenues generated. Recently, the formation of the Free Trade Area of

Americas7 (FTAA) and the ‘pros and cons’ about the insertion of Mercosur countries in this trade agreement have been discussed among the main policymakers and societies across the Mercosur countries.

Trade policy reforms are still being debated in Brazil and other South-American countries, and the process of import tariff reduction seems to be irreversible for them.

According to Winters (2002), developing countries can experience a higher degree of uncertainty due to trade liberalization, where the country becomes more vulnerable to trade shocks, such as commodity price booms and slumps or exchange rate changes, undermining policies to alleviate poverty8 and redistribute income. Another problem may come from the way that the government replaces, for example, the tariff revenues due to

7 The Free Trade Area of Americas will include all South, Central and North-American countries, and the main regulations and agreements in different sectors still in debate and negotiations.

8 It is true that the analysis of the poverty due to trade liberalization can be more general than the pattern of trade restrictions across countries. See Winters (2002) for more details. 9 import tariff reduction. Taxes can be raised, welfare reduced, and poverty increased. In general, economists often focus on the effects of trade policy reform on overall efficiency, ignoring equity effects (Harrison et al., 2003).

There are many studies dealing with the macroeconomic impacts of import tariff reduction in Brazil and other Latin American countries, but only a few evaluate the consequences of trade reforms on poverty and income inequality. Even though Brazil is the tenth largest economy in the world, the fifth in geographical size, and sixth in population, the relevance of studying the consequences of trade reforms in this country is of utmost importance. Almost 12 % of Brazil’s population lives in complete poverty, and it also has one of the highest levels of inequality in the distribution of income in the world (Barros et al., 2001). Brazil also has significant regional disparities, which contribute to income concentration and poverty. Since the expected implementation of the FTAA by 2005 implies reduction and harmonization of current tariffs, it is very important not only to analyze the overall economic results from tariff reduction in the

Brazilian economy, but also to consider its impacts on income distribution and poverty at the regional level.

Any general textbook in trade theory would emphasize the gains from trade, mainly in the long run, and it would indicate that a country removing any trade distortion would always gain from opening its economy. In general, trade reforms would bring gains for a country in the long run, since there would be enough time to have a better allocation and distribution of resources, improving the overall economy. The problem is

10 the uncertainty about short to medium run effects of trade reforms, mainly when there are prior regional disparities in poverty and income distribution as in Brazil, resulting in some households that win and others that lose from such reforms.

This study is devoted to assessing the economic impacts of a reduction in import tariffs on poverty and distribution of income, identifying a combined policy that can reduce possible negative impacts from trade reform on the poor, through a single-country multi-regional computable general equilibrium model (CGE) applied to Brazil.

The use of a single-country CGE model is justified by De Melo and Tarr (1992), who used a similar type of model to estimate effects of the removal of protection in the

United States. They argued that multi-country models might overestimate the terms of trade effects induced by a unilateral reduction in protection.

This chapter is organized as follows. The next section discusses the main issues of the study. Section 2.3 describes the main objectives of the study. Section 2.4 contains a literature review for selected trade CGE models and their main features and results, and the main CGE studies about trade applied to Brazil. Sections 2.5 and 2.6 introduce the model framework, data source and describe the main features of the model. Section 2.7 discusses the design of the simulations to be performed. The results and discussion are in section 2.8. Section 2.9 has the main conclusions of the study.

2.2. The Issue

The trade liberalization to be analyzed in this study is the elimination of import tariffs for many goods, and it can be considered as one of the main components of the structural adjustment policy measures in many developing countries. Not only the

11 traditional neo-classical theory indicates that a country benefits from free trade, but also some new arguments about spillover effects, economies of scale, or benefits from technological progress would also result in the same conclusion. The main argument is that the gains are obtained at the same moment the trade barriers are removed, as trade controls absorb government resources and cause net welfare losses.

A decrease in import tariffs will reduce the price of imported goods, implying that imports increase and the price of composite goods, which is given by domestic and imported commodities available for domestic consumers, is reduced due to the increased share of cheaper imports. The real exchange rate depreciates, which in turn improves the competitiveness of the export sector (Sorsa, 1999).

According to Mehlum (2002), the export sector experiences gains in relative prices with trade liberalization, which causes a short-term deficit in the current account balance. One of the reasons for this is the shift in demand away from domestic producers, which reduces capacity utilization, increases imports, and slow down exports growth.

However, provided that enough investments are taking place in the export sector, there may be improvements in the current account balance. The investments increase with the profits in the export sector, and the following periods show growth and improvement in the current account. Therefore, trade reform brings positive results only in the long run, with a positive investment response9.

Lopez and Panagariya (1992) say that if the import tariff reduction is

implemented for only a subset of goods, the results can be very different. They rely on a

9 Of course some other factors can affect the long-term responses of investments and the overall success of the trade reform as well, such as the economic and political environment of the country, since the degree of credibility of the reform plays an important role in this process. For more details, see Rodrik (1992) and Mehlum (2002). 12 theoretical result called the “concertina theorem”, which says that in a small open economy if the highest tariff rate is reduced to the next highest one, welfare will improve as long as the import demand for the good with the highest tariff presents gross substitutability with all other goods. Therefore, in the presence of a pure imported intermediate good the substitutability condition of the “concertina theorem” may be not satisfied, and the welfare results from the tariff removal become undefined.

According to Winters (2002), in the short run, trade liberalization puts great pressure on some economic agents and that, even in the long run, successful open regimes can leave some others in poverty. Even though there is a strong presumption that the long run effects from trade liberalization lead to pro-poor growth, the true effects differ among households and across countries.

A major policy concern here is the link between trade policy reform and poverty in Brazil. Although there are possible gains from trade in the long run, the general problem addressed in this study is to evaluate the consequences of import tariffs reduction in the short to medium run. Specifically in the case of Brazil, what are the main consequences of import tariff reduction in the presence of regional disparities, high poverty level and unequally distributed income? What would happen to the rural and urban poor? If there are some sectors after the trade reform that hurt the poor, should such sectors be excluded from reform? Would it be possible to implement any compensation scheme for those people hurt after the fall in the import tariff?

13 The questions posed represent important issues to be carefully analyzed by any government willing to implement trade reform based on import tariff reduction, since it is possible that the losses from such reform exceed the gains, worsening the overall welfare within the country, increasing income concentration and poverty.

The proportion of rural and urban people in Brazil varies according to the region.

In the North and Northeast, the rural population represents about 30 % of the total, and in the Southeast there is only 10 % of population in rural areas. Considering the whole country, the average proportion of rural population is around 20 % (IBGE, 2000a). The low-income people10 in the North and Northeast are, respectively, 64 % and 79 %. In contrast, in the Southeast this proportion is 48 %, which stress some of the regional disparities seen in Brazil. The North and Northeast are the most remote areas, and they are exactly those that have less infrastructure and health resources available, contributing to worsening the poverty problem. Although the Gini coefficient, which measures the degree of inequality in the distribution of income, has decreased in recent years, the

Brazilian income is still one of the most unequally distributed in the world11, with a coefficient around 0.58, as shows Figure 2.1.

10 According to the Demographic Census 2000 (IBGE, 2000a), the low-income here represents people whose total monthly earnings are less than two minimum wages, approximately US$ 140.

11 According to information from the World Bank, South Africa and Malawi are the countries with the highest degree of income inequality, with Gini coefficient respectively of 0.62 and 0.61. Brazil is the third in this list (Barros et al., 2001). 14 0.605

0.6

0.595

0.59

0.585

0.58

0.575

0.57 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001

Source: IBGE

Figure 2.1: Gini coefficient for the income distribution in Brazil, 1992-2001

According to Table 2.1, the Gini index reflects the distribution of income across regions and states, which emphasizes the regional inequalities in Brazil. The income is unequally distributed throughout the country, with very contrasting implications for households in each region or state. For instance, we can see the lower level of income inequality in Amapa (0.483), and the much higher one in Paraiba (0.644).

Since poverty is also an important problem in Brazil, Figure 2.2 gives a clear picture about the extent of poverty in this country. Poverty12 here means those people that are below the poverty line, which is defined as the cost of a basket of food that supplies the minimum calories needed for a person to live. The number and proportion of poor has

12 Some authors such as Barros et al. (2001) consider the concept of poverty used here as indigence.

15 increased substantially from 1977 until 1986, where the latter was the year which an important, but inefficient, economic plan was implemented, called the Cruzado Plan. In

1985 and 1986 Brazil grew at a rate higher than 7 % per year, but the drastic fall in the inflation rate in 1986 seems to be the main result from the Cruzado Plan, which could be the main cause of the reduction in number and proportion of poor for that year13.

However, the economic plan failed, and poverty increased in the subsequent years. In

1990 Brazil had more than 30 million people living below the poverty line (more than 20

% of the population), the inflation rate was above 2,700 %, and the economy was shrinking. This scenario continued until a fourth attempt to implement a sustainable economic plan in 1994, called Real Plan, dramatically changed Brazil’s economic performance. After 1994, the annual inflation rate fell to levels below 10 %, bringing growth and stability to the country, contributing to reduced poverty levels (Figure 2.2).

Although poverty was reduced in Brazil after 1995, its level is still very high, with the necessity to implement many actions to reduce it.

The slow process of import tariff reduction that has occurred in Brazil in recent years has important consequences for urban and rural households and also for poverty and income distribution. Due to the diversity of households in Brazil and to the disparities and distributional issues discussed so far, it is likely that any trade reform will bring unequal distribution of gains for households at least in the short run.

13 Barros et al. (2000) study the exact influence of macroeconomic instability (inflation, growth, employment) on poverty and income inequality. Among their findings, the inflation rate plays an important and essential role in determining the poverty and income inequality in Brazil. 16

Regions and states Gini index Regions and states Gini index

North 0.547 Sergipe 0.589

Rondonia 0.543 Bahia 0.558

Acre 0.588 Center-West 0.573

Amazonas 0.488 Mato Grosso do Sul 0.548

Roraima 0.493 Mato Grosso 0.528

Para 0.556 Goias 0.549

Amapa 0.483 Distrito Federal 0.595

Tocantins 0.560 Southeast 0.537

Northeast 0.587 Minas Gerais 0.549

Maranhao 0.592 Espirito Santo 0.549

Piaui 0.609 Rio de Janeiro 0.532

Ceara 0.598 Sao Paulo 0.514

Rio Grande do Norte 0.572 South 0.543

Paraiba 0.644 Parana 0.561

Pernambuco 0.586 Santa Catarina 0.504

Alagoas 0.529 Rio Grande do Sul 0.544

Source: IBGE (2002)

Table 2.1: Gini index for regional monthly labor income for people 10 years old or older in Brazil, 1999

25 ) s n o i l l i 20 m (

d n a

) % ( 15

5 1977 1979 1982 1984 1986 1988 1990 1993 1996 1998

Source: PNAD (IBGE) Number of Poor (millions) Proportion of Poor (%)

Figure 2.2: The poverty problem in Brazil, 1977 to 1999

One feature of the policy analysis to be examined in this study is exactly how to mitigate the negative and positive welfare effects on the poor. Since some sectors after the import tariff reduction can bring negative impacts on the poor, we have as an important goal to find the best and the worst trade reform alternatives with respect to total sectoral or partial liberalization of the Brazilian economy. As pointed out by Harrison et al. (2003), it can be dangerous to suggest sector-specific liberalization, as it could induce political lobbying by those sectors that have been protected through high import tariffs.

This study can be useful to verify the validity of lobbyist claims that some sectors should or should not be protected in helping the poor.

18 To address the issues discussed so far, we are going to use a single-country multiregional CGE model for Brazil, and perform different levels of reduction in import tariffs at overall and sectoral levels, in order to identify the trade policy that can bring gains for the poor and for the income distribution, and to design an alternative policy that can be combined with trade reform in order to reduce poverty and improve the distribution of income.

2.3. Objectives of the Study

After having defined that the main issue of this study is to address the consequences of import tariff reduction on poor households and inequality in the distribution of income in Brazil, taking into account the regional characteristics of the production sectors and factor allocation, we can establish the following as the main objectives of the study:

‹ Evaluate the effects of different levels of reduction of import tariffs on poverty

and income distribution in rural and urban areas of Brazil, and on the regional

production sectors and factor markets.

‹ Design an equity-efficiency policy to offset possible losses from the import tariffs

reduction in order to guarantee more equal opportunities of the gains from trade

for the population14, identifying the agricultural and non-agricultural sectors that

bring the most negative effects on poor rural and urban households through

impacts on welfare due to the import tariffs reduction.

The study will be divided in three stages:

14 The concept is based on the “theory of distributive justice”, or “Rawlsian equalitarian theory”. For more details, see Rawls (1971). 19 (i) Run the CGE model with elimination of import tariffs occurring at different

levels for all goods, in order to analyze the overall effect of trade reform on

poor households and inequality in the distribution of income. At this stage we

will try to identify those regional productive sectors that hurt the poor and

contribute to increased inequality in the distribution of income, accounting for

the overall gains and losses from the fall in import tariffs.

(ii) Evaluate the effects of a sector-specific import tariffs reduction on poor

households and income inequality, if possible trying to identify some elements

that could be important for designing policy alternatives that would give a

more equal distribution of the gains from removing the trade distortion

(import tariffs).

(iii) Evaluate the model under an equity-efficiency trade reform that can bring

more gains for the poor, reducing poverty and income inequality.

The three stages to be performed are represented by three different sets of

simulations. The first stage will be represented by scenario 1, which is composed by 50

% and 100 % overall reductions on import tariffs. The second stage will be performed by

scenario 2, which consists of import tariffs reductions of 50 % and 100 % in some

selected sectors. The third stage, to be performed by scenario 3, will be the

implementation of an equity-efficiency alternative reform, which will be represented by

overall or sectoral import tariffs reductions (50% and 100%), together with a 20%

increase in income tax rates15.

15 More details about the set of simulations to be used in this study can be found in section 2.7.

20 2.4. Literature Review

2.4.1. CGE and Walrasian Models

Computable general equilibrium models (CGE) can be defined as the fundamental macroeconomic general equilibrium links among incomes of many economic agents, demand, the balance of payments, and the production structure (Thissen, 1998). The term

“general equilibrium” refers to both a methodological viewpoint and a substantive theory.

Methodologically, the economy is considered as a closed and interrelated system in which a simultaneous solution for the equilibrium values of all variables is obtained.

Therefore, the presence of any exogenous shock to the system leads the equilibrium levels of the whole system to be recomputed16. From a substantive viewpoint, general equilibrium theory is referred to as the “Walrasian theory” (Mas-Collel et al., 1995).

According to Mas-Collel et al. (1995) the study of competitive market economies from a general equilibrium perspective was the origin of the “Walrasian theory of markets” (Walras, 1874), or “Walrasian general equilibrium theory”, which is a theory of the determination of equilibrium prices and quantities in a system of perfectly competitive markets. The Walrasian theory attempts to predict the complete vector of final demand and supply of goods using only the list of goods, preferences, endowments, technology, quoted prices for every good, and the price taking assumption for consumers and producers.

The Walrasian general equilibrium model has been considered as a pinnacle of achievements in economics comparable to that of theoretical physics (Schumpeter, 1954).

16 In the partial equilibrium approach the effect on endogenous variables that is not directly related to the shock is disregarded. The ceteris paribus feature of the partial equilibrium approach is not adequate when feedback effects from a particular policy change or a shock are considered to be significant. 21 According to Karunaratne (1998), this model was infused with life by the Leontief’s input-output empirics (Leontief, 1966). The proof of existence of equilibrium (Arrow and

Debreu, 1954) and the Walrasian system solution (Scarf, 1967; Scarf and Hansen, 1973) were responsible for the advance in both the theoretical and empirical foundations of general equilibrium theory.

CGE models are considered the numerical versions of the theoretical general equilibrium models, and they can be divided according to their origins, objectives and theoretical background in Walrasian CGE models and Macro CGE models (Robinson,

1989; Willenbockel, 1994).

Thissen (1998) defines a Walrasian CGE model as the attempt to make the general equilibrium approach of Walras operational, and its origin comes from applied welfare economics theory. Walrasian CGE models are simply the numerical counterparts of Walrasian general equilibrium models. The aim of this type of model is to convert the

Walrasian general equilibrium structure from an abstract representation of an economy into realistic models of actual economies to be used in evaluating policy options by specifying production and demand parameters and incorporating real data (Shoven and

Whalley, 1984). This type of modeling started with Harberger’s (1962) study on the incidence of taxation in a numerical two-sector model, and it was popularized by many others later. The work of Scarf (1967) made the determination of the equilibrium of a

Walrasian system possible.

Thissen (1998) defines a macro CGE model as an extension of Leontief’s input- output analysis and linear programming models. Shoven and Whalley (1984) consider these types of models as empirical Walrasian models based on fixed input-output

22 coefficients by including substitution effects in production and demand, and also including more than one consumer. Johansen’s (1960) model with simultaneous determination of prices and quantities in the Norwegian economy is considered the first model in this category of CGE models (Thissen, 1998). Other models included in this category are the ORANI/MONASH models (Powell and Lawson, 1990; Vincent, 1990).

The objective of Macro CGE models is to quantify short run income distribution and resource allocation, sectoral growth and trade balance effects of shocks or policy alternatives. These models may include ad-hoc specifications and the behavior of economic agents may not be derived from optimization behavior (Thissen, 1998).

According to Qiang (1999), there are three dominant schools in the field of applied economics: the Norwegian/Australian linearizers school, North American levels school, and the mathematical programming/development planning school proposed by

Powell and Lawson (1990). Linearizers follow the Johansen tradition of a linearized solution technique, in which the equations of the model are log-linearized to permit the model to be solved by inverting a single matrix. The ORANI model is an example of this approach. In the levels school the non-linear general equilibrium systems are solved in levels rather than in log-linear form. Hertel et al. (1992) say that although the

“linearized” approach has a more straightforward representation and produces results easier to explain, the “levels” approach offers a more natural starting point for expressing accounting identities. The CGE model to be implemented in this study will be the

“levels” approach.

23 2.4.2. CGE Models to Evaluate Changes in Trade Policy

The proliferation of CGE models since the pioneering studies of Harberger (1962) and Johansen (1960) has occurred in many areas, such as trade and development

(Adelman and Robinson, 1978; Dervis et. al, 1982; De Melo, 1988; Robinson, 1989), and recently many trade policy issues have been addressed using many different CGE models applied worldwide17.

Bautista and Thomas (1997) examined the impact of alternative trade policy adjustments on income and equity, focusing on low-income rural households in the

Philippines. Using a CGE model and a Social Accounting Matrix (SAM) for 1979, they simulated three different trade policies: import rationing, uniform surcharges on imports, and trade liberalization. Markets for goods, factors and foreign exchange were assumed to respond to changing demand and supply conditions. The model had five agricultural sectors, three rural and two urban households, and four primary factors. The technology used was represented by a set of nested CES and Leontief functions. The composite good was a CES aggregate. Their model assumed that consumers minimize the cost of obtaining the composite good, based on a Cobb-Douglas utility function, and producers maximize revenue from sales. The simulation results showed that with a 5 % uniform reduction in the import tariffs, there was a 50 % reduction in the current-account deficit, suggesting that this is an attractive policy reform. Results indicate that the worst possible situation for the economy as a whole would be to impose an import tariff. Trade liberalization seemed to be the best among the three policies in terms of both efficiency

17 For literature surveys see Shoven and Whalley (1984), Srinivasan and Whalley (1986) and De Melo (1988). 24 and equity concerns. The authors conclude that rural Philippine households were penalized by the imposition of import rationing and of general import surtax. Fast and equitable growth cannot happen with inappropriate trade policies.

Bautista et al. (2001) compared partial and general equilibrium approaches in evaluating the effects of the policy intervention effects in agriculture in Tanzania. They considered two assumptions regarding substitutability between domestically produced and imported goods: perfect versus imperfect substitutability of imports and domestically produced goods. The study had four simulations. The first was an import substitution industrialization strategy with an import tariff on non-agricultural goods. The second simulation was the same as the first with a fixed exchange rate. The third and fourth simulations imposed a tax on agricultural exports with free and fixed exchange rates, respectively. The general equilibrium results suggested that trade policies have a less negative effect on relative prices in agriculture than those indicated by partial equilibrium analysis. The non-agricultural tariff reduced the terms of trade for this sector. The imposition of an export tax on all agricultural sectors with a fixed exchange rate was responsible for a lower deterioration of the terms of trade in comparison with that of a free exchange rate.

Cattaneo et al. (1999) developed a CGE model for Costa Rica using a SAM for

1991. It consisted of 25 production sectors, seven types of households and one aggregate enterprise account. They simulated trade liberalization under fixed and free exchange rates, with possible compensation for the loss of tax revenue through an increase in taxation in the domestic market. The results obtained suggest that the changes in

25 domestic prices are significant due to trade liberalization. However, the effects on income were very small, because all households receive some type of capital income. With tariff reduction, there was an increase in GDP due to the increase in agricultural production18.

Davies et al. (1998) studied the short run consequences of trade liberalization in

Zimbabwe using a five-sector CGE model based on a SAM for 1985. Full liberalization would lead to an increase in intermediate imports that could increase the domestic production of final goods. Demand for imported final goods would increase more than demand for domestic final goods. To alleviate this problem, exchange rate devaluation could be undertaken. They conclude that trade liberalization creates short run problems19 and this is the main reason liberalization has been so controversial.

Chou et al. (1997) estimated a single-country CGE model for Taiwan to evaluate the consequences of joining GATT assuming mobility of production factors among sectors, and unilateral versus multilateral negotiation trade liberalization. The ten simulations performed using a 14-sector SAM were exactly through unilateral and multilateral trade liberalizations with import tariffs and non-tariff barriers elimination.

Results show that liberalization benefits the domestic economy significantly, with increases in GDP, consumption and welfare.

Greenaway et al. (2002) study the increasing wage inequality in United Kingdom through a CGE model employing two different labor-type aggregations. They consider many different categories of labor according to the skill level and productive sector. Two

18 Chou et al. (1997) also applied a single-country CGE to Taiwan and concluded, with no surprise, that the economic gains from trade liberalization are positive and with particular benefits for households in terms of income and consumption.

19 These problems include consumption booms, short run contractions, drops of savings, demand switching to foreign goods, and growing trade deficits. 26 simulations are performed. The first is a decrease in trade barriers and the second an increase in the economic size of developing regions. Results simply state that trade has a minor role to play in explaining wage inequality, and that the skill-based technical change is the main force contributing to such inequality.

In contrast to the numerous studies available that deal with general effects from policy reforms in many countries20, there are not many CGE studies that address the poverty and equity concerns to capture effects from trade policies on households and overall economy. The use of a CGE model to evaluate equity issues started from studies such as Adelman and Robinson (1978), and Piggot and Whalley (1985), but just recently more attention has been given to the impact of trade reform on poverty and distribution of income through a CGE model. According to Khan (1997), while there are many studies relating trade to relative wages, there are just a few incorporating assessments of the size of the distribution of income21.

Gelan (2002) uses a urban-rural CGE model to examine the impacts of trade liberalization on structural changes and overall growth in Ethiopia. Results suggest that trade liberalization depends on wage-setting conditions on urban areas, with gains for urban and rural areas when the urban wages are flexibly determined. Although poverty and equity are not the main focus of Gelan’s study, the concern about different responses from rural and urban areas to trade reforms is already a good insight in this direction, since the rural population tends to be poorer than the urban one in Ethiopia.

20 Piggot and Whalley (1985), Ballard et al. (1985) and Whalley (1985) are examples of CGE models that have specified many households but have not made much about the distributional effects incorporated in their models.

21 For example, Deardorff and Haveman (1991). 27 The study of Lofgren (1999) is interesting not just because it simulates reduction in trade barriers but also uses complementary policies to protect rural households. The study consists of a CGE model for Morocco to evaluate the short run equilibrium effects of alternative scenarios for reduced protection in agriculture and industry. The model has a detailed specification of agricultural and other rural production, the labor market and households (divided in four categories). The main simulation results show that less agricultural protection would produce overall welfare gains at a cost of worsening the rural poor. Simulation of trade liberalization together with government transfers to owners of agricultural resources provides gains more evenly distributed among all households.

Even though the paper of Konan and Maskus (2000) does not evaluate the impact of policy reforms directly on poverty and inequality, it is a study that addresses the issue of trade liberalization at the same time that allows domestic taxes to adjust endogenously to satisfy a real government revenue target. Through a CGE model for Egypt, they decompose the welfare gains into effects from tax reform, trade reform and from their interaction. Conclusions show that welfare effects depend on the type of tax selected to replace the loss of tax revenue. Trade and tax reforms are important, but neither dominates.

Indeed, the link between trade and tax reforms is very important to account for implementation of any of these policies. Most studies of the welfare impacts of trade reforms have ignored the interactions of these policies with existing economy-distorted taxes, whose negative impacts can be even larger than the positive ones from the trade reforms in the second-best world (Williams III, 1999).

28 Harrison et al. (2003) is a good example of a study that not only addresses the poverty and equity effects from trade liberalization, but also accounts for a value added tax adjustment to assure tax revenue neutrality and equity concerns. Without complementary reforms, it might be the case that no trade reform is possible to bring welfare gains, due to second best effects. The authors stress that there are not many studies that attempt to capture the equity effects of policy reforms22. Harrison et al.

(2003) use a CGE model for Turkey to evaluate the equity effects from trade reform.

They use a SAM disaggregated in 40 different household categories defined by income and urban status. They design a trade reform package that includes a revenue replacement, which mitigates the negative effects of trade reform on the poorest households. Results show that the sum of welfare gains over all households is positive, but some of the poorest households lose from the reform.

2.4.3. CGE Studies About Trade in Brazil

There are many studies that try to capture the impacts of trade policies and regional integration on the Brazilian economy. Some of them are partial equilibrium studies (Carvalho and Parente, 1999), which fail to consider the regional integration as a general equilibrium phenomenon, producing biased estimates. Other studies use a general equilibrium approach to study issues related to Mercosur policies, such as Campos-Filho

22 Studies like Fougere and Merette (2000), Harrison and Rutherford (1999), Keuschnigg and Kohler (2000), and Rutherford (2000). 29 (1998) and Flores (1997); and others, such as Haddad (1999), Haddad and Azzoni (2001), and Carneiro and Arbache (2002), analyze issues related to unilateral liberalization and their implications for resource allocation.

Carneiro and Arbache (2002) used a CGE model to analyze the labor market reactions to trade liberalization. They tried to assess, specifically, whether a rise in exports is likely to yield a rise in employment and income in Brazil, using data from a social accounting matrix of 1996. They implemented three simulations to investigate the impact of an overall tariff increase to 1990 levels, the results of a selective export promotion policy oriented to skilled-labor intensive sectors, and the impact of a productivity shock on the economy. Simulation results confirmed previous findings in the literature that trade liberalization has a limited capacity to affect labor market outcomes in Brazil, and suggests that exports do not necessarily raise the employment level of less- skilled workers, as expected. Results have shown that trade liberalization contributes to improved economic welfare by means of greater output, lower domestic prices, and higher labor demand, but the benefits of this economic improvement tend to be appropriated by the most skilled workers in the most trade-oriented sectors.

Haddad et al. (2002) evaluated different strategies of economic integration for the

Brazilian economy. They evaluated three different trade liberalization scenarios through an interregional model integrated to a CGE model and a national CGE model. Results show that the trade strategies tested are likely to increase the regional inequality in Brazil.

Their main concern was the consequences in the regional inequality due to the Brazilian trade liberalization. Although this study evaluates regional short run effects of trade

30 liberalization, it does not address the income inequality and poverty that are very heavily affected by the regional distribution of resources, population, and production sectors in the Brazilian economy.

Regional integration for Mercosur is analyzed by Monteagudo and Watanuki

(2001). They investigated the impact on Mercosur after two different free trade agreements: Free Trade Area of Americas (FTAA) and free trade with the European

Union (EU). According to these authors, these agreements are about to bring gains in both trade and GDP growth to Mercosur countries, and also many structural changes with relevant economic and political consequences. They used a multi-country CGE model for

Mercosur, including features such as trade-linked externalities and scale economies in manufacturing industries. Their findings suggest that with the removal of tariffs and non- tariff barriers, the FTAA seems to be a better option for Mercosur countries. Free trade with EU would be a better alternative when only tariffs were removed. The integration seems to have a strong effect in Brazil, stimulating export specialization in manufacturing industries relative to the primary sector.

Rodrik (1991) has an interesting viewpoint with respect to the trade liberalization in South America. Historically, trade liberalization that has been happening in South

America brought some positive changes in the structure of production and consumption and gains in efficiency. But the costs of productive restructuring have been high, contributing to high unemployment rates and low quality employment.

One of the reasons for these “bad reactions” from trade liberalization in South

America, according to Stiglitz (1999), is that these countries have comparative advantages in the wrong place. Vaillant and Ons (2002) also agree with this idea,

31 pointing out that these countries are mainly exporters of goods with intensive use of natural resources, basically agricultural commodities, facing a distorted international market by the protectionist policies of developed countries.

Flores (1997) uses a CGE model with imperfect competition to evaluate the gains from Mercosur for Argentina, Brazil, Paraguay, and Uruguay. The study focuses on induced trade flows, on changes in welfare sources in the imperfect competitive sectors, and on total welfare gains in each country. Simulations performed are variations on the level of tariffs in different scenarios of free trade agreements (including agreements of

Mercosur with NAFTA and European Union). The results, in general, show that the gains for Uruguay are more significant. Outcomes for Brazil and Argentina seem to be tightly linked.

The pioneering work of Taylor et al. (1980), and Lysy and Taylor (1980) that evaluate the income distribution in Brazil using a general equilibrium model are the only studies that consider the effects of economic policies and programs on the size distribution of income in Brazil. In Lysy and Taylor (1980) the investment is fixed in real terms, and all income earnings are fixed in nominal terms. The effect of devaluation is examined and found to increase government savings by reducing export subsidy requirements and increasing tariff revenues. With the increase in savings, effective demand falls and causes unemployment, reduction of wages, and deflation. The latter helps the income distribution. They conclude that trade improves the distribution of income, increasing the income of the poorest households.

32 Barros et al. (2000) is one of a few studies known so far that addresses the impact of trade liberalization on poverty in Brazil. They used a CGE model and simulated an increase of protection to the same level as in 1985. They conclude that trade liberalization is beneficial for the whole country, but mainly for both urban and rural poor households.

In order to close this section, it is interesting to note that there are very few studies dealing with the general equilibrium effects of trade liberalization on poverty and inequality in the distribution of income in the literature of CGE models applied for

Brazil. This can be viewed as ironic, because Brazil has one of the largest income inequalities in the world, and poverty has been a systematic problem as well.

2.5. Social Accounting Matrix (SAM)

According to Reinert and Holst (1997), a social accounting matrix (SAM) is a form of single-entry accounting, which for every receipt there is an equivalent expenditure, and it records transactions between accounts in a square tableau or matrix format. A SAM can provide a consistent and comprehensive record of the interrelationships of an economy at many levels, such as production sector, households, factors, government and foreign institutions. Table 2.2 shows a representation of a standard aggregated SAM with the main features of a SAM in order to be used in the standard CGE model (Lofgren et al., 2001).

The disaggregated Brazilian Social Accounting Matrix (SAM) to be used in this study was constructed for 1995-96 by Andrea Cattaneo, of the Economic Research

Service’s Resource and Environment Policy Branch (USDA) (Cattaneo, 1998), and it was primarily generated from 1995 Input-Output tables for Brazil (IBGE, 1997a), National

33 Accounts (IBGE, 1997b), as well as the Agricultural Census data for 1995-96 (IBGE,

1998). According to Cattaneo (1999), total labor, land and capital value added were allocated across the agricultural activities based upon the Agricultural Census. The description of the SAM is summarized in Table 2.3, and the aggregated numerical version can be seen in Appendix A (Table A.1). It captures both regional and small and large-scale productive technologies. Four agricultural categories (annuals, perennials, livestock, and other agriculture) are disaggregated by holder size (small and large). The

SAM also includes three manufacturing activities, three service activities, and 24 commodities. There are 18 labor categories; including 10 urban (further disaggregated by skill level and sector) and 8 agricultural (by skill level and region); 9 capital categories, 8 of which are agricultural and distinguished by holder size and region; and 12 land categories disaggregated by land type (arable, grassland, and forested) and region.

Finally, the SAM includes five household accounts (rural and urban by income level), three tax accounts, a savings as well as inventory account, and one account each for enterprises, government, and rest-of-world (ROW).

; s e s i r p

r l e t e n d e

o –

T E N E G

; s C

d l d o r h a e d s n u a o t h

–

e S

h t d U l

r n O o i

H w

d ; e s e r h s t o

t u f

c o e a

f t r

s u – e t

c T – u

C r t A W s F

O ; s R M

e ; i t t A i n S d e

o m a t

m s f e m o

v o n n c i

- o – s i

t g p M n i i r O v a c C s

; e s –

e d I i

- t l i S a v

i ; r t t e c n a n e

e – m

n G T r

: e C v 2 . A o

2 g :

e – e

r l e b V h a O T

W G

Activity Commodities produced Factors used

Annuals Corn, rice, beans, manioc, sugar, soy, Arable land, unskilled rural labor, production horticultural goods, and other annuals skilled rural labor, agricultural capital

Perennials Arable land, unskilled rural labor, Coffee, cocoa, other perennials production skilled rural labor, agricultural capital

Grassland, unskilled rural labor, Animal products Milk, livestock, poultry skilled rural labor, agricultural capital

Non-timber tree products, timber, and Forest land, unskilled rural labor, Forest products deforested land for agricultural purposes skilled rural labor, agricultural capital

Arable land, unskilled rural labor, Other agriculture Other agriculture skilled rural labor, agricultural capital

Food processing Food processing Urban skilled labor, urban unskilled labor, urban capital

Mining and oil Mining and oil Urban skilled labor, urban unskilled labor, urban capital

Industry Industry Urban skilled labor, urban unskilled labor, urban capital

Construction Construction Urban skilled labor, urban unskilled labor, urban capital

Trade and Trade and transportation Urban skilled labor, urban unskilled transportation labor, urban capital

Services Services Urban skilled labor, urban unskilled labor, urban capital Source: Cattaneo (1999).

Table 2.3: Summary of activities, commodities, and factors included in the 1995 Brazilian SAM

36 2.5.1. Regional Sectoral Disaggregation

In order to identify more accurately the sectors’ influence on households due to import tariff reduction, it is necessary to disaggregate the national accounts for manufacturing and services sectors, and capital and labor used in each sector in four different regions (North, Northeast, Center-West and Southeast-South) as accomplished in the original SAM.

The “tops-down” approach will be used to perform the disaggregation of national flows to regional levels, since the “bottoms-up” approach requires a great deal of data that are not fully available for Brazil23. It is assumed that each region always produces a fixed share of each sector’s national output (Higgs et al., 1988). The procedure is basically the same as the one performed in the ORANI Regional Equation System (Higgs et al., 1988), and also the one to obtain regional input-output tables described in Leontief

(1966).

The industry and services sectors will be disaggregated into four regions in three stages: regional intermediate consumption, regional value added (capital and labor), and taxes. The regional intermediate consumption will be calculated according to the regional participation on total intermediate consumption (IBGE, 2000b). The regional value added for capital purchases will be obtained through regional GDP participation (IBGE, 2000b), and labor purchases will be calculated by the regional proportion of people employed in each sector (IBGE, 2000c). The tax payments by each regional industry and services sector will be calculated through the regional participation on total value added (IBGE,

2000b). The flows of regional output for each disaggregated sector (industry and

23 See Liew (1984) for a good evaluation of both “tops-down” and “bottoms-up” approaches. Higgs et al. (1988) give a third procedure that consists of a hybrid of both “tops-down” and “bottoms-up” approaches. 37 services) will be obtained through the regional output shares of each sector. The household income from the regionalized labor categories used by the regional industry and services sectors will be obtained through the regional shares of people employed by each sector according to the income level (IBGE, 2001). Finally, the payments made to enterprises by the regionalized capital categories used in each regional industry and services sectors will be obtained from the regional shares of enterprises in each sector according to the value added participation (IBGE, 2000c).

2.5.2. The Balance Procedure

In order to produce a more disaggregated SAM for detailed policy analysis, data are often supplemented by information from a variety of sources, as cited before from the

Brazilian Statistics Bureau (IBGE). The problem is to find an efficient way to incorporate and reconcile information from many sources, keeping the SAM balanced, that is, every column sum must equal the corresponding row sum.

The regional disaggregation described before will surely make the original SAM unbalanced, which will demand a balance procedure in order to make each column total equal the corresponding row total.

There are many ways to balance a SAM, starting from a simple use of a standard spreadsheet, and ending up with more elaborate and complex methods, including the

“RAS” procedure and the “cross-entropy” procedure. We will briefly comment about the cross-entropy procedure, which is going to be employed in our study, and can be used not only to balance a SAM, but also to update it for a more recent year.

38 The RAS procedure is a very common procedure that basically constitutes a special case of the cross-entropy (CE) procedure, when treating column and row coefficients symmetrically, using a single cross-entropy measure, instead of using the sum of column cross-entropies (Robinson et al., 2000)24.

The disaggregated SAM, in which new columns and rows will be created, can be seen as an inconsistent SAM with incomplete knowledge about both row and column totals and flows within the SAM25. According to Robinson et al. (1998), the problem with the “RAS” procedure in this case would be that this procedure would assume the

SAM is consistent, with full knowledge about the flows within the SAM. But since it is common to start with an inconsistent SAM, with incomplete knowledge about rows or columns sums, or about some flows within the SAM, incompatible data sources, or lack of data, it is critical to have an approach to estimating a consistent SAM that not only uses the existing information efficiently, but also is flexible to include information about various parts of the SAM. The approach that responds to these requirements, which will be applied in this study, is called a “cross-entropy” (CE) approach.

The cross-entropy (CE) procedure26 is not only flexible, but also allows incorporating errors in variables, inequality constraints, and prior knowledge about some flows within the SAM.

24 “RAS” procedure is also equivalent to maximizing a weighted sum of the column-coefficient cross- entropies.

25 It is true in our case because some of the information to be used in the disaggregating is based on personal judgments and from different sources.

26 For more details and explanation about this approach, see Robinson et al. (1998), Robinson and El-Said (2000), and Robinson et al. (2000). 39 As demonstrated by Table 2.3, a SAM is a square matrix whose columns and rows represent the expenditure and receipt accounts of an economy. Each cell shows a payment from a column account to a row account. According to Robinson et al. (1998) and Robinson et al. (2000), we can define T as the matrix of SAM transactions, where ti,j is a payment from column account j to row account i. The balance condition implies that every row sum equals the corresponding column sum (total receipts = total expenditures), resulting in the following expression:

(2.1) ƒti, j = ƒt j,i = yi j j

Where yi is total receipts and expenditures of account i.

It is possible to get a SAM matrix A with coefficients such that:

ti, j (2.2) Ai, j = y j

In addition all column sums of A27 must equal one, A is singular, and according to

(2.1), we have the following matrix notation:

(2.3) y = Ay

According to Robilliard and Robinson (2001) and Arndt et al. (2001), the starting point for the estimation procedure is information theory, as developed by Shannon

(1948), who defined a function to measure uncertainty (entropy) of a collection of events, and Jaynes (1957a,b) who proposed maximizing that function subject to appropriate consistency relations, such as moment conditions. The maximum entropy (ME) and its sister formulation minimum cross-entropy (CE), are now widely used in many fields to

27 The columns of A can be useful for economic analysis and modeling, for example, the intermediate-input coefficients are Leontief input-output coefficients, and the same usefulness is true for other columns of A. 40 estimate and make inferences when information is incomplete, highly scattered, and/or inconsistent (Kapur and Kesavan, 1992). The main idea behind this approach is to use all, and only, the information available for the estimation problem under analysis. Theil

(1967) was one of the pioneer studies with this approach applied in economics.

Information used in entropy estimation comes, generally, in two forms. First, information about the system that imposes constraints on the values that some parameters can assume. Second, prior knowledge of some parameter values. According to Arndt et al. (2001), in the first case, the information is included by imposing constraint equations in the estimation. In the second, it is necessary to specify a discrete prior distribution and estimate it by minimizing the entropy distance between the estimated and prior distributions (the minimum CE approach).

Golan, Judge and Miller (1996) bring the general regression model into the entropy approach by specifying an error term for each equation, without assuming any specific form for the error distribution. They specify a support set for error distribution and a prior on the moments of the distribution, which in our case will be symmetric about zero. The result is a flexible estimation method, which supports the use of information in many ways with different degrees of confidence28. The real power of this approach is that it makes efficient use of scarce information in estimating parameters. The new flows of the disaggregated SAM will be adjusted through the CE framework, using scarce information from priors such as known cells and macro totals (GDP, imports, exports, and others), instead of the traditional “RAS” approach.

28 The CE procedure also allows statistical inference. For more details, see Imbens (1997) and Golan and Vogel (1997). 41 The CE method to be used is the stochastic approach. This approach assumes the problem of extracting results from data and economic relationships with noise. It is assumed that row and column sums have errors in measurement, and also that a matrix exists (which is not a balanced SAM), similar to A, that yields a SAM transaction matrix

T* to exist, based on a new SAM coefficient matrix A*. Therefore, from (2.2), we have:

* * * (2.4) ti,j = ai,j yj where y* are known new row and column sums.

The CE approach can be applied to find the new set of A coefficients which minimizes the cross entropy distance between the prior and the new estimated coefficient matrix. It is assumed prior knowledge of the standard error of the estimate of control totals (a Bayesian prior), those SAM flows which will be disaggregated in different regions. It can be assumed that the initial column sums in the SAM are the best prior estimate. Therefore, the estimated error in the ith control total can be represented as a weighted sum of elements in a specified error support set, such as:

(2.5) ei = ƒ wi, jwtν i, jwt jwt

where ei is the error value of control total; wi,jwt are the error weights estimated in the CE procedure, which sum equal to one; and ν i, jwt is the error support set.

The prior on the variance of these errors is given by:

2 2 (2.6) σ = ƒ wi, jwtν i , jwt jwt where wi, jwt are the prior weights on the error support set and ƒ wi, jwt = 1. jwt

42 We assume the case of five-weight error distribution, with five weights w to be estimated (Robinson and El-Said, 2000). The moments include variance, skewness and kurtosis. Assuming a prior normal distribution with mean zero and variance σ2, the prior on kurtosis is 3σ4. The estimation procedure allows posterior estimates of all the moments of the error distribution (Golan, Judge and Miller, 1996). This specification also yields posterior estimation of four moments: mean, variance, skewness, and kurtosis.

The use of macro control totals can be very useful information to find a consistent

SAM through the CE approach. Aggregate national accounts can be available for this purpose. For example, we can get updated information about value added, consumption, government, investment, exports, and imports, through a country’s statistical agency.

This information can be included as additional linear adding-up constraints on various elements of the SAM, while allowing the possibility that additional information might be measured with error. It is given by equation (2.7):

(k ) (k ) (2.7) ƒƒGi, j Ti, j = γ + e2k i j

where G is an n-by-n aggregator matrix, which has ones for cells in the aggregate and

(k) zeros otherwise. Ti,j is the SAM cell values. γ is the value of the aggregate. The error term e2k is associated with macro aggregates, which will be specified as GDP at factor cost, GDP at market prices, total consumption, imports, and exports.

According to Robinson et al. (1998), the CE measures reflect how much the added information has changed the solution away from the inconsistent prior, also accounting for the imprecision of the moments assumed to be measured with error.

43 Therefore, if the information constraints are binding, the distance from the prior will increase. If they are not binding, the cross entropy distance will be zero. The CE estimation equations are described in Appendix B.

2.6. The Standard CGE Model29

The CGE model that will be used in this study is a regional adaptation of the so-

called “standard CGE model”, which was first developed and distributed through a

study30 of the International Food Policy Research Institute (IFPRI). The model follows the neo-classical-structuralist (Chenery, 1975) modeling tradition that is presented in

Dervis, de Melo, and Robinson (1982), and includes important characteristics developed in recent years in research projects conducted at IFPRI. Such characteristics are of particular importance in developing countries, and include household consumption of non-marketed commodities, explicit treatment of transaction costs for commodities that enter the market, and a distinction between producing activities and commodities that permits any activity to produce multiple commodities and any commodity to be produced by multiple activities.

This model consists of a system of linear and nonlinear simultaneous equations, with a set of constraints that cover market and aggregated macroeconomic variables. The

29 Lofgren, Robinson and Thurlow (2002), Thurlow and Van Seventer (2002) and Wobst (2002) applied this standard CGE model, respectively, to Zambia, South Africa and five Southern African countries.

Mathematical description of the model can be seen in Appendix 3.

30 For more details about this model, see Lofgren et al. (2001). 44 main role played by the system of equations is the exact behavioral description of the agents in the economy. The following sections present the main features of the model and some regional specifications of the standard model are found in Appendix C.

2.6.1. Prices, Activities, Production, and Factor Markets

Assuming that producers in each region maximize profits subject to the technology, taking prices as given, Figure 2.3 shows that this technology is specified by a

Constant-Elasticity-of-Substitution (CES) or a Leontief function of the quantities of value added and aggregate intermediate input. Value added is a CES function of primary factors, and the aggregate intermediate input is a Leontief function of disaggregated intermediate inputs. The equations related to these quantities can be seen in the production and commodity block in Appendix C. Each regional activity produces one or more commodities, or any commodity can be produced by more than one activity as well.

Activities employ factors of production up to the point where the marginal revenue product of each factor is equal to the factor price. The factor market closure to be used in this study considers that the quantity supplied of each factor is fixed at the initial level

(SAM). Labor is considered to be mobile across sectors, which is a medium run assumption. Capital and land are considered sector-specific. Hence, we expect that the resources will be reallocated to more productive uses, after reduction in import tariffs.

The wage is free to vary to assure that the sum of demands from all activities equal the quantity supplied. Therefore, the regional activities pay an activity-specific wage that is the product of the economy-wide wage and a fixed activity-specific wage term. The main

45 price, production, and commodity equations31 for each region are given below, where r represents region, c represents commodities, f represents factors, and a represents activities. The complete CGE model includes the regional equations in addition to the equations in Appendix C.

Commodity outputs (fixed yield coefficients) in region r

Activity Level (CES/Leontief)

Value Added Intermediate (CES) (Leontief)

Primary Factors Composite commodities

Capital in Labor in region r region r Imported Domestic

Land in Good Good region r from from region r region s

Figure 2.3: Regional production technology in the standard CGE model for Brazil

31 Description of parameters and variables can be seen in Appendix C.

46 Regional prices:

(2.8) PAa,r = ƒθ ac,r .PXACac,r (Regional Activity Price) c∈C

r (2.9) PINTAa,r = ƒ PQc .icaca (Regional Intermediate Input Price) c∈C

(2.10) PAa,r .(1− taa ).QAa,r = PVAa,r .QVAa,r + PINTAa,r .QINTAa,r (Regional Activity Revenues and Costs)

Production and commodity regional equations:

1 a a −ρ a a −ρ a a a a ρa (2.11)QAa,r = α a .(δ a .QVAa,r + (1−δ a ).QINTAa,r ) (Regional CES Activity Production Function)

1 a a QVA ≈ PINTA ’1+ρa a,r ∆ a,r δ a ÷ (2.12) = ∆ a ÷ QINTAa,r « PVAa,r 1−δ a ◊ (Regional CES Value added-Intermediate-Input Ratio)

r (2.13) QVAa,r = ivaa .QAa,r (Demand for Regional Value added)

r (2.14) QINTAa,r = intaa .QAa,r (Demand for Regional Intermediate Input)

1 vq ≈ va ’ ρa va ∆ va −ρa ÷ (2.15) QVAa,r =α a .∆ ƒδ fa .QFfa,r ÷ (Regional Value added and Factor Demands) « f ∈F ◊

−1 ≈ va ’ va ∆ va −ρa ÷ va −ρa −1 (2.16) W f ,r .WFDIST fa,r = PVAa,r .(1− tvaa ).QVAa,r .∆ ƒδ fa .QFfa,r ÷ .δ fa .QFfa,r « f ∈F ' ◊ (Regional Factor Demand)

r (2.17) QINTca,r = icaca .QINTAa,r (Regional Intermediate Input Demand)

r (2.18) QXACac,r + ƒQHAach,r =θ ac .QAa,r h∈H (Regional Commodity Production and Allocation)

1 ac ≈ ac ’ ρc −1 ac ac −ρc (2.19) QX c = α c .∆ƒδ ac .QXACac,r ÷ (Regional Output Aggregation Function) « a∈A ◊

47 −1 ≈ ac ’ ac ∆ ac −ρc ÷ ac −ρc −1 (2.20) PXACac,r = PX c .QX c . ƒδ ac .QXACac,r .δ ac .QXACac,r « a∈A' ◊ (First-order Condition for Regional Output Aggregation Function)

2.6.2. Institutions

Institutions are households, government, enterprises, and rest of the world.

Households receive income from payments for the use of factors of production,

and transfers from other institutions. They use their income to pay taxes, consume, save,

and make transfers to other institutions. Household consumption covers marketed

commodities, purchased at market prices that include commodity taxes and transactions

costs, and home commodities32. Household consumption is allocated across different

commodities according to a Linear Expenditure System (LES) demand function (it is

assumed that each household maximizes a “Stone-Geary” utility function subject to a

consumption expenditure constraint).

Enterprises can receive direct payments from households and transfers from other

institutions. Since enterprises do not consume, they allocate their income to direct taxes,

savings, and transfers to other institutions.

Government receives taxes (fixed at ad valorem rates) and transfers from other

institutions, and uses this income for consumption and for CPI-indexed33 transfers to

other institutions. Government consumption is fixed in real terms (quantity) and savings

is a flexible residual.

32 Home consumption is not present in the Brazilian SAM, but generally it is represented by payments from households to activities. 33 Government transfers indexed to the CPI makes the model homogeneous of degree zero in prices.

48 Transfer payments from the rest of the world, domestic institutions, and factors

are all fixed in foreign currency. Foreign savings is the difference between foreign

currency spending and receipts.

The main changes in the standard model for the institution block and constraint

equations can be seen below, where i represents institutions, and h represents households.

Institutions:

(2.21) YFf ,r = ƒWF f ,r .WFDIST fa,r .QFfa,r (Regional Factor Income) a∈A

(2.22) YIFif ,r = shifif ,r .[(1− tf f ).YFf ,r − trnsfrrowf ,r .EXR] (Regional Institutional Factor Incomes)

System constraints:

(2.23) QFS f ,r = ƒQFfa,r (Regional Factor Market Equilibrium) a∈A

2.6.3. Commodity Markets

For marketed output, the first stage in Figure 2.4 consists of generating

aggregated domestic output from the regional output of different activities of a given

commodity. Such regional outputs are not perfect substitutes. A Constant-Elasticity-of-

Substitution (CES) function is used as the aggregation function. The demand for the

regional output of each activity is derived from the problem of minimizing the cost of

supplying a given quantity of aggregated output subject to this CES function. Aggregated

domestic output is allocated between exports and regional domestic sales, where

suppliers maximize sales revenue for any given aggregate output level, subject to

imperfect transformability between exports and regional domestic sales, through a

49 Constant-Elasticity-of-Transformation (CET). The price received by domestic suppliers

for exports is expressed in domestic currency and adjusted for the transactions cost and

export taxes. The supply price for domestic sales is equal to the price paid by domestic

demanders minus the transaction costs of domestic marketing per unit of domestic sales.

All domestic market demands are for a composite commodity made up of imports

and domestic output. It is assumed that domestic demanders minimize cost subject to

imperfect substitutability. This is also captured by a CES aggregation function

(Armington function)34. The derived demands for imported commodities are met by

international supplies that are infinitely elastic at given world prices. Import tariffs and

fixed transaction costs are included in the import prices paid by domestic demanders. The

derived demand for domestic output is also met by domestic suppliers, and the prices

paid by demanders include the cost of transaction services.

The value of the elasticity of substitution between imported and domestic

commodities is based on Tourinho, Kume and Pedroso (2002), which estimated the

Armington elasticities for 28 industrial sectors in Brazil for the period 1986 –2001. Other

elasticities are borrowed from Asano and Fiuza (2001).

2.6.4. Macroeconomic Closures

According to Kraybill (1989), since a CGE model is basically a system of

equations, it requires a basic mathematical sufficient condition to assure the existence of

a solution: that the number of equations must equal the number of variables. The problem

is that this condition is not always satisfied when models are based on static general

34 Based on Armington (1969).

50 equilibrium theory. To have such a condition satisfied, it is necessary to drop one or more

equations from the system (or equivalently treat some variables as determined

exogenously), which is required in order to “close” the model. But which equations

should be dropped to guarantee a unique solution is a question that has many

consequences to the mechanisms that rule output, employment, and income distribution

in the CGE model. Therefore, there are many comparative-static macroeconomic closures

to be used in a CGE model.

Sen (1963), who was the first to use the term “macroeconomic closure”,

demonstrated that there are many ways to “close” an economic-wide static model. This

study employs the closure summarized in Table 2.4, in order to give enough information

for the simulations to be described in the next section.

For the government balance, the closure used here considers that the government

savings35 is a flexible residual while all tax rates are fixed. Therefore, the government

consumption is fixed, either in real terms or as a share of nominal absorption. For the

external balance, the real exchange rate36 is considered flexible while foreign savings is

fixed. The trade balance is also fixed, since transfers between rest of the world and

domestic institutions are fixed. For the savings-investment balance, closure is investment-

driven, where real investment quantities are fixed. This implies that, in order to generate

savings that equal the cost of the investment bundle, the base-year savings rates of

selected non-government institutions are adjusted by the same proportion.

35 It is defined as the difference between current government revenues and current government expenditures.

36 The Brazilian exchange rate policy in recent years allows flexible exchange rate fluctuations within a band as range controlled and determined by the Central Bank under government decision. 51

Commodity output Commodity output from activity 1 in . . . from activity n in region r region s

CES

Aggregate output CET

Aggregate Domestic sales Aggregate exports in each region imports

CES

Composite commodity

Household consumption +

Government consumption + Investment + Intermediate use

Figure 2.4: Flows of regional marketed commodities in the standard CGE model

52 According to Lofgren et al. (2001), the choice of macroeconomic closures

depends on the context of the analysis. Since it is a single-period model, a closure chosen

here with fixed foreign savings, fixed real investment, and fixed real government

consumption may be preferable for simulations that explore the equilibrium welfare

changes of alternative policies, as is the case of our study.

Assumptions Closure

Government balance Fixed government consumption; flexible gov. savings; fixed direct tax rates Investment-driven savings Fixed aggregate real investment; scaled MPS(*) for domestic institutions Balance of trade Flexible exchange rate; fixed foreign savings

Labor is fully employed and mobile; Factor market capital is sector-specific; flexible factor price (wages, capital rents, land price) (*) MPS = marginal propensity to save.

Table 2.4: Main assumptions and macroeconomic closure of the Brazilian standard CGE model

2.6.5. Inequality and Welfare Measures

Following the theorems of Heckscher-Ohlin-Samuelson and Stolper-Samuelson,

the relationship between increase in international trade, wage distribution and level of

employment has led several economists to conclude that recent internationalization of

economies has contributed to the increase of wage inequality and unemployment

(Arbache, 2001). The theorems cited are still the main analytical tools to explain the

53 relationship between international trade and distribution of income, but the case of

developing countries has received less attention.

According to Arbache (2001), empirical studies have shown that trade

liberalization in some developing countries is associated with an increase of the returns to

human capital and a worsening of the wage distribution. This is a puzzling result, since

developing countries have abundant unskilled labor, and the standard theory of

international trade would suggest that developing countries should specialize in the

production of those unskilled labor goods, increasing relative demand for this factor and

reducing the wage differences. Although the explanations for these empirical facts are

preliminary, they suggest that the opening to trade in developing countries induces a

simultaneous process of technological modernization and increase of capital stock,

inducing a positive impact in the demand for skilled labor, increasing the wage dispersion

and the distribution of income.

A growing empirical evidence for developing countries shows that trade is being

associated with an increase in the relative demand for skilled workers and rising wage

inequality, rejecting the predictions of the traditional theory of trade. It seems that Latin

America and other countries have experienced an increase of wage dispersion after trade

liberalization (Arbache, 2001).

Robbins (1995) found that the increase of exports caused an increase in wage

differentials in Colombia, caused by a positive correlation between the increase of

imports of machines and equipments, new technologies, and by a rising demand for

skilled labor. Robbins (1994) accounted for growing returns of skilled labor after trade

liberalization in Chile. Beyer, Rojas and Vergara (1999) also found a long term

54 correlation between openness and wage inequality in Chile. Robbins and Gindling (1999)

found similar results for Costa Rica as a result of the trade reform and consequent

changes in the structure of labor demand. The most plausible explanation is that trade

liberalization brings an increase on imports of capital goods that are complementary to

skilled labor.

Barros et al. (2001) used a CGE model to evaluate the impacts of trade

liberalization on the labor market, and found no significant impact of openness on income

inequality. Menezes-Filho and Rodrigues (2001) used a decomposition analysis to verify

the increase in demand for skilled labor in the Brazilian manufacturing industries after

trade liberalization. Once again, the results showed that the introduction of new

technology requires more skilled labor. Arbache and Corseuil (2000) found that

employment shares in Brazilian manufacturing are negatively associated with import

penetration, with larger effect for industries intensive in unskilled labor. Finally, Maia

(2001) used an input-output analysis to investigate the impact of trade and technology on

skilled and unskilled labor in Brazil, before and after openness. She concluded that trade

destroys more unskilled labor and that technology created more demand for skilled jobs.

In order to verify the impacts of reduction in import tariffs on poor households

and on income inequality, and to design equity-efficiency policies to offset possible losses from this trade reform, we need to define what would be the tools to quantify such effects.

When policy simulations are carried out, factor prices, transfers, or other endogenous

variables may change, which modify not only the total households net income, but also

any representation of the distribution of income (Khan, 1997). Most studies employ

measures such as Gini coefficient, and Theil index, but there are many other measures

55 that could be employed such as the log variance of income, the Atkinson measure, and

others.

We will use the Gini coefficient and the Theil index as income inequality

measures for the overall economy. The Gini coefficient and the Theil index are

represented (Khan, 1997), respectively, as:

»1 ' ' ' ' ' ' ' ÿ (2.24) G = 1− 2ƒ … (f h − f h−1 )(θyh −θyh−1 )+ ( f h − f h−1 )θyh−1 Ÿ h 2 ⁄

' ' Where f h is the cumulated population share of the household groups, and θyh is the cumulated income share.

1 (2.25) log H y log − ƒhθ h θyh

Where yh is the average income of households in category h, and θyh is the income share of the hth household group.

The main measures of inequality to be used at the regional level are the Gini coefficient (index), through its decomposition, and some generalized entropy inequality measures such as Theil, Hirschman-Herfindahl, and Bourguignon indexes. We will use a decomposition of these four indexes in order to better evaluate the impacts of import tariff reduction on households at a regional level37.

According to Silber (1989), Dagum (1997a), and Mussard (2003), we can decompose the Gini index by factor components when detailed income sources are available, as is the case of our regional standard CGE model and the available SAM. It is

37 There is no CGE model known so far that has implemented this approach to verify detailed consequences from counterfactual simulations on households.

56 possible to break down the inequality into within and between classes inequality when there are groups with different income ranges. Since our data contain not only different household groups arranged by income, but also by location (urban and rural), or population subgroups, with income sources from activities from different regions, we can also have some interaction term38.

For Dagum (1997a) and Mussard (2003), the Gini index measured for a population P with n income units yi (i = 1,…, n) can be written as a variation of our previous expression (2.24) as:

n n

ƒƒ yi − yr (2.26) G = i=1 r=1 2n 2 µ

Where µ is the income average of P.

The Gini index within a subpopulation Pj (j = 1, …, k) is given by:

n j n j

ƒƒ yi − yr i=1 r=1 (2.27) G jj = 2 2n j µ j

The Gini index between groups of subpopulation Pj and Ph is given by:

n j nh

ƒƒ y ji − yhr i=1 r=1 (2.28) G jh = µ j + µh

38 It is also possible to decompose the Gini index by factor components, which in our case will not be feasible in the case of capital income, since the capital rents are paid to the enterprises account, and not directly to households. Therefore, we will only consider the case of decomposition when we have different income classes. 57 According to Dagum (1997b), the gross economic affluence, which represents the expected income difference between groups j and h, such as yji > yhr and µj > µh, is given by:

∞ y

(2.29) d jh = — dFj (y)— (y − x)dFh (x) ∀µ j > µh 0 0

Where F(.) is the cumulative distribution of income. Pj is in mean superior to Ph.

The first order moment of transvariation is the expected income difference between Pj and Ph, such that yji < yhr and µj > µh. That is, it is a weighted average of the income differences for all pairs of economic units, one taken from the h-th subpopulation such as yji < yhr. It is given by:

∞ y

(2.30) p jh = — dFh (y)— (y − x)dFj (x) ∀µ j > µh 0 0

According to Dagum (1980), (2.29) and (2.30) imply that:

(d jh − p jh ) (2.31) D jh = d jh + p jh

Equation (2.31) is the ratio between the net economic affluence and its maximum possible value. It represents the relative economic affluence, which is a normalized index that indicates the “distance” between Pj and Ph.

The product GjhDjh measures the net contribution to the total inequality of between-group inequality. It basically represents the inequalities derived from the non- overlap of the distributions j and h. Hence, Gjh(1-Djh) is the transvariation between Pj and Ph, which is the part of the inequality due to the overlap of the distributions of j and h (Mussard, 2003).

58 According to Mussard (2003), we can define the first component of the Gini decomposition as the net contribution of the between-group inequalities to the overall

Gini measured on P, which is given by Gb below:

k j−1

(2.32) Gb = ƒƒG jh D jh ( p j sh + ph s j ) j=2 h=1

Where pj is the proportion of people of Pj (pj = nj/n), and sj is the income share of the subpopulation j (sj = njµj/nµ).

The second component is the contribution of the transvariation between the subpopulations to G, given by Gt:

k j−1

(2.33) Gt = ƒƒG jh (1− D jh )( p j sh + ph s j ) j=2 h=1

The last component is the contribution of the within-group inequalities to G as given by:

(2.34) Gw = ƒG jj p j s j j=1

Therefore, the fundamental equation of the Gini decomposition is given by (2.32),

39 (2.33) and (2.34), such that G = Gb + Gt + Gw.

According to Dagum (1997b), Theil introduced in 1967 a new inequality measure coming from the second law of thermodynamics40, called “the entropy law”, which measures the contribution of the between and the within subpopulations inequalities to

39 If the k distributions are equally distributed with identical means, then Gjj = Gjh = 0, which implies that G = Gb = Gt = Gw = 0.

40 Originally “the entropy law” measures the disorder of a thermodynamics system.

59 the total inequality. The Theil, Hirschman-Herfindahl, and Bourguignon indexes to be used in the regional impacts of the import tariff reductions are three particular cases of the generalized entropy ratio given by:

β k n j y »≈ y ’ ÿ 1 ji …∆ ji ÷ Ÿ (2.35) I β = ƒƒ ∆ ÷ −1 β (1+ β )n j=1 i=1 µ …« µ ◊ ⁄Ÿ

Where β is a parameter representing a real number.

According to Mussard (2003), the Theil index T is the generalized entropy ratio when β tends towards zero. Therefore, the between-group contribution Tb and within- group contribution Tw are given by:

k µ j n j µ j (2.36) Tb = lim I β = ƒ log β →0 j=1 µ n µ

k n j µ j n j 1 y ji y ji (2.37) Tw = lim I β = ƒ ƒ log β →0 j=1 µ n n j i=1 µ j µ j

Where T = Tb + Tw.

The Hirschman-Herfindahl (H-H) index is the special case of Iβ when β tends towards one. The between-group contribution I1b and within-group contribution I1w are given by:

k µ n ≈ µ ’ Varµ 1 j j ∆ j ÷ j (2.38) I1b = lim I βb = ƒ −1 = β →1 ∆ ÷ 2 2 j=1 µ n « µ ◊ 2µ

k 2 k 2 1 µ j n j Var y j 1 n j µ j 2 (2.39) I1w = lim I βw = ƒ = ƒ CV (y j ) β →1 2 2 2 2 j=1 µ n µ j 2 j=1 nµ

Where I1 = I1b + I1w, Var is the variance, and CV is the coefficient of variation.

60 Dagum (1997b) demonstrates that the Bourguignon index (B) is the limit of the entropy index when β tends towards -1. The Bourguignon decomposition results in the within-group and between-group contributions, in the same way as in the Theil and H-H indexes. These contributions are given, respectively, by Bb and Bw as it follows:

k n j µ (2.40) Bb = lim I βb = ƒ log = log µ − log M gµj β →−1 j=1 n µ j

k n j (2.41) Bw = lim I βw = ƒ (log µ j − log M gj ) β →−1 j=1 n

Where Mgj and Mgµj are the geometric means measured, respectively, on Pj and on the vector µj. The equations (2.40) and (2.41) imply that B = Bb + Bw.

For individual household groups we will evaluate the gains and losses through

standard welfare measures such as the equivalent variation (EV), which is the Hicksian

exact measure of the change in consumer surplus, given by (De Melo and Tarr, 1992):

(2.42) EV = e[p0, v(p1,y1)] - e[p0, v(p0y0)]

The first term in (2.42) is the minimum income necessary to reach utility level

v(p1,y1), given prices at p0. The equivalent variation is illustrated in Figure 2.5.

Good Y

v(p0,y0) y1

EV y0

•

• • v(p1,y1)

p1 p0

Good X

Figure 2.5: Hicksian equivalent variation (EV).

2.7. Trade Policy Simulations

Model implementation follows two stages. In the first, the model is solved for the base without imposing any changes in parameters or exogenous variables. The base values are compared with the results of the simulations that are implemented in the second stage. In the second stage, a set of exogenous variables or parameters is modified

to illustrate a change in the trade policy or an exogenous shock on tradable goods prices.

The solutions of the modified model (simulations) and of the base model (benchmark) are

compared. Therefore, through the CGE model it is possible to account for short to medium run effects that the import tariff reductions will have on the welfare of

62 households (gains and losses), including complementary policies to trade reform in order to generate the greatest aggregate welfare gains that do not bring losses for the poor

households.

As described in the objectives section (section 2.3), this study will have three

stages (scenarios). Table 2.5 summarizes all three sets of simulations to be performed in

this study. Each stage will have a set of simulations to be performed in order to reach the

objectives of the study.

Scenarios Simulations

Scenario 1 50 % reduction in import tariffs in all sectors

100 % reduction in import tariffs in all sectors

Scenario 2 50 % reduction in import tariffs on selected sectors

100 % reduction in import tariffs on selected sectors

Scenario 3 50 % reduction in import tariffs on selected sectors plus 20 % increase in direct tax rates 100 % reduction in import tariffs on selected sectors plus 20 % increase in direct tax rates

Table 2.5: Description of the main sets of simulation for the Brazilian trade reform

63 Scenario 1: set of two simulations consisting of reductions in import tariffs for all sectors by 50 % and 100 %.

In general the average nominal import tariff in Brazil is around 13 %, as noted by

Estevadeordal et al. (2000), Leipziger et al. (1997), and Monteagudo and Watanuki

(2002). Table 2.6 shows average nominal import tariffs in the Brazilian economy for

different sectors and goods. Some sectors present, on average, low levels of protection,

but there are some specific products with very high import tariffs. For instance, the

industry average import tariff is around 10.6 %, but the import tariff for vehicles is 39 %,

and for clothing and shoes is 18.3 %.

The idea here is to find the regional short to medium run effects that the import

tariff reductions will have on the welfare of households (gains and losses), and to

evaluate the sectoral trade policy to identify which specific sectors affect the poor more.

A way to do such identification is to verify which sectors bring negative impacts to the

poor households after the import tariffs are reduced or eliminated.

Scenario 2: set of two simulations consisting of a reduction in import tariffs for specific sectors by 50 % and 100 %.

The rationale for this second set of simulations is to verify what would be the welfare improvements for households after having identified and excluded from the trade policy reform those sectors that bring negative outcomes for the poor.

With these two scenarios, we compare the impact of general trade reform

(reduction or elimination of import tariffs) to a reform that is limited in selected sectors.

Nominal Nominal Sectors import tariff Sectors import tariff (%) (%) Agriculture 2.4 Industry 10.6

Rice 2.7 Steel 8.5

Wheat 2.6 Machinery and equipment 7.8

Cotton 3.0 Tractors and equipment 11.2

Food processing 6.6 Electric material 9.7

Coffee products 10.8 Electronic material 6.9

Other vegetable 5.2 Vehicles (cars, trucks, and 39.0

products buses)

Milk 14.5 Other vehicles and parts 7.5

Dairy products 7.8 Rubber products 7.8

Other food products 6.7 Fertilizers 1.1

Mining and oil 7.4 Chemical products 9.0

Oil 11.3 Pharmaceutical products 5.2

Gasoline 11.1 Plastic products 12.0

Other oil products 8.6 Clothing and shoes 18.3

Source: Matriz de Insumo-Produto 1995 (Brazilian Input-Output Tables, 1995) (IBGE, 1997a). Author’s calculation (import duty/total imports ratio).

Figure 2.6: Average nominal import tariffs by sectors and goods in Brazil, 1995

65 At this point, it might be the case that even a sector-specific trade reform is not enough to guarantee equal and efficient welfare gains. According to Harrison et al.

(2003), there can be many ways to include complementary policies to trade reform in order to generate the greatest aggregate welfare gains and that do not bring losses for the poor households. The one to be analyzed will be the import tariff reduction together with a domestic tax reform41, which will be addressed in the next scenario.

Scenario 3: set of two simulations consisting of a reduction in import tariffs for specific sectors by 50 % and 100 %, and 20 % increase in direct (income) tax rates.

The direct use of sidepayments to compensate those households that lose through transfers from those that gain from the import tariff reduction is just the “compensation principle” in welfare economics. It may not be feasible in practice. Instead, we can use the direct taxation system to capture part of the earnings of the high-income households to be indirectly distributed to those poor households, at the same time that it would compensate for government revenue losses. In Brazil, the increase in direct tax rates would affect enterprises, medium-income households, and high-income households, since

41 Another possibility would be an increase in wages as a way to compensate households from losses due to the reduction in import tariff. The logic behind that is that the Brazilian government regulates the increments in the minimum wages paid most for the poor workers, whose labor contracts are generally indexed to the law-determined minimum wage. There would be many problems with the use of this policy. The first is its political appeal, since it is a very common practice in election times to increase the minimum wage. Second, the policy may not achieve a significant proportion of the population because of the size and composition of the formal and informal labor markets. The SAM used here does not have a specified informal labor market account. Third, depending on the labor/capital ratio and elasticity values used in the sectors of our CGE model, an increase in minimum wage does not guarantee an improvement in welfare for households, since the counterpart reaction of firms would be the reduction of production due to the increase in its costs (labor cost). Fourth, to perform this simulation, one of our closure rules should change, and this should be the factor market closure, implying that labor market, at least, should be considered as having fixed economy-wide wage and some unemployment. This could be an issue when comparing the results with other simulations under different labor market closure rules.

66 the poor do not pay direct taxes. Therefore, a combination of trade and tax reform might

be proposed through the third scenario, in order to improve welfare for all poor

households in rural and urban areas.

The direct tax system in Brazil has changed over the years. It is still a progressive

system, but with only three tax rate categories. Before 1989, however, there were more

than nine different tax rates compatible with the income level. After 1988’s Constitution,

there were many changes in the tax rates applied to the population. In 1996-1997, which

is the period our SAM was constructed, the direct tax rates were: 0% (for low income),

15 % (for medium income), and 25 % (for high income). From 1998 to now, people with

annual income less than R$ 10,800 do not pay income tax. Those with annual income

between R$ 10,800 and R$ 21,600 pay income tax at the rate of 15 %. People with annual

income larger than R$ 21,600 pay 27.5% as income tax. The rationale here is to increase

the tax rate for high-income people, since the tax rate of 25 % is very small in

comparison to the rate in place during the 70s and middle 80s42, which can help in the reduction of income concentration and inequality. Table 2.7 shows the maximum tax

rates in Brazil and in selected developed countries from 1986 to 1997. Many of the

developed countries reduced their ceiling tax rates over time, but their rates are still

higher than in Brazil in 1997.

The tax that the government uses to raise revenue affects the outcome, since the

direct tax chosen (due to operational features of the model) does not impose the least

42 One way to justify an increase in the high-income taxation would be to compare the tax rate applied to a person with annual income of R$ 24,000, who would pay the same tax rate as one that earns R$ 240,000 per year. Although there was a more complex system with more income categories with different and larger tax rates, before 1988, the system at that time was fairer than the one seen nowadays that allows this type of distortion.

67 marginal excess burden among the tax instruments available. There might be a risk in this

complementary policy that the loss due to the increase in domestic taxes can be larger

than the gains from the import tariffs reduction, but it needs to be empirically

investigated.

Country 1986 1990 1995 1996 1997 Australia 57 47 47 47 47 Austria 62 50 50 50 50 Germany 56 53 53 53 53 Belgium 72 55 55 56.6 56.6 Brazil 50 25 35 25 27.5 Canada 34 29 31.3 31.3 31.3 Denmark 45 40 34.5 34.5 34.5 England 60 40 40 40 40 Finland 51 43 39 39 38 France 65 57 56.8 56.8 54 Italy 62 50 51 51 51 Japan 70 50 50 50 50 Portugal 61 40 40 40 40 Spain 66 56 56 56 47.6 USA 50 28 39.6 39.6 39.6 Source: OECD on the internet - http://www.tax.org./ritp.nsf/2e553e534ac6bd2e8

Table 2.7: Maximum income tax rates for selected countries, in percent

68 As in many developing countries, there is a great proportion of the working population in the “informal” labor market in Brazil. The impact of trade reform can have

a multiplicative impact on reducing poverty when combined with policies that improve

the labor markets’ flexibility, which could help the poor to move into the formal sector

(Harrison et al., 2003). But it is difficult in terms of data sources and quality to make the distinction between the formal and informal labor markets in Brazil. Actually the proportion of workers that receive up to one minimum wage43 per month in Brazil was, on average, 52 % in 1995, and 43 % in 199944 (IBGE, 2002).

According to IBGE (1997c) Brazil has, on average, 60 % of the working population as unskilled workers, and the share of unskilled workers among the low-

income people is around 78 %. Therefore, it is expected that with the import tariffs

reduction, the unskilled labor and unskilled-endowed households will gain from such

reform. Following the Heckscher-Ohlin-Samuelson model (HOS), since Brazil protects

the capital-intensive sectors, after the import tariffs reduction these sectors should lose

and labor-intensive sectors should gain. Since almost 20 % of the low-income workers

are employed in agriculture, following HOS would lead to an increase in exports, and

trade reform should bring gains for unskilled workers in rural areas.

In the third scenario we try to combine policies such that no poor household is

harmed from a reduction in import tariffs, trying to identify the equity-efficiency

tradeoffs available in Brazil, and to indicate the most attractive alternative.

43 See footnotes 10 and 41.

44 In the state of Maranhao (Northeast) this proportion was approximately 75 % in 1999.

69 To close this section, in determining the effects of reduction in import tariffs on

poor households it is important to have a clear picture of the transmission mechanism,

and the behavior of the economic agents involved.

Figure 2.6 shows exactly the complexity of effects of a trade shock such as import

tariff reduction of a good. It illustrates the transmission of price shocks from world prices

to final consumers.

Quantities Exchange Tariffs, Tariff Revenue rate QR’s

Pass through, competition Border Price

Wholesale Price Taxes Spending Enterprises

Distribution, taxes, regulation Retail Price Factor Markets Cooperatives, technology, random shocks

male elderly Welfare

female young

Source: Winters (2002).

Figure 2.6: Transmission of trade shocks in the domestic market of a good

70 Winters (2002) exemplifies this transmission for an import good, where the

foreign price of the good, combined with the exchange rate and import tariff, define the

post-tariff border price. Once inside the country, the good price is affected by domestic

taxes, transportation costs, and even a compulsory procurement by the authorities.

Therefore, the resulting price is the wholesale price. From the distribution centre, the

good can be affected by additional taxes and costs, resulting in the retail prices of such

good to already be distributed to households. At each stage the institutions incur costs

and add mark-ups, defining the final price.

2.8. Results and Discussion45

2.8.1. Regional Disaggregated SAM

The original SAM was disaggregated into four regions as specified in section five.

From now on, many of the components of the regional disaggregated SAM will be

referred to by abbreviations whose meanings are described through tables in Appendix D.

Table D.1 describes all activities in the SAM to be used in the policy experiments. There

are 15 main activities divided into four regional groups, totaling 60 activities. Tables D.2

and D.3 show the main types of labor employed in the activities according to the four

regions, respectively in urban and rural areas. There are 40 urban categories of labor, and

eight categories of labor in rural areas. The main capital categories, a total of 32 for both

urban and rural areas, used in each region and activities are in table D.4. Table D.5 shows

all 12 main types of land in all regions.

45 See Appendix E for additional tables and detailed discussion. 71 The regional disaggregated SAM was balanced46 according to a cross entropy

(CE) procedure, which can be constrained by some known flows from the original SAM.

Therefore, according to the original SAM, some known totals were considered fixed in

this procedure, such as total activities expenditures, total demand, total capital income,

total government income, and total household income. The result is a balanced SAM that

provides an important and consistent set of relationships showing intermediate, final

demand, value added, and foreign transactions.

Table 2.8 shows a summary of the Brazilian economy for the period 1995/1996,

reflecting an important feature about the way that a SAM is built, which includes not only

information from input-output tables, but also macroeconomic data from national

accounts. According to the disaggregated SAM, the private consumption was responsible

for the larger part of the Brazilian GDP, followed by investments and government

consumption. Total trade represented 15.5% of GDP in 1995, when the balance of trade

was negative due to the overvalued exchange rate at that time. The net indirect tax

revenue represented approximately 15% of GDP. Although these numbers, which are not

different from those in IBGE (1997c), show that Brazil is not so dependent on the

external sector, trade can be a key for a better distribution of income within the country.

The Gini coefficient and the Theil index calculated for all five households were,

respectively, 0.505 and 0.634, for the period 1995/1996. For both coefficients, the higher

the value, the more unequal is the distribution of income.

46 The term “balanced” is very often employed in the CGE literature, and it means that total rows equal total columns.

72 The policy simulations were designed to verify whether trade liberalization alone can guarantee not only that the distribution of the income improves, but also that the poor

households will not lose from such policy. Before we start the next section with a

discussion of the simulation results, we need to explore some important information

about the main agents and their inter-relationships in the Brazilian economy that the

regional disaggregated SAM offers.

Macro-aggregates Value (1995 billions of R$) % of GDP

Private consumption 430 65.5

Fixed investment 126 19.1

Government consumption 110 16.7

Exports 46 7.1

Imports -55 -8.4

Absorption 657 100.0

Net indirect tax revenue 96 14.6

Source: author’s calculations from disaggregated regional SAM. Sum may not equal 100 due to rounding.

Table 2.8: Aggregated national accounts

The participation of the main commodities in value added and production is shown in Table 2.9, and we can note that the agricultural products represent less than

10% of the total value added in Brazil. Actually, services and industrial products have 73 been important in the Brazilian economy as their participation on total value added and

production is around 70%. Services are notably essential for employment, since they are

responsible for more than 56% of the employment in the country. Mining and oil,

processed food, and agricultural commodities have important shares on Brazil’s exports,

but the industrial products are the main goods imported as they represent 70% of imports.

Table 2.9 also confirms the data from IBGE (1997c), from which it is possible to verify

the large participation of services in GDP.

Value added Production Share in Exports Imports Commodities share share total share share employment (%) (%) (%) (%) (%)

9.8 7.9 7.4 18.4 5.6 Agricultural 23.4 34.6 19.6 9.8 69.7 Industrial 5.8 6.6 2.7 - - Construction 4.6 7.5 2.2 23.8 3.0 Processed food 2.1 3.1 1.8 35.4 7.7 Mining and oil 8.3 7.6 9.6 12.6 3.9 Transport and trade 46.0 32.7 56.7 - 10.1 Service Source: author’s calculations from the disaggregated regional SAM. Sum may not equal 100 due to rounding.

Table 2.9: Participation of commodities in value added, production, employment, exports, and imports shares

74 According to Table E.1 (Appendix E), there is an expected relative small importance of skilled labor in activities related to agricultural products, represented by the first eight rows. Land and capital are important market factors in these activities. The information compiled from the disaggregated SAM is also consistent with respect to the larger use of capital in large activities. This table also shows that the overall shares of factors of production employed in the Brazilian economy are: 30.6 % of skilled labor, 15

% of unskilled labor, 51 % of capital, and 3.4 % of land.

The existing regional differences in Brazil can be illustrated through Table 2.10, in which the total factors employed and also the factor endowments for each household are allocated in each of the four regions. The South and Southeast regions employ more than 74 % and 67 % of Brazil’s skilled and unskilled labor, respectively, and 75 % and 61

% of the country’s capital and land, respectively. Although the North is the largest region in size and the South/Southeast region is the smallest, the former employs only 8.7 % of the country’s land in economic activity. The uneven distribution of wealth is based on the high concentration of resources in the South/Southeast region, illustrated by Table 2.10.

According to Table E.2 (Appendix E), more than 60 % of the households in both rural and urban areas are in the South/Southeast region.

Table E.3 (Appendix E) shows the households’ budget shares spent on the main commodities specified in the disaggregated SAM. Low-income-urban households spend more of their income in services (57 %) and processed food (32 %). Low-income-rural households have larger shares of industrial goods (16 %) and agricultural goods (9 %), and smaller on services (46 %), in comparison to the low-income-urban households.

Regions Skilled labor Unskilled labor Capital Land (%) (%) (%) (%) North 4.2 6.4 4.2 8.7 Northeast 14.3 18.0 13.8 14.5 Center-West 7.4 8.4 7.0 15.8 South and SE 74.1 67.1 75.0 61.0 Source: author’s calculations from the disaggregated regional SAM. Sum may not equal 100 due to rounding.

Table 2.10: Proportion of Brazil’s total factors employed in each region

2.8.2. Overall Trade Liberalization (Scenario 1)

Macro impacts

The simulation results of eliminating tariffs on imported commodities for all sectors are shown in Table 2.11. The overall reduction in import tariffs causes a reduction

in the price of imported commodities relative to domestic goods, and a shift towards

imported goods and away from domestic production. Imports increase 5.8 % and 12.4 %,

respectively, in the cases of 50 % and 100 % reduction in the import tariffs. In order to

keep the fixed trade balance, exports must rise 6.7 % and 14.4 %, which is only achieved

after a depreciation of 2.1 % and 4.4 % of the real exchange rate.

After total elimination of the overall import tariffs, there was a significant

decrease in import prices of some commodities, such as corn, rice, perennial goods,

horticultural goods, forest products, and industrial goods. Consequently, there was also a

large increase in total imports of these commodities. 76

50 % reduction import tariff 100 % reduction import tariff Absorption 0.1 0.1 Private consumption 0.1 0.1 Exports 6.7 14.4 Imports 5.8 12.4 Real exchange rate 2.1 4.4

Share of GDP (%)

Investment -0.1 -0.2 Private savings 0.2 0.5 Foreign savings 0.1 0.1 Government savings -0.4 -0.9 Tariff revenue -0.4 -0.9 Direct tax revenue 0.0 0.1

Equivalent Variation (%)

Rural low inc. household 0.4 0.7 Rural medium income 0.3 0.7 household Urban low income -0.3 -0.7 household Urban medium income 0.0 0.0 household High income household 0.2 0.3 Total welfare 0.1 0.1 Gini coefficient -0.1 -0.2 Theil index -0.1 -0.3

Table 2.11: Simulations results for overall import tariffs reduction (scenario 1), % change from benchmark values

77 Lower prices of imported commodities reduce the cost of intermediate goods for domestic producers, which together with increased export demand, induces an increase in production. To illustrate this, horticultural, forest, and industrial commodities have large increases in exports after eliminating import tariffs. Lower import tariffs also reduce consumer prices, increasing real income and real absorption (0.1 %). Reduction in import tariffs causes a decrease in government revenue, implying in a reduction in government savings (-0.4 % and -0.9 %).

The overall welfare impacts from the import tariffs reductions were positive for both simulations (Table 2.11). The welfare increased for all households, but not for the low income urban households (hurblow). However, the poorest households, rural low and middle-income households (hrurlow and hrurmed), had their welfare improved after the trade reform. Not surprisingly, the Gini coefficient and the Theil index were reduced with the removal of the import tariffs. The Gini coefficient was reduced from 0.5054 (base) to

0.505 (partial removal of the import tariffs), and to 0.5045 (total removal of the import tariffs). The Theil index in the base was 0.6344 and, with the partial and complete elimination of the import tariffs, respectively, reduced to 0.6327 and 0.6336. These results emphasize that a concern about equity is not equivalent to a concern about poverty, since the trade simulation evaluated in this section resulted in greater equity, but with an increase in poverty for urban poor.

The expected results from the first scenario, simulations with 50 % and total elimination of the import tariffs, would be that trade liberalization would bring gains for all poor households, since there would be a shift of resources from capital intensive manufacturing toward unskilled labor intensive agriculture and less capital intensive

78 manufacturing, increasing the wage of unskilled labor to capital and skilled labor.

Simulation results show that the poorest households, rural low and rural medium, gain from trade reform, but poor households in urban areas do not.

The price changes due to trade liberalization affect the incentives to produce particular goods and the technologies they employ. The Stolper-Samuelson Theorem

(SST) predicts that, under particular conditions, an increase in the price of the commodity that is unskilled labor intensive in production will increase the unskilled real wage and decrease that of skilled labor. The results for the rural households confirm exactly the

SST. But what can be said about the results for urban poor households?

According to Harrison et al. (2003), due to the second best effects it might be the case that there is no trade reform that can improve welfare for the whole society without having a compensatory mechanism, which can imply that low income households in

urban areas may experience welfare gains only if an import tariff reduction is combined

with some alternative policy that compensates their losses from trade reform.

According to Winters (2002), despite its theoretical elegance, the SST is not

robust enough to totally explain the link between trade and poverty in the real world. One

of the problems cited that is relevant for our analysis is the dimensionality problem, since

the results are not so predictable when there are many sectors, commodities, and also

factors of production that are immobile. Another complication is that the prices of non-

traded goods are determined in order to clear the domestic market. In our case, trade

79 shocks induce changes in the real exchange rate47, and in case these goods have different factor intensities, there is an introduction of an extra source of factor market effects (Lal,

1986).

Brazil seems to be unskilled labor-abundant, so a reduction in the import tariffs

should improve workers’ welfare. However, within Brazil it is not clear that the least-

skilled workers, and thus the most likely to be poor, are the most intensively used factor

in the production of tradable goods, mainly in urban areas. According to Winters (2002),

the agricultural sector should be the one to certainly gain from free trade because this

sector has a higher proportion of unskilled workers. Therefore, results for rural

households, in Table 2.11, are coherent to the SST, and Table E.1 (Appendix E) confirms

that the agricultural activities allocate a large fraction of factors of production as

unskilled labor.

The urban poor households are harmed after the removal of the import tariffs, and

some possible explanations for this result were described in the previous paragraphs. In

section 2.6.5 we discussed some studies, such as Robbins (1994, 1995), Beyer, Rojas and

Vergara (1999), Robbins and Gindling (1999), and Arbache (2001), which claim that

trade liberalization can increase wage inequality, perhaps as a consequence of higher

technological modernization, increasing the demand for skilled labor. Other studies also

go against the predicted results given by the traditional theory of trade, such as Arbache

and Corseuil (2000), Barros et al. (2001), Menezes-Filho and Rodrigues (2001), and

Maia (2001), and their conclusions consist of an uncertain impact of trade openness on

labor market in Brazil.

47 The real exchange rate in our model is represented by the relative prices of traded and non-traded goods.

80 Regional impacts

The results obtained for the whole country were a consequence of the regional reallocation of resources after the counterfactual reduction in import tariffs. According to

Table E.1 (Appendix E), it is expected that the capital-intensive activities have their

domestic prices and sales reduced, after the import tariff reduction, which can bring

negative outcomes for urban households, mainly the poor ones, where labor income is

related to such activities.

The North, Northeast, and Center-West are very poor regions in Brazil, and the

effects of the import tariffs reduction are important to have a better picture of what would

happen if we consider an elimination of such import tariffs in these regions. The main

consequence of eliminating the import tariffs for most of these regions was the decrease

in production of agricultural annual commodities in large farms, industrial goods, and

goods in the construction sector. These activities have a large proportion of capital

employed in their production process, as demonstrated in Table E.1 (Appendix E).

The prices of capital and land for large farm annuals decrease substantially

(Tables E.5, E.8, E.14, Appendix E). Labor and capital prices and income are reduced in

the industry and construction sectors, with larger negative impact on low income urban

households (Hurblow), who are more dependent on capital-intensive goods (Tables E.5,

E.6, E.8, E.9, E.11,E.12, E14, E15, Appendix E).

The results seem to suggest that the main changing component for the factor

prices was the effect on capital prices, reducing the final price and production for large

farm annuals, industry, and construction.

81 The effect of trade liberalization on agriculture brings welfare gains for all rural households, with a higher increase in wages for skilled workers (Table E.6, E.9, E.12,

E.15, Appendix E). It confirms the findings of some Brazilian studies discussed before, in which the possibility to import a capital good at a lower price can increase production together with a larger demand for skilled labor to gain advantage of the new technologies to be implemented. For instance, the Brazilian industries for tractors and fertilizers are very concentrated, and they act as an oligopoly, charging high prices from farmers. The reduction of import tariffs on these goods can motivate small farmers to buy a tractor and use fertilizers at lower prices than before, improving the potential production. The increase in the realized production, however, would be possible only upon the use of qualified labor in order to extract the maximum yield of such technology, in this case tractor and fertilizers. Therefore, the demand for skilled labor should increase in this example.

The South/Southeast is the most developed and wealthy region in Brazil. Most of the industry and agriculture is located in this region, which make it responsible for more than 90 % of the GDP produced. This region has the largest proportion of households, factor endowment, skilled labor and capital shares than any other region (Table 2.10, and

Tables E.1, and E.2, Appendix E). The share of the industry in production and employment is significant, and the participation of all factors of production and households’ categories are very large in this region. Although unskilled labor wages have a larger increase than the skilled labor ones, it is not enough to offset the losses from the industry, which is the main income supplier for urban low-income households.

82 The simulation showed that all four regions experienced similar impacts from 100

% reduction in the import tariffs, with some regional specificity, generating small

differences in activity prices and production, and resources allocation.

A comparison about the main consequences for labor income distribution among

households in all four regions can be seen in Table 2.12. The larger labor income48 gains are obtained in the North and Center-West, mainly for rural households. In terms of poverty, the results seem to indicate that labor income does not contribute to more urban

poor, but capital income is the main factor that contributes to reduce welfare for urban

low-income households, as was seen at aggregate level.

Even though the inequality in the distribution of income is high in the

South/Southeast and Center-West, trade liberalization in Brazil brings a better regional

distribution of income within regions, as seen through reduction in all four indexes after

simulation (Table 2.13).

Although Table 2.13 shows that the income inequality is slightly reduced after

eliminating import tariffs, the question becomes what are the main changes between

regions? Table 2.14 points out some elements to answer this question. In this table we

have the multi-decomposition of the four inequality measures (indexes) used so far. The

overall results from simulations do not change the structure of how the labor income is

distributed within and between regions. The largest part of the overall inequality seems to

48 It was also considered land income, with larger gains for households in the South/Southeast. Income inequality measures do not alter the direction of the changes considering land and labor incomes together. However, the consideration of capital income is not possible at the regional level, since capital income payments are made from sectors to enterprises, and therefore to households, but without information about the regional origin of such income.

83 come from the inequality in labor income among the four Brazilian regions49. According to the Gini index, 78.6 % of the total labor income inequality is due to the inequality among regions. Only the Gini coefficient can provide the intensity of transvariation (4.8

%), which represents the part of the between-regions disparities issued from the overlap

among the distributions50. Therefore, the simulation does not modify the structure of the inequality within and among regions in Brazil, and the inequality among regions is more important than within regions.

Rural low Rural Urban low Urban High Regions income medium income medium income household income household income household household household North 3.0 3.0 1.3 1.2 1.2

Northeast 1.7 1.8 1.1 0.9 1.1

Center-West 2.9 3.0 1.0 1.0 1.1

South/Southeast 1.9 1.9 0.7 0.7 0.6

Table 2.12: Regional impacts from an overall elimination of the import tariffs in household’s labor income (% change from benchmark values)

49 H-H index was the only index to indicate that the within-region inequality is the most important component to explain the overall inequality.

50 The low value for transvariation was not surprising due to the SAM disaggregation, since the labor income comes from activities specified by region, with no overlap from sources of income.

Indexes North Northeast Center-West South/Southeast Base(*) Sim(**) Base Sim Base Sim Base Sim Gini 0.258 0.255 0.353 0.352 0.402 0.400 0.475 0.474 Theil 0.115 0.113 0.229 0.227 0.275 0.272 0.390 0.388 H-H 0.106 0.104 0.201 0.200 0.275 0.273 0.388 0.386 Bourguignon 0.139 0.136 0.310 0.308 0.342 0.337 0.526 0.522 (*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.13: Regional income inequality measures before and after an overall elimination of the import tariffs

The results from the multi-decomposition of the four inequality indexes also show

that regions North, Northeast, and Center-West contribute to reducing the overall

inequality among regions. Region South/Southeast has the most important contribution

not only to increase the overall inequality among regions, but also within this region.

According to Table 2.15, we can see the relative importance of all four regions for the

inequality within a region. The main contribution for this type of inequality seems to

come from the South/Southeast. For instance, according to the Gini index, around 13 %

of the overall inequality originates from the inequality within region South/Southeast.

% of the within-region % of the between- % of transvariation Indexes component regions component

Base(*) Sim(**) Base Sim Base Sim Gini 16.6 16.6 78.6 78.6 4.8 4.8

Theil 40.2 40.2 59.8 59.8 - -

H-H 58.2 58.1 41.8 41.9 - -

Bourguignon 37.5 37.4 62.5 62.6 - -

(*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.14: Contribution of the four decompositions to overall labor income inequality before and after simulation

North Northeast Center-West South/Southeast Indexes Base(*) Sim(**) Base Sim Base Sim Base Sim Gini (%) 0.5 0.5 2.0 2.1 1.2 1.2 12.9 12.8 Theil (%) 0.7 0.6 4.2 4.2 2.5 2.6 32.8 32.8 H-H (%) 0.07 0.07 1.4 1.4 0.5 0.5 56.2 56.2 Bourguignon (%) 3.9 3.9 8.8 8.8 9.7 9.7 15.0 15.0 (*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.15: Regional contribution to overall labor income inequality before and after simulation

2.8.3. Sectoral Trade Liberalization (Scenario 2)

The set of simulations to be performed in scenario 2 consists of 50 % and 100 % reduction in import tariff for some specific sectors. The sectors considered are divided in five groups: (i) agriculture (AGR), which is composed of corn, rice, soybeans, beans,

perennial commodities, annual commodities, horticultural products, forest products, cattle

meat, poultry meat, milk, sugar, and other agricultural commodities; (ii) annual (ANN),

which is composed of corn, rice, soybeans, beans, annual commodities, horticultural

products, and other agricultural commodities; (iii) perennial (PER), which is represented

by coffee, cocoa, manioc, perennial commodities, and forest products; (iv) industrial

(IND), which is composed of industrial commodities, mining and oil goods, and

processed foods; and (v) the last group is given by a combination of industry and

agriculture (MIX), which Brazil is more likely to import such as corn, rice, perennial

commodities, annual commodities, forest products, milk, cattle meat, other agricultural

commodities, processed foods, mining and oil goods, and industrial products.

In this section, our main goal is to verify the possibility of finding a sectoral

reduction in import tariffs that does not harm poor households. As seen in overall trade

liberalization, poor urban households are likely to experience welfare losses after

reduction in the import tariffs. If there is no sectoral trade liberalization that can bring

gains for all households’ categories, then it may be instructive to find an efficiency-equity

combination of policies not only to reduce the protection of domestic sectors in Brazil,

but also to bring welfare improvements for all households.

87 The main findings from a reduction in the import tariffs for some specific sectors, respectively, 50 % and 100 % differ only in magnitude, but the direction of change is the

same for both simulations (Tables E.16, Appendix E, and 2.16).

The sectoral trade liberalization in the agricultural sector51 does not bring considerable modifications in the economy in the short to medium run. The impacts on trade are small, without any substantial change in the inequality measures. However, the poorest people lose, which is not surprising, as we can see by the decrease in welfare for

rural households. In this case, resources from agriculture would be reallocated in the most

capital-intensive sectors, and it would even bring gains for urban households when the

import tariffs are totally eliminated, as in Table 2.16.

The simulations considering trade liberalization on annual and perennial

agricultural commodities are just special cases of the most general agricultural sector

analysis made in the previous paragraph. Table 2.16 shows that impacts from the import

tariffs reduction in the perennial agricultural commodities are not that expressive.

However, the same impacts from the annual agricultural commodities opening are just

about the same as those from AGR. They show, once again, that poor households in rural

areas are the main losers from trade liberalization. After removing tariff distortion from

the labor intensive sector, with fixed capital supply, capital-intensive sectors become

better off, as predicted by the Stolper-Samuelson theorem, since the capital/labor ratio

decreases making labor less productive.

51 Even though agriculture is composed of many different activities (sectors) in four different regions in the SAM, we are referring to the agricultural sector and agricultural sectors interchangeably.

100 % reduction import tariff AGR ANN PER IND MIX Absorption - - - 0.1 0.1 Private consumption - - - 0.1 0.1 Exports 1.3 0.9 0.4 13.1 14.1 Imports 1.3 0.8 0.5 11.2 12.1 Real exchange rate 0.2 0.2 0.1 4.2 4.3

Share of GDP (%)

Investment - - - -0.2 -0.2 Private savings - - - 0.5 0.5 Foreign savings - - - 0.1 0.1 Government savings - - - -0.8 -0.8 Tariff revenue -0.1 - - -0.9 -0.9 Direct tax revenue - - - 0.1 0.1

Equivalent Variation (%)

Rural low inc. household -0.4 -0.4 -0.02 1.1 1.0 Rural medium income -0.4 -0.3 -0.03 1.0 0.9 household Urban low income 0.2 0.1 0.02 -0.8 -0.7 household Urban medium income 0.1 0.1 0.03 -0.2 -0.1 household High income household - - - 0.3 0.3 Total welfare 0.02 0.01 - 0.1 0.1 Gini coefficient - - - -0.2 -0.2 Theil index - - - -0.4 -0.3

Table 2.16: Simulation results for sectoral elimination of the import tariffs (scenario 2), % change from benchmark values

89 As expected, the industrial sector plays the most important role in the Brazilian attempt to open its economy due to the existence of a high degree of protection in this

sector for many decades. The results from trade liberalization for agriculture stressed the

importance of the industry in the Brazilian liberalization process in such a way, that the

overall import tariffs reduction discussed in the previous section was not that different

from the results obtained from the industry sector only reduction in the import tariffs.

Results show a substantial increase in trade, with a devaluation on the real exchange rate.

The tax revenue is reduced and so is investment. Private savings increase. Although the

level of inequality falls through a reduction in the Gini and Theil indexes, the main

negative impact seems to be once again on the urban poor households through their

welfare reduction. As expected, rural poor households win with the reduction or

elimination of the protection in the capital-intensive sectors. However, this result can be

seen as a potential danger in policy making because it can be an invitation to strategic

lobbying by the industrial sector members.

The elimination of the import tariffs in agriculture does not improve inequality in

the distribution of income in any region (Table 2.17). This is a strong result against

sectoral trade liberalization in Brazil.

North Northeast Center-West South/Southeast Indexes Base(*) Sim(**) Base Sim Base Sim Base Sim Gini 0.258 0.259 0.353 0.354 0.402 0.403 0.475 0.476 Theil 0.115 0.116 0.229 0.231 0.275 0.276 0.390 0.391 H-H 0.106 0.106 0.201 0.203 0.275 0.276 0.388 0.389 Bourguignon 0.139 0.140 0.310 0.315 0.342 0.344 0.526 0.528 (*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.17: Regional income inequality measures before and after elimination of the import tariffs in agriculture

Elimination of an import tariff in the industry harms urban low and medium income households instead of rural households as seen in the case of AGR. According to

Table E.20 (Appendix E), rural households are those that gain from trade reform in the

industry sector, allowing substantial increase in their wages. Although urban households

lose with sectoral trade liberalization in the industry, the distribution of income within

regions improves (Table 2.18).

Sectoral elimination of the import tariffs in agriculture and industry produced opposite welfare outcomes for low and medium income households, in both rural and

urban areas. The elimination of import tariffs as a combination of agricultural and

industrial sectors (MIX) brings welfare losses for urban low and medium income

households (Table 2.16).

North Northeast Center-West South/Southeast Indexes Base(*) Sim(**) Base Sim Base Sim Base Sim Gini 0.258 0.255 0.353 0.350 0.402 0.400 0.475 0.474 Theil 0.115 0.112 0.229 0.225 0.275 0.272 0.390 0.387 H-H 0.106 0.103 0.201 0.198 0.275 0.272 0.388 0.385 Bourguignon 0.139 0.135 0.310 0.304 0.342 0.336 0.526 0.520 (*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.18: Regional income inequality measures before and after elimination of the import tariffs in industry

Even though the welfare implications from this combined sectoral trade reform do

not bring good outcomes for urban households (Table 2.16), the inequality of the regional

distribution of income improves (Table 2.19). However, the values do not differ

significantly from those from Table 2.18, under industrial removal of the import tariffs.

Section 2.8.3 emphasized the main overall and regional consequences of

removing import tariffs in some specific sectors and combination of sectors. The results

suggest that Brazil should find another type of policy to be combined with the import

tariffs reduction in order to achieve welfare improvements for all households in all

regions. This is the main task to be pursued in the next section.

North Northeast Center-West South/Southeast Indexes Base(*) Sim(**) Base Sim Base Sim Base Sim Gini 0.258 0.256 0.353 0.351 0.402 0.400 0.475 0.474 Theil 0.115 0.113 0.229 0.226 0.275 0.272 0.390 0.387 H-H 0.106 0.104 0.201 0.199 0.275 0.272 0.388 0.386 Bourguignon 0.139 0.136 0.310 0.305 0.342 0.336 0.526 0.521 (*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.19: Regional income inequality measures before and after elimination of the import tariffs in a combination of agriculture and industry

2.8.4. Equity-Efficiency Trade Liberalization (Scenario 3)

Since sectoral trade reform does not bring any substantial improvement in

households’ welfare, compared to overall trade reform, we consider the overall reduction

of the import tariffs as the main instrument of trade liberalization in our analysis in this

section. Therefore, overall reduction in the import tariffs is combined with a different

policy in order to improve welfare for all poor households. This policy is the increase in

direct tax rates, which was discussed in section 2.7.

Direct taxes can be an important instrument of redistribution of income in case of

a reduction of the import tariffs. Increases in direct tax rates would promptly affect

medium to high income households and enterprises, without affecting the poor. This

section is twofold. First, the main overall results from the combined policies are

93 investigated, in order to conclude if there exists an equity-efficiency trade liberalization

policy for Brazil. Second, the regional consequences are described in details to close the

results discussion of the regional CGE model.

Scenario 3: Trade and Direct Tax Reform

Even though our main focus is on the link between trade policy changes and

poverty and distribution of income in Brazil, our previous findings showed that overall

and sectoral reduction in the import tariffs do not improve welfare for urban poor

households. But the question becomes whether there is a combination of trade policy and

direct tax policy to achieve more efficiency and equity in Brazil. Therefore, the challenge

becomes to find a “win-win” combination of policy reforms for all poor households. One

word of caution is needed here since we are in a second-best world. Although the

alternative policy to be considered is a simple tax reform52 that will bring more distortion to the economy, it consists of an increase of tax rates for medium to high income

households that will serve as a compensatory scheme to offset poor households’ losses

after reduction in the import tariffs. The use of sidepayments or lump-sum taxes as

options of policies is not considered in our analysis.

The combined reduction of import tariff/increase in direct tax rates improves overall income, welfare, production for some selected sectors, and brings a better distribution of income. The direct tax rates for the base year and for the equity-efficiency

52 According to the discussion in section 2.7, a possible politically appealing alternative could be an increase in the minimum wage that is determined by the Brazilian government. However, as expected and discussed in that section, the results of the simulations accounting for this type of policy bring welfare losses for all households when combined with reduction of the import tariffs. Due to space constraint, the explicit and detailed results were omitted from our results discussion in this section.

94 combined policies can be seen in Figure 2.7. Note that the level of direct tax rate for urban medium-income households is very low, since the household income categories in

the SAM do not coincide to those in the official Brazilian direct tax rate schedule.

Enterprises and high-income households are key agents to serve as instruments of income

re-distribution in the proposed combined trade/tax reform (scenario 3).

e t 12 a r

x ) a t

% 10

( t c e r i 8 d

2 SCENARIO 3 0

ENTERPRISE BASE URBAN MEDIUM INCOME HOUSEHOLDS HIGH INCOME HOUSEHOLDS

BASE SCENARIO 3

Figure 2.7: Direct tax rates at the base year and for the simulation in scenario 3 (in %)

95 The main result from these combined policies is that the trade balance improves, at the price of real exchange rate devaluation (Table 2.20). Investment and private

savings fall, but the government savings increase in order to balance the government

account. Direct tax revenues increase 2.6 % in both simulations, as a result of the 20 %

increase in the direct tax rates. The overall and individual household’s welfare improve

after both simulations, except for high income households, who will pay more taxes after

the implementation of the combined policies. The distribution of income also improves

substantially with the simulations. To be more specific, the values for the Gini and Theil

indexes for the base (0.5054 and 0.6344, respectively) become 0.5048 and 0.6333, in a

partial elimination of the import tariffs. The total removal of the import tariffs reduces

these indexes to 0.5043 and 0.6324.

Figure 2.8 summarizes all sets of simulations performed by the three scenarios.

Scenario 1, given by the overall reduction in the import tariffs, hurts urban poor

households. The figure also shows that the sectoral import tariffs reduction (scenario 2)

does bring negative outcomes for poor urban households (low and medium income) for

the import tariffs reduction in industry (IND) and in the combination of industry and

agriculture (MIX), and also for rural households (low and medium income) under the

sectoral reduction of the import tariffs in the agricultural sector (AGR). Finally, it is

possible to see the effects of the combined trade and tax reforms (scenario 3), under

which we could verify that the high-income households are the only ones to lose from

such policy.

50 % reduction import 100 % reduction import tariff tariff + 20 % increase direct + 20 % increase direct tax tax Absorption - 0.1 Private consumption 0.1 0.1 Exports 6.4 14.1 Imports 5.5 12.2 Real exchange rate 2.0 4.3

Share of GDP (%)

Investment -0.1 -0.2 Private savings -2.3 -2.1 Foreign savings 0.1 0.1 Government savings 2.1 1.7 Tariff revenue -0.4 -0.9 Direct tax revenue 2.6 2.6

Equivalent Variation (%)

Rural low inc. household 2.1 2.4 Rural medium income household 2.1 2.4 Urban low income household 1.3 0.9 Urban medium income household 1.3 1.3 High income household -1.1 -1.0 Total welfare 0.1 0.1 Gini coefficient -0.1 -0.2 Theil index -0.2 -0.3

Table 2.20: Simulation results for overall import tariffs reduction combined with 20 % in direct tax rates (scenario 3), % change from benchmark values

97 The results seem to suggest that the specific combination of trade and tax reform can improve overall poverty and income inequality in Brazil, with few differences with

respect to the level of reduction of the import tariffs, since the qualitative differences

between partial or total elimination of import tariffs were very small. Therefore, it is

possible to have an equity-efficiency policy that can bring openness and larger welfare

gains for the poor with smaller income inequality.

2.4

1.9 100% reduction 100% reduction import import tariffs in tariffs in Industry and Industry only (IND) Agriculture (MIX)

1.4 Overal 100% reduction import tariffs )

% 0.9 100% reduction import (

s tariffs Agriculture only e

g (AGR) n a

h 0.4 c

e r a f l e -0.1 w

-0.6

-1.1 Overall 100% reduction import tariffs + 20 % increase direct taxes -1.6

Rural Low Income Rural Medium Income Urban Low Income Urban Medium Income High Income

Figure 2.8: The main effects of different simulations on household’s welfare changes from base (%)

98 But it is interesting to note how an increase in direct tax rates plus an elimination of the import tariffs can help urban poor households to overcome welfare losses by eliminating only the import tariffs. Table 2.21 shows a comparison of consumption

expenditure changes for all household categories, for scenarios 1 and 3. Although high

income households in rural and urban areas are worse off than any of the scenarios

analyzed, poor households in both urban and rural areas are better off under scenario 3.

Scenario 3 can be considered as a combination of policies that is at the same time equity-

efficient because under these trade/tax reform all poor households in both rural and urban

areas become better off.

Rural low Rural Urban low Urban High income medium income medium income Scenarios household income household income household (%) household (%) household (%) (%) (%) 100% import tariffs 0.14 0.09 1.22 0.78 -0.03 (scenario 1)

100% import tariffs 1.85 1.41 2.98 2.50 -1.37 + 20 % direct tax rates (scenario 3)

Table 2.21: Main changes in consumption expenditures by households for scenarios 1 and 3

99 Regional impacts

We concentrate our attention at the regional level on the total elimination of the import tariffs combined with the 20 % increase in direct tax rates (combined trade/tax reform). Under combined trade/tax reform, prices and output in all regions are very similar to those found in section 2.8.2 (scenario 2). The changes in factor prices and factor income are identical for some labor categories. The main change for the factor prices is the effect on capital prices, reducing the final price and production, for large farm annuals, industry, and construction, for most regions.

All regions have a similar pattern of change for payments of factor of production as in section 2.8.2. Capital and land payments have larger changes in their payments under the combined policies, when considered only the reduction in the import tariffs

(scenario 1). In the same way, labor payments are larger (Tables E.23, E.24, E.25, and

E.26, Appendix E), showing a larger appreciation for unskilled labor wages relatively to skilled ones. Factor payments for unskilled labor increased more relatively to skilled labor. Most of capital and land used by sectors have some increase in their prices, but with the same direction of change as in section 2.8.2. Labor income for unskilled workers has a relative larger change than those for skilled workers in each agricultural activity.

All four regions experience many similar impacts from a reduction in the import tariffs combined with an increase in the direct tax rates. Some regional differences can be seen in Table 2.22. Once again, larger labor income gains are obtained in the North and

Center-West, mainly for rural households.

100

Rural low Rural Urban low Urban High Regions income medium income medium income household income household income household household household North 3.1 3.1 1.3 1.2 1.3

Northeast 1.7 1.8 1.2 1.0 1.1

Center-West 3.0 3.1 1.1 1.1 1.2

South/Southeast 2.0 2.0 0.7 0.7 0.6

Table 2.22: Overall regional impacts from an elimination of the import tariffs combined with an increase in the rate of direct tax on household’s labor income (% change from benchmark values)

Table 2.23 shows that the main capital income losses are in industry, caused by

elimination of the import tariffs, which removes protection for those capital-intensive

sectors. Construction capital income falls, but the losses are very small. After the

combined trade/tax policies changes, in general, land income increases, with the

exception of forest land.

Table 2.24 shows that the combination of the import tariffs reduction with an

increase in the direct tax rates in Brazil can improve regional distribution of income

within regions, as seen through a small reduction in all four indexes after simulation.

101

Factors North Northeast Center-West South/Southeast

Small farm 4.27 2.05 1.43 2.27 agricultural capital Large farm 2.37 -2.25 4.38 0.46 agricultural capital Food processing 4.05 3.35 3.59 3.25 capital Mining and oil 8.57 8.60 8.03 8.52 capital Industry capital -2.05 -1.99 -1.95 -1.96 Construction capital -0.69 -0.68 -0.68 -0.64 Transport and trade 3.06 2.64 2.73 2.79 capital Services capital 1.20 1.09 1.02 1.02 Arable land 1.47 -0.19 3.31 0.05 Grassland 4.18 4.78 4.10 4.04 Forest land 0.80 -1.86 -4.23 -5.10

Table 2.23: Overall regional impacts from an elimination of the import tariffs combined with an increase in the rate of direct tax on capital and land incomes (% change from benchmark values)

102

North Northeast Center-West South/Southeast Indexes Base(*) Sim(**) Base Sim Base Sim Base Sim Gini 0.258 0.255 0.353 0.352 0.402 0.400 0.475 0.474 Theil 0.115 0.112 0.229 0.228 0.275 0.272 0.390 0.388 H-H 0.106 0.103 0.201 0.200 0.275 0.273 0.388 0.386 Bourguignon 0.139 0.136 0.310 0.309 0.342 0.337 0.526 0.522 (*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.24: Regional income inequality measures before and after an overall elimination of the import tariffs combined with an increase in the rate of direct tax

The equity-efficiency trade/tax policies proposed do not bring important changes

in the income inequality measures seen in previous sections. Although Table 2.24 shows

that the income inequality is slightly reduced after using the combined trade/tax policies,

the overall results from simulation do not change the structure of how the labor income is

distributed within and between regions. The largest part of the overall inequality seems to

come from the inequality in labor income among the four Brazilian regions. According to

the Gini index, 78.6 % of the total labor income inequality is due to the inequality among

regions. The intensity of transvariation still is 4.8 %; as in Table 2.14, it represents the

part of the between-regions disparities issued from the overlap among the distributions.

This simulation does not modify the structure of the inequality within and among regions

in Brazil, in comparison to the simulation accounting only for the import tariffs

reduction.

103 As seen before in Table 2.15, the four inequality indexes also show that regions

North, Northeast, and Center-West contribute to reducing the overall inequality among

regions. Region South/Southeast has the most important contribution not only to increase

the overall inequality among regions, but also within this region.

If we consider only capital income, Table 2.25 shows that the decomposition of

capital income follows the same pattern as that of labor income. However, the proposed

combined trade/tariff policy seems to increase the inequality between regions and,

consequently, improves inequality of capital income within regions. As seen before with

labor income, most of the bad distribution of capital income in Brazil is due to substantial

differences among regions. This result is not surprising since it was also obtained by

Haddad et al. (2002), which found that trade liberalization through free trade area

agreements can lead to an increase in regional inequalities in Brazil.

% of the within-region % of the between- % of transvariation Indexes component regions component Base(*) Sim(**) Base Sim Base Sim Gini 17.9 17.8 77.5 77.7 4.5 4.4

Theil 45.7 45.2 54.3 54.8 - -

H-H 67.0 66.7 33.0 33.3 - -

Bourguignon 38.3 37.6 61.6 62.3 - -

(*) Base indicates values at the benchmark solution (**) Sim refers to values after simulation

Table 2.25: Contribution of the four decompositions to overall capital income inequality before and after simulation

104

2.9. Conclusions

Brazil has been negotiating trade agreements with the European Union, countries from all Americas in the creation of the free trade area of Americas (FTAA), and with

many other individual countries in many bilateral negotiations. Many, if not all, of these

agreements imply reduction in the protection of some sectors in the Brazilian economy.

Due to the complex structure of the Brazilian economy, with its highly protected capital-

intensive sectors, the impacts of such trade liberalization need to be scrutinized in order

to evaluate their impacts on the country as a whole, and also at its regional level.

The level of poverty in Brazil is high, with one of the world’s worst distributions

of income, which is very high within regions, but even worse among regions. It is

assumed that trade policy results in gains for households in the long run but, due to the

diversity and composition of the Brazilian economy, it is likely that some households can

lose mainly in the short to medium run. Our major policy concern was the interaction

between trade policy changes and poverty and income distribution in Brazil. The main

challenge of our research was not only to find an efficient trade policy in the Brazilian

trade liberalization process, but also to find an efficient instrument of policy that has, at

the same time, equity concerns, without hurting poor and reducing income inequality.

Brazil has a progressive direct tax rate system, but with very few categories and a

low level for the maximum rate. This study found an equity-efficiency policy based on a

combination of import tariff and an increase in the direct tax rate, in order to compensate

“losers” from considering only reduction in the import tariffs. A single country, static,

CGE model was used to evaluate trade policy experiments in Brazil under three different

105 scenarios, through a top-down-regionalized social accounting matrix (SAM) with 60

sectors divided in four regions and five household categories. Although the model used

has some limitations, it is not dynamic and multi-country, it is a very standard and

flexible type of model that can be extended to incorporate many other important features.

The model experiments were divided into three stages. In the first stage the model considered only the overall reduction in the import tariffs. The following stage consisted of sectoral import tariff reductions. The third and last stage was based on the attempt of

finding a complementary policy in order to compensate losers, mainly poor households,

to the import tariff reduction.

The main overall and regional consequences of a reduction in import tariffs

showed the following main conclusions:

(i) There was an overall welfare gain from trade reform;

(ii) Urban poor households lose, which indicates the presence of a trade-off

between aggregate welfare gains and the welfare gains to the urban poor from

reduction in import tariffs, as found by Harrison et al. (2003) for Turkey;

(iii) Overall and regional income inequality is reduced among households,

contrary to what was found in Haddad (1999) and Haddad et al. (2002);

(iv) The reduction or elimination of the import tariffs is not enough to change the

structure of the inequality in the distribution of the regional income. The

inequality among regions is the most important component that contribute to

the overall inequality in Brazil;

(v) South/Southeast has the most important weight in determining the inequality

of income among the regions in Brazil;

106 (vi) Although there were some small differences among regions, the main regional

impacts from trade reform indicate a similar pattern for the whole country, in

which industry had suffered the main negative impacts, consequently reducing

income and welfare of poor households employed in this sector;

(vii) There were positive and negative impacts on production of agricultural

commodities, such as forest products, manioc, soybeans, cocoa, corn,

horticultural products, rice, beans, coffee, poultry, cattle, and others,

depending on the region;

(viii) The mining and oil sector experienced the largest gains in all regions, since

this sector is very dependent on trade. This sector exports many mineral

commodities and imports a considerable amount of oil used for many

purposes.

In the second stage, the main results from the sectoral reduction in import tariffs seemed to follow the Heckscher-Ohlin-Samuelson model and Stolper-Samuelson theorem. The trade reform in agriculture showed that rural households have welfare losses, with opposite results for urban households from trade reform in the industry.

Therefore, a mix of import tariff reduction in agriculture and industry was simulated in an attempt to find a policy that would not hurt the poor. The results from such a policy were very similar to those in the simulation in the first stage, which showed that the urban poor get harmed and the regional income inequality became worse after the trade liberalization. Therefore, the search for a limited trade policy reform which could achieve

107 equity goals was not successful in finding any sectoral policy to do so. However, this search was important in concluding that there is a need for designing another policy

instrument to be combined to the import tariff reduction.

The third stage showed that it is possible to find an equity-efficiency policy

combination through import tariff reduction and an increase in the direct tax rates. The

simulation results showed that all households gain from the combined policies in the

short to medium run, with an overall improvement in the distribution of income. GDP,

exports, and imports increased, at the macro level. At the regional level, there was an

improvement in the distribution of labor income within and among regions. However, the

distribution of capital income among regions became more unequal.

Most of the trade policies evaluated in our study resulted in a distribution of the

gains in a way that poorest households (rural low and medium income) obtained the

largest proportion of increase in their incomes. It occurred because the shift of resources

from capital intensive sectors toward labor intensive agriculture and less capital intensive

manufacturing, which induced a larger increase in wages relative to other factor

payments.

In the next rounds of free trade negotiations, the Brazilian government should

consider the importance of interregional differences for a better understanding of the

consequences of those agreements at the national and regional levels. There should be

more options for public policy that can be used together with different strategies of trade

reforms, such as the tax reform proposed in this study, in order to generate a more

108 efficient and equitable relationship between producers and consumers, enhancing the

outcomes of such policies and even increasing Brazilian competitiveness in international

markets.

The results obtained in our study are conditional on the data and model used in all

simulations discussed in previous sections. It is expected that the consequent impacts on

the poor and on income distribution are sensitive to the design of the simulation

performed and to the main assumptions of the model. Therefore, there are many features

of the data and model that can be modified to examine more deeply the effects of trade

liberalization in Brazil, which can serve as a possible research agenda for future studies.

109 CHAPTER 3

AN EXAMINATION OF EXCHANGE RATE VOLATILITY IN THE MERCOSUR AND IN THE PROPOSED FREE TRADE AREA OF THE AMERICAS: SECTORAL TRADE IMPACTS IN BRAZIL

3.1. Introduction

The consequences of trade liberalization and market integration for developing

countries have become very interesting issues with the creation of free trade areas such as

the North American Free Trade Agreement (NAFTA), the European Union (EU) and

Common Market of the Southern Cone (Mercosur). On the economic side, there have

been many important changes in trade, macroeconomic policies, public sector and

regulatory policies because of different trade agreements worldwide. Important trade and

policy debates will continue in the future, not only with the creation and consolidation of

new integrated markets, but also with new agreements such as the Free Trade Area of the

Americas (FTAA) and the free trade agreement between Mercosur and the

EuropeanUnion (EU)53. The Mercosur and the FTAA, a proposed free trade area of

North, Central and South Americas, are our main interests of this research proposal.

53 Negotiations for the bilateral trade liberalization between Mercosur and the EU were launched during the Latin American and Caribbean – EU Summit in Rio de Janeiro in June 1999. According to Devlin (2000), the first meeting of the Bi-regional Negotiations Committee, to discuss organization, calendar and contents of the negotiations, took place in Buenos Aires in April 2000.

110 There are many questions and few studies54 about the consequences of the creation of the FTAA, and the role of macroeconomic instability in the Mercosur and in

the FTAA. The stability of the Mercosur bloc (Argentina, Brazil, Paraguay and Uruguay)

is in doubt due to the many recent problems of the countries. These include major

economic crises in recent years, the large devaluation of the Brazilian currency (Real) in

January 1999, the worldwide recession in 2000, and the recent Argentinean and Brazilian

crises, in 2002. The period 1999 to 2002, in particular, was a very difficult time for the

Mercosur countries, with many political and economic negative outcomes for this

economic bloc. In 2001 the main members of Mercosur, Brazil and Argentina, were

obligated to rely on new financial arrangements with the International Monetary Fund

(IMF) that had important political consequences, mainly in Argentina. Mercosur achieved

one of the highest levels of integration in Latin America, but the latest economic

slowdown has brought some uncertainty to the future of this regional integration

agreement, which was responsible for faster trade growth and promoted trade

diversification among its members.

The integration that occurred in the Mercosur might be seen as a weak one,

caused mainly by a lack of actual and continuous coordination of macroeconomic

policies of the four member countries. Since 1991, many different economic plans were

designed and implemented in these countries, aiming only their own economic stability

and bringing some doubts about the future of the Mercosur. This study does not have the

ambitious goal of stating that a country’s economic policies have to be only directed

54 Haddad (2002) and Diao et al. (2002) are two recent examples of studies that address the FTAA issue using a different approach than the one to be applied in this study.

111 towards the economic integration of the free trade area, without taking care of the

domestic macroeconomic instability that might happen. What will be considered here is

that the lack of a minimum of coordination of macroeconomic policies across countries,

within the same free trade area, can lead to the devastation of trade. Therefore, the

emphasis on stabilization programs rather than trade integration can cause tax

disagreements among countries in the free trade area, which would cause an excess flow

of capital within and from outside the region to countries with lower taxes, but with less

comparative advantage.

According to Baer et al. (2001), major swings in the real exchange rate may strongly affect the returns on investments, resulting in changes in the location of new production plants and reallocation of the existing ones.

It is our belief that the different economic stabilization plans adopted at different times, and implemented by different countries in the Mercosur and in the proposed

FTAA, can be responsible for most of the medium to long term real exchange rate volatility (from now on real ER volatility). The rationale behind this belief is in the fact

that long swings in the real exchange rate, caused by country-specific economic

stabilization plans, can increase the level of uncertainty among domestic and foreign

(trade partners) economic agents. One would not be able to hedge against this

uncertainty, since the price of this long term risk is not possible to obtain. According to

De Grauwe and Bellefroid (1986), it is most likely that the high degree of uncertainty

comes from these long swings in the real exchange rates.

This study will investigate Brazil’s main trade determinants in the Mercosur and

in the proposed FTAA, accounting for the possibility that the lack of stable

112 macroeconomic policies might hurt Mercosur trade, and it would be a problem for the

implementation of the FTAA as well. This study will also address the possibility that

Mercosur countries will be included in the FTAA, which is expected to be signed in

January of 2005. The main focus will be the different effects of medium to long run

exchange rate volatility on different sectors, including agricultural trade, since agriculture

is the least protected sector in the Latin American countries. In the empirical trade

literature, exchange rate volatility is responsible for negative effects in agricultural trade

(Cho et al., 2002; Maskus, 1986).

The objectives of this study are to specify and estimate a gravity model to

evaluate:

(i) the impacts of medium to long run exchange rate volatility, on different

sectors in the Mercosur and FTAA configurations; and

(ii) the Brazilian trade flow pattern in the Mercosur and FTAA configurations,

looking at the impacts that the changes in physical border, distance, tariffs and

income have in trade flows among these countries.

Section 3.2 specifies the problem to be addressed in this chapter. Section 3.3

examines the literature using gravity models to estimate and evaluate trade issues, and the

studies addressing the proposed FTAA. Section 3.4 describes and discusses the set of

data to be used in the empirical section. The gravity model to be estimated is defined in

section 3.5. Results and discussions are in section 3.6. Section 3.7 concludes the chapter.

113 3.2. Specification of the problem

In 1990, Brazilian President Collor de Melo implemented a program that transformed the whole economic structure of the country. For decades, the economy had

been characterized by its industrial bias through import substitution policies, credit

support, and fiscal incentives. The new program was basically a unilateral trade

liberalization with a generalized reduction of a complex tariff and non-tariff barrier

system. The Brazilian imports increased so much, that in 1995 the trade balance was

negative for the first time in more than 15 years.

The transition from an import substitution policy to a market-oriented policy in

the economy was not the only important structural change in the Brazilian external sector

in the 1990’s. The creation of the free trade agreement of the South-American countries

(Argentina, Brazil, Paraguay, and Uruguay), Mercosur, was an important factor in

consolidating the economic opening process that had started in Brazil. A gradual

reduction of tariffs was agreed upon for Mercosur countries between 1991 and 1994.

Trade within Mercosur countries increased substantially after 1995 (Table 3.1). Mercosur

has been an important instrument for macroeconomic coordination and attempts for

economic stabilization in these countries. At the same time that trade and integration

within countries increased, two of the largest countries of the region, Argentina and

Brazil, experienced many domestic crises over the whole decade.

The lack of macroeconomic coordination among Mercosur countries was one of

the main causes of the divergent and numerous price and exchange rate fluctuations,

affecting international trade and allocation of investments in the Mercosur. Since 1999,

114 both Argentina and Brazil have announced changes in nominal exchange rates55, directly affecting the returns on investments, and inducing shifts in the location of new production

plants and the reallocation of existing ones. Figure 3.1 shows the main exchange rate

devaluations in Brazil in February 199956, September 2001 and October 2002.

Country Total Exports Exports to Mercosur Total Imports Imports from Mercosur

Argentina 8.5 % 19.0 % 25.3 % 30.5 %

Brazil 6.0 % 22.9 % 11.8 % 15.5 %

Source: Ministerio do Desenvolvimento, Industria e do Comercio; Inter-American Development Bank.

Table 3.1: Average annual growth rate of trade in Argentina and Brazil for the period 1991-2000

According to Baer et al. (2001), the lack of macroeconomic coordination affects

international trade through two main channels: the international transactions risk channel,

and the political economy channel. The first is characterized by the increase of risk in

international transactions, affecting producer decisions on trade and resulting in a

55 Argentina continued with its fixed exchange rate regime, and Brazil had changed from a pegged regime to a more free oriented one, with free rates between minimum and maximum values (“moving bands”) defined by central bank.

56 Actually this is part of a large devaluation that started in November 1998.

115 different resource allocation than what would be expected from comparative advantage.

The second channel is also influenced by uncoordinated policies, which would promote

lobbying as a way to protect domestic markets when there is an increase in the import

penetration ratio (Trefler, 1993).

141

131

121

111

101

-99 -99 -99 -00 -00 -00 -01 -01 -01 -02 -02 -02 -03 Jan May Sep Jan May Sep Jan May Sep Jan May Sep Jan

Source: FGV

Figure 3.1: Index of Brazilian real exchange rate (Brazil/USA), Jan/99 = 100, period January 1999 to April 2003.

A direct consequence of the previous paragraph can be seen in the Mercosur,

whose members eliminated most trade barriers between 1991 and 1995 at a progressive

pace. They established a common external tariff (CET) structure in 1995, ranging from

zero to 20 %, applied to almost 85 % of total trade. But the tariffs were not totally

eliminated, and all countries were allowed to have lists of products sensitive to foreign

116 competition, which could be protected until 1999, for Argentina and Brazil, and to 2001,

for Paraguay and Uruguay57. Some sensitive products were totally excluded from the free trade area, and there are some capital goods, computers and related products, and

telecommunications equipments, that were not yet included in the CET regime.

Therefore, each country could have their own tariff rate on these products58.

The unstable and unsystematic implementation of the CET with so many arbitrary measures on the matter of tariff rates, resulted in commercial divergences among

Mercosur partners. This is another example of the lack of coordinated macroeconomic policies to promote actual trade integration. There is no doubt that trade improved

significantly after the creation of the Mercosur (table 3.1), but this success was a

fortunate59 consequence of a set of economic policies implemented by Mercosur

countries aimed at achieving domestic macroeconomic stability, seeking sustainable

growth and controlled low inflation, with no trade integration bias whatsoever. Since

1991, Argentina and Brazil have experienced trade deficits or surpluses caused by

different domestic economic actions that led them to adopt more protectionist or liberal

measures, which included taxes, tariffs, quotas and restrictions on import credit to

selected products, including those from Mercosur partners.

57 According to Averbug (1998), in Baer et al. (2001), there were 29 products in the Brazilian list, 212 in the Argentinean, 432 in the Paraguayan, and 963 in the Uruguayan one.

58 Tariffs were supposed to converge to 14 % by January 2001 for capital goods, for Argentina and Brazil, and by January 2006 for Paraguay and Uruguay. For the other products, their tariff rates are supposed to converge to 16 % by 2006. However, by middle of July 2001, Argentina decreased its extra-regional import tariffs for goods and computer equipment, causing some diplomatic divergences between Argentina and Brazil. (Baer et al., 2001)

59 Eichengreen (1998) is an example of this view.

117 There were many different economic events that increased the Mercosur partners’

divergences after 2001, which emphasize our argument about the negative effects of a

lack of coordinated macroeconomic policies on trade. Among them, we can cite

Argentina’s Convertibility Plan60 in 1991, the Brazil’s Real Plan in 1994, the large devaluations of the Real in 1999, 2001 and 2002, and the Argentinean Peso devaluation

in 2002. In general, these economic policies were implemented at different times, with

unusual consequences in terms of trade between the two main Mercosur partners. To

illustrate this, after the large devaluation in 1999, Brazil implemented a floating exchange

rate system. Argentina’s economy minister, Domingo Cavallo, mentioned the following

as a reference to the periodic devaluations of the Brazilian currency that was reducing

Argentina’s trade surplus with Brazil:

“…those who devaluate their currency are stealing their neighbors’ house.”

(Revista Veja, March 30th 2001, in Baer et al., 2001)

Our discussion basically implies that disharmonized macroeconomic policies

cause too many price and exchange rate swings, which affect trade in two ways. First, the

impact on domestic importers and exporters, due to an increase in the ER volatility, leads

risk-averse exporters and importers to reduce their supply and demand of traded goods

because they face more risk and uncertainty about their profits from overseas; or second,

they lobby for protection of import sectors.

60 With the Convertibility Plan the Argentinean Peso was pegged by law to the U.S. dollar at a rate of one to one. More details about this and the other macroeconomic issues related to Mercosur, see Eichengreen (1998).

118 Therefore, it is interesting to verify the consequences of such exchange rate instability on different sectors in these countries. There are many studies addressing the influence that volatility in exchange rates has on a country’s economy. Many of them claim that exchange rate volatility reduces the level of trade (Hooper and Kohlhagen,

1978; Thursby and Thursby, 1987; Cushman, 1988; Frankel and Wei, 1993; Eichengreen and Irwin, 1995; Rose, 2000). But as pointed out by Sauer and Bohara (2001), factors such as degree of risk aversion, hedging opportunities, the currency used in contracts, or the presence of other types of business risk, the direction and magnitude between exchange rate uncertainty and trade is an empirical question that needs to be investigated.

Exchange rate volatility and macroeconomic instability not only can affect the future of Mercosur, but may also affect the proposed FTAA. The FTAA was launched in the Miami Summit in December 1994 and is expected to start in January 2005. After its implementation, the FTAA will be the largest free trade agreement in the world, and it is important to know what the main determinants are of the pattern of trade within the proposed FTAA, considering the exchange rate instability in countries like Argentina and

Brazil.

This study investigates the effects of exchange rate volatility on trade under

Mercosur and FTAA scenarios. The main focus is to estimate the trade flow patterns of

Brazil in both configurations, and to verify how the trade responds when there exist changes in exchange rates and in other trade determinants such as tariffs, distance, GDP, and third country exchange rate volatility (third country effect).

119 Some questions to be answered in this study are: What would be the consequences for Brazilian sectors under the FTAA set up? What would happen to the trade flows if the exchange rate becomes more volatile? Would this volatility bring positive or negative effects on trade for Brazil? How much would the trade flows change as a result of a reduction in tariffs, or of an increase of a country’s GDP?

There are some additional issues that can be addressed in this analysis. The first is the fact that the United States and Brazil/Argentina have economies that are very different in size, structure and composition, which can bring interesting insights about the interpretation of the empirical results. The second issue is that the difference in factor endowments between these economies can characterize their production under different environments. Specialization and increasing returns to scale can bring different results from the empirical estimation. The third is a possible simultaneity bias, since the exchange rate regimes are not constant over time. They are often affected by the national government or central bank, when attempting to stabilize the economy, and/or the balance of trade between major trade partners. The fourth and last issue is a consequence of the previous one, since the volatility of the exchange rate in other countries can affect the trade flows between two countries. This is a particularly interesting issue, mainly in developing countries, that have witnessed important adverse consequences from economic crisis worldwide. Even though all previous issues are important, we will only approach the last two. Although it is relevant and interesting to address the first two issues, it would be necessary to test some different theoretical models under different assumptions, which is not a goal of this research.

120 Several distinguishing features of our approach are important contributions of this

study: (i) the level of disaggregation and the sample size are larger than those used in

other studies (we use 2-digit SITC61, rather than 1-digit SITC trade data as in most other studies); (ii) the evaluation of impacts of medium to long run exchange rate volatility, instead of short-run volatility; (iii) the sectoral effects of exchange rate volatility, including agricultural trade in Brazil and in other Mercosur countries (literature shows

that the emphasis has been on U.S. agricultural trade flows); (iv) the use of fixed

(random) effects to capture the trade flows patterns on both Mercosur and FTAA

specifications.

3.3. Literature Review

3.3.1. Gravity models

To analyze such important issues, a powerful tool of analysis like the empirically successful gravity model (Tinbergen, 1962), can be used to determine the bilateral trade

patterns among Mercosur countries and to predict this pattern for Brazil in the proposed

FTAA configuration. A gravity model can account not only for trade flows, but also for

border effects (such as transport costs, trade barriers, location, contiguity, etc),

population, countries’ GDP, and exchange rate effects, which has been a major

macroeconomic variable that has influenced the bilateral trade flows in Mercosur in

recent years. This seems to be a legitimate issue to be studied due to the currency

devaluations in recent years in Argentina and Brazil.

61 SITC stands for Standard Industrial Trade Classification.

121 The international trade literature has studied the determinants of bilateral trade

flows focusing on the Linder hypothesis, the gravity models, and the effect of exchange

rate variability. The bilateral approach of Linder’s hypothesis62 is that trade of manufactured goods between two countries will be inversely proportional to the difference in their per capita income. Actually there exists a high proportion of bilateral

trade that occurs between countries with similar income63.

Although the Heckscher-Ohlin-Samuelson model (HOS), which is basically a factor-endowment model, has been popular in the international theory of trade, it is known that the differences in relative factor endowments are not the only cause of trade

(Markusen, 1986). Studies such as Krugman (1979), Helpman (1981) and Ethier (1982) about bilateral trade in manufactured goods between similar countries added imperfect competition, scale economies, and product differentiation to test Linder’s hypothesis.

Therefore, we have two main types of trade explained by the international trade theory.

The first is the inter-industry trade, which is the basic HOS model based on differences in

factor endowments. The second is the intra-industry64 trade that occurs in manufactured goods and services.

Gravity models have been employed to evaluate many different issues related to bilateral trade flows in the international trade literature. As Feenstra et al. (2001) point out, a gravity equation can be used to describe international trade flows as a log-linear specification of the income and distance between trading partners. According to Rose

62 For different ways of evaluating the Linder’s hypothesis see Blejer (1978) and Markusen (1986). 63 Deardorff (1984), however, rejected the Linder hypothesis after controlling for transport costs.

64 For a theoretical discussion about intra-industry trade see Grubel and Lloyd (1975).

122 (2000), the gravity model of international trade is a very successful model used in

economics. As described by Leamer and Levinshon (1995, p.1384), a gravity model

provides: “Some of the clearest and most robust empirical findings in economics”.

Anderson (1979, p.1) also stresses the qualities of the gravity models in international trade: “Probably the most successful empirical trade device of the last

twenty five years is the gravity equation”.

A gravity model, or gravity equation, is a reduced form equation from a general

equilibrium system of international trade in final goods, and the model assumes that trade

between two countries is dependent on their size, stage of development, market openness,

and proximity. Trade is directly proportional to the size of the countries, and it is

negatively correlated to the distance between the countries. It was based on the “gravity

theory” from physics, which says that the “force of gravity” between two objects or

planets is directly proportional to the size and inversely proportional to the distance

between them. Analogously, the “trade flow” between two countries is a function of

income and distance, and other variables (population, contiguity, language, transport

costs, tariffs, etc).

The first econometric studies of trade using gravity models were Tinbergen

(1962) and Poyhonen (1963). In both studies the gravity equation was specified using

only intuitive justification. Due to its strong empirical explanatory power, the gravity

model became a very popular tool of bilateral trade flow analysis. But for many years

there was a lack of theoretical foundations for the gravity model usage and specification,

which was the main reason for its “poor reputation among reputable economists”

(Baldwin, 1994). Contrary to what was thought, gravity models do have such theoretical

123 foundations. Anderson (1979), Krugman (1979), Helpman and Krugman (1985),

Bergstrand (1985, 1989, 1990), and Evenett and Keller (2002) added these foundations.

Linnemann (1966) included more variables and tried to justify theoretically the model

through a Walrasian general equilibrium system. Leamer and Stern (1970) derived the

gravity model from a probability model of transactions, but they did not make any

linkage to the HOS model. Leamer (1974) joined the gravity equation and the HOS

model to specify the empirical model, but he did not integrate the two approaches

theoretically.

There were many other attempts to formally derive the gravity equation. The first

to assume product differentiation was Anderson (1979). He used an Armington

assumption, which implies that products are differentiated by country of origin.

Bergstrand (1985) used the same assumption and found empirical support for the

assumption that goods were not perfect substitutes, and that imports were closer

substitutes for each other than for domestic goods. Since the HOS model was inconsistent

with some of the empirical findings of Deardorff (1984), Helpman (1987) found that the

factor-proportions theory contributes little to determine the volume of trade among and

within groups of countries. Helpman also tested a simple gravity equation, and found

empirical support for the monopolistic competition model. Bergstrand (1989, 1990)

assumed the Dixit and Stiglitz (1977) monopolistic competition and product

differentiation among firms rather than among countries to derive a gravity equation and

to examine intra-industry trade.

Deardorff (1998) shows that a gravity model can be consistent with the HOS

model with non-homothetic preferences without any role for monopolistic competition,

124 as in Bergstrand (1989). When Deardorff incorporates transport costs, the distance

between two countries reduces trade, and trade is sensitive to the relative distance

between importer and exporter countries relative to the average of all demanders’ relative

distances from the exporter country.

Therefore, the success of gravity models cannot be considered as evidence of any

trade theories with imperfect competition and scale economies as suggested by Helpman

(1987). Deardorff (1998) and Evenett and Keller (2002)65 conclude that since specialization is the “force of gravity” that is responsible for the empirical success of

gravity models, it is not necessary to mention any trade model to derive a gravity

equation. If countries are specialized, consumers will want to buy those things that are

not available in their home country. The more things consumers do not have available,

the more they want to buy from other countries. Therefore, the output of other countries

determines the trade flow. The home country’s output has similar behavior, since the

more income the consumers have, the more they want to buy. When each good is

produced in each country (complete specialization) and preferences are identical and

homothetic, the elasticity of trade with respect to each country’s income is equal to one.

This is true no matter which theoretical consideration one takes to explain the

specialization, be they increasing returns to scale in differentiated products, technology

differences in Ricardian trade, large factor endowments in HOS trade, or transport costs

on any type of trade based on endowments.

65 This study examines the HOS theory and the increasing returns of scale theory to explain the empirical success of the gravity equation. Since both theories can predict the gravity equation, they estimated pure and mixed versions of both theories for a cross-section data for 58 countries. Their findings suggest that predictions of a model with imperfect specialization that is based only on differences on factor endowments find support in the data.

125 Feenstra (2002) estimates a gravity model to evaluate trade between and within

Canada and U.S. using a monopolistic competition model with constant elasticity of

substitution (CES) allowing for transport costs and trade barriers to be incorporated in the

analysis. When there are transport costs and trade barriers, which are called border

effects, prices are no longer equalized across countries, and there is a necessity of

building more complex gravity models.

Brown and Anderson (2002) study a similar problem considering the potential for

further economic integration among Canadian and American regions through an

estimated constrained gravity model derived from microeconomic foundations. Results

show that after controlling for changes in output, distance, wages, productivity, and

localization economies, the border remains an important barrier to trade.

There are many other studies that explore the gravity model as an international

trade application. Some of them explore the effects of exchange rate variability on trade

flows66. Rose (2000) uses a panel gravity model to study the effects of exchange rate volatility and currency unions on the volume of bilateral international trade. The results

show that two countries sharing the same currency trade three times as much as they

would with different currencies. Rose and van Wincoop (2001), and Glick and Rose

(2001) also investigate the issues related to currency unions and trade through a gravity

model.

66 Such as Hooper and Kohlhagen (1978), Cushman (1983), Kenen and Rodrik (1984), Thursby and Thursby (1987), Frankel and Wei (1993), Eichengreen and Irwin (1995), and others.

126 According to Cho et al. (2002), there are very few studies that account for impacts of exchange rate variability on agricultural trade. Some of the first attempts to investigate

such effects are Schuh (1974), Batten and Belongia (1986), Haley and Kissoff (1987),

and Bessler and Babula (1987).

Some studies deal with the impacts of short-run67 exchange rate volatility on agricultural trade. Pick (1990) does not find any effect of the exchange rate risk on U.S. trade flows to developed countries, but he finds a negative effect on trade flows to developing countries. Klein (1990) finds negative impacts of short-run exchange rate

volatility on U.S. agricultural trade.

Cho et al. (2002) estimate a gravity model for many developed countries to evaluate the effect of exchange rate uncertainty on agricultural trade. Their results show

that real exchange rate uncertainty has had a negative impact on agricultural trade for the

period from 1974 to 1995.

The competitiveness of a country is reduced from an overvaluation of its currency

and vice-versa. Tweeten (1989) finds that the appreciation of the U.S. dollar during the

1980s had negative effects on U.S. agricultural exports. Cho (2001) argues that, due to

the loss of competitiveness, some sectors can lose domestic and foreign markets resulting

in reduction of employment and output. This outcome can result in lobbying for

protection by those groups that lost with the exchange rate overvaluation. If a

protectionist measure is adopted by the government due to the lobby, it is not easily

removed when exchange rate depreciation occurs. In the same way, some industries gain

67 Peree and Steinherr (1989) consider as short-run exchange rate volatility if one takes the exchange rate uncertainty for a period less than one year.

127 from the undervaluation, which can induce resources to enter such industries. However,

when the undervaluation disappears the industries can seek import barriers or subsidies.

This sequence of overvaluation and undervaluation can ratchet up the level of protection.

Countries that experience many fluctuations in their exchange rates for a long time period

are likely to have a reduction on their trade growth (De Grauwe, 1988). Pick and Vollrath

(1994) find that movements of exchange rates in developing countries have negatively

affected the competitiveness of the agricultural sector.

Cho (2001) finds that the empirical research about the effect of the long run

exchange rate variability on international trade flows is very sparse. The reason is

because the random walk hypothesis of real exchange rates was recently rejected68.

3.3.2. The proposed FTAA

There are many studies addressing issues about Mercosur, but few that analyze the proposed FTAA. Among them, there are very few studies that use the gravity model as an analytical tool to evaluate this free trade agreement. No study was found that investigates the effects of exchange rate variability on trade flows for the FTAA. Baer et al. (2001) was the only study that did this analysis for Mercosur.

Watanuki and Monteagudo (2001), Diao et al. (2001), Valls Pereira (2001),

Tourinho and Kume (2002), Haddad et al. (2002), and Harrison et al. (2002), analyze and compare the Mercosur under the FTAA and European Union agreements through a static computable general equilibrium (CGE) model. There are many partial equilibrium

models applied to evaluate the FTAA, such as Carvalho and Parente (1999), Carvalho et

68 More details about different methods of determining real exchange rate see Mark (2001).

128 al. (1999), Nonnenberg and Mendonca (1999), and Mattson and Koo (2003). Canuto et al. (2003) make a descriptive analysis of the effects of the FTAA on foreign investments in services in Brazil. Castilho (2002) explores the main studies about the effects of the

FTAA in the Brazilian economy.

There are few studies applying gravity models to investigate the effects of free

trade agreements in the Brazilian economy. For instance, Castilho (2001) tries to identify

which sectors and products should be given more attention in the negotiations between

Mercosur and the European Union. Azevedo (2002) estimates a gravity model to verify

and compare the differences on trade patterns of Mercosur countries before and after the

creation of Mercosur. Piani and Kume (2000) estimate a gravity equation to evaluate the

influence of six trade agreements on trade flows for 44 countries. They emphasize the

effects of NAFTA and Mercosur, and the data set is basically aggregated. Porto and

Canuto (2002) assess the impact of Mercosur on Brazilian regional development through

a gravity model. The results suggest that Mercosur contributed to an increase of the

regional inequality in Brazil. Porto uses aggregated regional data. Barcellos Neto et al.

(2002) study the effects of FTAA on selected blocs, Mercosur, NAFTA and the Andean

Pact69. Through a gravity model they examined the effects on trade flows attributed to each bloc formation, separating these from other factors influencing trade. They created

scenarios from the results to make inferences about the FTAA. They used aggregated

data for bilateral trade.

69 The Andean Pact is the oldest free trade agreement in Latin America, and it includes Bolivia, Colombia, Ecuador, Peru, and Venezuela.

129 Baer et al. (2001) investigated the lack of macroeconomic policy coordination

between Argentina and Brazil, which caused a “battle” for attracting foreign direct

investments. They also find that bilateral exchange rate volatility has negative effects on

trade between these countries. To avoid the endogeneity problem between trade barriers

and import penetration ratio, they estimated a fixed effects panel data model, concluding

that large swings in bilateral exchange rate indirectly generates barriers of trade within

Mercosur.

3.3.3. Effects of exchange rate volatility on different sectors

The absence of a well managed and stable exchange rate system can be an

important source of misalignment70, mainly for those countries that have pegged their currencies to the US dollar. Argentina and Brazil experienced pegged exchange rate

regimes during the 1990’s, which brought substantial and persistent deviation of nominal

exchange rates from their macroeconomic fundamentals. Even though the inflation rates

in these countries dropped from over 1000 % per year with their domestic economic

stabilization programs, the reduced inflation rates were still larger than the ones observed

in the United States.

The size of such misalignments, from now on we call it the long run real

exchange rate variability, is an important factor that affects international trade, and it is

explained by the hysteresis model (Baldwin, 1988; Baldwin and Krugman, 1989).

70 This term means the departure of nominal exchange rates from long run equilibrium level or economic fundamentals. More detailed economic consequences of these misalignment problems can be found in Tweeten (1989). The long run exchange rate volatility can be considered as a proxy of the size of misalignment (De Grauwe and Bellefroid, 1986).

130 The hysteresis effect of exchange rate movements says that an unexpected bilateral exchange rate misalignment can cause a permanent change in market structure

(Baldwin, 1988; Baldwin and Krugman, 1989). For example, many exporters can enter a

country that overvalued its currency, changing permanently the country’s market

structure. But after the misalignment problem disappears, the foreign firms will stay in

the country because of the high amount of initial investment. According to Cho (2002),

economists accept that there is some range of inertia of exporters’ entry and exit

decisions due to changes in exchange rate movements.

According to Cho (2001), the hysteresis model and empirical evidence show that

exchange rate changes bring different impacts on different sectors in an economy, due to

specific characteristics of each industry. These characteristics could be given by different

levels of initial investment (Baldwin, 1988), the level of substitutability of goods

(Dornbusch, 1987), or whether the products are durable or not (Froot and Klemperer,

1989). Therefore, the impacts of exchange rate volatility can be very different from one

sector to another. There are very few studies addressing the sectoral impacts of exchange

rate volatility in the literature (Cho, 2001).

One of the theories that explains the exporters’ reaction to exchange rate volatility

is suggested by Baldwin (1988). The model assumes there is an exporting firm that is

operating in a foreign market, and that this firm has spent large initial sunk costs to start

its business. If there is an overvaluation of the exchange rate of the exporter’s currency,

the firm would stay in the foreign market as long as the exchange rate changes were

within a specific range that covers the costs of exit and the initial sunk costs. Under

131 imperfect competitive markets, or imperfect substitute products sold by the exporter,

there would be a partial passing through of the exchange rate through price

discrimination in the foreign market.

Baldwin’s model is only a specific attempt to explain the exporters’ behavior

when facing a change in the exchange rate. Froot and Klemperer (1989), for example,

add the possibility that the cost of losing market share in a foreign market is larger than

the cost of staying in the market even with substantial overvaluation of the exporter’s

currency. Dornbusch (1987), using a Cournot model, suggests that the exporter’s price

reaction depends on the market concentration, number and relative concentration of

foreign and domestic firms, and also on the degree of substitution between domestic and

foreign products.

Although many theories try to explain how misalignments happen and affect

international trade, there is no empirical consensus about which theory gives the best

explanation to this phenomenon. MacDonald and Taylor (1992), for example, use the

fully revealing rational expectation hypothesis to test if foreign exchange markets are

fully efficient, which would make it impossible for traders to earn excess returns using

their private information71. This is one reason why recent theory deviates from the traditional macroeconomic assumptions that all participants have the same expectations,

focusing more on the heterogeneous expectations of the economic agents to explain the

exchange rate movements (Frankel and Rose, 1995).

71 There is some evidence that the forward foreign exchange rate is a biased and inefficient predictor of future spot exchange rates, putting in doubts the efficiency in foreign exchange markets (Cho, 2001).

132 Another view of the negative effects of the exchange rate volatility

(misalignment) on trade comes from the “political economy view” of the misalignment

(McKinnon, 1988), or from the theory of endogenous protection (Mayer 1984). In case

that a currency is overvaluated, the country’s competitiveness is reduced72, which motivates some sectors that lose domestic and foreign markets to lobby to pass protectionist legislation. Analogously to the Baldwin and Krugman’s hysteresis model, the protectionist legislation is not eliminated after the currency depreciates. The same

problem would happen, but in the opposite direction, in the case of the currency

undervaluation.

Due to so many theoretical and empirical ambiguities about the relationship

between long run exchange rate volatility and trade, it is not surprising that the empirical

literature about this topic is so limited. Facing the theoretical difficulties that the scarce

literature has presented about the sectoral impacts of the long run exchange rate volatility,

it is our objective to examine the main impacts of such volatility on trade across sectors

without trying to find a theoretical explanation for them. However, the Baldwin and

Krugman’s hysteresis model may help to interpret the results obtained for the different

sectors. According to this model, sectors with large amounts of initial investment would

be less susceptible to the shock-inducing structural change discussed before, suffering

less from the exchange rate volatility. In the case of those sectors that do not need much

initial investment, these sectors would tend to be more sensitive to the exchange rate

volatility. But whether these impacts are positive, neutral, or negative, is an empirical

question that needs to be investigated.

72 It can also be represented through an increase in the import penetration ratio.

133 3.4. Data and Issues

The main data to be used will consist of bilateral trade, and a simple average of

the tariffs between Brazil and other 17 countries in the South, Central and North

Americas (Appendix E), for the period 1989 to 2002, from the TRAINS(Trade Analysis

and Information System)/WITS (World Integrated Trade Solution) (UNCTAD)73 package. This is a pooled dataset consisting of the nominal value of exports from one country to the other, for each sector (agriculture, chemicals, livestock, mining and oil, manufactured, and all sectors together)74, at a 2-digit SITC75 code. The aggregated sample will consist of 15,232 observations (17 countries times 64 different products times 14 years).

Since the focus of the study is to evaluate the effects of exchange rate variability on Brazilian agricultural trade on Mercosur and the proposed FTAA, the data are

converted into the exporting country’s currency using nominal exchange rates76 and deflated by the consumer price index of the exporting country, from the International

Financial Statistics (IFS). Other data requirements are nominal GDP values, and

population, given by International Financial Statistics (IFS), and distance, which is the

great circle distance between economic centers77, given by Soloaga and Winters (2001).

73 United Nations Conference on Trade and Development.

74 The list of products in each sector can be seen in Appendix F.

75 SITC stands for Standard International Trade Classification.

76 We used end-of-period exchange rate from IMF’s International Financial Statistics (IFS).

77 The great circle method is given by the weighted average of the latitudes and longitudes of the main economic centers.

134 GDP was also deflated as described before. The real exchange rates78 are calculated through the following expression:

≈ ’ ∆ CPI s,t ÷ (3.1) RERis,t = NERis,t ∆ ÷ « CPI i,t ◊

Where RERis,t and NERis,t are the real and nominal exchange rates for country i with

respect to the country’s s currency at time t. The expression 3.1 shows how the real exchange rate is calculated for country i using the 1995 U.S. dollars as common foreign

currency from country s (the United States). The CPIs,t reflects the consumer price index in the United States at time t. The CPIi,t reflects the consumer price index in country i at time t. Therefore, the bilateral real exchange rate (Xij,t) for each country can be obtained by the ratios between each of the 17 countries’ real exchange rates and the Brazilian real exchange rate (j).

The medium to long run exchange rate uncertainty79 is essential for our study. It can be obtained using two different procedures as proxies for the long run exchange rate uncertainty, the moving standard deviation and the Perre and Steinherr volatility measures80.

78 The main reason to use real exchange rate in this study is because the nominal and real exchange rates are expected to be highly correlated, but the real ER volatility is expected to be larger than the nominal ER movements. De Grauwe and Bellefroid (1986) explain that when a currency depreciates by some proportion, it is likely that the real exchange rate will change by a smaller amount than the initial depreciation, due to the inflation changes will in general offset the nominal initial depreciation. These differences between real and nominal exchange rates can become important when medium to long run variability are investigated, which is what we have in our study.

79 For detailed discussion about measures of exchange rate volatility, see Lanyi and Suss (1982), Brodsky (1984), and Kenen and Rodrik (1986).

80 Because the volatility measures are used as proxy measures for the actual long run exchange rate uncertainty, volatility and uncertainty are used interchangeably throughout this study.

135 The moving standard deviation of the log differences of the real bilateral ER is a modification of the standard deviation usually employed in many studies using cross-

section or time-series data, such as Kenen and Rodrik (1986), De Grauwe and Bellefroid

(1986) and Dell’Ariccia (1999). The moving standard deviation is used here because it

has to be time varying due to the time-series feature of the panel data we have, as in Cho et al. (2002).

The moving standard deviation of the log differences of the bilateral real ER (Sijt) is given by:

k 2 ƒ(xij,t−l − xij,t ) (3.2) S = u = l=1 ij,t ij,t k −1

Where Xij,t is the bilateral real exchange rate, xij,t = ln(Xij,t) – ln (Xij,t-1), and k = 2, 4, 6, 8,

81 and 9 years . xij,t is the mean of xij,t over the past k years.

The other measure of real ER volatility to be used is based on Peree and Steinherr

(1989), which assumes that the uncertainty of the economic agents is defined by previous experiences about the maximum and minimum values, which are adjusted through the experience of the last year relative to an “equilibrium” exchange rate. Therefore, large changes in the past generate expected volatility. They proposed the following measure of exchange rate uncertainty:

t t » k ÿ max X − min X X ij,t − X ij,t (3.3) V = u = ij,t−k ij,t−k + …1+ Ÿ ij.t ij,t t … k Ÿ min X ij,t−k X ij,t ⁄

81 The time period covered is arbitrarily chosen to investigate the robustness of the results.

136 t Where k is the period length; min X ij,t is the minimum value of the absolute value of the

t bilateral real exchange rate in the last k periods; max X ij,t is the maximum value of the

k absolute value of the bilateral real exchange rate in the last k periods; X ij,t is the mean of

the absolute value of the bilateral real exchange rate over the last k periods. It is a proxy

for the long run bilateral real exchange rate equilibrium. Each period in our analysis is

equivalent to each year. The reason is that our emphasis is on effects of medium to long

run exchange rate uncertainty.

According to Dell’Ariccia (1999), and Cho et al.(2002), the first term can be

considered as the “accumulated experience” term, since agents remember very well the

extreme values reached by the exchange rate in the past, even when the differences

become small. The second term is the recent information in each period t and represents

t k 82 how far each real exchange rate (X i,j,t) is from the “equilibrium” exchange rate (X i,j,t) .

Figures 3.2 and 3.3 show the two measures of bilateral real exchange rate

volatility for Mercosur using an 8-year time window. According to the moving standard

deviation measure, the bilateral real ER volatility between the Argentinean peso and the

Brazilian real has relatively high levels of volatility. However, after 1997 the uncertainty

was strongly reduced due to constant depreciation of the Brazilian currency later on. We

can note that the Paraguayan guarany/Brazilian real volatility is the most stable in

Mercosur. The stability of the ER volatility for the period 1992 to 1997 can be due to the

Brazilian exchange rate policy adopted to be an anchor to control domestic inflation

(Figure 3.2).

82 According to Mark (1995), there is no way to accurately measure the long run equilibrium exchange rate. For this reason we adopted a simple mean for the whole sample period to obtain a proxy of such equilibrium measure.

137 The Peree and Steinherr measure of ER volatility (Figure 3.3) is characterized by

a decreasing behavior from 1990 to 1999, probably because of the “accumulated

experience” feature of this measure of volatility, which takes into account the exchange

rates from the past83. After the large devaluation of the Brazilian real in 1999, the ER

volatility grew mainly for the Argentinean peso/Brazilian real bilateral volatility, which

became the largest volatility among Mercosur countries only after the Real Plan was

implemented in the middle of 1994. In general, the Peree and Steinherr measure has the

largest volatility relative to the moving standard deviation measure.

0.35 ) t j i S (

e r 0.30 u s a

e Argentina/Brazil M

y t i l i t a l 0.25 o V

e Paraguay/Brazil t a R

e g n

a 0.20 h c x E

l a r e t a l i 0.15 B

a Uruguay/Brazil e R

0.10 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Figure 3.2: Bilateral real exchange rate volatility (moving standard deviation measure) in Mercosur, 1989 – 2002.

83 The inflation rates were very high in the beginning of the 1990’s in all these countries, which also affected the behavior of the exchange rates in this period. The relationship between inflation rates and exchange rates will be clarified later in this section through the monetary model of the exchange rates.

138

4 ) t j i V (

3.5 e r

u Paraguay/Brazil s a e M

t 3 i l i t a l o V

e t

a 2.5 R

e g n

a Argentina/Brazil h c x 2 E

a Uruguay/Brazil r e t a l i B

l 1.5 a e R

1 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002

Figure 3.3: Bilateral real exchange rate volatility (Peree and Steinherr measure) in Mercosur, 1989 – 2002.

There exist other two issues that can affect the way we will use the data. The first issue is the potential presence of simultaneity bias, which is due to the endogeneity of exchange rate in case central banks could intervene to stabilize the bilateral exchange rate with their trade partners. The exchange rate volatility and trade would still be negatively correlated, but it would not be clear which direction the causality would occur.

Consequently the OLS regression would produce biased estimates, since it would not be possible to distinguish between the effects of investor’s risk aversion and the effects of

139 central bank policies. Since it is not clear if there is, or if it would be available, an

instrumental variable that could be used to solve the problem, Dell’Ariccia (1999)

proposes a solution based on panel data models. The idea is that the central banks try to

smooth the behavior of the bilateral exchange rate against their main trade partners.

Dell’Ariccia considers that the exchange rate uncertainty becomes a function of the trade

shares between the two countries relative to the total trade between them:

Tij,t T ji,t (3.4) uij,t = λij,t −φ −ϕ +ηij,t Tit T jt

Where φ and ϕ represent the stabilization effort functions of the two central banks or

governments, λ is a constant term, and η is an error term. The term Tij,t/Tit (Tji,t/Tjt) is the

exports from country i to j relative to i's total exports. In case the bilateral trade shares are

relatively constant over time, we can rewrite (3.4) as:

(3.5) uij,t = λij,t +φij +ηij,t

Equation (3.5) means that the central bank effect is treated as a country-pair fixed effect.

Therefore, the fixed effect model would produce unbiased estimates.

According to our data, the Brazilian export shares to Argentina, for example,

increased significantly from 1991 to 1998, decreasing after that. For Paraguay and

Uruguay, the changes on the Brazilian export shares were very small. For many other

countries in many different sectors, there were some changes over time, but not very

large84, which brings the question of the importance of a formal test to address the

84 For instance, considering all sectors together, the shares of the Brazilian trade with the United States were around 35 % from 1991 to 1995. It fell to 17 % from 1996 to 1998, rising to 25 % from 1999 to 2000, and increasing again to 38 % after that. 140 simultaneity bias problem. As in Dell’Ariccia (1999), we will use an instrumental

variable procedure to test the simultaneity bias through a Hausman test. The null

hypothesis is an absence of simultaneity causality bias through the fixed (random) effects

estimators against the instrumental variable (IV) estimator. If the hypothesis is rejected,

the fixed (random) effects estimators are biased, although consistent. In the case we do

not reject the null hypothesis, the fixed (random) effects estimators are unbiased and

consistent. In case the fixed effects estimation show to be superior to the random effects

estimations, through use of an analogous Hausman test for that, the fixed effect model

will also take care of the simultaneity bias through the country-pair dummies, since the

trade shares between Brazil and all other countries seem to be relatively constant.

But the question becomes, what instrumental variable could be used for the

bilateral real ER volatility? Frankel and Wei (1993) used the standard deviation of the

money supply as an instrument for the exchange rate volatility. Unfortunately, as pointed

out by Dell’Ariccia (1999), exchange rates can be used to determine the monetary policy

of countries in the European Union, which is similar to the Latin American countries in

the 1990’s, when exchange rates were an anchor for the monetary policies

implemented85.

85 The Argentinean Convertibility Plan in 1991, and the Brazilian Real Plan in 1994 can be some examples.

141 This study tries to find a theoretical model that provides a good instrument for the

exchange rate volatility. According to Hallwood and MacDonald (2000) and Mark

(2001), we can start with a monetary model under flexible exchange rates86, where

money demand functions for domestic and foreign markets are given by:

d (3.6) mt - pt = φyt - λit

d* * * * (3.7) mt - pt = φyt - λit

Where 0 < φ < 1 is the income elasticity of money demand, λ > 0 is the interest rate semi-

87 d d* elasticity of money demand , mt (mt ) is the logarithmic domestic (foreign) money

* * demand, pt (pt ) is the logarithmic domestic (foreign) prices level, yt (yt ) is the

* logarithmic domestic (foreign) real income level, and it (it ) is the domestic (foreign)

nominal interest rate.

With flexible exchange rates and equilibrium in the money market, money stock

is exogenous. Then we can combine both expressions to get:

* * * * (3.8) pt - pt = mt - mt - φ (yt - yt ) + λ (it - it )

Under purchasing-power parity and uncovered interest parity, international capital

market equilibrium is given by:

* * (3.9) it – it = Etst+1 - st = Etst+1 – pt – pt

Where Etst+1 ≡ (Etst+1/It) is the rational expectation of the exchange rate at period t+1,

conditioned on full information set (It) available at period t.

86 The fact that Argentina and Brazil adopted pegged exchange rate regimes for a large part of our sample does not cause any problem for this model because the exchange rate volatility is mainly a problem originated from flexible regimes.

87 In order to simplify the algebra, φ and λ are considered the same across countries.

142 Combining equations (3.8) and (3.9), we obtain what we call economic

“fundamentals”, given by the following expression:

* * (3.10) ft ≡ (mt - mt ) - φ(yt - yt )

Expression (3.10) defines the economic fundamentals, whose variability will be used as the instrumental variable for the exchange rate volatility. Although the weak empirical evidence of the purchasing-power parity and uncovered interest parity would suggest that the fundamentals are not good instruments for the exchange rates, I will use the Flood and Rose (1999) argument that the fundamentals are relevant for exchange rates at low frequencies or when inflation is high, which is exactly the case of this research.

From previous expressions, the main relationship between the exchange rate and the fundamentals can be obtained as:

(3.11) st = ft + λ(Etst+1 - st)

The forward solution for st gives:

j k+1 1 λ 1 k ≈ λ ’ ≈ λ ’ (3.12) st = ft + Et st+1 = ƒ∆ ÷ Et ft+ j + ∆ ÷ Et st+k+1 1+ λ 1+ λ 1+ λ j=0 «1+ λ ◊ «1+ λ ◊

Equation (3.12) is the solution for the first-order stochastic difference equation of the monetary model. The interpretation for the first two terms says that expectations of future values of the exchange rate are included in the current exchange rate. High relative money growth in the domestic market promotes a depreciation of the domestic currency, while high relative income growth leads to an appreciation of the domestic currency

(Mark, 2001). The last term of the expression disappears under the transversality condition, as k → ∞ and (λ/1+λ) < 1, limiting the rate at which the exchange rate can

143 grow asymptotically. In this case, the exchange rate is the discounted present value of

expected future values of the fundamentals. The correlation between bilateral real

exchange rate and economic fundamentals (Figure 3.4) is positive for most of the FTAA

countries in our data, with the exceptions of Argentina, Chile and Guatemala.

The variability of the fundamentals will be obtained through the moving standard

deviation of the fundamentals, using two time windows of 4 and 8 years. The variables

used to construct the fundamentals were the real money supply88 and real GDP for each

country from IMF’s International Financial Statistics with respect to Brazil’s real money

supply and real GDP.

The second issue is the effect of the bilateral real exchange rate volatility of a

third country on the bilateral trade under analysis. The “third country effect” was

investigated in studies such as Wei (1996), Dell’Ariccia (1999), and Cho et al. (2002),

using a measure that takes into account the exchange rate volatility for all other countries

excluding trade between the two countries under analysis. However, our approach is

slightly different than the one employed in these studies. First, unlike previous studies,

whose measure of third country effect was calculated for total trade and used the total

trade shares of other countries as weights to get the measure of the third country effect

volatility, our proposed measure will be differentiated by sector, accounting for sector

specific trade shares as weights. Second, previous studies considered the trade shares

based on a specific year with the justification that theses shares were relatively constant

over time. Despite the large time and data requirements, we consider sector specific trade

88 The real money supply was obtained through the sum of the “currency money (currency outside banks)” plus the “demand deposits” deflated for the country’s currency, and then converted to 1995 U.S. dollars.

144 shares for each year, based on the fact that some changes might have occurred from one

year to another in the sample, characterizing different responses from trade flows to

exchange rates movements.

The proposed measure of third country real ER volatility did not present a

collinearity problem with the bilateral real ER volatility, as noted by Wei (1996),

Dell’Ariccia (1999), and Cho et al. (2002). In the case of Dell’Ariccia, he found a

correlation above 0.9 between these two measures.

The proposed third country real exchange rate volatility measure (u3ij,t) is given

by:

g g g (3.13) u3ij,t = ƒuij,t wij,t + ƒu ji,t w ji,t i≠ j j≠i

where uij,t (uji,t) is the measure of bilateral real ER volatility, either the moving standard

deviation measure (Sij,t) or the Peree and Steinherr measure (Vij,t), defined by equations

(3.2) and (3.3); g = 1, … , 5, where 1 is for the livestock sector, 2 is for agriculture, 3 is

g g for chemicals, 4 is for manufactured, and 5 is for mining and oil; and the wij,t and wji,t

are the sector specific trade shares of other countries. This measure enters the gravity

equation as another variable. It is expected that the coefficient sign for the third effect

variable will be positive as found by Wei (1996). However, Dell’Ariccia (1999) found it

to be negative and not significant, and Cho et al. (2002) found that the coefficient was

positive and negative for different sectors.

145

0.8 HND USA

0.7 BOL SLV CAN 0.6

0.5 CRI URY PER VEN 0.4 COL n o i t ECU a l 0.3 e r r o MEX C 0.2

PRY 0.1

0 ARG GTM -0.1 CHL -0.2 Country

Figure 3.4: Correlation between bilateral real exchange rate and economic fundamentals between Brazil and other countries, 1989-2002.

3.5. The Gravity model

The theoretical model

The theoretical model of the gravity equation to be used in this study is based on

Deardorff (1998), and the basic assumptions are also the same as those used in Anderson

(1979), Feenstra (2002), and Anderson and Van Wincoop (2003). Considering that all

goods are differentiated by place of origin, each country is specialized in the production

of only one good. Under perfect competition, the supply of each good is fixed, and the

146 price received by exporters of country i is given by pi net of trade costs, or “free on

board” (f.o.b.). There are “iceberg” transport costs with the transport factor between

countries i and j being tij, where the amount (tij –1) “melts” along the away (Samuelson,

1952). Buyers from country j pay pij including the transport cost. Then pij = tijpi.

Assuming identical homothetic preferences, the CES utility function for country j is

defined by:

σ /(σ −1) j ≈ (σ −1) /σ ’ (3.14) U = ∆ƒ β icij ÷ « i ◊

where β is a positive distribution parameter, σ > 0 is the elasticity of substitution between

any pair of countries’ products, and cij is the consumption of any product sent from

country i to country j. The consumers of country j maximizes (3.14) subject to their

income Yj = pjxj from producing xj. Therefore, the demand for each product cij is given

by:

≈ ’1−σ 1 tij pi c Y ∆ ÷ (3.15) ij = j β i ∆ l ÷ tij pi « p j ◊

l Where pj is an overall CES price index of landed prices in country j, defined as:

1/(1−σ ) l ≈ 1−σ 1−σ ’ (3.16) p j = ∆ƒ βitij pi ÷ « i ◊

The f.o.b. value of exports from country i to country j is represented by:

≈ ’1−σ 1 tij pi T fob Y ∆ ÷ (3.17) ij = j β i ∆ l ÷ tij « p j ◊

147 Considering that θi is the country i's share of world income, the general

equilibrium structure of the model imposes market clearance, which implies that:

≈ ’1−σ ≈ ’1−σ Y p x 1 tij pi tij pi i i i p x ∆ ÷ ∆ ÷ (3.18) θ i = w = w = w ƒ β i j j ∆ l ÷ = βi ƒθ j ∆ l ÷ Y Y Y j « p j ◊ j « p j ◊

Solving (3.18) for β, we have89:

Yi 1 (3.19) β i = Y w ≈ ’1−σ ∆ tij pi ÷ ƒθ j ∆ l ÷ j « p j ◊

Combining (3.19) with (3.17), we have:

» 1−σ ÿ ≈ t ’ … ∆ ij ÷ Ÿ … ∆ l ÷ Ÿ fob YiY j 1 « p j ◊ (3.20) Tij = w … 1−σ Ÿ Y tij … ≈ ’ Ÿ ∆ tih ÷ …ƒθ h ∆ l ÷ Ÿ … h « ph ◊ ⁄Ÿ

Deardorff (1998) simplifies (3.20) selecting units of goods so that pi = 1.

l Therefore, pj becomes a CES index of country j’s transport factors as an importer. Define the average distance from suppliers as:

1/(1−σ ) s ≈ 1−σ ’ (3.21) δ j = ∆ƒ β itij ÷ « i ◊

89 The theoretical models of Feenstra (2002), and Anderson and Van Wincoop (2003) are a departure from the Deardorff (1998) approach when they solve for the scaled price βpi, instead of solving for β.

148 For consumers in country j what is important is the transport factor relative to the relative distance from suppliers, given by:

tij (3.22) ρij = s δ j

Using (3.22) and (3.20) we get the following gravity equation:

» ÿ 1−σ fob YiY j 1 … ρij Ÿ (3.23) Tij = … Ÿ Y w t 1−σ ij …ƒθ h ρih Ÿ h ⁄

Equation (3.23) has a more direct interpretation than (3.20), since the trade flows between countries i and j is determined not only by the relative distance between these two countries, but also by the relative distance of all other importers from country j. It is an interesting result that nests a case where the relative distance is the same for all importers, resulting in the frictionless gravity model (without transport costs). The distance between i and j reduces trade. Trade is influenced by the relative distance of these two countries relative to the average of all other importer countries relative distances from i. This result is very similar to those obtained by Anderson (1979) and

Bergstrand (1989).

The econometric specification

The gravity equation, given by equation (3.23), will be estimated by two different econometric specifications90 for both Mercosur and FTAA analysis.

90 It is common practice to include a dummy variable for language as one of the independent variables of the model, and also other dummies to capture other qualitative characteristics of the trade flows. In the case of language, it would equal 1 if the countries share the same language, and zero otherwise. For other examples, see Rose (2000). Frankel (1997) suggests the addition of another variable, a measure of land area, to account for natural resources of the countries.

149 The econometric specification for the Mercosur analysis is given by the following expression:

(3.24)

g g g g g g lnTij ,t = α i + γ 1 ln(YitYjt ) + γ 2 (Popit Pop jt ) + γ 3 (uij ,t ) + γ 4 ln(Dij ) +

g g g g ij ,t γ 5 ln(1+Tariff ) + γ 6 (u3ij ,t ) + ε ij,t

g where Tij,t is the gross bilateral trade between countries i and j in each sector g, YitYjt is the product of the countries GDP in period t, and its coefficient is expected to be positive.

PopitPopjt is the product of countries’ population in period t, which can be thought to reduce trade between countries as population of both countries i and j increases, since the demand for domestic production increases, reducing the amount of goods to be traded; its coefficient is expected to be negative. The variable uij,t is the measure of bilateral real ER volatility, either the moving standard deviation measure (Sij,t) or the Peree and Steinherr measure (Vij,t), defined by equations (4.2) and (4.3), and it is expected to have a negative coefficient. Dij is the distance between countries i and j, which represents a proxy for transportation costs and it should reduce bilateral trade91. Tariff is the simple mean of tariffs within the product category between countries i and j, and it is expected to have a negative coefficient, implying larger trade when there are lower tariffs. The variable u3ij,t

91 Linnemann (1966) pointed out that the effect of distance on trade comes from three sources: 1) transport costs; 2) time (perishability, adaptation to market conditions, irregularities in supply, interest costs); and 3) “psychic” distance, which includes familiarities with laws, institutions, and culture. Linnemann’s idea about the comprehensive meaning of the variable distance is also pointed out by Frankel et al. (1998), who noted that physical shipping costs may not be the most important component of costs associated with distance. Transport costs should be seen as transaction costs, which include not only the cost of physical transportation of goods, but also costs of communications and the fact that countries tend to have a better understanding of their close neighbors and institutions.

150 is the third country real ER volatility (third country effect) for all countries other than countries i and j. Its expected sign is ambiguous, as pointed out by Wei (1996) and Cho et al. (2002).

The proposed FTAA analysis will be performed through the following econometric specification:

(3.25)

g g g g g g ln Tij ,t = α i + γ 1 ln(Yit Y jt ) + γ 2 (Pop it Pop jt ) + γ 3 (u ij ,t ) + γ 4 ln( Dij ) + g g g g g g ij,t γ 5 ln(1+Tariff ) + γ 6 (u3ij,t ) + γ 7 Bij + γ 8 FTAij,t + ε ij,t

where Bij is a border dummy which equals 1 if countries share a common land border and zero otherwise, and its coefficient is expected to be positive. FTAij,t is a dummy variable that represents whether or not the countries are part of a free trade agreement, which equals 1 if a country is member and zero otherwise. If two countries i and j are members of the same free trade area, it is expected to get a positive coefficient for this dummy variable. Other variables are defined as in the Mercosur specification.

Both gravity equations (3.24) and (3.25) will be estimated under two different specifications according to the measure of the exchange rate volatility (uij,t) to be used: the moving standard deviation measure (Sij,t) and/or the Peree and Steinherr measure

(Vij,t).

According to Winters (1997), the FTA dummy can capture the excess trade attributed to the economic bloc agreement, which is a property that has made the gravity models preferred in relation to other econometric-based trade models.

According to Egger (2002), the choice of the econometric set-up is of great relevance for the calculation of bilateral trade flows. Therefore, the estimation procedure

151 in this study will be a panel data econometrics, taking advantage of the panel data available. The cross-section approach could be used mainly to capture the long run real exchange rate variability since this approach treats it as a time-invariant variable. The justification for treating real exchange rate variability as a time-invariant variable comes from empirical evidence of long run PPP (Purchasing Power Parity) which indicates that the real exchange rate is a stationary process for most developed countries (Cho, 2001), which might not be the case for the developing countries that are included in our sample.

If those real exchange rates are non-stationary, their variances are time dependent, then measuring long run real exchange rate variability becomes impossible. This is one of the reasons why there have been so few studies of this type in the literature. In addition to the stationarity problem just mentioned, the cross-section approach eliminates all important time-series information and, due to this, can incorporate small sample bias. The advantages of the panel data approach include more reliable estimates, reduces the multicollinearity problem, increases the degrees of freedom, and allows for the inclusion of real exchange rate volatility in the model, which does not make sense in a cross- section approach92. There are some studies that address the problems of misspecification of gravity models in cross-section approaches, such as Matyas (1997) and Egger (2002).

The study will estimate two different panel data models: fixed and random effects models. The Hausman test will be performed to evaluate which model should be used to represent the model specified in equations (3.24) and (3.25). The complete estimation of

92 Cho (2001) included real exchange rate variable and volatility in his cross-sectional and panel data approaches. However, the inclusion of the real exchange rate in the cross-section estimation makes no sense since it would not provide any information as to whether the currency is undervalued or overvalued.

152 these equations will be possible only under the random effects model, since the fixed effects specification eliminates the coefficients for time-invariant variables, such as distance, and free trade area and border dummies.

The fixed effect (FE) model is basically specified in vector form as:

g g g g (3.26) Tij,t = αi + Xitγ + εit

g where Tij,t is a vector of dependent variables in each sector g (gross bilateral trade

g between countries i and j), Xit represent a vector of all explanatory variables, and εit is

2 the associated vector of disturbances with zero mean and variance σε . Note that there is no cross-sectional subscript in the vector of coefficients γg. In this specification there is an unobservable cross-sectional specific latent effect that can be represented by common border, distance, or common language among countries. According to Greene (1997), this type of model can be viewed as an inter-country comparison, which may well include the full set of countries for which it is reasonable to assume that the model is constant.

This general representation can be specified in two more ways. The first is called

“deviations from the group means” and produces what is called “the within groups estimators”. It is basically a specification of (3.26) without the individual fixed effect parameter and considers the deviations of dependent and independent variables, and disturbances, from their means. The second alternative is called “group means representation”, which produces “the between-groups estimators”. In this specification the fixed effect parameter is not eliminated, but the dependent and independent variables, and disturbances, are replaced by their respective means.

153 The random effect (RE) approach is generally represented by:

g g g g (3.27) Tij,t = α + Xitγ + ηit

g g g g where ηit = νi + εit , and νi is the random disturbance representing the ith observation

g 2 g and is constant through time. νi has zero mean and variance σν . The term α is not only considered constant over time, but also constant among countries in each sector g. Greene

(1997) justifies this approach in case that the individual specific constant terms are viewed as randomly distributed across cross-sectional units. It is appropriate if we believe that sampled cross-section units were drawn from a large population.

3.6. Results and discussion

This section presents the results from the econometric estimations of the gravity equations for both free trade areas: Mercosur and FTAA. The central idea is to capture the effects of medium to long run bilateral real exchange rate volatility on Brazilian sectoral trade, and also to determine the impacts of other important factors that contribute to Brazil’s total trade. The absence of macroeconomic policy coordination among

Mercosur and potential FTAA partners is verified through the exchange rate volatility coefficients estimated using fixed- (random-) effects type of models. This section is twofold: the first section shows the main results for the Mercosur gravity equation estimations; the second section presents the same analysis for the FTAA configuration.

154 All gravity equations were estimated under two specifications of bilateral and third country real ER volatility, the Moving Standard Deviation measure (MSD), and the Perre and Steinherr measure (P&S). The results under many different time periods for such variables and also using instrumental variables specifications are not reported due to space constraints.

3.6.1. The Mercosur analysis

The Mercosur model presents the econometric results for five sectors plus all sectors together. The main results indicate that Brazil’s trade is negatively affected not only by its own exchange rate movements, but also by its Mercosur partners’ exchange rate volatility as well. The population variable was not included in the final estimations due to high correlation with the countries’ income.

The total trade in the agricultural sector is composed of 17 different groups of products, and the main results (Table 3.2) were different for both specifications of the ER measure used in the econometric estimations. According to the Hausman test, the specification using the Moving Standard Deviation measure (MSD) is better represented through a fixed effects model (FE). The Hausman test did not reject the random effects model (RE) for the specification using the Perre and Steinherr measure (P&S) of exchange rate volatility. Therefore, each model specification was estimated differently from the other. A similar test was performed against the hypothesis that the instrumental variable estimators are superior to those from the fixed (random) effects model, and the results were favorable to the latter.

155 The main results show that GDP has an important role in agricultural trade, with a large coefficient (4.63) in the MSD specification (Table 3.2). An increase of 1 % in both countries’ income (Brazil and its partner) improves trade to 4.63 %. Tariffs and the bilateral real ER volatility were significant but larger in size for the MSD specification.

These coefficients affect negatively the bilateral agricultural trade in the Mercosur. Once again, the lack of stable macroeconomic policies can reduce bilateral trade in this free trade area. In the case of the third country effect, only the MSD specification was statistically significant, with a large positive coefficient. Although the expected sign for this coefficient can be ambiguous (Wei, 1996; Cho et al., 2002), it seems to indicate that third country uncertainty increases trade between Brazil and another Mercosur partner in the agricultural sector. Not surprisingly, the random effects results for the P&S specification show an estimated coefficient for distance as negative and statistically significant at the 10 % level93.

93 The use of different time periods for the ER volatility variables in the MSD specification was consistent with the results from Table 3.2. The main changes were with respect to the magnitude of the GDP coefficient and with the statistical significance for the bilateral real ER volatility. Under specific time periods, the GDP coefficient was a lot smaller, but significant, and the bilateral real ER volatility was not significant. The P&S specification was robust with different time windows used to account for the ER volatility, but some of the time periods showed a significant negative third country effect, as opposite to the results from the MSD specification.

156

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (RE) GDP 4.63* 1.84** (6.29) (2.51) Distance - -4.83*** (-1.82) Average tariffs -7.43* -4.97* (-6.49) (-4.12) Real ER volatility -3.22** -0.59** (-2.41) (-2.24) Third country real 12.74* -0.13 ER volatility (5.76) (-0.33) t = 14; n = 676; i (product groups) = 17 Note: All values in parentheses are t- and z-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.2: Fixed and random effects estimations for trade in the agricultural sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

The livestock sector results (Table 3.3) show that the main factors contributing to trade are tariffs and a country’s GDP, the only statistically significant coefficients. Both

ER volatility measures (Sij,t and Vij,t) produced non-significant coefficients. The bilateral real ER volatility proved not to be important for trade in this sector, and this result was robust for many different specifications under different time windows.

157

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 1.02* 1.09* (2.73) (2.95) Average tariffs -5.73** -6.22* (-1.97) (-2.11) Real ER volatility 0.30 0.10 (0.08) (0.24) Third country real 0.49 0.29 ER volatility (0.14) (0.46) t = 14; n = 164; i (product groups) = 4 Note: All values in parentheses are t-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.3: Fixed effects estimations for trade in the livestock sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

Brazilian trade in the livestock sector seems to be very sensitive to changes in tariffs and country’s income94. Results show that a reduction of 1% in tariffs would improve trade in this sector by approximately 6%. The main implication drawn from the results in this sector is that the coordination of macroeconomic policies does not contribute to improved trade among Mercosur countries through reduction on ER volatility among these trade partners.

The results from the chemicals and agricultural sectors were different in terms of magnitude and statistical significance of their estimations, but they were very similar with respect to the Hausman test results. In the MSD specification, the random effects model was rejected in favor of the fixed effects model (Table 3.4). The opposite occurred

94 The results for the livestock sector were also robust when compared to the estimation using instrumental variables (not reported), such as the economic fundamentals, and its moving standard deviation with time periods of 4, and 8 years.

158 with the P&S specification. GDP and bilateral real ER volatility were statistically significant in both specifications. In absolute values, all coefficients were larger in the

MSD specification than in the P&S one. It is curious that the coefficient of tariffs was not significant in the P&S specification, which could lead one to think that this might be caused by almost no variation on the level of tariffs across this sector. But according to the data, there was a large variation on the level of tariffs across products over the years.

In general, the tariffs were reduced for the period 1989 to 1994, increasing after that for most of the 9 products that compose this sector. The P&S specification results show that distance is not important for trade in this sector, and that the third country effect is an important trade obstacle in the Mercosur95. According to the MSD specification, a reduction of 10% in the bilateral ER volatility would increase trade to 10.1%96. Under the

P&S specification, the same reduction in the bilateral ER volatility would improve trade to 21% (Table 3.4).

The largest individual sector analyzed in our study was the manufactured sector, with 23 different categories of products, with a total of 905 observations. The fixed effects model was estimated for the MSD and P&S specifications, since the random effects model was rejected through the Hausman test97.

95 The results for both specifications were consistent under different time periods. However, when using a 6-year time period the negative coefficients for tariffs and third country effects became significant. The Hausman test did not reject the fixed (random) effects coefficients as unbiased and consistent in comparison to the use of instrumental variables. 96 The average bilateral ER volatility used to obtain this interpretation was 0.189 and 1.99, for the MSD and P&S specifications, respectively. 97 The results (Table 3.5) were robust through different combinations of time periods for both specifications of ER volatility used in the estimations. The use of instrumental variables for the ER volatility measures did not improve the main results found here.

159 Exchange rate volatility measure Variable MSD specification (FE) P&S specification (RE) GDP 1.05* 0.80* (4.76) (3.92) Distance - -0.67 (-0.92) Average tariffs -2.61*** 0.93 (-1.90) (0.76) Real ER volatility -5.81* -1.09* (-4.24) (-4.54) Third country real -2.47 -1.32* ER volatility (-1.13) (-3.33) t = 14; n = 361; i (product groups) = 9 Note: All values in parentheses are t- and z-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.4: Random and fixed effects estimations for trade in the chemicals sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

According to Table 3.5, although GDP positively affects bilateral trade in this sector, tariffs, bilateral ER volatility, and third country effects reduce bilateral trade. The results from the MSD specification were larger than those from the P&S specification, with the third country effect being the only exception. The estimates indicate that a 1% reduction of the bilateral ER volatility would increase trade around 1.6% and 1.1%, respectively, under the MSD and the P&S specifications98. The estimated coefficients for bilateral and third country real ER volatility seem to stress the idea that the lack of macroeconomic policy coordination brings adverse effects on manufactured trade among the Mercosur partners.

98 To interpret the impact of the bilateral real ER volatility on total trade as elasticity, it is necessary to use the average of the MSD and P&S bilateral ER volatility measures, whose values are, respectively, 0.207 and 1.97.

160

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 1.94* 1.48* (13.64) (11.37) Average tariffs -4.94* -4.04* (-5.71) (-4.56) Real ER volatility -7.74* -0.56* (-5.36) (-3.34) Third country real -0.40 -1.44* ER volatility (-0.28) (-5.94) t = 14; n = 905; i (product groups) = 23 Note: All values in parentheses are t-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.5: Fixed effects estimations for trade in the manufactured sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

The mining and oil sector results do not respond to changes in tariffs. The tariff coefficient (Table 3.6) was not significant for either specification. GDP was once more very important to explain bilateral trade in this sector. According to both specifications, a

10 % increase in a countries’ income would improve trade between 11 and 16.5 %.

Bilateral and third country real ER volatility coefficients were negatively correlated with bilateral trade. These coefficients were larger in the MSD specification than in the P&S one, as they were in the results from other sectors99.

99 Although not reported, the results for the mining and oil sector were robust under different time periods and, once again, the use of instrumental variables was rejected through the Hausman test.

161

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (RE) GDP 1.65* 1.11* (7.13) (4.74) Distance - -1.86** (-0.91) Average tariffs 1.88 3.02 (0.78) (1.24) Real ER volatility -7.79* -0.89* (-3.00) (-2.65) Third country real -2.76 -1.36* ER volatility (-1.13) (-3.69) t = 14; n = 334; i (product groups) = 10 Note: All values in parentheses are t- and z-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.6: Random and fixed effects estimations for trade in the mining and oil sector between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

The last set of econometric estimations performed in the Mercosur analysis considered all sectors together, which was called “total trade”, and the results (Table 3.7.) show that all estimated coefficients were significant at the 1% level, and they have the expected signs, in both specifications100. The main difference between specifications was the larger magnitude of the estimated coefficients in the MSD specification.

Uncoordinated macroeconomic policies among Mercosur countries seem to be an

100 The results were robust under different time periods for the two ER volatility measures (not reported). The Hausman test was not significant to reject the fixed effects model, and the instrumental variables estimation was not superior to the fixed effect model.

162 obstacle for the total trade among these countries. A 10% reduction in the third country real ER volatility would result in an increase on trade around 4.4% and 20%, respectively for the P&S and the MSD specifications101.

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 1.24* 1.15* (11.29) (13.25) Average tariffs -4.55* -3.02* (-7.43) (-5.01) Real ER volatility -5.88* -0.65* (-7.08) (-5.94) Third country real -2.40* -0.99* ER volatility (-5.69) (-6.62) t = 14; n = 2440; i (product groups) = 63 Note: All values in parentheses are t- and z-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.7: Fixed effects estimations for total trade between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

Table 3.8 summarizes the results obtained using two different specifications for the ER volatility for the Mercosur analysis. The results for the bilateral real exchange rate were very different in terms of magnitude for all sectors in the MSD specification, except for manufactured and mining and oil sectors, which presented a very large coefficient for the bilateral ER volatility. In the livestock sector, the bilateral ER volatility was not

101 The mean values for the third country ER volatility used to obtain this interpretation were 0.185 and 2.02, respectively for the MSD and P&S specifications.

163 important in determining the trade pattern among the Mercosur partners. The bilateral ER volatility was not only very important for the chemicals sector, but also presented the largest coefficient in the P&S specification.

The third country real exchange rate volatility (third country effect), had a very different set of results for both specifications across sectors. Considering the MSD measure of real ER volatility, the third country effect was statistically different than zero only for agriculture, and when all sectors were considered together. In all sectors the third country ER effect was negative. In the agriculture sector, this variable was not only large, but also positive, which shows that the uncertainty in the other Mercosur members contributes to more Brazil trade within Mercosur.

The third country effect under the P&S measure of exchange rate volatility was more stable across sectors, since agriculture and livestock were the only sectors in which this variable was not significant. In all others the coefficient was negative and varied only from -0.99 to -1.44. This means that the exchange rate uncertainty for other members of the Mercosur is very important and contributes to reducing the trade between Brazil and any other partner for chemicals, manufacturing, mining and oil, and for all sectors.

The results for Mercosur show that the use of the two different real exchange rate

volatility measures can produce similar results in terms of signs and interpretation for the

econometric estimations of the gravity equations. The results were ambiguous only when considering the third country effect.

164

Exchange rate volatility measure MSD specification P&S specification Sectors ER volatility Third country ER volatility Third country (Sij,t) ER volatility (Vij,t) ER volatility (S3ij,t) (V3ij,t) Agriculture -3.22** 12.74* -0.59** - Livestock - - - - Chemicals -5.81* - -1.09* -1.32* Manufactured -7.74* - -0.56* -1.44* Mining and oil -7.79* - -0.89* -1.36* Total (all sectors) -5.88* -2.40* -0.65* -0.99* (*) statistically significant at the 1% level; (**) statistically significant at the 5% level.

Table 3.8: Summary of the statistically significant coefficients for the sectoral trade between Brazil and Mercosur partners, 1989 – 2002 by exchange rate volatility measure.

The use of the MSD measure of ER volatility produced not only very different estimates for both bilateral ER volatility and third country effect coefficients in comparison to the estimates from the P&S measure within sectors, but also contrasting results across sectors. The differences in magnitude between estimated coefficients in both specifications were expected since the measures of ER volatility used are very different. Therefore, the conclusions from the Mercosur analysis indicate that the country’s income and level of tariffs are important determinants of trade for the member countries, but volatile exchange rates also adversely affect trade and can be accounted for by the Mercosur countries’ governments. The negative and significant impacts of the

165 bilateral and third country real ER volatility in this common market seem to be a result of the lack of macroeconomic policy coordination among all Mercosur members. The policy implications from the results suggest that, with common and stable implementation of policies to promote macro coordination, it is possible to reduce the impact of exchange rate volatility in the Mercosur trade. The likelihood of political lobbying to increase trade barriers when the import penetration ratio increases would be reduced by a more stable and smooth exchange rate regime.

3.6.2. The FTAA analysis

The FTAA analysis includes 18 countries (Appendix F, Table F.2), with more than 10,000 observations across five sectors and 64 different product categories102. This analysis used the same MSD and P&S specifications for the measure of the long run bilateral and third country (third country effect) real exchange rate volatility as in the

Mercosur analysis. The same robust checking for the real ER volatility measures with different time periods was performed, but not reported due to space restriction.

As in the Mercosur analysis, the Hausman test did not reject the fixed (random) effects model in favor of the instrumental variable model for all sector estimations under both specifications (not reported). This means that in the FTAA analysis, as in the

Mercosur, there is no simultaneity bias or endogeneity problem. This result is not surprising based on our evidence about the lack of macroeconomic policy coordination between Brazil and its 17 main trade partners. In this matter, the central bank decisions in

102 Sectors and product categories are the same as in the Mercosur analysis, see Appendix F.

166 any pair of countries in the FTAA configuration do not take into account the main trade impacts from their decisions. The central banks’ actions seem to be directed only to address the domestic economic issues, but with strong consequences in the terms of trade.

Table 3.9 shows that Brazil’s trade in agriculture is very sensitive to changes in

GDP, tariffs and in the third country ER volatility. The potential FTAA agreement could bring a large increase in agricultural trade between Brazil and the other 17 countries with a reduction in the level of tariffs. For instance, a 10% reduction in the level of tariffs would increase trade about 52%. Although statistically significant, the GDP coefficient was very small. It is interesting to note that the bilateral ER volatility was not very important in explaining the trade pattern in the agricultural sector. However, the third country effect was shown to be important in both specifications of exchange rate volatility measures. This third country coefficient was over three times larger under the

MSD measure. Therefore, through implementation of the FTAA with efficient and lasting macroeconomic policy coordination, agricultural trade among these countries could increase significantly. The use of different time windows and instrumental variables did not change the main findings (Table 3.9).

167

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 0.06* 0.06* (4.82) (4.68) Average tariffs -5.29* -5.21* (-7.65) (-7.80) Real ER volatility -1.37 -0.09 (-1.44) (-0.97) Third country real -1.76* -0.47** ER volatility (-2.59) (-1.84) t = 14; n = 2684; i (product groups) = 17 Note: All values in parentheses are t-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.9: Fixed effects estimations for trade in the agricultural sector between Brazil and 17 Potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

One could expect that trade in the livestock sector would be similar to the agriculture results, but Table 3.10 shows that Brazil’s trade in the livestock sector is very sensitive to changes only in GDP and tariffs. Brazil’s trade in the livestock sector seems to be insensitive to any source of exchange rate volatility (Table 3.9). The GDP variable was statistically significant in both specifications, but the average of tariffs was important to explain trade only in the P&S specification. The GDP estimates suggest that an increase of 1% in GDP for Brazil and any other trade partner would improve trade among these countries between 0.9% and 1.2% in this sector. The results were robust using ER volatility measures with different time periods.

168

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 1.19* 0.93* (4.03) (2.44) Average tariffs -2.07 -2.88*** (-1.27) (-1.77) Real ER volatility 1.28 0.34 (0.53) (1.29) Third country real -1.19 -0.93 ER volatility (-0.41) (-1.55) t = 14; n = 554; i (product groups) = 4 Note: All values in parentheses are t-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.10: Fixed effects estimations for trade in the livestock sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

In the chemicals sector, both specifications show that the level of tariffs and the bilateral ER volatility negatively affect trade in the proposed FTAA103 (Table 3.11).

Tariffs, in particular, play an important role in explaining trade in this sector. A reduction of 1% in the level of tariffs would bring an increase in trade between 7.3% and 9.2%, respectively under the P&S and MSD specifications. Although the weak macroeconomic policy interactions between Brazil and the potential FTAA partners, the reduction in bilateral ER volatility can bring trade improvements in this sector. Third country effects, however, seem to increase bilateral trade for the nine product categories of this sector.

103 The results in the chemicals sector were robust after using different time periods and instrumental variables, in order to verify the main changes in the econometric estimations (not reported).

169 Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 1.02* 1.02* (7.25) (5.89) Average tariffs -9.17* -7.26* (-10.70) (-9.84) Real ER volatility -2.66* -0.45* (-2.61) (-4.40) Third country real 7.03* 0.88* ER volatility (4.16) (2.97) t = 14; n = 1609; i (product groups) = 9 Note: All values in parentheses are t-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.11: Fixed effects estimations for trade in the chemicals sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

The econometric results for the manufactured sector come from a random effects model, since the Hausman test was not significant. Therefore, it was possible to estimate the most complete gravity equation specification for manufacturing, including distance and border dummy (expression 3.25, from section 5). The dummy variable for Mercosur

(FTA) was not included in Table 3.12 due to its high correlation with distance.

The main results from the manufactured sector estimations suggest that the border dummy is not significant to explain manufactured trade in this FTAA configuration. This result is not surprising since, in the 1989 - 2002 period, between 57 and 80% of Brazil’s total manufactured trade has occurred with the United States, which is far away from the

Brazilian border. The GDP coefficients for manufacturing were larger than those estimated in other sectors, reflecting the relatively large income elasticity of exports

(imports) of more value added products. For the first time distance was an important

170 determinant of trade in the proposed FTAA. The coefficient was not only significant, but also relatively large. But the most important variable in explaining trade in this sector was the average level of tariffs, which is a very strong instrument of protection in the

Brazilian economy, and the results just confirm that. The FTAA can imply a large increase in Brazil’s manufactured trade with a substantial reduction in tariffs. To illustrate this, a reduction of 1% in the level of tariffs with the implementation of the

FTAA would increase trade by approximately 12%.

The two specifications of real exchange rate volatility resulted in relatively close estimates for all variables in the manufactured sector. The only exceptions were the coefficients for bilateral and third country real ER volatility, with ambiguous signs and lack of statistical significance. According to the MSD specification, there should not be any major concern about the lack of macroeconomic policies among the FTAA countries, since there is no significant negative impact of ER volatility of any kind on trade.

However, the P&S specification shows that the bilateral ER volatility contributes to reduced trade between Brazil and the potential FTAA partners, which is probably a consequence of an uncoordinated set of policies.

171

Exchange rate volatility measure Variable MSD specification (RE) P&S specification (RE) GDP 1.50* 1.27* (14.96) (10.44) Distance -3.49* -2.55* (-10.17) (-5.79) Border 0.17 0.23 (0.87) (1.18) Average tariffs -12.18* -11.65* (-22.22) (-21.58) Real ER volatility -0.92 -0.32* (-1.26) (-4.42) Third country real 3.80* -0.09 ER volatility (3.40) (-0.49) t = 14; n = 4156; i (product groups) = 24 Note: All values in parentheses are z-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.12: Random effects estimations for trade in the manufactured sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

The results for the mining and oil sector (Table 3.13) show that GDP and tariffs were significant and presented the expected signs. The average level of tariffs seems to be an important obstacle for trade in this sector. Their coefficients are even larger than those from the manufactured sector. With the implementation of the FTAA, there would be a significant trade increase in this sector. The bilateral ER volatility seems to be important only under the P&S specification, with a significant negative coefficient.

According to the P&S specification, a reduction of 10% in the bilateral ER volatility would improve bilateral trade by 14.3%104. For both specifications of ER volatility

104 The average bilateral ER volatility used to obtain this interpretation was 2.564.

172 measures, the third country effect is also another important determinant of trade in this sector; whenever there is an exchange rate uncertainty in other FTAA countries, there is an increase in trade between Brazil and another FTAA partner.

Exchange rate volatility measure Variable MSD specification (FE) P&S specification (FE) GDP 0.89* 0.87* (4.85) (4.11) Average tariffs -14.92* -12.95* (-9.01) (-8.26) Real ER volatility -2.21 -0.56* (-1.54) (-2.63) Third country real 7.76* 0.76** ER volatility (3.20) (2.01) t = 14; n = 1486; i (product groups) = 11 Note: All values in parentheses are t-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.13: Fixed effects estimations for trade in the mining and oil sector between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

When all sectors are considered at the same time, once again the random effects model is superior to the fixed effects one. The results in Table 3.14 emphasize the overall trade in the proposed FTAA, considering all sectors. As the estimated coefficients show, there are few differences between both specifications. GDP, border, distance, and tariffs seem to be important variables in explaining the total trade in the FTAA. The results for both bilateral and third country ER volatility show a significant negative influence of bilateral ER volatility and a positive impact of third country effect on total trade.

173

Exchange rate volatility measure Variable MSD specification (RE) P&S specification (RE) GDP 1.14* 1.06* (17.27) (13.22) Distance -2.45* -2.16* (-10.88) (-7.38) Border 0.35* 0.23*** (2.69) (1.82) Average tariffs -8.83* -8.36* (-23.83) (-23.59) Real ER volatility -1.48* -0.26* (-2.97) (-5.18) Third country real 2.89* 0.24*** ER volatility (3.89) (1.77) t = 14; n = 10489; i (product groups) = 64 Note: All values in parentheses are z-values, (*) statistically significant at the 1 percent level; (**) statistically significant at the 5 percent level; (***) statistically significant at the 10 percent level.

Table 3.14: Random effects estimations for total trade between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

The FTAA results discussed so far were reasonable in the sense that there were no inconsistencies across different specifications, showing that countries’ income level

(GDP), level of tariffs and, in some sectors, distance and border are important determinants of trade. However, there were strong differences in the estimations of the influence of the bilateral and third country ER volatility on trade. According to Table

3.15, if we consider the use of the MSD measure of ER volatility, there is an influence of bilateral ER volatility only on the chemicals sector and on all sectors together. The third country effect is the only ER uncertainty that affects trade in the FTAA on all sectors

174 (except in the livestock sector). In the case of the P&S specification, there is not only the influence of the third country ER volatility, but also the bilateral ER volatility on trade across sectors.

The use of the P&S measure of real ER volatility seems to make the most sense because the agents’ uncertainty is based on an accumulation of past experience plus the level of misalignment of the exchange rate. The use of the MSD measure of ER volatility can be a good option to capture the influence of the third country effect. It might not be so effective to capture the influence of the bilateral ER volatility, as noted by

Dell’Ariccia (2000) and Cho et al. (2002). Another drawback for the MSD specification was the fact that when the gravity equations were estimated without the presence of the third country effect, the bilateral ER volatility coefficient became significant and negative.

Even though the FTAA is still in negotiation among the countries’ authorities, our study addressed the main trade determinants of this possible free trade agreement in the near future through an ex-ante econometric analysis, which used trade information and other economic variables for the period 1989 to 2002. The pattern of trade in the proposed FTAA found GDP and the level of tariffs to be important variables. The main findings also suggest that the bilateral real exchange rate volatility has a negative effect on trade among these countries, which would suggest a need for coordinated macroeconomic policies across FTAA countries. However, the third country real exchange rate volatility seems to increase trade, without bringing any threat to trade for the FTAA. This means that the presence of this type of exchange rate volatility does not represent an obstacle to trade among FTAA countries.

175

Exchange rate volatility measure MSD specification P&S specification Sectors ER volatility Third country ER volatility Third country (Sij,t) ER volatility (Vij,t) ER volatility (S3ij,t) (V3ij,t) Agriculture - -1.76* - -0.47** Livestock - - - - Chemicals -2.66* 7.03* -0.45* 0.88* Manufactured - 3.80* -0.32* - Mining and oil - 7.76* -0.56* 0.76** Total (all sectors) -1.48* 2.89* -0.26* 0.24*** (*) statistically significant at the 1% level; (**) statistically significant at the 5% level; (***) statistically significant at the 5% level.

Table 3.15: Summary of the statistically significant coefficients for the sectoral trade between Brazil and 17 potential FTAA partners, 1989 – 2002 by exchange rate volatility measure.

3.7. Conclusions and implications

Mercosur

Brazil has been negotiating with Mercosur partners the fate of the agreements to implement the common external tariffs by 2006, and also to improve their trade flows with stable and lasting multilateral actions in favor of a more integrated free trade area.

The economic crisis in recent years brought new obstacles to their trade policies within

Mercosur. The uncoordinated domestic macroeconomic policies, adopted mainly by

Argentina and Brazil, have been a threat to the future of this economic bloc.

176 The main results indicate that Brazil’s trade is negatively affected not only by its own exchange rate movements, but also by its Mercosur partner’s exchange rate volatility. The impacts of exchange rate volatility varied across sectors, but there was little evidence of the reasons for such responses being due to the Baldwin and Krugman’s hysteresis model, since the results have shown that sectors with presumably large sunk costs were relatively sensitive to bilateral ER volatility. The mining and oil and manufactured sectors were the ones that presented the highest coefficient for the bilateral real exchange rate volatility, which following the hysteresis model would predict that these sectors should have a small sunk cost, which is exactly the opposite in these sectors.

The results for the bilateral real exchange rate were very similar in terms of sign and magnitude for all sectors, except for livestock, mining and oil, and manufactured sectors. In the livestock sector, this variable was not important in determining the trade pattern among the Mercosur partners. The third country real exchange rate volatility

(third country effect), had a very different set of results for both specifications across sectors. Considering the moving standard deviation as a measure of real exchange rate volatility, the third country effect was statistically different from zero only for agriculture, and for all sectors combined. In agriculture, this variable was not only large, but also surprisingly positive, which shows that the uncertainty in the other Mercosur members contributes to more Brazil trade within Mercosur. The third country effect under the second measure of exchange rate volatility, Perre and Steinherr, was more stable across sectors, but in the agriculture and livestock sectors this variable was not significant. In the other sectors, this coefficient was negative and varied only from -0.99 to -1.44.

177 The Mercosur analysis indicates that a lack of macroeconomic policy coordination between Brazil and its trade partners, together with the role of tariffs and a country’s GDP, were the main empirical findings of this study. The fact that Argentina and Brazil, the two main partners of the Mercosur, have pursued different and divergent macroeconomic policies for many years was analyzed in this study through the impact of bilateral and third country real exchange rate volatilities. The results suggest that these disharmonized policies cause substantial price and exchange rates movements, which bring negative impacts on bilateral trade due to behavior of risk averse economic agents, and due to the overall protectionism caused by them. The policy implications from our results suggest that, with a common and stable implementation of policies to promote macro coordination, it is possible to reduce the secondary impact of the exchange rate volatility in the Mercosur trade, since political lobbying to increase barriers when the import penetration ratio increases would be reduced due to a more stable and smooth exchange rate regime. These findings can also be seen in the ex-ante analysis of the proposed FTAA, emphasizing the importance of the main work that is occurring to implement this free trade area in 2005.

In the case of the Mercosur, the harmonization of macroeconomic and exchange rate policies seem to favor a monetary union among these countries. However, as

Eichengreen (1998) says, Mercosur countries do not achieve some important conditions needed to have a monetary union. These conditions include an independent central bank that is not vulnerable to “political business cycle”, wage and price flexibility, a strong financial sector, and the presence of barriers of exit from Mercosur. Eichengreen’s view shows that a monetary union is not impossible, but it would require some time for these

178 countries to achieve those conditions. For instance, these countries already created some conditions for their central banks to operate politically and economically independently, but stronger financial systems and enhanced labor market flexibility need more time to be entirely achieved.

While the main conditions for a monetary union are far away from Mercosur countries, the search for short run solutions for the problem of uncoordinated macroeconomic policies must continue, in order to have strong and balanced trade integration in the long run.

FTAA

The future of the proposed FTAA has also been discussed due to disagreements among its potential members. The most sensitive and important issue to be solved seems to be agriculture subsidies in production and in exports, but there are other differences to be negotiated, such as market access, creation of environmental standards, government purchases, and property rights (Bouzas, 2001).

The procedure for the FTAA analysis was the same as for the Mercosur specification105 with the exception that the model included some additional variables. The

Hausman test did not reject the fixed (random) effects model in favor of the instrumental variable model for all sector estimations under the MSD and P&S specifications. This means that in the FTAA setup, as in the Mercosur, there is no simultaneity bias or endogeneity problem. This result is not surprising based on our beliefs about the lack of macroeconomic policy coordination between Brazil and its 17 main trade partners.

105 The main differences in the specification are when the random effects model is estimated. See section 3.5 for more details.

179 Central bank decisions in any pair of countries in the FTAA configuration seem to be independent of the main trade impacts followed from their decisions. The central bank actions seem to be directed only to address the domestic economic issues.

The sectoral results once again did not confirm the Baldwin and Krugman’s hysteresis model. For the agriculture sector, for example, the implementation of the

FTAA with efficient and lasting macroeconomic policy coordination would increase trade significantly. In the chemicals sector, the level of tariffs and the bilateral exchange rate volatility negatively affect trade in the proposed FTAA. Tariffs, in particular, play an important role explaining trade in this sector. Despite the absence of macroeconomic policy interactions between Brazil and the potential FTAA partners, a reduction on bilateral exchange rate volatility can bring trade improvements in this sector. The econometric results from the manufactured sector show that GDP coefficients (two specifications) were larger in the manufactured sector than those estimated in other sectors, reflecting the relatively large income elasticity of exports (imports) of more value added products. But the most important variable affecting trade in this sector was the average level of tariffs, which is a very strong instrument of protection in the Brazilian economy. The FTAA results imply a large expansion in Brazil’s manufactured trade with a substantial reduction in tariffs. The two specifications of real exchange rate volatility resulted in relatively similar estimates for all variables in the manufactured sector. The only exceptions were the coefficients for bilateral and third country real ER volatility.

The former was negative and significant only under the MSD specification. The latter was positive and significant only under the S&P specification.

180 The proposed FTAA was analyzed through different specifications of the real exchange rate volatility for different sectors. The random effects model proved to be superior when all sectors are considered at the same time. The results for both bilateral and third country ER volatility show the significant negative influence of bilateral ER volatility and the positive impact of the third country effect on total trade. The results discussed so far were reasonable in the sense that there were no inconsistencies across different specifications, showing that countries’ income level (GDP), level of tariffs and, in some sectors, border and distance are important determinants of trade in the expected future largest free trade area in the world. However, there were strong differences in the estimations of both specifications, when accounting for the influence of the bilateral and third country exchange rate volatilities on trade. The use of the P&S measure of real exchange rate volatility seems to make more sense because the agents’ uncertainty is based on an accumulation of past experience plus the level of misalignment of the exchange rate. The use of the MSD measure of ER volatility can be a good option to capture the influence of the third country effect, and it might not be so effective to capture the influence of the bilateral ER volatility, as noted by Dell’Ariccia (2000) and

Cho et al. (2002).

The main findings suggest that the bilateral real exchange rate volatility has a negative effect on trade among these countries, which would suggest a need for coordinated macroeconomic policies across FTAA countries. However, the third country real exchange rate volatility seems to increase trade, without bringing any threat to trade for the FTAA.

181 In some ways the results obtained in this study reflect the findings of other empirical studies. The negative impact of the bilateral exchange rate volatility on trade, and the experience with the exchange rate volatility has been different across countries

(Kenen and Rodrik, 1986), which in our case was also different across sectors (Cho et al,

2002). The lack of macroeconomic policy coordination expressed by the bilateral exchange rate volatility is one of the main obstacles to Brazil’s trade within the proposed

FTAA (Eichengreen, 1998; Baer et al, 2001).

For the proposed FTAA, however, there is a need to achieve a general agreement among the main countries in the negotiations of the challenging topics for this free trade area. Some important information to help the discussion agenda was investigated in this study. The lack of macroeconomic policy coordination was one of them. Other findings that could serve for the negotiations were the importance of tariffs and a country’s level of income.

Further research should look at more disaggregated data, and some other proxies as measures of the exchange rate uncertainty, since some of the sectoral responses found in our empirical analysis were ambiguous in terms of sign and magnitude. The search for a better instrumental variable for the exchange rate volatility measure to test the presence of the simultaneity bias should also be included in future studies.

182

APPENDIX A

BRAZILIAN SOCIAL ACCOUNTING MATRIX (SAM) (Cattaneo, 1999), 1995-96 AGGREGATED VERSION

183

l ; ; s a d r l e

r s u ) i t o l r $ u w p

r c R i e e

r t i h t g n b

a e

f -

5 o n =

t 9

o s 9 n T e

r 1 N

= (

E =

n D ; s o W N i d I l s O r o C R

h e

; ; e v t y s

t n i u d e d o e o h m t t

a m s n e g a m v e b o r r n c i u g -

l s g a = g r a

u i B t 6 l v R 9 a u - s c U

i 5 r = H

9 g

I ; a 9 - s

S 1 d =

l ;

, o s ) R h e 9 e x G s 9 a t A u 9

t o C 1 c h

; ,

e l r y o i a t i r e d

v u n i r = t a

c t = t a X

a l A R a C T r U ( u

Y t R

) l ; u H s

f c M ; i f i r d r A g n a a S t a -

l (

n =

= x o i

n R r

D t A = a N

L ; m D s

N e g I

x ; n a l A i t

t a

; t t i n y c t p u e i a r o v i c i

c d t = c c n

i a a

l = l

A a a r

i C X . u

c t ; e A l r o g u o T s n c

b i a A a r

n h l ;

g t a c

i a = n l

y e i r = B z o

m t a A n R n r r L e

e G ; B v v

y A n : o t i i

1 A g . d

= :

o = A e

r K m e e V T l m h O S b o

a W t c G D

T t

184

APPENDIX B

CROSS-ENTROPY EQUATIONS

185

Equation Description » ÿ (B.1) min I = …ƒƒ Ai, j ln Ai, j −ƒƒ Ai, j ln AŸ {A,W1 ,W2 } i j i j ⁄ » ÿ cross-entropy minimand + …ƒƒW1i, jwt lnW1i, jwt −ƒƒW1i, jwt lnW 1i, jwt Ÿ i jwt i jwt ⁄ » ÿ + …ƒƒW 2k, jwt lnW 2k, jwt −ƒƒW 2k, jwt lnW 2k, jwt Ÿ k jwt k jwt ⁄ SAM equation (B.2) Ti, j = Ai, j .(X i + e1i ) row/column sum consistency (B.3) Yi = X i + e1i error definition (B.4) e1i = ƒW1i, jwt .v1i, jwt jwt sum of weight on errors (B.5) ƒW1i, jwt = 1 jwt row sum (B.6) ƒTi, j = Yi j column sum (B.7) ƒTi, j = X i + e1i i sum of column coefficients (B.8) ƒ Ai, j = 1 and 0 < Ai, j < 1 i sum of weights on errors (B.9) ƒWi,w = 1 and 0 < Wi,w < 1 w (k ) (k ) additional constraints (B.10) ƒƒGi, j Ti, j =γ + e2k i j

Notation

Set Parameters i and j SAM accounts Ai, j prior SAM coefficient matrix w weights on error support set (k ) Gi, j kth aggregator matrix (k) γ kth control total Variables n number of elements in set w

Ai,j SAM coefficient matrix vi, jwt error support values and bounds e error variable i X i fixed value of column sum I Cross-entropy measure Ti,j transactions SAM Wi,w error weight Yi row sum

186

APPENDIX C

THE MODIFIED STANDARD CGE MODEL (Lofgren et al., 2001)

187

Sets a ∈ A activities (agricultural and non-agricultural) a ∈ ACES(⊂C) activities w/CES function at the top of the technology nest a ∈ ALEO(⊂C) activities w/Leontief function at the top of the techn. nest c ∈ C commodities (agricultural and non-agricultural) c ∈ CD (⊂C) commodities with domestic sales of domestic output c ∈ CDN (⊂C) commodities not in CD c ∈ CE(⊂C) exported commodities c ∈ CEN(⊂C) commodities not in CE c ∈ CM (⊂C) imported commodities c ∈ CMN (⊂C) commodities not in CM c ∈ CT(⊂C) transaction service commodities c ∈ CX(⊂C) commodities with domestic production f ∈ F factors of production (capital, labor, and land) i ∈ INS institutions (domestic and ROW) i ∈ INSD(⊂INS) domestic institutions i ∈ INSDNG(⊂INSD) domestic non-government institutions h ∈ H (⊂INSDNG) households (rural and urban) r ∈ R regions (NO, NE, CW, and S-SE)

Parameters

a αa efficiency parameter in the CES activity function va αa efficiency parameter in the CES value added function ac αc shift parameter for domestic commodity aggregation function q αc shift parameter for the Armington function t αc shift parameter for the CET function h βach marginal share of consumption spending on home commodity c from activity a for household h

188

m βch marginal share of consumption spending on marketed commodity c for household h a δa CES activity function share parameter ac δac share parameter for domestic commodity aggregation function q δc Armington function share parameter t δc CET function share parameter va δfa CES value added function share parameter for factor f in activity a m γch subsistence consumption of marketed commodity c for household h h γach subsistence consumption of home commodity c from activity a for household h

θac yield of output c per unit of activity a

θac,r yield of output c per unit of activity a in region r a ρa CES production function exponent va ρa CES value added function exponent ac ρc domestic commodity aggregation function exponent q ρc Armington function exponent t ρc CET function exponent cwtsc weight of commodity c in the consumer price index dwtsc weight of commodity c in the producer price index pwec export price

pwmc import price qdstc quantity of stock change

qg c base-year quantity of government demand

qinv c base-year quantity of private investment demand

shifif share for domestic institution i in the income from f shiiii’ share of net income of i’ to i (i’ ∈ INSDNG’; i ∈ INSDNG) taa tax rate for activity a tec export tax rate

tff direct tax rate for factor f

189

tins c exogenous direct tax rate for domestic institution i tins01i 0-1 parameter with 1 for institutions with potentially flexed direct tax rates

tmc import tariff rate

tqc rate of sales tax trnsfrif transfer from factor f to institution i trnsfrif,r transfer from factor f to institution i in region r

tvaa rate of value added tax for activity a icaca amount of c used as intermediate input per unit of final output in a

tvaa,r rate of value added tax for activity a in region r icaca,r c used as intermediate input per unit of final output in a in region r icdcc’ input per unit of c’ produced and sold domestically

icecc’ amount of c as trade input per exported unit of c’

icmcc’ amount of c as trade input per imported unit of c’ intaa amount of aggregate intermediate input per activity unit ivaa amount of aggregate value added input per activity unit intaa,r amount of aggregate intermediate input per activity unit in region r ivaa,r amount of aggregate value added input per activity unit in region r

mpsi base savings rate for domestic institution i mps01i 0-1 parameter with 1 for institutions with potentially flexed direct tax rates

Variables CPI consumer price index DTINS change in domestic institution tax share (for base year = 0) FSAV foreign savings

GADJ government consumption adjustment factor IADJ investment adjustment factor MPSADJ savings rate scaling factor (for base = 0)

QFS f quantity supplied of factor

QFS f ,r quantity supplied of factor in region r

190

TINSADJ direct tax scaling factor (for base = 0)

WFDIST fa wage distortion factor for factor f in activity a

WFDIST fa ,r wage distortion factor for factor f in activity a in region r DMPS change in domestic institution savings rates (for base = 0) DPI producer price index for domestically marketed output EG government expenditures

EHh consumption spending for household h EXR foreign exchange rate GOVSHR government consumption share in nominal absorption GSAV government savings INVSHR investment share in nominal absorption

MPSi marginal propensity to save for domestic non-government institution

PAa price of activity a

PAa,r price of activity a in region a

PDDc demand price for commodity produced and sold domestically

PDSc supply price for commodity produced and sold domestically

PEc export price

PINTAa aggregate intermediate input price for activity a

PINTAa,r aggregate intermediate input price for activity a in region r

PMc import price

PQc composite commodity price

PXc producer price

PVAa value added price of a

PVAa,r value added price of a in region r

PXACac producer price of commodity c for activity a

PXACac,r producer price of commodity c for activity a in region r

QDc quantity of domestic output sold domestically

QEc quantity of exports

QMc quantity of imports

QAa level of activity a

191

QAa,r level of activity a in region r

QFfa demand for factor f from activity a

QFfa,r demand for factor f from activity a in region r

QGc government consumption demand for c

QHch consumption of c by household h

QHAach household home consumption of c from activity a by household h

QINTAa quantity of aggregate intermediate input

QINTAa,r quantity of aggregate intermediate input in region r

QINTca quantity of commodity c as intermediate input to activity a

QINTca,r quantity of commodity c as intermediate input to activity a in region r

QINVc quantity of investment demand for commodity c

QQc quantity supplied of composite good

QTc quantity of commodity demanded as trade input

QVAa quantity of aggregate value added

QVAa,r quantity of aggregate value added in region r

QXc quantity of aggregate domestic output

QXACac quantity of output of commodity c from activity a

QXACac,r quantity of output of commodity c from activity a in region r TABS total nominal absorption

TINSi direct tax rate for institution i (i ∈ INSDNG)

TRIIii’ transfers from institution i’ to i (both ∈ INSDNG)

WFf average price of factor f

WFf,r average price of factor f in region r

YFf income of factor f

YFf,r income of factor f in region r YG government revenue

YIi income of domestic non-government institution

YIFif income to domestic institution i from factor f

YIFif,r income to domestic institution i from factor f in region r

192

Equations

Prices Block

c ∈ CM (Import Price) (C.1) PM c = (1+ tmc ).EXR.pwmc + ƒ PQc'icmc'c c'∈CT c ∈ CE (Export Price) (C.2) PEc = (1 − tec ).EXR.pwec + ƒ PQc'icec'c c'∈CT c ∈ CD (Demand Price of (C.3) PDDC = PDS C + ƒ PQc'icd c'c c'∈CT Domestic Non- traded Goods) (Absorption) (C.4) PQc .(1 − tqc ).QQc = PDDc .QDc + PM c .QM c c ∈ (CD∪CM) (Marketed Output (C.5) PX c .QX c = PDS c .QDc + PEc .QEc c ∈ CX Value) a ∈ A, r ∈ R (Activity Price) (C.6) PAa = ƒ PAa,r r∈R a ∈ A, r ∈ R (Aggregate (C.7) PINTAa = ƒ PINTAa,r r∈R Intermediate Input Price) (C.8) a ∈ A (Activity Revenues and Costs) PAa.(1− taa ).QAa = PVAa .QVAa + PINTAa .QINTAa

r r ∈ R (Aggregate (C.9) CPI = ƒƒ PQc .cwtsc r∈R c∈C Consumer Price Index) r r ∈ R (Producer Price (C.10) DPI = ƒƒ PDSc .dwtsc r∈R c∈C Index for Non- traded Market Output) a ∈ A, r ∈ R (Aggregate (C.11) PXACac = ƒ PXACac,r r∈R Activity Producer Price)

193

Production and Commodity Block a ∈ ACES, (Aggregate CES (C.12) QAa = ƒQAa,r r∈R Activity Production r ∈ R Function) 1 a ∈ ACES (CES Value added- a a QVA ≈ PINTA δ ’1+ ρa Intermediate-Input (C.13) a = ∆ a a ÷ ∆ a ÷ Ratio) QINTAa « PVAa 1− δ a ◊ a ∈ ALEO (Aggregate Demand (C.14) QVAa = ƒQVAa,r r∈R for Value added) a ∈ ALEO (Demand Aggregate (C.15) QINTAa = ƒQINTAa,r r∈R for Intermediate Input) 1 a ∈ A (Value added and vq ≈ va ’ ρ a va ∆ va − ρ a ÷ Factor Demands) (C.16) QVAa = α a .∆ ƒδ fa .QFfa ÷ « f ∈F ◊ f ∈ F, a ∈ A (Factor Demand) (C.17) W f .WFDIST fa = −1 ≈ va ’ va ∆ va −ρa ÷ va −ρa −1 PVAa .(1− tvaa ).QVAa .∆ ƒδ fa .QFfa ÷ .δ fa .QFfa « f ∈F ' ◊ a ∈ A, r ∈ R (Aggregate (C.18) QINTa = ƒQINTa,r r∈R Intermediate Input Demand) c ∈ CX; a ∈ (Aggregate (C.19) QXACac = ƒQXACac,r r∈R A; r ∈ R Commodity Production and Allocation) 1 c ∈ CX (Output Aggregation ac ≈ ac ’ ρ c −1 ac ac − ρc Function) (C.20) QX c = αc .∆ ƒδ ac .QXACac ÷ « a∈A ◊

1 t t c ∈ (CET Function) t t ρ c t ρc t ρc (C.21) QX c = αc.(δ c .QEc + (1− δ c ).QDc ) (CE∩CD) 1 c ∈ (Export-Domestic ≈ t ’ ρ t −1 QEc ∆ PEc 1− δ c ÷ c (CE∩CD) Supply Ratio) (C.22) = ∆ . t ÷ QDc « PDSc δ c ◊ c ∈ (Output (C.23) QX c = QDc + QEc (CE∩CEN) Transformation for ∪ Nonexported (CE∪CDN) Commodities)

194

−1 q q c ∈ (Armington Function) q q − ρc q − ρc q ρc (C.24) QQc = αc .(δ c .QM c + (1− δ c ).QDc ) (CM∩CD) 1 c ∈ (Import-Domestic ≈ q ’1+ ρ q QM c ∆ PDDc δ c ÷ c (CM∩CD) Demand Ratio) (C.25) = ∆ . q ÷ QDc « PM c 1− δ c ◊ c ∈ (Composite Supply for (C.26) QQc = QDc + QM c (CD∩CMN) Nonimported ∪ Commodities) (CM∩CDN) (C.27) c ∈ CT (Demand for QTc = ƒ(icmcc'.QM c'+ icecc'.QEc'+ icdcc'.QDc') Transactions Services) c'∈C' (Aggregate Average Price of Factors) (C.28) W f = ƒW f ,r r∈R (Aggregate Demand for Factors) (C.29) QFfa = ƒQFfa,r r∈R (Aggregate Wage Distortion Factor) (C.30) WFDIST fa = ƒWFDIST fa,r r∈R Institution Block f ∈ F; r ∈ R (Aggregate Factor (C.31) YFf = ƒYFf ,r r∈R Income) f ∈ F; r ∈ R (Aggregate (C.32) YIFf = ƒYIFf ,r r∈R Institutional Factor Incomes) (C.33) i ∈ INSDNG (Income of Domestic, Non-government YIi = ƒYIFif + ƒTRIIii'+trnsfrigov .CPI + trnsfrirow. f ∈F i'∈INSDNG' Institutions)

i ∈ INSDNG; (Intra-Institutional (C.34) TRIIii'= shiiii'.(1− MPSi').(1− TINSi').YIi' i' ∈ Transfers) INSDNG’ (C.35) h ∈ H (Household ≈ ’ Consumption EH h = ∆1− ƒ shiiih ÷.(1− MPSh ).(1− TINSh ).YIh Expenditure) « i∈INSDNG ◊ (Household (C.36) QH ch = γ ch + c ∈ C, h ∈ H, Consumption Demand ≈ ’ r ∈ R β m.∆ EH − ƒ PQ .γ m − ƒƒƒ PXAC .γ h ÷ for Marketed ch « h c' c'h ac',r ac'h ◊ Commodities) c'∈C r∈R a∈A c'∈C PQc

195

h a ∈ A; c ∈ C, (Household (C.37) QHAach = γ ach + h ∈ H, r ∈ R Consumption Demand ≈ ’ for Home h ∆ m h ÷ βach. EH h − ƒ PQc'.γ c'h − ƒƒƒ PXACac',r .γ ac'h Commodities) « ◊ c'∈C r∈R a∈A c'∈C PXACac c ∈ CINV (Invested Demand) (C.38) QINVc = qinvc.IADJ

(C.39) QG = qg .GADJ c ∈ C (Government c c Consumption Demand) (Government (C.40) YG = ƒTINSi .YIi + EXR.trnsfrgovrow + ƒtf f .YFf + i∈INSDNG f ∈F Revenue)

ƒtqc .PQc .QQc + ƒtvaa .PVAa .QVAa + c∈C a∈A ƒtaa .PAa .QAa + ƒ tmc .EXR.pwmc .QM c + ƒtec .EXR.pwec .QEc + ƒYFgovf a∈A c∈CM c∈CE f ∈F (Government Expenditures) (C.41) EG = ƒtrnsfrigov.CPI + ƒ PQc.QGc i∈INSDNG c∈C r ∈ R (Aggregate Transfers (C.42) trnsfrrow, f = ƒtrnsfrrowf ,r r∈R from Factors to ROW)

Constraint Block f ∈ F; r ∈ R (Aggregate Factor (C.43) QFS f = ƒQFS f ,r r∈R Market Equilibrium)

(C.44) c ∈ C (Composite Commodity QQc = ƒQH ch + ƒQINTca + QINVc + qdstc + QTc h∈H a∈A Equilibrium)

(C.45) (Current Account Balance for ROW) ƒ pwec .QEc + ƒtrnsfri,row + FSAV = ƒ pwmc .QM c + ƒtrnsfrrow, c∈C i∈INSD c∈CM f ∈F

(C.46) YG = EG + GSAV (Government Balance) (C.47) i ∈ INSDNG (Direct Institutional Tax Rates) TINSi = tinsi .(1+ TINSADJ.tins01i )+ DTINS.tins01i

(C.48) i ∈ INSDNG (Institutional Savings Rates) MPSi = mpsi .(1+ MPSADJ.mps01i )+ DMPS.mps01i

196

(Savings-Investment (C.49) ƒ MPSi .(1− TINSi ).YIi + GSAV +EXR.FSAV = i∈INSDNG Balance)

ƒ PQc .QINVc + ƒ PQc .qdstc c∈C c∈C (Total Absorption) (C.50) TABS = ƒƒ PQc .QH ch + ƒƒ ƒ PXACac .QHAach h∈H c∈C a∈A c∈C h∈H

+ ƒ PQc .QGc + ƒ PQc .QINVc + ƒ PQc .qdstc c∈C c∈C c∈C (C.51) (Ratio of Investment to Absorption)

INVSHR.TABS = ƒ PQc .QINVc + ƒ PQc.qdstc c∈C c∈C (Ratio of Government Consumption to (C.52) GOVSHR.TABS = ƒ PQc .QGc c∈C Absorption)

197

APPENDIX D

MAIN DISAGGREGATED SAM COMPONENTS

198

Abbreviation Meaning Abbreviation Meaning

smallholder annuals in large farm other agric aannl1s Amazon aotrag3l products in Center-West smallholder annuals in large farm other agric aannl2s Northeast aotrag4l products in South and SE smallholder annuals in aannl3s Center-West alogdef1 forest products in Amazon smallholder annuals in aannl4s South and SE alogdef2 forest products in Northeast large farm annuals in forest products in Center- aannl1l Amazon alogdef3 West large farm annuals in forest products in South and aannl2l Northeast alogdef4 SE large farm annuals in aannl3l Center-West aprfood1 food processing in Amazon large farm annuals in South aannl4l and SE aprfood2 food processing in Northeast smallholder perennials in food processing in Center- aperen1s Amazon aprfood3 West smallholder perennials in food processing in South aperen2s Northeast aprfood4 and SE smallholder perennials in aperen3s Center-West aminpet1 mining and oil in Amazon smallholder perennials in aperen4s South and Se aminpet2 mining and oil in Northeast large farm perennials in mining and oil in Center- aperen1l Amazon aminpet3 West large farm perennials in mining and oil in South and aperen2l Northeast aminpet4 SE large farm perennials in aperen3l Center-West Aindust1 industry in Amazon large farm perennials in aperen4l South and Se Aindust2 industry in Northeast Smallholder Livestock In alivst1s Amazon Aindust3 industry in Center-West smallholder livestock in alivst2s Northeast Aindust4 industry in South and SE

Continued

Table D.1: Description of the main activities in the disaggregated SAM

199

Table D.1 continued

smallholder livestock in alivst3s Center-West aconst1 construction in Amazon smallholder livestock in alivst4s South and SE aconst2 construction in Northeast large farm livestock in alivst1l Amazon aconst3 construction in Center-West large farm livestock in construction in South and alivst2l Northeast aconst4 SE large farm livestock in trade and transportation in alivst3l Center-West atrantd1 Amazon large farm livestock in trade and transportation in alivst4l South and SE atrantd2 Northeast smallholder other agric trade and transportation in aotrag1s products in Amazon atrantd3 Center-West smallholder other agric trade and transportation in aotrag2s products in Northeast atrantd4 South and SE smallholder other agric aotrag3s products in Center-West asvc1 services in Amazon smallholder other agric aotrag4s products in South and SE asvc2 services in Northeast large farm other agric aotrag1l products in Amazon asvc3 services in Center-West large farm other agric aotrag2l products in Northeast asvc4 services in South and SE

200

Abbreviation Meaning Abbreviation Meaning urban skilled food processing in urban unskilled light industry urbsfd1 Amazon urbult1 in Amazon urban skilled food processing in urban unskilled light industry urbsfd2 Northeast urbult2 in Northeast urban skilled food processing in urban unskilled light industry urbsfd3 Center-West urbult3 in Center-West urban skilled food processing in urban unskilled light industry urbsfd4 South and SE urbult4 in South and SE urban unskilled food processing urban skilled construction in urbufd1 in Amazon urbscn1 Amazon urban unskilled food processing urban skilled construction in urbufd2 in Northeast urbscn2 Northeast urban unskilled food processing urban skilled construction in urbufd3 in Center-West urbscn3 Center-West urban unskilled food processing urban skilled construction in urbufd4 in South and SE urbscn4 South and SE urban skilled heavy industry in urban unskilled construction in urbshv1 Amazon urbucn1 Amazon urban skilled heavy industry in urban unskilled construction in urbshv2 Northeast urbucn2 Northeast urban skilled heavy industry in urban unskilled construction in urbshv3 Center-West urbucn3 Center-West urban skilled heavy industry in urban unskilled construction in urbshv4 South and SE urbucn4 South and SE urban unskilled heavy industry urban skilled services in urbuhv1 in Amazon urbssv1 Amazon urban unskilled heavy industry urban skilled services in urbuhv2 in Northeast urbssv2 Northeast urban unskilled heavy industry urban skilled services in urbuhv3 in Center-West urbssv3 Center-West urban unskilled heavy industry urban skilled services in South urbuhv4 in South and SE urbssv4 and SE urban skilled light industry in urban unskilled services in urbslt1 Amazon urbusv1 Amazon urban skilled light industry in urban unskilled services in urbslt2 Northeast urbusv2 Northeast urban skilled light industry in urban unskilled services in urbslt3 Center-West urbusv3 Center-West urban skilled light industry in urban unskilled services in urbslt4 South and SE urbusv4 South and SE

Table D.2: Description of the main types of urban labor in the disaggregated SAM

201

Abbreviation Meaning agskl1 rural skilled in Amazon agskl2 rural skilled in Northeast agskl3 rural skilled in Center-West agskl4 rural skilled in South and SE agunsk1 rural unskilled in Amazon agunsk2 rural unskilled in Northeast agunsk3 rural unskilled in Center-West agunsk4 rural unskilled in South and SE

Table D.3: Description of the main types of rural labor in the disaggregated SAM

Abbreviation Meaning lndar1 arable land in Amazon lndar2 arable land in Northeast lndar3 arable land in Center-West lndar4 arable land in South and SE lndgrs1 grassland in Amazon lndgrs2 grassland in Northeast lndgrs3 grassland in Center-West lndgrs4 grassland in South and SE lndfr1 forested land in Amazon lndfr2 forested land in Northeast lndfr3 forested land in Center-West lndfr4 forested land in South and SE

Table D.4: Description of the main types of land in the disaggregated SAM

202

Abbreviation Meaning Abbreviation Meaning small farm agricultural capag1s capital in Amazon capind1 industry capital in Amazon small farm agricultural industry and oil capital in capag2s capital in Northeast capind2 Northeast small farm agricultural industry and oil capital in capag3s capital in Center-West capind3 Center-West small farm agricultural industry and oil capital in capag4s capital in South and SE capind4 South and SE large farm agricultural construction capital in capag1l capital in Amazon capcon1 Amazon large farm agricultural construction capital in capag2l capital in Northeast capcon2 Northeast large farm agricultural construction capital in capag3l capital in Center-West capcon3 Center-West large farm agricultural construction capital in capag4l capital in South and SE capcon4 South and SE food processing capital in trade and transportation capfp1 Amazon captd1 capital in Amazon food processing capital in trade and transportation capfp2 Northeast captd2 capital in Northeast food processing capital in trade and transportation capfp3 Center-West captd3 capital in Center-West food processing capital in trade and transportation capfp4 South and SE captd4 capital in South and SE mining and oil capital in capmin1 Amazon capsvc1 services capital in Amazon mining and oil capital in services capital in capmin2 Northeast capsvc2 Northeast mining and oil capital in services capital in Center- capmin3 Center-West capsvc3 West mining and oil capital in services capital in South capmin4 South and SE capsvc4 and SE

Table D.5: Description of the main types of capital in the disaggregated SAM

203

APPENDIX E

ADDITIONAL RESULTS AND DISCUSSION FROM CHAPTER 2

204

E1 – Disaggregated regional SAM

Table E.1 displays the aggregated demand for factors of production for each sector. We aggregated all 60 regional activities into 15 activities. Labor is divided into two categories: skilled and unskilled labor. We can draw some interesting information here, such as the expected relative small importance of skilled labor in activities related to agricultural products, represented by the first eight rows. Land and capital are important market factors in these activities. The information compiled from the disaggregated SAM is also consistent with respect to the larger use of capital in large activities. The only exceptions are the livestock activities (ALIVSTS and ALIVSTL), since land is what matters for large livestock farms. The same is true for forest products activity (ALOGDEF), where land represents about 61% of the total factors demanded.

Sectors such as processed food (APRFOOD), mining and oil (AMINPET), industry

(AINDUST), and construction (ACONST), are very similar in terms of factor shares, where capital participation is above 60% and skilled labor is the main type of labor used.

The last two sectors, trade and transportation (ATRANTD), and services (ASVC), are the most skilled-labor sectors in the economy, with more than 34 % of factors employed, and less than 50 % of capital. The most labor-intensive sectors are services, transport and trade, and perennial farms.

205

Skilled labor Unskilled labor Capital Land Activities (%) (%) (%) (%) AANNS 3.2 32.4 32.3 32.0 AANNL 4.7 22.6 50.4 22.2 APERENS 10.9 47.6 30.4 11.1 APERENL 15.8 37.1 34.0 13.1 ALIVSTS 1.9 30.5 30.5 37.1 ALIVSTL 5.7 20.8 20.8 52.6 AOTRAGS 4.3 37.6 27.8 30.5 AOTRAGL 14.1 23.7 34.3 27.9 ALOGDEF 4.2 15.9 19.1 60.7 APRFOOD 13.3 8.7 78.0 - AMINPET 25.6 12.7 61.7 - AINDUST 31.1 7.1 61.8 - ACONST 11.5 9.8 78.7 - ATRANTD 43.4 9.5 47.1 - ASVC 37.8 18.4 43.7 - Regions: North 27.1 20.5 46.1 6.2 Northeast 30.0 18.5 48.1 3.4 Center-West 29.7 16.4 46.9 7.0 South and SE 31.0 13.8 52.3 2.8 Total Brazil 30.6 15.0 51.0 3.4 Source: author’s calculations from the disaggregated regional SAM. Sum may not equal 100 due to rounding.

Table E.1: Quantity of factors employed by each sector and region

206

The bottom of Table E.1 shows the factor shares in each region, and we can note that the South/Southeast region is the one that employs more capital and skilled labor, and the least land among all regions. These proportions are not surprising since this region is the smallest in geographical terms, but the richest in absolute and relative income. The land participation shows that the Amazon and Center-West regions use relatively more land, because the main forest activities are in the former, and many large agricultural properties are located in the latter. Table E.2 shows that more than 60 % of the households in both rural and urban areas are in the South/Southeast region.

hurblow hurbmed hrurlow hrurmed hhigh Regions (%) (%) (%) (%) (%) 5.4 4.4 9.4 7.9 4.2 North 22.4 13.5 18.0 16.4 12.9 Northeast 7.7 7.7 9.8 9.2 7.6 Center-West 64.6 74.4 62.7 66.5 75.3 South and SE Source: author’s calculations from the disaggregated regional SAM. Where: hurblow = urban low income household; hurbmed = urban medium income household; hrurlow = rural low income household; hrurmed = rural medium income household; and hhigh = rural and urban high income household. Sum may not equal 100 due to rounding.

Table E.2: Regional distribution of factor endowments for each type of household

Table E.3 shows the budget shares spent on the main commodities specified in the disaggregated SAM. Low-income-urban households spend more of their income in services (57 %) and processed food (32 %). Medium-income-urban households have a

207

different pattern of expenditure, in which consumption of industrial goods (31 %) represents a larger share than consumption of processed food (21 %). The share of services still is very high in this category of household (44 %). Low-income-rural households have larger shares of industrial goods (16 %) and agricultural goods (9 %), and smaller on services (46 %), in comparison to the low-income-urban households.

Medium-income-rural households have very similar budget shares to those in urban areas. The last category of households are those with high income in both urban and rural areas, and their budget shares are mostly represented by expenditure on industrial goods

(48 %) and services (39 %), with just 10 % spent on processed food.

hurblow hurbmed hrurlow hrurmed hhigh Commodities (%) (%) (%) (%) (%) Processed food 31.8 21.5 29.2 21.5 10.0

Mining and oil 0.4 0.4 - 0.5 0.3

Agricultural goods 4.4 3.5 8.8 4.4 2.5

Industrial goods 6.1 30.7 15.7 30.0 48.2

Services 57.3 44.0 46.3 43.6 39.0

Source: author’s calculations from the disaggregated regional SAM. Where: hurblow = urban low income household; hurbmed = urban medium income household; hrurlow = rural low income household; hrurmed = rural medium income household; and hhigh = rural and urban high income household. Sum may not equal 100 due to rounding.

Table E.3: Budget share for commodities by households

208

E.2 - Overall Trade Liberalization (Scenario 1)

Regional Impacts

Region North

Table E.4 shows that, in the Region North, there were reductions in domestic prices of the activities: large farm annuals (aannl1l), industry (aindust1), and construction

(aconst1). For large farm annuals, price reductions occurred for rice, sugar, beans, and horticultural products. In general, the main price increases for agricultural products were those of corn, manioc, annual commodities, and soybean106.

Since we are assuming full employment and a mobile labor market, Tables E.5 and E.6 show that the changes in factor price and factor income are identical for some labor categories. The results seem to suggest that the main changing component for the factor prices was the effect on capital prices, reducing the final price and production for large farm annuals, industry, and construction (Table E.4). The prices of capital and land employed in the sector “other small agricultural goods” (Aotrag1s) also decreased after the import tariff removal, but this reduction was not large enough to reduce its price

(Table E.4). Its production was probably reduced due to the substitution effects caused by the increase in the production of other agricultural commodities specified in the model

(previously discussed).

The prices of capital and land for large farm annuals decrease substantially (Table

E.5). Labor and capital prices and income are reduced in the industry and construction

106 It is important to note that the output price and domestic production for some goods increased or decreased in a static general equilibrium framework, which would be unusual in a partial equilibrium analysis, under a ceteris paribus assumption.

209

sectors, with larger negative impact on low income urban households (Hurblow), who are more dependent on capital-intensive goods (Tables E.5 and E.6).

The prices of capital and land employed in the sector “other small agricultural goods” (Aotrag1s) also decreased after the import tariff removal, but this reduction was not large enough to reduce its price (Table E.4). Its production was probably reduced due to the substitution effects caused by the increase in the production of other agricultural commodities specified in the model (previously discussed).

The prices of capital and land for large farm annuals decrease substantially (Table

E.5). Labor and capital prices and income are reduced in the industry and construction sectors, with larger negative impact on low income urban households (Hurblow), who are more dependent on capital-intensive goods (Tables E.5 and E.6).

Region Northeast

After eliminating import tariffs in the Northeast, the output prices and production fall for both small and large farms that produce annual agricultural commodities (Table

E.7). Prices of intermediate aggregate and value added fall for these activities as well.

Following the same logic as in the analysis of Region North, we can use Table E.1 to help to understand the results obtained in Table E.7. According to Table E.1, the

Northeast uses more skilled labor and capital than the North, and these two agricultural activities employ a large amount of capital and land. The same happens with other activities, such as forest products (Alogdef2), industry (Aindust2), and construction

(Aconst2). Although the latter does have a decrease in the activity price, it is not large enough to reduce its production.

210

Activities Output price Domestic Price of Price of value production intermediate added aggregate Small farm 1.92 2.82 0.39 3.51 annuals Large farm -1.31 -7.22 -0.78 -2.03 annuals Small farm 2.13 3.43 -0.22 4.22 perennials Large farm 2.29 3.9 -0.23 4.73 perennials Small livestock 2.80 1.13 0.86 3.67 Large livestock 3.04 1.46 1.05 4.17 Small other 1.15 -0.95 0.70 1.70 agricultural Large other 1.03 -0.49 0.75 1.61 agricultural Forest products 0.97 - 0.33 1.55 Food processing 1.79 2.26 1.03 3.88 Mining and oil 3.00 6.12 0.06 8.72 Industry -1.26 -0.47 -1.12 -1.69 Construction -0.63 0.08 -0.68 -0.56 Trade and 0.66 2.67 -0.75 2.29 transportation Services 0.59 0.48 -0.19 1.18

Table E.4: Simulation results for the Region North after an overall 100 % reduction in the import tariffs (% change from benchmark values)

211

Activities Skilled Labor Unskilled Labor Capital Land

Small farm annuals - 3.04 4.76 4.76

Large farm annuals - 3.04 -9.06 -9.06

Small farm 3.89 3.04 5.55 5.55 perennials Large farm 3.89 3.04 6.18 6.18 perennials Small livestock 3.89 3.04 3.89 3.89

Large livestock 3.89 3.04 4.46 4.46

Small other 3.89 3.04 -0.04 -0.04 agricultural Large other 3.89 3.04 0.35 0.35 agricultural Forest products 3.89 3.04 0.85 0.85

Food processing 3.88 3.88 3.88 -

Mining and oil 8.72 8.72 8.72 -

Industry -1.69 -1.69 -1.69 -

Construction -0.55 -0.55 -0.55 -

Trade and 1.38 1.38 3.27 - transportation Services 1.38 1.38 0.99 -

Table E.5: Factor prices by each activity in the Region North after an overall 100 % reduction in the import tariffs (% change from benchmark values)

212

Labor Use Hrurlow Hrurmed Hurblow Hurbmed Hhigh

Food processing

Skilled - - 3.88 3.88 3.88

Unskilled - - 3.88 3.88 3.88

Heavy industry(*)

Skilled - - 8.72 8.72 8.72

Unskilled - - 8.72 8.72 8.72

Light industry(**)

Skilled - - -1.69 -1.69 -1.69

Unskilled - -1.69 -1.69 -1.69 -1.69

Construction

Skilled - - -0.55 -0.55 -0.55

Unskilled - -0.55 -0.55 -0.55 -0.55

Services

Skilled - - 1.38 1.38 1.38

Unskilled 1.20 1.20 1.20 1.20 1.20

Agriculture

Skilled 3.89 3.89 - - -

Unskilled 3.04 3.04 - 3.04 3.04

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.6: Household’s labor income in the Region North after an overall 100 % reduction in the import tariffs (% change from benchmark values)

213

As in the region North, rice, sugar, beans, and horticultural products have their prices reduced, but in smaller magnitude. Corn, manioc, soybean, other annual commodities, cocoa, other perennial commodities, cattle and poultry production, have an increase in price.

The reduction in output prices and production in industry and construction sectors is due to these activities being so capital intensive, with the reallocation of resources occurring towards labor-intensive sectors. In the case of sectors producing forest products, the main resource used is land, responsible for more than 60 % of the total resources used by this sector107.

The prices paid for the use of factors of production can be seen in Table E.8, in which we can note that the labor payments are higher after the elimination of import tariffs for all sectors, except for the industry and construction sectors. This table also helps to explain the results from Table E.7, since the price of land and capital fall for small and large farms producing annual agricultural commodities, and also for the forest products sector.

Workers in the industry and construction sectors lose with overall elimination of import tariffs (Table E.9). Even though small and large farms producing annual agricultural goods, and forest products sector, experience a fall in prices and output, substantially reducing the returns from capital and land in these sectors, the effects on the labor market in these activities are positive for all households.

107 In the case of North, the sector forest products (Alogdef1) did not have reduction in production and price (Table E.4) because price and quantity of the commodity “deforestation“ increase as a result from trade reform, increasing the area deforested to be allocated in agricultural activities. The Amazon forest is located in this region, and it is the main reason why this commodity is “produced” only in this region.

214

Price of Output Price Domestic Price of Value Activities Intermediate Production Added Aggregate Small farm annuals -0.96 -3.39 -0.50 -1.32 Large farm annuals -1.62 -5.47 -0.02 -5.10 Small farm 2.45 4.91 -0.02 4.36 perennials Large farm 2.16 5.45 -0.27 4.61 perennials Small livestock 2.82 1.90 0.74 3.72 Large livestock 2.42 2.10 0.56 3.80 Small other 0.89 0.05 0.57 1.22 agricultural Large other 0.48 1.67 -0.19 1.95 agricultural Forest products -0.95 -0.49 -1.44 -0.64 Food processing 2.04 1.26 1.28 3.21 Mining and oil 3.00 6.15 0.04 8.74 Industry -1.28 -0.40 -1.14 -1.64 Construction -0.63 0.10 -0.71 -0.54 Trade and 0.80 2.08 -0.79 2.05 transportation Services 0.62 0.33 -0.19 1.05

Table E.7: Simulation results for the Region Northeast after an overall 100 % reduction in the import tariffs (% change from benchmark values)

215

Activities Skilled Labor Unskilled Labor Capital Land

Small farm 2.80 1.65 -4.61 -4.61 annuals Large farm 2.80 1.65 -6.84 -6.84 annuals Small farm 2.80 1.65 7.62 7.62 perennials Large farm 2.80 1.65 7.80 7.80 perennials Small livestock 2.80 1.65 4.80 4.80 Large livestock 2.80 1.65 4.52 4.52 Small other 2.80 1.65 0.88 0.88 agricultural Large other 2.80 1.65 2.02 2.02 agricultural Forest products 2.80 1.65 -1.55 -1.55 Food 3.21 3.21 3.21 - processing Mining and oil 8.74 8.74 8.74 - Industry -1.64 -1.64 -1.64 - Construction -0.54 -0.54 -0.54 - Trade and 1.21 1.04 2.85 - transportation Services 1.21 1.04 0.90 -

Table E.8: Factor prices by each activity in the Region Northeast after an overall 100 % reduction in the import tariffs (% change from benchmark values)

216

Region Center-West

The Center-West is a region of many and constant changes, where the industry and agriculture are expanding their production. Because most of the land areas of the

South and Southeast are already occupied for many purposes, Center-West is known as the “new agriculture frontier” in Brazil, since it is a large region with many arable and flat areas. This is the region where the largest soybeans, corn and cotton farms are located.

Elimination of import tariffs has important impacts on the Center-West. The results from simulation show, in Table E.10, that the output prices of small farms producing annual commodities (Aannl3s), forest products (Alogfdef3), industry

(Aindust3), and construction (Aconst3) falls. Except for construction, however, production decreases for other agricultural commodities (small and large farms). The main changes in prices follow the same pattern as in the previous regions, with the main fall in prices occurring for rice, beans, sugar, and horticultural commodities.

Commodities that have their prices increased follow exactly the same behavior as those from the North.

According to Table E.1, the Center-West has a very similar distribution of market factors as that from the Northeast. The main difference is the amount of land as input in production, which is more than two times larger than that found in the Northeast. Due to the characteristics of the agriculture in this region, small farms producing annual commodities reduce production, but the opposite occurs for large farms, since this region has a very high concentration of large properties.

217

Labor Use Hrurlow Hrurmed Hurblow Hurbmed Hhigh

Food processing

Skilled - - 3.21 3.21 3.21

Unskilled 3.21 3.21 3.21 3.21 3.21

Heavy industry(*)

Skilled - - 8.74 8.74 8.74

Unskilled 8.74 8.74 8.74 8.74 8.74

Light industry(**)

Skilled - - -1.64 -1.64 -1.64

Unskilled -1.64 -1.64 -1.64 -1.64 -1.64

Construction

Skilled - - -0.54 -0.54 -0.54

Unskilled -0.54 -0.54 -0.54 -0.54 -0.54

Services

Skilled - - 1.21 1.21 1.21

Unskilled 1.04 1.04 1.04 1.04 1.04

Agriculture

Skilled 2.80 2.80 - - -

Unskilled 1.65 1.65 - 1.65 1.65

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.9: Household’s labor income in the Region Northeast after an overall 100 % reduction in the import tariffs (% change from benchmark values)

218

The main increases in sectoral production occur in the mining and oil sector, and in large farms producing annual commodities, such as soybeans, since it is the main agricultural product in this region.

Table E.11 shows that the prices paid for the use of factors of production help to understand the results from Table E.10, since the price of land and capital fall for forest production, and small farms, and increase for large farms producing annual agricultural commodities. According to Table E.12, once again people working in the light industry and construction sectors lose with overall elimination of import tariffs.

Region South/Southeast

Import tariff elimination causes output prices for forest products (Alogfdef4), industry (Aindust4), and construction (Aconst4) to fall (Table E.13). There is no reduction for any agricultural activity. However, there is a small decrease in the production of annual agricultural commodities for small farms (Aannl4s). Production also decreases for forest products, industry, and other agricultural commodities produced by large farms (Aotrag4l). Commodity prices and production follow the same pattern as in the previous regions.

219

Price of Output Price Domestic Price of Value Activities Intermediate Production Added Aggregate Small farm -0.90 -1.68 -1.14 -0.49 annuals Large farm 1.33 4.60 -0.24 5.02 annuals Small farm 1.87 2.17 -0.29 3.63 perennials Large farm 2.05 2.78 -0.32 4.05 perennials Small livestock 2.50 0.82 0.74 3.28 Large livestock 2.43 1.63 0.41 3.99 Small other 1.05 -0.58 0.72 1.48 agricultural Large other 0.93 -0.13 0.59 1.52 agricultural Forest products -1.58 -2.21 -0.23 -2.12 Food 1.95 1.61 1.20 3.44 processing Mining and oil 3.20 5.31 0.05 8.17 Industry -1.29 -0.35 -1.13 -1.61 Construction -0.63 0.09 -0.70 -0.54 Trade and 0.77 2.19 -0.78 2.02 transportation Services 0.64 0.24 -0.18 1.01

Table E.10: Simulation results for the Region Center-West after an overall 100 % reduction in the import tariffs (% change from benchmark values)

220

Activities Skilled Labor Unskilled Labor Capital Land

Small farm 3.44 3.27 -2.84 -2.84 annuals Large farm 3.44 3.27 5.68 5.68 annuals Small farm 3.44 3.27 3.94 3.94 perennials Large farm 3.44 3.27 4.75 4.75 perennials Small livestock 3.44 3.27 3.28 3.28 Large livestock 3.44 3.27 4.22 4.22 Small other 3.44 3.27 0.30 0.30 agricultural Large other 3.44 3.27 0.66 0.66 agricultural Forest products 3.44 3.27 -3.87 -3.87 Food 3.44 3.44 3.44 - processing Mining and oil 8.17 8.17 8.17 - Industry -1.61 -1.61 -1.61 - Construction -0.54 -0.54 -0.54 - Trade and 1.17 1.02 2.94 - transportation Services 1.17 1.02 0.84 -

Table E.11: Factor prices by each activity in the Region Center-West after an overall 100 % reduction in the import tariffs (% change from benchmark values)

221

Labor Use Hrurlow Hrurmed Hurblow Hurbmed Hhigh

Food processing

Skilled - - 3.44 3.44 3.44

Unskilled - - 3.44 3.44 3.44

Heavy industry(*)

Skilled - - 8.17 8.17 8.17

Unskilled - - 8.17 8.17 8.17

Light industry(**)

Skilled - - -1.61 -1.61 -1.61

Unskilled -1.61 -1.61 -1.61 -1.61 -1.61

Construction

Skilled - - -0.54 -0.54 -0.54

Unskilled - -0.54 -0.54 -0.54 -0.54

Services

Skilled - - 1.17 1.17 1.17

Unskilled 1.02 1.02 1.02 1.02 1.02

Agriculture

Skilled 3.44 3.44 - - -

Unskilled 3.27 3.27 - 3.27 3.27

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.12: Household’s labor income in the Region Center-West after an overall 100 % reduction in the import tariffs (% change from benchmark values)

222

The main increase in production is the mining and oil sector (Aminpet4) (Table

E.13). Large farms producing annual commodities do not have a large increase in production as seen in the Center-West, since the agricultural areas in the South/Southeast have already been used at their full capacity for many years.

The fall in production of forest products and industry is related to the large amount of reduction in the capital price (Tables E.14 and E.15). The mining and oil sector has a significant increase in production even with a high increase in labor and capital prices (value-added), which are transmitted for all households’ categories. Urban low-income households seem to lose a large proportion of their income because of the losses from the light industry sector. The rationale here is that these households are the main labor force used by the industry in urban areas which, in the case of the

South/Southeast, represents a large proportion of the country, as noted in Tables E.1 and

E.2.

223

Price of Output Price Domestic Price of Value Activities Intermediate Production Added Aggregate Small farm annuals 0.24 -0.25 0.01 0.50 Large farm annuals 1.32 1.44 -3.14 -0.24 Small farm 1.28 1.20 -0.40 2.17 perennials Large farm 1.00 2.00 -0.76 2.42 perennials Small livestock 2.80 1.00 1.65 3.25 Large livestock 2.20 1.63 0.61 3.30 Small other 1.05 0.11 0.53 1.33 agricultural Large other 0.91 -0.03 0.70 1.28 agricultural Forest products -2.43 -2.42 -1.03 -3.94 Food processing 2.07 1.12 1.67 3.11 Mining and oil 3.02 6.03 0.03 8.66 Industry -1.28 -0.38 -1.09 -1.63 Construction -0.64 0.14 -0.74 -0.51 Trade and 0.75 2.27 -0.84 2.08 transportation Services 0.64 0.25 -0.07 1.02

Table E.13: Simulation results for the Region South/Southeast after an overall 100 % reduction in the import tariffs (% change from benchmark values)

224

Activities Skilled Labor Unskilled Labor Capital Land

Small farm annuals 1.91 2.01 -0.09 -0.09 Large farm annuals 1.91 2.01 -1.07 -1.07 Small farm 1.91 2.01 2.44 2.44 perennials Large farm 1.91 2.01 2.92 2.92 perennials Small livestock 1.91 2.01 3.84 3.84 Large livestock 1.91 2.01 3.81 3.81 Small other 1.91 2.01 0.90 0.90 agricultural Large other 1.91 2.01 0.80 0.80 agricultural Forest products 1.91 2.01 -4.72 -4.72 Food processing 3.11 3.11 3.11 - Mining and oil 8.66 8.66 8.66 - Industry -1.63 -1.63 -1.63 - Construction -0.51 -0.51 -0.51 - Trade and 1.29 1.08 3.00 - transportation Services 1.29 1.08 0.84 -

Table E.14: Factor prices by each activity in the Region South/Southeast after an overall 100 % reduction in the import tariffs (% change from benchmark values)

225

Labor Use Hrurlow Hrurmed Hurblow Hurbmed Hhigh

Food processing

Skilled - - 3.11 3.11 3.11

Unskilled 3.11 3.11 3.11 3.11 3.11

Heavy industry(*)

Skilled - - 8.66 8.66 8.66

Unskilled 8.66 8.66 8.66 8.66 8.66

Light industry(**)

Skilled - - -1.63 -1.63 -1.63

Unskilled -1.63 -1.63 -1.63 -1.63 -1.63

Construction

Skilled - - -0.51 -0.51 -0.51

Unskilled -0.51 -0.51 -0.51 -0.51 -0.51

Services

Skilled - - 1.29 1.29 1.29

Unskilled 1.08 1.08 1.08 1.08 1.08

Agriculture

Skilled 1.91 1.91 - - -

Unskilled 2.01 2.01 - 2.01 2.01

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.15: Household’s labor income in the Region South/Southeast after an overall 100 % reduction in the import tariffs (% change from benchmark values)

226

E.3 – Sectoral Trade Liberalization (Scenario 2)

50 % reduction import tariff AGR ANN PER IND MIX Absorption - - - - - Private consumption - - - 0.1 0.1 Exports 0.6 0.4 0.2 6.1 6.6 Imports 0.5 0.4 0.2 5.2 5.7 Real exchange rate 0.1 0.1 - 2.0 2.1

Share of GDP (%)

Investment - - - -0.1 -0.1 Private savings - - - 0.2 0.2 Foreign savings - - - 0.1 0.1 Government savings - - - -0.4 -0.4 Tariff revenue - - - -0.4 -0.4 Direct tax revenue - - - - -

Equivalent Variation (%)

Rural low inc. household -0.2 -0.2 - 0.5 0.5 Rural medium income -0.2 -0.1 - 0.5 0.5 household Urban low income - - - -0.4 -0.3 household Urban medium income - - - - - household High income household - - - 0.2 0.2 Total welfare - - - 0.06 0.07 Gini coefficient 0.02 0.02 0.002 -0.1 -0.09 Theil index 0.04 0.04 0.004 -0.2 -0.1

Table E.16: Simulations results for 50 % sectoral import tariffs reduction (scenario 2), % change from benchmark values

227

Agricultural Sector

The sectoral import tariff reduction, composed of three main simulations, import tariff reduction in agriculture (AGR), industry (IND), and a combination of industrial and agricultural sectors (MIX) affects the regions differently.

The elimination of the import tariffs in the agricultural sector does not bring better welfare outputs for rural households, due to a reduction in labor payments in the agricultural sector. Table E.17 shows exactly how much reduction rural households had in their wages after the import tariff reduction, with a larger decrease for unskilled workers, except in the Center-West. In addition to these results, there were reductions in domestic sales for those goods that experienced tariff reduction, and vice-versa for commodities such as soybeans, coffee, cocoa, sugar, milk, and cattle and poultry meat.

The reduction of import tariffs in agriculture shows that rural households have a negative and substantial reduction in their labor income, mainly in the Northeast (Table

E.18).

228

Labor Use North Northeast Center-West South/Southeast

Food processing

Skilled 0.56 0.20 0.31 0.11

Unskilled 0.56 0.20 0.31 0.11

Heavy industry(*)

Skilled 0.32 0.31 0.30 0.31

Unskilled 0.32 0.31 0.30 0.31

Light industry(**)

Skilled 0.27 0.21 0.19 0.15

Unskilled 0.27 0.21 0.19 0.15

Construction

Skilled 0.12 0.12 0.12 0.12

Unskilled 0.12 0.12 0.12 0.12

Services

Skilled 0.22 0.21 0.21 0.21

Unskilled 0.21 0.21 0.20 0.21

Agriculture

Skilled 0.35 -0.95 -0.72 -0.51

Unskilled -0.42 -1.77 -0.32 -0.60

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.17: Household’s labor income after elimination of the import tariffs in agriculture (% change from benchmark values)

229

Rural low Rural Urban low Urban High Regions income medium income medium income household income household income household household household North -0.3 -0.3 0.2 0.2 0.2

Northeast -1.4 -1.5 0.2 0.2 0.2

Center-West -0.3 -0.4 0.2 0.2 0.2

South/Southeast -0.5 -0.5 0.2 0.2 0.2

Table E.18: Regional changes in household’s labor income after elimination of the import tariffs in agriculture (% change from benchmark values)

Industry Sector

Elimination of an import tariff in the industry harms urban low and medium income households instead of rural households as seen in the case of AGR. Table E.19 points out that labor income increases substantially in the heavy industry labor used in the mining and oil production. However, the labor income in the light industry decreases in all four regions, contributing to reducing the urban household’s welfare, since capital payments also decrease in all regions for the industry sector.

230

Labor Use North Northeast Center-West South/Southeast

Food processing

Skilled 3.29 2.99 3.11 2.99

Unskilled 3.29 2.99 3.11 2.99

Heavy industry(*)

Skilled 8.44 8.47 7.90 8.39

Unskilled 8.44 8.47 7.90 8.39

Light industry(**)

Skilled -1.94 -1.85 -1.79 -1.77

Unskilled -1.94 -1.85 -1.79 -1.77

Construction

Skilled -0.67 -0.66 -0.66 -0.63

Unskilled -0.67 -0.66 -0.66 -0.63

Services

Skilled 1.17 1.01 0.97 1.08

Unskilled 0.99 0.84 0.82 0.88

Agriculture

Skilled 3.50 3.70 4.10 2.39

Unskilled 3.51 3.39 3.54 2.58

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.19: Household’s labor income after elimination of the import tariffs in industry (% change from benchmark values)

231

The overall regional labor income for all categories of households confirms that the improvements in labor payments for urban households are not large enough to overcome the reduction on capital payments in the industry (table E.20). Rural households are those that gain from trade reform in the industry sector, allowing substantial increase in their wages.

Rural low Rural Urban low Urban High Regions income medium income medium income household income household income household household household North 3.3 3.3 1.0 1.0 1.0

Northeast 3.1 3.3 0.9 0.7 0.9

Center-West 3.2 3.4 0.8 0.8 0.9

South/Southeast 2.4 2.4 0.5 0.5 0.4

Table E.20: Regional changes in household’s labor income after elimination of the import tariffs in industry (% change from benchmark values)

Agriculture and Industry

Sectoral elimination of import tariffs in agriculture and industry produced opposite welfare outcomes for low and medium income households, in both rural and

232

urban areas. The elimination of import tariffs as a combination of agricultural and industrial sectors (MIX) brings welfare losses for urban low and medium income households (Table E.16).

The regional impacts of this combination can be observed in the next tables. The effects from removing the import tariff in the industry seem to overcome the effects from the agricultural sector, which shows a substantial reduction in labor payments in the light industry labor, negatively affecting urban households’ income (Table E.21).

Similarly to Table E.20, Table E.22 shows once again that the overall labor income gains for urban low and medium income households were not enough to overcome the losses in capital payments. Rural households in the Center-West gain the largest increase in labor income among the regions.

233

Labor Use North Northeast Center-West South/Southeast

Food processing

Skilled 3.62 3.10 3.29 3.11

Unskilled 3.62 3.10 3.29 3.11

Heavy industry(*)

Skilled 8.67 8.69 8.12 8.61

Unskilled 8.67 8.69 8.12 8.61

Light industry(**)

Skilled -1.82 -1.72 -1.69 -1.67

Unskilled -1.82 -1.72 -1.69 -1.67

Construction

Skilled -0.56 -0.55 -0.55 -0.52

Unskilled -0.56 -0.55 -0.55 -0.52

Services

Skilled 1.27 1.11 1.07 1.18

Unskilled 1.09 0.94 0.92 0.98

Agriculture

Skilled 4.01 3.63 3.90 2.23

Unskilled 2.75 3.03 3.58 2.40

(*) Labor used by “mining and oil” activity (**) Labor used by “industry” activity.

Table E.21: Household’s labor income after elimination of the import tariffs in a combination of agriculture and industry (% change from benchmark values)

234

Rural low Rural Urban low Urban High Regions income medium income medium income household income household income household household household North 2.8 2.7 1.1 1.0 1.1

Northeast 2.8 3.0 1.0 0.8 1.0

Center-West 3.2 3.4 0.9 0.9 1.0

South/Southeast 2.3 2.3 0.6 0.6 0.5

Table E.22: Regional changes in household’s labor income after elimination of the import tariffs in a combination of agriculture and industry (% change from benchmark values)

E.4 – Equity-Efficiency Trade Liberalization (Scenario 3)

In the North the magnitude of some of the input payments changed from section

2.8.2. Table E.23 shows the percentage change of the prices for the main factors of production used in the activities in the North. Once again, capital and land payments for large farm annuals (Aannl1l) have a substantial decrease of almost 9 % with the implementation of the combined policies, but smaller than the results from section 2.8.2.

The industry and construction sectors have the same direction of change for labor and

235

capital payments as in scenario 2. It is interesting to note, however, that small farms producing other agricultural commodities (Aotrag1s) have positive land and capital payments under the combined policies scenario.

The Northeast region also has a similar pattern of change for payments of factor of production as in section 2.8.2 (Table E.24). Capital and land payments have larger changes in their payments under the combined policies, in comparison to Table E.8, when considering only the reduction in the import tariffs (scenario 1). In the same way, labor payments are larger in Table E.24, showing a larger appreciation for unskilled labor wages relative to skilled ones.

In the Center-West, factor payments for unskilled labor increased more relatively to skilled labor (Table E.25). Most of capital and land used by sectors have some increase in their prices, but with the same direction of change as in section 2.8.2.

The main changes in the South/Southeast for payments for factors of production are in the size of the changes, since the direction of changes is the same as in section

2.8.2. Labor income for unskilled workers has a relatively larger change than those for skilled workers in each agricultural activity (Table E.26).

236

Activities Skilled Labor Unskilled Labor Capital Land

Small farm annuals - 3.16 5.04 5.04 Large farm annuals - 3.16 -8.90 -8.90 Small farm 3.89 3.16 5.52 5.52 perennials Large farm 3.89 3.16 6.17 6.17 perennials Small livestock 3.89 3.16 3.97 3.97 Large livestock 3.89 3.16 4.38 4.38 Small other 3.89 3.16 0.32 0.32 agricultural Large other 3.89 3.16 0.71 0.71 agricultural Forest products 3.89 3.16 0.80 0.80 Food processing 4.05 4.05 4.05 - Mining and oil 8.58 8.58 8.58 - Industry -2.05 -2.05 -2.05 - Construction -0.69 -0.69 -0.69 - Trade and 1.51 1.37 3.06 - transportation Services 1.51 1.37 1.20 -

Table E.23: Factor prices by each activity in the Region North after combining trade/tax reform (% change from benchmark values)

237

Activities Skilled Labor Unskilled Labor Capital Land

Small farm annuals 2.78 1.70 -4.53 -4.53 Large farm annuals 2.78 1.70 -7.46 -7.46 Small farm 2.78 1.70 7.56 7.56 perennials Large farm 2.78 1.70 7.79 7.79 perennials Small livestock 2.78 1.70 4.92 4.92 Large livestock 2.78 1.70 4.62 4.62 Small other 2.78 1.70 1.25 1.25 agricultural Large other 2.78 1.70 2.48 2.48 agricultural Forest products 2.78 1.70 -1.85 -1.85 Food processing 3.35 3.35 3.35 - Mining and oil 8.60 8.60 8.60 - Industry -2.00 -2.00 -2.00 - Construction -0.68 -0.68 -0.68 - Trade and 1.34 1.20 2.64 - transportation Services 1.34 1.20 1.09 -

Table E.24: Factor prices by each activity in the Region Northeast after combining trade/tax reform (% change from benchmark values)

238

Activities Skilled Labor Unskilled Labor Capital Land

Small farm annuals 3.53 3.38 -2.77 -2.77 Large farm annuals 3.53 3.38 5.89 5.89 Small farm 3.53 3.38 3.99 3.99 perennials Large farm 3.53 3.38 4.77 4.77 perennials Small livestock 3.53 3.38 3.50 3.50 Large livestock 3.53 3.38 4.21 4.21 Small other 3.53 3.38 0.65 0.65 agricultural Large other 3.53 3.38 1.06 1.06 agricultural Forest products 3.53 3.38 -4.23 -4.23 Food processing 3.59 3.59 3.59 - Mining and oil 8.03 8.03 8.03 - Industry -1.95 -1.95 -1.95 - Construction -0.68 -0.68 -0.68 - Trade and 1.29 1.17 2.73 - transportation Services 1.29 1.17 1.02 -

Table E.25: Factor prices by each activity in the Region Center-West after combining trade/tax reform (% change from benchmark values)

239

Activities Skilled Labor Unskilled Labor Capital Land

Small farm annuals 2.00 2.11 -0.03 -0.03 Large farm annuals 2.00 2.11 -1.29 -1.29 Small farm 2.00 2.11 2.54 2.54 perennials Large farm 2.00 2.11 3.09 3.09 perennials Small livestock 2.00 2.11 4.08 4.08 Large livestock 2.00 2.11 3.99 3.99 Small other 2.00 2.11 1.29 1.29 agricultural Large other 2.00 2.11 1.15 1.15 agricultural Forest products 2.00 2.11 -5.09 -5.09 Food processing 3.25 3.25 3.25 - Mining and oil 8.52 8.52 8.52 - Industry -1.96 -1.96 -1.96 - Construction -0.64 -0.64 -0.64 - Trade and 1.39 1.22 2.79 - transportation Services 1.39 1.22 1.02 -

Table E.26: Factor prices by each activity in the Region South/Southeast after combining trade/tax reform (% change from benchmark values)

240

APPENDIX F

PRODUCT DISAGGREGATION ACROSS SECTORS

241

Brazilian Exports (%) Brazilian Imports (%) 87- Other vehicles 14.00 87- Other vehicles 21.79 84 - Nuclear reactors 11.60 27 – Mineral fuels 15.84 85 - Electrical machinery 7.91 10 – Cereals 15.49 39 – Plastics and articles thereof 5.66 39 – Plastics and articles thereof 5.45 48 – Paper and paperboard 5.60 84 – Nuclear reactors 4.20 72 – Iron and steel 3.01 85 - Electrical machinery 2.44 73 – Articles of iron and steel 2.83 12 – Oilseeds and oleaginous fruits 2.02 64 – Footwear 2.66 29 – Organic chemicals 2.01 29 – Organic chemicals 2.64 7 – Edible vegetables 1.93 40 – Rubber and articles thereof 2.58 4 – Dairy produce, birds eggs, natural honey 1.89 Source: Ministry of Development, Industry and International Trade (MDIC), Haddad et al. (2002)

Table F.1: Sectoral participation in trade between Brazil and Mercosur partners, 2001

Country Country Argentina ARG Guatemala GTM Bolivia BOL Honduras HND Brazil BRA Mexico MEX Canada CAN Nicaragua NIC Chile CHL Paraguay PRY Colombia COL Peru PER Costa Rica CRI United States USA Ecuador ECU Venezuela VEN El Salvador SLV Uruguay URY

Table F.2: Main countries considered in the proposed FTAA analysis

242

Livestock Sector Products 00 – Live animals except fish 01 – Meat preparations 02 – Dairy products and eggs 03 – Fish, shellfish, and others

Table F.3: Main products included in the livestock sector for the Mercosur and the FTAA analysis

Agricultural Sector Products 4 - Cereals/cereal preparation 5 – Vegetables and fruits 6 – Sugar, sugar preparation, honey 7 – Coffee, tea, cocoa, spices 9 – Miscellaneous food products 11 – Beverages 12 – Tobacco, manufactures 21 – Raw skin, fur 22 – Oil seeds, oil fruits 23 – Crude synthetic rubber 24 – Cork and wood 25 – Pulp and waste paper 26 – Textile fibers 41 – Animal oil, fat 42 – Fixed vegetable oils, fats 43 – Animal/vegetable oils, process. “d”

Table F.4: Main products included in the agricultural sector the Mercosur and the FTAA analysis

243

Chemical Sector Products 51 – Organic chemicals 52 – Inorganic chemicals 53 – Dyeing, tanning, color materials 54 – Pharmaceutical products 55 – Perfume, cosmetic products 56 – Manufactured fertilizers 57 – Plastics in primary form 58 - Plastics non-primary form 59 – Chemical materials

Table F.5: Main products included in the chemical sector the Mercosur and the FTAA analysis

Manufactured Sector Products 61 – Leather manufactures 77 – Electrical equipment 62 – Rubber manufactures 78 – Road vehicles 63 – Cork/wood manufactures 79 – Railway/tramway equipment 64 – Paper/paperboard materials 81 – Building fixtures, others 65 – Textile yarn, fabric 82 – Furniture, furnishings 66 – Non-metal mineral manufactures 83 – Travel goods, handbag, others 71 – Power generating equipment 84 – Apparel, clothing, accessories 72 – Industry special machine 85 - Footwear 73 – Metalworking machinery 87 – Scientific instruments, others 74 – Industrial equipment 88 – Photographic equipments, clocks 75 – Office machines 89 – Miscellaneous manufactures 76 – Telecommunication equipments

Table F.6: Main products included in the manufactured sector the Mercosur and the FTAA analysis

244

Mining and Oil Sector Products 27 – Crude fertilizer/mineral 28 – Metal ores/metal scrap 32 – Coal, coke, briquettes 33 – Petroleum and products 34 – Gas natural/manufactured 67 – Iron and steel 68 – Non-ferrous metals 69 – Metal manufactures 97 – Gold ore

Table F.7: Main products included in the mining and oil sector the Mercosur and the FTAA analysis

245

LIST OF REFERENCES

Chapter 1:

Barros de Castro, A. and Souza, P.F. A Economia Brasileira em Marcha Forçada. Rio de Janeiro: Paz e Terra, 1985.

Benevides, M.B. O Governo Kubistchek: Desenvolvimento Econômico e Estabilidade Política.. Rio de Janeiro: Paz e Terra, 1976.

Bértola, L, Bittencourt, M. and Porcile, J.G. Does Specialisation Matter? Structural and Income Convergence in Argentina, Brazil and Uruguay, 1963-1996. Paper published in the Anais do 26º Encontro Nacional de Economia - ANPEC. Vitória-ES, December, 1998.

Bértola, L., Bittencourt, M.V.L., Porcile, J.G. The Thirlwall Law Revisited: Aplication of a VAR Model to the Brazilian Economy in the Pos-War Period. Proceedings, 3rd. Brazilian Congress of Economic History, Curitiba, Parana, 1999.

Fishlow, A. The Latin American State. The Journal of Economic Perspectives 4 (3), 1990, pp. 61-74.

Leopoldi, M. “ Crescendo em Meio a Incerteza: A Política Econômica do Governo JK”, in Castro Gomes, A. (org.) O Brasil de JK, Rio de Janeiro: FGV, CPDOC, 1991.

Porcile, J.G., Higachi, H., and Bittencourt, M. Balance-of-Payments-Constrained Growth in Brazil: A Cointegration Test of the Thirlwall’s Law. Annual Conference of the Latin American Cliometric Association (LACLIO Conference), Stanford University, November 17-18, 2000.

246

Chapter 2:

Adelman, I. and S. Robinson. Income distribution policy in developing countries: a case study of Korea. Stanford University Press, Stanford, 1978.

Armington, P.A. A theory of demand for products distinguished by place of production. IMF Staff Papers 16 (1), 1969, 159-178.

Arndt, C., Robinson, S., and F. Tarp. Parameter estimation for a computable general equilibrium model: a maximum entropy approach. Discussion Paper 40. Trade and Macroeconomics Division, IFPRI. 2001.

Arrow, K. and G. Debreu. Existence of an equilibrium for a competitive economy. Econometrica 22, 1954, 265-90.

Asano, S. and E. Fiuza. Estimation of the Brazilian consumer demand system. Discussion Paper IPEA 793, Rio de Janeiro, 2001.

Ballard, C., Fullerton, D., Shoven, J. and J. Whalley. A general equilibrium model for tax policy evaluation. University of Chicago Press, Chicago, 1985.

Barros, R.P., Corseuil, C.H., Cury, S., and P. Leite. Openness and income distribution in Brazil. Proceedings of the Workshop on Trade Liberalization and the Labor Market in Brazil. Brasília: UnB/IPEA, 2001.

Barros, R. P., Henriques, R., and Mendonca, R. A estabilidade inaceitavel: desigualdade e pobreza no Brasil. Texto para Discussao 800, IPEA, Rio de Janeiro, 2001.

Bautista, R., Robinson, S., Tarp, F. and Wobst, P. Policy bias and agriculture: partial and general equilibrium measures. Review of Development Economics 5 (1), 2001, 89-104.

Bautista, R. and Thomas, M. Income effects of alternative trade policy adjustments on Philippine rural households: a general equilibrium analysis. Discussion Paper 22. Trade and Macroeconomics Division, IFPRI. 1997.

Beyer, H., Rojas, P., and R. Vergara. Trade liberalization and wage inequality. Journal of Development Economics 59, 1999, 103-123.

Campos-Filho, L. Unilateral liberalization and Mercosul: Implications for resource allocation. Unpublished Ph.D. Dissertation, University of London, 1998.

Carneiro, F. G. and J.S. Arbache. The impacts of trade on the Brazilian labor market: A CGE model approach. Export-Led Research Project of the UNDP Office in New York. Preliminary version, 2002.

247

Carvalho, A., and A. Parente. Impactos comerciais da ALCA. IPEA Discussion Papers 635, Brazil, 1999.

Cattaneo, A. The interaction between economic incentives, deforestation, and land degradation in Brazil. In: The Impact of Macroeconomic policy on Deforestation: A Comparative Study of Indonesia and Brazil. Progress Report, International Food Policy Research institute, Washington, DC., 1998.

Cattaneo, A. Deforestation in the Brazilian Amazon: Comparing the impacts of macroeconomic shocks, land tenure, and technological change. Paper presented at the workshop on “Technological Change in Agriculture and Deforestation”, Costa Rica. Resource and Environmental Policy Branch Economic Research Service, USDA, Washington, D.C., 1999

Cattaneo, A., Hinojosa-Ojeda, R., and Robinson, S. Costa Rica trade liberalization, fiscal imbalances, and macroeconomic policy: a computable general equilibrium analysis. The North American Journal of Economics and Finance 10, 1999, 39-67.

Chenery, H.B. The structuralist approach to development policy. American Economic Review 65, (2), 1975, 310-16.

Chou, J., Chu, Y. and Wang, S. Effects of trade liberalization on Taiwan: a computable general equilibrium analysis. Asian Economic Journal 11 (21), 1997, 169-85.

Dagum, C. Inequality Measures between income distributions with applications. Econometrica 48, 1980, 1791-1803.

Dagum, C. A new approach to the decomposition of the Gini income inequality ratio. Empirical Economics 22(4), 1997a, 515-531.

Dagum, C. Decomposition and interpretation of Gini and the generalized entropy inequality measures. Proceedings of the American Statistical Association, Business and Economic Statistics Section, 157th Meeting, 1997b, 200-205.

Davies, R., Rattso, S. and Torvik, R. Short run consequences of trade liberalization: A computable general equilibrium model of Zimbabwe. Journal of Policy Modeling 20 (3), 1998, 302-33.

Deardorff, A.V. and J.D. Haveman. The effects of U.S. trade laws on poverty in America. University of Michigan. Mimeo. 1991.

De Melo, J. Computable general equilibrium models for trade policy analysis in developing countries: a survey. The World Bank Working Paper 53, 1988.

248

De Melo, J. and Tarr, D. A general equilibrium analysis of US foreign trade policy. MIT Press. 1992.

Dervis, K., J. de Melo and S. Robinson. General equilibrium models for development policy. New York: Cambridge University Press. 1982.

Estevadeordal, A., Goto, J. and R. Saez. The new regionalism in the Americas: The case of Mercosur. Working Paper 5, INTAL-ITD, Inter-American Development Bank, 2000.

Flores, R. G. The gains from Mercosul: A general equilibrium, imperfect competition evaluation. Journal of Policy Modeling, 19 (1), 1997, 1–18.

Fougere, M. and M. Merette. Population aging, intergenerational equity, and growth: analysis with an endogenous growth overlapping-generations model. In Harrison, G., Jensen, S., Pedersen, L. and T. Rutherford. (eds.) Using dynamic general equilibrium models for policy analysis. North-Holland, Amsterdam, 2000, 325-61.

Gelan, A. Trade liberalisation and urban-rural linkages: a CGE analysis for Ethiopia. Journal of Policy Modeling 24, 2002, 707-38.

Golan, A., G. Judge, and Miller, D. Maximum entropy econometrics: robust estimation with limited data. Chichester: John Wiley & Sons. 1996.

Golan, A., and S. J. Vogel. Estimation of stationary and non-stationary accounting matrix coefficients with structural and supply-side information. ERS/USDA. Unpublished. 1997

Gopinath, M. and Roe, T. Modeling inter-sectoral growth linkages: an application to U.S. agriculture. Agricultural Economics 21, 1999, 131-44.

Greenaway, D., Reed, G. and N. Winchester. Trade and rising wage inequality in the U.K.: results from a CGE model. Research Paper 29, University of Nottingham, UK, 2002.

Haddad, E. A. Regional inequality and structural changes: Lessons from the Brazilian economy. Aldershot: Ashgate. 1999.

Haddad, E. A., and C.R. Azzoni. Trade and location: geographical shifts in the Brazilian economic structure. In J. J. M. Guilhoto and G. J. D. Hewings (Eds.), Structure and structural change in the Brazilian economy. Aldershot: Ashgate. 2001.

Haddad, E. A., Domingues, E. and F. Perobelli. Regional effects of economic integration: the case of Brazil. Journal of Policy Modeling 24, 2002, 453-82.

Harberger, A. The incidence of the corporate income tax. Journal of Political Economy 70, 1962, 215-40.

249

Harrison, G. and T. Rutherford. Burden sharing, joint implementation, and carbon coalitions. In Carraro, C. (ed.) International environmental agreements on climate change. Kluwer Academic Publishing, Amsterdam, 1999, 77-108.

Harrison, G., Rutherford, T. and D. Tarr. Trade liberalization, poverty and efficient equity. Journal of Development Economics 71, 2003, 97-128.

Hertel, T., Horridge, J. and K. Pearson. Mending the family tree: a reconciliation of the linearization and levels school of AGE modelling. Economic Modelling 9, 1992, 385- 407.

Higgs, P., Parmenter, B. and R. Rimmer. A hybrid top-down, bottom-up regional computable general equilibrium model. International Regional Science Review 11 (3), 1988, 317-28.

IBGE. Matriz de insumo-produto Brasil 1995. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 1997a.

IBGE. Sistema de contas nacionais Brasil 1990-1995/96. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 1997b.

IBGE. Anuario estatistico do Brasil 1997. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 1997c.

IBGE. Censo agropecuário 1995/1996. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 1998.

IBGE. Censo demografico 2000. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 2000a.

IBGE. Contas regionais 1985-2000. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 2000b.

IBGE. Cadastro central de empresas 2000. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 2000c.

IBGE. Pesquisa nacional por amostra de domicilios 2001. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 2001.

IBGE. Indicadores de desenvolvimento sustentavel 2002. Instituto Brasileiro de Geografia e Estatistica: Rio de Janeiro, 2002.

Imbens, G. One-step estimators for over-identified generalized method of moment models. Review of Economic Studies 64, 1997, 359-383.

250

Jaynes, E.T. Information theory and statistical mechanics II. Physics Review 108, 1957a, 171-190.

Jaynes, E.T. Information theory and statistical mechanics I. Physics Review 106, 1957b, 620-630.

Johansen, L. A multi-sectoral study of economic growth. North-Holland, Amsterdam, 1960.

Kapur, J.N and H.K. Kesavan. Entropy optimization principles with applications. Academic Press. 1992.

Karunaratne, N. Macroeconomics insights on the liberalised trading regime of Thailand. International Journal of Social Economics 25 (6,7,8), 1998, 1142-159.

Keuschnigg, C. and W. Kohler. Eastern enlargement of the EU: a dynamic general equilibrium perspective. In Harrison, G., Jensen, S., Pedersen, L. and T. Rutherford. (eds.) Using dynamic general equilibrium models for policy analysis. North-Holland, Amsterdam, 2000, 119-170.

Khan, F. Household disaggregation. In Francois, J. F. and Reinert, K. (eds.) Applied Methods for Trade Policy Analysis: A Handbook. New York: Cambridge University Press. 1997, 300-27.

Konan, D. and K. Maskus. Joint trade liberalization and tax reform in a small open economy: the case of Egypt. Journal of Development Economics 61, 2000, 365-92.

Kraybill, D. A computable general equilibrium analysis of regional impacts of macro- shocks in the 1980s. Ph.D. Dissertation, Virginia Tech., 1989.

Leipziger, D. M., Frischtak, C., Kharas, H. and Normand, J. F. Mercosur: integration and industrial policy. The World Economy 20 (5), 1997, 585-603.

Leontief, W. Input-output economics. Oxford University Press, New York. 1966.

Liew, L.H. Tops-down versus bottoms-up approaches to regional modeling. Journal of Policy Modeling 6 (3), 1984, 351-67.

Lofgren, H. Trade reform and the poor in Morocco: a rural-urban general equilibrium analysis of reduced protection. TMD Disscusion Paper 38, IFPRI, 1999.

Lofgren, H., Harris, R. and Robinson, S. A standard computable general equilibrium (CGE) model in GAMS. TMD Disscusion Paper 75, IFPRI, 2001.

251

Lofgren, H., Robinson, S. and Thurlow, J. Macro and micro effects of recent and potential shocks to copper mining in Zambia. TMD Disscusion Paper 99, IFPRI, 2002.

Lopez, R. and A. Panagaryia. On the theory of piecemeal tariff reform: the case of pure imported intermediate inputs. American Economic Review 82, (3), 1992, 615-25.

Lysy, L. and L. Taylor. A computable general equilibrium model for the functional distribution of income: experiments for Brazil, 1959-71. In Taylor, L., Bacha, E., Cardoso, E. and F. Lysy (eds) Models of growth and distribution for Brazil. Oxford University Press. 1980.

Maia, K. The impact of trade, technology and final demand on the structure of employment in Brazil, 1985-1995. Proceedings of the Workshop on Trade Liberalization and the Labor Market in Brazil. Brasília: UnB/IPEA, 2001.

Mas-Colell, A., Whinston, M. and J. Green. Microeconomic theory. New York: Oxford University Press, 1995.

Mehlum, H. Zimbabwe: Investments, credibility, and the dynamics following trade liberalization. Economic Modelling 19, 2002, 565-84.

Menezes-Filho, N. and M. Rodrigues Jr. Openness, technology and skills: evidence from the Brazilian manufacturing. Proceedings of the Workshop on Trade Liberalization and the Labor Market in Brazil. Brasília: UnB/IPEA, 2001.

Monteagudo, J. and M. Watanuki. Regional trade agreements for Mercosur: the FTAA and the FTA with the European union. Paper prepared for presentation at the Conference Impacts of Trade Liberalization Agreements on Latin America and the Caribbean, organized by the Inter-American Development Bank and the Centre d’Etudes Prospectives et d’Information Internationales, Washington, DC., 2001.

Monteagudo, J. and Watanuki, M. Evaluating agricultural reform under the FTAA and Mercosur-EU FTA for Latin America A quantitative CGE assessment. Paper prepared for presentation at the Agricultural Liberalization and Integration: What to expect from the FTAA and the WTO? Hosted by the Special Initiative on Integration and Trade, Integration and Regional Programs Department, Inter-American Development Bank, Washington D.C., 2002.

Mussard, S., Seyte, F., and M. Terraza. Decomposition of Gini and the generalized entropy inequality measures. Economics Bulletin 4(7), 2003, 1-6.

Piggott, J. and J. Whalley. UK Tax policy and applied general equilibrium analysis.Cambridge University Press, New York, 1985.

252

Powell, A. and T. Lawson. A decade of applied general equilibrium modelling. In Bergman, D. and E. Zalai (eds) General equilibrium modeling and economic policy analysis. Basil Blackwell, Oxford, 1990, 241-90.

Qiang, Y. CGE modelling and Australian economy. Ph.D. Thesis, The University of Western Australia., 1999.

Rawls, J. A Theory of Justice. Cambridge: Harvard University Press, 1971.

Reinert, K., and Roland-Holst, D. Social accounting matrices. In Francois, J. F. and Reinert, K. (eds.) Applied Methods for Trade Policy Analysis: A Handbook. New York: Cambridge University Press. 1997, 94-121

Robbins, D.J. Worsening relative wage dispersion in Chile during trade liberalization, and its causes: is supply at fault? Development Discussion Papers 484, Harvard University: Harvard Institute for International Development, 1994.

Robbins, D.J. Wage dispersion and trade in Colombia: an analysis of greater Bogota: 1876-1989. Harvard University: Harvard Institute for International Development, 1995, mimeo.

Robbins, D. J., and T. H. Gindling. Trade liberalization and the relative wages for more- skilled workers in Costa Rica. Review of Development Economics 3, 1999, 140-154.

Robilliard, A., and S. Robinson. Reconciling household surveys and national accounts data using a cross entropy estimation method. Discussion Paper 50. Trade and Macroeconomics Division, IFPRI. 2001.

Robinson, S. Multisector models. Handbook of Development Economics. North-Holland, Amsterdam, 1989, 885-947.

Robinson, S., Cattaneo, A., and M. El-Said. Updating and estimating a social accounting matrix using cross entropy methods. Discussion Paper 58. Trade and Macroeconomics Division, IFPRI. 2000.

Robinson, S., Cattaneo, A., and M. El-Said. Estimating a social accounting matrix using cross entropy methods. Discussion Paper 33. Trade and Macroeconomics Division, IFPRI. 1998.

Rodrik, D. Policy uncertainty and private investment in developing countries. Journal of Development Economics 36, 1991, 229-42.

Rodrik, D. The limits of trade policy reform in developing countries. Journal of Economic Perspectives 6 (1), 1992, 87-105.

253

Rutherford, T. Carbon abatement, technical change and intergenerational burden sharing. In Harrison, G., Jensen, S., Pedersen, L. and T. Rutherford. (eds.) Using dynamic general equilibrium models for policy analysis. North-Holland, Amsterdam, 2000, 79-117.

Scarf, H. The approximation of fixed points of a continuous mapping. SIAM Journal of Applied Mathematics 15 (5), 1967, 1328-343.

Scarf, H. and T. Hansen. The computation of economic equilibria. Yale University Press, New Haven. 1973.

Schumpeter, J. A. A history of economic analysis. Oxford University Press, New York. 1954.

Sen, A. Neo-classical and neo-Keynesian theories of distribution. Economic Record 39, 1963, 45-64.

Shannon, C. E. A mathematical theory of communication. Bell System Technical Journal 27, 1948, 379-423.

Shoven, J. B. and J. Whalley. Applied general equilibrium models of taxation and international trade: an introduction and survey. Journal of Economic Literature XXII, 1984, 1007-1051.

Silber J. Factor components, population subgroups and the computation of the Gini index of inequality. Review of Economics and Statistics 71, 1989, 107-115.

Sorsa, P. Algeria: the real exchange rate, export diversification, and trade protection. IMF Working Paper 99-49. 1999.

Srinivasan, T. and J. Whalley. General equilibrium trade policy modelling. MIT Press, Cambridge, 1986.

Stiglitz, J. Two principles for the next round or how to bring developing countries in from the cold. WTO, Geneva, September 1999.

Taylor, L., Bacha, E., Cardoso, E. and F. Lysy. Models of growth and distribution for Brazil. Oxford University Press. 1980.

Theil, H. Economics and information theory. Rand McNally. 1967

Thissen, M. Two decades of CGE modelling: lessons from models for Egypt. SOM Research Report 99C02, University of Groningen, The Netherlands. 1998.

Thurlow, J. and van Seventer, D. E. A standard computable general equilibrium model for South Africa. TMD Disscusion Paper 100, IFPRI, 2002.

254

Tourinho, O., Kume, H. and A. C. Pedroso. Elasticidades de Armington para o Brasil:1986-2001. Discussion Paper IPEA 901, Rio de Janeiro, 2002.

Vaillant, M and Ons, A. Preferential trading arrangements between the European Union and South America: the political economy of free trade zones in practice. The World Economy 25 (10), 2002, 1433-68.

Vincent, D. P. Applied general equilibrium modelling in the Australian industries assistance commission: perspectives of a policy analyst. In Bergman, D. and E. Zalai (eds) General equilibrium modeling and economic policy analysis. Basil Blackwell, Oxford, 1990, 291-347.

Walras, L. Elements d’economie politique pure. 4me Edition Definitive et Augmentee par l’Auteur (1926), R. Pichon et R. Durand-Auzias Editeurs, Paris. 1874.

Whalley, J. Trade liberalization among major trading areas. MIT Press, Cambridge, 1985.

Willenbockel, D. Applied general equilibrium modeling, imperfect competition and European integration. Wiley, Chichester, U.K. 1994.

Williams III, R. Revisiting the cost of protectionism: the role of tax distortions in the labor market. Journal of International Economics 47, 1999, 429-47.

Winters, L. A. Trade liberalisation and poverty: what are the links?. The World Economy 25 (9), 2002, 1339-67.

Wobst, P. The impact of domestic and global trade liberalization on five Southern African countries. TMD Disscusion Paper 92, IFPRI, 2002.

Chapter 3:

Anderson, J.E. A theoretical foundation for the gravity equation. The American Economic Review 69 (1),1979, 106-116.

Anderson, J.E. and E. van Wincoop. Gravity with gravitas: A solution to the border puzzle. The American Economic Review 93 (1),2003, 170-92.

Averbug, A. Mercosul: conjuntura e perspectives. Revista do BNDES, Dezembro, 1998.

255

Azevedo, A.F. What has been the real effect of Mercosur on trade? A gravity model approach. Discussion Paper 1 PPGE/UFRS, Porto Alegre, 2002.

Baer, W., Silva, P. and T. Cavalcanti. Economic Integration without Policy Coordination: the case of Mercosur. Manuscript. University of Illinois, Champaign-Urbana. 2001.

Baldwin, R.E. Hysteresis in import prices: the beachhead effect. The American Economic Review 78, 1988, 773-785.

Baldwin, R.E. Toward an integrated Europe. Centre for Economic Policy Research, London, 1994.

Baldwin, R.E. and P. Krugman. Persistent trade effects of large exchange rate shocks. Quarterly Journal of Economics 104, 1989, 635-655.

Barcellos Neto, P.C., Azevedo, A.F. and M. Portugal. Impactos comerciais da area de livre comercio das Americas: uma aplicacao do modelo gravitacional. Discussion Paper 9 PPGE/UFRS, Porto Alegre, 2002.

Batten, D.S. and M.T. Belongia. Monetary policy, real exchange rates and U.S. agricultural exports. American Journal of Agricultural Economics 68,1986, 422-27.

Bergstrand, J.H. The gravity equation in international trade: some microeconomic foundations and empirical evidence. Review of Economics and Statistics 67(3), 1985, 474-481.

Bergstrand, J.H. The generalized gravity equation, monopolistic competition, and the factor-proportions theory in international trade. Review of Economics and Statistics 71(1),1989, 143-53.

Bergstrand, J. H. The Heckscher-Ohlin-Samuelson model, the Linder hypothesis, and the determinants of bilateral intra-industry trade. Economic Journal 100(4), 1990, 1216-229.

Bessler, D.A. and R.A. Babula. Forecasting wheat exports: Do exchange rates matter?. Journal of Business, Economics and Statistics 5,1987, 397-406.

Blejer, M. Income per capita and the structure of industrial exports: an empirical study. Review of Economics and Statistics 69, 1987, 555-61.

Bouzas, R. Mercosur’s economic agenda: short- and medium-term policy challenges. Integration and Trade 1, 1997, 57-79.

Bouzas, R. El processo del ALCA: incertidumbres y desafios. Indcadores Economicos FEE 29, 2001, 9-24.

256

Brodsky, D. Fixed versus flexible exchange rates, and the measurement of exchange rate instability. Journal of International Economics 16, 1984, 295–306.

Brown, W. and Anderson, W. Spatial markets and the potential for economic integration between Canadian and U.S. regions. Papers in Regional Science 81,2002, 99-120.

Canuto, O., Lima, G.T. and M. Alexandre. Investimentos externos em servicos e efeitos potenciais da negociacao da ALCA. Discussion Paper IPEA 942, Rio de Janeiro, 2003.

Carvalho, A. and M.A. Parente. Impactos comerciais da área de livre comércio das Américas. Discussion Paper IPEA 635, Rio de Janeiro, 1999.

Carvalho, A., Parente, M. A., Lerda, S. and S. Miyata. Impactos da integração comercial Brasil - Estados Unidos. Discussion Paper IPEA 646, Rio de Janeiro, 1999.

Castilho, M. O acesso das exportações do Mercosul ao mercado europeu. Discussion Paper IPEA 851, Rio de Janeiro, 2001.

Castilho, M. Impactos de acordos comerciais sobre a economia Brasileira: resenha dos trabalhos recentes. Discussion Paper IPEA 936, Rio de Janeiro, 2002.

Cho, G. D. Real exchange rate movements and agricultural trade. Ph.D. Dissertation, The Ohio State University, 2001.

Cho, G., Sheldon, I. M. and McCorriston, S. Exchange rate uncertainty and agricultural trade. American Journal of Agricultural Economics 84 (4),2002, 932-942.

Cushman, D.O. US bilateral trade flows and exchange risk during the floating period. Journal of International Economy 24, 1988, 317-30.

Deardorff, A. Testing trade theories and predicting trade flows. In Jones, R. and P. Kenen (eds.), Handbook of International Economics vol. 1, North-Holland, Amsterdam, 1984, 467-517.

Deardorff, A. Determinants of bilateral trade: does gravity work in a Neoclassical world?. In J. Frankel (ed.), In the Regionalization of the World Economy. Chicago: University of Chicago Press, 1998.

De Grauwe, P. Exchange rate variability and the slowdown in growth of international trade. IMF Staff Papers 24, 1988, 317-30.

De Grauwe, P. and B. De Bellefroid. Long-run exchange rate variability and international trade. In S. Arndt and J. D. Richardson (Editors), Real-Financial Linkages Among Open Economies. Cambridge: MIT Press. 1986.

257

Dell’Ariccia, G. Exchange rate fluctuations and trade flows. IMF Staff Papers 46, 2000, 293-314.

Devlin, R. The free trade area of the Americas and Mercosur-European Union free trade processes: Can they learn something from each other? Intal ITD Occasional Paper 6, Inter-American Development Bank, Buenos Aires, 2000.

Diao, X., Bonilla, E. and S. Robinson. Scenarios for trade integration in the Americas. IFPRI/TMD Discussion Paper 90, Washington, D.C., 2002.

Dixit, A. and J. Stiglitz. Monopolistic competition and optimum product diversity. The American Economic Review 67, 1977, 297-308.

Dornbusch, R. Exchange rates and prices. The American Economic Review 77, 1987, 93- 106.

Egger, P. An econometric view on the estimation of gravity models and the calculation of trade potentials. The World Economy 25(2), 2002, 297-312.

Eichengreen, B. Does Mercosur need a single currency?. NBER Working Paper 6821. 1998.

Eichengreen, B. and D.A. Irwin. Trade blocs and the reorientation of trade in the 1930s. Journal of International Economy 38, 1995, 1-24.

Ethier, W. National and international returns of scale in the modern theory of international trade. The American Economic Review 72, 1982, 389-405.

Evenett, S.J. and W. Keller. On theories explaining the success of the gravity equation. Journal of Political Economy 110(2), 2002, 281-316.

Feenstra, R. C. Border effects and the gravity equation: consistent methods for estimation. Journal of Political Economy 49 (5), 2002, 491-506.

Feenstra, R. C., Markusen, J. and Rose, A. Using the gravity equation to differentiate among alternative theories of trade. Canadian Journal of Economics 34 (2), 2001, 430- 447.

Flood, R.P. and A. Rose. Understanding exchange rate volatility without the contrivance of macroeconomics. Economic Journal 109, 1999, 660-672.

Frankel, J. Regional trading blocs in the world economic system. Washington DC: Institute for International Economics. 1997.

258

Frankel, J. and A. Rose. Empirical research on nominal exchange rates. In G. Grossman and K. Rogoff (editors), Handbook of International Economics, Vol. III., 1995, 1689- 1729.

Frankel, J. and S. Wei. Trade blocs and currency blocs. NBER Working Paper 4335, Cambridge, MA. NBER, 1993.

Frankel, J., Stein, E. and S. Wei. Continental trading blocs: are they natural or supernatural? In J. Frankel (ed.), In the Regionalization of the World Economy. Chicago: University of Chicago Press, 1998.

Froot, K. A. and P. D. Kemplerer. Exchange rate pass-through when market share matters. The American Economic Review 79, 1989, 637-654.

Glick, R. and A. Rose. Does a currency union affect trade? The time series evidence. NBER Working Paper 8396, Cambridge, MA. NBER, 2001.

Greene, W. Econometric analysis. Prentice-Hall, New Jersey. 3rd edition. 1997.

Grubel, H.G. and P.J. Lloyd. Intra-industry trade: the theory and measurement of international trade in differentiated products. London: Macmillan, 1975.

Haddad, E., Domingues, E. and F. Perobelli. Regional effects of economic integration: the case of Brazil. Journal of Policy Modeling 24, 2002, 453-82.

Haley, S.L. and B. Krissoff. U.S. grain exports and the value of the U.S. dollar. Journal of Agricultural Economics Research 39,1987, 12-21.

Hallwood, C. P. and R. MacDonald. International money and finance. Blackwell Publishers, third edition. 2000.

Harrison, G., Rutherford, T., Tarr, D., and A. Gurgel. Regional, multilateral and unilateral trade policies of Mercosur for growth and poverty reduction in Brazil. Paper presented at Conference of DIMAC, Rio de Janeiro: IPEA, 2002.

Helpman, E. International trade in the presence of product differentiation, economies of scale, and monopolistic competition. Journal of International Economics 11, 1981, 305- 40.

Helpman, E. Imperfect competiton and international trade: evidence from fourteen industrial countries. Journal of the Japanese and International Economies 1, 1987, 62- 81.

Helpman, E. & Krugman, P. Market structure and foreign trade. MIT Press, Cambridge, MA. 1985.

259

Hooper, P. and S.W. Kohlhagen. The effect of exchange rate uncertainty on the prices and volume of international trade. Journal of International Economy 8, 1978, 483-511.

Kenen, P. and D. Rodrik. Measuring and analysing the effects of short-term volatility in real exchange rates. Review of Economics and Statistics 68, 1986.

Klein, M.W. Sectoral effects of exchange rate volatility on United States exports. Journal of International Money and Finance 9,1990, 299-308.

Krugman, P. Increasing returns, monopolistic competition, and international trade. Journal of International Economics 9,1979, 469-79.

Kume, H. and G. Piani. Fluxos bilaterais de comércio e blocos regionais: Uma aplicação do modelo gravitacional. Discussion Paper IPEA 749, Rio de Janeiro, 2001.

Lanyi, A. and E. Suss. Exchange rate variability: alternative measures and interpretation. IMF Staff Papers 29, 1982, 527–60.

Leamer, E. E. The commodity composition of international trade in manufactures: an empirical analysis. Oxford Economic Papers 26, 1974, 350-74.

Leamer, E.E. and J. Levinsohn. International trade theory: the evidence. In Grossman, G. and K. Rogoff (eds.), Handbook of International Economics vol. 3, North-Holland, Amsterdam, 1995.

Leamer, E.E. and R.M. Stern. Quantitative international economics. Boston: Allyn and Bacon, 1970.

Linnemann, H. An econometric study of international trade flows. Amsterdam: North- Holland, 1966.

MacDonald, R. and M. P. Taylor. Exchange rate economics: a survey. IMF Staff Paper 39, 1992, 1-57.

Mark, N. Exchange rates and fundamentals: evidence on long-horizon predictability. The American Economic Review 85, 1995, 201-18.

Mark, N. International macroeconomics and finance. Blackwell Publishers. 2001.

Markusen, J. Explaining the volume of trade: an eclectic approach. The American Economic Review 76(5), 1986, 1002-011.

260

Mattson, J.W. and W. Koo. U.S. agricultural trade with Latin American countries and effects of the Free Trade Area of the Americas on specific commodities. Agribusiness and Applied Economics Report 510, Fargo, ND. North Dakota State University, 2003.

Matyas, L. Proper econometric specification of the gravity model. The World Economy 20(3), 1997, 363-68.

Maskus, K. Exchange rate risk and U.S. trade: a sectoral analysis. Federal Reserve Bank of Kansas Economic Review 3, 2002, 16-28.

Mayer, W. Endogenous tariff formation. The American Economic Review 74, 1984, 970- 985.

McKinnon, R. I. Monetary and exchange rate policies for international financial stability: a proposal. Journal of Economic Perspectives 2, 1988, 83-103.

Nonnenberg, M. and M. Mendonça, M. Criação e desvio de comércio no Mercosul: o caso dos produtos agrícolas. Discussion Paper IPEA 631, Rio de Janeiro, 1999.

Peree, E. and A. Steinherr. Exchange rate uncertainty and foreign trade. European Economic Review 33,1989, 1241-64.

Pick, D.H. Exchange rate risk and U.S. agricultural trade flows. American Journal of Agricultural Economics 72,1990, 694-700.

Pick, D.H. and T.L. Vollrath. Real exchange rate misalignment and agricultural export performance in developing countries. Economic Development and Cultural Change 42,1994, 555-71.

Porto, P.S. and O. Canuto. Mercosur: gains from regional integration and exchange rate regimes. Revista Economia Aplicada 6(4),2002.

Poyhonen, P. A tentative model for the volume of trade between countries. Weltwirtschaftliches Archiv 90, 1963, 93-100.

Rose, A. One money, one market: the effect of common currencies on trade. Economic Policy 30, 2000, 7-46.

Rose, A. and E. Van Wincoop. National money as a barrier to international trade: the real case for currency union. The American Economic Review, Papers and Proceedings 91(2), 2001, 386-90.

Samuelson, P. The transfer problem and transport costs: the terms of trade when impediments are absent. Economic Journal 62, 1952, 278-304.

261

Sauer, C. and A. K. Bohara. Exchange rate volatility and exports: regional differences between developing and industrialized countries. Review of International Economics 9 (1), 2001, 133-152.

Schuh, G. E. The exchange rate and U.S. agriculture. American Journal of Agricultural Economics 56,1974, 1-13.

Soloaga, I. and A. Winters. Regionalism in the nineties: What effect on trade?. The North American Journal of Economics and Finance 12, 2001, 1-29.

Thursby, J.G. and M.C. Thursby. Bilateral trade flows, the Linder hypothesis and exchange rate risk. Review of Economics and Statistics 69, 1987, 488-95.

Tinbergen, J. Shaping the world economy. Twenty-Century Fund, New York, 1962.

Tourinho, O. and H. Kume. Os impactos setoriais de acordos de comércio: análise com um modelo CGE da economia brasileira. Rio de Janeiro: IPEA, 2002 (mimeo).

Trefler, D. Trade liberalization and the theory of endogenous protection: an econometric study of U.S. import policy. Journal of Political Economy 101 (1), 1993, 138-160.

Tweeten, L. Farm policy analysis. Boulder, CO: Westview Press, 1989.

Valls Pereira, L. Impactos do Mercosul, da Alca e do acordo MS-UE na economia brasileira e argentina: um mapeamento preliminar. IBRE, 2001 (mimeo).

Watanuki, M. and J. Monteagudo. Regional trade agreements for Mercosur: the FTAA and the FTA with the European Union. Paper presented in the Impacts of Trade Liberalization Agreements on Latin America and the Caribbean Meeting. Washington: CEPII e BID (orgs.), 2001.

Wei, S.J. Intra-national versus international trade: How stubborn are nations in global integration?. NBER Working Paper 5531, Cambridge, MA. NBER, 1996.

Winters, L.A. Regionalism and the rest of the world: The irrelevance of the Kemp-Wan theorem. Oxford Economic Papers 49, 1997, 228-34.

262